Rehabilitation for ankle fractures in adults

Abstract

Background

Ankle fracture is one of the most common lower limb fractures. Whilst immobilisation of the ankle can support and protect the fracture site during early healing, this also increases the risk of ankle weakness, stiffness, and residual pain. Rehabilitation aims to address the after‐effects of this injury, to improve ankle function and quality of life. Approaches are wide‐ranging and include strategies to improve ankle joint movement, muscle strength, or both. This is an update of a Cochrane review last published in 2012.

Objectives

To assess the effects of rehabilitation interventions following surgical or non‐surgical management of ankle fractures in adults.

Search methods

We searched CENTRAL, MEDLINE, Embase, three other databases, and two clinical trials registers in May 2022, and conducted additional searches of CENTRAL, MEDLINE, and Embase in March 2023. We also searched reference lists of included studies and relevant systematic reviews.

Selection criteria

We included randomised controlled trials (RCTs) and quasi‐RCTs comparing any rehabilitation intervention delivered to adults with ankle fracture. Interventions could have been given during or after the initial fracture management period (typically the first six weeks after injury), which may or may not have included surgical fixation. We excluded participants with multi‐trauma, pathological fracture, or with established complications secondary to ankle fracture.

Data collection and analysis

We used standard methodological procedures expected by Cochrane. We collected data for five outcomes: activity limitation (ankle function), health‐related quality of life (HRQoL), participant satisfaction with treatment, pain, and adverse events (we focused on re‐operation, defined as unplanned return to theatre). We report the findings up to six months after injury.

Main results

We included 53 studies (45 RCTs, 8 quasi‐RCTs) with 4489 adults with ankle fracture. In most studies, orthopaedic management included surgical fixation but was non‐surgical in five studies, and either surgical or non‐surgical in six studies. Here, we summarise the findings for three common rehabilitation comparisons; these included the most data and were the most clinically relevant. Because of different intervention approaches, we sometimes included a study in more than one comparison. Data for other less common comparisons were also available but often included few participants and were imprecise.

All studies were unavoidably at high risk of performance and detection bias. We downgraded the certainty of all evidence for this reason. We also downgraded for imprecision and when we noted inconsistencies between studies that precluded meta‐analysis of data.

Early (within 3 weeks of surgery) versus delayed weight‐bearing (12 studies, 1403 participants)

Early weight‐bearing probably leads to better ankle function (mean difference (MD) 3.56, 95% confidence interval (CI) 1.35 to 5.78; 5 studies, 890 participants; moderate‐certainty evidence); however, this does not include a clinically meaningful difference. Early weight‐bearing may offer little or no difference to HRQoL compared to delayed weight‐bearing (standardised mean difference (SMD) 0.15, 95% CI ‐0.01 to 0.30; 5 studies, 739 participants; low‐certainty evidence); when translated to the EQ‐5D scale (a commonly‐used HRQoL questionnaire), any small difference was not clinically important. We were unsure whether there were any differences in participant satisfaction or pain because these outcomes had very low‐certainty evidence. For adverse events, there may be little or no difference in re‐operation (risk ratio (RR) 0.50, 95% CI 0.09 to 2.68; 7 studies, 1007 participants; low‐certainty evidence).

Removable versus non‐removable ankle support (25 studies, 2206 participants)

Following surgery, using a removable ankle support may lead to better ankle function (MD 6.39, 95% CI 1.69 to 11.09; 6 studies, 677 participants; low‐certainty evidence). This effect included both a clinically important and unimportant difference. There is probably an improvement in HRQoL with a removable ankle support, although this difference included both a clinically important and unimportant difference when translated to the EQ‐5D scale (SMD 0.30, 95% CI 0.11 to 0.50; 3 studies, 477 participants; moderate‐certainty evidence). No studies reported participant satisfaction. We were unsure of the effects on pain because of very low‐certainty evidence (1 study, 29 participants). There may be little or no difference in re‐operations (RR 1.20, 95% CI 0.39 to 3.71; 6 studies, 624 participants; low‐certainty evidence).

Following non‐surgical management, there may be little or no difference between removable and non‐removable ankle supports in ankle function (MD 1.08, 95% CI ‐3.18 to 5.34; 3 studies, 399 participants), and HRQoL (SMD ‐0.04, 95% CI ‐0.24 to 0.15; 3 studies, 397 participants); low‐certainty evidence. No studies reported participant satisfaction. We were unsure of the effects on pain (2 studies, 167 participants), or re‐operation because of very low‐certainty evidence (1 study, 305 participants).

Physical therapy interventions versus usual care or other physical therapy interventions (9 studies, 857 participants)

Types of interventions included the use of active controlled motion, a spring‐loaded ankle trainer, an antigravity treadmill, and variations of enhanced physiotherapy (e.g. additional stretching, joint mobilisation, neuromuscular exercises), delivered during or after the initial fracture management period. We were unable to pool data because of the differences in the design of interventions and their usual care comparators. Studies often included very few participants. The certainty of the evidence for all outcomes in this comparison was very low, and therefore we were unsure of the effectiveness of these therapies. No studies in this comparison reported re‐operation.

Authors' conclusions

Early weight‐bearing may improve outcomes in the first six months after surgery for ankle fracture, but the difference is likely to be small and may not always be clinically important. A removable ankle support may also provide a better outcome, but again, the difference may not always be clinically important. It is likely that neither approach increases the re‐operation risk. We assume that the findings for these comparisons are applicable to people with closed ankle fractures, and that satisfactory fracture stabilisation had been achieved with surgery. For people who have non‐surgical treatment, there is no evidence that either a removable or non‐removable ankle support may be superior. We were uncertain whether any physical therapy interventions were more effective than usual care or other physical therapy interventions. We encourage investigators of future studies on rehabilitation interventions for ankle fracture to use a core outcome set.

Plain language summary

Which approaches help with recovery after a broken ankle in adults?

Key messages

‐ Putting weight on a broken ankle within three weeks of surgery could improve recovery. But it might not improve recovery enough to matter.

‐ Using a removable ankle support (which allows someone to stretch and exercise their injured ankle) during the first six weeks after surgery may improve recovery. But, again, the difference might not be important enough.

‐ There isn't enough evidence to assess the value of physical therapies in recovering after a broken ankle.

Broken ankles

A broken ankle is one of the most common types of broken bone. Sometimes surgery is needed to help fix the broken bone, and sometimes the ankle just needs support for a few weeks while it heals. It usually takes about six weeks for the bone to heal, but can take much longer for someone to get back to their normal activities. Different approaches can be used during this recovery period to increase ankle movement or muscle strength, or both.

What approaches are used to improve recovery?

‐ Early weight‐bearing or delayed weight‐bearing. People may be encouraged to start putting weight on their ankle within three weeks of surgery. Alternatively, they may be asked to completely avoid putting weight on their injured ankle for the first six weeks or so.

‐ Removable ankle support or non‐removable ankle support. A brace, splint, or other removable ankle support may be placed around the ankle – this allows people to regularly remove the support to stretch and move their ankle. Alternatively, a cast could be fitted, so that there is no ankle movement.

‐ Physical therapies. These include different kinds of ankle exercises, perhaps guided by a physiotherapist (or other physical therapist), and people may need to use specialist equipment.

What did we want to find out?

We wanted to find out if any of these approaches were better than the other to improve:

‐ ankle function (how well a person can do activities involving the ankle, and issues like ankle pain or swelling);

‐ quality of life;

‐ treatment satisfaction (how satisfied someone is with their treatment); and

‐ pain.

We also wanted to find out if any of these approaches were associated with an increased risk of having extra surgery (at a later date) on the ankle.

What did we do?

We searched for studies that looked at different approaches for improving recovery after an ankle fracture. We compared and summarised the results of studies and rated our confidence in the evidence, based on factors such as study methods and sizes.

What did we find?

We found 53 studies involving 4489 adults with a broken ankle. Most people in the studies had surgery to help fix the bone, and others used only ankle supports (with no surgery) in the first few weeks after injury.

Main results

Early versus delayed weight‐bearing. Early weight‐bearing after surgery for a broken ankle probably leads to very slightly better ankle function (though the difference might not be big enough to matter). Although it may also lead to a better quality of life, the difference might not be big enough to matter. We could not tell if there were differences between the two approaches for people's treatment satisfaction or pain. There may be little or no difference between approaches in how many people need extra ankle surgery.

Removable ankle support or non‐removable ankle support (such as a cast).

‐ For people who have surgery after their injury, a removable ankle support may lead to better ankle function (though for some people it may not be a big enough difference to be important). Similarly, people are probably more likely to have a better quality of life (though the difference might not always be important). We could not tell if there were differences between types of ankle support in terms of pain. There may be little or no difference between ankle supports in the need for extra ankle surgery.

‐ For people who didn't need surgery, there may be little or no difference between types of ankle support in ankle function or quality of life. We could not tell if there were differences between ankle supports in terms of pain or the need for ankle surgery at a later date.

No studies in this comparison reported treatment satisfaction.

Physical therapies. Studies examined types of therapy that were all very different from one another (including different exercise programmes or different equipment to exercise the ankle). Studies were often small, and we could not tell if any of the approaches were better than another. These studies did not report how many people needed later surgery.

What are the limitations of the evidence?

We could not always be confident in the evidence because people in the studies always knew which approach was being used for their ankle. Some findings showed that there may have been little or no difference. Some studies were very small, and sometimes the approaches were very different from one another.

How current is this evidence?

The evidence is current to March 2023.

Summary of findings

Summary of findings 1. Early weight‐bearing compared with delayed weight‐bearing for ankle fracture in adults.

Early weight‐bearing compared with delayed weight‐bearing for ankle fracture in adults
Population: adults treated surgically for ankle fractures Setting: hospitals, orthopaedic clinics Intervention: early weight‐bearing (i.e. weight‐bearing commenced typically within three weeks of injury) Comparison: delayed weight‐bearing (i.e. no weight‐bearing during initial fracture management period, which is typically the first six weeks after injury)
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with delayed weight‐bearinga	Risk with early weight‐bearing
Activity limitation Measured using OMAS (100‐point scale, higher scores indicate a better outcome) Follow‐up: short‐term ≤ 6 months. In the included studies, follow‐up was at 12 weeks, 4 months, or 6 months.	In the delayed weight‐bearing group, mean OMAS scores ranged from 61.2 to 80.9	MD 3.56 higher (1.35 higher to 5.78 higher)	‐	890 (5 studies)	Moderateb	Although people had less activity limitation with early weight‐bearing, this did not include a clinically meaningful difference, based on an MCID of 10.
HRQoL Measured using SF‐36 (100‐point scale), SF‐12 (100‐point scale), EQ‐5D (VAS; 100‐point scale) and EQ‐5D‐5L (utility scale; ‐0.224 to 1). Higher scores in all scales indicate a better outcome Follow‐up: short‐term ≤ 6 months. In the included studies, follow‐up was at 10 weeks, 12 weeks, 4 months and 6 months.	In the delayed weight‐bearing group, mean scores were: 72.4 to 74.9 (SF‐36), 49 (SF‐12), 76.3 (EQ‐5D VAS), and 0.69 (EQ‐5D‐5L)	SMD 0.15 higher (0.01 lower to 0.3 higher)	‐	739 (5 studies)	Lowc	We re‐expressed these data onto the EQ‐5D scale (MD 0.03, 95% CI ‐0.00 to 0.06). This effect estimate was imprecise. Based on an MCID for EQ‐5D of 0.075, this did not include a clinically important difference.
Participant satisfaction Measured using a 10‐point VAS (higher scores indicate greater satisfaction) Follow‐up: short‐term ≤ 6 months. In the included study, follow‐up was at 6 months.	Mean satisfaction score in the delayed weight‐bearing group was 9.2	MD ‐0.10 lower (0.99 lower to 0.79 higher)	‐	45 (1 study)	Very lowd
Pain Measured using a 10‐point VAS (higher scores indicate less pain) Follow‐up: short‐term follow‐up ≤ 6 months. In the included studies, follow‐up was at 10 weeks and 6 months.	In 1 study, mean pain scores were 0. In another study, mean pain scores were 0.6	See comment	‐	88 (2 studies)	Very lowe	We did not pool data owing to substantial statistical heterogeneity. In 1 study, people in the delayed weight‐bearing group may have had less pain (MD 1.00, 95% CI 0.23 to 1.77; 43 participants). In another study, there may have been little or no difference in pain between groups (MD ‐0.10, 95% CI ‐0.61 to 0.41). In both cases, the evidence was very uncertain.
Adverse events Unplanned return to theatre	Study population		RR 0.50 (0.09 to 2.68)	1007 (7 studies)	Lowc
	55 per 1000	28 per 1000 (5 to 149)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; EQ‐5D(‐5D): EuroQol‐5 Dimension; HRQoL: health‐related quality of life; MCID: minimal clinically important difference; MD: mean difference; OMAS: Olerud and Molander Ankle Score; RR: risk ratio; SF‐36 (12): Short Form Health Survey ‐ 36‐(or 12‐)item; SMD: standardised mean difference; VAS: visual analogue scale
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aData derived from delayed weight‐bearing groups in the included studies
^bWe downgraded by one level for risk of bias because the included studies were at high risk of bias in at least one domain.
^cWe downgraded by one level for risk of bias because the included studies were at high risk of bias in at least one domain. We also downgraded one level for imprecision because the CI included the possibility of benefit or no benefit for both early and delayed weight‐bearing.
^dWe downgraded by one level for risk of bias because the included study was at high risk of detection bias. We downgraded by two levels for imprecision because the data were derived from very few participants and the CI included the possibility of benefit or no benefit for both early and delayed weight‐bearing. 
^eWe downgraded by one level for risk of bias because the included studies were at high risk of detection bias. We downgraded by one level for imprecision because the data were derived from very few participants. We also downgraded by one level for inconsistency because the findings in each study were not consistent; we noted substantial levels of statistical heterogeneity and we did not pool these data.

Summary of findings 2. Removable ankle support compared with non‐removable ankle support for ankle fracture in adults.

Removable ankle support compared with non‐removable ankle support for ankle fracture in adults
Population: adults with ankle fracture Setting: hospitals, orthopaedic clinics Intervention: removable ankle support Comparison: non‐removable ankle support (such as a cast)
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with non‐removable ankle support (cast)	Risk with removable ankle support
Activity limitation Measured using OMAS (100‐point scale, higher scores indicate better function) Follow‐up: short‐term follow‐up, range from 3 to 6 months	After surgical fracture management		‐	677 (6 studies)	Lowa	This effect estimate included both a clinically important and unimportant improvement with a removable ankle support.b
	For non‐removable ankle support, mean scores using OMAS ranged from 48.8 to 86	MD 6.39 higher (1.69 higher to 11.09 higher)
	After non‐surgical fracture management		‐	399 (3 studies)	Lowc	This effect estimate did not include a clinically important difference between treatments.
	For non‐removable ankle support, mean scores using OMAS ranged from 69 to 79.55	MD 1.08 higher (3.18 lower to 5.34 higher)
Health‐related quality of life Measured using EQ‐5D (scale from ‐0.224 to 1, or unknown modified scale), and SF‐36 (100‐point scale); higher scores in all scales indicate better HRQoL Follow‐up: short‐term follow‐up, range from 3 to 6 months	After surgical fracture management		‐	477 (3 studies)	Moderated	We re‐expressed this onto the EQ‐5D scale (MD 0.05, 95% CI 0.02 to 0.08; favours removable support). This effect estimate included both a clinically important and unimportant improvement with a removable ankle support.e
	For non‐removable ankle support, mean score using EQ‐5D ranged from 0.14 to 0.21. Mean score with SF‐36 was 75.5	SMD 0.30 higher (0.11 higher to 0.50 higher)
	After non‐surgical fracture management		‐	397 (3 studies)	Lowf	We re‐expressed this onto the EQ‐5D scale (MD ‐0.01, 95% CI ‐0.04 to 0.03; favours non‐removable support). This effect estimate did not include a clinically important difference.e
	For non‐removable ankle support, mean score using EQ‐5D was 0.175. Mean score with a modified EQ‐5D was 6.27. Mean score with RAND‐36 was 59	SMD 0.04 lower (0.24 lower to 0.15 higher)
Participant satisfaction	‐	Not estimable	‐	‐	‐	No studies reported this outcome
Pain Measured after surgical fracture management as number of people with pain. Measured after non‐surgical fracture management using a VAS (10‐point or 100‐point scale; lower scores indicate less pain). Follow‐up: short‐term follow‐up, range from 3 to 6 months	Study population after surgical fracture management		RR 0.74 (0.52 to 1.07)	29 (1 study)	Very lowg
	923 per 1000	683 per 1000 (480 to 988)
	After non‐surgical fracture management		‐	167 (2 studies)	Very lowg	We re‐expressed this onto a 100‐point scale (MD ‐2.2, 95% CI ‐16.06 to 11.66; favours removable support).
	For non‐removable ankle support, mean score using a 10‐point VAS was 2.18. Mean score using a 100‐point VAS was 21.	SMD 0.10 lower (0.73 lower to 0.53 higher)
Adverse events (Unplanned return to theatre)	Study population after surgical fracture management		RR 1.20 (0.39 to 3.71)	624 (6 studies)	Lowc
	81 per 1000	97 per 1000 (31 to 299)
	Study population after non‐surgical fracture management		RR 0.99 (0.14 to 6.96)	305 (1 study)	Very lowg
	13 per 1000	13 per 1000 (2 to 92)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; EQ‐5D: EuroQol‐5 Dimensions; HRQoL: health‐related quality of life; MCID: minimal clinically important difference; MD: mean difference; OMAS: Olerud and Molander Ankle Score; RR: risk ratio; SF‐36: Short Form Health Survey – 36‐item; SMD: standardised mean difference; VAS: visual analogue scale
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe downgraded by one level for risk of bias and one level for inconsistency because of unexplained statistical heterogeneity (I² = 63%).
^bBased on an MCID of approximately 10 points for OMAS.
^cWe downgraded by one level for risk of bias and one level for imprecision because the CI derived from these studies included both benefit and harm.
^dWe downgraded by one level for risk of bias in the included studies.
^eBased on an MCID of 0.075 points for EQ‐5D.
^fWe downgraded by one level for risk of bias and one level for imprecision because the CI from these small studies included both benefit and no benefit for both treatments.
^gWe downgraded the certainty of the evidence by one level for risk of bias, and two levels for imprecision because the effect estimate was derived from very few participants and included the possibility of benefit and harm for both treatments.

Summary of findings 3. Physical therapy interventions versus usual care or other physical therapy interventions for ankle fracture in adults.

Physical therapy interventions versus usual care or other physical therapy interventions for ankle fracture in adults
Population: adults with ankle fracture Setting: hospitals or physical therapy centres (with therapy sometimes completed at home) Intervention: physical therapy interventions applied either during the initial fracture management period of typically six weeks (after people had surgical treatment for their fracture), or after the initial fracture management period (after people had either surgical or non‐surgical treatment for their fracture) Comparison: usual care or other physical therapy interventions
Outcomes	Impact	№ of participants in each study	Certainty of the evidence (GRADE)	Comments
Activity limitation Ankle function measured using Mazur score, OMAS, FAOS (100‐point scales), or LEFS (80‐point scale). Higher scores in all scales indicate better ankle function. Follow‐up: short‐term ≤ 6 months. In the included studies, follow‐up was at 3 months, 6 months.	During initial fracture management period (after surgical treatment)	Study 1: 44 Study 2: 106 Study 3: 41	Very lowa
	In Study 1, people may have had better ankle function when they used ACM than only usual care physiotherapy (MD 10.80, 95% CI 3.40 to 18.20). In Study 2, people may have had better ankle function after usual care than a spring‐loaded ankle trainer (MD ‐7.10, 95% CI ‐13.64 to ‐0.56). In Study 3, there may have been little or no difference in ankle function for people who used an antigravity treadmill or usual care rehabilitation (MD ‐3.00, 95% CI ‐14.35 to 8.35). In all cases, the evidence was very uncertain.
	After initial fracture management period (after surgical or non‐surgical treatment)	Study 4: 52 Study 5: 95 Study 6: 136 Study 7: 170 Study 8: 105	Very lowb
	In Study 4, there may have been little or no difference in ankle function for people who used a 16‐week progressive exercise programme compared or best practice advice (MD ‐3.84, 95% CI ‐12.63 to 4.95). In Study 5, there may have been little or no difference in ankle function for people who used enhanced physiotherapy exercise sessions or usual care physiotherapy (MD ‐2.00, 95% CI ‐6.95 to 2.95). In Study 6, there may have been little or no difference in ankle function for people who used additional stretches as part of physiotherapy programme or standard home exercise programme (MD 0.81, 95% CI ‐3.83 to 5.45). In Study 7, there may have been little or no difference in ankle function for people who attended a supervised exercise class or advice only (MD 0.00, 95% CI ‐3.59 to 3.59). In Study 8, there may have been little or no difference in ankle function for people who used a 12‐week programme of bi‐weekly physiotherapy sessions or usual care (MD ‐0.80, 95% CI ‐9.68 to 8.08). In all cases, the evidence was very uncertain.
HRQoL Measured using FAOS subscale (100‐point scale), EQ‐5D (‐0.224 to 1), AQoL (45‐point scale), SF‐36 (100‐point scale). Higher scores better quality of life. Follow‐up: short‐term ≤ 6 months. In the included studies, follow‐up was at 3 to 6 months.	During initial fracture management period (after surgical treatment)	Study 3: 41 participants	Very lowc
	In Study 3, there may have been little or no difference in quality of life for people who used an antigravity treadmill or usual care rehabilitation (MD ‐2.00, 95% CI ‐15.17 to 11.17), but the evidence was very uncertain.
	After initial fracture management period (after surgical or non‐surgical treatment)	Study 4: 56 Study 5: 94 Study 7: 169 Study 8: 105	Very lowa
	In Study 4, there may have been little or no difference in quality of life for people who used a 16‐week progressive exercise programme or best practice advice (MD 0.04, 95% CI ‐0.07 to 0.15). In Study 5, people who used enhanced physiotherapy sessions may have had better quality of life than those who used usual care physiotherapy (MD 1.49, 95% CI 0.14 to 2.84). In Study 7, there may have been little or no difference in quality of life for people who attended a supervised exercise class or advice only (MD 0.10, 95% CI ‐1.12 to 1.32). In Study 8, there may have been little or no difference in quality of life for people who used a 12‐week programme of bi‐weekly physiotherapy sessions or usual care (MD 1.70, 95% CI ‐2.45 to 5.85). In all cases, the evidence was very uncertain.
Participant satisfaction Measured using VAS (100‐point scale; higher scores indicate greater satisfaction) Follow‐up: short‐term ≤ 6 months. In the included studies, follow‐up was at the end of study follow‐up.	During initial fracture management period (after surgical treatment)	‐	‐	No studies reported this outcome
	‐
	After initial fracture management period (after surgical or non‐surgical treatment)	Study 5: 94 Study 6: 138	Very lowa
	In Study 5, there may have been little or no difference in people's satisfaction with treatment between those who used enhanced physiotherapy exercise sessions or usual care physiotherapy (MD 6.50, 95% CI ‐0.12 to 12.26). In Study 6, there may have been little or no difference in people's satisfaction with treatment between those who used additional stretches as part of a physiotherapy programme or standard home exercise programme (MD 2.20, 95% CI ‐4.35 to 8.75). In both cases, the evidence was very uncertain.
Pain Measured using VAS (100‐point scale, and 10‐point scale), FAOS subscale (100‐point scale). Lower scores indicate less pain Follow‐up: short‐term ≤ 6 months. In the included studies, follow‐up was at 3 months and 6 months.	During initial fracture management period (after surgical treatment)	Study 2: 106 Study 3: 41	Very lowd
	In Study 2, there may have been little or no difference in pain for people who used a spring‐loaded ankle trainer versus usual care (MD 0.00, 95% CI ‐0.59 to 0.59). In Study 3, there may have been little or no difference in pain for people who used an antigravity treadmill versus usual care rehabilitation (MD ‐4.00, 95% CI ‐16.86 to 8.86). In both cases, the evidence was very uncertain.
	After initial fracture management period (after surgical or non‐surgical treatment)	Study 4: 53 Study 5: 94 Study 6: 135 Study 7: 171	Very lowb
	In Study 4, there may have been little or no difference in pain for people who used a 16‐week progressive exercise programme compared to best practice advice (MD 1.70, 95% CI ‐8.26 to 11.66). In Study 5, there may have been little or no difference in pain on equal weight‐bearing for people who used enhanced physiotherapy exercise sessions versus usual care physiotherapy (1.10, 95% CI ‐3.90 to 6.10). In Study 6, there may have been little or no difference in pain on stair descent for people who used additional stretches as part of physiotherapy programme versus standard home exercise programme (MD ‐2.10, 95% CI ‐8.08 to 3.88). In Study 7, there may have been little or no difference in pain on equal weight‐bearing for people who attended a supervised exercise class versus advice only (MD 0.00, 95% CI ‐0.30 to 0.30). In all cases, the evidence was very uncertain.
Adverse events Unplanned return to theatre	During initial fracture management period (after surgical treatment)	‐	‐	No studies reported this outcome
	‐
	After initial fracture management period (after surgical or non‐surgical treatment)	‐	‐	No studies reported this outcome
	‐
Study 1: regular ACM using CAMOped device (20 minutes a day in own home) alongside usual care physiotherapy versus usual care physiotherapy. Study 2: spring‐loaded ankle trainer (3 daily sessions with 10 minutes of use, for first 3 weeks after surgery) versus usual care physiotherapy. Study 3: antigravity treadmill (20 minutes, 2 to 3 times a week for 6 weeks) versus standard physiotherapy. Study 4: progressive exercise programme (up to 6 face‐to‐face physiotherapy sessions, with home exercise and advice programme for 16 weeks) versus best practice advice. Study 5: joint mobilisation using enhanced physiotherapy (2 sessions a week for 4 weeks, with progression as advised by physical therapist) versus usual care physiotherapy. Study 6: stretching exercises of either a long duration (30 minutes of stretches a day) or short duration (6 minutes of stretches per day) in addition to home exercise programme (with up to 5 sessions with a physiotherapist) versus home exercise programme. Study 7: individually tailored, prescribed, supervised and progressed exercise programme versus advice only. Study 8: physiotherapy exercise programme using neuromuscular exercises (2 physiotherapy sessions a week for 12 weeks, plus home exercise advice) versus usual care. ACM: active controlled motion; AQoL: Assessment of Quality of Life score; CI: confidence interval; EQ‐5D: EuroQol 5‐Dimensions; FAOS: Foot and Ankle Outcome Score; LEFS: Lower Extremity Functional Scale; MD: mean difference; OMAS: Olerud and Molander Ankle Score; SF‐36: Short Form Health Survey – 36‐item; VAS: visual analogue score
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aWe downgraded by one level for risk of bias, one level for imprecision because the data for each therapy type were derived from few participants, and one level for inconsistency because differences between treatments and comparisons meant that we could not pool these data.
^bWe downgraded by one level for risk of bias, and two levels for imprecision because the data for each therapy type were derived from few participants and the CI included benefits for both intervention and usual care treatments. These data also include inconsistencies in design, meaning that we were unable to combine data in meta‐analysis.
^cWe downgraded by one level for risk of bias and two levels for imprecision because the evidence for this outcome included very few participants.
^dWe downgraded by one level for risk of bias, and two levels for imprecision because the data for each therapy type were derived from few participants and the CI included benefits for both intervention and usual care treatments.

Background

Description of the condition

Ankle fracture is a term used to describe a fracture of the distal tibia, fibula, or both (Wire 2023). It is one of the most common lower limb fractures affecting both sexes and people of all ages, with the incidence reported increasing over the past few decades (Rydberg 2023). Studies report the incidence of ankle fractures as being between 42 and 187 per 100,000 people (Court‐Brown 2006; Scheer 2020). The most common causes of ankle fracture are twisting injuries and falls, followed by sports injuries (Scheer 2020; Wire 2023). The peak prevalence of ankle fractures occurs in adolescents and young adults before rising again at 55 years (Rydberg 2023; Scheer 2020). It is also expected that, due to the ageing population, ankle fractures will increase by 300% in the elderly (So 2023).

Depending mainly on their severity, ankle fractures are treated surgically or conservatively (i.e. non‐surgically), usually followed by a period of ankle immobilisation (Wire 2023). Ankle fractures can be classified as stable or unstable (Lampridis 2018). An unstable fracture is where the ankle bones dislocate and the talus no longer remains aligned under the tibia. Unstable fractures typically undergo surgical management, which generally comprises open reduction to reposition the fractured parts followed by internal fixation of the fracture using osteosynthesis material such as wires, screws, and other devices. After surgery, the ankle can be immobilised in a cast, splinted, bandaged, or left unsupported, depending on the surgeon's preference.

People with stable fractures are usually treated with non‐surgical management, where the fracture is just supported by a cast or splint immobilisation while it heals. However, immobilisation can result in decreased range of motion or muscle atrophy, which can impede recovery and rehabilitation (Yamauchi 2017; Yoshiko 2018). Even short periods of immobilisation, such as five days, can cause substantial muscle atrophy (Wall 2014).

In addition, ankle fracture can be accompanied by other injuries, which may further impair recovery. These injuries include damage to the cartilage and ligaments, the existence of free bodies within the intra‐articular space, and diastasis of the distal tibiofibular joint (Williams 2021). It has been suggested that these concomitant injuries lead to poor functional outcomes, pain, joint stiffness, and swelling of the ankle or lower limb, and reduced participation in sport (Gonzalez 2016).

Description of the intervention

Several rehabilitation interventions are used to address the sequelae of ankle fracture and immobilisation (holding the ankle still while it heals). These interventions may begin during the initial fracture management period, which is typically the first six weeks after injury and whilst the ankle may still be immobilised. Alternatively, rehabilitation interventions may commence only after this initial period.

Methods of ankle immobilisation include casts (custom‐made immobilisation made of plaster or synthetic material), backslabs (half casts), and orthoses (commercially‐available braces). We considered these as methods of rehabilitation for this review. These interventions may facilitate or restrict the commencement of other rehabilitation interventions, such as weight‐bearing and exercise, during the initial fracture management period.

Rehabilitation protocols exist which may recommend early weight‐bearing (defined as weight‐bearing within three weeks of injury). People are often offered crutches or walking aids during this initial period. Different exercises can be performed, ranging from passive or active range of movement, or structured exercise programmes, and people may be encouraged to perform exercises either during or after the initial fracture management period.

How the intervention might work

The initial surgical or non‐surgical management of an ankle fracture aims to restore the anatomy and stability of the ankle. Ankle immobilisation can support and protect the fracture site during early healing. It is used to minimise the risk of non‐union or malunion at the fracture site. However, ankle immobilisation also increases the risk of ankle stiffness, muscle weakness, ankle and lower limb swelling, and residual ankle pain. Whilst immobilised, weight‐bearing is often either restricted or not permitted, with consequent reduction in general mobility and functioning. Rehabilitation is aimed at improving participation in work, caring and leisure activities, restoring mobility, and reducing pain and other impairments.

Why it is important to do this review

This is an update of a Cochrane review first published in 2008 and updated in 2012 (Lin 2008a; Lin 2012). In the most recent version of this review, we identified 38 studies with limited evidence supporting early weight‐bearing and the use of a removable ankle support to allow exercise during the initial fracture management period after surgical fixation. There was little evidence for interventions given during the initial fracture management period with non‐surgical approaches, and no evidence for physical therapy interventions compared to usual care.

In this review, we incorporate new studies to update the evidence for rehabilitation after ankle fracture in adults.

Objectives

To assess the effects of rehabilitation interventions following surgical or non‐surgical management of ankle fractures in adults.

Methods

Criteria for considering studies for this review

Types of studies

We included only randomised controlled trials and quasi‐randomised controlled trials (in which the method of allocating participants to a treatment is not strictly random, e.g. by date of birth, hospital record number, or alternation).

Types of participants

We included studies involving adults of either sex who had attended a hospital or community setting for rehabilitation following an ankle fracture. Allocation to the treatment group must have been within three months of the ankle fracture, and participants could have had either surgical or non‐surgical orthopaedic management. We anticipated that the timing of group allocation was likely to differ according to the type of intervention and the time at which the intervention was likely to take place. For studies investigating early weight‐bearing or removable ankle supports, we expected that allocation would occur within two weeks of initial orthopaedic management (i.e. during the initial fracture management period, which is typically the first six weeks of injury). For other interventions, it is possible that group allocation could occur during or after the initial fracture management period.

We excluded studies conducted exclusively on participants with multi‐trauma, pathological fracture, or established complications secondary to ankle fracture (e.g. non‐union). We also excluded studies conducted exclusively on people with tibial pilon or plafond fractures, which are defined as intra‐articular fracture of the distal tibia, or fracture of the talus. We included studies with mixed populations of adults and children if data for the adults were reported separately.

Types of interventions

We included any rehabilitation intervention (e.g. early weight‐bearing, different orthoses, ankle exercises, manual therapy, stretching, electrotherapy) employed by any health professional (e.g. physiotherapist, doctor) started at or after orthopaedic management (i.e. surgical or non‐surgical fracture management).

Included interventions could be compared with placebo, no treatment, or another rehabilitation intervention. We excluded studies if they occurred before orthopaedic management, or examined pharmacological or surgical intervention alone.

In this review, we were interested in the following intervention comparisons.

• Early weight‐bearing versus delayed weight‐bearing

• Removable ankle support versus non‐removable ankle support (e.g. cast)

• Physical therapy interventions versus usual care or other physical therapy interventions

We stratified studies according to whether participants had surgical or non‐surgical fracture management. If studies did not limit their criteria by fracture management (i.e. included participants who had either type of fracture management), we also planned to include stratification for a mixed fracture management approach.

For physical therapy interventions, we also stratified the data according to whether the interventions were given during the initial treatment management period (usually the first six weeks after injury when people may have a removable or non‐removable ankle support) or after the initial treatment management (usually starting after six weeks when a non‐removable ankle support may be removed).

We also collected data for other interventions that did not fit into these main comparison groups. We grouped the data for these interventions under the subheading 'Other rehabilitation interventions', which we then divided into electrotherapy and other interventions.

Types of outcome measures

Primary outcomes

• Activity limitation as a measure of ankle function, using questionnaires or performance tests. Questionnaires were our preferred primary outcome measure in instances where activity limitation was measured by a questionnaire and a performance test, because questionnaires tend to have more robust measurement properties (Andersson 2010). Possible questionnaires include the self‐reported Olerud and Molander Ankle Score (OMAS; Olerud 1984) and the Lower Extremity Functional Scale (LEFS; Binkley 1999). Possible performance tests include timed walking or stair‐climbing tasks (Moseley 2005).

Secondary outcomes

• Health‐related quality of life, measured with questionnaires such as the EuroQol‐5 Dimensions (EQ‐5D), Short Form Health Survey (SF‐36; Ware 1992), and Assessment of Quality of Life (AQoL; Hawthorne 2001).

• Participant satisfaction, measured with questionnaires as defined by study authors.

• Pain, measured with questionnaires such as the numerical rating scale (Downie 1978), visual analogue scales (VAS; Huskisson 1974), or McGill Pain Questionnaire (Melzack 1975).

• Adverse events. We included any adverse event reported by study authors (e.g. malunion or non‐union, pain resulting from cessation of intervention). In this review update, we also noted whether any participants had an unplanned return to theatre. If we suspected that withdrawals or dropouts may have been related to adverse events, we contacted study authors for more information.

We collected outcome data at the following three time points for both primary and secondary outcomes, where relevant.

• At the end of treatment (i.e. the time during which the intervention was applied)

• In the short term (≤ six months after initial injury)

• In the long term (> six months after initial injury)

Economic data

We tabulated economic data if available, but did not analyse these data due to differences in costs and currencies over time and between countries.

Search methods for identification of studies

Electronic searches

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) via CRS‐Web (29 March 2023, Issue 3); MEDLINE Ovid (MEDLINE and Epub Ahead of Print, In‐Process, In‐Data‐Review & Other Non‐Indexed Citations, Daily and Versions (1946 to 29 March 2023)); Embase Ovid (1980 to 29 March 2023); CINAHL (Cumulative Index to Nursing and Allied Health Literature) Plus (5 May 2022); AMED (Allied and Complementary Medicine; 5 May 2022); and SPORTDiscus (5 May 2022). For this update, the searches were limited to 2011 onwards. We did not place any restrictions on language or publication status. At the time of the March 2023 searches, CENTRAL was fully up‐to‐date with all records from the Cochrane Bone, Joint and Muscle Trauma (BJMT) Group Specialised Register and so it was not necessary to search this database separately.

Search strategies are shown in Appendix 1 for CENTRAL, MEDLINE, Embase, CINAHL, AMED, and SPORTDiscus. For MEDLINE, the subject specific search was combined with the sensitivity‐maximising version of the 'Cochrane Highly Sensitive Search Strategy for identifying randomised trials' (Lefebvre 2023). In this update, we amended our original search strategies slightly with assistance from the Cochrane BJMT Group due to changes in MeSH terms or keywords (for CENTRAL, MEDLINE, Embase, and AMED) or the database platform (for CINAHL and SPORTDiscus). Details of the search strategies used in the previous version of the review are reported in Lin 2012.

We also searched for unpublished or ongoing studies via the WHO International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov (29 March 2023). See Appendix 1.

Searching other resources

We also checked the reference lists of included studies and relevant systematic reviews.

Data collection and analysis

Selection of studies

Two review authors (SL, RP, or MP) independently and in duplicate screened the titles and abstracts of all records identified through databases (and other) searches. We obtained the full text for all studies that appeared to meet the inclusion criteria, or for which there were insufficient data in the title and abstract to make a clear decision. Two review authors (SL, RP or MP) independently assessed the full texts to establish whether the studies met the inclusion criteria. We resolved disagreements through discussion and, if necessary, would have involved a third person for resolution.

Data extraction and management

Working independently, two review authors (SL, RP, or MP) performed data extraction, using a form designed for this review. Again, we resolved disagreements through discussion and, if necessary, would have involved a third person for resolution. Review authors did not extract data from studies in which they were involved; these studies were evaluated by other review authors.

Assessment of risk of bias in included studies

We assessed the risk of bias in the included studies using the Cochrane risk of bias tool (Higgins 2011). We assessed the following domains.

• Sequence generation (selection bias)

• Allocation concealment (selection bias)

• Blinding of participants, personnel (performance bias)

• Blinding of outcome assessors (detection bias)

• Incomplete outcome data (attrition bias)

• Selective reporting (reporting bias)

• Other risks of bias

For each domain, two review authors (SL, RP, or MP) judged whether study authors made sufficient attempts to minimise bias in their design. For each domain, we made judgements using three measures – high, low, or unclear risk of bias – and we recorded these judgements in risk of bias tables. We made these judgements independently and reached consensus through discussion; if necessary, we would have involved a third person for resolution. Review authors did not evaluate the risk of bias in studies in which they were involved; these studies were evaluated by other review authors.

Measures of treatment effect

For dichotomous outcomes, we calculated risk ratios (RRs) and 95% confidence intervals (95% CIs). For continuous outcomes, we calculated mean differences (MDs) and 95% CIs if studies used the same outcome measure. If studies used different measures to assess the same outcome, we calculated standardised mean differences (SMDs) and 95% CIs.

In order to aid interpretation of SMDs with statistically significant effect estimates (with a P value < 0.05), we expressed effect estimates as MDs using the most commonly used scale in the pooled studies and guided by Section 15.5.3.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021).

Unit of analysis issues

We anticipated that potential unit of analysis issues would most commonly arise due to having more than two treatment groups in one study, and reporting of adverse events by the number of events rather than number of participants with these outcomes. If a study had more than two groups within a meta‐analysis, we combined groups to create a single pair‐wise comparison and, if appropriate, split the data from a shared comparison group to create more than one smaller comparison group. For adverse events, we used the number of participants as the unit of measurement in the meta‐analysis.

Dealing with missing data

We attempted to contact study authors where additional information was required. In the case of missing data, we performed available‐case analysis. For studies with missing standard deviations (SDs), we followed the advice of the Cochrane Handbook on methods of imputing data (Higgins 2023). For studies that did not report the group allocation of dropouts, we planned to impute the number of participants used in meta‐analysis by evenly distributing dropouts between groups. If studies did not report dropouts, we assumed there were none.

Assessment of heterogeneity

We assessed statistical heterogeneity by visually inspecting the forest plots and considering the I² statistic and Chi² tests. We used the following interpretations of I² values: 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; and 75% to 100% considerable heterogeneity (Higgins 2021).

Assessment of reporting biases

We planned to investigate the potential for publication bias and explore possible small‐study biases using funnel plots. However, we had insufficient studies (fewer than 10 studies) for all outcomes (Sterne 2017).

To assess outcome reporting bias, we screened clinical trials registers for protocols and registration documents of included studies that were prospectively published, and we sourced all clinical trials register documents that were reported in the study reports of included studies. We used evidence of prospective registration, and the details within these registration documents, to judge whether studies were at risk of selective reporting bias.

Data synthesis

Where possible, and if the studies were clinically homogenous, we performed a meta‐analysis. We used a random‐effects model for all meta‐analyses to account for the likely differences between participants. If studies were clinically and methodologically heterogenous, we did not pool the data and narratively reported the results for each study.

Whenever appropriate, we stratified the analyses according to whether participants had received surgical or non‐surgical fracture management. Although we presented the data according to these two subgroups, we did not test for subgroup interactions. If calculating a different effect measure (MD or SMD) for the two fracture management approaches, we presented the data in different analysis tables.

Subgroup analysis and investigation of heterogeneity

As described above, we stratified data into distinct groups according to the fracture management approach. We did not conduct any formal assessment of subgroup interactions in this review.

Sensitivity analysis

In sensitivity analysis, we explored the impact of decisions made during the review process. We therefore used sensitivity analysis to remove studies from the primary analysis and compare the two sets of findings. Sensitivity analyses were limited to comparisons in which data were pooled (i.e. early weight‐bearing versus delayed weight‐bearing; and removable ankle support versus non‐removable ankle support). We excluded studies that had:

• high or unclear risk of selection bias (for random sequence generation or allocation concealment, or both);

• high risk of detection bias (assessed separately according to whether outcomes were participant‐ or personnel‐assessed);

• high risk of attrition bias (according to outcome‐level and relevant time point data).

In addition, we assessed the impact of using a random‐effects model in analyses and re‐calculated the pooled primary analyses using a fixed‐effect model. Also, we assessed the impact of using a cut‐off of six months (rather than three months) to distinguish between short‐term and long‐term follow‐up (see Differences between protocol and review). Finally, we visually inspected the data for any outliers and excluded these studies in sensitivity analyses.

We reported the results of sensitivity analyses in the review when these results indicated a different interpretation of the effect estimate.

Summary of findings and assessment of the certainty of the evidence

Two review authors (SL and MP) used the GRADE system to assess the certainty of the body of evidence associated with the following outcomes, measured in the short term (≤ six months).

• Activity limitation

• HRQoL

• Participant satisfaction

• Pain

• Adverse events (in particular, we presented the evidence for unplanned return to theatre)

The GRADE approach assesses the certainty of a body of evidence based on the extent to which we can be confident that an estimate of effect or association reflects the item being assessed. Evaluation of the certainty of a body of evidence considers within‐study risk of bias (study limitations), directness of the evidence (indirectness), heterogeneity of the data (inconsistency), precision of the effect estimates (imprecision), and risk of publication bias. The certainty of the evidence could be high, moderate, low, or very low, being downgraded by one or two levels depending on the presence and extent of concerns in each of the five GRADE domains. We used footnotes to describe reasons for downgrading the certainty of the evidence for each outcome, and we used these judgements when drawing conclusions in the review.

Using the GRADE profiler software (GRADEpro GDT), we constructed summary of findings tables for the following three comparisons.

• Early weight‐bearing compared with delayed weight‐bearing

• Removable ankle support compared with non‐removable ankle support

• Physical therapy interventions versus usual care or other physical therapy interventions

Results

Description of studies

Results of the search

The search for this update (search period July 2011 to May 2022; and to March 2023 for CENTRAL, MEDLINE, and Embase) produced 9224 references from the following databases: CENTRAL (1527); MEDLINE (1802); Embase (2998); CINAHL (1283); AMED (99); SPORTDiscus (900 records); the WHO ICTRP (297); and ClinicalTrials.gov (318). We identified no potentially eligible studies from other sources. Once duplicates had been removed, we had a total of 5998 records.

We sourced the full text of 117 records. From these full‐text records, we excluded 43 records. We included 15 new studies (with 48 records), and combined these studies with those previously included (Lin 2012). For this update, we included a total of 53 studies (with 104 records). Eighteen studies (with 18 records) are ongoing, and seven studies (with eight records) are awaiting classification. See Figure 1.

1.

Study flow diagram

Included studies

See Characteristics of included studies. We contacted the study authors to clarify that the following studies were unique: Ahl 1986; Ahl 1987; Ahl 1988; Ahl 1993. We contacted 24 study authors for additional information regarding their study methods, results, or both (Brink 1996; DiStasio 1994; Dogra 1999; Egol 2000; Fitzgerald 1994; Franke 2008; Handolin 2005; Hernandez 2006; Honigmann 2007; Lin 2008b; Matthews 2020; Moseley 2005; Nilsson 2009; Rasmussen 2006; Reed 1998; Romero Zepeda 2008; Seiger 2009; Sondenaa 1986; Stuart 1989; Tropp 1995; Van den Berg 2018; Van Laarhoven 1996; Wetzler 1991; Wilson 1991).

Types of studies and setting

All except eight of the included studies were randomised controlled trials; these eight studies were quasi‐randomised (Davies 1991; DiStasio 1994; Seiger 2009; Siddique 2005; Stockle 2000; Van Laarhoven 1996; Vioreanu 2007; Wetzler 1991). Only one study used a cross‐over design (Keene 2016).

Fifteen were multicentre studies (Bretherton 2024; Brink 1996; Dehghan 2016; Ginandes 1999; Kearney 2021; Keene 2022; Kortegangas 2019; Lin 2008b; Moseley 2005; Moseley 2015; Palke 2022; Park 2021; Siddique 2005; Smeeing 2020; Van Laarhoven 1996). Three studies were pilot studies (Ginandes 1999; Keene 2022; Wilson 1991).

Studies were published between 1986 and 2022, with less than a third published since 2010 (Dehghan 2016; Gorodetskyi 2010; Jansen 2018; Kearney 2021; Keene 2016; Kortegangas 2019; Matthews 2020; Molund 2020; Moseley 2015; Palke 2022; Park 2021; Schubert 2020; Smeeing 2020; Van den Berg 2018). For Bretherton 2024, we included unpublished data supplied by the study authors.

Seven studies were published only as abstracts and, therefore, did not always contain adequate study details (Fitzgerald 1994; Hernandez 2006; Losch 2002; Rasmussen 2006; Reed 1998; Venesmaa 2004; Wetzler 1991).

Types of participants

In total, 4489 participants were recruited in the included trials. Nearly half of the studies reported that they included only participants with closed ankle fracture. Only one study provided information to confirm that a small number of participants had open fractures (4/561 participants in Bretherton 2024). Other studies did not report this information. The size of trials ranged from 12 to 669 participants. Participants' age ranged from a mean of 26 years to 63 years. Keene 2022 recruited only adults who were at least 50 years of age; no studies limited participation to older adults (e.g. > 65 years). Where gender was reported, studies included both female and male participants, with 54% being female participants. We note that participants in DiStasio 1994 were active duty military personnel.

Whilst participants in most studies had surgical fracture management, participants in five studies had non‐surgical fracture management (Brink 1996; Ginandes 1999; Kortegangas 2019; Stuart 1989; Van den Berg 2018), and participants had either surgical or non‐surgical fracture management in six studies (Kearney 2021; Keene 2022; Lin 2008b; Moseley 2005; Moseley 2015; Wilson 1991).

Types of interventions

We reported study data in comparison groups (see below) according to the type of intervention. Five studies were multi‐arm studies (Finsen 1989; Keene 2016; Kortegangas 2019; Moseley 2005; Smeeing 2020; Van Laarhoven 1996), and five studies fit into more than one of our comparison groups (Ahl 1993; Dehghan 2016; Finsen 1989; Smeeing 2020; Van Laarhoven 1996).

Early weight‐bearing versus delayed weight‐bearing

Twelve studies compared participants who were allowed to weight‐bear within the first three weeks after injury (early weight‐bearing) to those who delayed weight‐bearing until the end of the study period, which was typically six weeks after injury (Ahl 1986; Ahl 1987; Ahl 1988; Ahl 1993; Dehghan 2016; Finsen 1989; Honigmann 2007; Park 2021; Schubert 2020; Smeeing 2020; Van Laarhoven 1996; Bretherton 2024).

All of these studies included participants who had surgical fracture management.

Removable ankle support versus non‐removable ankle support

Twenty‐five studies compared the effectiveness of a removable ankle support (which allowed for some mobilisation or range of motion during the initial fracture management period, typically the first six weeks after injury) with a non‐removable ankle support (typically a cast). Of these 25 studies, we included two studies in which no immobilisation was used rather than a removable ankle support (Siddique 2005; Van Laarhoven 1996), and two studies in which no immobilisation was used after the removal of a cast or backslab three days after injury (Finsen 1989; Sondenaa 1986).

Of the studies in this comparison, 21 included participants who had surgical fracture management (Ahl 1993; Davies 1991; Dehghan 2016; DiStasio 1994; Dogra 1999; Egol 2000; Finsen 1989; Franke 2008; Hedstrom 1994; Lehtonen 2003; Losch 2002; Matthews 2020; Rasmussen 2006; Siddique 2005; Sondenaa 1986; Stockle 2000; Tropp 1995; Van Laarhoven 1996; Venesmaa 2004; Vioreanu 2007; Wetzler 1991), and three had non‐surgical fracture management (Kortegangas 2019; Stuart 1989; Van den Berg 2018). One study recruited participants who had either surgical or non‐surgical fracture management (Kearney 2021).

In most studies, participants who used a removable ankle support were encouraged to perform range‐of‐motion exercises each day. In five studies, participants in this group attended additional physiotherapy sessions (Davies 1991; DiStasio 1994; Matthews 2020; Sondenaa 1986; Stockle 2000). In Davies 1991, participants in the intervention group had continuous passive motion for 12 to 15 hours a day for an average of six days before hospital discharge. In Losch 2002, participants in both groups received physiotherapy input, but this was structured in the intervention group and non‐structured for participants with a cast. In Franke 2008, we noted a difference in physical therapy in each group, with additional physiotherapy sessions for participants who used a non‐removable ankle support. We noted that exercise was either not described or was comparable in both intervention and comparison groups in 10 studies (Ahl 1993; Finsen 1989; Fitzgerald 1994; Siddique 2005; Stockle 2000; Stuart 1989; Van den Berg 2018; Van Laarhoven 1996; Venesmaa 2004; Wetzler 1991).

Physical therapy interventions versus usual care or other physical therapy interventions

Nine studies compared different physical therapy interventions. The types of interventions and comparisons in these studies all differed, and we describe them briefly here.

• Jansen 2018: regular active‐controlled motion (ACM) using a CAMOped device alongside usual care physiotherapy versus usual care physiotherapy in participants who had surgical fracture management. Training with ACM started two to five days postoperatively and participants were asked to use the device for 20 minutes a day in their own home.

• Keene 2022: progressive exercise programme with up to six face‐to‐face physiotherapy sessions in outpatient clinics versus best practice advice with one face‐to‐face session with participants who had either surgical or non‐surgical fracture management. Sessions started after the initial fracture management period, and the intervention included a home exercise and advice programme delivered over a 16‐week period.

• Lin 2008b: joint mobilisation using enhanced physiotherapy versus usual care physiotherapy in participants who had either surgical or non‐surgical fracture management. In the intervention group, participants attended two outpatient physiotherapy sessions a week for four weeks. Exercises were individually prescribed and progressed by increasing the number of repetitions, force of application, and range of dorsiflexion.

• Molund 2020: spring‐loaded ankle trainer versus usual care physiotherapy in participants who had surgical fracture management. The intervention started on the first postoperative day and continued for the first three weeks after surgery (at which point, conventional rehabilitation was continued). Participants in the intervention group attended three daily sessions with a physical therapist, with 10 minutes of exercise use; participants were able to adjust the resistance of the ankle trainer to increase the pressure as pain resolved.

• Moseley 2005: stretching exercises of either a long duration (30 minutes of stretches a day) or short duration (six minutes of stretches per day) versus a home exercise programme in participants who had either surgical or non‐surgical fracture management. The intervention included five sessions with a physiotherapist and was in addition to a home exercise programme with prescription, instruction, monitoring, and progression of exercise.

• Moseley 2015: tailored, supervised exercise programme versus advice only for participants who had either surgical or non‐surgical fracture management. Exercises in the intervention group were conducted at home but also in the hospital outpatient physical therapy service, with the intervention individually‐tailored, prescribed, monitored, and progressed by a physical therapist.

• Nilsson 2009: physiotherapy exercise programme using neuromuscular exercises compared with usual care for participants who had surgical fracture management. Participants in the intervention group attended two physiotherapy sessions a week for 12 weeks, and were also advised to complete daily home exercises.

• Palke 2022: antigravity treadmill versus standard physiotherapy for participants who had surgical fracture management. The intervention included manual lymphatic drainage and cryotherapy, with exercise using an antigravity treadmill for 20 minutes two to three times per week for six weeks.

• Wilson 1991: joint mobilisation using enhanced physiotherapy versus usual care physiotherapy for participants who had either surgical or non‐surgical fracture management. The intervention consisted of Kaltenborn‐based manual therapy to the talocrural and talocalcaneal joints, plus other lower limb joints which had been assessed as hypomobile by the physiotherapist; three physiotherapy sessions per week were applied for five weeks.

Other rehabilitation interventions

We also included in the review other comparisons of rehabilitation interventions.

• Five studies assessed electrotherapy interventions. These interventions included inferential therapy (Christie 1990), neurostimulation (Gorodetskyi 2010), ultrasound (Handolin 2005), or electrical stimulation (Hernandez 2006). Participants in these studies received the intervention during the initial treatment period (usually six weeks) after surgical fracture management. In Seiger 2009, thermal pulsed diathermy was compared with non‐thermal pulsed diathermy in participants during the rehabilitation period six to 12 weeks after surgical fracture management.

• Brink 1996: Aircast air‐stirrup versus DonJoy orthosis in participants who had surgical fracture management.

• Ginandes 1999: hypnosis intervention given over 12 weeks versus cast immobilisation to participants who had non‐surgical fracture management.

• Fitzgerald 1994: compression stocking used with a below‐knee plaster cast versus below‐knee plaster cast for six weeks in participants who had surgical fracture management.

• Keene 2016: three‐arm cross‐over study in which participants who had surgical fracture management were treated with each of three types of removable ankle supports (Tubigrip, stirrup brace, and walker boot).

• Reed 1998: wool and crêpe bandage versus below‐knee backslab for up to 48 hours postoperatively, with pain measured at one to two days.

• Romero Zepeda 2008: circular (Robert Jones) bandage versus a backslab for 24 hours after injury, with pain and swelling measured at 24 hours.

Outcomes

In our three main comparison groups (early weight‐bearing versus delayed weight‐bearing; removable ankle support versus non‐removable ankle support; physical therapy interventions versus usual care), data were available from at least one study for most of our review outcomes. Only in the removable versus non‐removable ankle support comparison were no data available for participant satisfaction.

No studies comparing electrotherapy interventions reported quality of life or participant satisfaction. For our other comparison groups, no studies reported quality of life. Data were available, however, for all other outcomes for these comparison groups.

In two of the multi‐arm studies, it was possible to use more than one study arm within the same comparison group (Finsen 1989; Smeeing 2020). To manage this, we selected the intervention for the main analysis which we judged to fit best, as follows.

• In the early weight‐bearing comparison group:

  • for Finsen 1989, we included data in the main analysis for interventions described by study authors as 'cast with weight‐bearing vs cast with no weight‐bearing' and we presented data for 'cast with weight‐bearing vs exercise' in Appendix 2;

  • for Smeeing 2020, we included data in the main analysis for interventions described as 'unprotected weight‐bearing vs unprotected non‐weight‐bearing' and we presented data for 'protected weight‐bearing vs unprotected non‐weight‐bearing' in Appendix 2.

• In the removable ankle support comparison:

  • for Finsen 1989, we included in the main analysis data for 'exercise vs cast with no weight‐bearing' and we presented data for 'exercise vs cast with weight‐bearing' in Appendix 2;

  • in Smeeing 2020, we included in the main analysis data for 'unprotected weight‐bearing vs protected weight‐bearing' and we presented data for 'unprotected non‐weight‐bearing vs protected weight‐bearing' in Appendix 2.

In another multi‐arm study (Moseley 2005), we judged that the intervention arms (stretching exercises for a short duration or for a long duration in addition to a home exercise programme and sessions with a physiotherapist) were sufficiently similar relative to their control group (home exercise programme and physiotherapy sessions without stretching exercises); we combined data for the two stretching groups in analyses.

Data for a third study arm in Kortegangas 2019 did not fit easily into our comparisons, and we presented data for 'three‐week cast versus six‐week cast' in this study in Appendix 2.

We contacted study authors for Kearney 2021 who provided data separately according to the fracture management approach. This allowed us to report outcome data separately for participants who had surgical fracture management and those who had non‐surgical fracture management.

Ten studies reported some data that we could not use to calculate an effect estimate; for example, because data were reported as medians or were reported without distribution/dispersion values (Ahl 1993; Brink 1996; DiStasio 1994; Franke 2008; Hedstrom 1994; Honigmann 2007; Van den Berg 2018; Van Laarhoven 1996; Venesmaa 2004; Wetzler 1991). For completeness, we included these data in appendices. Two studies contributed no data to the review outcomes (Christie 1990; Keene 2016).

Sources of funding and declarations of interest

Four studies reported that they received no commercial or external funding (Jansen 2018; Schubert 2020; Smeeing 2020; Van den Berg 2018), and 17 studies reported funding from independent sources such as research foundations (Bretherton 2024; Dehghan 2016; DiStasio 1994; Ginandes 1999; Gorodetskyi 2010; Kearney 2021; Keene 2016; Keene 2022; Kortegangas 2019; Lehtonen 2003; Lin 2008b; Matthews 2020; Molund 2020; Moseley 2005; Moseley 2015; Nilsson 2009; Palke 2022). In Stuart 1989, the intervention devices (Air‐stirrups) were provided by a manufacturing company. In Handolin 2005, investigators received independent funding and declared that the intervention devices (ultrasound equipment) were provided by a manufacturing company. We could not determine whether the inferential unit in Christie 1990 was also supplied by the manufacturer because of unclear reporting. Fourteen studies declared that the study investigators had no conflicts of interest (Dehghan 2016; Jansen 2018; Keene 2016; Lehtonen 2003; Molund 2020; Moseley 2005; Moseley 2015; Nilsson 2009; Palke 2022; Park 2021; Schubert 2020; Smeeing 2020; Van den Berg 2018; Van Laarhoven 1996), and two studies reported declarations which we did not expect to include conflicts of interest (Kearney 2021; Kortegangas 2019). We noted that the lead author in Gorodetskyi 2010 was a paid consultant for the manufacturer who provided partial funding for the trial. The remaining studies reported no information about their funding sources nor provided declarations about conflicts of interest.

Excluded studies

We excluded 43 articles on assessment of the full texts (see Figure 1). Here, we report the details of 10 key excluded studies (see Characteristics of excluded studies).

We excluded six studies that were listed on clinical trials registers. Two potentially eligible studies had no updates on progress in the last 10 years; we were unsuccessful in attempts to reach the contact authors for these studies (ACTRN12611001096943; ACTRN12613000631707). We identified a completed study of an eligible intervention (orthosis), but contacted the study authors, who confirmed that no participants with ankle fracture were included in the study (DRKS00021487). Three studies were included in the previous version of the review as ongoing, but there have been no updates on the clinical trials register since then, and we excluded them because we assumed that the studies were never undertaken (ACTRN12610000557033; N0055190984; NCT01127776).

Additionally, we excluded two studies because they examined eligible rehabilitation interventions but in a mixed population with lower limb fractures; in these studies, data were not reported separately for ankle fracture (Albin 2019; Gan 2014). We also excluded two studies because participants were not randomised to treatment groups (Büker 2019; Güngör 2022).

Studies awaiting classification

We found seven studies awaiting classification with a total of 784 participants (see Characteristics of studies awaiting classification).

ISRCTN15497399 compared a removable support boot (allowing range of movement) with a below‐knee plaster cast for four weeks. In Khojaly 2021, a walking boot (with instruction to walk weight‐bearing as tolerated immediately and to remove the walking boot four times a day) was compared with instruction to perform ankle range‐of‐motion exercises until referred to outpatient physiotherapy. NCT02609347 compared manual therapy with sham manual therapy. NCT02923479 compared general rehabilitation, and specific ankle rehabilitation using the ARBOT device, with general rehabilitation, specific ankle rehabilitation by a physiotherapist, specific ankle rehabilitation by Biodex System 3 dynamometer, and the ProKin PK254 platform. NCT04235907 evaluated the effects of telerehabilitation with exercise instruction delivered via website (pictures and videos) and weekly follow‐up calls compared with paper‐based instruction and a prescription to visit a physical therapist. Shin 2000 compared early weight‐bearing (with exercise) with late weight‐bearing (cast immobilisation). NCT02779244 compared early weight‐bearing with standard treatment, using a controlled ankle motion (CAM) boot in both groups.

We contacted the study authors of ISRCTN15497399, but they were unable to share summarised data until publication. For Khojaly 2021, NCT02609347, and NCT02923479, we contacted study authors but received no reply. Shin 2000 is published in Korean, and we await translation before inclusion in future review updates.

Ongoing studies

We identified 18 ongoing studies with an estimated recruitment of 1650 participants (see Ongoing studies).

Eight of these studies (estimated recruitment of 699 participants) compare early weight‐bearing with delayed weight‐bearing (IRCT20201124049484N1; NCT03970603; NCT04028414; NCT04429217; NCT04872296; NCT05587842; NTR7143; TCTR20220221004). One study with an estimated 100 participants is comparing removable ankle support with fixed ankle support (NTR4469). Seven studies (estimated recruitment of 746 participants) are comparing variations of physical therapy interventions (ChiCTR2100045058; DRKS00024489; DRKS00024506; ISRCTN11830323; JPRN‐UMIN000047296; NCT05280639; Stinner 2022). The remaining two studies include electrotherapy interventions (Guo 2022; JPRN‐UMIN000039630). The estimated recruitment in Guo 2022 is 60 participants, and is not specified in the other study.

Risk of bias in included studies

Figure 2 presents a summary of the review authors' judgements about each risk of bias item for the included studies; and Figure 3 presents a summary of the risk of bias items across all included studies. We did not assess the risk of bias in Christie 1990 and Keene 2016 because we did not include outcome data for these studies in the review. Additionally, there were no personnel‐assessed outcomes in Bretherton 2024. Thus, the figures include blank spaces for these studies.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study. We did not assess risk of bias in Christie 1990 and Keene 2016 because we did not include outcome data for these studies in the review, and there were no outcomes assessed by personnel in Bretherton 2024; thus, these studies include blank spaces.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

A third of studies did not describe methods used to allocate participants to treatments and the risk of selection bias for sequence generation was unclear. Eight studies (15%) were quasi‐randomised trials, and we judged these to have a high risk of selection bias for sequence generation (Davies 1991; DiStasio 1994; Seiger 2009; Siddique 2005; Stockle 2000; Van Laarhoven 1996; Vioreanu 2007; Wetzler 1991). We judged methods of sequence generation to be adequate in the remaining studies, and judged their risk of selection bias to be low.

For bias owing to methods of allocation concealment, we judged all eight quasi‐randomised trials to have a high risk of bias as well as an additional study in which study authors said that an open random number table had been used (Romero Zepeda 2008). Of the remaining studies, half used methods that were likely to conceal allocation (low risk of bias) whilst the other half reported insufficient information for us to judge the risk of bias.

Blinding

The nature of the interventions meant that studies that involved the participants wearing immobilisation devices, weight‐bearing, exercising, or receiving most forms of physical therapy (e.g. hypnosis, manual therapy) were unable to blind the participant and the therapist; we judged the risk of performance bias to be high in all of these studies. Only two studies implemented both participant and therapist blinding. These studies investigated ultrasound and pulsed shortwave diathermy, respectively (Handolin 2005; Seiger 2009). In one study investigating non‐invasive, interactive neurostimulation, blinding of participants, but not therapists, was possible and implemented (Gorodetskyi 2010).

For detection bias, we judged risk according to whether outcomes were measured by participants or other outcome assessors. In most studies, participants were aware of their treatment group and thus, the risk of detection bias for participant‐reported outcomes was high. As investigators in Gorodetskyi 2010, Handolin 2005, and Seiger 2009 masked participants to their treatment, the risk of detection bias for participant‐reported outcomes was low. For assessment of other outcomes, we judged 12 studies to be at high risk of detection bias because outcome assessors were aware of treatment groups (Ahl 1993; Davies 1991; Egol 2000; Franke 2008; Ginandes 1999; Hernandez 2006; Jansen 2018; Kearney 2021; Matthews 2020; Molund 2020; Palke 2022; Van Laarhoven 1996). Of the remaining studies, half reported that blinding of outcome assessors had been achieved (low risk of bias) whilst the other half reported insufficient information for us to judge the risk of bias.

Incomplete outcome data

Most studies had few participant losses, and losses were sufficiently explained or balanced between groups, or both; we judged attrition bias to be low risk in these studies. In 20 studies, we noted a high percentage of losses (> 10%) which were sometimes not balanced between groups, and we judged the risk of attrition bias to be high in these studies (Ahl 1988; Bretherton 2024; Davies 1991; Finsen 1989; Fitzgerald 1994; Hedstrom 1994; Hernandez 2006; Kearney 2021; Keene 2022; Kortegangas 2019; Lehtonen 2003; Losch 2002; Matthews 2020; Moseley 2005; Nilsson 2009; Palke 2022; Park 2021; Reed 1998; Van den Berg 2018; Wilson 1991). We were less certain of the risk of attrition bias (unclear risk) in three studies because of limitations in reporting (Brink 1996; Dehghan 2016; Wetzler 1991).

Selective reporting

Only nine studies had a protocol or prospective clinical trial registration and had reported outcomes in their published reports as planned (Bretherton 2024; Kearney 2021; Keene 2022; Kortegangas 2019; Lin 2008b; Moseley 2015; Palke 2022; Park 2021; Smeeing 2020); we judged these studies to have a low risk of selective reporting bias. We judged the risk of bias to be high in four studies because the methods in their published reports indicated that outcomes were measured, but data for these outcomes were not reported in the results section of their reports (Ahl 1986; Ahl 1987; Ahl 1988; Nilsson 2009). We were unable to effectively judge the risk of selective reporting bias in the remaining studies because they had no protocols or prospective trial registration.

Other potential sources of bias

We judged eight studies to be at risk of other bias (Fitzgerald 1994; Hernandez 2006; Losch 2002; Matthews 2020; Rasmussen 2006; Reed 1998; Venesmaa 2004; Wetzler 1991). In particular, these studies were all published only as abstracts with some additional information provided from direct contact with study authors. As these studies had not undergone an independent peer‐review process, we judged that the risk of bias was high.

Effects of interventions

See: Table 1; Table 2; Table 3

See Appendix 3 for the range, scores, and direction of effect of the measurement scales used by study authors in these comparisons.

Early weight‐bearing versus delayed weight‐bearing

All studies in this comparison group included participants who were given the intervention after surgical fracture management. We constructed a summary of findings table for data in the short term, at six months or less (Table 1). We did not assess the certainty of the evidence for other time points.

In the following analyses, we did not include data for some study arms from multi‐arm studies that did not easily fit our review criteria for this comparison (Finsen 1989; Smeeing 2020). For completeness, we have included the effect estimates for these data in Appendix 2.

1. Activity limitation

Activity limitation was measured at the end of treatment in five studies (Bretherton 2024; Dehghan 2016; Park 2021; Schubert 2020; Smeeing 2020). At this time point, early weight‐bearing led to improved outcome according to the Olerud and Molander Ankle Score (OMAS; mean difference (MD) 9.14, 95% confidence interval (CI) 4.60 to 13.68; I² = 69%; 5 studies, 901 participants; Analysis 1.1). Based on a minimal clinically important difference (MCID) of approximately 10 points for the OMAS (Penning 2022), this included both a clinically important and unimportant benefit of early weight‐bearing.

1.1. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 1: Activity limitation (using OMAS; at the end of treatment)

The same studies also measured this outcome using OMAS at a short‐term follow‐up of 12 weeks (Park 2021), four months (Bretherton 2024), and six months (Dehghan 2016; Schubert 2020; Smeeing 2020). Findings indicated that there was probably continued improvement after early weight‐bearing (MD 3.56, 95% CI 1.35 to 5.78; I² = 0%; 5 studies, 890 participants; moderate‐certainty evidence; Analysis 1.2). Based on an MCID of approximately 10 points for the OMAS (Penning 2022), this difference did not include a clinically important benefit of early weight‐bearing. We downgraded the certainty of this evidence by one level for risk of bias because the included studies had high risk of bias in at least one domain.

1.2. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 2: Activity limitation (using OMAS; short‐term follow‐up, ≤ 6 months)

At 12 months, there was no longer a difference between groups in activity limitation according to OMAS (SMD 0.10, 95% CI ‐0.07 to 0.27; I² = 27%; 5 studies, 865 participants; Analysis 1.3). These data were reported by Bretherton 2024, Dehghan 2016, Finsen 1989, Park 2021, and Smeeing 2020. We re‐expressed the data onto the OMAS, and found that this effect included no clinically important difference according to weight‐bearing approaches (MD 1.60, 95% CI ‐1.28 to 4.31).

1.3. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 3: Activity limitation (using different measurement tools; long‐term follow‐up, > 6 months)

We did not include data from Ahl 1993, Honigmann 2007, and Van Laarhoven 1996 in these analyses. These studies reported data as medians or as mean values without distribution data. For completeness, we report these data in Appendix 4.

2. Health‐related quality of life (HRQoL)

Quality of life was measured using the physical component score of the SF‐36 (Dehghan 2016; Smeeing 2020), the SF‐12 (Honigmann 2007), as well as the EQ‐5D visual analogue scale (VAS) (Schubert 2020), and the EQ‐5D‐5L utility score (Bretherton 2024).

People had better quality of life at the end of treatment after early weight‐bearing (SMD 0.39, 95% CI 0.24 to 0.55; I² = 0%; 4 studies, 655 participants; Analysis 1.4). In this analysis, we included data from Bretherton 2024, Dehghan 2016, Honigmann 2007, and Schubert 2020. We re‐expressed this onto the EQ‐5D‐5L scale (MD 0.09, 95% CI 0.05 to 0.12). Based on an MCID for EQ‐5D of 0.075 (Walters 2005), this includes both a clinically important and unimportant improvement in HRQoL with early weight‐bearing.

1.4. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 4: HRQoL (using different measurement tools; at end of treatment)

In the short term, this outcome was measured at 10 weeks (Honigmann 2007), 12 weeks (Smeeing 2020), four months (Bretherton 2024), and six months (Dehghan 2016; Schubert 2020). At this time point, there may be little or no difference in quality of life between treatment groups (SMD 0.15, 95% CI ‐0.01 to 0.30; I² = 5%; 5 studies, 739 participants; low‐certainty evidence; Analysis 1.5). We re‐expressed this onto the EQ‐5D‐5L scale (MD 0.03, 95% CI ‐0.00 to 0.06). Based on an MCID for EQ‐5D of 0.075 (Walters 2005), any difference did not include a clinically important benefit of early weight‐bearing. We downgraded the certainty of the evidence by one level for risk of bias because the included studies had high risk of bias in at least one domain, and one level for imprecision because the CI included the possibility of benefit or no benefit for both early and delayed weight‐bearing.

1.5. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 5: HRQoL (using different measurement tools; short‐term follow‐up, ≤ 6 months)

At 12 months, there was no longer a difference between groups in quality of life (SMD 0.11, 95% CI ‐0.21 to 0.43; I² = 64%; 3 studies, 641 participants; Analysis 1.6). In this analysis, we included data from Bretherton 2024, Dehghan 2016, and Smeeing 2020. We re‐expressed this onto the EQ‐5D‐5L scale, which showed no clinically important differences according to time of weight‐bearing (MD 0.02, 95% CI ‐0.04 to 0.09).

1.6. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 6: HRQoL (using different measurement tools; long‐term follow‐up, > 6 months)

We did not include data from Franke 2008 in these analyses, as this study reported data using median values. For completeness, we report these data in Appendix 4.

3. Participant satisfaction

One study reported participant satisfaction using a VAS (Schubert 2020). There may be little or no difference in participant satisfaction at the end of treatment (MD 0.40, 95% CI ‐0.48 to 1.28; 48 participants), or after a short‐term follow‐up of six months, but the evidence is very uncertain (MD ‐0.10, 95% CI ‐0.99 to 0.79; 45 participants; very low‐certainty evidence). See Analysis 1.7. We downgraded the certainty of the evidence for the short‐term follow‐up data by one level for risk of bias because the included study was at high risk of detection bias, and two levels for imprecision because the data were derived from very few participants and the CI included the possibility of benefit or no benefit for both early and delayed weight‐bearing.

1.7. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 7: Participant satisfaction (VAS, /10; end of treatment, and short‐term follow‐up of ≤ 6 months)

No studies reported this outcome at a longer‐term follow‐up.

4. Pain

Two studies reported pain using a VAS (Honigmann 2007; Schubert 2020). Findings were imprecise when measured at the end of treatment (MD 0.16, 95% CI ‐0.28 to 0.59; I² = 0%; 2 studies, 92 participants; Analysis 1.8).

1.8. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 8: Pain (VAS; end of treatment)

At short‐term follow‐up, data included substantial levels of statistical heterogeneity (I² = 82%), which we were unable to explain, and we did not pool these data. In Honigmann 2007, people in the delayed weight‐bearing group may have had less pain at 10 weeks than those in the early weight‐bearing group (MD 1.00, 95% CI 0.23 to 1.77; 43 participants). In Schubert 2020, there may have been little or no difference between groups in pain at six months (MD ‐0.10, 95% CI ‐0.61 to 0.41; 45 participants). In both cases, the evidence is very uncertain. See Analysis 1.9. We judged the evidence at short‐term follow‐up to be of very low certainty. We downgraded by one level for risk of bias because the included studies were at high risk of detection bias. We also downgraded by one level for imprecision because the data were derived from very few participants, and one level for inconsistency owing to differences in findings between studies and unexplained statistical heterogeneity.

1.9. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 9: Pain (VAS; short‐term follow‐up, ≤ 6 months)

No studies reported this outcome at a longer‐term follow‐up.

5. Adverse events

We found imprecise findings for all types of adverse events. We used the GRADE approach for unplanned return to theatre, and included these data in the summary of findings table. We downgraded the certainty of the evidence for this adverse event by one level for risk of bias, and one level for imprecision because the effect estimate included the possibility of benefit and harm for both treatments. See Analysis 1.10.

1.10. Analysis.

Comparison 1: Early versus delayed weight‐bearing, Outcome 10: Adverse events

• Unplanned return to theatre (re‐operation; RR 0.50, 95% CI 0.09 to 2.68; I² = 68%; 7 studies, 1007 participants; low‐certainty evidence).

• Redislocation (RR 0.73, 95% CI 0.04 to 13.21; I² = 44%; 3 studies, 137 participants).

• Delayed union (Not estimable due to no events in either group; 1 study, 167 participants)

• Non‐union (RR 0.33, 95% CI 0.01 to 8.18; 2 studies, 728 participants).

• Malunion (RR 3.01, 95% CI 0.12 to 73.59; 1 study, 561 participants)

• Loss of fixation (RR 4.01, 95% CI 0.45 to 35.69; 2 studies, 663 participants).

• Ankle metalwork or implant breakage (RR 3.01, 95% CI 0.12 to 73.59; 1 study, 561 participants).

• Loss of reduction (Not estimable due to no events in either group; 2 studies, 269 participants).

• Deep vein thrombosis (DVT; RR 1.25, 95% CI 0.28 to 5.67; I² = 0%; 3 studies, 687 participants).

• Symptomatic pulmonary embolism (RR 3.01, 95% CI 0.12 to 73.59; 1 study, 561 participants).

• Vascular injury (RR 3.01, 95% CI 0.12 to 73.59; 1 study, 561 participants).

• Deep infection (Not estimable due to no events in either group; 3 studies, 200 participants).

• Surgical site infection (RR 1.25, 95% CI 0.60 to 2.63; 1 study, 561 participants).

• Low grade infection or superficial infection (RR 1.83, 95% CI 0.89 to 3.76; I² = 0%; 8 studies, 1017 participants).

• Wound dehiscence (RR 1.33, 95% CI 0.47 to 3.80; I² = 0%; 3 studies, 749 participants).

• Wound healing complications (RR 3.06, 95% CI 0.33 to 28.46; 1 study, 107 participants).

• Prominent implant or metalwork irritation (RR 1.16, 95% CI 0.56 to 2.39; 1 study, 561 participants).

• Swelling, skin irritation, and/or blistering (RR 0.95, 95% CI 0.10 to 8.78; I² = 0%; 2 studies, 74 participants).

• Dystrophy/atrophy (RR 0.29, 95% CI 0.01 to 7.02; 2 studies, 122 participants).

• Complex regional pain syndrome (RR 0.11, 95% CI 0.01 to 2.06; 1 study, 561 participants).

• Fall (RR 0.33, 95% CI 0.01 to 8.18; 1 study, 561 participants).

• Re‐fracture of the same ankle (RR 2.01, 95% CI 0.18 to 22.01; 1 study, 561 participants).

• Nerve palsy (RR 0.75, 95% CI 0.17 to 3.33; 1 study, 561 participants).

Economic data

One study reported costs for this comparison (Bretherton 2024); we briefly summarise these in Appendix 5.

Sensitivity analysis

Risk of bias. For each outcome, we excluded studies at high or unclear risk of selection bias. We also, separately, excluded studies at high risk of attrition bias. All studies were at high risk of detection bias and this sensitivity analysis was not feasible. For Analysis 1.3, we found that the effect estimate favoured early weight‐bearing when including only studies at low risk of both random sequence generation and allocation concealment (SMD 0.16, 95% CI 0.01 to 0.32; 2 studies, 646 participants). We found no other differences in our interpretation of findings as a result of this sensitivity analysis.

Effects model. We re‐analysed all pooled data using a fixed‐effect model, and found no difference in our interpretation of the findings of the primary analyses.

Different thresholds for short‐term and long‐term data. We re‐analysed all pooled data using a threshold of three months or less for short‐term follow‐up, and assessment after three months for long‐term data. We found no difference in our interpretation of the findings of the primary analyses.

Outliers. Pooled data for this comparison included no outlying data.

Removable ankle support versus non‐removable ankle support

In this comparison, we report the data separately according to the type of fracture management approach. We constructed a summary of findings table for data in the short term at six months or less (Table 2). We did not assess the certainty of the evidence for other time points.

In the following analyses, we did not include some study arms from multi‐arm studies that did not easily fit our review criteria for this comparison (Finsen 1989; Kortegangas 2019; Smeeing 2020). For completeness, we have included the effect estimates for these data in Appendix 2.

1. Activity limitation

Interventions given after surgical fracture management

This outcome was measured using the OMAS in most studies, except for Egol 2000 (Mazur score), Tropp 1995 (modified score), and Rasmussen 2006 (type of activity questionnaire was not specified). We noted one obvious outlier on visual inspection of the initial pooled data (Rasmussen 2006). As this study was published only as an abstract and the type of questionnaire was unknown, we investigated the impact of including this study in analysis using sensitivity analysis and judged that the data were more statistically consistent without Rasmussen 2006. We therefore report these data without Rasmussen 2006.

At the end of treatment, people had less activity limitation after treatment with removable ankle support (SMD 0.36, 95% CI 0.12 to 0.59; I² = 51%; 7 studies, 682 participants; Analysis 2.1). We re‐expressed this data onto the OMAS (MD 6.60, 95% CI 2.20 to 10.81). Based on an MCID of approximately 10 points for the OMAS (Penning 2022), this included both a clinically important and unimportant difference when a removable ankle support was used.

2.1. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 1: Activity limitation after surgical fracture management (using different measurement tools; end of treatment)

This outcome was also measured in the short term at 12 weeks (Matthews 2020; Siddique 2005; Vioreanu 2007), 16 weeks (Kearney 2021), 24 weeks (Rasmussen 2006), and six months (Dehghan 2016; Smeeing 2020). For consistency with decisions made regarding the inclusion of Rasmussen 2006 in the analysis at end of treatment, we also excluded Rasmussen 2006 from this meta‐analysis. At short‐term follow‐up, people may have a better outcome after treatment with a removable ankle support (MD 6.39, 95% CI 1.69 to 11.09; I² = 63%; 6 studies, 677 participants; low‐certainty evidence; Analysis 2.2). Based on an MCID of approximately 10 points (Penning 2022), this effect estimate included both a clinically important and unimportant difference when a removable ankle support was used. We downgraded the certainty of this evidence by one level for risk of bias and one level for unexplained statistical heterogeneity. Kearney 2021 measured this outcome using two instruments, and in this analysis we included data from the OMAS. Data from Kearney 2021 using the Manchester Oxford Foot questionnaire is in Appendix 6. Losch 2002 reported this outcome as an ability to climb 12 stairs independently (see Appendix 6).

2.2. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 2: Activity limitation (using different measurement tools; short‐term follow‐up, ≤ 6months)

When measured at long‐term follow‐up, there was no evidence of a difference between groups (SMD 0.11, 95% CI ‐0.10 to 0.31; I² = 0%; 6 studies, 380 participants; Analysis 2.3). This was measured at 12 months in all studies, except in Lehtonen 2003 in which the outcome was measured at two years. We re‐expressed this effect estimate onto the OMAS (MD 1.47, 95% CI ‐1.33 to 4.13).

2.3. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 3: Activity limitation (using different measurement tools; long‐term follow‐up, > 6 months)

We did not include data from five studies in these analyses (DiStasio 1994; Franke 2008; Hedstrom 1994; Venesmaa 2004; Wetzler 1991). These studies reported data as medians, mean values without distribution data, or categorical data with an undefined rating scale. For completeness, we report these data in Appendix 7.

Interventions given after non‐surgical fracture management

All studies measured this outcome using OMAS. Kortegangas 2019 also reported data for this outcome using the Foot and Ankle Outcome Score (FAOS) questionnaire, and Van den Berg 2018 also reported data using the American Academy of Orthopedic Surgeons (AAOS) Foot and Ankle Score. We reported data using these measurement tools in Appendix 6. At the end of treatment, there was no evidence of a difference in outcome according to type of ankle support (MD ‐0.89, 95% CI ‐6.74 to 4.96; I² = 42%; 3 studies, 423 participants; Analysis 2.4).

2.4. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 4: Activity limitation after non‐surgical fracture management (using OMAS; end of treatment)

This outcome was also measured in the short term at 12 weeks (Kortegangas 2019), 16 weeks (Kearney 2021), and six months (Van den Berg 2018). There may be little or no difference in activity limitation according to type of support at this time point (MD 1.08, 95% CI ‐3.18 to 5.34; I² = 0%; 3 studies, 399 participants; low‐certainty evidence; Analysis 2.2). All studies used the OMAS. In addition, Kearney 2021 reported this outcome using the Manchester Oxford Foot questionnaire (see Appendix 6). We downgraded by one level for risk of bias and one level owing to imprecision because the CI indicated both benefit and harm for each treatment.

This outcome was measured in Kortegangas 2019 and Van den Berg 2018 at 12 months, with no evidence of a difference between groups at long‐term follow‐up (SMD 0.18, 95% CI ‐0.13 to 0.48; I² = 0%; 2 studies, 168 participants; Analysis 2.3). We re‐expressed this effect estimate onto the OMAS (MD 3.24, 95% CI ‐2.34 to 8.64).

2. HRQoL

Interventions given after surgical fracture management

Only two studies measured this outcome at the end of treatment, using SF‐36 (Dehghan 2016) and EQ‐5D (Kearney 2021). We found that quality of life was better at the end of treatment for people who used a removable ankle support (SMD 0.32, 95% CI ‐0.00 to 0.64; I² = 52%; 2 studies, 370 participants; Analysis 2.5). We re‐expressed this effect estimate onto the EQ‐5D scale (MD 0.09, 95% CI ‐0.00 to 0.17). Based on an MCID for EQ‐5D of 0.075 (Walters 2005), this includes both a clinically important and unimportant difference in HRQoL with a removable ankle support.

2.5. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 5: HRQoL (using different measurement tools; end of treatment)

In the short term, at 12 weeks (Matthews 2020; Smeeing 2020), and 16 weeks (Kearney 2021), quality of life was probably better for people who used a removable ankle support (SMD 0.30, 95% CI 0.11 to 0.50; I² = 9%; 3 studies, 477 participants; moderate‐certainty evidence; Analysis 2.6); this was measured with EQ‐5D (Matthews 2020; Kearney 2021), and SF‐36 (Smeeing 2020). We re‐expressed this effect estimate onto the EQ‐5D scale (MD 0.05, 95% CI 0.02 to 0.08). Based on an MCID for EQ‐5D of 0.075 (Walters 2005), this includes both a clinically important and unimportant difference in HRQoL with a removable ankle support. We downgraded the certainty of the evidence by one level for risks of bias in the included studies.

2.6. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 6: HRQoL (using different measurement tools; short‐term follow‐up, ≤ 6 months)

At a longer follow‐up, 12 months after surgery, there was no evidence of a difference between groups in quality of life (MD 2.60, 95% CI ‐5.12 to 10.31; I² = 63%; 2 studies, 178 participants; Analysis 2.7). In both studies, this was measured using SF‐36.

2.7. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 7: HRQoL (using different measurement tools; long‐term follow‐up, > 6 months)

Interventions given after non‐surgical fracture management

There was no evidence of a difference in quality of life at the end of treatment when measured using EQ‐5D (Kearney 2021; Van den Berg 2018) or Rand‐36 (Kortegangas 2019) (SMD 0.02, 95% CI ‐0.17 to 0.21; I² = 0%; 3 studies, 424 participants; Analysis 2.5). We re‐expressed this effect estimate onto the EQ‐5D scale (MD 0.02, 95% CI ‐0.16 to 0.19).

There may be little or no difference in quality of life in the short term (SMD ‐0.04, 95% CI ‐0.24 to 0.15; I² = 0%; 3 studies, 397 participants; low‐certainty evidence; Analysis 2.6); this was measured at 12 weeks (Kortegangas 2019), 16 weeks (Kearney 2021), and six months (Van den Berg 2018). We re‐expressed this effect estimate onto the EQ‐5D scale, which did not include a clinically important difference (MD ‐0.01, 95% CI ‐0.04 to 0.03). We downgraded the certainty of the evidence by one level for risk of bias, and one level for imprecision because the CI derived from these small studies included both benefit and no benefit for both treatments.

Only Kortegangas 2019 measured this outcome in the longer term, at 12 months, with no evidence of a difference between groups (MD ‐1.00, 95% CI ‐11.59 to 9.59; 1 study, 137 participants; Analysis 2.7).

3. Participant satisfaction

No studies reported this outcome.

4. Pain

Interventions given after surgical fracture management

Rasmussen 2006 found that participants who used a removable ankle support had less pain at the end of the treatment period (RR 0.12, 95% CI 0.03 to 0.44; 1 study, 40 participants; Analysis 2.8).

2.8. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 8: Pain after surgical fracture management (numbers with pain; all time points)

After a short‐term follow‐up of six months in Davies 1991, there may have been little or no difference in pain between the two groups, but the evidence is very uncertain (RR 0.74, 95% CI 0.52 to 1.07; 1 study, 29 participants; very low‐certainty evidence; Analysis 2.8). We downgraded the certainty of the evidence by one level for risk of bias, and two levels for imprecision because the effect estimate was derived from very few participants and included the possibility of benefit and harm for both treatments.

Three studies reported long‐term data at 12 months (Davies 1991; Sondenaa 1986), and 24 months (Rasmussen 2006). There was little or no difference in pain between groups at this time point (RR 0.63, 95% CI 0.35 to 1.14; I² = 34%; 3 studies, 116 participants; Analysis 2.8).

We did not include data from Wetzler 1991 in analyses; this study reported data without distribution values. For completeness, we report these data in Appendix 7.

Interventions given after non‐surgical fracture management

Two studies measured pain using a visual analogue scale (VAS) (Kortegangas 2019; Van den Berg 2018); the length of scale differed in each study (10‐point or 100‐point VAS). We reported unpooled data when measured at the end of treatment because of substantial statistical heterogeneity (I² = 86%). In Kortegangas 2019, people had less pain at the end of treatment when a non‐removable ankle support was used (MD 11.00, 95% CI 4.32 to 17.68; 1 study, 145 participants). In Van den Berg 2018, there was little or no difference in pain between the two groups (MD ‐1.10, 95% CI ‐2.60 to 0.40; 1 study, 39 participants). See Analysis 2.9.

2.9. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 9: Pain after non‐surgical fracture management (VAS; end of treatment)

Short‐term pain was measured at 12 weeks (Kortegangas 2019), and six months (Van den Berg 2018). There may have been little or no difference in pain between groups, but the evidence is very uncertain (SMD ‐0.10, 95% CI ‐0.73 to 0.53; I² = 57%; 2 studies, 167 participants; very low‐certainty evidence; Analysis 2.10). We re‐expressed this onto a 100‐point VAS (MD ‐2.2, 95% CI ‐16.06 to 11.66). We downgraded the certainty of the evidence by one level for risk of bias, and two levels for imprecision because the effect estimate was derived from very few participants and included the possibility of benefit and harm for both treatments.

2.10. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 10: Pain after non‐surgical fracture management (VAS; short‐term and long‐term follow‐up)

Long‐term pain was measured at 12 months in both studies, with little or no difference between groups (SMD ‐0.07, 95% CI ‐0.37 to 0.23; I² = 0%; 2 studies, 169 participants; Analysis 2.10). We re‐expressed this onto a 100‐point VAS (MD ‐1.23, 95% CI ‐6.48 to 4.03).

5. Adverse events

Interventions given after surgical fracture management

We found imprecise findings for all types of adverse events. We used the GRADE approach for unplanned return to theatre, and included these data in the summary of findings table. We downgraded the certainty of the evidence for this adverse event by one level for risk of bias, and one level for imprecision because the effect estimate included the possibility of benefit and harm for both treatments. See Analysis 2.11.

2.11. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 11: Adverse events (after surgical fracture management)

• Unplanned return to theatre (RR 1.20, 95% CI 0.39 to 3.71; I² = 54%; 6 studies, 624 participants; low‐certainty evidence).

• Non‐union (RR 0.63, 95% CI 0.21 to 1.87; 1 study, 364 participants).

• Loss of fixation (RR 3.00, 95% CI 0.13 to 71.92; 3 studies, 257 participants).

• Loss of reduction (not estimable due to no events; 4 studies, 222 participants).

• Fixation/implant failure (RR 3.10, 95% CI 0.13 to 73.16; 2 studies, 113 participants).

• Refracture of the lateral malleolus (RR 3.00, 0.13 to 71.92; 1 study, 100 participants).

• Deep vein thrombosis (DVT; RR 0.60, 95% CI 0.14 to 2.59; I² = 9%; 4 studies, 604 participants).

• Pulmonary embolism (RR 0.63, 95% CI 0.08 to 5.05; I² = 0%; 2 studies, 419 participants).

• Deep infection (RR 6.38, 95% CI 0.80 to 51.17; I² = 0%; 4 studies, 330 participants).

• Wound dehiscence (RR 2.50, 95% CI 0.34 to 18.55; I² = 0%; 3 studies, 288 participants).

• Low grade/superficial infection (RR 1.64, 95% CI 0.92 to 2.95; I² = 0%; 9 studies, 627 participants).

• Wound healing complications (RR 2.28, 95% CI 0.47 to 11.18; I² = 0%; 3 studies, 186 participants).

• Local skin necrosis (RR 3.00, 95% CI 0.13 to 71.92; 1 study, 100 participants).

• Pressure sore/ulcer (RR 0.63, 95% CI 0.21 to 1.87; 1 study, 364 participants).

• Dystrophy (not estimable due to no events; 1 study, 78 participants).

• Numbness at side of foot (RR 0.84, 95% CI 0.59 to 1.20; 1 study, 364 participants).

• Skin irritation (RR 2.00, 95% CI 0.19 to 21.36; 1 study, 100 participants).

• Dysesthesia of the skin (RR 3.00, 95% CI 0.32 to 27.87; 1 study, 100 participants).

• Chronic allodynia (RR 3.00, 95% CI 0.13 to 71.92; 1 study, 100 participants).

Interventions given after non‐surgical fracture management

We found imprecise findings for all types of adverse events except for DVT, for which there were fewer events after treatment with a removable ankle support (RR 0.20, 95% CI 0.04 to 0.93; I² = 0%; 3 studies, 306 participants). We used the GRADE approach for unplanned return to theatre, and included these data in the summary of findings table. We downgraded the certainty of the evidence for this adverse event by one level for risk of bias, and two levels for imprecision because the effect estimate was derived from few participants and included the possibility of benefit and harm for both treatments. See Analysis 2.11.

• Unplanned return to theatre (RR 0.99, 95% CI 0.14 to 6.96; 1 study, 305 participants; very low‐certainty evidence).

• Non‐union (RR 1.24, 95% CI 0.34 to 4.54; 2 studies, 469 participants).

• Loss of congruity (not estimable due to no events; 1 study, 164 participants).

• Nerve compression (RR 2.10, 95% CI 0.19 to 22.71; 1 study, 164 participants).

• Pulmonary embolism (RR 0.33, 95% CI 0.01 to 8.07; 1 study, 305 participants).

• Numbness at the side of the foot (RR 0.84, 95% CI 0.39 to 1.82; 1 study, 305 participants).

• Pressure sore/ulcer (RR 0.25, 95% CI 0.03 to 2.20; 1 study, 305 participants).

• Skin irritation/ulceration (RR 0.75, 95% CI 0.06 to 9.67; 2 studies, 201 participants).

Economic data

Three studies reported costs for this comparison (Franke 2008; Kearney 2021; Stockle 2000); we summarise these data in Appendix 5.

Sensitivity analysis

We report the results of sensitivity analyses when we noted a change in our interpretation of the primary analyses in Appendix 8.

Physical therapy interventions versus usual care or other physical therapy interventions

We stratified this comparison according to the time at which the intervention took place and the type of fracture management that participants received:

• during the initial fracture management period (typically six weeks, during which people may wear a removable or non‐removable ankle support) after surgical treatment;

• after the initial fracture management period (typically after a removable or non‐removable ankle support is removed, approximately six weeks after injury) after surgical or non‐surgical treatment.

We constructed a summary of findings table for data in the short term at six months or less (Table 3). We did not assess the certainty of the evidence for other time points.

1. Activity limitation

Interventions given during the initial fracture management period after surgical treatment

Jansen 2018 measured activity limitation using more than one measurement tool. For this analysis, we included data measured using the Mazur score; see Appendix 9 for data using other measurement tools in this study.

Jansen 2018 found an improvement in outcome at the end of the first six weeks after surgery when active controlled motion (ACM) was used alongside physiotherapy (MD 7.70, 95% CI 0.88 to 14.52; 1 study, 48 participants). This positive effect was also found in Molund 2020, measured with OMAS after three weeks of using a spring‐loaded ankle trainer (MD 5.60, 95% CI 0.93 to 10.27; 1 study, 112 participants). Another study found little or no difference in outcome when use of an antigravity treadmill was compared to usual care rehabilitation (MD ‐2.00, 95% CI ‐11.91 to 7.91; 1 study, 43 participants); this was measured using the Foot and Ankle Outcome Score (FAOS) six weeks postoperatively in Palke 2022. See Analysis 3.1.

3.1. Analysis.

Comparison 3: Physical therapy, Outcome 1: Activity limitation (using different measurement tools; at end of treatment)

In the short term, measured at 12 weeks in three studies (Jansen 2018; Molund 2020; Palke 2022), there may still have been a better outcome for people who received the ACM intervention, but the evidence was very uncertain (MD 10.80, 95% CI 3.40 to 18.20; 1 study, 44 participants; Jansen 2018). In Molund 2020, there may no longer have been an improvement in outcome with the spring‐loaded ankle trainer; in fact, at 12 weeks, evidence seemed to indicate a better outcome if usual care physiotherapy (stretching exercises with a non‐elastic band) was undertaken, but the evidence was very uncertain (MD ‐7.10, 95% CI ‐13.64 to ‐0.56; 106 participants). In Palke 2022, there may still have been little or no difference in outcome with the antigravity treadmill intervention at 12‐week follow‐up, but the evidence was very uncertain (MD ‐3.00, 95% CI ‐14.35 to 8.35; 1 study, 41 participants). See Analysis 3.2. We downgraded the certainty of this evidence to very low. We downgraded by one level for risk of bias, one level for imprecision because the data for each therapy type were derived from few participants, and one level for inconsistency because differences between treatments and comparisons meant that we could not pool these data.

3.2. Analysis.

Comparison 3: Physical therapy, Outcome 2: Activity limitation (using different measurement tools; short‐term follow‐up, ≤ 6 months)

Palke 2022 and Molund 2020 also reported this outcome at 12 months. There was no evidence of a difference in outcome when measured using OMAS at 12 months in Molund 2020 (MD ‐2.10, 95% CI ‐6.85 to 2.65; 1 study, 101 participants), or using FAOS in Palke 2022 (MD 3.00, 95% CI ‐9.07 to 15.07; 1 study, 41 participants). See Analysis 3.3.

3.3. Analysis.

Comparison 3: Physical therapy, Outcome 3: Activity limitation (using different measurement tools; long‐term follow‐up, > 6 months)

Interventions given after the initial fracture management period after either surgical or non‐surgical treatment

We found no evidence of a difference in outcome for any of the physical therapy interventions when measured at the end of the treatment period using the Lower Extremity Functional Scale (LEFS) in four studies (Lin 2008b; Moseley 2005; Moseley 2015; Wilson 1991). There was little or no difference in outcome at the end of four weeks of enhanced physiotherapy exercise sessions (MD ‐2.41, 95% CI ‐7.80 to 2.98; 1 study, 94 participants; Lin 2008b), or four weeks of additional daily stretching on top of the usual home physiotherapy programme in Moseley 2005 (MD 2.28, 95% CI ‐2.58 to 7.14; 1 study, 141 participants). There was also little or no difference in outcome at the end of the four‐week supervised exercise programme in Moseley 2015 (MD ‐0.40, 95% CI ‐4.59 to 3.79; 1 study, 194 participants), or after five weeks of Kaltenborn‐based manual therapy in Wilson 1991 (MD 21.00, 95% CI ‐1.40 to 43.40; 1 study, 10 participants). See Analysis 3.1.

There may be little or no difference in activity limitation between the two groups in the short term, when measured using LEFS in four studies (Keene 2022; Lin 2008b; Moseley 2005; Moseley 2015), or using OMAS in a fifth study (Nilsson 2009), but the evidence is very uncertain. Keene 2022 found that a 16‐week progressive exercise programme compared with best practice advice may have made little or no difference in outcome at six months (MD ‐3.84, 95% CI ‐12.63 to 4.95; 1 study, 52 participants). Lin 2008b found that enhanced physiotherapy exercise sessions may have made little or no difference in outcome 24 weeks after injury (MD ‐2.00, 95% CI ‐6.95 to 2.95; 1 study, 94 participants). Similarly, there may have been little or no difference in outcome at 12 weeks when additional stretches were included in a physiotherapy programme (MD 0.81, 95% CI ‐3.83 to 5.45; 1 study, 136 participants; Moseley 2005), or at six months when participants attended a supervised exercise class as part of their rehabilitation (MD 0.00, 95% CI ‐3.59 to 3.59; 1 study, 170 participants; Moseley 2015). There may also have been little or no difference in OMAS scores six months after injury when participants had attended a 12‐week programme of bi‐weekly physiotherapy sessions (MD ‐0.80, 95% CI ‐9.68 to 8.08; 1 study, 105 participants; Nilsson 2009). See Analysis 3.2. Keene 2022 also reported this outcome using OMAS, and we have included these data in Appendix 9. We downgraded the certainty of this evidence to very low. We downgraded one level for risk of bias, and two levels for imprecision because the data for each therapy type were derived from few participants and the CIs included benefits for both intervention and usual care treatments. These data also include inconsistencies in design, meaning that we were unable to combine data in meta‐analysis.

There was little or no difference in outcome 12 months after participants had attended a 12‐week programme of bi‐weekly physiotherapy sessions (MD 3.00, 95% CI ‐5.42 to 11.42; 1 study, 96 participants; Nilsson 2009; Analysis 3.3); this was measured using OMAS.

2. HRQoL

Interventions given during the initial fracture management period after surgical treatment

Palke 2022 reported this outcome using the quality of life subscale of the FAOS. At the end of the treatment period, there may have been little or no difference in quality of life between the antigravity treadmill group and the control group (MD ‐7.00, 95% ‐19.13 to 5.13; 1 study, 43 participants; Analysis 3.4). Findings were similar when reported in the short term at 12 weeks (MD ‐2.00, 95% CI ‐15.17 to 11.17; 1 study, 41 participants; very low‐certainty evidence; Analysis 3.5), and the long‐term at 12 months (MD 2.00, 95% CI ‐16.06 to 20.06; 1 study, 41 participants; Analysis 3.6), but the evidence was very uncertain. We downgraded the certainty of the evidence by one level for risk of bias and two levels for imprecision because the evidence for this outcome included very few participants.

3.4. Analysis.

Comparison 3: Physical therapy, Outcome 4: HRQoL (using different measurement tools; at end of treatment)

3.5. Analysis.

Comparison 3: Physical therapy, Outcome 5: HRQoL (using different measurement tools; short‐term follow‐up, ≤ 6 months)

3.6. Analysis.

Comparison 3: Physical therapy, Outcome 6: HRQoL (using different measurement tools; long‐term follow‐up, > 6 months)

Interventions given after the initial fracture management period after either surgical or non‐surgical treatment

There was no evidence of a difference in quality of life when measured at the end of the treatment period using Assessment of Quality of Life (AQoL) scores. There was little or no difference in quality of life at the end of four weeks of enhanced physiotherapy exercise sessions (MD 1.17, 95% CI ‐0.46 to 2.80; 1 study, 94 participants; Lin 2008b), or at the end of the four‐week supervised exercise programme in Moseley 2015 (MD 0.70, 95% CI ‐0.70 to 2.10; 1 study, 193 participants). See Analysis 3.4.

Four studies reported short‐term data for HRQoL using EQ‐5D (Keene 2022), AQoL (Lin 2008b; Moseley 2015), and SF‐36 (Nilsson 2009). Lin 2008b found that enhanced physiotherapy exercise sessions may improve quality of life 24 weeks after injury when compared to usual care physiotherapy (MD 1.49, 95% CI 0.14 to 2.84; 1 study, 94 participants). Keene 2022 found that there may have been little or no difference in quality of life scores at six months when participants took part in a progressive exercise and advice programme (MD 0.04, 95% CI ‐0.07 to 0.15; 56 participants). Moseley 2015 also found that there may have been little or no difference in quality of life scores at six months after injury when participants took part in a supervised exercise programme (MD 0.10, 95% CI ‐1.12 to 1.32; 1 study, 169 participants). Similarly, at six months after injury, Nilsson 2009 found that there may have been little or no difference in quality of life for participants who had attended a 12‐week programme of bi‐weekly physiotherapy sessions (MD 1.70, 95% CI ‐2.45 to 5.85; 1 study, 105 participants). In all cases, the evidence was very uncertain. See Analysis 3.5. We downgraded the certainty of this evidence to very low, downgrading by one level for risk of bias, one level for imprecision because the data for each therapy type were derived from few participants, and one level for inconsistency because differences between treatments and comparisons meant that we could not pool these data.

Only Nilsson 2009 reported long‐term data for this outcome, with little or no difference in quality of life 12 months after injury (MD 0.40, 95% CI ‐3.54 to 4.34; 1 study, 96 participants; Analysis 3.6).

3. Participant satisfaction

Interventions given during the initial fracture management period after surgical treatment

No studies in this comparison group measured this outcome.

Interventions given after the initial fracture management period after either surgical or non‐surgical treatment

Lin 2008b and Moseley 2005 both used a 100‐point VAS to measure people's satisfaction with their treatment. Lin 2008b found that there may have been little or no difference in satisfaction according to whether people used enhanced physiotherapy exercise sessions or usual care physiotherapy sessions (MD 6.50, 95% CI ‐0.12 to 12.26; 94 participants). Moseley 2005 also found that there may have been little or no difference in satisfaction when additional stretches were used in a physiotherapy session (MD 2.20, 95% CI ‐4.35 to 8.75; 1 study, 138 participants). In both cases, the evidence was very uncertain. See Analysis 3.7. We downgraded one level for risk of bias, one level for imprecision because the data for each therapy type were derived from few participants, and one level for inconsistency because differences between treatments and comparisons meant that we could not pool these data.

3.7. Analysis.

Comparison 3: Physical therapy, Outcome 7: Participant satisfaction (100‐mm VAS; end of treatment)

4. Pain

Interventions given during the initial fracture management period after surgical treatment

Molund 2020 reported this outcome using a 10‐point VAS, and Palke 2022 reported this outcome using the pain subscale of the FAOS.

At the end of the treatment period, there was little or no difference in levels of pain between those in the spring‐loaded ankle trainer group and the control group in Molund 2020 (MD ‐0.20, 95% CI ‐0.89 to 0.49; 1 study, 112 participants), and between those in the antigravity treadmill or control group in Palke 2022 (MD ‐8.00, 95% CI ‐20.56 to 4.56; 1 study, 43 participants). See Analysis 3.8.

3.8. Analysis.

Comparison 3: Physical therapy, Outcome 8: Pain (using different measures; at end of treatment

Results were similar at short‐term follow‐up (12 weeks postoperatively). There may have been little or no difference between groups in pain scores reported in Molund 2020 (MD 0.00, 95% CI ‐0.59 to 0.59; 1 study, 106 participants), or in Palke 2022 (MD ‐4.00, 95% CI ‐16.86 to 8.86; 1 study, 41 participants), but the evidence was very uncertain. See Analysis 3.9. We downgraded the certainty of this evidence to very low, downgrading by one level for risk of bias, and two levels for imprecision because the data for each therapy type were derived from few participants and CIs included benefits for both intervention and usual care treatments.

3.9. Analysis.

Comparison 3: Physical therapy, Outcome 9: Pain (using different measurement tools; short‐term follow‐up, ≤ 6 months)

With a long‐term follow‐up of 12 months in both Molund 2020 and Palke 2022, there was similarly little or no difference in pain scores when a spring‐loaded ankle trainer was used (MD 0.30, 95% CI ‐0.27 to 0.87; 1 study, 101 participants), or an antigravity treadmill (MD ‐8.00, 95% CI ‐20.18 to 4.18; 1 study, 41 participants). See Analysis 3.10.

3.10. Analysis.

Comparison 3: Physical therapy, Outcome 10: Pain (using different measurement tools; long‐term follow‐up, > 6 months)

Interventions given after the initial fracture management period after either surgical or non‐surgical treatment

Lin 2008b and Moseley 2015 reported pain on equal weight‐bearing at the end of treatment, with little or no difference in pain after enhanced physiotherapy exercise sessions (MD 0.20 95% CI ‐5.74 to 6.14; 1 study, 94 participants; Lin 2008b) or after a supervised exercise programme (MD 0.10, 95% CI ‐0.44 to 0.64; 1 study, 194 participants; Moseley 2015). In Moseley 2005, data were available for pain on stair descent, with little or no difference in pain after additional stretching exercises for four weeks (MD ‐6.72, 95% CI ‐14.39 to 0.95; 1 study, 139 participants). See Analysis 3.8. These outcomes were measured using a 100‐point VAS in Lin 2008b, and a 10‐point VAS in the other studies.

These three studies also reported pain after a short‐term follow‐up, as did Keene 2022. Lin 2008b found that there may have been little or no difference in pain on equal weight‐bearing at 24 weeks (1.10, 95% CI ‐3.90 to 6.10; 1 study, 94 participants), as did Moseley 2015 at six months (MD 0.00, 95% CI ‐0.30 to 0.30; 171 participants). Moseley 2005 reported pain on stair descent 12 weeks after injury, again finding that there may have been little or no difference in pain scores (MD ‐2.10, 95% CI ‐8.08 to 3.88; 1 study, 135 participants). At six months, Keene 2022 found that there may have been little or no difference in pain at rest after a progressive exercise and advice programme as measured on a 100‐point VAS (MD 1.70, 95% CI ‐8.26 to 11.66; 53 participants). In all cases, the evidence was very uncertain. See Analysis 3.9. Keene 2022 also reported data for pain on walking, and we have included these data in Appendix 9. We downgraded the certainty of this evidence to very low, downgrading by one level for risk of bias, and two levels for imprecision because the data for each therapy type were derived from few participants and CIs included benefits for both intervention and usual care treatments. These data also include inconsistencies in design, meaning that we were unable to combine data in meta‐analysis.

No studies reported pain at longer‐term follow‐up.

Pain was reported using different measures in Keene 2022, Lin 2008b, and Moseley 2015. For completeness, we have included these data in Appendix 9.

5. Adverse events

Interventions given during the initial fracture management period after surgical treatment

We found imprecise findings for all types of adverse events, which are reported in Analysis 3.12. Most events were described in Palke 2022 and often included only one event for each type of adverse outcome across both the intervention and control groups. No studies reported unplanned return to theatre as an adverse event.

3.12. Analysis.

Comparison 3: Physical therapy, Outcome 12: Adverse events (during initial fracture management period after surgical treatment)

Interventions given after the initial fracture management period after either surgical or non‐surgical treatment

We found imprecise findings for all types of adverse events, which are reported in Analysis 3.13. No studies reported unplanned return to theatre as an adverse event.

3.13. Analysis.

Comparison 3: Physical therapy, Outcome 13: Adverse events (after initial fracture management period after either surgical or non‐surgical treatment)

Economic data

Two studies reported costs for this comparison (Lin 2008a; Moseley 2015); we summarise these data in Appendix 5.

Other rehabilitation interventions

Electrotherapy versus control

In this comparison group, all participants had surgical fracture management. We stratified the results according to whether the intervention was given during the initial fracture management period (i.e. usually within the first six weeks after the initial injury) or after the initial fracture management period (i.e. starting at least six weeks after injury). Because of the differences between treatment types in this comparison group, we did not pool data.

1. Activity limitation

Interventions given during the initial fracture management period after surgical treatment

This outcome was measured using OMAS in all studies (Analysis 4.1). At the end of a six‐week treatment period, Hernandez 2006 found little or no difference in activity limitation between participants given electrical stimulation and those given no treatment (MD 8.86, 95% CI ‐10.86 to 28.58; 21 participants).

4.1. Analysis.

Comparison 4: Other rehabilitation interventions: electrotherapy versus control, Outcome 1: Activity limitation (interventions given during initial fracture management period; OMAS; all time points)

Similarly, after a short‐term follow‐up 12 weeks after randomisation, there remained little or no difference in activity limitation between those in the electrical stimulation and no treatment groups in Hernandez 2006 (MD 11.27, 95% CI ‐4.17 to 26.71; 21 participants).

Handolin 2005 measured the effectiveness of ultrasound in the long term (18 months after randomisation), and found little or no difference in activity limitation between those given ultrasound and sham treatment (MD ‐1.25, 95% CI ‐7.67 to 5.17; 16 participants).

Interventions given after the initial fracture management period after surgical treatment

No studies measured or reported this outcome.

2. HRQoL

No studies measured or reported this outcome.

3. Participant satisfaction

No studies measured or reported this outcome.

4. Pain

Interventions given during the initial fracture management period after surgical treatment

This outcome was measured using a 10‐point VAS in two studies (Gorodetskyi 2010; Hernandez 2006). Gorodetskyi 2010 found that pain was reduced at the end of a 10‐day treatment with non‐invasive interactive neurostimulation compared with a sham treatment (MD ‐2.10, 95% CI ‐2.36 to ‐1.84; 60 participants). There was little or no difference in pain at the end of a six‐week treatment period of electrical stimulation in Hernandez 2006 (MD ‐0.75, 95% CI ‐2.74 to 1.24). See Analysis 4.2.

4.2. Analysis.

Comparison 4: Other rehabilitation interventions: electrotherapy versus control, Outcome 2: Pain (10‐cm VAS; end of treatment)

Hernandez 2006 also measured pain in the short term (12 weeks after initial injury), and similarly found little or no difference according to whether participants had received electrical stimulation or no treatment (MD ‐1.05, 95% CI ‐2.57 to 0.47; 21 participants; Analysis 4.3).

4.3. Analysis.

Comparison 4: Other rehabilitation interventions: electrotherapy versus control, Outcome 3: Pain (interventions given during the initial fracture management period; VAS; short‐term follow‐up, ≤ 6 months)

Handolin 2005 reported the number of people experiencing pain in the long term (18 months after ultrasound or sham treatment), with little or no difference between groups (RR 3.00, 95% CI 0.39 to 23.07; 16 participants; Analysis 4.4).

4.4. Analysis.

Comparison 4: Other rehabilitation interventions: electrotherapy versus control, Outcome 4: Pain (numbers with pain; long‐term follow‐up, > 6 months)

Interventions given after the initial fracture management period after surgical treatment

Seiger 2009 found little or no difference between thermal and non‐thermal pulsed shortwave therapy given to participants twice a week for 12 sessions (MD 0.20, 95% CI ‐0.93 to 1.33; 12 participants; Analysis 4.2).

5. Adverse events

Interventions given during the initial fracture management period after surgical treatment

Handolin 2005 reported that there were no healing complications after ultrasound or sham treatment, and Hernandez 2006 reported that there were no adverse effects of treatment. No other studies reported adverse events.

Interventions given after the initial fracture management period after surgical treatment

Seiger 2009 reported that there were no adverse reactions from thermal pulsed shortwave therapy.

Other interventions

Here, we report other comparisons that did not fit into the above comparison groups. We did not combine data for these studies because of the differences between comparisons. None of these studies reported HRQoL or participant satisfaction.

1. Activity limitation

Interventions given during the initial fracture management period after surgical treatment

We report data from Brink 1996 at end of treatment and at short‐term follow‐up of 12 weeks in Appendix 10. These data were reported as median values and we did not calculate effect estimates.

Interventions given during the initial fracture management period after non‐surgical treatment

No studies reported activity limitation in this comparison.

2. HRQoL

No studies in either comparison group measured or reported this outcome.

3. Participant satisfaction

No studies in either comparison group measured or reported this outcome.

4. Pain

Interventions given during the initial fracture management period after surgical treatment

In Reed 1998, participants were asked to judge their levels of pain after using a wool and crêpe bandage or below‐knee backslab for the first 24 hours after internal fixation. There was little or no difference between treatment groups in pain as measured on a 10‐point VAS (MD ‐0.28, 95% CI ‐2.04 to 1.48; 34 participants; Analysis 5.1).

5.1. Analysis.

Comparison 5: Other rehabilitation interventions, Outcome 1: Pain (using different measurement instruments; at end of treatment)

In Romero Zepeda 2008, there was little or no difference in pain after using a circular (Robert Jones) bandage or backslab for 24 hours after injury as measured on a 10‐point VAS with a median score of 2 points on a 10‐point VAS.

Interventions given during the initial fracture management period after non‐surgical treatment

Ginandes 1999 found little or no difference in pain at the end of the 12‐week intervention period of hypnosis or no treatment (MD ‐0.43, 95% CI ‐1.49 to 0.63; 11 participants; Analysis 5.1); this was measured using a 10‐point numerical scale.

5. Adverse events

Interventions given during the initial fracture management period after surgical treatment

Fitzgerald 1994 reported that there were no adverse events in either group.

Brink 1996 reported little or no difference in fracture displacement after either an Aircast Air‐stirrup or an orthosis was used after surgical fixation (RR 0.97, 95% CI 0.06 to 14.70; 57 participants), and little or no difference in skin irritation (RR 2.90, 95% CI 0.12 to 68.33; 57 participants). See Analysis 5.2.

5.2. Analysis.

Comparison 5: Other rehabilitation interventions, Outcome 2: Adverse events (interventions during initial fracture management with surgical treatment)

Interventions given during the initial fracture management period after non‐surgical treatment

No studies reported adverse events in this comparison.

Economic data

Brink 1996 reported economic data, which we summarise in Appendix 5.

Discussion

Summary of main results

Here, we summarise outcomes when measured after a short‐term follow‐up, which we defined as up to six months after the initial injury.

Early weight‐bearing versus delayed weight‐bearing. For this comparison, we included 12 studies with 1403 participants who had surgical fracture management. For activity limitation, we found that early weight‐bearing within three weeks of surgery probably leads to an improved outcome. However, based on an MCID of 10 points for the OMAS scale on which this outcome was measured, we noted that this result does not include a clinically meaningful difference. We also found there may be little to no difference in quality of life between those who had early versus delayed weight‐bearing. As these data were pooled using SMD (because of different outcome measures), we re‐expressed the data onto a commonly used scale (EQ‐5D) and found the result for early‐weight bearing included both a clinically important and unimportant improvement. We were unsure whether there were any differences in participant satisfaction or pain because these data had very low‐certainty evidence. For adverse events, we found low‐certainty evidence that there may be little or no difference in unplanned return to theatre (i.e. re‐operation).

Removable versus non‐removable ankle support. For this comparison, we included 25 studies with 2206 participants. We reported the data separately according to whether participants had surgery as part of the fracture management. When used after surgical fracture management, we found that using a removable ankle support may lead to better ankle function. Based on an MCID of 10 points, we noted that this low‐certainty evidence included both a clinically important and unimportant difference. We found moderate‐certainty evidence that there is probably an improvement in HRQoL, but again, this difference included both a clinically important and unimportant difference. No studies reported participant satisfaction, and we were unsure of the effects of treatment on pain because of very low‐certainty evidence. We found low‐certainty evidence that there may be little or no difference in the number of people who have an unplanned return to theatre.

When used after non‐surgical fracture management, we found low‐certainty evidence that there may be little or no difference between types of ankle support in ankle function, and HRQoL. No studies reported participant satisfaction. We were unsure of the effects of treatment on pain and unplanned return to theatre because of very low‐certainty evidence.

Physical therapy interventions versus usual care or other physical therapy interventions. Overall, we included nine studies (with 857 participants) that investigated physical therapy interventions. We collected data separately according to whether physical therapy interventions were given during the initial fracture management period of typically six weeks (after surgical treatment), or after this period (regardless of whether surgery was used). In each study, the type of intervention varied, and even the comparison groups (often described as usual care) were likely to be different from one another. It was not possible to combine data from these studies in any meaningful way because of the differences between these studies. In addition, all the studies included few or very few participants. We judged the certainty of the evidence for each outcome in this comparison to be very low, and therefore, we were unsure of the effectiveness of these physical therapies.

Other rehabilitation interventions. For completeness, we reported in the review other types of interventions that were given to people after ankle fracture to assist with their rehabilitation. A small group of studies evaluated electrotherapy interventions but as each type of intervention differed, we did not pool data from these studies. We also found studies that compared different types of ankle supports, and one study that evaluated the use of hypnosis after ankle fracture. These were small studies with no similarities in study design. We did not include data from these studies in summary of findings tables because we did not think that these individual comparisons were clinically important. We did not evaluate the certainty of these data using GRADE but expected that the certainty of evidence for reported outcomes was likely to be very low in all cases.

Overall completeness and applicability of evidence

Overall, we found 53 studies, with most of the studies contributing evidence to the three main comparisons in this review. Studies included adults who were mostly younger than 65 years of age; no studies specifically targeted interventions at older adults (> 65 years of age). Most participants had surgical orthopaedic management for ankle fracture, although we also included some studies in which participants had non‐surgical fracture management. It was beyond the scope of this review to explore the impact on the results of different types of ankle fracture (or related soft tissue injuries), or surgical approaches and fixation methods, or both. Because it was not always clearly reported, we could not confirm definitively that surgery was limited to displaced or unstable fractures. Although more than half the studies did not describe whether study populations were limited to closed fractures, we assumed the evidence in this review was largely applicable to closed fractures. We could not account for differences in fracture management between earlier studies (published before 2010) and more recent studies; we noted that two‐thirds of the evidence was from studies published before 2010. For the physical therapy interventions, there were many differences between the studies' approaches, including in terms of the frequency of treatment, progression of exercise, and the setting. Some physical therapy studies gave scant details of the interventions used, and we were unable to pool their data. It is likely that the evidence for physical therapy interventions is incomplete because of the variation between the small studies that we included in this review.

Certainty of the evidence

The certainty of the evidence for outcomes measured in the short term (≤ six months after injury) ranged from moderate to very low. All the studies in the main comparison groups were at unavoidably high risk of performance and detection bias, and we downgraded all the evidence owing to serious risk of bias. The effect estimates for some outcomes were imprecise, with CIs that included the possibility of benefit for either the intervention or comparison. The evidence was sometimes derived from very few participants; in these cases, we also downgraded the certainty of the evidence for imprecision. We did not pool the data for pain for people who had early or delayed weight‐bearing because of statistical heterogeneity between study results which we could not explain, and we downgraded in this case for inconsistency. The evidence for physical therapy interventions included such differences in rehabilitation approaches that it was not appropriate to attempt to pool these data. Given this, we also downgraded this evidence for inconsistency.

All the participants and interventions provided direct evidence for this review. Although we did not detect publication bias, we could not rule out the possibility that the evidence included publication bias.

Potential biases in the review process

We conducted a thorough search, and two review authors independently assessed study eligibility, extracted data, and assessed the risk of bias in the included studies before reaching consensus between the two review authors. In the event that review authors were involved in the design or conduct of included studies, other review authors extracted data and assessed the risk of bias for these studies to reduce risk of bias. Whilst this update included some changes stemming from adherence to current methodological and reporting expectations for Cochrane reviews (e.g. reporting of summary of findings tables), we do not expect that these changes would have introduced bias to the review process.

In this update, we split the review into three distinct comparison groups (early versus delayed weight‐bearing; removable versus non‐removable ankle supports; and physical therapy interventions). This decision was driven by the current main clinical questions in rehabilitation management, and allowed us to present distinct summary of findings tables for these different approaches. We assessed the intervention components of each study to categorise the interventions into these comparison groups. In some cases, this meant that we were able to use a study in more than one comparison group, thus maximising the data set. However, there was insufficient information reported in other studies, and we could not rule out the possibility that data from some studies could have been used in additional comparisons.

We reduced the number of outcomes in this update, no longer reporting data for ankle dorsiflexion and plantarflexion range of motion, strength, and swelling, which we considered to be less clinically relevant. Data for these outcomes from studies published before 2011 are reported in Lin 2012. We included an additional outcome (unplanned return to theatre) which was not in the previous version of the review, but which we believed to be an important adverse event. However, we acknowledge that other adverse events may be more common (such as wound complications or surgical site infections for those who have surgical fixation), and we have included in the review all adverse events reported by study authors. In addition, we note that some adverse events may be related to participants performing the intervention (for example, falls). We did not distinguish between adverse events or adverse effects that were related or likely to be related to the intervention. We also used a different threshold of six months (rather than three months) for short‐term and long‐term follow‐up when collecting outcome data; this threshold was based on Beckenkamp 2014. However, we used sensitivity analyses to explore this decision and found that there were no differences in our interpretation of the pooled effect estimates.

We could not always use data in analysis; for example, because data were reported as medians or without distribution values. This limited the amount of data in our analyses. However, for completeness, we included these data in appendices.

In most studies including surgical fracture management, randomisation was conducted after surgery. Sometimes, study authors specified that participants were only included after stable internal fixation (Davies 1991; DiStasio 1994; Finsen 1989; Lehtonen 2003; Matthews 2020; Park 2021; Reed 1998; Smeeing 2020; Tropp 1995; Van Laarhoven 1996; Vioreanu 2007). However, other studies did not clearly specify whether participants were only included after successful stabilisation, and we could not rule out the possibility of potential bias in these studies. We did not address this factor of study conduct in our risk of bias assessment or in any sensitivity or subgroup analyses. Therefore, we cannot confirm whether our findings in surgical populations are only applicable to interventions delivered in fractures with stable fixation.

Agreements and disagreements with other studies or reviews

This review included two large multicentre studies. The findings of Bretherton 2024 for early versus delayed weight‐bearing, and Kearney 2021 for removable versus non‐removable ankle supports, provide substantial data which are consistent with the findings in this review.

We found few other well‐conducted systematic reviews evaluating early weight‐bearing or the use of a removable ankle support. Reviews by Keene 2014 and Smeeing 2015 both evaluated the use of a removable ankle support (to allow mobilisation). These reviews included fewer studies than this Cochrane review, and most were smaller single‐centre studies, reflecting the published trials at the time of their searches. Time points differed in their analyses of shorter‐term follow‐ups, and their findings for function were less precise than those of this review. A more recent systematic review of both early weight‐bearing and removable ankle supports also did not capture data from the larger multicentre studies that we have included (Sharma 2022), and they reported imprecise findings for their shorter‐term follow‐ups, with more statistical heterogeneity than our own more comprehensive analyses. We are not aware of other recent, well‐conducted systematic reviews, particularly evaluating different physical therapy approaches for the rehabilitation of ankle fractures.

Authors' conclusions

Implications for practice.

For people who have surgery for ankle fracture, early weight‐bearing within three weeks of surgery may improve outcomes in the first six months after injury. However, the pooled estimate for function indicated only a small improvement after early weight‐bearing, which is unlikely to include a clinically important difference, and the improvement for health‐related quality of life after early weight‐bearing included the possibility of both a clinically meaningful and non‐meaningful difference. It is likely that neither early nor delayed weight‐bearing increase the risk of re‐operation.

A removable ankle support, which allows the opportunity for gentle exercise or mobilisation of the ankle, may also provide better function than a non‐removable ankle support (such as a plaster cast) for people who have had surgery, but again, the difference may not always be clinically important. There may be no difference between removable and non‐removable ankle supports for the risk of re‐operation. We assume that the findings for this review comparison are applicable to people with a closed ankle fracture who had surgery that achieved satisfactory fracture stabilisation. For people who have non‐surgical treatment, neither a removable nor non‐removable ankle support may be superior in terms of function or health‐related quality of life in the first six months after injury. However, we were unsure if either approach increases the risk of needing later surgery. As for the early versus delayed weight‐bearing comparison, we assume that these findings are applicable to people with a closed ankle fracture who had surgery that achieved satisfactory fracture stabilisation.

We were also uncertain whether any physical therapy interventions were more effective than usual care or other physical therapy interventions. The types of physical therapy interventions differed in the included studies, precluding meta‐analysis. In addition, the sample sizes in this dataset were often small.

Implications for research.

Since the previous version of this Cochrane review, several well‐designed multicentre trials have been conducted. The inclusion of Bretherton 2024 in this update, with more than 500 participants from multiple centres, provides a more certain result for function after early weight‐bearing. Further research may not yield important information to change this finding. However, we note that there are already ongoing studies for this comparison, which we hope will supplement this review finding.

Similarly, we included the large multicentre study by Kearney and colleagues, evaluating the effectiveness of using a removable ankle support during the early rehabilitation period (Kearney 2021). Whilst the data for people who had surgery in Kearney 2021 were supported by several other studies, the evidence for people who had non‐surgical fracture management was derived from fewer participants. For this latter group, data from large, well‐conducted studies would increase the certainty of the evidence. As randomised controlled trials are feasible in this field, we encourage these types of study designs, with clear reporting of methods used to reduce bias (including attrition for participant‐reported outcomes).

Although we are aware of ongoing studies of other rehabilitation interventions (including types of physical therapy), we recognise the difficulties of providing any pooled evidence in this research field when intervention components are likely to differ. Most of these studies include a 'usual care' comparator, but differences between resources and protocols in each healthcare setting mean that there is unlikely to be a standard rehabilitation process across settings. Detailed reporting of intervention components in future studies will increase the potential for pooling of interventions across common components, as well as translation of the research findings into real‐world application. This review provided little data on the costs of interventions, and it is likely that enhanced physical therapy interventions may have a higher economic impact on healthcare provision. We encourage future researchers for this comparison to provide an economic evaluation of their interventions.

A core outcome set was unavailable when we conducted this Cochrane review. However, we are aware that a core outcome set is in development for ankle fracture, and we encourage investigators to include these outcomes in future studies (Pearson 2021). We will include these outcomes in future updates of this Cochrane review.

What's new

Date	Event	Description
23 September 2024	New search has been performed	Review authors: we added six new authors (SL, MP, RP, HS, DK, CB) and removed four authors (ND, KR, KK, AM). Searches and data extraction: we updated and re‐ran the search for studies, extracted data on new studies, conducted risk of bias assessments on all included studies, and incorporated new data into the review. We reduced the number of outcomes and added unplanned return to theatre as an adverse event. We reported short‐term and long‐term follow‐up at ≤ 6 months and > 6 months, respectively. We reported the data according to distinct comparison groups (early vs delayed weight‐bearing; removable vs non‐removable ankle supports; physical therapies; other rehabilitation interventions), stratified according to the initial surgical or non‐surgical orthopaedic treatment. Results: this review update includes an additional 15 studies. Conclusions: we made changes to the conclusions to reflect review findings.
23 September 2024	New citation required and conclusions have changed	We made changes to the conclusions to reflect review findings

History

Protocol first published: Issue 1, 2006
Review first published: Issue 3, 2008

Date	Event	Description
1 October 2012	New citation required and conclusions have changed	1. New evidence on electrotherapy during and after the immobilisation period, and manual therapy and exercise after the immobilisation period necessitated changes to the conclusions. 2. Three new authors were added to the review team.
1 October 2012	New search has been performed	For this update, published in Issue 11, 2012, the following changes were made: 1. The search was updated to 5 July 2011. 2. Seven new studies were included 3. New evidence on electrotherapy during and after the immobilisation period, and manual therapy and exercise after the immobilisation period necessitated changes to the results and conclusions. 4. In line with Collaboration policy, included trials were assessed using the 'Risk of bias' tool. Since the studies included in the previous version of this review were originally assessed using the PEDro scale, analyses using the 'Risk of bias' tool were included for these studies in this update.
9 May 2008	Amended	Converted to new review format.

Appendices

Appendix 1. Search strategies (2011 to present)

CENTRAL (CRS‐Web)

#1 MESH DESCRIPTOR Ankle Fractures AND CENTRAL:TARGET (191)
#2 MESH DESCRIPTOR Ankle AND CENTRAL:TARGET (136)
#3 MESH DESCRIPTOR Ankle Joint AND CENTRAL:TARGET (783)
#4 MESH DESCRIPTOR Ankle Injuries AND CENTRAL:TARGET (766)
#5 (ankle or malleol*):AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (11416)
#6 (distal NEAR2 (tibia* or fibula*)):AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (424)
#7 #2 OR #3 OR #4 OR #5 OR #6 (11720)
#8 MESH DESCRIPTOR Fractures, Bone AND CENTRAL:TARGET (2430)
#9 (fracture*):AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (25526)
#10 #8 OR #9 (25525)
#11 #7 AND #10 (1410)
#12 #1 OR #11 (1410)
#13 01/07/2011_TO_05/05/2022:CRSCREATED AND CENTRAL:TARGET (1865270)
#14 #12 AND #13 (1409)

MEDLINE (Ovid)

1 Ankle Fractures/ (2091)
2 Ankle/ or Ankle Joint/ or Ankle Injuries/ (33811)
3 (ankle or malleol*).tw. (67538)
4 (distal adj2 (tibia* or fibula*)).tw. (5091)
5 or/2‐4 (78142)
6 Fracture, Bone/ (68251)
7 fracture*.tw. (284215)
8 6 or 7 (300977)
9 5 and 8 (12519)
10 1 or 9 (12780)
11 randomized controlled trial.pt. (567952)
12 controlled clinical trial.pt. (94866)
13 randomi?ed.ab. (670034)
14 placebo.ab. (228053)
15 drug therapy.fs. (2487722)
16 randomly.ab. (381812)
17 trial.ab. (598940)
18 groups.ab. (2347636)
19 or/11‐18 (5364144)
20 exp animals/ not humans.sh. (5003616)
21 19 not 20 (4670493)
22 10 and 21 (2053)
23 (201107* or 201108* or 201109* or 201110* or 201111* or 201112* or 2012* or 2013* or 2014* or 2015* or 2016* or 2017* or 2018* or 2019* or 2020* or 2021* or 2022*).ed,dt. (13737007)
24 22 and 23 (1476)

Embase (Ovid)

1 ankle fracture/ or distal tibia fracture/ (6319)
2 ankle/ or ankle injury/ (43400)
3 (ankle or malleol*).tw. (87945)
4 (distal adj2 (tibia* or fibula*)).tw. (6363)
5 or/2‐4 (100091)
6 fracture/ (87245)
7 fracture*.tw. (323148)
8 6 or 7 (341972)
9 5 and 8 (13476)
10 1 or 9 (15564)
11 Randomized controlled trial/ (702495)
12 Controlled clinical study/ (465521)
13 Random*.ti,ab. (1767801)
14 randomization/ (93544)
15 intermethod comparison/ (282464)
16 placebo.ti,ab. (333943)
17 (compare or compared or comparison).ti. (542486)
18 ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. (2475347)
19 (open adj label).ti,ab. (96430)
20 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. (248384)
21 double blind procedure/ (191424)
22 parallel group*1.ti,ab. (29212)
23 (crossover or cross over).ti,ab. (113272)
24 ((assign* or match or matched or allocation) adj5 (alternate or group*1 or intervention*1 or patient*1 or subject*1 or participant*1)).ti,ab. (374867)
25 (assigned or allocated).ti,ab. (441609)
26 (controlled adj7 (study or design or trial)).ti,ab. (402122)
27 (volunteer or volunteers).ti,ab. (260271)
28 trial.ti. (350641)
29 or/11‐28 (5279249)
30 (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) (6400107)
31 29 not 30 (4586698)
32 10 and 31 (3296)
33 (2011* or 2012* or 2013* or 2014* or 2015* or 2016* or 2017* or 2018* or 2019* or 2020* or 2021* or 2022*).dc,yr. (18189232)
34 32 and 33 (2530)

CINAHL Plus

S1 MH ankle fractures OR MH Tibial Fractures OR MH Fibula Fractures (6,356)
S2 MH Ankle OR MH Ankle Joint OR MH Ankle Injuries OR MH Tibial OR MH Fibula (24,901)
S3 TX ankle OR TX malleol* (44,295)
S4 TX (distal n2 (tibia* or fibula*)) (2,184)
S5 S2 OR S3 OR S4 (54,824)
S6 MH fractures (21,014)
S7 TX fracture* (98,488)
S8 S6 OR S7 (98,488)
S9 S5 AND S8 (9,121)
S10 S1 OR S9 (11,837)
S11 MH randomized controlled trials (128,429)
S12 MH double‐blind studies (53,004)
S13 MH single‐blind studies (15,555)
S14 MH random assignment (73,517)
S15 MH pretest‐posttest design (49,041)
S16 MH cluster sample (4,953)
S17 TI (randomised OR randomized) (126,963)
S18 AB (random*) (372,152)
S19 TI (trial) (164,004)
S20 MH (sample size) AND AB (assigned OR allocated OR control) (4,332)
S21 MH (placebos) (13,670)
S22 PT (randomized controlled trial) (141,798)
S23 AB (control W5 group) (133,295)
S24 MH (crossover design) OR MH (comparative studies) (457,671)
S25 AB (cluster W3 RCT) (452)
S26 MH animals+ (101,909)
S27 MH (animal studies) (146,727)
S28 TI (animal model*) (3,511)
S29 S26 OR S27 OR S28 (239,772)
S30 MH (human) (2,547,929)
S31 S29 NOT S30 (206,495)
S32 S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23 OR S24 OR S25 (958,249)
S33 S32 NOT S31 (912,531)
S34 S10 AND S33 (1,805)
S35 EM 20110701‐20220503 (4,419,538)
S36 S34 AND S35 (1,283)

AMED (Ovid)

1 Tibial Fractures/ (190)
2 Ankle/ or Ankle Joint/ or Ankle Injuries/ or Tibia/ or Fibula/ (7408)
3 (ankle or malleol*).tw. (10790)
4 (distal adj2 (tibia* or fibula*)).tw. (440)
5 2 or 3 or 4 (11754)
6 Fractures, Bone/ (1645)
7 fracture*.tw. (6260)
8 6 or 7 (6260)
9 5 and 8 (1601)
10 1 or 9 (1710)
11 Randomized controlled trial.pt. (5470)
12 Controlled clinical trial.pt. (70)
13 Randomized Controlled Trials/ (2954)
14 Random Allocation/ (335)
15 Double‐Blind Method/ (870)
16 clinical trial.pt. (1264)
17 exp Clinical trials/ (4934)
18 (clinic* adj25 trial*).tw. (9078)
19 ((singl* or doubl* or trebl* or trip*) adj (mask* or blind*)).tw. (3739)
20 Placebos/ (746)
21 placebo*.tw. (3966)
22 random*.tw. (23718)
23 exp Research design/ (21410)
24 (latin adj square).tw. (28)
25 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 (43692)
26 exp Animals/ not Humans/ (14603)
27 25 not 26 (42287)
28 10 and 27 (247)
29 (2011* or 2012* or 2013* or 2014* or 2015* or 2016* or 2017* or 2018* or 2019* or 2020* or 2021* or 2022*).up,yr. (115703)
30 28 and 29 (99)

SPORTDiscus

S1 DE "ANKLE fractures" (732)
S2 DE "ANKLE" OR DE "ANKLE joint" OR DE "ANKLEBONE" OR DE "TIBIA" OR DE "FIBULA" (10,604)
S3 TX ankle or TX malleol* (26,541)
S4 TX (distal n2 (tibia* or fibula*)) (917)
S5 S2 OR S3 OR S4 (29,546)
S6 DE "FRACTURES" (2,331)
S7 TX fracture* (27,458)
S8 S6 OR S7 (27,458)
S9 S5 AND S8 (4,116)
S10 S1 OR S9 (4,116)
S11 TX ((clinical or controlled or comparative or placebo or prospective or randomi?ed) and (trial or study)) (181,430)
S12 TX (random* and (allocat* or allot* or assign* or basis* or divid* or order*)) (24,300)
S13 TX ((singl* or doubl* or trebl* or tripl*) and (blind* or mask*)) (10,535)
S14 TX ( crossover* or 'cross over' ) or TX cross n1 over (8,759)
S15 TX ((allocat* or allot* or assign* or divid*) and (condition* or experiment* or intervention* or treatment* or therap* or control* or group*)) (37,883)
S16 S11 OR S12 OR S13 OR 14 OR 15 (208,4536)
S17 S10 AND S16 (1,331)
S18 S10 AND S16 Date limit: 01012011‐05052022 (900)

WHO ICTRP

ankle* AND fracture* OR malleol* AND fracture* (297)

ClinicalTrials.gov

(fracture OR fractures) AND (ankle OR malleolar OR malleolus) | First posted from 07/01/2011 to 05/05/2022 (318)

Appendix 2. Multi‐arm studies. Data for comparisons not included in the main analysis tables

Outcome (measurement tool; time point)	Study ID	Comparison (as described by study authors)	Effect estimatea
Activity limitation (OMAS; end of treatment)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD ‐6, 95% CI ‐12.17 to 0.17; 156 participants
Activity limitation (FAOS ‐ ADL subscore; end of treatment)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 1.00, 95% CI ‐3.74 to 5.74; 150 participants
Activity limitation (OMAS; end of treatment)	Smeeing 2020	Protected weight‐bearing vs unprotected non‐weight‐bearing After surgical management	MD 6.00, 95% CI ‐3.82 to 15.82; 73 participants
Activity limitation (OMAS; short‐term follow‐up ‐ 12 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 4.00, 95% CI ‐2.89 to 10.89; 145 participants
Activity limitation (FAOS ‐ ADL subscore; short‐term follow‐up ‐ 12 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 3.00, 95% CI ‐1.71 to 7.71; 141 participants
Activity limitation (OMAS; short‐term follow‐up ‐ 6 months)	Smeeing 2020	Protected weight‐bearing vs unprotected non‐weight‐bearing After surgical management	MD 5.10, 95% CI ‐2.27 to 12.47; 73 participants
Activity limitation (questionnaire; long‐term follow‐up ‐ 52 weeks)	Finsen 1989	Plaster cast with weight‐bearing vs exercise	MD 0.00, 95% CI ‐0.42 to 0.42; 27 participants
Activity limitation (OMAS; long‐term follow‐up ‐ 52 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 4.00, 95% CI ‐1.09 to 9.09; 146 participants
Activity limitation (FAOS ‐ ADL subscore; long‐term follow‐up ‐ 52 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 3.00, 95% CI ‐0.20 to 6.20; 145 participants
Activity limitation (OMAS; long‐term follow‐up ‐ 52 weeks)	Smeeing 2020	Protected weight‐bearing vs unprotected non‐weight‐bearing After surgical management	MD 0.40, 95% CI ‐5.72 to 6.52; 73 participants
Quality of life (RAND‐36 physical role function score; end of treatment)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 3.00, 95% CI ‐5.89 to 11.89; 155 participants
Quality of life (RAND‐36 physical role function score; short‐term follow‐up ‐ 12 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 12.00, 95% CI ‐2.06 to 26.06; 144 participants
Quality of life (SF‐36 physical subscale; short‐term follow‐up ‐ 12 weeks)	Smeeing 2020	Protected weight‐bearing vs unprotected non‐weight‐bearing After surgical management	MD 0.60, 95% CI ‐8.87 to 10.07; 73 participants
Quality of life (RAND‐36 physical role function score; long‐term follow‐up ‐ 52 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 0.00, 95% CI ‐9.64 to 9.64; 144 participants
Quality of life (SF‐36 physical subscale; long‐term follow‐up ‐ 52 weeks)	Smeeing 2020	Protected weight‐bearing vs unprotected non‐weight‐bearing After surgical management	MD ‐2.60, 95% CI ‐8.82 to 3.62; 73 participants
Pain (100 point VAS; end of treatment)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD 9.00, 95% CI 2.37 to 15.63; 147 participants
Pain (100 point VAS; short‐term follow‐up ‐ 12 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD ‐3.00, 95% CI ‐9.91 to 3.91; 142 participants
Pain (100 point VAS; long‐term follow‐up ‐ 52 weeks)	Kortegangas 2019	3‐week cast vs 6‐week cast; after non‐surgical management	MD ‐3.00, 95% CI ‐8.00 to 2.00; 144 participants

ADL: activities of daily living; CI: confidence interval; FAOS: Foot and Ankle Outcome Score; MD: mean difference; OMAS: Olerud and Molander Ankle Score; RAND‐36: SF‐36 score developed by organisation called RAND; SF‐36: Short Form Health Survey; VAS: visual analogue scale

Appendix 3. Scales used in outcome measures

Outcome	Scale	Commonly used abbreviation	Range	Direction of scale
Activity limitation	Olerud & Molander Ankle Score (Olerud 1984)	OMAS	0 to 100	Higher scores indicate better outcomes and fewer symptoms except for Kearney 2021 where scores range from 0 for totally impaired to 100 for completely unimpaired.
	Foot and Ankle Outcome Score (Roos 2001)	FAOS	0 to 100	Higher scores indicate better function
	American Academy of Orthopaedic Surgeons Foot and Ankle score (Johanson 2004)	AAOS Foot and Ankle Questionnaire	0 to 100	Higher scores indicate better function
	Mazur score (Mazur 1979)	‐	0 to 100	Higher scores indicate better function
	Philip score (Phillips 1985)	‐	0 to 100	Lower scores indicate better function
	American Orthopaedic Foot and Ankle Society (Kitaoka 1994)	AOFAS	0 to 100	Higher scores indicate better function
	Visual Analogue Scale for Foot and Ankle (Richter 2006)	VAS FA	0 to 100	Higher scores indicate better function
	Manchester Oxford Foot Questionnaire (Morley 2013)	‐	0 to 100	Higher scores indicate better function
	Lower Extremity Functional Scale (Binkley 1999)	LEFS	0 to 80	Higher scores indicate better function
Participant satisfaction	Visual Analogue Scale of patient satisfaction	VAS	0 to 10 or 0 to 100	Higher scores indicate higher satisfaction
Quality of life	RAND‐36 physical role function score (Aalto 1999)	RAND‐36	0 to 100	Higher scores indicate better quality of life
	EuroQol‐5 Dimension (EQ‐5D)	EQ‐5D	−0.224 to 1	Higher scores indicate better quality of life
	Short Form Health Survey (‐12 or ‐36) (Ware 1992)	SF‐12, SF‐36	0 to 100	Higher scores indicate better quality of life
	Visual Analogue Scale for Quality of Life	VAS	0 to 100	Higher scores indicate better quality of life
	Foot and Ankle Outcome Score qualify of life subscore (Roos 2001)	FAOS	0 to 100	Higher scores indicate better quality of life
	Assessment of Quality of Life 45 (Hawthorne 2001)	AQoL	0 to 45	Higher scores indicate better quality of life
Pain	Visual Analogue Scale of pain (Chapman 1985)	VAS	0 to 10 or 0 to 100	Lower scores indicate less pain
	Foot and Ankle Outcome Score pain subscore (Roos 2001)	FOAS	0 to 100	Lower scores indicate less pain

Appendix 4. Early vs delayed weight‐bearing. Not included in analysis because data were insufficient to calculate effect estimates

Outcome	Study ID	Early weight‐bearing	Late weight‐bearing	Between‐group differencea
Activity limitation; measured using OMAS at end of treatment	Van Laarhoven 1996	Median = 65	Median = 50	P = 0.03
Activity limitation; measured using OMAS at end of treatment	Honigmann 2007	Median (range) = 72 (35 to 95)	Median (range) = 70 (45 to 90)	Difference = 1.9
Activity limitation; measured using OMAS at short‐term follow‐up (10 weeks)	Honigmann 2007	Median (range) = 80 (40 to 100)	Median (range) = 85 (40 to 100)	Difference = ‐5.0
Activity limitation; measured using OMAS at long‐term follow‐up (18 months)	Ahl 1993	Mean = 90	Mean = 86	MD = 4
Activity limitation; measured using OMAS at long‐term follow‐up (12 months)	Van Laarhoven 1996	Median (range) = 95 (0 to 100)	Median range = 95 (35 to 100)	P = 0.94
Quality of life; measured using SF‐12 (MCS) at end of treatment	Franke 2008	Median = 59.9	Median = 52	Difference = 7.2; P = 0.008

^aP values as reported by study authors; OMAS: Olerud and Molander Ankle Score; SF‐12 (MCS): Short Form Health Survey‐12 (mental component score)

Appendix 5. Economic data

Study ID	Parameters for calculations	Costs
Brink 1996	Cost of device (brace or walker)	Aircast ankle brace: US dollars (USD) 40 (Scandinavia) DonJoy RoM Walker: USD 110 (Scandinavia)
Franke 2008	Time spent treating the participant (for a sample of participants); costs for the cast, dressings, orthosis, thrombosis prophylaxis, and physiotherapy	Removable ankle support: 381 euros Non‐removable ankle support: 419 euros
Kearney 2021	Resource use and QALYs calculations	Removable brace was more costly (MD £67, 95% CI ‐9 to 139), but increased quality of life (MD 0.01 QALYs, 95% CI 0.004 to 0.015) at 16 weeks. The incremental cost‐effectiveness ratio was £6887/QALY (in favour of removable brace) £ = GBP (pounds sterling)
Lin 2008a	Utilisation of healthcare resources and participants' out‐of‐pocket costs, collected every 4 weeks by survey. Direct healthcare costs included: outpatient physiotherapy, medical specialists, GP, emergency department, hospitalisation, investigations, PBS medication; private health providers, gap costs of visits to medical specialists and GP as well as investigations and PBS medication, non‐PBS medication and other purchases. Direct non‐healthcare costs included: public transport, tolls, parking, use of private vehicle	Direct healthcare costs for enhanced physiotherapy: AUD (Australian dollar) 492.88 Direct non‐healthcare costs for enhanced physiotherapy: AUD 577.04 Direct health care costs for usual care physiotherapy: AUD 676.64 Direct non‐healthcare costs for usual care physiotherapy: AUD 295.52
Moseley 2015	Direct costs to the health system and out‐of‐pocket costs to the participants (over 6 months) included: visits to the hospital physical therapist, visits to hospital or private physical therapists, medical specialists, primary care physicians, community services or alternative or complementary health practitioners, emergency department visits and hospital admissions, medications, and equipment, number of days away from paid work and unpaid activities	Direct costs to healthcare system for tailored supervised exercises, mean (SD): AUD (Australian dollar) 500.86 (1526.94) Out‐of‐pocket costs for tailored supervised exercises, mean (SD): AUD 245.57 (431.39) Direct costs to healthcare system for advice, mean (SD): AUD 192.10 (888.23) Out‐of‐pocket costs for advice, mean (SD): AUD 292.97 (579.79)
Stockle 2000	Cost of device	Removable ankle support (orthosis). Rental for 4 weeks: DM 503; rental for 6 to 8 weeks: DM 607; purchase price: DM 798 Non‐removable ankle support. Material costs: DM 165; treatment costs: DM 126; cleaning and waste costs: DM 40
Bretherton 2024	Costs to NHS and PSS, calculated using estimates of resource inputs associated with the broader utilisation of hospital and community‐based health and social care services, out‐of‐pocket medical expenses, and values of work losses and additional care costs associated with ankle fracture surgery	Early weight‐bearing, mean costs from NHS and PSS perspective: £725 Delayed weight‐bearing, mean costs from NHS and PSS perspective: £785 £ = GBP (pounds sterling)

CI: confidence interval; DM: Deutschmark (German currency up to introduction of euro in 2002); GP: general practitioner; MD: mean difference; NHS: National Health Service; PBS: Pharmaceutical Benefits Scheme; PSS: personal social services; QALYs: quality‐adjusted life years; SD: standard deviation

Appendix 6. Removable ankle support vs non‐removable ankle support. Outcome data for outcomes measured using alternative measurement tools

Outcome (measurement tool; time point)	Study ID	After surgical or non‐surgical fracture management	Effect estimatea
Activity limitation (Manchester Oxford Foot Questionnaire; 16 weeks)	Kearney 2021	After surgical management	MD ‐2.97, 95% CI ‐9.04 to 3.09; 234 participants
Activity limitation (Manchester Oxford Foot Questionnaire; 16 weeks)	Kearney 2021	After non‐surgical management	MD ‐0.62, 95% CI ‐6.98 to 5.73; 217 participants
Activity limitation (able to climb 12 steps independently; 10 weeks)	Losch 2002	After surgical management	RR 2.25, 95% CI 1.26 to 4.03; 33 participants
Activity limitation (FAOS ‐ ADL subscore; end of treatment)	Kortegangas 2019	After non‐surgical management	MD ‐1.00, 95% CI ‐6.66 to 4.66; 147 participants
Activity limitation (FAOS ‐ ADL subscore; short‐term follow‐up ‐ 12 weeks)	Kortegangas 2019	After non‐surgical management	MD ‐3.00, 95% CI ‐8.36 to 2.36; 137 participants
Activity limitation (FAOS ‐ ADL subscore; long‐term follow‐up ‐ 52 weeks)	Kortegangas 2019	After non‐surgical management	MD 0.00, 95% CI ‐4.58 to 4.58; 137 participants
Activity limitation (AAOS Foot and Ankle Score; long‐term follow‐up ‐ 52 weeks)	Van den Berg 2018	After non‐surgical management	MD 4.83, 95% CI ‐2.39 to 12.05; 29 participants

^aCalculated with the calculator in Review Manager (RevMan 2024).

AAOS: American Academy of Orthopedic Surgeons; CI: confidence interval; FAOS ‐ ADL: Foot and Ankle Outcome Score ‐ activities of daily living; MD: mean difference; RR: risk ratio

Appendix 7. Removable ankle support vs non‐removable ankle support. Not included in analysis because data were insufficient to calculate effect estimates

Outcome	Study ID	Removable ankle support	Non‐removable ankle support	Between‐group difference
Activity limitation; measured using 10‐point activity limitation questionnaire at end of treatment	Wetzler 1991	Mean = 8.7	Mean = 6.8	MD = 1.9; reported as "statistically significant"
Activity limitation; measured using activity questionnaire (number with mild, moderate, or severe symptoms) at short‐term follow‐up (9 weeks)	Venesmaa 2004	Mild: 13/18 Moderate: 5/18	Mild: 4/14 Moderate: 9/14 Severe: 1/14	P = 0.013
Activity limitation; measured using Maryland Foot Score at end of follow‐up (6 months)	DiStasio 1994	Mean = 93	Mean = 87	P = 0.027
Activity limitation; measured using OMAS at end of follow‐up (18 months)	Hedstrom 1994	Median (range) = 100 (80 to 100)	Median (range) = 88 (55 to 100)	Reported as "not statistically significant"
Activity limitation; measured using OMAS at end of follow‐up (10 weeks)	Franke 2008	Median = 95	Median = 75	P = 0.02
Activity limitation; measured using 10‐point activity limitation questionnaire at end of follow‐up (12 months)	Wetzler 1991	No data reported	No data reported	Reported as "not statistically significant"
Pain; measured using 10‐point scale at end of treatment	Wetzler 1991	Mean = 3.2	Mean = 5.5	MD = 2.3; reported as "statistically significant"
Pain; measured using 10‐point scale at end of follow‐up (12 months)	Wetzler 1991	No data reported	No data reported	Reported as "not statistically significant"

OMAS: Olerud and Molander Ankle Score; MD: mean difference

Appendix 8. Removable ankle support vs non‐removable ankle support: sensitivity analyses

Outcome	Studies removed from primary analysis	Interpretation of effect estimate after sensitivity analysis
Analysis 2.2 After surgical fracture management	High or unclear risk of selection bias:Dehghan 2016; Siddique 2005; Smeeing 2020; Vioreanu 2007 Risk of attrition bias:Kearney 2021; Matthews 2020	High or unclear risk of selection bias: imprecise effect estimate (MD 8.26, 95% CI ‐0.66 to 17.18; I2 = 73%; 2 studies, 401 participants) Risk of attrition bias: imprecise effect estimate (MD 5.12, 95% CI ‐1.71 to 11.96; I2 = 69%; 4 studies, 276 participants)
Analysis 2.5 After surgical fracture management	High or unclear risk of selection bias:Dehghan 2016 Risk of attrition bias:Kearney 2021	High or unclear risk of selection bias: imprecise effect estimate (MD 0.05, 05% CI ‐0.01 to 0.11; 1 study, 266 participants) Risk of attrition bias: now favours treatment with non‐removable ankle support (MD 8.50, 95% CI 2.33 to 14.67; 1 study, 104 participants)
Analysis 2.5 After non‐surgical fracture management	Risk of attrition bias:Kearney 2021	Risk of attrition bias: still imprecise but favours alternative treatment (SMD ‐0.06, 95% CI ‐0.35 to 0.22, favours fixed support; 2 studies, 191 participants)
Analysis 2.6 After surgical fracture management	Risk of attrition bias:Kearney 2021; Matthews 2020	Risk of attrition bias: imprecise effect estimate (MD 1.30, 95% CI ‐7.62 to 10.22; 1 study, 78 participants)
Analysis 2.10 After non‐surgical fracture management End of treatment	High or unclear risk of selection bias:Van den Berg 2018	Risk of attrition bias: favours non‐removable ankle support (MD 11, 95% CI 4.32 to 17.68; 1 study, 145 participants) Effects model: using a fixed effect model resulted in a more precise result favouring non‐removable ankle support (SMD 0.32, 95% CI 0.03 to 0.62; 2 studies, 184 participants)
Analysis 2.11 Adverse events after surgical fracture management	High or unclear risk of selection bias (unplanned return to theatre):Davies 1991; Dehghan 2016; Hedstrom 1994; Stockle 2000 High risk of attrition bias (unplanned return to theatre):Davies 1991; Hedstrom 1994; Kearney 2021 High or unclear risk of selection bias (detection bias):Lehtonen 2003; Smeeing 2020; Vioreanu 2007	High or unclear risk of selection bias (unplanned return to theatre): now favours non‐removable ankle support (RR 1.94, 95% CI 1.01 to 3.73; 2 studies, 387 participants) High risk of attrition bias (unplanned return to theatre): effect estimate still imprecise but favours alternative treatment (RR 0.50, 95% CI 0.07 to 3.69; I2 = 45%; 3 studies, 166 participants) High or unclear risk of selection bias (detection bias): effect estimate still imprecise but favours alternative treatment (RR 1.50, 95% CI 0.25 to 8.87, favours fixed ankle support; 1 study 164 participants) Effects model: using a fixed‐effect model resulted in a more precise result favouring non‐removable ankle support (RR 1.71, 1.00 to 2.93; 9 studies, 627 participants)
Analysis 2.12 Adverse events after non‐surgical fracture management	High risk of attrition bias (DVT):Kearney 2021; Kortegangas 2019	High risk of attrition bias (DVT): imprecise effect estimate (RR 0.19, 95% CI 0.01 to 3.71; 1 study, 37 participants)

2.12. Analysis.

Comparison 2: Removable ankle support versus non‐removable ankle support (e.g. cast), Outcome 12: Adverse events (after non‐surgical fracture management)

CI: confidence interval; DVT: deep vein thrombosis; MD: mean difference; RR: risk ratio; SMD: standardised mean difference

Appendix 9. Physical therapy. Outcome data for outcomes measured using multiple measurement tools

Outcome (measurement tool; time point)	Study ID	Comparison(s)	Effect estimatea
Activity limitation (Philip score; end of treatment)	Jansen 2018	ACM and usual care vs usual care During initial fracture management period after surgical management	MD 6.70, 95% CI ‐1.33 to 14.73; 48 participants
Activity limitation (AOFAS; end of treatment)	Jansen 2018	ACM and usual care vs usual care During initial fracture management period after surgical management	MD 7.60, 95% CI 1.67 to 13.53; 48 participants
Activity limitation (VAS FA score; end of treatment)	Jansen 2018	ACM and usual care vs usual care During initial fracture management period after surgical management	MD 15.40, 95% CI 8.49 to 22.31; 48 participants
Activity limitation (Philip score; short‐term follow‐up ‐ 12 weeks)	Jansen 2018	ACM and usual care vs usual care During initial fracture management period after surgical management	MD 19.00, 95% CI 8.85 to 29.15; 44 participants
Activity limitation (AOFAS; short‐term follow‐up ‐ 12 weeks)	Jansen 2018	ACM and usual care vs usual care During initial fracture management period after surgical management	MD 12.30, 95% CI 6.40 to 18.20; 44 participants
Activity limitation (VAS FA score; short‐term follow‐up ‐ 12 weeks)	Jansen 2018	ACM and usual care vs usual care During initial fracture management period after surgical management	MD 16.30, 95% CI 7.38 to 25.22; 44 participants
Activity limitation (OMAS; short‐term follow‐up ‐ 6 months)	Keene 2022	Progressive exercise and advice vs best practice advice After initial fracture management period after surgical or non‐surgical management	MD 3.73, 95% CI ‐7.96 to 15.42; 54 participants
Pain (on walking using 100‐point VAS; short‐term follow‐up at 6 months)	Keene 2022	Progressive exercise and advice vs best practice advice After initial fracture management period after surgical or non‐surgical management	MD ‐1.50, 95% CI ‐12.93 to 9.93; 53 participants
Pain (on stair descent using 100‐point VAS; short‐term follow‐up at 24 weeks)	Lin 2008b	Joint mobilisation using enhanced physiotherapy vs usual care physiotherapy After initial fracture management period after surgical or non‐surgical management	MD 0.1, 95% CI –5.1 to 5.3; 94 participantsb
Pain (on stair descent using 10‐point VAS; short‐term follow‐up at 6 months)	Moseley 2015	Supervised exercise programme and advice vs advice only After initial fracture management period after surgical or non‐surgical treatment	MD ‐0.10, 95% CI ‐0.55 to 0.35; 171 participants

^aEffect estimates calculated using calculator in RevMan (unless otherwise stated; RevMan 2024)
^bEffect estimate as reported by study authors

ACM: active controlled motion; AOFAS: American Orthopedic Foot and Ankle Society score; CI: confidence interval; MD: mean difference; OMAS: Olerud and Molander Ankle Score; VAS (FA): visual analogue scale (foot and ankle score)

Appendix 10. Other comparisons. Not included in analysis because data were insufficient to calculate effect estimates

Outcome	Study ID	Aircast	DonJoy walker	Between‐group difference
Activity limitation measured using "Inflammatory score" questionnaire (out of 48) at end of treatment. Lower scores indicate better function	Brink 1996	Median = 16	Median = 10	P value < 0.01
Activity limitation measured using "Inflammatory score" questionnaire (out of 48) at short‐term follow‐up (12 weeks). Lower scores indicate better function	Brink 1996	Median = 5	Median = 2	Not reported
Participant satisfaction (scores out of 5) at end of treatment. Higher scores indicate greater satisfaction.	Brink 1996	Median = 5	Median = 4	P value < 0.05
Pain (scores out of 10) at end of treatment. Lower scores indicate less pain	Brink 1996	Median = 0.9	Median = 0.3	P value < 0.05

Data and analyses

Comparison 1. Early versus delayed weight‐bearing.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Activity limitation (using OMAS; at the end of treatment)	5	901	Mean Difference (IV, Random, 95% CI)	9.14 [4.60, 13.68]
1.2 Activity limitation (using OMAS; short‐term follow‐up, ≤ 6 months)	5	890	Mean Difference (IV, Random, 95% CI)	3.56 [1.35, 5.78]
1.3 Activity limitation (using different measurement tools; long‐term follow‐up, > 6 months)	5	865	Std. Mean Difference (IV, Random, 95% CI)	0.10 [‐0.07, 0.27]
1.4 HRQoL (using different measurement tools; at end of treatment)	4	665	Std. Mean Difference (IV, Random, 95% CI)	0.39 [0.24, 0.55]
1.5 HRQoL (using different measurement tools; short‐term follow‐up, ≤ 6 months)	5	739	Std. Mean Difference (IV, Random, 95% CI)	0.15 [‐0.01, 0.30]
1.6 HRQoL (using different measurement tools; long‐term follow‐up, > 6 months)	3	641	Std. Mean Difference (IV, Random, 95% CI)	0.11 [‐0.21, 0.43]
1.7 Participant satisfaction (VAS, /10; end of treatment, and short‐term follow‐up of ≤ 6 months)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.7.1 End of treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.7.2 Short‐term follow‐up (6 months)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.8 Pain (VAS; end of treatment)	2	92	Mean Difference (IV, Random, 95% CI)	0.16 [‐0.28, 0.59]
1.9 Pain (VAS; short‐term follow‐up, ≤ 6 months)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.10 Adverse events	12		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.10.1 Re‐operation	7	1007	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.09, 2.68]
1.10.2 Re‐dislocation	3	137	Risk Ratio (M‐H, Random, 95% CI)	0.73 [0.04, 13.21]
1.10.3 Delayed union	1	167	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.10.4 Non‐union	2	728	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.01, 8.18]
1.10.5 Malunion	1	561	Risk Ratio (M‐H, Random, 95% CI)	3.01 [0.12, 73.59]
1.10.6 Loss of fixation	2	663	Risk Ratio (M‐H, Random, 95% CI)	4.01 [0.45, 35.69]
1.10.7 Ankle metalwork or implant breakage	1	561	Risk Ratio (M‐H, Random, 95% CI)	3.01 [0.12, 73.59]
1.10.8 Loss of reduction	2	269	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.10.9 DVT	3	687	Risk Ratio (M‐H, Random, 95% CI)	1.25 [0.28, 5.67]
1.10.10 Symptomatic pulmonary embolism	1	561	Risk Ratio (M‐H, Random, 95% CI)	3.01 [0.12, 73.59]
1.10.11 Vascular injury	1	561	Risk Ratio (M‐H, Random, 95% CI)	3.01 [0.12, 73.59]
1.10.12 Deep infection	3	200	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
1.10.13 Surgical site infection	1	561	Risk Ratio (M‐H, Random, 95% CI)	1.25 [0.60, 2.63]
1.10.14 Low grade infection or superficial infection	8	1017	Risk Ratio (M‐H, Random, 95% CI)	1.83 [0.89, 3.76]
1.10.15 Wound dehiscence	3	749	Risk Ratio (M‐H, Random, 95% CI)	1.33 [0.47, 3.80]
1.10.16 Wound healing complications	1	107	Risk Ratio (M‐H, Random, 95% CI)	3.06 [0.33, 28.46]
1.10.17 Prominent implant or metalwork irritation	1	561	Risk Ratio (M‐H, Random, 95% CI)	1.16 [0.56, 2.39]
1.10.18 Swelling, skin irritation and/or blistering	2	78	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.10, 8.78]
1.10.19 Dystrophy/atrophy	2	122	Risk Ratio (M‐H, Random, 95% CI)	0.29 [0.01, 7.02]
1.10.20 Complex regional pain syndrome	1	561	Risk Ratio (M‐H, Random, 95% CI)	0.11 [0.01, 2.06]
1.10.21 Fall	1	561	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.01, 8.18]
1.10.22 Re‐fracture of the same ankle	1	561	Risk Ratio (M‐H, Random, 95% CI)	2.01 [0.18, 22.01]
1.10.23 Nerve palsy	1	561	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.17, 3.33]

Comparison 2. Removable ankle support versus non‐removable ankle support (e.g. cast).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Activity limitation after surgical fracture management (using different measurement tools; end of treatment)	7	682	Std. Mean Difference (IV, Random, 95% CI)	0.36 [0.12, 0.59]
2.2 Activity limitation (using different measurement tools; short‐term follow‐up, ≤ 6months)	8		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.2.1 After surgical fracture management	6	677	Mean Difference (IV, Random, 95% CI)	6.39 [1.69, 11.09]
2.2.2 After non‐surgical fracture management	3	399	Mean Difference (IV, Random, 95% CI)	1.08 [‐3.18, 5.34]
2.3 Activity limitation (using different measurement tools; long‐term follow‐up, > 6 months)	8		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.3.1 After surgical fracture management	6	380	Std. Mean Difference (IV, Random, 95% CI)	0.11 [‐0.10, 0.31]
2.3.2 After non‐surgical fracture management	2	168	Std. Mean Difference (IV, Random, 95% CI)	0.18 [‐0.13, 0.48]
2.4 Activity limitation after non‐surgical fracture management (using OMAS; end of treatment)	3	423	Mean Difference (IV, Random, 95% CI)	‐0.89 [‐6.74, 4.96]
2.5 HRQoL (using different measurement tools; end of treatment)	4		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.5.1 After surgical fracture management	2	370	Std. Mean Difference (IV, Random, 95% CI)	0.32 [‐0.00, 0.64]
2.5.2 After non‐surgical fracture management	3	424	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.17, 0.21]
2.6 HRQoL (using different measurement tools; short‐term follow‐up, ≤ 6 months)	5		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.6.1 After surgical fracture management	3	477	Std. Mean Difference (IV, Random, 95% CI)	0.30 [0.11, 0.50]
2.6.2 After non‐surgical fracture management	3	397	Std. Mean Difference (IV, Random, 95% CI)	‐0.04 [‐0.24, 0.15]
2.7 HRQoL (using different measurement tools; long‐term follow‐up, > 6 months)	3		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.7.1 After surgical fracture management	2	178	Mean Difference (IV, Random, 95% CI)	2.60 [‐5.12, 10.31]
2.7.2 After non‐surgical fracture management	1	137	Mean Difference (IV, Random, 95% CI)	‐1.00 [‐11.59, 9.59]
2.8 Pain after surgical fracture management (numbers with pain; all time points)	3		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
2.8.1 End of treatment	1	40	Risk Ratio (M‐H, Random, 95% CI)	0.12 [0.03, 0.44]
2.8.2 Short‐term follow‐up ≤ 6 months	1	29	Risk Ratio (M‐H, Random, 95% CI)	0.74 [0.52, 1.07]
2.8.3 Long‐term follow‐up > 6 months	3	116	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.35, 1.14]
2.9 Pain after non‐surgical fracture management (VAS; end of treatment)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.10 Pain after non‐surgical fracture management (VAS; short‐term and long‐term follow‐up)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.10.1 Short‐term follow‐up ≤ 6 months	2	167	Std. Mean Difference (IV, Random, 95% CI)	‐0.10 [‐0.73, 0.53]
2.10.2 Long‐term follow‐up > 6 months	2	169	Std. Mean Difference (IV, Random, 95% CI)	‐0.07 [‐0.37, 0.23]
2.11 Adverse events (after surgical fracture management)	15		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
2.11.1 Unplanned return to theatre	6	624	Risk Ratio (M‐H, Random, 95% CI)	1.20 [0.39, 3.71]
2.11.2 Non‐union	1	364	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.21, 1.87]
2.11.3 Loss of fixation	3	257	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.13, 71.92]
2.11.4 Loss of reduction	4	222	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
2.11.5 Fixation/implant failure	2	113	Risk Ratio (M‐H, Random, 95% CI)	3.10 [0.13, 73.16]
2.11.6 Refracture of the lateral malleolus	1	100	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.13, 71.92]
2.11.7 DVT	4	604	Risk Ratio (M‐H, Random, 95% CI)	0.60 [0.14, 2.59]
2.11.8 Pulmonary embolism	2	419	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.08, 5.05]
2.11.9 Deep infection	4	330	Risk Ratio (M‐H, Random, 95% CI)	6.38 [0.80, 51.17]
2.11.10 Wound dehiscence	3	288	Risk Ratio (M‐H, Random, 95% CI)	2.50 [0.34, 18.55]
2.11.11 Low grade/superficial infection	9	627	Risk Ratio (M‐H, Random, 95% CI)	1.64 [0.92, 2.95]
2.11.12 Wound healing complications	3	186	Risk Ratio (M‐H, Random, 95% CI)	2.28 [0.47, 11.18]
2.11.13 Local skin necrosis	1	100	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.13, 71.92]
2.11.14 Pressure sore/ulcer	1	364	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.21, 1.87]
2.11.15 Dystrophy	1	78	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
2.11.16 Numbness at side of foot	1	364	Risk Ratio (M‐H, Random, 95% CI)	0.84 [0.59, 1.20]
2.11.17 Skin irritation	1	100	Risk Ratio (M‐H, Random, 95% CI)	2.00 [0.19, 21.36]
2.11.18 Dysesthesia of the skin	1	100	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.32, 27.87]
2.11.19 Chronic allodynia	1	100	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.13, 71.92]
2.12 Adverse events (after non‐surgical fracture management)	3		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
2.12.1 Unplanned return to theatre	1	305	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.14, 6.96]
2.12.2 Non‐union	2	469	Risk Ratio (M‐H, Random, 95% CI)	1.24 [0.34, 4.54]
2.12.3 Loss of congruity	1	164	Risk Ratio (M‐H, Random, 95% CI)	Not estimable
2.12.4 Nerve compression	1	164	Risk Ratio (M‐H, Random, 95% CI)	2.10 [0.19, 22.71]
2.12.5 DVT	3	506	Risk Ratio (M‐H, Random, 95% CI)	0.20 [0.04, 0.93]
2.12.6 Pulmonary embolism	1	305	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.01, 8.07]
2.12.7 Numbness at side of foot	1	305	Risk Ratio (M‐H, Random, 95% CI)	0.84 [0.39, 1.82]
2.12.8 Pressure sore/ulcer	1	305	Risk Ratio (M‐H, Random, 95% CI)	0.25 [0.03, 2.20]
2.12.9 Skin irritation/ulceration	2	201	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.06, 9.67]

Comparison 3. Physical therapy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Activity limitation (using different measurement tools; at end of treatment)	7		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.1.1 During initial fracture management period after surgical treatment	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.1.3 After the initial fracture management period after either surgical or non‐surgical treatment	4		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.2 Activity limitation (using different measurement tools; short‐term follow‐up, ≤ 6 months)	8		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.2.1 During initial fracture management period after surgical treatment	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.2.2 After the initial fracture management period after either surgical or non‐surgical treatment	5		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.3 Activity limitation (using different measurement tools; long‐term follow‐up, > 6 months)	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.3.1 During initial fracture management period after surgical treatment	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.3.2 After initial fracture management period after surgical or non‐surgical treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.4 HRQoL (using different measurement tools; at end of treatment)	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.4.1 During initial fracture management period after surgical treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.4.3 After initial fracture management period after either surgical or non‐surgical treatment	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.5 HRQoL (using different measurement tools; short‐term follow‐up, ≤ 6 months)	5		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.5.1 During initial fracture management period after surgical treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.5.2 After initial fracture management period after either surgical or non‐surgical treatment	4		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.6 HRQoL (using different measurement tools; long‐term follow‐up, > 6 months)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.6.1 During initial fracture management period after surgical treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.6.2 After initial fracture managment period after either surgical or non‐surgical treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.7 Participant satisfaction (100‐mm VAS; end of treatment)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.7.3 After initial fracture management period after either surgical or non‐surgical treatment	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.8 Pain (using different measures; at end of treatment	5		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.8.1 During initial fracture management period after surgical treatment	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.8.2 After initial fracture management period after either surgical or non‐surgical treatment	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.9 Pain (using different measurement tools; short‐term follow‐up, ≤ 6 months)	6		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.9.1 During initial fracture management period after surgical treatment	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.9.2 After initial fracture management period after either surgical or non‐surgical treatment	4		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.10 Pain (using different measurement tools; long‐term follow‐up, > 6 months)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.10.1 During initial fracture management period after surgical treatment	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
3.11 Pain (numbers with pain; long‐term/end of follow‐up	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.11.1 After initial fracture management period after either surgical or non‐surgical treatment	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12 Adverse events (during initial fracture management period after surgical treatment)	2		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.2 Set screw breakage	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.3 Loosening hem around large fragment screw	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.4 Deep infection	2		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.5 Low grade infection/superficial infection	2		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.6 Supination trauma	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.7 Sensitivity disorders	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.8 Achilles tendon pain	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.9 Neuropathic pain from hip to ankle	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.10 Full body rash and pruritus	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.11 Rheumatic episode	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.12 Sacral burn	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.13 Depression	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.14 Metatarsal fracture (contralateral foot)	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.16 Scar impingement	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.17 Pseudarthrosis of the tibia shaft	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.18 Material fracture in pseudarthrosis treatment on the tibia shaft	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.19 Syndesmosis insufficiency	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.20 Neuritis in the head	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.21 Tendinopathy M.tibialis posterior tendon	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.22 Tear of the anterior cruciate ligament after a fall	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.23 Wedge‐shaped bone defect distal tibial joint surface ventral	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.24 Subjective change in skin colour in ankle and foot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.12.25 Hardly adjustable hypertension	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13 Adverse events (after initial fracture management period after either surgical or non‐surgical treatment)	5		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13.1 Total adverse events (type not specified)	2		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13.2 DVT	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13.3 Pain	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13.4 Recent falls (in previous 3 months)	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13.5 Increase in pain	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
3.13.6 Condition worsened	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

3.11. Analysis.

Comparison 3: Physical therapy, Outcome 11: Pain (numbers with pain; long‐term/end of follow‐up

Comparison 4. Other rehabilitation interventions: electrotherapy versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Activity limitation (interventions given during initial fracture management period; OMAS; all time points)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.1.1 End of treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.1.2 Short‐term follow‐up	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.1.3 Long‐term follow‐up	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.2 Pain (10‐cm VAS; end of treatment)	3		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.2.1 During initial fracture management period	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.2.2 After initial fracture management period	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.3 Pain (interventions given during the initial fracture management period; VAS; short‐term follow‐up, ≤ 6 months)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.4 Pain (numbers with pain; long‐term follow‐up, > 6 months)	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
4.4.1 During immobilisation after surgical management	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Comparison 5. Other rehabilitation interventions.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Pain (using different measurement instruments; at end of treatment)	2		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
5.1.1 During initial fracture management period after surgical management	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
5.1.2 During initial fracture management period after non‐surgical treatment	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
5.2 Adverse events (interventions during initial fracture management with surgical treatment)	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
5.2.1 Fracture displacement	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
5.2.2 Skin irritation/ulceration	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Ahl 1986.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 44 Source population: single centre, Department of Orthopaedics, Danderyd Hospital, Sweden Inclusion criteria: closed dislocated fibular fracture with ruptured anterior tibiofibular ligament; surgical fixation required; > 18 years of age; able to cooperate (e.g. alcohol and drug addiction, senility excluded); no concomitant injuries, except fracture of the posterior tibial margin Exclusion criteria: those who preferred not to participate; < 18 years of age; unable to cooperate (e.g. alcoholics, drug addicts, people with senility); with other concomitant injuries interfering with the postoperative programme Overall baseline characteristics Age, mean: 44 years Sex (m/f): 22/24 Surgery: all had surgery Baseline characteristics for intervention group Fracture severity: Weber B/C: 18/6 Other data not reported by group Baseline characteristics for control group Fracture severity: Weber B/C: 18/4 Other data not reported by group
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: below‐knee plaster cast for 7 weeks Intervention group Number randomised: 24; losses: 2 (at 6 months, lost to follow‐up); number analysed at each time point: 3 months (n = 18); 6 months (n = 22) Treatment management: weight‐bearing from first postoperative day Control group Number randomised: 22; losses: 3 (at 6 months, reasons not reported); number analysed at each time point: 3 months (n = 17); 6 months (n = 19) Treatment management: weight‐bearing from fourth postoperative week
Outcomes	Timing of assessments: 3, 6, and 18 months after randomisation End of treatment: nil End of follow‐up: 6‐month data used as more complete than 18‐month data Included in this review: adverse events Other outcomes measured in the study: activity limitation questionnaire (measured but not reported), calf circumference, radiographs, roentgen stereo‐photogrammetric analysis, bone mineral content, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: March 1984 to March 1985 Funding sources: not reported Declarations of interest: not reported Note: we received additional information from the study author, T Ahl (regarding the relationship between this study and other included references by Ahl and colleagues)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "each patient was randomly allocated." Comment: method of randomisation not described. No baseline data reported to allow assessment of study group comparability owing to effective randomisation
Allocation concealment (selection bias)	Unclear risk	Quote: "by instruction in a sealed envelope." Comment: insufficient information provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "Forty‐six patients were included in this study. Two patients were lost to follow‐up." Comment: lost participants were in the early weight‐bearing group. Because losses were few, we did not expect that this attrition would influence the data.
Selective reporting (reporting bias)	High risk	Study protocol or clinical trials registration was not reported and we could not use these documents to effectively assess risk of selective reporting bias. We noted from the Methods section of the report that activity limitation questionnaire was measured, but not reported.
Other bias	Low risk	We identified no other sources of bias.

Ahl 1987.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 53 Source population: single centre, Department of Orthopaedics, Danderyd Hospital, Sweden Inclusion criteria: closed dislocated bimalleolar or trimalleolar fracture with ruptured anterior tibiofibular ligament; surgical fixation required, adults, able to cooperate (e.g. alcohol and drug addiction, senility excluded); no other injuries which could interfere with rehabilitation, except fracture of the posterior tibial margin Exclusion criteria: children; people with open fractures or injuries interfering with the rehabilitation program; those unable to cooperate (e.g. alcoholics and drug addicts) Overall baseline characteristics Age, mean: 57 Surgery: all had surgery Baseline characteristics for intervention group Sex (m/f): 10/15 Fracture severity: Weber B/C: 12/13 Baseline characteristics for control group Sex (m/f): 6/22 Fracture severity: Weber B/C: 15/13
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: below‐knee plaster cast for 7 weeks Intervention group Number randomised: 25; losses: 1 (reason not reported); number analysed at each time point: 3 months (n = 24); 6 months (n = 24) Treatment management: weight‐bearing from first postoperative day Control group Number randomised: 28; losses: 3 (1 withdrew, 1 died, reason for other participant is not clearly reported); number analysed at each time point: 3 months (n = 23); 6 months (n = 25) Treatment management: weight‐bearing from fourth postoperative week
Outcomes	Timing of assessments: 3, 6, and 18 months after randomisation End of treatment: nil End of follow‐up: 6‐month data used as more complete than 18‐month data Included in this review: adverse events Other outcomes measured in the study: activity limitation questionnaire (measured but not reported), calf circumference, radiographs, roentgen stereo‐photogrammetric analysis, bone mineral content, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: March 1984 to April 1986 Funding sources: not reported Declarations of interest: not reported Note: we received additional information from the study author, T Ahl (regarding the relationship between this study and other included references by Ahl and colleagues)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "all the fractures were randomly allocated." Comment: method of randomisation not described. No baseline data reported to allow assessment of study group comparability owing to effective randomisation
Allocation concealment (selection bias)	Unclear risk	Method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "of 53 patients included in this study, 2 patients were lost to follow‐up." Comment: both lost participants were from the late weight‐bearing group. Reasons for losses were death in one participant and withdrawal from the study by the other participant. We did not expect this small number of losses to impact the outcome data.
Selective reporting (reporting bias)	High risk	Study protocol or clinical trials registration was not reported, and we could not use these documents to effectively assess risk of selective reporting bias. We noted from the Methods section of the report that activity limitation questionnaire was measured, but not reported.
Other bias	Low risk	We identified no other sources of bias.

Ahl 1988.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 51 Source population: not reported Inclusion criteria: closed displaced lateral malleolar fracture with ruptured anterior tibiofibular ligament; surgical fixation required; adults; able to cooperate (e.g. alcohol and drug addiction, senility excluded); no other injuries which could interfere with rehabilitation, except fracture of the posterior tibial margin Exclusion criteria: children; people with open fractures; those with other injuries interfering with the rehabilitation program; and those unable to cooperate (e.g. alcoholics, drug addicts, and people with senility) Baseline characteristics for intervention group Age, mean (range): 47 (18 to 74) years Sex (m/f): 13/13 Surgery: all had surgery Fracture severity: Weber B/C: 19/7 Baseline characteristics for control group Age, mean (range): 39 (18 to 74) years Sex (m/f): 12/13 Surgery: all had surgery Fracture severity: Weber B/C: 15/10
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: active unloaded dorsiflexion and plantarflexion exercises 5 times/day from the second to seventh postoperative weeks Intervention group Number randomised: 26; losses: 5 (reasons not reported); number analysed at each time point: 3 months (n = 22); 6 months (n = 21) Treatment management: below‐knee plaster cast in the first postoperative week, then orthosis for 7 weeks, weight‐bearing after first postoperative week Control group Number randomised: 25; losses: 3 (reasons not reported); number analysed at each time point: 3 months (n = 21); 6 months (n = 22) Treatment management: below‐knee plaster cast in the first postoperative week, then dorsal splint for 7 weeks, non‐weight‐bearing
Outcomes	Timing of assessments: 3 and 6 months after randomisation End of treatment: nil End of follow‐up: 6‐month data used Included in this review: adverse events Other outcomes measured in the study: activity limitation questionnaire (measured but not reported), calf circumference, radiographs, roentgen stereo‐photogrammetric analysis, work capacity, capacity for sports and leisure, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from the study author, T Ahl (regarding the relationship between this study and other included references by Ahl and colleagues)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "the patients were randomly allocated." Comment: method of randomisation not described. No baseline data reported to allow assessment of study group comparability owing to effective randomisation
Allocation concealment (selection bias)	Unclear risk	Method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information provided
Incomplete outcome data (attrition bias) All outcomes	High risk	Overall loss at 6 months of 16%. Losses are reasonably balanced between groups, but reasons for losses were not reported.
Selective reporting (reporting bias)	High risk	Study protocol or clinical trials registration was not reported, and we could not use these documents to effectively assess risk of selective reporting bias. We noted from the Methods section of the report that activity limitation questionnaire was measured, but not reported.
Other bias	Low risk	We identified no other sources of bias.

Ahl 1993.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparisons: early weight‐bearing versus delayed weight‐bearing; removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 43 Source population: not reported Inclusion criteria: closed displaced bimalleolar and trimalleolar fracture; surgical fixation required; adults; able to cooperate; no other injuries which could interfere with rehabilitation, except fracture of the posterior tibial margin Exclusion criteria: children; people with open fractures or other injuries interfering with the rehabilitation program; those unable to cooperate (e.g. alcoholics and drug addicts) Baseline characteristics for intervention group Age, mean (range): 55 (20 to 76) years Sex (m/f): 4/17 Surgery: all had surgery Fracture severity: Lauge‐Hansen supination‐eversion IV/pronation‐abduction III/pronation‐eversion IV: 15/3/3 Baseline characteristics for control group Age, mean (range): 55 (22 to 77) years Sex (m/f): 3/16 Surgery: all had surgery Fracture severity: Lauge‐Hansen supination‐eversion IV/pronation‐abduction III/pronation‐eversion IV: 13/3/3
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: active unloaded dorsiflexion and plantarflexion exercises 5 times/day from the second to seventh postoperative weeks Intervention group Number analysed: 21; 3 participants were excluded after randomisation; 1 participant did not want to participate in the study and the other 2 were mistakenly given a walking cast postoperatively. However, these losses were not reported by group, and we could not determine with certainty how many were randomised to each group. Treatment management: plaster cast in the first postoperative week, then orthosis for 7 weeks, weight‐bearing after first postoperative week Control group Number analysed: 19; 3 participants were excluded after randomisation; 1 participant did not want to participate in the study and the other 2 were mistakenly given a walking cast postoperatively. However, these losses were not reported by group, and we could not determine with certainty how many were randomised to each group. Treatment management: plaster cast in the first postoperative week, then dorsal splint for 7 weeks, non weight‐bearing
Outcomes	Timing of assessments: 3, 6, and 18 months after randomisation End of treatment: nil End of follow‐up: 18‐month data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: work capacity, capacity for sports and leisure, calf circumference, radiographs, roentgen stereo‐photogrammetric analysis, ankle dorsiflexion and plantarflexion range of motion Note: we could not use outcome data for superficial infection and local wound necrosis because these data were not reported by intervention group.
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from the study author, T Ahl (regarding the relationship between this study and other included references by Ahl and colleagues).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "the patients were randomly allocated." Comment: method of randomisation not described. No baseline data reported to allow assessment of study group comparability owing to effective randomisation
Allocation concealment (selection bias)	Unclear risk	Method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Insufficient information on whether assessor blinding was implemented. We assumed therefore that no attempts were made to blind outcome assessors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Some participants were excluded after randomisation, and it was not clear to which group these participants belonged. Because loss was < 10%, we did not expect this to impact on the results.
Selective reporting (reporting bias)	Unclear risk	Study protocol or clinical trials registration was not reported, and we could not use these documents to effectively assess risk of selective reporting bias.
Other bias	Low risk	We identified no other sources of bias.

Bretherton 2024.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 561 Source population: 23 NHS hospitals, UK Inclusion criteria: adults (≥ 18 years of age), undergone operative fixation for an unstable ankle fracture, surgery performed within 14 days of injury, participant may benefit from early weight‐bearing in the opinion of the treating surgeon, able and willing to give consent Exclusion criteria: lack of protective sensation (e.g. peripheral neuropathy), unable to adhere to trial procedures; bilateral operatively‐treated ankle fractures; already in a trial for ankle fracture; received a hindfoot nail to treat index fracture Baseline characteristics for intervention group Age, mean (SD): 48.5 (± 16.4) years Sex (m/f): 101/179 Surgery: all had surgery Fracture severity: not reported Baseline characteristics for control group Age, mean (SD): 48.1 (± 16.6) years Sex (m/f): 101/180 Surgery: all had surgery Fracture severity: not reported
Interventions	Timing of randomisation: average of 2 weeks after surgery Treatment management for both groups: all participants non‐weight‐bearing in first 2 weeks in immediate postoperative period. Type of ankle support and how much unloaded ankle exercise was decided by the treating clinician. Intervention group Number randomised: 280 Treatment management: weight‐bearing on the injured leg, starting at 2 weeks postoperatively, and continuing for 4 weeks Control group Number randomised: 281 Treatment management: asked to remain non‐weight‐bearing for an additional 4 weeks
Outcomes	Timing of assessments: baseline, 6 weeks, 4 months, 12 months End of treatment: 6 weeks postoperatively End of follow‐up: 12 months Included in this review: activity limitation (OMAS), HRQoL (EQ‐5D‐5L), adverse events Other outcomes measured in the study: none
Notes	Study dates: 28 December 2019 to 28 October 2021 (recruitment period) Funding sources: NIHR Research for Patient Benefit Programme. Funding sources for individual study authors included NIHR Clinical Scientist Award, Royal College of Surgeons of England/Rosetrees Research Fellowship Declarations of interest: study authors declare they have no conflicts of interest
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote (from protocol): "randomly allocated to one of two intervention groups (1:1) using a secure (encrypted) remote computer randomisation service administered by the Oxford Clinical Trials Research Unit"
Allocation concealment (selection bias)	Low risk	Quote (from protocol): "randomly allocated to one of two intervention groups (1:1) using a secure (encrypted) remote computer randomisation service administered by the Oxford Clinical Trials Research Unit"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to treatments
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants aware of group allocation
Incomplete outcome data (attrition bias) All outcomes	High risk	Although reasonably balanced between groups, losses at end of follow‐up were approximately 15%.
Selective reporting (reporting bias)	Low risk	Prospective clinical trials registration (ISRCTN12883981). Reported outcomes are consistent with those listed in the clinical trials register
Other bias	Low risk	We identified no other sources of bias.

Brink 1996.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions
Participants	Total number of randomised participants: 66 Source population: multicentre, emergency departments at Aarhus University Hospital and Silkeborg Hospital, Denmark Inclusion criteria: acute (< 24 hours), stable, isolated supination‐eversion II fracture of the lateral malleolus; > 18 years of age; no tenderness on medial side of ankle Exclusion criteria: if tenderness was present on the medial side of the ankle, suggesting a possible injury to the deltoid ligament, and thus a stage IV fracture Baseline characteristics for intervention group Age, mean (range): 45 (18 to 82) years Sex (m/f): 4/29 Surgery: none had surgery Fracture severity: all had stable lateral malleolar fractures Baseline characteristics for control group Age, mean (range): 45 (18 to 84) years Sex (m/f): 15/18 Surgery: none had surgery Fracture severity: all had stable lateral malleolar fractures
Interventions	Timing of randomisation: within 24 hours of presentation Treatment management for both groups: full weight‐bearing within limits of pain and using crutches if necessary, rest and elevation for 3 to 5 days for swelling, brace could be removed for bathing, when resting in chair, and at night after 4 weeks Intervention group Number randomised: 33; losses: 2 (minor skin reaction (Aircast not worn over a sock) = 1; lost to follow‐up = 1); number analysed at each time point: 1‐week (n = 31); 4‐week (n = 30); 6‐week (n = 18) (only cases without clinically united fracture at 4 weeks were examined after 6 weeks); 12‐week (n = 29) Treatment management: Aircast air‐stirrup for an average of 39 days Control group Number randomised: 33; losses: 1 (emigrated); number analysed at each time point: 1‐week (n = 32); 4‐week (n = 32); 6‐week (n = 13) (only cases without clinically united fracture at 4 weeks were examined after 6 weeks); 12‐week (n = 28) Treatment management: DonJoy orthosis with hinges unlocked for an average of 35 days
Outcomes	Timing of assessments: 1, 4, and 12 weeks after randomisation (and 6 weeks if not clinically united at 4 weeks) End of treatment: 4‐week data used End of follow‐up: 12‐week data used Included in this review: activity limitation questionnaire, participant satisfaction, pain, adverse events Other outcomes measured in the study: radiographs, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: additional information received from O Brink (method of allocation, assessor blinding, adverse events)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomized by random number table." Comment: we noted that there was a higher proportion of females
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by consecutively numbered, opaque envelopes.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Only participants without fracture union were examined at 6 weeks. We could not be certain of the impact on data for this time point.
Selective reporting (reporting bias)	Unclear risk	Study protocol or clinical trials registration was not reported, and we could not use these documents to effectively assess risk of selective reporting bias.
Other bias	Low risk	We identified no other sources of bias.

Christie 1990.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions (electrotherapy)
Participants	Total number of randomised participants: 24 Source population: single centre, Royal Adelaide Hospital, Australia Inclusion criteria: acute closed ankle fracture; surgical fixation required; 15 to 61 years of age; none of the following: peripheral vascular disease, history of bleeding disorders, ankle swelling on uninvolved side, medication likely to affect swelling, previous treatment or contraindication with interferential Baseline characteristics for intervention group Age, mean (range): 33 (19 to 52) years Sex (m/f): 8/4 Surgery: all had surgery Fracture severity: uni‐/bi‐/trimalleolar: 6/5/1 Baseline characteristics for control group Age, mean (range): 37 (15 to 61) years Sex (m/f): 6/6 Surgery: all had surgery Fracture severity: uni‐/bi‐/trimalleolar: 4/3/5
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: electrodes (4) placed on the medial and lateral sides of the ankle, and the medial and lateral sides of the shank. Plaster backslab for 2 to 4 days after surgery, elevation, twice daily supervised dorsiflexion and plantarflexion exercises, then below‐knee cast for an unspecified period Intervention group Number randomised: 12; losses: 0; number analysed at each time point: 12 Treatment management: interferential ‐ frequency 0 to 100 Hz, intensity 20 mA, 20 minutes/day from the physiotherapist for 2 to 4 days until below knee cast application Control group Number randomised: 12; losses: 0; number analysed at each time point: 12 Treatment management: sham interferential ‐ electrodes applied but no current, 20 minutes/day from the physiotherapist for 2 to 4 days until below‐knee cast application
Outcomes	Timing of assessments: 2 to 4 days after randomisation End of treatment: 2 to 4‐day data used End of follow‐up: nil Included in this review: nil Other outcomes measured in the study: swelling
Notes	Study dates: not reported Funding sources: loan of inferential unit from FSE (definition not supplied by study authors) Declarations of interest: not reported Notes: as we included no data from this study in the review, we did not complete a risk of bias assessment for this study.

Davies 1991.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 41 Source population: single centre, East Glamorgan General Hospital, UK Inclusion criteria: fracture of the medial, lateral or posterior malleolus involving the articular surface; surgical fixation required; < 50 years of age; no co‐existing fractures of either leg; could have lateral, deltoid, or inferior tibiofibular ligament injuries Baseline characteristics for intervention group Age, mean: 31 years Sex (m/f): 16/5 Surgery: all had surgery Fracture severity: uni‐/bi‐/trimalleolar: 7/11/3 Baseline characteristics for control group Age, mean: 30 years Sex (m/f): 10/10 Surgery: internal fixation Fracture severity: uni‐/bi‐/trimalleolar: 11/6/3
Interventions	Timing of randomisation: immediately after surgery; "all fractures accurately reduced, firmly and rigidly internally fixed, using AO/ASIF techniques" Treatment management for both groups: injured leg elevation for 3 to 4 days post‐operatively, non‐weight‐bearing with crutches until radiological union Intervention group Number randomised: 21; losses: 6 (reasons not reported) Treatment management: below‐knee backslab for 24 hours, then retained for night use only for an unspecified period, continuous passive motion 12 to 15 hours/day for an average of 6 days (range of motion set to that achieved actively by the participant and then increased daily within pain limits), regular active hip, knee, toe movements, physiotherapy after hospital discharge for an average of 6 weeks Control group Number randomised: 20; losses: 5 (reasons not reported) Treatment management: below‐knee plaster cast for an average of 57 days, physiotherapy (if referred) after removal of plaster for an average of 3 weeks
Outcomes	Timing of assessments: before hospital discharge (timing not reported), 6 and 12 months after randomisation End of treatment: nil End of follow‐up: 12‐month data used Included in this review: pain, adverse events Other outcomes measured in the study: ability to resume former hobbies, return to work or full activity, subtalar range of motion, calf circumference, radiographs, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: 1 January 1987 to 31 December 1988 Funding sources: not reported Declarations of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "subject selection depended on which consultant was responsible for the care of the patient." Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	The assessor was aware of the study‐group allocation of the participants.
Incomplete outcome data (attrition bias) All outcomes	High risk	Whilst losses were balanced between groups, the reasons for loss were not reported and the overall number of losses was high.
Selective reporting (reporting bias)	Unclear risk	Study protocol or clinical trials registration was not reported, and we could not use these documents to effectively assess risk of selective reporting bias.
Other bias	Low risk	We identified no other sources of bias.

Dehghan 2016.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparisons: early weight‐bearing versus delayed weight‐bearing; removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 110 Source population: 2 x level 1 trauma centres (Sunnybrook Health Sciences Centre and St. Michael’s Hospital) in Toronto, Canada Inclusion criteria: acute unstable unilateral ankle fracture that was considered by the attending staff surgeon to require surgical fixation, including: isolated lateral malleolus fracture with talar shift; vertical shear medial malleolus fractures; bimalleolar fractures; trimalleolar fractures that did not require posterior fragment fixation; closed, grade I, or grade II open fractures Exclusion criteria: participant‐related exclusion factors included: skeletal immaturity, previous ipsilateral ankle surgery, nonambulatory status before injury, inability to comply with postoperative protocol (i.e. advanced dementia), medical comorbidity precluding operative intervention, and people with workers' compensation; people with polytrauma. Fracture‐related exclusion factors included: surgical fixation > 14 days from time of injury; grade III open fractures; tibial plafond fractures; syndesmotic injuries/fixation; posterior malleolar fractures requiring fixation. No exclusion criteria with regard to participant age, presence of osteoporosis/osteopenia, or BMI Baseline characteristics for intervention group Age, mean (SD): 41.7 (± 15.1) years Sex (m/f): 32/24 Surgery: all had surgery Fracture severity: unimalleolar: 26; bimalleolar: 25; trimalleolar: 5. OTA: 44A1: 1; 44A2: 2; 44B1: 17; 44B2: 20; 44B3: 11; 44C1: 4; 44C2: 1 Baseline characteristics for control group Age, mean (SD): 42.1 (± 15.4) years Sex (m/f): 27/27 Surgery: all had surgery Fracture severity: unimalleolar: 18; bimalleolar: 27; trimalleolar: 9. OTA: 44A1: 1; 44A2: 0; 44B1: 14; 44B2: 27; 44B3: 9; 44C1: 2; 44C2: 1
Interventions	Timing of randomisation: at the first postoperative visit (2 weeks postoperatively) Treatment management for both groups: ORIF of unstable ankle fractures with rigid osteosynthesis techniques using standard surgical procedures. Postoperatively, all participants were placed in a below‐knee posterior plaster slab and instructed not to bear weight. Posterior slab removed at 2‐week visit before randomisation. For boot orthosis, all participants were given a night splint to use in bed while not wearing the orthosis. Intervention group Number randomised: 56; losses at 6 weeks: 3; at 3 months: 7; at 6 months: 10; at 12 months: 6 (reasons for losses not explained); number analysed at 6 weeks: 53; at 3 months: 49; at 6 months: 56; at 12 months: 50 Treatment management: participants in the early WB group had their ankle placed in a boot orthosis at the 2‐week postoperative visit and instructed to start full weight‐bearing as tolerated. They were also instructed to remove the orthosis 4 times a day and perform ankle dorsiflexion, plantarflexion, inversion, and eversion exercises. Participants were seen by a physiotherapist, who gave them instructions regarding weight‐bearing and range of motion exercises. At the 6‐week postoperative visit, participants were instructed to wean from the boot orthosis over the next 2 to 4 weeks. Control group Number randomised: 54; losses at 3 months: 3; at 6 months: 8; at 12 months: 2 (reasons for losses not explained; number analysed at 6 weeks: 54; at 3 months: 51; at 6 months: 46; at 12 months: 52 Treatment management: participants in the late WB group were placed in a below‐knee fibreglass cast at 2 weeks postoperatively and instructed to remain non‐weight‐bearing for a further 4 weeks (6 weeks total). At the 6‐week postoperative visit, the cast was removed, and participants were allowed to perform ankle range of motion exercises and begin full weight‐bearing with the use of a boot orthosis. Participants were seen by a physiotherapist, who discussed instructions regarding weight‐bearing and range of motion exercises. They were also instructed to wean from the orthosis over the next 2 to 4 weeks.
Outcomes	Timing of assessment: clinical follow‐up at 2 weeks, 6 weeks, 3 months, 6 months, and 12 months End of treatment: 8 to 10 weeks End of follow‐up: 12 months Included in this review: activity limitation, QoL, adverse events Other outcomes measured in the study: time to return to work, ankle dorsiflexion range of motion, plantarflexion range of motion
Notes	Study dates: 2010 to 2014 Funding sources: research grants from several sources: the COA; the OTA's Resident Research Grant; the COTS Young Investigator Award; and a PSI Resident Research Grant Declarations of interest: quote: "There are no conflict of interests for any of the authors with regard to this study"
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Low risk	Quote: "use of concealed sequentially numbered, opaque, sealed envelopes" and "Allocation was concealed from patients and investigators until the randomization was completed"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants or personnel to treatments
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants aware of treatment allocation
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Study authors do not report whether outcome assessors were blinded.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Losses are reported, and reasonably balanced between groups, but reasons for losses are not reported, and it is not clear why some participants were re‐included at 12 months.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

DiStasio 1994.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 61 Source population: active‐duty military personnel Inclusion criteria: isolated closed ankle fractures, surgical fixation required Baseline characteristics Surgery: all had surgery No other baseline characteristics reported
Interventions	Timing of randomisation: immediately after surgery; "randomised into two groups once stable fixation was achieved" Treatment management for both groups: partial weight‐bearing or full weight‐bearing in air‐stirrup after 6 weeks, syndesmosis screws (if used) removed at 6 to 8 weeks postoperatively Intervention group Number randomised: 30; losses: 1; number analysed at each time point: 29 Treatment management: DonJoy orthosis for 6 weeks, physiotherapy from the first postoperative week Control group Number randomised: 31; losses: 3; number analysed at each time point: 28 Treatment management: below‐knee cast for 6 weeks, physiotherapy after cast removal
Outcomes	Timing of assessments: 6 weeks, termination of physiotherapy (timing not reported), 3 and 6 months after randomisation End of treatment: nil End of follow‐up: 6‐month data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: activity limitation test (only measured in 21/61 participants), ankle range of motion (only measured in 21/61 participants), strength (only measured in 21/61 participants), swelling (only measured in 21/61 participants), time to return to full duty Note: we could not use data for some types of adverse events because they were not reported by intervention groups (re‐operation for hardware removal, injury to superficial peroneal nerve, syndesmosis)
Notes	Study dates: 1990 to 1992 Funding sources: Bureau of Medicine and Surgery Clinical Investigation Program Declarations of interest: not reported Note: additional information received from A DiStasio (method of allocation)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Additional information obtained from author: randomisation was performed by order of presentation. Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few losses which are balanced between groups
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Dogra 1999.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 52 Source population: single centre, Hartlepool District General Hospital, UK Inclusion criteria: bimalleolar fracture; surgical fixation required; stable fixation, 16 to 65 years of age Exclusion criteria: people with previous ankle disease or other concomitant skeletal injury Overall baseline characteristics Age, mean (range): 42.7 (18 to 65) years Sex (m/f): 27/25 Surgery: all had surgery Fracture severity: all had bimalleolar fracture Baseline characteristics not reported by group
Interventions	Timing of randomisation: within 24 hours after surgery Treatment management for both groups: below‐knee walking cast for 4 weeks after 2 weeks in backslab, graduated weight‐bearing, advice on ankle remobilisation after 6 weeks postoperatively Intervention group Number randomised: 26; losses: 0; number analysed at each time point: 26 Treatment management: active ankle dorsiflexion and plantarflexion movements starting 24 hours postoperatively, 4 times 10 minutes/day for 2 weeks. Ankle resting in below‐knee backslab at all other times Control group Number randomised: 26; losses: 0; number analysed at each time point: 26 Treatment management: ankle continuously in below‐knee backslab for 2 weeks
Outcomes	Timing of assessments: 12 weeks after randomisation End of treatment: nil End of follow‐up: 12‐week data Included in this review: activity limitation questionnaire, pain, ankle range of motion, adverse events Other outcomes measured in the study: none Note: we could not use data for one case of superficial infection because these data were not reported by intervention group
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: additional information received from A Rangan (regarding method of randomisation, source population)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: randomisation by computer random number function.
Allocation concealment (selection bias)	Unclear risk	Quote: "sealed opaque randomization envelopes." Comment: insufficient information provided about number of the envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "all patients were assessed by a single observer who was 'blind' to the post operative regime." Comment: assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "All patients attended for follow‐up until completion of the study."
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Egol 2000.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 60 Source population: single centre, tertiary care centre, USA Inclusion criteria: closed, isolated, unstable ankle fracture, surgical fixation required, complete skeletal growth, non‐neuropathic joint Exclusion criteria: not reported Baseline characteristics for intervention group Age, mean (SD): 39.5 (± 17.2) years Sex (m/f): 13/14 Surgery: all had surgery Fracture severity: Lauge‐Hansen supination‐external rotation/supination‐adduction/pronation‐external rotation/pronation‐abduction: 24/1/1/1 Baseline characteristics for control group Age, mean (SD): 45.6 (± 17.5) years Sex (m/f): 10/18 Surgery: all had surgery Fracture severity: Lauge‐Hansen supination‐external rotation/supination‐adduction/pronation‐external rotation/pronation‐abduction: 22/1/1/4
Interventions	Timing of randomisation: after surgery Treatment management for both groups: weight‐bearing started 6 weeks postoperatively (8 weeks for those with syndesmosis screw) Intervention group Number analysed: 27. Overall losses, as below. Number allocated to each was not reported Treatment management: plaster splint 2 to 3 days postoperatively, then Aircast removable functional brace for 6 weeks, active and passive ankle and subtalar exercises 3 times/day from 3 to 4 days postoperatively Control group Number analysed: 28. Overall losses, as below. Number allocated to each was not reported Treatment management: plaster splint 2 to 3 days postoperatively, then fibreglass short‐leg cast for 6 weeks, physiotherapy after 6 weeks postoperatively Overall losses: not reported by group. Moved abroad = 1; developed multiple sclerosis with severe progression of disease = 1; lost to follow‐up = 3
Outcomes	Timing of assessments: 6, 12, 26 and 52 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 52‐week data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: quality of life (incomplete raw data and no between‐group differences given), radiographs, return to work
Notes	Study dates: 30 June 1995 to 30 June 1997 Funding sources: not reported Declarations of interest: not reported Note: we received additional information received from study author, K Egol (regarding method of allocation, source population)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: randomisation by random numbers table
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by sealed, opaque and consecutively numbered envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Quote: "The operating surgeon performed the follow‐up rating score"
Incomplete outcome data (attrition bias) All outcomes	Low risk	It is not clear to which group lost participants belonged. Reasons for loss, however, are clearly reported and the overall number of losses is < 10%.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Finsen 1989.

Study characteristics
Methods	RCT, parallel design, 3‐armed trial Review comparisons: early weight‐bearing versus delayed weight‐bearing; removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 56 Source population: Trondheim University Hospital, Norway Inclusion criteria: displaced ankle fracture (must include fracture of the lateral malleolus); surgical fixation required; operation within 1 week of fracture; stable osteosynthesis allowing active ankle movement but not unprotected weight‐bearing; no concomitant injury; could attend follow‐up visits Exclusion criteria: not specified. However, we noted that participants were excluded for the following reasons: discharged before measurement of bone mineral, refusal to participate, did not attend follow‐up appointments, could not manage weight‐bearing routine (owing to epilepsy) Baseline characteristics for weight‐bearing group Age, mean (SE): 43 (± 5.2) years Sex (m/f): 5/14 Surgery: all had surgery Fracture severity: Weber A/B/C, uni‐/bi‐/trimalleolar:0/16/3, 8/4/7 Baseline characteristics for exercise group Age, mean (SE): 43 (± 3.2) years Sex (m/f): 8/10 Surgery: all had surgery Fracture severity: Weber A/B/C, uni‐/bi‐/trimalleolar: 0/15/3, 8/4/6 Baseline characteristics for control group Age, mean (SE): 40 (± 3.3) years Sex (m/f): 8/11 Surgery: all had surgery Fracture severity: Weber A/B/C, uni‐/bi‐/trimalleolar: 0/16/3, 8/2/9
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: full weight‐bearing after 6 weeks postoperatively Weight‐bearing group Number randomised: 19; losses: 5 (reasons not explained by group, see below); number analysed for activity limitation: 14; number analysed for adverse events: 19 Treatment management: below‐knee plaster cast with rubber walker for 6 weeks, weight‐bearing as tolerated Exercise group Number randomised: 18; losses: 5 (reasons not explained by group, see below); number analysed for activity limitation: 13; number analysed for adverse events: 18 Treatment management: below‐knee plaster cast for 3 days, ankle and subtalar range of motion exercises daily Control group Number randomised: 19; losses: 4 (reasons not explained by group, see below); number analysed for activity limitation: 15; number analysed for adverse events: 19 Treatment management: below‐knee lightweight plaster cast for 6 weeks, non weight‐bearing Overall losses: 14 (4 lost to follow‐up at 8 months, a further 10 lost to follow‐up at 1 year)
Outcomes	Timing of assessments: 9, 18, 36, 52 and 104 weeks after randomisation End of treatment: nil End of follow‐up: 52‐week data used as more complete than 104‐week data Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: pain (categorical scale), return to work, subtalar range of motion, radiographs, bone mineral content Note: some adverse event data were not described by intervention group and we could not include these data in the review (deep infection and superficial infection), ankle dorsiflexion and plantarflexion range of motion (categorical scale), swelling (categorical scale)
Notes	Study dates: November 1983 and June 1985 Funding sources: not reported Declarations of interest: not reported Notes: We have used data from this study for the following comparison groups in the review: early versus delayed weight‐bearing: cast with weight‐bearing versus cast with no weight‐bearing removable support versus fixed support: exercise versus cast with no weight‐bearing We reported data for other possible ways to compare these study groups in Appendix 2.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "patients were randomly assigned." Comment: method of randomisation not described
Allocation concealment (selection bias)	Unclear risk	Method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Quote: "'each patient was examined by one of us." Comment: insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Large number of losses which are balanced between groups, although reasons for losses are not reported. Because loss at 12 months is 25%, it is possible that this could impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Fitzgerald 1994.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions
Participants	Total number of randomised participants: 20 Source population: single centre, emergency department of a general hospital, Ireland Inclusion criteria: ankle fracture, surgical fixation required, no other injuries or pathologies, adults Baseline characteristics for intervention group Age, mean (range): 36.5 (25 to 60) years Sex (m/f): 6/4 Surgery: all had surgery Fracture severity: bi‐/trimalleolar: 8/2 Baseline characteristics for control group Age, mean (range): 34.8 (21 to 64) years Sex (m/f): 7/3 Surgery: all had surgery Fracture severity: bi‐/trimalleolar: 7/3
Interventions	Timing of randomisation: after surgery Intervention group Number randomised: 10; losses: 3 (reasons not reported); number analysed: 7 Treatment management: compression stocking (18 mmHg at the ankle, 8 mmHg below the knee) and below‐knee plaster cast for 6 weeks Control group Number randomised: 10; losses: 0; number analysed: 10 Treatment management: below‐knee plaster cast for 6 weeks
Outcomes	Timing of assessments: baseline, 6, 12 and 18 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 18‐week data used Included in this review: adverse events (zero events) Other outcomes measured in the study: air plethysmography, swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Notes: We received additional information from study author, P Fitzgerald (regarding methods of randomisation and allocation, assessor blinding, source population, adverse events, dropouts, additional data) We did not include a third group (early weight‐bearing) from the study in this review as allocation to this group was a clinical decision and not by randomisation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: randomisation by shuffling envelopes
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by sealed, opaque and consecutively numbered envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Small study with loss in one group of 33%. No losses in the comparison group
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study is published only as an abstract, with information supplied by the study author. Not published in a peer‐reviewed journal

Franke 2008.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 27 Source population: single centre, Berufsgenossenschaftlichte Unfallklinik (Liability Insurance Traumatology Clinic), Ludwigshafen, Germany Inclusion criteria: closed, isolated, surgically‐treated ankle fractures; 18 to 65 years of age; simple or bimalleolar fractures Exclusion criteria: open fractures, multi‐fragmentary fractures and fractures requiring a positioning screw; and people with any other disorders involving restriction of mobility, limited ambulation on forearm crutches, or situations affecting the healing process (e.g. joint disease, neurological disorder, amputation) Baseline characteristics for intervention group Age, median (range): 44.3 (20.3 to 59.4) years Sex (m/f): 7/7 Surgery: all had surgery Fracture severity: simple Weber B/bimalleolar Weber B: 13/1 Baseline characteristics for control group Age, median (range): 40.8 (25 to 64.1) years Sex (m/f): 8/5 Surgery: all had surgery Fracture severity: simple Weber B/bimalleolar Weber B: 13/0
Interventions	Timing of randomisation: not reported Treatment management for both groups: full weight‐bearing allowed from day 15 Intervention group Number randomised: 14; losses: 1 (personal reasons); number analysed at each time point: 13 Treatment management: dynamic vacuum orthosis for 6 weeks. Partial weight‐bearing and ankle exercises (10 degrees dorsi‐ and plantarflexion) from the second postoperative day. Thrombosis prophylaxis during the period of limited mobility Control group Number randomised: 13; losses: 2 (personal reasons); number analysed at each time point: 11 Treatment management: circular cast (with a window cut in it on the second post‐operative day to permit ankle plantarflexion). After replacement by a supportive bandage, partial weight‐bearing of 20 kg was prescribed from the time of wound healing to the 14th postoperative day. Participants attended physiotherapy 3 times a week for 4 weeks following their 6‐week immobilisation. Thrombosis prophylaxis for the duration of immobilisation
Outcomes	Timing of assessments: 6 and 10 weeks End of treatment: 10‐week data used End of follow‐up: nil Included in this review: activity limitation questionnaire, quality of life, participant satisfaction, adverse events Other outcomes measured in the study: ankle dorsiflexion range of motion (categorical data), ankle plantarflexion range of motion (categorical data), time to return to work, economic parameters
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author, J Franke (regarding method of allocation, assessor blinding, intention to treat)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "computer‐generated block‐randomisation sequence."
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: opening a closed, opaque envelope randomised off‐site Comment: randomisation occurred off‐site
Blinding of participants and personnel (performance bias) All outcomes	High risk	Additional information obtained from author: neither participants, therapists, nor assessors were blinded
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Additional information obtained from author: neither participants, therapists, nor assessors were blinded
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Additional information obtained from author: neither participants, therapists, nor assessors were blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	Additional information obtained from author: the analyses were conducted following the intention‐to‐treat principle. Comment: 3 dropouts, which are reasonably balanced between groups. We did not expect these losses to impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Ginandes 1999.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions
Participants	Total number of randomised participants: 12 Source population: multicentre, Orthopaedic Emergency Department of Massachusetts General Hospital, USA, and McLean Hospital, USA Inclusion criteria: acute non‐displaced lateral malleolar fracture, 19 to 49 years of age, no previous fracture or arthritic condition involving the injured ankle, no coexisting illness, condition, medication or substance use that would influence bone healing or contraindicate hypnosis, fluency in English, available for the treatment period Overall baseline characteristics Sex (m/f): 3/9 No other baseline characteristics reported by group
Interventions	Timing of randomisation: within 48 hours of presentation Treatment management for both groups: cast immobilisation for 6 weeks, orthopaedic follow‐up clinics at 1, 3, 6, 9, 12 weeks after injury, no physiotherapy or rehabilitation Intervention group Number randomised: 6; losses: 1 (withdrew prior to intervention); number analysed at each time point: 5 Treatment management: 6 individual hypnotic interventions over 12 weeks with a psychologist trained in clinical hypnosis, hypnotic audiotapes for daily home practice Control group Number randomised: 6; losses: 0; number analysed at each time point: 6 Treatment management: cast immobilisation for 6 weeks, orthopaedic follow‐up clinics at 1, 3, 6, 9, 12 weeks after injury, no physiotherapy or rehabilitation
Outcomes	Timing of assessments: 1, 3, 6, 9 and 12 weeks after randomisation End of treatment: 12‐week data used End of follow‐up: nil Included in this review: pain Other outcomes measured in the study: use of analgesics, tenderness on palpation, radiographs, hypnotic induction profile, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: not reported Funding sources: supported by a grant from the NIH's Office of Alternative Medicine Declarations of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomized by order of presentation using a list of random numbers."
Allocation concealment (selection bias)	Unclear risk	Comment: method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "both the hypnotherapist and the orthopedist were aware of subject assignment" Comment: participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Quote: "The participating radiologist who reviewed all records was blinded to subject assignment." Comment: however, assessor blinding (the orthopedist and the radiologist conducted the outcome assessment) was not fully implemented.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "an intention‐to‐treat approach was used." Quote: "the 1 subject who dropped out with no follow‐up data could not be included." Comment: because participant loss was small, we did not expect this to impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Gorodetskyi 2010.

Study characteristics
Methods	RCT, parallel design, 3‐armed trial Review comparison: other rehabilitation interventions (electrotherapy)
Participants	Total number of randomised participants: 60 Source population: single centre; Moscow, Russia Inclusion criteria: surgically treated bimalleolar, AO type B2 ankle fracture with comminution and displacement of fragments, age 20 to 60 years, able to begin therapy within 24 hours of the initial procedure, compliance with the ongoing regimen of care Exclusion criteria: limitation that could interfere with delivery of electrical stimulation, history of epilepsy or seizure, current pregnancy, active malignancies, inability to consent to participation Overall baseline characteristics Baseline characteristics for intervention group Age, mean (range): 35.3 (21 to 57) years Sex (m/f): 16/14 Surgery: all had surgery Fracture severity: all had bimalleolar, AO type B2 ankle fracture with comminution and displacement of fragments Baseline characteristics for control group Age, mean (SD): 38.4 (22 to 58) years Sex (m/f):17/13 Surgery: all had surgery Fracture severity: all had bimalleolar, AO type B2 ankle fracture with comminution and displacement of fragments
Interventions	Timing of randomisation: no more than 24 hours after surgery Treatment management for both groups: standard interdisciplinary postoperative care and daily physiotherapy sessions focused on exercise to increase range of motion and mobility. Non‐narcotic analgesic (Ketorolac) as needed up to 3 times per day. Intervention group Number randomised: 30; no losses Treatment management: twice‐daily noninvasive interactive neurostimulation (NIN) for 10 consecutive days, delivered to the tissue at 7 different sites in the foot/ankle region via a pair of concentric electrodes placed in direct contact with the target area Control group Number randomised: 30; no losses Treatment management: twice‐daily sham NIN for 10 consecutive days
Outcomes	Timing of assessments: daily (measured after the morning treatment session) from day 1 to 10 of treatment. End of treatment: 10‐day data used End of follow‐up: nil Included in this review: pain, range of motion Other outcomes measured in the study: swelling
Notes	Study dates: not reported Funding sources: partial funding from Neuro Resource Group Declarations of interest: 1 study author is a paid consultant for Neuro Resource Group; no potential conflicts of interest for other study authors
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "fixed randomisation scheme." Comment: insufficient information on method of randomisation
Allocation concealment (selection bias)	Unclear risk	Quote: "with sealed envelopes." Comment: insufficient information on method of allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "although health care personnel delivering postoperative therapy with the NIN device were necessarily aware of treatment status, all of the patients, evaluating physicians and nurses (including individuals responsible for collection of outcome measures) were blinded to group assignment." Comment: participant blinding implemented but personnel were not blinded
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	Low risk	Quote: "although health care personnel delivering postoperative therapy with the NIN device were necessarily aware of treatment status, all of the patients, evaluating physicians and nurses (including individuals responsible for collection of outcome measures) were blinded to group assignment." Comment: participant blinding implemented
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "although health care personnel delivering postoperative therapy with the NIN device were necessarily aware of treatment status, all of the patients, evaluating physicians and nurses (including individuals responsible for collection of outcome measures) were blinded to group assignment." Comment: assessor blinding implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	CONSORT flowchart showed all participants were followed up and included in the data analysis.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Handolin 2005.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions (electrotherapy)
Participants	Total number of randomised participants: 16 Source population: single centre; Department of Orthopaedics and Traumatology, Helsinki University Central Hospital, Finland Inclusion criteria: closed isolated Weber B lateral malleolar ankle fracture (bimalleolar or trimalleolar fracture excluded), surgical fixation required, age 18 to 65 years Exclusion criteria: bimalleolar and trimalleolar fractures Baseline characteristics for intervention group Age, mean (range): 43.3 (28 to 66) years Sex (m/f): 3/5 Surgery: all had surgery Fracture severity: Weber B unimalleolar Baseline characteristics for control group Age, mean (range): 41.8 (22 to 59) years Sex (m/f): 4/4 Surgery: all had surgery Fracture severity: Weber B unimalleolar
Interventions	Timing of randomisation: 2 weeks after surgery Treatment management for both groups: removable soft cast brace for 6 weeks, partial weight‐bearing at 2 weeks postoperatively, full weight‐bearing at 4 weeks postoperatively Intervention group Number randomised: 8; no losses; analysed: 8 Treatment management: active ultrasound 20 minutes/day for 6 weeks from 2 weeks postoperatively, to the lateral side of the ankle over the fracture line Control group Number randomised: 8; no losses; analysed: 8 Treatment management: sham ultrasound 20 minutes/day for 6 weeks from 2 weeks postoperatively, to the lateral side of the ankle over the fracture line
Outcomes	Timing of assessment: 18 months after randomisation (radiographs also taken at 2, 6, 9 and 12 weeks) End of treatment: nil End of follow‐up: 18‐month data used Included in this review: activity limitation questionnaire, pain, adverse events Other outcomes measured in the study: radiographs, computed tomography, clinical assessment of wound and ankle stability, bone mineral content, ankle dorsiflexion, range of motion, swelling
Notes	Study dates: 7 November 2001 to 1 January 2002 Funding sources: research grants from the Foundation for Orthopaedical and Traumatological Research in Finland, Helsinki University Central Hospital, and the Academy of Finland. Ultrasound devices supplied by Exogen Inc. Declarations of interest: not reported Note: we received additional information from study author, L Handolin (regarding methods of randomisation and allocation, details of assessment, adverse events, dropouts, additional data)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: insufficient information on method of randomisation
Allocation concealment (selection bias)	Low risk	Additional information obtained from the author: randomisation was performed off‐site.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "The patients were provided randomly in a double‐blind manner with either an active... or a sham ... ultrasound device".
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	Low risk	Quote: "The patients were provided randomly in a double‐blind manner with either an active... or a sham ... ultrasound device". Quote: "All the analyses were performed blind to the ultrasound treatment."
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "All the analyses were performed blind to the ultrasound treatment."
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Hedstrom 1994.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 53 Source population: single centre; Sweden Inclusion criteria: closed lateral malleolar fracture dislocated > 2 mm, surgical fixation required, adults Exclusion criteria: open fractures, children, unable to cooperate Baseline characteristics for intervention group Age, mean (range): 44 (15 to 70) years Sex (m/f): 13/15 Surgery: all had surgery Fracture severity: Lauge Hansen supination‐eversion II/supination‐eversion III/supination‐eversion IV/pronation‐abduction III/pronation‐eversion III/pronation‐eversion IV: 13/2/8/1/1/3 Baseline characteristics for control group Age, mean (SD): 41 (16 to 71) years Sex (m/f): 11/14 Surgery: all had surgery Fracture severity: Lauge Hansen supination‐eversion II/supination‐eversion III/supination‐eversion IV/pronation‐abduction III/pronation‐eversion III/pronation‐eversion IV: 10/5/7/1/1/1
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: weight‐bearing Intervention group Number randomised: 28; losses at 18 months: 2 (lost to follow‐up) Treatment management: orthosis for an unspecified period, unloaded dorsiflexion and plantarflexion exercises without orthosis at least 5 times/day Control group Number randomised: 25; losses at 18 months: 4 (lost to follow‐up) Treatment management: walking cast for an unspecified period
Outcomes	Timing of assessment: 3, 6 and a minimum of 18 months after randomisation End of treatment: nil End of follow‐up: 18‐month data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: radiographs, ankle dorsiflexion, and plantarflexion range of motion
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "the patients were randomly allocated." Comment: insufficient information on method of randomisation
Allocation concealment (selection bias)	Unclear risk	Insufficient information on method of allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "at 18 months, six patients were lost to follow‐up." Comment: study is small, and loss of 6 participants equates to loss of 11%. More losses were from the control group, and it is possible that this could impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Hernandez 2006.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions (electrotherapy)
Participants	Total number of randomised participants: 24 Source population: single centre; Israel Inclusion criteria: closed ankle fracture, surgical fixation required, adults Exclusion criteria: conditions preventing mobilisation or immobilisation, history of cardiomyopathy, pacemaker use, peripheral vascular disease, females of childbearing age, and medication affecting bone metabolism Baseline characteristics for intervention group Age, mean (SD): 41.5 (± 13.4) years Sex (m/f): 8/4 Surgery: all had surgery Fracture severity, uni‐/bi/trimalleolar: 2/9/1 Baseline characteristics for control group Age, mean (SD): 43.5 (± 14.1) years Sex (m/f): 8/4 Surgery: all had surgery Fracture severity, uni‐/bi/trimalleolar: 5/5/2
Interventions	Timing of randomisation: 24 hours after surgery Treatment management for both groups: walking cast for 6 weeks, weight‐bearing as tolerated, passive ankle range of motion exercises for 24 hours 1 to 2 days after surgery, instructions in cast care, walking, and stairs Intervention group Number randomised: 12; losses: 2 (reasons not reported) Treatment management: electrical stimulator (Myospare) at an intensity sufficient to elicit a muscular twitch to gastrocnemius, 1 minute on/4 minutes off, for an unspecified length of time daily for 6 weeks Control group Number randomised: 12; losses: 1 (reasons not reported) Treatment management: same as 'treatment for both groups'
Outcomes	Timing of assessments: 2, 6, and 12 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 12‐week data used Included in this review: activity limitation questionnaire, pain, adverse events Other outcomes measured in the study: ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: May 2004 to November 2011 Funding: not reported Declarations of interest: not reported Note: we received additional information from study author, E Leibner (regarding additional data, dropouts).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "patients were randomised." Comment: insufficient information on method of randomisation
Allocation concealment (selection bias)	Unclear risk	Quote: "using pre‐prepared, identical, sealed opaque envelopes." Comment: insufficient information provided about numbering of envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "No blinding was used." Comment: participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Quote: "No blinding was used."
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Assessor blinding not implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "Two patients did not return for scheduled follow‐up visits and one patient did not comply with the cast treatment protocol. These patients were not included in the analysis." Comment: small study and loss represented a large percentage. We could not be certain of the impact of this loss on data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study reported only in an abstract and is not published in a peer‐reviewed journal.

Honigmann 2007.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 45 Source population: single centre; Switzerland Inclusion criteria: isolated and displaced Weber A or B ankle fracture, surgical fixation required, age 16 to 65 years, BMI < 35 kg/m2 Exclusion criteria: not reported Baseline characteristics for intervention group Age, median (range): 42.5 (17.3 to 61.9) years Sex (m/f): 14/9 Surgery: all had surgery Fracture severity, Weber A/B: 1/22 Baseline characteristics for control group Age, median (range): 38.1 (18.5 to 65.7) years Sex (m/f): 9/13 Surgery: all had surgery Fracture severity, Weber A/B: 0/22
Interventions	Timing of randomisation: after surgery Treatment management for both groups: partial weight‐bearing with crutches, ankle range of motion exercises Intervention group Number randomised: 23; losses: 2 (lost to follow‐up at 3 weeks, and 6 weeks); analysed: 21 Treatment management: VACOped (removable immobilisation in a vacuum orthosis) for 6 weeks, full weight‐bearing after 14 days, walking without crutches allowed at 3 weeks, orthosis off for ankle range of motion exercises and sleep Control group Number randomised: 22; no losses; analysed: 22 Treatment management: plaster splint 2 to 4 days, then bandage only, full weight‐bearing after 6 weeks
Outcomes	Timing of assessments: 6 and 10 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 10‐week data used Included in this review: activity limitation questionnaire, quality of life, participant satisfaction, pain, adverse events Other outcomes measured in the study: return to work, duration of hospitalisation, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author, P Honigmann (method of randomisation, sex of participants, confirmation of results reported)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: randomisation sequence provided by external independent investigator
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by sealed, opaque and consecutively numbered envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although loss was only in the intervention group, we did not expect the loss of 2 participants to impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Jansen 2018.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care (or another physical therapy intervention)
Participants	Total number of randomised participants: 50 Source population: single centre, Level 1 trauma centre; Department of Trauma, Hand, Plastic and Reconstructive Surgery, University Clinics Wuerzburg, Germany Inclusion criteria: operatively treated unstable ankle fractures as the only injury of the lower extremity with a type B‐ or C‐fracture in the Weber classification, the need of partial weight‐bearing for 6 weeks, age > 18 years, possibility to perform physiotherapy and ACM and no problems in walking before the injury Exclusion criteria: psychiatric comorbidities; unable to perform partial weight‐bearing Baseline characteristics for intervention group Age: 46 (22 to 73) years (note: study authors did not state the type of average or distribution value) Sex (m/f): 14/11 Surgery: all had surgery Fracture severity, Weber B/C: 15/10 Baseline characteristics for control group Age: 53 (22 to 73) years (note: study authors did not state the type of average or distribution value) Sex (m/f): 13/11 Surgery: all had surgery Fracture severity, Weber B/C: 12/12 (note: data are missing for 1 participant)
Interventions	Timing of randomisation: after surgery Treatment management for both groups: if required, lag screw removed after 6 weeks (prior to full weight‐bearing). Additional boot applied for first 6 weeks postoperatively, depending on surgeon preference. Intervention group Number randomised: 25; losses 3 (refused further participation); number analysed at 12 weeks: 22 Treatment management: ACM (+PT as below) ‐ using CAMOped device (Camoped©, OPED® GmbH, Germany). Training started 2 to 5 days postoperatively, depending on condition of soft tissue. For use 20 minutes a day in the participant's home. Control group Number randomised: 25; losses 3 (1 excluded due to protocol violation; 2 refused further participation); number analysed at 12 weeks: 22 Treatment management: physiotherapy. Started on POD1, daily treatment of 20 minutes during hospitalisation with focus on mobilisation of participant using 2 crutches and ability to maintain partial weight‐bearing. After discharge, physiotherapy performed for 20 minutes, 2 to 3 times a week for 6 weeks, focusing on treatment of oedema and regaining ROM, avoiding forced dorsiflexion
Outcomes	Timing of assessments: clinical follow‐up was performed 6 and 12 weeks after the operation End of treatment: 6 weeks End of follow‐up: 12 weeks Included in this review: activity limitation Other outcomes measured in the study: able to fully weight‐bear at 6 weeks, return to work, dynamic pedobarography, ankle dorsiflexion range of motion
Notes	Study dates: not reported Funding sources: none Declarations of interest: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "randomized by lot to the two groups". Comment: insufficient information
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Not possible to blind participants
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Quote: "Clinical follow‐up was performed 6 and 12 weeks after the operation by two, not blinded, examiners."
Incomplete outcome data (attrition bias) All outcomes	Low risk	Losses were few, balanced between groups and well explained.
Selective reporting (reporting bias)	Unclear risk	Retrospectively registered with a clinical trials register (DRKS00009979). It is not feasible to effectively assess risk of selective reporting bias from this document because it was submitted after the start of the study.
Other bias	Low risk	We identified no other sources of bias.

Kearney 2021.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical or non‐surgical fracture management)
Participants	Total number of randomised participants: 669 Source population: multicentre, 20 trauma units; NHS Trust sites, UK Inclusion criteria: ≥ 18 years of age, closed ankle fracture. After 3 weeks of usual care treatment, further assessment for eligibility was made ‐ the treating clinician decided that immobilisation (with cast or brace) was required Exclusion criteria: fracture secondary to known metastatic disease, complex intra‐articular fracture (e.g. Pilon fracture), wound complications contraindicating a removable bracing, pre‐existing neuropathic joint disease, previous ankle fracture already randomised in the present trial, unable to adhere to trial follow‐up procedures, or required close contact casting Baseline characteristics for intervention group Age, mean (SD): 45.9 (± 16) years Sex (m/f): 152/183 Surgery: 182/153 Fracture severity: not specified Baseline characteristics for control group Age, mean (SD): 46.7 (± 17) years Sex (m/f): 136/198 Surgery: 182/152 Fracture severity: not specified
Interventions	Timing of randomisation: either immediately at presentation to the trauma centre (for non‐operatively managed fractures) or up to 3 weeks after injury Treatment management for both groups: treating clinical would usually treat the fracture in a cast for a minimum of 3 weeks, whether treated operatively or non‐operatively. Any adjunctive treatments (e.g. antibiotics) at the discretion of attending clinicians Intervention group Number randomised: 335; losses: 75 (11 withdrew, 64 missing); number analysed at final time point: 260 Treatment management: removable braces of fixed‐angle design (brand was not standardised and was dependent on local procedures). Participants encouraged (in verbal and written format) to take off the brace to complete active unloaded ankle ROM exercises as often as pain allowed, with a recommendation for each movement to be performed 10 times, 3 times a day Control group Number randomised: 334 ; losses: 92 (16 withdrew, 76 missing); number analysed at final time point: 242 Treatment management: standard below‐knee cast. Participants started active unloaded ankle ROM exercises once the cast was removed.
Outcomes	Timing of assessments: 6, 10,16 weeks. 6, 12, 18 and 24 months End of treatment: 6 weeks End of follow‐up: 24 months Included in this review: activity limitation, QoL, adverse events, economic data Other outcomes measured in the study: fracture healing, strength
Notes	Study dates: October 2017 to September 2019 Funding sources: NIHR as part of a personal fellowship, and NIHR Oxford Biomedical Research Centre Declarations of interest: all study authors had no financial relationships with any organisations that might have an interest in the submitted work in the previous three years. Some authors declared membership of NIHR panels/boards or receipt of other NIHR funding
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Use of a secure web‐based system
Allocation concealment (selection bias)	Low risk	Use of an independent randomisation team
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to treatment
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded to treatment
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Personnel (outcome assessors) were not blinded to treatment
Incomplete outcome data (attrition bias) All outcomes	High risk	High participant loss and reasons for losses are not fully explained. Study authors explored these losses in sensitivity analyses but concluded that missingness did not impact on final interpretation of the findings.
Selective reporting (reporting bias)	Low risk	Study prospectively registered with a clinical trials register (ISRCTN15537280) and protocol also published. Reported outcomes are mostly consistent with planned outcomes.
Other bias	Low risk	We identified no other sources of bias.

Keene 2016.

Study characteristics
Methods	RCT, 3‐armed cross‐over design Comparison group: other rehabilitation interventions
Participants	Total number of randomised participants: 18 Source population: single centre; John Radcliffe NHS Hospital, Oxford, UK Inclusion criteria: adults ≥ 18 years of age, undergone internal fixation surgery for isolated ankle fractures Exclusion criteria: AO type C/suprasyndesmotic ankle fracture, inability to walk outdoors unaided prior to fracture, severe mental health disorder, any previous severe lower‐limb fracture Baseline characteristics (overall) Age, mean (SD): 47 (± 4) years Sex (m/f): 10/8 Surgery: all had surgery Fracture severity, Weber A/B; uni‐/bi‐/trimalleolar: 7/11; 4/5/9
Interventions	Timing of randomisation: 1 hour after removal of cast at 6 weeks postoperatively Treatment management for all groups: supports applied according to manufacturer guidelines by the lead investigator, an experienced physical therapist Intervention groups Treatment management 1: stirrup brace Treatment management 2: Tubigrip compression bandage Treatment management 3: removable below‐knee walker
Outcomes	Timing of assessments: after removal of each treatment option End of treatment: after removal of each treatment option End of follow‐up: after removal of each treatment option Included in this review: no outcomes. No data available from first study time point. Outcomes reported in the study report: pain/comfort, ankle dorsiflexion and plantarflexion, adverse events, gait measures, step length, single support time symmetry index, step width
Notes	Study dates: June 2012 to March 2013 Funding sources: NIHR Declarations of interests: study authors declared no conflicts of interest Note: data were not available from the first study period, and we therefore did not collect any data. As we included no data from this study in the review, we did not complete a risk of bias assessment for this study. Clinical trials registration ISRCTN84536917

Keene 2022.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care (or another physical therapy intervention)
Participants	Total number of randomised participants: 61 Source population: 5 NHS hospitals; UK Inclusion criteria: adults ≥ 50 years of age with an ankle fracture; undergoing surgical fixation, or conservative management involving ankle immobilisation for ≥ 4 weeks Exclusion criteria: unable to understand spoken and written English; do not have the capacity to consent to study participation; not ambulatory prior to the injury; considered inappropriate for referral to physiotherapy (in the opinion of the clinician); unable to attend outpatient physiotherapy at a participating centre; serious concomitant disease (such as terminal illness); avulsion fractures or other fractures not requiring surgery or immobilisation for definitive management; bilateral lower limb fractures; displaced femoral neck fracture, diagnosed by CT or X‐ray, > 75 years of age with a history of injury Baseline characteristics for intervention group Age, mean (SD): 63.7 (± 10.4) years Sex (m/f): 11/20 Surgery (Y/N): 15/13 (data for 3 are missing) Fracture severity: not reported Baseline characteristics for control group Age, mean (SD): 63.3 (± 7.2) years Sex (m/f): 8/22 Surgery (Y/N): 12/15 (data for 3 are missing) Fracture severity: not reported
Interventions	Timing of randomisation: once the participant could mobilise with unrestricted weight‐bearing and complete ankle exercises, usually at 6 to 8 weeks after injury Treatment management in both groups: individualised education and advice, with other aspects of health and social care continuing as normal Intervention group Number randomised: 31; 2 lost to follow‐up; 5 did not complete final follow‐up assessment; data reported for variable participant numbers for each outcome and time point Treatment management: progressive exercise group; home exercise and advice programme overseen by a physiotherapist over ≤ 6 sessions within 16 weeks Control group Number randomised: 30; 2 lost to follow‐up; 2 did not complete final follow‐up assessment; data reported for variable participant numbers for each outcome and time point Treatment management: best practice advice group; home exercise and advice programme initiated during single face‐face sessions with a physiotherapist, and then performed unsupervised at home
Outcomes	Timing of assessments: 3 and 6 months End of treatment: 16 weeks End of follow‐up: 6 months Included in this review: activity limitation (OMAS, lower extremity function), pain (VAS), HRQoL (EQ‐5D), adverse events Other outcomes measured in the study: fear of falls; self‐efficacy; return to desired activities; walking aid use and distance; exercise adherence; health resource use; ankle joint range; muscle strength; mobility and balance
Notes	Study dates: October 2018 to August 2019 Funding sources: NIHR post‐doctoral fellowship; NIHR Biomedical Research Centre; NIHR Collaboration for Leadership in Applied Health Research and cAre Declarations of interest: institutions of the study authors have received research grant funding from NIHR, European Union, Royal College of Surgeons England, and industry Note: pilot study. Follow‐on study is currently in 'Ongoing studies'
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "centralised computer randomisation service provided by the Oxford Clinical Trials Research Unit"
Allocation concealment (selection bias)	Low risk	Quote: "centralised computer randomisation service provided by the Oxford Clinical Trials Research Unit"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to treatments
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants aware of treatment allocation
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "Physiotherapists or training research associates independent of the clinical team collected the objective outcome measures at the 6‐month follow‐up time point"
Incomplete outcome data (attrition bias) All outcomes	High risk	Although few participants were lost to follow‐up, there were more losses at 6 months in the progressive exercise group. Given that the sample size was small, we were unclear if these few losses may have impacted the results.
Selective reporting (reporting bias)	Low risk	Prospective clinical trials registration (ISRCTN16612336). Outcomes reported in the published study report are consistent with outcomes in the clinical trials register.
Other bias	Low risk	We identified no other sources of bias.

Kortegangas 2019.

Study characteristics
Methods	RCT, parallel design, 3‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after non‐surgical fracture management)
Participants	Total number of randomised participants: 247 Source population: multicentre; 2 major trauma centres in Finland Inclusion criteria: all skeletally mature participants (≥ 16 years) with an isolated Weber B type fibular fracture and congruent ankle mortise (medial clear space < 4 mm and ≤ 1 mm wider than the superior clear space) in the primary radiographs visiting the emergency departments; ability to walk unaided before the current trauma and enrolment < 7 days after the injury Exclusion criteria: previous deltoid ligament injury (a history of ankle twisting injury requiring cast treatment), other important fracture in the affected ankle or foot area, current bilateral ankle fracture, concomitant tibial fracture, pathological fracture, diabetic or other neuropathy, inadequate cooperation, or permanent residence outside the catchment area Baseline characteristics for intervention group (3‐week orthosis) Age, mean (SD): 46 (± 19) years Sex (m/f): 36/44 Surgery: none had surgery Fracture severity: not reported Baseline characteristics for intervention group (3‐week cast) Age, mean (SD): 46 (± 17) years Sex (m/f): 43/40 Surgery: none had surgery Fracture severity: not reported Baseline characteristics for control group Age, mean (SD): 45 (± 18) years Sex (m/f): 46/38 Surgery: none had surgery Fracture severity: not reported
Interventions	Timing of randomisation: in the emergency department Treatment management for both groups: participants received guidance from a physiotherapist on walking with crutches. Weight‐bearing was permitted in all 3 groups immediately after the application of the cast or orthosis. Participants received written and verbal instructions on how to cope with the ankle fracture and support (cast and orthosis groups) and how to remove and reapply the support (orthosis group) Intervention group Number randomised: 80; overall losses: 14 (3 lost contact, 11 withdrew); number analysed at 12 weeks: 73; number analysed at 12 months: 66 Treatment management: orthosis (Dynacast/Ortho‐Glass; BSN medical, Rutherford College, NC) for 3 weeks. Participants in the 3‐week orthosis group were instructed to wear the brace as necessary or according to comfort. They could take off the orthosis when needed and move their ankle freely. At 3 weeks, the orthosis was removed. Participants were instructed to continue weight‐bearing as tolerated with no need for further bracing or support. Intervention group Number randomised: 83; overall losses 10 (2 lost contact, 7 withdrew, 1 died); number analysed at 12 weeks: 75; number analysed at 12 months: 73 Treatment management: below‐the‐knee cast for 3 weeks. At 3 weeks, the cast was removed. Participants were instructed to continue weight‐bearing as tolerated with no need for further bracing or support. Control group Number randomised: 84; overall losses 11 (4 lost contact, 7 withdrew); number analysed at 12 weeks: 73; number analysed at 12 months: 73 Treatment management: cast for 6 weeks
Outcomes	Timing of assessments: 6 and 12 weeks, and at 52 weeks (additional clinical follow‐up also at 3 weeks) End of treatment: 6 weeks End of follow‐up: 52 weeks Included in this review: activity limitation, QoL, impairment measures of pain, adverse events. Other outcomes measured in the study: fracture union, ankle dorsiflexion range of motion, plantarflexion range of motion Note: in the primary analysis, we compared the 3‐week orthosis group with the 6‐week cast group
Notes	Study dates: December 2012 to June 2016 Funding sources: supported by state funding for university level health research (Oulu University Hospital and Tampere University Hospital) Declarations of interest: 1 study author declared an honorarium for a lecture. Other study authors declare no support from any organisation for the submitted publication, no financial relationships with any organisation that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "computerised random number generator"
Allocation concealment (selection bias)	Low risk	A biostatistician who was not involved in the clinical care of participants prepared sealed, consecutively numbered and opaque envelopes.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to the treatment
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded to treatment group
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "the outcome assessors (trained study physiotherapists) of other study outcomes and doctors carrying out the follow‐up visits were blinded to group allocation when possible"
Incomplete outcome data (attrition bias) All outcomes	High risk	Losses were clearly reported and reasonably balanced between groups. However, loss for 12‐month data was 24% overall.
Selective reporting (reporting bias)	Low risk	Prospectively registered with clinical trials register (NCT01758835). Reported outcomes are consistent with planned outcomes
Other bias	Low risk	We identified no other sources of bias.

Lehtonen 2003.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 100 Source population: single centre; Section of Orthopaedics, Tampere University Hospital, Finland Inclusion criteria: acute (< 72 hours), displaced or unstable Weber A or B ankle fractures, surgical fixation required, stable for early mobilisation, ability to cope with either protocol, no open fractures, pilon fracture, Weber C fracture needing syndesmosis screw or other severe injuries Exclusion criteria: fracture considered to be insufficiently stable for early mobilisation, open fracture, pilon fracture of the tibia, Weber type‐C fracture of the ankle requiring screw placement across the syndesmosis, other severe injuries, and unable to cope with either of the 2 postoperative protocols Baseline characteristics for intervention group Age, mean (SD): 41 (± 13) years Sex (m/f): 25/25 Surgery: all had surgery Fracture severity, Weber A/B; uni‐/bi‐/trimalleolar: 2/48; 30/14/6 Baseline characteristics for control group Age, mean (SD): 41 (± 13) years Sex (m/f): 31/19 Surgery: all had surgery Fracture severity, Weber A/B; uni‐/bi‐/trimalleolar: 2/48; 29/14/7
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: FoamWalker leg brace until randomisation, non‐weight‐bearing and crutches for the first 2 weeks, partial weight‐bearing for the next 4 weeks, but full weight‐bearing allowed at 4 weeks, strength and balance exercises 10 times/exercise, 5 to 10 times/day after 6 weeks, taping after 6 weeks as required, running, stairs and sports after achieving full ankle range of motion, orthosis for sports or strenuous activities for 3 to 6 months after injury Intervention group Number randomised: 50; losses: 4 at 2‐year follow‐up (reasons not reported by group, see below); analysed: at 2 and 6 weeks ‐ 50; at 12 weeks ‐ 49; at 52 weeks ‐ 49; at 2 years ‐ 46 Treatment management: air‐stirrup for 6 weeks, daily active and passive ankle and subtalar range of motion exercises without air‐stirrup Control group Number randomised: 50; losses: 8 at 2‐year follow‐up (reasons not reported by group, see below); analysed: at 2 and 6 weeks ‐ 50; at 12 weeks ‐ 50; at 52 weeks ‐ 47; at 2 years ‐ 42 Treatment management: below‐knee plaster cast for 2 weeks, then fibreglass walking cast for the next 4 weeks Overall reasons for losses: unwilling to continue in the study (4); moved to another area (3); lost to follow‐up owing to alcohol‐related problems (2); severe knee ligament injury (1); herniated lumbosacral disc (1); complicated rheumatoid arthritis (1)
Outcomes	Timing of assessments: 2, 6, 12, 52 and 104 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 104‐week data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: calf circumference, return to work, radiographs, length of hospitalisation, ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: November 1995 to April 1998 Funding sources: funding from The Medical Research Fund of the Finnish Office for Health Care Technology Assessment (Fin OHTA), Helsinki, Finland Declarations of interest: quote: "None of the authors received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, educational institution, or other charitable or nonprofit organization with which the authors are affiliated or associated"
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "patients were randomly allocated." Comment: insufficient information on method of randomisation
Allocation concealment (selection bias)	Unclear risk	Quote: "use of sealed envelopes." Comment: insufficient information on numbering or opacity of envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	The participants and therapists were not blinded.
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	The participants were not blinded.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "the follow‐up examinations were performed by the same independent physician who had not been involved in the actual treatment of any of the patients." Comment: The assessor was blinded.
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "A total of 12 patients were lost to follow‐up." Comment: losses at 2 years were higher in the control group. Reasons for losses were explained and justified. We could not be certain, however, of the impact of these losses on the overall outcome data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Lin 2008b.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 94 Source population: multicentre; 3 large teaching hospitals, Australia Inclusion criteria: ankle fracture treated with cast immobilisation with or without surgical fixation (average length of immobilisation = 43 days), cast removed in preceding 7 days, approval to weight‐bear as tolerated or partial weight‐bear, referral to outpatient physiotherapy for treatment, at least 2 out of 10 pain (VAS) in the affected ankle on equal weight‐bearing at cast removal, available for the 24‐week follow‐up period, no concurrent pathologies Exclusion criteria: not reported Baseline characteristics for intervention group Age, mean (SD): 42.5 (± 14.3) years Sex (m/f): 21/26 Surgery (Y/N): 30/17 Fracture severity, uni‐ or bimalleolar/trimalleolar: 30/17 Baseline characteristics for control group Age, mean (SD): 40.8 (± 15.1) years Sex (m/f): 30/17 Surgery (Y/N): 26/21 Fracture severity, uni‐ or bimalleolar/trimalleolar: 31/16
Interventions	Timing of group allocation: within 7 days of cast removal Treatment management for both groups: a physiotherapy programme including exercise, gait re‐training, progression of walking aids, advice on prognosis and return to activities and ice, elevation and compression if required. After 4 weeks, participants in both groups could progress to exercise other than those from the standardised exercise programme. Intervention group Number randomised: 47; losses: 1 (withdrew consent); analysed: 47 (for effectiveness data), and 46 (for economic data) Treatment management: 2 sessions a week for 4 weeks applying 3 sets of 60 seconds of large amplitude (grade III) anterior‐posterior glides of the talus. The treatment could be progressed by increasing the number of repetitions, the force of application and by increasing the range of dorsiflexion in which the treatment was performed. Control group Number randomised: 47; losses: 2 (withdrew consent, lost to follow‐up); analysed: 47 (for effectiveness data), and 46 (for economic data) Treatment management: 2 sessions in the first week, followed by 1 session a week for the next 3 weeks
Outcomes	Timing of assessments: 4, 12 and 24 weeks after randomisation End of treatment: 4‐week data used End of follow‐up: 24‐week data used Included in this review: activity limitation questionnaire, activity limitation test, quality of life, participant satisfaction, pain, adverse events Other outcomes measured in the study: return to work, return to sports and leisure activities, global perceived effect of treatment, number of days to pain‐free walking, ankle dorsiflexion range of motion
Notes	Study dates: November 2004 to January 2007 Funding: funded by the Motor Accidents Authority of New South Wales, Australia, and the New South Wales Institute of Sport, Australia. CL received a scholarship from the National Health and Medical Research Council, Australia. Declarations of interest: not reported Note: we received additional information received from study author, C Lin (regarding additional data)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "the randomization sequence was computer‐generated a priori by an independent investigator to ensure concealment."
Allocation concealment (selection bias)	Low risk	Quote: "sealed, opaque and consecutively numbered envelopes."
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "outcome assessments were conducted by an assessor blinded to treatment allocation." Comment: assessor blinding implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "statistical analysis was based on intention‐to‐treat principles" Quote: "missing data were replaced by the last known value carried forward or by the means of the allocated group if no data were available"
Selective reporting (reporting bias)	Low risk	All expected (mentioned in protocol and pre‐specified) outcomes included
Other bias	Low risk	We identified no other risks of bias.

Losch 2002.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 40 Source population: single centre, Germany Inclusion criteria: isolated Weber C ankle fracture, surgical fixation required Exclusion criteria: not reported Baseline characteristics for intervention group Age, mean: 37 years Sex (m/f): 8/12 Surgery: all had surgery Fracture severity: Weber C Baseline characteristics for control group Age, mean: 38 years Sex (m/f): 8/12 Surgery: all had surgery Fracture severity: Weber C
Interventions	Timing of group allocation: not stated Treatment management for both groups: no details Intervention group Number randomised: 20; losses: 1 (reasons not reported) Treatment management: combi‐cast orthosis for an unspecified period, 3 hours of functional rehabilitation programme given by the physiotherapist 3 times/week for an unspecified period after discharge from hospital Control group Number randomised: 20; losses: 6 (reasons not reported) Treatment management: walking cast for 6 weeks (but weight‐bearing status not reported), non‐structured physiotherapy
Outcomes	Timing of assessments: 10 weeks after randomisation End of treatment: nil End of follow‐up: 10‐week data used Included in this review: activity limitation questionnaire Other outcomes measured in the study: length of absence from work, radiographs, ankle dorsiflexion and plantarflexion range of motion
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: published in German. Study eligibility assessed by the German Cochrane Centre, data extraction performed by Erwin Scherfer and Jutta Jablonski. Risk of bias assessed by Arianne Verhagen and Nicole Donkers.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method of randomisation not described
Allocation concealment (selection bias)	Unclear risk	Method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Seven out of 40 participants were lost to follow‐up (17.5%). Study authors do not report reasons for losses, and we noted that most loses were in the control group.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study published only as an abstract and we assumed that this study was not peer‐reviewed.

Matthews 2020.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 157 Source population: single centre; Nottingham University Hospitals, Nottingham, UK Inclusion criteria: people with isolated closed Weber B type (AO44 B1, B2, B3) ankle fractures, which are stable following ORIF; ≥ 18 years of age, living independently, capable of independently reading and completing study paperwork in English Exclusion criteria: inability to provide informed consent or declining to participate; co‐morbidities (taking prescription medication for diabetes, non‐healing leg/foot ulcers, oral or IV administered steroid users, pre‐existing ankle arthritis, contralateral lower limb injury/condition, neurological disorders); at 2‐week clinic visit, judgement of notable risk to satisfactory healing after early wound movement; unable to commit to weekly clinic visits Baseline characteristics for intervention group Age, mean (SD): 43.5 (± 15.4) years Sex (m/f): 35/45 Surgery: all had surgery Fracture severity: all Weber B Baseline characteristics for control group Age, mean (SD): 42.5 (± 14) years Sex (m/f): 41/36 Surgery: all had surgery Fracture severity: all Weber B
Interventions	Timing of randomisation: 2 weeks after surgery Treatment management for both groups: same treatment for first 2 weeks postoperatively. All participants were non‐weight‐bearing throughout the study until the 6‐week review. Intervention group Number randomised: 80; losses at short‐term follow‐up: 12 (reasons not given); analysed: 68 Treatment management: EMADE (early motion and directed exercise) physiotherapy. Combines non‐weight‐bearing progressive home exercises with manual therapy, advice and education between weeks 2 and 4 postoperatively. Conducted by participants up to 6 times a day. Fitted with a removable below‐knee cast with Velcro straps. Weekly sessions at 30 minutes, including administration time Control group Number randomised: 77; losses at short‐term follow‐up: 15 (reasons not given); analysed: 62 Treatment management: below‐knee cast. Non‐weight‐bearing up to the 6‐week review
Outcomes	Timing of assessments: 2, 6, 12, 24 and 52 weeks End of treatment: 6 weeks postoperatively End of follow‐up: 52 weeks postoperatively. Data were only available for this review at 12 weeks. Included in this review: activity limitation, quality of life, adverse events Other outcomes measured in the study: return to work and driving, X‐ray findings, ankle dorsiflexion and plantarflexion
Notes	Study dates: not reported Funding sources: Centre for Sport, Exercise and Osteoarthritis Research Versus Arthritis. AO Foundation. NIHR via the Portfolio. CSP Charitable Trust. Nottingham Hospitals Charity Declarations of interest: not reported Note: some information taken from abstract and protocol publications. Outcome data and baseline characteristics were provided by study authors.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Online computer service
Allocation concealment (selection bias)	Low risk	Quote: "facility generated the codes concealed from the research team"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to treatment
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Outcome assessors were not blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	Large number of participants lost
Selective reporting (reporting bias)	Unclear risk	Because we did not have information from a final published report, it was not feasible to assess risk of selective reporting bias against clinical trials registration of published protocols.
Other bias	High risk	We obtained data from study authors. Data has not yet been published in a peer‐reviewed journal.

Molund 2020.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 113 Source population: single centre, level 1 trauma centre; Norway Inclusion criteria: 18 to 65 years with dislocated or unstable closed Weber B ankle fractures, who were operated on according to the hospital’s guidelines. Only participants who were considered to have good compliance and could follow the rehabilitation protocol were included. Exclusion criteria: previous trauma to the tibia or fibula, previous ankle or foot surgery, symptomatic arthritis or arthrosis in the foot or ankle, generalised joint disease, multi‐trauma and other diseases that impair walking, or pathological fractures Baseline characteristics for intervention group Age, mean (SD): 43.8 (± 12.2) years Sex (m/f): 21/35 Surgery: all had surgery Fracture severity: Weber B Baseline characteristics for control group Age, mean (SD): 41.3 (± 11.8) years Sex (m/f): 23/34 Surgery: all had surgery Fracture severity: Weber B
Interventions	Timing of randomisation: day after surgery Treatment management for both groups: both groups instructed to use crutches on stairs and on flat surfaces the day after surgery, as well as general postoperative exercises for muscular control and circulation Intervention group Number randomised: 56; losses: 4 (lost to follow‐up); number analysed at 52 weeks: 52 Treatment management: spring‐loaded ankle trainer from first postoperative day. Participants attended 3 daily sessions with 10 x 10 mins of exercise. Ankle trainer used for first 3 weeks after surgery. Then conventional rehabilitation was continued. Control group Number randomised: 57; losses: 8 (lost to follow‐up); number analysed at 52 weeks: 49 Treatment management: use of a non‐elastic band (conventional physiotherapy) for stretching the ankle in dorsal motion. Daily sessions with 10 repetitions of a 30‐second hold in maximum ankle dorsal movement in 3 sets.
Outcomes	Timing of assessments: 3, 6, 12 and 52 weeks End of treatment: 6 weeks End of follow‐up: 52 weeks Included in this review: activity limitation, impairment measures of pain Other outcomes measured in the study: length of hospital stay, analgesic use, use of antibiotics, ankle dorsiflexion range of motion
Notes	Study dates: November 2014 to December 2017 Funding sources: Oslo University Hospital, Norway Declarations of interest: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Use of a web‐based "randomizer"
Allocation concealment (selection bias)	Low risk	Whilst randomisation was "done by the physiotherapist including the patient", we assumed that the code could not be manipulated because it was delivered by an external source (Norwegian University of Science and Technology).
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants and personnel not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	High for all participant‐reported outcomes because of lack of blinding
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	For other outcomes, detection bias depended on timing of data collection ‐‐ outcomes at 3 weeks were collected by unblinded physiotherapist (high risk), but later outcomes were collected by a blinded physiotherapist (low risk).
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although there were more losses at 12 months in the control group (8 losses compared with 4 losses), we judged the risk of attrition bias to be low because the loss was less than 10%.
Selective reporting (reporting bias)	Unclear risk	Retrospectively registered with a clinical trials register (NCT04168307). We could not use this information to effectively assess risk of selective reporting bias.
Other bias	Low risk	We identified no other sources of bias.

Moseley 2005.

Study characteristics
Methods	RCT, parallel design, 3‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 150 Source population: multicentre; plaster clinics of 2 large teaching hospitals, Australia Inclusion criteria: ankle fracture treated with cast immobilisation with or without surgical fixation (average length of immobilisation = 45 days), cast removed in preceding 5 days, approval to weight‐bear as tolerated or partial weight‐bear, decreased passive ankle dorsiflexion, complete skeletal growth, no concurrent pathologies Exclusion criteria: not reported Baseline characteristics for intervention group (long‐duration stretch) Age, mean (SD): 47 (± 15) years Sex (m/f): 24/27 Surgery (Y/N): 24/27 Fracture severity, Weber A/B/C/missing: 9/31/3/8 Baseline characteristics for intervention group (short‐duration stretch) Age, mean (SD): 47 (± 15) years Sex (m/f): 23/26 Surgery (Y/N): 26/23 Fracture severity, Weber A/B/C/missing: 11/30/5/3 Baseline characteristics for control group Age, mean (SD): 49 (± 15) years Sex (m/f): 25/25 Surgery (Y/N): 34/16 Fracture severity, Weber A/B/C/missing: 9/30/7/4
Interventions	Timing of randomisation: within 5 days of cast removal Treatment management for all groups: home exercise programme, up to 5 sessions with the physiotherapist (gait re‐training and advice, ice, compression and elevation if required) Intervention group (long‐duration stretch) Number randomised: 51; losses: 7 (reasons for losses not reported by group, see below); analysed: at 4 weeks ‐ 46; at 3 months ‐ 44 Treatment arrangement: 30 minutes of stretches/day Intervention group (short‐duration stretch) Number randomised: 49; losses: 5 (reasons for losses not reported by group, see below); analysed: at 4 weeks ‐ 46; at 3 months ‐ 44 Treatment arrangement: 12 sets of 30 second stretches/day, i.e. a total of 6 minutes/day Control group Number randomised: 50; losses: 4 (reasons for losses not reported by group, see below); analysed: at 4 weeks ‐ 47; at 3 months ‐ 46 Treatment arrangement: same as treatment for 'all groups' Overall reasons for losses: declined assessment, unable to contact, partial data only
Outcomes	Timing of assessments: 4 weeks and 3 months after randomisation End of treatment: 4‐week data used End of follow‐up: 3‐month data used Included in this review: activity limitation questionnaire, activity limitation test, participant satisfaction, pain, adverse events Other outcomes measured in the study: return to work, return to sports and leisure activities, perceived effect of treatment, duration of physiotherapy, ankle dorsiflexion range of motion
Notes	Study dates: not reported Funding sources: supported by University of Sydney, and National Health and Medical Research Council, Australia Declarations of interest: "no commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organisation with which the author(s) is/are associated" Notes: We received additional information from study author, A Moseley (regarding method of randomisation, adverse events, additional data). We pooled data from the long‐duration stretch group and the short‐duration stretch group as a single intervention group for data analysis.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by computer‐generated random sequence.
Allocation concealment (selection bias)	Low risk	Quote: "the randomization was concealed by using consecutively numbered, sealed, opaque envelopes."
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "All measurements were made by assessors who were blind to group allocation." Comment: assessor blinding implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Whilst losses were reasonably balanced between groups, loss was high for a 3‐month follow‐up and reasons for loss were not provided.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess the risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Moseley 2015.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 214 Source population: multicentre, Sydney, Australia: Royal North Shore Hospital, Royal Prince Alfred Hospital, Blacktown Mount Druitt Hospital, Prince of Wales Hospital, Ryde Hospital, Mona Vale Hospital, and Hornsby Ku‐ring‐gai Hospital. Although participants were screened at Hornsby Ku‐ring‐gai Hospital, no participants were recruited from this site. Inclusion criteria: isolated ankle fracture treated with immobilisation (with or without surgical fixation), immobilisation removed on the day of recruitment, approval received from the orthopaedic surgeon to bear weight as tolerated or bear partial weight, reduced ankle dorsiflexion range of motion (≥ 30 mm less motion compared with the non‐fractured leg, using the weight‐bearing lunge method), ankle pain ≥ 2 of 10 when 50% of body weight was borne through the affected leg, completed skeletal growth (no evidence of epiphyseal cartilage in the tibia in radiographs taken for fracture management), no concurrent pathologies (e.g. symptomatic osteoarthritis, stroke, other fractures) that would affect the ability to perform everyday tasks or the measurement procedures used in this trial; written informed consent Exclusion criteria: not specified Baseline characteristics for intervention group Age, mean (SD): 43.1 (± 16.5) years Sex (m/f): 43/63 Surgery: ORIF, n: 47 Fracture severity: less severe (1 malleoli fracture): 64; more severe (2 or 3 malleoli fractures or presence of dislocation regardless of number of malleoli fractures): 42 Baseline characteristics for control group Age, mean (SD): 41.3 (± 15.2) years Sex (m/f): 51/57 Surgery: ORIF, n: 51 Fracture severity: less severe: 60; more severe: 48
Interventions	Timing of randomisation: immediately after removal of immobilisation Intervention group Number randomised: 106 (note that 15 received out‐of‐trial physical therapy); losses: 22 (15 lost contact, 7 withdrew); number analysed at final time point: 84 Treatment management: participants in the rehabilitation group received the same advice in the fracture clinic and also participated in a supervised exercise program individually tailored, prescribed, monitored, and progressed by a physical therapist in the hospital outpatient physical therapy service. These participants were encouraged to perform the exercises at home. Three types of exercises were prescribed in the supervised exercise program: (1) ankle mobility and strengthening exercises, (2) stepping exercises, and (3) weight‐bearing and balancing on the affected leg. Exercise cards were provided to participants. Participants also received gait training and advice about returning to usual work and leisure activities. Control group Number randomised: 108 (note that 39 received out‐of‐trial physical therapy); losses: 22 (18 lost contact, 4 withdrew); number analysed at final time point: 86 Treatment management: participants in the advice group were provided with a single session of self‐management advice about exercise and return to activity. The advice was given by a physical therapist in the fracture clinic after removal of immobilisation. Exercise involved ankle movement in non‐weight‐bearing positions. Participants were given a handout summarising this advice with text and illustrations.
Outcomes	Timing of assessments: 1, 3, and 6 months End of treatment: 1 month End of follow‐up: 6 months Included in this review: activity limitation, QoL, participant satisfaction, impairment measures of pain, adverse events, economic data Other outcomes measured in the study: number of days to pain‐free walking, the number of days to return to full pre‐fracture work, return to pre‐fracture work and leisure activities, pain, walking speed, physical activity, ankle dorsiflexion range of motion
Notes	Study dates: December 2010 to December 2013 Funding sources: National Health and Medical Research Council, Australia Declarations of interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was concealed using a central telephone‐based randomisation service provided by the Australian National Health and Medical Research Council Clinical Trials Centre.
Allocation concealment (selection bias)	Low risk	Randomisation was concealed using a central telephone‐based randomisation service provided by the Australian National Health and Medical Research Council Clinical Trials Centre.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Primary outcomes (activity limitation and quality of life) were self‐reported by unblinded participants.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Attempts made, where possible, to blind outcome assessors to some outcomes by using different physical therapists than those who had conducted the therapy.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Study authors used intention‐to‐treat analysis when reporting findings for primary analysis. We note, however, that losses ranged from approximately 10% at 1 month to 20% at 6 months. However, losses were balanced between groups and explained in the report.
Selective reporting (reporting bias)	Low risk	Prospectively published protocol and clinical trials registration. Outcomes reported in the final study report were consistent with those in the protocol documents.
Other bias	Low risk	We identified no other sources of bias.

Nilsson 2009.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 110 Source population: single centre; Department of Orthopaedics, University Hospital, Lund, Sweden Inclusion criteria: surgically treated ankle fracture, between 18 and 64 years old, living within an area of 50 kilometres from the hospital, proficient in the Swedish language, no co‐existing fracture on the other or the same leg, no psychiatric diagnosis or history of drug abuse, no symptomatic osteoarthritis in the lower extremity, no rheumatic or other systemic diseases, no delayed surgery due to complications Exclusion criteria: disease that might influence the physical function of the lower extremity or adherence to the randomisation process Baseline characteristics for intervention group Age, mean (SD): males ‐ 34 (± 22) years; females ‐ 51 (± 22) years Sex (m/f): 19/31 Surgery: all had surgery Fracture severity, uni‐/bi‐/trimalleolar: 28/19/3 Baseline characteristics for control group Age, mean (SD): males ‐ 32 (± 26) years, females ‐ 51 (± 19) years Sex (m/f): 24/31 Surgery: all had surgery Fracture severity, uni‐/bi‐/trimalleolar: 36/18/1
Interventions	Timing of group allocation: randomised the day after surgery, treatment started after plaster removal Intervention group Number randomised: 52; losses: 4 (2 ‐ another disease, 2 ‐ dropped out); analysed: at 6 months ‐ 50; at 12 months ‐ 48 Treatment management: 2 physiotherapy sessions a week for 12 weeks, plus daily home exercises. The programme was based on neuromuscular principles, standardised and progressed, including range of motion, strength and balance, and stretching, weight‐bearing and walking exercises Control group Number randomised: 58; losses: 10 (dropped out); analysed: at 6 months ‐ 55; at 12 months ‐ 48 Treatment management: usual care, consisting of instruction from the physician to start walking and return to normal function as soon as possible. A referral to physiotherapists was in some cases given, based on the physicians' judgement, and participants were free to seek physiotherapy if they chose.
Outcomes	Timing of assessments: 6 and 12 months End of treatment: nil End of follow‐up: 12‐month data used Included in this review: activity limitation questionnaire, activity limitation test, quality of life, adverse events Other outcomes measured in the study: ankle dorsi‐ and plantarflexion range of motion, strength
Notes	Study dates: January 2003 to October 2005 Funding sources: Faculty of Medicine at Lund University and the Skane County Council, Primary Healthcare Research Declarations of interest: none Note: we received additional information from study author, G Nilsson (regarding method of randomisation, dropouts, adverse events)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by shuffling envelopes
Allocation concealment (selection bias)	Unclear risk	Additional information obtained from author: by sealed envelopes without codes on the outside Comment: insufficient information on numbering and opacity of envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "all subjects were examined by the same physiotherapist who was blinded to the allocation group." Comment: assessor blinding implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Loss of 12% participants, with more losses in the control group
Selective reporting (reporting bias)	High risk	The published protocol doesn't state the SF‐36 will only be partly used. Also, a few of the reported outcomes weren't pre‐specified in the published protocol.
Other bias	Low risk	We identified no other sources of bias.

Palke 2022.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 53 Source population: multicentre; 3 x level 1 trauma centres, Leipzig, Germany Inclusion criteria: 18 to 65 years and had undergone surgery because of an isolated closed tibial plateau or ankle fracture and received postoperative rehabilitation with partial weight‐bearing of 20 kg for 6 weeks after surgical fixation Exclusion criteria: weight > 100 kg, serious illness or poor general health that may influence rehabilitation, as judged by a physician, open fractures (> 1° according to Gustilo and Anderson), severe injury, alcohol or drug addiction, pregnancy, neuromuscular disorder, or pre‐existing muscle atrophy Baseline characteristics for intervention group Age, mean (SD): 43 (± 12) years Sex (m/f): 12/12 Surgery: all had surgery Fracture severity, AO Type B/C: 11/13 Baseline characteristics for control group Age, mean (SD): 40 (± 13) years Sex (m/f): 10/11 Surgery: all had surgery Fracture severity, AO Type B/C: 12/9
Interventions	Timing of randomisation: after surgery Treatment management for both groups: during the hospital stay, all participants received the same therapy, including manual lymphatic drainage, cryotherapy, and physiotherapy with mobilisation under partial weight‐bearing using crutches. The participants were allowed to move their affected joint freely while maintaining partial weight‐bearing. Intervention group Number randomised: 27; losses: 5 (3 did not receive intervention, 1 discontinued treatment, 1 lost to follow‐up); number analysed at final time point: 22 Treatment management: manual lymphatic drainage, cryotherapy, and exercise for 20 minutes on an antigravity treadmill 2 to 3 times per week, according to a predefined schedule, for 6 weeks Control group Number randomised: 26; losses: 7 (5 did not receive intervention, 1 discontinued treatment, 1 lost to follow‐up); number analysed at final time point: 19 Treatment management: standard physiotherapy, including manual lymphatic drainage, cryotherapy, and 20‐minute physiotherapy sessions 2 to 3 times per week for 6 weeks. Physiotherapy included passive movement and mobilisation of the operated joint and adjacent joints and gait training on crutches. Cryotherapy and lymphatic drainage were provided for 20 minutes until the swelling of the soft tissue in the affected area had subsided.
Outcomes	Timing of assessments: 6 weeks, 12 weeks, 12 months End of treatment: 6 weeks End of follow‐up: 12 months after operation Included in this review: activity limitation Other outcomes measured in the study: muscular atrophy of the thigh and lower leg
Notes	Study dates: August 2016 to June 2018 Funding sources: supported by German Social Accident Insurance and the Ambulante Reha Centren, Germany Declarations of interest: none Note: study also includes tibial plateau fractures, but we have not included any data for these fracture types.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation
Allocation concealment (selection bias)	Low risk	Quote: "The patients were assigned to the respective group by the Centre for Clinical Trials through blinded faxes sent to each participating institution."
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Individuals assessing outcomes (participants) were aware of their allocation to either intervention or control group.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Individuals assessing outcomes (personnel) were aware of their allocation to either intervention or control group.
Incomplete outcome data (attrition bias) All outcomes	High risk	A relatively high number of losses (10 losses overall from 53 randomised participants). Reasons for losses are explained, and include those who did not receive the allocated treatment.
Selective reporting (reporting bias)	Low risk	Prospective clinical trials registration (NCT02790229). Most reported outcomes are consistent with those in the clinical trials register. However, we note that data from the SF‐36 are not reported. Because quality of life is measured by the other measurement tools used in this study, we did not judge this to be a high risk of selective reporting bias.
Other bias	Low risk	We identified no other sources of bias.

Park 2021.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 194 Source population: multicentre, 6 university hospitals, Korea Inclusion criteria: 18 to 65 years and unstable ankle fractures, including unimalleolar fractures and bimalleolar fractures, that were stabilised after satisfactory reduction and firm internal fixation Exclusion criteria: ankle fractures whose stability after ORIF was questioned by the operating surgeon, fractures that required syndesmotic screw fixation, and cases with residual widening of the medial clear space or syndesmotic clear space after ORIF. Trimalleolar fractures, open fractures, pathologic fractures, and concurrent injuries that precluded following the rehabilitation protocol were also criteria for exclusion. People with obesity (BMI > 30; weight > 100 kg), diabetes or neuroarthropathy, and other conditions that would hinder following the study protocol were excluded. Baseline characteristics for intervention group Age, mean (SD): 42.7 (± 14.2) years Sex (m/f): 54/41 Surgery: all had surgery Fracture severity, lateral‐/medial‐/bi‐malleolus: 48/16/31 Baseline characteristics for control group Age, mean (SD): 43.1 (± 14.2) years Sex (m/f): 60/39 Surgery: all had surgery Fracture severity, lateral‐/medial‐/bi‐malleolus: 54/11/34
Interventions	Timing of randomisation: at 2 weeks postoperative visit Treatment management for both groups: all participants underwent ORIF. Then postoperative immobilisation with posterior short‐leg splint and maintained in non‐weight‐bearing state (with crutches) for 2 weeks. At 6 weeks, the cast or splint was removed from both groups and more active exercise was recommended. Intervention group Number randomised: 95; overall losses: 12 (did not turn up for final follow‐up appointment); number analysed: 95 (ITT analysis used) Treatment management: EWB. At 2 weeks postoperatively, fitted with a removable walking cast and allowed to bear weight as tolerated for the next 4 weeks. Daily active and passive ROM exercises encouraged Control group Number randomised: 99; losses: 15 (did not turn up for final follow‐up appointment); number analysed: 99 (ITT analysis used) Treatment management: NWB, maintained in a non‐weight‐bearing state with removable splint with crutches until 6 weeks postoperatively. ROM started 2 weeks postoperatively, similar to EWB group
Outcomes	Timing of assessments: 6, 8, 12 weeks, and 12 months End of treatment: 6 weeks End of follow‐up: 12 months Included in this review: activity limitation, participant satisfaction, adverse events Other outcomes measured in the study: time to return to pre‐injury activities, loss of reduction, implant failure
Notes	Study dates: 2014 to 2017 Funding sources: not reported Declarations of interest: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation
Allocation concealment (selection bias)	Low risk	Quote: "A biostatistician who was not involved in the clinical care of the patients prepared sealed envelopes using a computerized random number generator. The group allocation was concealed from the investigators, and the consecutively numbered sealed envelopes were opened at the 2‐week postoperative visit to allocate patients into either group".
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to treatment
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Attempts were made to conceal group allocation from those assessing other outcome measures.
Incomplete outcome data (attrition bias) All outcomes	High risk	Large number of losses at 12 months. More losses in the control group and reasons for these losses were not explained.
Selective reporting (reporting bias)	Low risk	Prospectively registered with clinical trials register (NCT02029170). Reported outcomes were consistent with outcomes in trial register.
Other bias	Low risk	We identified no other sources of bias.

Rasmussen 2006.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 40 Source population: single centre, Hvidovre University Hospital, Denmark Inclusion criteria: bimalleolar ankle fractures, Lauge‐Hansen supination eversion type IV or AO classification type B2.2 injury, surgical fixation required, 18 to 55 years of age, no rupture of the tibiofibular syndesmosis, in full working capacity Baseline characteristics for intervention group Age, mean (SD): 35.3 (± 12.1) years Sex (m/f): 8/12 Surgery: all had surgery Fracture severity: all had bimalleolar fracture Baseline characteristics for control group Age, mean (SD): 37.5 (± 9.5) years Sex (m/f): 10/10 Surgery: all had surgery Fracture severity: all had bimalleolar fracture
Interventions	Timing of randomisation: immediately after surgery Treatment management for both groups: full weight‐bearing with 2 crutches Intervention group Number randomised: 20; losses: 0; number analysed: 20 Treatment management: below‐knee plaster cast for 1 day after surgery, then pneumatic brace for 6 weeks, early active range of motion exercises, compression from pneumatic brace Control group Number randomised: 20; losses: 1 (reason unknown); number analysed: 19 Treatment management: below‐knee plaster cast for 6 weeks
Outcomes	Timing of assessments: 6, 8 and 12 weeks, and 24 months after randomisation End of treatment: 6‐week data used End of follow‐up: 24‐month data used Included in this review: activity limitation questionnaire, pain Other outcomes measured in the study: days of hospitalisation, radiographs, return to work, days of using crutches, ankle dorsiflexion and plantarflexion range of motion
Notes	Study dates: November 1997 to February 1999 Funding sources: not reported Declarations of interest: not reported Note: additional information received from S Rasmussen (method of allocation, assessor blinding, adverse events, dropouts, details of intervention, additional data)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The patients were allocated to the two groups according to a computer generated randomized list."
Allocation concealment (selection bias)	Low risk	Quote: "concealed in a set of numbered envelopes." Comment: insufficient information on numbering or opacity of envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and therapists were not blinded.
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Additional information obtained from author: assessor was blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study reported only in an abstract, with other information supplied after contact with study authors, and is not published in a peer‐reviewed journal.

Reed 1998.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions
Participants	Total number of randomised participants: 55 Source population: single centre; district general hospital, UK Inclusion criteria: ankle fractures, surgical fixation required, stable surgical fixation, adults Exclusion criteria: children, unsafe to leave out of backslab because of unstable fixation Baseline characteristics for intervention group Age, mean (SD): 41.3 (± 18.3) years Sex: not reported Surgery: all had surgery Fracture severity, Weber A/B/C: 2/15/10 Baseline characteristics for control group Age, mean (SD): 40.3 (± 14.0) years Sex: not reported Surgery: not reported Fracture severity, Weber A/B/C: 3/21/4
Interventions	Timing of randomisation: immediate after surgery Treatment management for both groups: physiotherapy after removal of immobilisation Intervention group Number randomised: 27; losses: 9 (reasons not reported) Treatment management: wool and crêpe bandage for 1 to 2 days Control group Number randomised: 28; losses: 13 (reasons not reported) Treatment management: below‐knee backslab for 1 to 2 days
Outcomes	Timing of assessments: 1 to 2 days after randomisation End of treatment: 1 to 2‐day data used End of follow‐up: nil Included in this review: pain Other outcomes measured in the study: ankle dorsiflexion range of motion, days of hospitalisation
Notes	Study dates: October 1995 to February 1997 Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author M Reed (regarding methods of randomisation and allocation, assessor blinding, details of assessment and intervention, additional data)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by shuffling envelopes
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by sealed, opaque, and consecutively numbered envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Assessor blinding implemented
Incomplete outcome data (attrition bias) All outcomes	High risk	Large participant loss which is not explained
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study reported only in an abstract, with additional information supplied by study author, and is not published in a peer‐reviewed journal.

Romero Zepeda 2008.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions
Participants	Total number of randomised participants: 50 Source population: single centre; emergency department of a general hospital, Mexico Inclusion criteria: ankle fracture (confirmed through x‐ray), within 12 hours of fracture, both genders, > 15 years of age, participants had to be able to rest for 24 hours after the bandage was applied Exclusion criteria: dislocation, associated fracture of the other leg, open fracture, multi‐trauma, vascular or neurological diseases of the lower limb, compartment syndrome Baseline characteristics for intervention group Sex (m/f): 22/4 Surgery: not clearly reported. For the purpose of this review, we have assumed that all participants were intended for surgical management, given that the intervention was an immediate intervention to reduce swelling. However, the study does not report how many participants subsequently underwent surgery. Fracture severity, Weber A/B/C/missing: 2/18/6/0 Baseline characteristics for control group Sex (m/f): 14/10 Surgery: not clearly reported (see above) Fracture severity, Weber A/B/C/missing: 0/14/10/0 Overall baseline characteristics Age, mean (SD): 39.6 (± 15.39) years
Interventions	Timing of randomisation: within 12 hours after fracture Treatment management for both groups: resting with the lower limb at body level for 24 hours. Metamizol (15 mg/kg/dosis) every 6 hours or ketorolac (30 mg fixed dosage) every 8 hours in case of allergies. Intervention group Number randomised: 26; no losses Treatment management: circular (Robert Jones) bandage for 24 hours Control group Number randomised: 24; no losses Treatment management: backslab for 24 hours
Outcomes	Timing of assessments: baseline, 24 hours after randomisation End of treatment: 24‐hour data used End of follow‐up: nil Included in this review: pain Other outcomes measured in the study: swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Notes: Published in Spanish (with English abstract). Study eligibility assessed by Paula Beckenkamp and Daniel Steffens; data extraction performed by Paula Beckenkamp and Rafael Zambelli Pinto. We received additional information from study author, E Romero Zepeda (regarding method of randomisation and allocation, surgery, dropouts, adverse events)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by drawing of lots
Allocation concealment (selection bias)	High risk	Additional information obtained from author: using an open random list of numbers
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Assessor blinding implemented (an independent assessor measured swelling)
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Schubert 2020.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 50 Source population: single centre, level I trauma centre, Australia Inclusion criteria: unimalleolar, bimalleolar, or trimalleolar ankle fractures with instability and/or dislocation requiring surgical fixation Exclusion criteria: open fractures, significant osteoporotic fractures, syndesmotic injuries requiring syndesmotic fixation, skeletally immature, nonambulatory status before injury, inability to comply with postoperative protocol, or polytrauma Baseline characteristics for intervention group Age, mean (SD): 46 (± 14) years Sex (m/f): 13/12 Surgery: all had surgery Fracture severity, uni‐/bi‐/tri‐malleolar/AO‐44B1/44B2/44b3/other: 11/10/4/7/7/7/4 Baseline characteristics for control group Age, mean (SD): 42 (± 16) years Sex (m/f): 16/9 Surgery: all had surgery Fracture severity, uni‐/bi‐/tri‐malleolar/AO‐44B1/44B2/44b3/other: 16/7/2/10/10/5/0
Interventions	Timing of randomisation: 2 weeks postoperative Treatment management for both groups: all participants were treated with ORIF. Then treated with below‐knee posterior splint and kept non‐weight‐bearing with crutches for the first 2 weeks until wound healing. The stitches were removed and participants placed in a controlled ankle motion (CAM) walking boot. Daily active and passive ROM exercises of ankle and subtalar joints with the brace were encouraged in both groups. Intervention group Number randomised: 25; losses: 1 (lost to follow‐up); number analysed at 6 months: 24 Treatment management: weight‐bearing in CAM walking boot. Allowed to bear weight as tolerated within the CAM boot from 2 weeks postoperatively Control group Number randomised: 25; losses: 4 (lost to follow‐up); number analysed at 6 months: 21 Treatment management: non‐weight‐bearing in CAM walking boot. Not allowed to bear weight for 6 weeks. Allowed ROM exercises 3 times a day out of their CAM boot.
Outcomes	Timing of assessments: 2, 6, 12, 26 weeks postoperatively End of treatment: 6 weeks End of follow‐up: 26 weeks postoperatively Included in this review: activity limitation, QoL, participant satisfaction, impairment measures of pain, adverse events Other outcomes measured in the study: none
Notes	Study dates: 2012 to 2016 Funding sources: none Declarations of interest: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Use of random number generator
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "follow‐up examinations and rating scores were completed by 1 of 2 orthopedic nurse practitioners not involved in patients’ care"
Incomplete outcome data (attrition bias) All outcomes	Low risk	Small number of losses which were clearly reported
Selective reporting (reporting bias)	Unclear risk	Protocol or clinical trials registration not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Seiger 2009.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: other rehabilitation interventions (electrotherapy)
Participants	Total number of randomised participants: 14 Source population: single centre; local orthopaedic surgeons' offices, Provo, Utah, USA Inclusion criteria: ankle fracture, 6 to 12 weeks postoperative, wearing a walking boot and allowed to periodically remove the boot, able to bear weight as tolerated, implanted orthopaedic metal at the fracture site, active dorsiflexion ROM < 10 degrees Exclusion criteria: cardiac pacemaker, neuro‐stimulator or fine wire implant, neurological or metabolic disorder, impaired mental cognition, severe osteoporosis, non‐union of the fracture site, altered or absent temperature sensation in the lower limb, receiving concurrent physical therapy, chiropractic or osteopathic treatment for the fractured ankle Baseline characteristics for intervention group Age, mean (SD): 45.7 (± 10.6) years Sex (m/f): 2/5 Surgery: all had surgery Fracture severity, Weber A/B/C/missing: 0/2/2/3 Baseline characteristics for control group Age, mean (SD): 37.0 (± 14.4) years Sex (m/f): 2/5 Surgery: all had surgery Fracture severity, Weber A/B/C/missing: 0/1/2/4
Interventions	Timing of randomisation: 6 to 12 weeks after surgical fixation, after participants were allowed to weight‐bear as tolerated. Treatment management for both groups: ankle and foot joint mobilisations, cold gel pack, home exercises twice a day Intervention group Number randomised: 7; losses: 1 (stopped treatment for unknown reasons) Treatment management: thermal pulsed short‐wave diathermy (27.12 MHz; 400 μsec; 800 pulses per second (pps); 48 W) 20 minutes twice a week for 12 sessions Control group Number randomised: 7 Treatment management: non‐thermal pulsed short‐wave diathermy (27.12 MHz; 20 μsec; 50 pps) 20 minutes twice a week for a total of 12 sessions
Outcomes	Timing of assessments: 2 to 3 times during each of the 12 treatment sessions End of treatment: 6‐week data used End of follow‐up: nil Included in this review: pain, adverse events Other outcomes measured in the study: ankle dorsiflexion and plantarflexion range of motion, swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author, C Seiger (regarding method of randomisation, source population)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Participants were paired according to their entry date. The first of the pair was randomly assigned to groups by drawing a number out of a hat. The second in each pair was assigned to the other group. Comment: randomisation sequence generated by a rule based on date of entry.
Allocation concealment (selection bias)	High risk	Allocation of the second participant in each pair was by alternation.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "The same therapist obtained all measurements. This therapist and all subjects were blinded to group assignment."
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	Low risk	Quote: "The same therapist obtained all measurements. This therapist and all subjects were blinded to group assignment."
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "The same therapist obtained all measurements. This therapist and all subjects were blinded to group assignment."
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "One subject in the treatment group completed 7 of 12 treatment sessions then stopped participation." Comment: data were not reported for this one participant at end of treatment and reason is not explained. However, loss is small
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents. We noted that one measurement (pain during shortwave diathermy) was not pre‐specified in the Methods.
Other bias	Low risk	We identified no other sources of bias.

Siddique 2005.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 44 Source population: multicentre; 4 emergency departments within the catchment area of Merlin Park Regional Hospital, Ireland Inclusion criteria: isolated Weber B lateral malleolar or bimalleolar ankle fracture, no syndesmosis injury, surgery within 24 hours, age 16 to 60 years, no cognitive difficulties, available for follow‐up Exclusion criteria: not reported Baseline characteristics not reported by group All had surgery All had Weber B uni‐ or bimalleolar fractures No data reported for age and sex
Interventions	Timing of randomisation: on admission to the emergency department Treatment management for both groups: partial weight‐bearing at 4 weeks postoperatively, with gradual progression to full weight‐bearing Intervention group Number randomised: 22; no reported losses Treatment management: no immobilisation Control group Number randomised: 22; no reported losses Treatment management: below‐knee plaster cast for 6 weeks
Outcomes	Timing of assessments: 6 and 12 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 12‐week data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: participant satisfaction (data incomplete), pain (data incomplete), radiographs, ankle dorsiflexion and plantarflexion range of motion
Notes	Study dates: 15 January 2001 to 30 November 2001 Funding sources: not reported Declarations of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "patients admitted under the care of consultant A were included to the immobilized group. Patients admitted under the care of consultant B were included to the mobilized group." Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and therapists were not blinded.
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "An independent and blinded physiotherapist assessed all patients at 6 and 12 weeks post‐surgery." Comment: assessor was blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Smeeing 2020.

Study characteristics
Methods	RCT, parallel design, 3‐armed trial Review comparison: early weight‐bearing versus delayed weight‐bearing
Participants	Total number of randomised participants: 115 Source population: multicentre, conducted in 4 teaching hospitals in the Netherlands: Diakonessenhuis (level 2 trauma centre), St. Antonius Hospital (level 2 trauma centre), Haaglanden Medisch Centrum (level 1 trauma centre) and the Elisabeth‐Twee Steden hospital (level 1 trauma centre) Inclusion criteria: 18 to 65 years with a Lauge‐Hansen supination exorotation type 2, 3 and 4 ankle fractures, articular discongruity of > 2 mm on X‐ray Exclusion criteria: pre‐existent impaired morbidity, polytrauma, expected insufficient stable fracture fixation with standard surgical procedure, necessity for syndesmosis screw, tertius fragment requiring operative treatment, Gustillo 2 and 3 open fractures, cognitive disability, BMI > 30, and diabetes mellitus, inability to comply with non‐weight‐bearing mobilisation, inability to comply with follow‐up Baseline characteristics for intervention group (unprotected weight‐bearing) Age, mean (SD): 37.8 (± 15.1) years Sex (m/f): 22/20 Surgery: all had surgery Fracture severity, Lauge‐Hansen classification. Supination external rotation type 2/3/4: 14/0/28 Baseline characteristics for intervention group (unprotected non‐weight‐bearing) Age, mean (SD): 37.8 (± 13.7) years Sex (m/f): 23/14 Surgery: all had surgery Fracture severity, Lauge‐Hansen classification. Supination external rotation type 2/3/4: 14/2/21 Baseline characteristics for control group (protected weight‐bearing) Age, mean (SD): 41.5 (± 14.2) years Sex (m/f): 16/20 Surgery: all had surgery Fracture severity, Lauge‐Hansen classification. Supination external rotation type 2/3/4: 9/4/23
Interventions	Timing of randomisation: on the day of, and immediately prior to, surgery Treatment management for both groups: all participants consulted a physical therapist postoperatively or after cast removal to learn exercises and received advice on how to start mobilising according to their specific postoperative care regimen. Participants were treated on the same day as admission if possible. According to their postoperative care regimen, participants received a cast, exercise instructions, and a physiotherapist referral letter. Intervention group Number randomised: 42; losses: 0; number analysed: 42 (ITT) Treatment management: unprotected weight‐bearing, which included a pressure dressing in the first 24 hours postoperatively. After the first 24 hours, weight‐bearing was allowed as tolerated by the participant. Intervention group Number randomised: 37; losses: 0; number analysed: 37 (ITT) Treatment management: unprotected non‐weight‐bearing, which included a pressure dressing in the first 24 hours postoperatively. After 6 weeks, weight‐bearing was allowed. Control group Number randomised: 36; losses: 0; number analysed: 36 (ITT) Treatment management: protected weight‐bearing, which included a 10‐day backslab plaster splint. After 10 days, the participant received a weight‐bearing cast for the remaining first 6 weeks.
Outcomes	Timing of assessments: clinical follow‐up at 6 weeks, 12 weeks, 6 months, and 1 year. End of treatment: 6 weeks End of follow‐up: 1 year Included in this review: activity limitation, QoL, adverse events Other outcomes measured in the study: time to return to work, time to return to sports, number of complications
Notes	Study dates: January 2013 to October 2016 Funding sources: none Declarations of interest: none Note: 3 participants switched from protected weight‐bearing to the unprotected non‐weight‐bearing group because of cast irritation. ITT analysis used.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not possible to blind participants and personnel to treatment
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	No information is reported on whether outcome assessors were blinded to treatment allocation
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Low risk	Prospective clinical trials registration (NTR3727). Outcomes relevant to this review are included in the trial register documents. Whilst most outcomes are reported in the published report, we note that no data are reported for pain (using a VAS).
Other bias	Low risk	We identified no other sources of bias.

Sondenaa 1986.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 43 Source population: single centre; Norway Inclusion criteria: ankle fracture, surgical fixation required, satisfactory fracture reduction and osteosynthesis Exclusion criteria: compound or compression fracture Baseline characteristics for intervention group Age, median (range): 35 (16 to 59) years Sex (m/f): 9/11 Surgery: all had surgery Fracture severity, Weber A/B/C: 2/8/10 Baseline characteristics for control group Age, median (range): 37 (16 to 66) years Sex (m/f): 9/14 Surgery: all had surgery Fracture severity, Weber A/B/C: 0/15/8
Interventions	Timing of randomisation: 1 day after surgery Treatment management for both groups: full weight‐bearing after 6 weeks Intervention group Number randomised: 20; no losses Treatment management: below‐knee backslab for 3 days, active ankle exercises after removal of backslab, 12 appointments of physiotherapy in the first 6 weeks Control group Number randomised: 23; no losses Treatment management: below‐knee plaster cast for 6 weeks, non‐weight‐bearing, 12 appointments of physiotherapy after 6 weeks
Outcomes	Timing of assessments: 6, 12, 18, and 52 weeks after randomisation End of treatment: 6‐week data used End of follow‐up: 52‐week data used Included in this review: pain, adverse events Other outcomes measured in the study: strength (measured but not reported), swelling (subjective data), radiographs, ankle dorsiflexion and plantarflexion range of motion
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author, K Sondenaa (regarding method of randomisation, assessor blinding, details of assessment and intervention, adverse events, comorbidities)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by mixing envelopes
Allocation concealment (selection bias)	Unclear risk	Additional information obtained from author: sealed envelopes Comment: insufficient information on numbering or opacity of envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Stockle 2000.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 40 Source population: single centre; Germany Inclusion criteria: isolated and first incidence of ankle fracture, surgical fixation required, age 18 to 65 years Exclusion criteria: not reported Baseline characteristics for intervention group Fracture severity, Weber B/C:15/5 Baseline characteristics for control group Fracture severity, Weber B/C: 12/8 Overall baseline characteristics Age, mean (range): 45 (20 to 65) Sex (m/f): 21/19 Surgery: all had surgery
Interventions	Timing of randomisation: after surgery Treatment management for both groups: partial weight‐bearing with crutches, heparin Intervention group Number randomised: 20; no reported losses Treatment management: immobilised in VACOped (removable immobilisation in a vacuum orthosis) for 6 weeks, which could be removed for wound check, hygiene, and intermittent physiotherapy. Details of physiotherapy not reported Control group Number randomised: 20; no reported losses Treatment management: below‐knee plaster cast for 6 weeks
Outcomes	Timing of assessments: 6 weeks and 3 months after randomisation End of treatment: 6‐week data used End of follow‐up: 3‐month data used Included in this review: adverse events Other outcomes measured in the study: ankle dorsiflexion and plantarflexion range of motion, subtalar range of motion, calf circumference
Notes	Study dates: September 1996 to July 1997 Funding sources: not reported Declarations of interest: not reported Note: published in German. Study eligibility was assessed by the German Cochrane Centre. English translation was available and used as the primary source of data extraction. Data extraction from German article by Erwin Scherfer and Jutta Jablonski
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "the assignment took place randomly after order of admittance to the emergency ward." Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Stuart 1989.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after non‐surgical fracture management)
Participants	Total number of randomised participants: 40 Source population: single centre; Newcastle General Hospital Accident Department, UK Inclusion criteria: stable supination‐eversion stage II lateral malleolar fractures, adults Exclusion criteria: tenderness on the medial side of the ankle Baseline characteristics not reported by group Surgery: none had surgery Fracture severity: all had supervision‐eversion stage II Age and sex: not reported
Interventions	Timing of group allocation: immediately on attendance Treatment management for both groups: weight‐bearing as tolerated, leg elevation while resting Intervention group Number randomised: 20; losses: 1 (requested change to alternative treatment); analysed: 19 Treatment management: air‐stirrup for 4 weeks, then plaster cast for another 2 weeks if required Control group Number randomised: 20; losses: 2 (inadequate follow‐up); analysed: 18 Treatment management: below‐knee walking cast for 4 weeks, then plaster cast for another 2 weeks if required, non‐weight‐bearing for 2 days to allow cast to dry
Outcomes	Timing of assessments: 24 hours, 1, 4 weeks (and 6 weeks if required), 3 months after randomisation End of treatment: 4‐ and 6‐week data used End of follow‐up: nil Included in this review: adverse events Other outcomes measured in the study: ankle dorsiflexion and plantarflexion range of motion, participant satisfaction (not measured at end of treatment or follow‐up), swelling (not measured at end of treatment or follow‐up), time to union, radiographs
Notes	Study dates: not reported Funding sources: air‐stirrups supplied by Orthopaedic Systems of Widnes, Cheshire, UK Declarations of interest: not reported Note: we received additional information from study author, P Stuart (regarding method of randomisation, method of allocation, no participant had surgery)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "the patients were allocated to one of two groups using a table of random numbers."
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by sealed, opaque and consecutively numbered envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Small loss of 3 participants; loss balanced between groups and reasons for losses explained
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Tropp 1995.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 15 Source population: single centre; Sweden Inclusion criteria: Weber B or C ankle fracture, surgical fixation required, stable fracture fixation and satisfactory reduction, age 18 to 60 years Exclusion criteria: large posterior tibial fragments Baseline characteristics (overall) Age, mean (range): 26 (19 to 60) years Sex: not reported Surgery: all had surgery Fracture severity: Weber B or C Baseline characteristics for each group were not reported.
Interventions	Timing of randomisation: after surgery Treatment management for both groups: weight‐bearing Intervention group Number randomised: 15; no losses Treatment management: brace for 10 weeks, dorsiflexion and plantarflexion movements, strength and functional exercises starting immediately, crutches for at least the first 2 weeks Control group Number randomised: 15; no losses Treatment management: walking cast for 6 weeks, crutches used during immobilisation and at least 2 to 4 weeks after removal of cast, mobility, strength and functional exercises starting after 6 weeks
Outcomes	Timing of assessments: 10 weeks and 12 months after randomisation End of treatment: 10‐week data used End of follow‐up: 12‐month data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: calf circumference, radiographs, ankle dorsiflexion and plantarflexion range of motion, strength through range (data at 60 degrees/second used for analysis), swelling
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author, H Tropp (regarding methods of randomisation and allocation, assessor blinding, fracture severity, details of intervention, dropouts, adverse events)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by shuffling envelopes
Allocation concealment (selection bias)	Low risk	Additional information obtained from author: by sealed, opaque, and consecutively numbered envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Additional information obtained from author: assessor was blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Van den Berg 2018.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after non‐surgical fracture management)
Participants	Total number of randomised participants: 44 Source population: single centre, Level 1 centre, HMC Haaglanden, Netherlands Inclusion criteria: 18 to 70 years of age, unimalleolar fibular fracture at the level of the syndesmosis, a congruent joint and fracture of < 2 mm on both mortise and lateral ankle X‐rays Exclusion criteria: previous ankle fracture on the same side, multi‐trauma patients, follow‐up elsewhere, no proper understanding of the Dutch language, comorbidities that limit mobilisation, already existing fractures (> 2 weeks) Baseline characteristics for intervention group Age, mean (SD): 44.3 (± 14.9) years Sex (m/f): 13/10 Surgery: none Fracture severity: all unimalleolar Baseline characteristics for control group Age, mean (SD): 41.9 (± 16.3) years Sex (m/f): 11/10 Surgery: none Fracture severity: all unimalleolar
Interventions	Timing of randomisation: 1 week after initial treatment (on confirmation of a congruent joint with progression to dislocation) Treatment management for both groups: initial treatment with a splint. After application of cast/brace, both groups were allowed to weight‐bear as tolerated, using crutches as long as necessary. At the 6‐week review, cast/brace was removed dependent on radiographs. After the fracture was considered healed, both groups were encouraged to do stabilisation exercises and return to sport as tolerated. Referral to a physiotherapist was at the discretion of the participant and the treating surgeon. Intervention group Number randomised: 23; losses: 9 (death or did not respond to follow‐up request); number analysed at end of treatment: 19, and at 6 months: 14 Treatment management: removable brace (no additional details in the study report) Control group Number randomised: 21; losses 10 (death or did not respond to follow‐up request); number analysed at end of treatment: 20, at 6 months: 11 Treatment management: regular cast. This group were given thromboprophylaxis during the immobilisation period.
Outcomes	Timing of assessments: 3, 6, 12, 26, 52 weeks End of treatment: 6 weeks End of follow‐up: 52 weeks Included in this review: activity limitation, QoL, impairment measures of pain Other outcomes measured in the study: congruency of ankle joint, amount of dislocation of fibular fracture at 3 and 6 weeks, ankle dorsiflexion range of motion, plantarflexion range of motion, strength
Notes	Study dates: March 2013 to May 2015 Funding sources: none Declarations of interest: None Notes: Study ended early (after 26 months) because of low recruitment Data reported as median values. We contacted study authors by email for outcome data as mean (SD)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method of sequence generation is not clearly reported
Allocation concealment (selection bias)	Low risk	After permission and written informed consent, participants were randomised into two groups using sealed opaque envelopes.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants and personnel not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Independent assessors were used, and were blinded to study treatment
Incomplete outcome data (attrition bias) All outcomes	High risk	Attrition for both groups was even and reasons for attrition are provided. However, loss at 12 months was more than 10%, and we could not be certain whether this impacted the outcome data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Van Laarhoven 1996.

Study characteristics
Methods	Quasi‐RCT, parallel design, 2‐armed trial Review comparisons: early weight‐bearing versus delayed weight‐bearing; removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 81 Source population: multicentre; 2 hospitals, Netherlands Inclusion criteria: ankle fracture, surgical fixation required, stable fracture after operation suitable for early mobilisation, able to cope with either regime Exclusion criteria: pilon fracture, open injury to physeal plate or grade II or III open fractures Baseline characteristics for intervention group Age, median (range): 35.5 (17 to 77) years Sex (m/f): 24/17 Surgery: all had surgery Fracture severity, uni‐/bi‐/trimalleolar:17/24 Baseline characteristics for control group Age, median (range): 37 (15 to 77) years Sex (m/f):21/19 Surgery: all had surgery Fracture severity, uni‐/bi‐ or trimalleolar: 16/24
Interventions	Timing of randomisation: 3 to 5 days after surgery Treatment management for both groups: below‐knee plaster cast and range of motion exercises until randomisation Intervention group Number randomised: 41 Treatment management: walking cast for 6 weeks, weight‐bearing, 9/41 participants received physiotherapy after 6 weeks Control group Number randomised: 40 Treatment management: no immobilisation, non‐weight‐bearing with crutches, 14/40 participants received physiotherapy after 6 weeks Study authors report that 2 participants were lost to follow‐up, but group allocation is not reported
Outcomes	Timing of assessments: 10 to 16 days, 6 weeks, 3 months and 1 year after randomisation End of treatment: 6‐week data used End of follow‐up: 1‐year data used Included in this review: activity limitation questionnaire, adverse events Other outcomes measured in the study: dropouts (data incomplete), radiographs, return to work, ankle dorsiflexion range of motion Note: we could not include some adverse event data (osteitis, secondary dislocation, delayed union, Sudeck's dystrophy, further surgery) because data for these were not reported by intervention group.
Notes	Study dates: April 1991 to April 1993 Funding sources: not reported Declarations of interest: "No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article" Notes: We received additional information from study author, C van Laarhoven (regarding comorbidities). We did not include data for activity limitation in the review as this was reported as median scores with distribution values.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "we randomized using an odd or even date of accident." Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	High risk	Quote: "it was not possible to undertake blind assessment of the treatment groups." Comment: assessor blinding possible but not implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although group allocation of lost participants is not reported or the reason for loss, the number of losses is few and unlikely to impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Venesmaa 2004.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 32 Source population: single centre; Kuopio University Hospital, Finland Inclusion criteria: low energy uni‐ or bi‐malleolar fracture Exclusion criteria: not reported Baseline characteristics for intervention group Age, median (range): 41 (20 to 63) years Sex (m/f): 8/10 Surgery: all had surgery Fracture severity: all had uni‐ or bimalleolar fractures Baseline characteristics for control group Age, median (range): 48 (19 to 69) years Sex (m/f): 6/8 Surgery: all had surgery Fracture severity: all had uni‐ or bimalleolar fractures
Interventions	Timing of randomisation: not reported Treatment management for both groups: not specified Intervention group: Number randomised:18; no reported losses Treatment management: Aircast for 6 weeks Control group Number randomised: 14; no reported losses Treatment management: plaster cast for 6 weeks
Outcomes	Timing of assessments: 9 and 26 weeks after randomisation End of treatment: nil End of follow‐up: 26‐week data used Included in this review: activity limitation questionnaire Other outcomes measured in the study: ankle dorsiflexion range of motion Note: we did not include complications data because the type of complication was not specified; we note, however, that "all fractures healed without complications"
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method of allocation not described
Allocation concealment (selection bias)	Unclear risk	Method of allocation not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	No apparent losses
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study published only as an abstract, not in a peer‐reviewed journal

Vioreanu 2007.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 66 Source population: single centre; Ireland Inclusion criteria: acute, closed ankle fracture, surgical fixation required, age 14 to 65 years, fibular displacement > 2 mm, non‐neuropathic joint Exclusion criteria: other severe injuries, pilon fracture, isolated medial malleolar fracture, diabetes mellitus, insufficiently stable fracture fixation Baseline characteristics for intervention group Age, mean (SD): 37.2 (± 12.9) years Sex (m/f): 21/10 Surgery: all had surgery Fracture severity, Weber B/C: 29/4 Baseline characteristics for control group Age, mean (SD): 34.9 (± 16.0) years Sex (m/f): 20/9 Surgery: all had surgery Fracture severity, Weber B/C: 21/8
Interventions	Timing of randomisation: after surgery Treatment management for both groups: prophylactic antibiotics pre‐operatively, dorsal splint for 10 to 14 days, non‐weight‐bearing with crutches, partial weight‐bearing and physiotherapy after 6 weeks, full weight‐bearing after 8 weeks Intervention group Number analysed: 33 (see below) Treatment management: removable fibreglass cast for 6 weeks, ankle range of motion exercise 3 times for 10 minutes/day Control group Number analysed: 29 (see below) Treatment management: fibreglass cast (non‐removable) for 6 weeks Note: number randomised to each group is not reported. Overall, there were 4 dropouts (1 moved abroad, 1 did not return to follow‐up appointments, 2 non‐compliant with weight‐bearing status)
Outcomes	Timing of assessments: 2, 6, 9, 12, 24 weeks after randomisation (activity limitation measured at 9 and 12 weeks only) End of treatment: 6‐week data used End of follow‐up: 12‐week data used as included activity limitation Included in this review: activity limitation questionnaire, quality of life (measured at 24 weeks only), adverse events Other: ankle dorsiflexion range of motion, swelling, radiographs, return to work, muscle atrophy, time to return to work and pre‐injury activity
Notes	Study dates: August 2004 to March 2005 Funding sources: not reported Declarations of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "patients were randomly assigned, using their odd or even day of date of birth." Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few losses, with reasons provided, although it is not clear to which group the losses were allocated
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

Wetzler 1991.

Study characteristics
Methods	Q‐RCT, parallel design, 2‐armed trial Review comparison: removable ankle support versus non‐removable ankle support (after surgical fracture management)
Participants	Total number of randomised participants: 45 Source population: single centre, Department of Orthopaedic Surgery, Boston City Hospital, USA Inclusion criteria: closed ankle fracture, surgical fixation required, adults Baseline characteristics for intervention group Surgery: all had surgery Baseline characteristics for control group Surgery: all had surgery
Interventions	Timing of randomisation: after surgery Treatment management for both groups: weight‐bearing Intervention group Number randomised: 20; losses: not reported; number analysed: unclear in abstract Treatment management: pneumatic walker for 1 to 2 weeks, then pneumatic ankle brace for an unspecified period Control group Number randomised: 25; losses: not reported; number analysed: unclear in abstract Treatment management: below‐knee plaster cast for 6 weeks
Outcomes	Timing of assessments: 1, 2, 4, 6, 12 weeks, 6 months and 1 year after randomisation End of treatment: 6‐week data used End of follow‐up: 1‐year data used Included in this review: activity limitation questionnaire, pain Other outcomes measured in the study: ankle dorsiflexion and plantarflexion range of motion
Notes	Study dates: April 1988 to November 1989 Funding sources: not reported Declarations of interest: not reported Note: additional information received from M Wetzler (method of allocation, weight‐bearing status)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Additional information obtained from author: allocation by birth dates Comment: quasi‐random procedure
Allocation concealment (selection bias)	High risk	Quasi‐random procedure
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participant and therapist blinding not possible
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participant blinding not possible
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Unclear risk	Insufficient information on whether assessor blinding was implemented
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Because of insufficient reporting in the abstract, we could not be certain of participant losses.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	High risk	Study published only as an abstract, and we assumed that this study was not peer‐reviewed.

Wilson 1991.

Study characteristics
Methods	RCT, parallel design, 2‐armed trial Review comparison: physical therapy interventions versus usual care or other physical therapy interventions
Participants	Total number of randomised participants: 12 Source population: single centre; Fracture Clinic at Rotorua Hospital, New Zealand Inclusion criteria: ankle fracture treated with or without surgery, age 16 to 65 years, immediate referral from fracture clinic to physiotherapy after cast removal (length of immobilisation = 6 weeks), able to comply with treatment protocol, full mobility of unaffected ankle Exclusion criteria: pre‐existing joint pathology or severe leg injury/pathology Baseline characteristics for intervention group Age, mean (SD): 47.2 (± 16.3) years Sex (m/f): 4/1 Surgery (Y/N): 1/4 Fracture severity, Weber B/C: 4/1 Baseline characteristics for control group Age, mean (SD): 40.6 (± 15.4) years Sex (m/f): 3/2 Surgery (Y/N): 2/3 Fracture severity, Weber B/C: 4/1
Interventions	Timing of randomisation: immediately after cast removal Treatment management for both groups: whirlpool exercises, individual exercise programme, home exercises Intervention group Number randomised: 7; losses: 2 (did not complete treatment because fit to return to work at 3 weeks) Treatment management: 3 physiotherapy sessions/week for 5 weeks of Kaltenborn‐based manual therapy to the talocrural and talocalcaneal joints, plus other lower limb joints assessed to be hypomobile by the physiotherapist Control group Number randomised: 5 Treatment management: 3 physiotherapy sessions/week for 5 weeks
Outcomes	Timing of assessments: 5 weeks after randomisation End of treatment: 5‐week data used End of follow‐up: nil Included in this review: activity limitation questionnaire Other: ankle dorsiflexion and plantarflexion range of motion
Notes	Study dates: not reported Funding sources: not reported Declarations of interest: not reported Note: we received additional information from study author, F Wilson (regarding methods of randomisation and allocation)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Additional information obtained from author: by shuffling envelopes
Allocation concealment (selection bias)	Unclear risk	Additional information obtained from author: by sealed opaque envelopes selected by receptionist/physiotherapy assistant Comment: insufficient information on numbering of envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and therapists were not blinded.
Blinding of outcome assessment (detection bias) Participant‐assessed outcomes	High risk	Participants were not blinded.
Blinding of outcome assessment (detection bias) Personnel‐assessed outcomes	Low risk	Quote: "A 'blind' assessor recorded measurements."
Incomplete outcome data (attrition bias) All outcomes	High risk	Both participants were lost from the intervention group. Because the study is so small, this could impact the data.
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or protocol not reported. It is not feasible to effectively assess risk of selective reporting bias without these documents.
Other bias	Low risk	We identified no other sources of bias.

ACM: active controlled motion; AO: Arbeitsgemeinschaft für Osteosynthesefragen (German for "Association for the Study of Internal Fixation" (ASIF); system for classification of fractures); BMI: body mass index; CAM: controlled ankle motion; COA: Canadian Orthopaedic Association; COTS: Canadian Orthopaedic Trauma Association; CSP: The Chartered Society of Physiotherapy; CT: computerised tomography; EQ‐5D(‐5L): EuroQol 5‐dimensions; EWB: early weight‐bearing; HRQoL: health‐related quality of life; ITT: intention to treat; IV: intravenous(ly); m/f: male/female; n: number of participants; NHS: National Health Service; NIH: National Institutes of Health; NIHR: National Institute for Health and Care Research; NWB: non‐weight‐bearing; OMAS: Olerud and Molander Ankle Score; ORIF: open reduction internal fixation; OTA: Orthopaedic Trauma Association; POD1: first postoperative day; PSI: Physician Services Incorporation; PT: physiotherapy; QoL: quality of life; RCT/Q‐RCT: randomised controlled trial/quasi‐randomised controlled trial; ROM: range of motion; SD: standard deviation; SE: standard error; SF‐36: Short Form Health Survey (36‐item); VAS: visual analogue scale; WB: weight‐bearing; Y/N: yes/no

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
ACTRN12610000557033	Clinical trials registration for a randomised controlled trial about early mobilisation after ankle fracture fixation. The previous version of this review included this study as an ongoing study. Because the clinical trials register has not been updated since 2010 and ethics approval has still not been submitted, we have excluded this study because we assumed that the study was not undertaken.
ACTRN12611001096943	Randomised controlled trial comparing functional outcomes in participants performing weightbearing at 2 weeks versus 6 weeks, following surgical fixation of ankle fracture. We identified this potentially ongoing study in our search of clinical trials registers. We excluded the study because there have been no updates on trial progress since 2011. We attempted email contact with the contact person, but the address was no longer valid.
ACTRN12613000631707	Randomised controlled trial comparing protected early postoperative range of motion exercises versus cast immobilisation in ankle fractures requiring surgery. We identified this potentially ongoing study in our search of clinical trials registers. We excluded the study because there have been no updates on trial progress since 2013. We attempted email contact with the contact person, but the address was no longer valid.
Albin 2019	Randomised controlled trial of physical therapy interventions. We excluded this study because it included people with both ankle and hindfoot fractures, and did not report data separately according to fracture type.
Büker 2019	Comparison of physical therapy interventions. We excluded this study because allocation to treatment groups was not randomised or quasi‐randomised.
DRKS00021487	Randomised controlled trial comparing an orthosis (iWalk 2.0) with crutches. Described as completed in the clinical trials register. We contacted the study authors for further information about the study before results are published. We excluded this study because no participants had ankle fractures.
Gan 2014	Randomised controlled trial of electrotherapy interventions for lower limb stress fractures. We excluded this study because data were not reported separately for ankle fracture.
Güngör 2022	Participants had supervised exercise and early weight‐bearing after ankle fracture. We excluded this study because it was not randomised and involved only retrospective collection of data.
N0055190984	Randomised controlled trial of postoperative management of ankle fractures treated with open reduction and internal fixation (ORIF), six weeks in cast versus no cast and active early physiotherapy. The previous version of this review included this study as an ongoing study. We excluded it because, to our knowledge, it was not completed, and no results are available.
NCT01127776	Randomised controlled trial for APOS System effects in postoperative bi/trimalleolar ankle fractures (APOS System not defined in the clinical trials register). The previous version of this review included this study as an ongoing study. Because the clinical trials register has not been updated since 2012, we excluded this study because we assumed that it was not conducted.

Characteristics of studies awaiting classification [ordered by study ID]

ISRCTN15497399.

Methods	RCT, parallel design
Participants	Target number of participants: 246 Inclusion criteria: surgery for fixation of unstable ankle fractures; informed consent Exclusion criteria: < 16 years old (skeletally immature); poor skin condition at operation site; serious concomitant disease (e.g. stroke, osteoporosis, arthritis); diabetic neuropathy/other sensory neuropathy (lack of sensation); non‐ambulatory prior to injury; active leg ulceration; unable to understand the study information or unable to complete the outcome questionnaires; surgeon concerned about quality of fixation/integrity of wound; fracture requiring further stabilisation in/around the ankle (e.g. syndesmosis); open ankle fracture (bone broken through skin); participant in other concurrent interventional research which may over‐burden the participant or confound data collection; concomitant injuries which will have a confounding effect on rehabilitation in the opinion of the investigator Setting: UK
Interventions	Intervention: removable support boot with range of movement for 4 weeks Control: below‐knee plaster cast for 4 weeks
Outcomes	OMAS; ankle functional data (range of movement, weight‐bearing); quality of life (EQ‐5D), healing status, complications, return to usual activities
Notes	Study is also called 'ART' Study is completed. We have contacted the study authors who are unable to share summarised data until publication.

Khojaly 2021.

Methods	RCT, parallel design
Participants	Target number of participants: 140 Inclusion criteria: all skeletally mature, acute ankle fractures that were treated with anatomical reduction and stable internal fixation Exclusion criteria: skeletal immaturity; Grade‐III open fractures; tibial plafond fractures; polytrauma; non‐ambulatory status before injury; expected insufficient stable fracture fixation with standard surgical technique; pre‐existent cognitive disability, inability to comply with non‐weight‐bearing mobilisation and inability to comply with follow‐up Setting: Ireland
Interventions	Intervention: walking boot, instructed to walk weight‐bearing as tolerated immediately, also instructed to remove walking boot 4 times a day and perform ankle range of motion exercises until referred to outpatient physiotherapy Control: cast, instructed to walk non‐weight‐bearing using crutches/frame for 6 weeks
Outcomes	OMAS; range of motion; SF‐36; time to return to work; wound complications and complications related to fracture fixation and bone healing
Notes	According to information in the clinical trials register, this study is now complete. We attempted to contact the study authors for information about their trial and completed trial results, but we received no reply. We await publication of the full report.

NCT02609347.

Methods	RCT, parallel design
Participants	Number of randomised participants: 72 Inclusion criteria: 18 to 70 years of age, stiffness > 5 cm side to side difference as measured by the ankle lunge test, surgical management of ankle or hindfoot fracture, full weight‐bearing, able to read and speak sufficient English to complete outcome tools Exclusion criteria: any prior foot/ankle surgery, deformity, or injury on the unaffected side which would affect gait, weight‐bearing dorsiflexion or balance, unable to attend follow‐up visits, any cognitive impairments which would prevent participant from being able to complete or understand questionnaires, received any previous manual therapy for their current ankle and/or hindfoot fracture, known non/malunion, additional fractures of the spine, hip, or knee that would likely affect their weight‐bearing or gait, avascular necrosis as demonstrated on radiographs, syndesmotic screw intact, open reduction internal fixation >12 months previously Setting: USA
Interventions	Intervention: manual therapy with 3 treatment sessions of joint mobilisation based on clinical decision‐making Control group: sham manual therapy consisting of soft tissue mobilisation and Grade 1 mobilisations at proximal tibia/fibia joint
Outcomes	Range of motion, spatial gait parameters, balance, characteristics that predict outcomes, foot mobility, muscle stiffness, stance time, single limb stance time
Notes	According to information in the clinical trials register, this study is now complete. We attempted to contact the study authors for information about their trial and completed trial results, but we received no reply. We await publication of the full report.

NCT02779244.

Methods	RCT, parallel design
Participants	Expected number of participants: 132 Source population: hospital, USA Inclusion criteria: operative unstable ankle fractures; 18 to 70 years of age Exclusion criteria: requirement of syndesmotic fixation; pilon fractures; open fractures; people with diabetic neuropathy; BMI > 40; people with polytrauma; people who are unable to comply with non‐weight‐bearing; people who are unable to follow up
Interventions	Intervention group: early weight‐bearing plus CAM boot Control group: standard treatment CAM boot
Outcomes	OMAS
Notes	According to information in the clinical trials register, this study is now complete. We await publication of the full report.

NCT02923479.

Methods	RCT, parallel design
Participants	Number of randomised participants: 32 Inclusion criteria: functional limitation following work‐related ankle injuries; post‐immobilisation phase; time interval since fracture event < 12 months; informed consent Exclusion criteria: non‐compliance with study requirements, pregnancy or breastfeeding, current or prior history of malignancy, open skin at level of patient‐device interface, sensory deficit at the level of patient‐device interface, ankle motor deficit secondary to peroneal or tibial neuropathy, acute inflammatory arthritis of the ankle, other pathological conditions inducing lower limb pain or dysfunction Setting: Italy
Interventions	Intervention: general rehabilitation, and specific ankle rehabilitation using the ARBOT (robot) device Control: general rehabilitation, specific ankle rehabilitation by physiotherapist, specific ankle rehabilitation by Biodex System 3 dynamometer and ProKin PK254 platform
Outcomes	Dorsiflexion range of motion, isometric peak torque at 0° and 10° of plantar flexion, isometric plantar‐flexion torque, 2‐minute walk test, side effects, timed stair‐climbing test, proprioceptive tests, pain (VAS), LEFS, AOFAS, participant satisfaction (VAS)
Notes	According to information in the clinical trials register, this study is now complete. We attempted to contact the study authors for information about their trial and completed trial results, but we received no reply. We await publication of the full report.

NCT04235907.

Methods	RCT, parallel design
Participants	Expected number of participants: 60 Source population: hospital, USA Inclusion criteria: bimalleolar or trimalleolar fracture; repaired with open reduction internal fixation technique; must have access to computer, tablet, or smartphone with internet access Exclusion criteria: severe soft tissue damage associated with injury; physical or mental conditions that will affect the participant's ability to participate in independent rehabilitation; non‐English speaking
Interventions	Intervention group Treatment management: participants will be given access to a website containing instructions with pictures and videos for the exercises they are to complete as part of their rehabilitation. Participants will not be given a prescription for physical therapy. Participants in this group will receive weekly calls from a physical therapist during weeks 6 to 12 postoperatively. Control group Treatment management: participants will be given a prescription to see a physical therapist and a paper‐based rehabilitation protocol. Participants will be allowed to receive physical therapy at a location of their choosing. This arm represents the current standard of care.
Outcomes	Change in self‐reported function (at 2, 6, and 12 weeks, and 6 and 12 months) measured by the LEFS; change in self‐reported quality of life (at 2, 6 and 12 weeks, and 6 and 12 months) measured using the Veterans RAND 12‐Item Health Survey; change in pain as assessed by the Brief Pain Inventory short form (at 2, 6 and 12 weeks, and 6 and 12 months); participant satisfaction as assessed by a 0 to 10 satisfaction scale (at 6 months); out‐of‐pocket cost (at 6 months); travel time (at 6 months); fear of pain with movement as assessed by the Tampa Kinesiophobia Scale (at 2 weeks); self‐efficacy for home exercise programme as assessed by a 12‐Item questionnaire (at 2 weeks)
Notes	According to information in the clinical trials register, this study is now complete. We await publication of the full report.

Shin 2000.

Methods	RCT, parallel group
Participants	Number of randomised participants: 42 Inclusion criteria: ankle fractures treated with rigid internal fixation Exclusion criteria: not reported Setting: South Korea
Interventions	Intervention: ankle exercise with early weight‐bearing Control: cast immobilisation with late weight‐bearing
Outcomes	Postoperative displacement; Meyer's criteria
Notes	Study published in Korean. We are awaiting translation for additional information and to establish if study reports review outcomes. All information noted here was taken from the English abstract.

AOFAS: American Orthopedic Foot and Ankle Society; ARBOT: Ankle Rehabilitation Robot; BMI: body mass index; CAM: controlled ankle motion; EQ‐5D: EuroQol 5‐dimensions; LEFS: Lower Extremity Functional Scale; OMAS: Olerud‐Molander Ankle Score; RAND: quality of life tool from RAND Corporation; RCT: randomised controlled trial; SF‐36: Short Form Health Survey (36 items); VAS: visual analogue scale

Characteristics of ongoing studies [ordered by study ID]

ChiCTR2100045058.

Study name	The role of rehabilitation therapy in functional recovery after ankle surgery: a prospective randomized controlled study
Methods	RCT, parallel design
Participants	Expected number of participants: 100 Source population: hospital, China Inclusion criteria: people with successful ankle spared‐joint surgery without postoperative complications; patients have no heart disease, central nervous system injury and other basic diseases; no primary mental disorder and dementia disease; can consciously cooperate with the completion of cognitive function assessment; agree to participate in the study and sign the informed consent; the follow‐up time was ≥ 6 months, and the clinical data and follow‐up data were complete; the treatment period cannot accept physical therapy, traditional Chinese massage, acupuncture and other treatment; participants themselves or their family members have and can use smartphones skilfully
Interventions	Intervention group: outpatient rehabilitation Comparison group: telerehabilitation
Outcomes	AOFAS score, VAS score, TAS, tibiotalar angle, TLS, ankle range of motion
Starting date	1 April 2022
Contact information	Meng Wei, scudavidpapers@163.com, West China Hospital, Sichuan University, China
Notes

DRKS00024489.

Study name	Standard postoperative care vs. multimodal teletherapy (CASPAR) for rehabilitation following surgical treated ankle fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 74 Source population: University Medical Centre, Germany Inclusion criteria: distal fibula fractures with and without injury of the syndesmosis; fracture of the malleolus medialis; bi‐ and trimalleolar fracture; 18 to 99 years of age Exclusion criteria: pathological fracture; implant removal due to infection; < 18 years of age; prisoners; pregnant women; people with a legal guardian
Interventions	Intervention group Treatment management: postoperative aftercare with the multimodal telerehabilitation programme over a period of 3 months following hospital treatment. Regular ambulatory visits for clinical and radiological follow‐up 6 weeks, 3 months, and 1 year after surgery Control group Treatment management: postoperative standard aftercare with ambulatory physiotherapy over a period of 3 months following hospital treatment. Regular ambulatory visits for clinical and radiological follow‐up 6 weeks, 3 months, and 1 year after surgery
Outcomes	FAOS; patient‐reported survey of patient health: SF‐36, patient‐reported functional outcome of the ankle (Munich Ankle Questionnaire, Karlsson‐Peterson Score, OMAS); patient satisfaction with the postoperative aftercare rated on a Likert scale
Starting date	14 January 2021
Contact information	Mr. Dr. med. Patrick Pflüger, patrick.pflueger@tum.de, Klinik und Poliklinik für Unfallchirurgie, Germany
Notes

DRKS00024506.

Study name	Effect of a modern biodynamic feedback system on the postoperative aftercare following surgical treated ankle fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 108 Source population: University Medical Centre, Germany Inclusion criteria: all people treated operatively at the University's Department of Trauma. Fibula fracture with and without syndesmotic injury, fracture of the malleolus medialis, bi‐ and trimalleolar ankle fracture Exclusion criteria: < 18 years of age, pathological ankle fractures, implant removal due to an infection, prisoners, pregnant women, people with a legal guardian
Interventions	Intervention group Treatment management: participants will be treated with a lower leg walking boot with integrated pedographic insole for continuous load monitoring. When the preset load is exceeded, participants will get direct audible and visual feedback. There will be regular ambulatory clinical‐radiological follow‐up visits at 6 weeks, 3 months, and 1 year postoperatively. Control group Treatment management: participants will be treated with a lower leg walking boot with integrated pedographic insole for continuous load monitoring. The data will be recorded, but the participant will not get any kind of feedback (standard care). There will be regular ambulatory clinical‐radiological follow‐up visits at 6 weeks, 3 months, and 1 year postoperatively.
Outcomes	Ratio of correct partial weight‐bearing of the operated limb (number of steps with proper partial weight‐bearing/number of all steps) during the first 6 postoperative weeks; FAOS; SF‐36; Munich Ankle Questionnaire; Karlsson‐Peterson score; OMAS; participant satisfaction (Likert scale)
Starting date	28 September 2020
Contact information	Mr. Dr. med. Patrick Pflüger, patrick.pflueger@tum.de, Klinik und Poliklinik für Unfallchirurgie, Germany
Notes

Guo 2022.

Study name	Effect of transcutaneous electrical acupoint stimulation on bone metabolism in patients with immobilisation after foot and ankle fracture surgery: a randomised controlled trial study protocol
Methods	RCT, parallel design
Participants	Expected number of participants: 60 Source population: China Inclusion criteria: "satisfaction with key diagnosis points and immobilisation needs of foot and ankle injury; aged between 18 and 45 years old; with clear consciousness; the length of internal fracture fixation does not exceed 15 cm above the ankle joint; participant is informed and agrees to sign an informed consent form" Exclusion criteria: "people with endocrine diseases (thyroid, parathyroid, gonad, adrenal diseases, etc.) that affect bone metabolism, immune diseases (such as rheumatoid arthritis, digestive tract, liver and kidney diseases) that affect calcium and vitamin D absorption, and malignant diseases (such as multiple myeloma); long‐term use of glucocorticoids, antiepileptic drugs, oestrogen, heparin, progesterone and other drugs affecting bone metabolism; various congenital and acquired bone metabolic disorders; people who have taken calcium, calcitonin, vitamin D and other drugs affecting bone metabolism within 30 days; people with severe infection of the limbs, or pathological, old fractures, or other fractures of the limbs, and severe heart, liver, kidney, lung, blood and endocrine system diseases; people with histories of severe digestive system disease, kidney disease, connective tissue disease, or malignant tumour, or other disease affecting bone metabolic; pregnant or lactating women; with a stent implanted in their body or a history of joint replacement; with known heart disease and people with pacemakers; who have been treated by this study within the past 30 days; with allergies, poor compliance or mental disorders who cannot cooperate with the study; newly diagnosed people who need calcium intervention"
Interventions	Intervention group Treatment management: based on the conventional orthopaedic treatment in control group, TEAS treatment will be adopted bilaterally at SP36, BL23 and ST36 (acupuncture locations). Control group Treatment management: conventional orthopaedic treatment: (a) no weight‐bearing lower limb joint and muscle training for 4 weeks after operation, including active and passive exercises; (b) routine X‐ray examinations 1 month and 2 months after operation will be performed to observe fracture healing and internal components (plates, screws)
Outcomes	Mean changes in bone turnover marker and parathyroid hormone from baseline to 8 weeks. Secondary outcomes: bone mineral density; blood phosphorus and calcium in plasma
Starting date	November 2020
Contact information	Professor Baixiao Zhao; baixiao100@vip.sina.com
Notes

IRCT20201124049484N1.

Study name	The effects of early weight‐bearing and mobilization on the outcomes of patients with unstable ankle fractures after internal fixation surgery
Methods	RCT, parallel design
Participants	Expected number of participants: 50 Source population: Iran Inclusion criteria: 18 to 65 years of age; unstable fractures of lateral malleolus with or without talar shift; unstable fractures of medial malleolus or vertical fractures of the medial malleolus; bimalleolar fractures; trimalleolar fractures (fractures involved < 25% of posterior malleolus and do not require posterior malleolus surgery); closed fractures and open fractures of grade 1 and 2 Exclusion criteria: immature skeleton; previous ankle surgery; the need for medical care preventing other practical interventions; impossibility to follow the protocol after the operation; surgery after > 14 days from the time of injury; concurrent posterior malleolus fracture with > 25% involvement; grade 3 open fractures; concurrent syndesmosis injury
Interventions	Intervention group Treatment management: weight‐bearing and mobilisation will begin 2 weeks after internal fixation surgery of unstable ankle fractures (n = 25) Control group Treatment management: weight‐bearing and mobilisation will begin 6 weeks after internal fixation surgery of unstable ankle fractures (n = 25)
Outcomes	Time required to return to work; time required to bear full body weight; OMAS; number of required PT sessions
Starting date	9 January 2021
Contact information	Davood Dehghani Ashkezari, Davooddehghani@kaums.ac.ir, Shahid Beheshti Hospital, Iran
Notes

ISRCTN11830323.

Study name	AFTER – Ankle fracture treatment: enhancing rehabilitation trial
Methods	RCT, parallel design
Participants	Expected number of participants: 344 Source population: UK Inclusion criteria: aged ≥ 50 years with an ankle fracture undergoing surgical fixation or non‐surgical management; provided with a cast or orthotic boot (non‐removable or removable for non‐weight bearing ankle movement) for ≥ 4 weeks and no longer than 10 weeks; has capacity to consent to trial participation within 14 days of removal of the cast/boot Exclusion criteria: participant is deemed unable to adhere to trial procedures or complete questionnaires; not ambulatory before the injury; has contraindications to participation in an exercise programme
Interventions	Intervention group Treatment management: "participants will receive an advice booklet available in paper or online format from the fracture clinic containing key information on early recovery after removal of the cast/boot and basic initial exercises that they can start ahead of seeing a physiotherapist. Participants will have 4 to 6 one‐to‐one sessions with a physiotherapist, spread over 3 months from the initial session. The first session will be as soon as possible after the referral, and no later than 3 weeks from randomisation. Sessions will be delivered via face‐to‐face or telephone/videoconference, whichever mode of physiotherapy delivery would usually be provided for people with ankle fracture. Exercise progression will be individualised by progressing and regressing the volume and load in line with each participant’s capabilities and preferences. Participants will be asked to identify their goals and, with the physiotherapist’s help, write an action plan for where and when they will perform their home exercises and a contingency plan for managing difficulties. Participants will receive a personal exercise guide and diary." Control group Treatment management: self‐directed rehabilitation
Outcomes	Ankle function (OMAS) at baseline, 2, 4 and 6 months; HRQoL (EQ‐5D‐5L) at baseline, 2, 4 and 6 months; pain measured by pain sub‐scales of the EQ‐5D‐5L and OMAS at baseline, 2, 4 and 6 months; physical function (PROMIS Physical Function) at baseline, 4 and 6 months; self‐efficacy (Self‐Efficacy Exercise Score) at baseline, 4 and 6 months; exercise adherence (self‐reported exercise frequency) at 2, 4 and 6 months; complications (complications questionnaire and CRF) at 2, 4 and 6 months; cost‐effectiveness of interventions (health economics questionnaire) at 2 and 6 months
Starting date	15 September 2022
Contact information	David Keene Email: after@ndorms.ox.ac.uk
Notes	Due for completion: 31 July 2024

JPRN‐UMIN000039630.

Study name	Effect of neuromuscular electrical stimulation on individual quadriceps muscle atrophy in patients with non‐weight bearing: a randomized control study
Methods	RCT, parallel design
Participants	Expected number of participants: not reported Source population: Japan Inclusion criteria: 20 to 70 years of age; females and males Exclusion criteria: participants with a history of leg fracture and surgery, and hip and/or knee osteoarthritis
Interventions	Intervention group Treatment management: during non‐weight‐bearing, home exercises are performed by themselves; open kinetic chain exercises including knee flexion and extension, and hip abduction and adduction 30 times per day. In addition to the home exercises, neuromuscular electrical stimulation is performed during knee flexion and extension exercise for 10 minutes twice per week at hospital during non‐weight‐bearing. Control group Treatment management: not reported
Outcomes	Individual QM thicknesses are measured 1 week after starting non‐weight‐bearing, immediately before starting weight‐bearing, and the end of rehabilitation. Clinical symptoms are assessed with participant‐reported outcome measures: the SAFE‐Q at the end of rehabilitation. Strengthening and tightness of QM are measured at the end of rehabilitation
Starting date	March 2020
Contact information	koun_mail_world@yahoo.co.jp
Notes	We could not determine if data will be reported separately for participants with ankle fracture. We will assess this information once the study is completed and published.

JPRN‐UMIN000047296.

Study name	Effects of ankle joint exercises with visual and auditory feedback on pain, motor function, and quality of life in patients with ankle or foot fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 64 Source population: Japan Inclusion criteria: adults between 18 and 85 years with foot or ankle fractures Exclusion criteria: participants with serious arrhythmia; with a tendency to bleed; with obvious inflammatory symptoms (redness, heat, swelling, too much pain); with malignant tumour; if there is sensory disturbance; if there is infection; unstable circulatory dynamics; peripheral circulatory disturbance
Interventions	Intervention group Treatment management: "after surgery, isometric ankle plantar flexion exercise using visual and auditory feedback is performed in addition to normal rehabilitation from the start of loading to 4 weeks after the start of loading. Isometric ankle plantar flexion exercises using visual and auditory feedback are performed 5 times a week, 10 times for 10 seconds, 3 sets" Control group Treatment management: "only normal rehabilitation is performed after surgery and after the start of loading"
Outcomes	VAS; Short‐Form McGill Pain Questionnaire; Pain Catastrophizing Scale; Hospital Anxiety and Depression Scale; Neglect Like Symptom Score; Tampa Scale for kinesophobia; pressure pain threshold; two‐point identification; range of motion; muscle strength; 10 m walking time; walking distance for six minutes; self‐administered foot evaluation questionnaire
Starting date	3rd November 2021
Contact information	Hideki Kataoka; kataoka‐hide@kii.bbiq.jp
Notes

NCT03970603.

Study name	Syndesmosis ankle study
Methods	RCT, parallel design
Participants	Expected number of participants: 80 Source population: hospital, USA Inclusion criteria: ≥ 18 years of age; ankle fracture with suspected syndesmotic disruption that would likely be treated with suture button fixation operative intervention; ability to ambulate preoperatively Exclusion criteria: < 18 years of age; length unstable syndesmotic injury; neuropathic diabetics; weight ≥ 325 lbs; prisoner; pregnant or lactating; end‐stage renal disease; inability to follow commands (dementia, TBI, etc.); people with polytrauma
Interventions	Intervention group Treatment management: early weight‐bearing. Being directed to bear weight on the affected ankle 2 weeks from suture button fixation for syndesmotic disruption Control group Treatment management: delayed/late weight‐bearing. Being directed to bear weight on the affected ankle 6 weeks from suture button fixation for syndesmotic disruption
Outcomes	Measure of the syndesmotic change; pain scores; surgical experience; work productivity and activity impairment; foot and ankle outcomes
Starting date	15 October 2019
Contact information	Stacee Baker, bakersa@umsystem.edu, University of Missouri Health System, USA
Notes

NCT04028414.

Study name	Effects of early weight bearing on rehabilitation outcomes in patients with traumatic ankle and tibial plateau fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 100 Source population: 20 locations, USA Inclusion criteria: "≥ 18 years of age; surgically treated open or closed fractures of the ankle (OTA Codes 44A, B, C) meeting one of the following criteria: 1. bimalleolar fracture requiring fixation of any two malleoli (i.e. lateral/medial, medial/posterior, lateral/posterior); 2. bimalleolar equivalent SER4 fractures (lateral malleolus fracture requiring fixation and no medial or posterior malleolus fixation); 3. trimalleolar fracture with fixation of at least two malleoli (i.e. lateral/medial, medial/posterior, lateral/posterior); 4. operative treatment of an isolated lower extremity unicondylar tibial plateau fracture (AO/OTA 41B1, Schatzker Type 1 or 4) without joint impaction" Exclusion criteria: "Gustilo Type III injuries or soft tissue injury of either lower extremity that would contraindicate immediate or delayed weight‐bearing; syndesmotic injuries/fixation; osteoporosis as defined by treatment with a bisphosphonate and/or other osteoporosis medications, including Prolia and Forteo; or a prior fragility fracture (e.g., spine compression, proximal humerus, distal radius, femoral neck/intertrochanteric fracture); neuropathy, defined as diagnosis of peripheral neuropathy in medical record, neuropathic foot ulcer, or diminished or absent plantar sensation to light touch; pathologic fracture related to neoplasm; fractures and dislocations to the ipsilateral or contralateral lower limb that prevent weight‐bearing as tolerated after first postoperative visit; other contraindication to immediate or delayed weight‐bearing (e.g. ipsilateral fracture effecting weight‐bearing status such as a calcaneus fracture); fractures of the ipsilateral lower extremity that would not affect weight‐bearing status may be included (e.g. proximal femur, femoral shaft or tibia shaft fractures); any upper limb injury that would limit upper extremity weight‐bearing (e.g. surgical neck fracture of proximal humerus); injuries to other body systems that would affect the ability to comply with either weight‐bearing protocol (e.g. spinal cord injury; severe TBI; major abdominal or chest injury); person has third degree burns on >10% total surface area affecting the study limb; tibial plateau fractures that are required to wear a locking brace or fractures that require fixed immobilisation beyond 4 weeks (e.g. cast); non‐ambulatory pre‐injury; pre‐injury limitation to ROM of ipsilateral hip, knee or ankle; morbidly obese (BMI ≥ 40); documented psychiatric disorder requiring admission in perioperative period; severe problems with maintaining follow‐up (e.g. prisoners or homeless at the time of injury or who are intellectually challenged without adequate family support)"
Interventions	Intervention group Treatment management: "participants with ankle fractures will be instructed to weight‐bear as tolerated while in a boot with a heel to toe normal gait and wean from walker or crutches to a cane or no support device. At the 6‐week postoperative visit, participants with ankle fractures will be instructed to wean from the boot and continue full weight‐bearing as tolerated until full weight‐bearing is achieved. Participants with plateau fractures will be instructed to begin weight‐bearing as tolerated until full weight‐bearing is achieved." Control group Treatment management: "participants with ankle fractures will be instructed to touch‐down (toe touch or foot flat) weight‐bear (approximately 10% of body weight) while in the boot. Participants will be instructed to keep their foot off of the floor or set the ball of the foot or heel on the ground for balance, using walker or crutches at all times. After the 6‐week postoperative visit, participants may begin weight‐bearing as tolerated. Participants with tibial plateau fractures will be instructed to touch down (toe touch or foot flat) weight‐bear (approximately 10% of body weight) for at least 6 weeks. After the 6‐week postoperative visit, participants may begin weight‐bearing as tolerated until full weight‐bearing is achieved."
Outcomes	Number of study related complications experienced; return to usual major activity and work productivity; self‐reported measures of function; pain score; pain interference; joint ROM
Starting date	1 January 2020
Contact information	Elizabeth Wysocki, ewysock2@jhu.edu
Notes

NCT04429217.

Study name	A study to compare two different managements after ankle surgery: immediate and delayed weight‐bearing
Methods	RCT, parallel design
Participants	Expected number of participants: 100 Source population: hospital, Switzerland Inclusion criteria: "male and female adults; 18 to 70 years of age; signed informed consent; ankle fracture requiring surgical management classified as: 1. Weber A fracture pattern (AO 44.A1, 44.A2, 44.A3) with or without association with a medial or posterior lesion/fracture, or 2. Weber B fracture pattern (AO 44.B1, 44.B2, 44.B3) with or without association with a medial or posterior lesion/fracture; willingness and ability to participate in the trial" Exclusion criteria: "Weber C fracture pattern; bilateral fractures or fractures in other locations; maisonneuve‐associated lesion; BMI < 18.5 and ≥ 30; metabolic diseases influencing fracture healing processes; medical comorbidities precluding operative intervention; inability to follow the procedures of the study, e.g., due to language problems, psychological disorders, dementia, general physical problem of the participant, etc.; inability or contraindications to undergo the investigated intervention; other clinically significant concomitant disease states (e.g., renal failure, hepatic dysfunction, cardiovascular disease, etc.); pregnant women"
Interventions	Intervention group Treatment management: experimental postoperative rehabilitation intervention consisting of immediate weight‐bearing Control group Treatment management: control postoperative rehabilitation intervention consisting of delayed weight‐bearing
Outcomes	OMAS; FADI; pain on the NRS; daily pain and function; change in quality of life; return to previous daily activities; objective ankle functional recovery; ankle joint flexibility; objective activity level; radiological assessment; pain drug consumption; participant satisfaction
Starting date	28 April 2020
Contact information	Enrique Testa, Ente Ospedaliero Cantonale, Bellinzona
Notes

NCT04872296.

Study name	Weight bearing for fibular fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 64 Source population: hospital, USA Inclusion criteria: > 18 years of age; undergoing surgical treatment for isolated lateral malleolar fracture with foot and ankle surgeons at the Rothman Orthopaedic Institute Exclusion criteria: people requiring fixation of other fractures, deltoid ligament repair, or syndesmosis fixation; revision ankle fracture cases; pre‐existing impaired mobility; pre‐existing cognitive disability; open fractures; BMI > 40; diabetes mellitus
Interventions	Intervention group Treatment management: participants treated for an ankle fracture will be allowed to weight‐bear early after surgery starting 2 weeks postoperatively Control group Treatment management: participants treated for an ankle fracture will be treated with standard protocol of non‐weight‐bearing for 6 weeks postoperatively.
Outcomes	Participant satisfaction using the SF‐12; participant satisfaction using the FAAM score; participant pain using the VAS score for pain
Starting date	20 December 2018
Contact information	Not specified
Notes

NCT05280639.

Study name	Simplified post op rehabilitation for ankle and pilon fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 30 Source population: hospital, USA Inclusion criteria: "18 to 65 years of age; surgically treated open or closed fractures of the ankle or tibial plafond" Exclusion criteria: "contralateral lower extremity injuries that would limit weight‐bearing after 6 weeks; severe injury requiring flap coverage or vascular reconstruction (Gustilo‐Anderson Type IIIB and C, respectively); neurological deficits that would impede ability to stand safely unassisted for home exercise regiment; desire to participate in formal physical therapy program; additional injury that would compromise subjects ability to follow either Home Exercise Program; non‐ambulatory prior to injury; previous ankle or tibial plafond injury on ipsilateral extremity; BMI > 50; severe problems maintaining follow up; previous ankle/tibial plafond fracture; prisoners; neurological impairments that impair balance"
Interventions	Intervention group: "participants will perform a simplified postoperative rehabilitation program using a simplified wooden block protocol" Control group: "participants will perform formal physical therapy or a home exercise program consistent with AAOS standards"
Outcomes	Ankle ROM (at 2 and 6 weeks, and 6 months after surgery); self‐reported participant compliance with exercise programme: number of days completed (at 2 and 6 weeks after surgery); LEFS score (at 2 and 6 weeks, and 6 months after surgery); pain (at 2 and 6 weeks, and 6 months after surgery)
Starting date	Not yet recruiting; estimated starting date 31 December 2022
Contact information	Eric McVey, edm9u@hscmail.mcc.virginia.edu, University of Virginia Orthopaedic Surgery
Notes

NCT05587842.

Study name	Early vs delayed weightbearing after surgical fixation of unstable ankle fractures with syndesmosis disruption
Methods	RCT, parallel design
Participants	Expected number of participants: 150 Source population: USA Inclusion criteria: unilateral ankle injury with unstable syndesmotic rupture, in the setting of isolated syndesmotic sprain, fibular fracture, bimalleolar, or trimalleolar equivalent ankle fracture; surgical fixation of syndesmosis and/or ankle within 2 weeks of injury; closed or low grade open (Gustilo‐Anderson grade I & II low‐energy injuries without gross contamination) fractures; skeletally mature with closed physis Exclusion criteria: previous ipsilateral ankle surgery; bilateral ankle fractures or concurrent other lower extremity injuries/major injuries that would affect recovery time; inability to cooperate with postoperative protocol (advanced dementia, polytrauma, developmental delay, etc.); non‐ambulatory pre‐injury; tibial plafond fractures including articular impaction requiring elevation; peripheral neuropathy; BMI > 45
Interventions	Intervention group Treatment management: participants will be instructed to be weight‐bearing, as tolerated, on their ankle starting 2 weeks after operation Control group Treatment management: participants will be instructed to be non‐weight‐bearing on their ankle until at least 6 weeks. After 6 weeks, the participant will be instructed to be weight‐bearing as tolerated.
Outcomes	Physical Exam Functional Outcomes; X‐Ray Functional Outcomes; loss of reduction of the syndesmosis; OMAS; SF‐36; participant satisfaction survey; difference in rates of adverse events. Follow‐up at 1 year
Starting date	10 February 2021
Contact information	Anna Valencia; atvalencia@arizona.edu Stephanie Marsh; slmarsh@arizona.edu
Notes

NTR4469.

Study name	The treatment of stable ankle fractures: brace versus cast immobilization
Methods	RCT, parallel design
Participants	Expected number of participants: 100 Source population: hospital, Netherlands Inclusion criteria: people with a stable ankle fracture (type Weber B and < 2 mm dislocation); 18 to 75 years of age Exclusion criteria: multiple fractures; intellectual disability; not living in the right region (follow‐up takes place in a different medical centre); people who do not speak Dutch fluently
Interventions	Intervention group: participants treated with cast immobilisation for a period of 5 weeks Control group: participants treated with a functional ankle brace for a period of 5 weeks
Outcomes	Primary outcome: OMAS at 6 weeks Secondary outcome: pain; comfort of cast/brace; ROM; long‐term function (using the AAOS foot and ankle score); HRQoL (using EQ‐5D); fracture dislocation; presence of osteoarthritis at 1 year
Starting date	1 February 2013
Contact information	NL Weil, n.l.weil@lumc.nl, Medical Center Haaglanden ‐ Bronovo Hospital
Notes	The clinical trials register was last updated in May 2022, and we therefore assumed that this study was still recruiting despite its starting date in 2013.

NTR7143.

Study name	Prospective randomized controlled trial: early weight bearing after conservative treatment of Weber B ankle fractures
Methods	RCT, parallel design
Participants	Expected number of participants: 142 Source population: hospital, Netherlands Inclusion criteria: ≥ 16 years of age; stable ankle fracture (non‐dislocated Weber B or Lauge‐Hansen supination‐eversion stage 2‐4A fracture, isolated malleolus tertius fracture); conservative treatment with a dorsal lower limb cast Exclusion criteria: < 16 years of age; Weber A ankle fracture; unstable ankle fracture; fractures involving both lower extremities; operative treatment of the ankle fracture; post‐traumatic period > 10 days; amputation of upper leg, lower leg or foot
Interventions	Intervention group Treatment management: participants receive a walker with which they can start permitted weight‐bearing Control group Treatment management: participants receive a circular lower limb plaster cast and instructions for non‐weight‐bearing and immobilisation until the next visit. After 6 weeks, the plaster cast will be removed. The participants may then extend weight‐bearing.
Outcomes	Functional outcome score of the ankle; dislocation of the fracture
Starting date	9 April 2018
Contact information	S Franssen, s.franssen@zuyderland.nl, Zuyderland Medisch Centrum
Notes

Stinner 2022.

Study name	Early advanced weight‐bearing after periarticular fractures: a randomized trial comparing antigravity treadmill therapy versus standard of care
Methods	RCT, parallel design
Participants	Expected number of participants: 26 Source population: hospital, USA Inclusion criteria: "18 to 55 years of age; closed or open (Gustilo type I, II, or IIIA) fractures meeting at least 1 of the following criteria: 1. distal femur (OTA/AO 33B, 33C); 2. tibial plateau (OTA/AO 41B, 41C); 3. distal tibia (OTA/AO 43B, 43C); definitively treated with a plate; ambulatory before injury; able to participate in rehabilitation starting 28 days after definitive fixation" Exclusion criteria: "does not meet height or weight requirements for AlterG (4 feet 8 inches to 6 feet 4 inches and 85 lbs to 400 lbs); medical or psychological disease that would preclude safe functional testing (severe TBI, stroke, heart disease, etc.); previous joint trauma or disease of the operative extremity that resulted in pain, stiffness, or other limitation to mobility; injury to the contralateral limb, upper extremities, or axial skeleton that would influence single‐limb weight‐bearing or use of an assistive device; pregnancy; unable to speak English; severe problems with maintaining follow‐up (e.g, prisoners, homeless at the time of injury, who are intellectually challenged without adequate family support, or who have documented psychiatric disorders); unable to provide informed consent"
Interventions	Intervention group: antigravity treadmill therapy Comparison group: standard care
Outcomes	Primary outcome: participant‐reported function using the KOOS for plateau and distal femur fractures and the AOS for distal tibia fractures Secondary outcomes: physical function; psychosocial health; return to work; ROM and strength; physical performance; fracture healing; fixation loss is defined as loss of reduction or hardware failure that requires reoperation or a change in weight‐bearing status; fracture alignment; other injury‐related complications including infections and wound healing are documented from the medical record
Starting date	Not specified
Contact information	Daniel J Stinner: daniel.j.stinner@vumc.org
Notes

TCTR20220221004.

Study name	Early weight bearing after ankle fracture fixation rapid returns of driving: a randomized controlled trial
Methods	RCT, parallel design
Participants	Expected number of participants: 58 Source population: Thailand Inclusion criteria: people who underwent right sided ankle fracture fixation classified as Lauge‐Hansen SER‐II and SER‐IV at the Department of Orthopedic Surgery Siriraj Hospital; aged between 18 and 65 years; a driver with a driver's license Exclusion criteria: associated other right lower extremity injuries; unable to pass the ability test that included Snellen chart for visual acuity, Trails making tests for cognitive ability and 3‐D block construction for perceptual ability testing; < IV of the ankle of plantar flexion muscle strength grading by the Oxford scale; unstable fixation
Interventions	Intervention group Treatment management: early weight‐bearing group receive a standard functional boot and allowed weight‐bearing as tolerated Control group Treatment management: conventional rehabilitation protocol group receive a short leg slab and non‐weight‐bearing postoperative protocol
Outcomes	Brake response time at 12 months after end of operation; ankle ROM at 12 months after end of operation; FAAM score at 12 months after end of operation
Starting date	1 September 2019
Contact information	Theerawoot Tharmviboonsri; th.woot@gmail.com Paphon Hirunyachoke; moopaphon@gmail.com
Notes

AAOS: American Academy of Orthopaedic Surgeons; AO/OTA: Arbeitsgemeinschaft für Osteosynthesefragen (German for "Association for the Study of Internal Fixation")/ Orthopaedic Trauma Association; AOFAS: American Orthopaedic Foot & Ankle Society; AOS: Ankle Osteoarthritis Score; BMI: body mass index; CRF: case report form; EQ‐5D(‐5L): EuroQol‐5 dimension; FAAM: Foot and Ankle Ability Measure; FADI: Foot and Ankle Disability Index; FOAS: Foot and Ankle Outcome Score; HRQoL: health‐related quality of life; KOOS: Knee Injury and Osteoarthritis Outcome Score; LEFS: Lower Extremity Functional Scale; NRS: Numerical Rating Scale; OMAS: Olerud & Molander Ankle Score; ORIF: open reduction internal fixation; PROMIS: Patient Reported Outcomes Measurement Information System; PT: physiotherapy; QM: quadriceps muscle; RAND: corporation responsible for research and production of SF‐12; organisation RCT: randomised controlled trial; ROM: range of motion; SAFE‐Q: Self‐Administered Foot Evaluation Questionnaire; SER: supination ‐ external rotation; SF‐12/SF‐36: Short Form Health Survey (12/36 item); TAS: distal tibial surface angle; TBI: traumatic brain injury; TEAS: transcutaneous electrical acupoint stimulation; TLS: tibial lateral surface angle; VAS: visual analogue scale

Differences between protocol and review

Here, we note changes to the review since the previous version (Lin 2012).

Review authors: five additional authors contributed to this review update (SL, MP, RP, DK, CB). The following authors did not contribute to this update (Nicole Donkers, Kathryn Refshauge, Paula Beckenkamp, Kriti Khera); see Acknowledgements.

Types of interventions: we split the review into three distinct comparison groups, which we then stratified according to surgical or non‐surgical fracture management. The decision to use these comparison groups was driven by the current main clinical questions in rehabilitation management.

Types of outcome measures

• In this update, we did not exclude studies based on outcome measures.

• We did not include the following outcomes in this update because these outcomes were less important for this review question: ankle dorsiflexion and plantarflexion range of motion, strength, and swelling.

• For adverse events, we included any adverse event reported by study authors. We also collected data for unplanned return to theatre (for any adverse event), which we considered to be the most important adverse event for this review question.

• We reported the results at three distinct time points, and specified a threshold of six months to distinguish short‐ from long‐term follow‐up.

Search methods: we did not search PEDro database in this review update.

Measures of treatment effect: we interpreted standardised mean differences (SMDs) in the review by re‐expressing effect estimates as mean differences (MDs) using the most commonly used scale.

Assessment of reporting biases: we specified an intention to formally assess the potential for publication bias through assessment of funnel plots. We also described methods to assess outcome reporting bias in each individual study.

Data synthesis

• We used a random‐effects model, rather than a fixed‐effect model, in order to account for likely differences in participants.

• We stratified all data analysis, where appropriate, according to fracture management (surgical or non‐surgical approach).

Sensitivity analysis: we assessed the impact of using a random‐effects model in meta‐analysis. Also, we assessed the impact of using different thresholds for follow‐up time points. Finally, we excluded any outlying data during sensitivity analysis.

Summary of findings and assessment of the certainty of the evidence: the previous version of the review did not include summary of findings tables. In this update, we constructed summary of findings tables for each of the three main comparison groups. Through discussion, we agreed a set of outcomes which we judged to be most clinically relevant and relevant to a person recovering from ankle fracture. We reported data in summary of findings tables for short‐term follow‐up (≤ six months).

Contributions of authors

SL (systematic reviewer): co‐ordinated the work and progress of the review, sifted and identified included studies, extracted study data, interpreted the findings, and drafted the review

MP (systematic reviewer): sifted and identified included studies, extracted study data, interpreted the findings, and drafted the review.

RP (systematic reviewer): sifted and identified included studies, extracted study data, interpreted the findings, and drafted the review.

PB (content expert): reviewed and approved the final draft.

HS (systematic reviewer): drafted the review, and reviewed and approved the final draft.

DK (content expert): reviewed and approved the final draft.

CB (content expert): reviewed and approved the final draft.

C‐WL (guarantor and content expert): conceived the original review and wrote the previous versions of this version, reviewed and approved the final draft of this version.

Sources of support

Internal sources

• Queen Mary University of London, UK

  Support to the Cochrane Bone, Joint and Muscle Trauma Group up to 31 March 2023

• Faculty of Health Sciences, The University of Sydney, Australia

  Support to author (Christine Lin)

External sources

• NIHR, UK

  This work acknowledges the support of the National Institute for Health Research Barts Biomedical Research Centre (NIHR203330)

• NIHR, UK

  The National Institute for Health Research (NIHR) provided infrastructure funding to the Cochrane Bone, Joint and Muscle Trauma Group up to 31 March 2023.

• National Health and Medical Research Council, Australia

  Christine Lin is supported by a fellowship (Investigator Grant APP1193939)

• NIHR, UK

  David Keene’s institutions have received grants from the National Institute for Health and Care Research (NIHR) for research projects related to this work. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.

Declarations of interest

Four of the review authors (PB, DK, CB, C‐WL) were involved in included studies. These review authors remained independent of data extraction, risk of bias, and any interpretation of findings for these studies.

SL is the former Deputy Co‐ordinating Editor of the Cochrane Bone, Joint and Muscle Trauma Group; she was not involved in the editorial process for this review, and has declared that she has no conflict of interest.

MP has declared that he has no conflict of interest.

RP has declared that she has no conflict of interest.

HS has declared that he has no conflict of interest.

PB has declared that she has no conflict of interest.

DK has declared that he has no conflict of interest.

CB has declared that he has no conflict of interest.

C‐WL has declared that she has no conflict of interest.

New search for studies and content updated (conclusions changed)