Are pre-operative exercise interventions for joint arthroplasty effective at improving peri‑operative outcomes? A systematic review and meta-analysis of randomized controlled trials

Highlights

• •Preoperative exercise before total knee arthroplasty and total hip arthroplasty was found to improve patient-reported health-related quality of life, pain and function at pre- and post-surgery.
• •No significant effects were found on hospital length of stay.
• •Most interventions combined strength training alongside other modalities and were delivered face-to-face, typically in clinical settings.
• •The main behavior change techniques of the interventions included instruction, self-monitoring, and behavioral demonstration. Of these, only instruction on how to perform the behavior was significantly associated with improved functional outcomes, according to meta-regression analysis.

Abstract

Background

Evidence on pre-operative physical activity before hip and knee arthroplasty is limited and heterogeneous. Intervention components and behavior change techniques remain underexplored. This review examined the effectiveness of pre-operative physical activity interventions on patient and surgical outcomes in elective hip and knee arthroplasty up to 12 weeks post-surgery.

Methods

A systematic search of 8 databases up to August 8, 2024 identified randomized controlled trials of physical activity interventions before total hip and knee arthroplasty. Quality of evidence was evaluated with Grading of Recommendations, Assessment, Development and Evaluation framework. Random-effects models were used for meta-analyses. The Template for Intervention Description and Replication (TIDieR) was used for detailing the included interventions.

Results

Forty trials were included. Significant mean differences in favor of the intervention groups were found at pre- and post-surgery in 4 outcomes: health-related quality of life (Knee pre-surgery standardized mean difference (SMD) = −0.5, 95% confidence interval (95%CI): −1.0 to −0.1 and Hip and Knee post-surgery SMD = −0.4, 95%CI: −0.6 to −0.1), pain (Hip and Knee pre-surgery SMD = −0.4, 95%CI: −0.6 to −0.1 and Knee post-surgery SMD = −0.3, 95%CI: −0.6 to −0.1), function (Hip and Knee pre-surgery SMD = −0.5, 95%CI: −0.8 to −0.2 and Knee post-surgery SMD = −0.6, 95%CI: −1.0 to −0.2), and timed-up-and-go (Hip and Knee MD = −1.2 s, 95%CI: −2.0 to −0.3 and Hip and Knee MD = −1.3 s, 95%CI: −1.7 to −0.8). Half of the interventions reported over 75% of the TIDieR items, while behavior change techniques reporting was limited.

Conclusion

Pre-operative exercise improves health-related quality of life, pain, and function pre- and post-surgery in elective hip and knee arthroplasty. Standardized reporting is needed for establishing effective intervention components.

Graphical abstract

1. Introduction

As of July 2024, over 850,000 patients were waiting for orthopedic surgery (having been seen by a consultant) in England,¹ with a median wait time of 24 weeks and approximately 5% of patients waiting over 52 weeks for treatment.¹ However, these statistics do not reflect the time patients have been waiting overall to be seen by a consultant after referral from a general practitioner, which could be up to 2 years in some cases.² Prolonged wait times are problematic for patients as their health-related quality of life (HRQOL), including their physical, mental, and social health, are compromised.^3,4 Similarly, physical deterioration and comorbidities increase the risk of adverse events and complications resulting in reduced independence and increased social care needs.^3,5,6 Evidence suggests that longer waiting times negatively impact health gain after joint replacement surgery as well as makes surgery riskier due to comorbidities.^5,7, 8, 9

Optimizing the health of patients awaiting surgery through physical activity interventions may reduce the risks associated with surgical complications and improve HRQOL. Intervening with patients during the pre-operative period is known as prehabilitation. Prehabilitation can consist solely of physical activity (including exercise) or become multimodal by incorporating psychological and nutrition components. Notably, prehabilitation may play a multifaceted role in influencing outcomes. Enhanced physiological readiness could reduce length of hospital stay and post-operative complications while also promoting faster and more complete recovery.^10,11

Current evidence on orthopedic prehabilitation is heterogeneous due to a lack of high-quality evidence and different protocols.12, 13, 14 A systematic review of orthopedic prehabilitation interventions found improvements in pre-operative function and strength in total knee replacement (TKR), and HRQOL and muscle strength in total hip replacement (THR).¹⁵ Another systematic review narratively synthesized the evidence for prehabilitation and suggested it may result in increased strength, reduced length of stay, and may not lead to increased harms, but it may be comparable to no prehabilitation in terms of pain, range of motion, and activities of daily living.¹⁶ These reviews investigated a variety of prehabilitation interventions, encompassing unimodal approaches, which focused solely on exercise, education or cognitive behavioral therapy, and multimodal approaches that integrated these components with each other or with alternative therapies. Furthermore, Punnoose et al.¹⁵ excluded non-English texts, and Konnyu et al.¹⁶ only examined outcomes post-operatively. Thus, none of the reviews specifically explored physical activity interventions, thoroughly outlined intervention components, or the behavior change techniques (BCTs) involved.

Exercise can improve aerobic capacity and increase lean muscle mass, while physical activity, when performed at sufficient intensities and frequencies or by deconditioned individuals, can help maintain or slow the deterioration of physiological function.17, 18, 19, 20, 21, 22, 23, 24 Optimizing physiological function helps patients withstand surgical stress demands and thus have better post-operative outcomes.^12,17,18,20 For this review, physical activity is defined as “any bodily movement produced by skeletal muscles that results in energy expenditure above resting levels”, while exercise is a subset of that, defined as: “physical activity that is planned, structured, and repetitive to improve or maintain physical fitness”.²⁵ All physical activity interventions, including exercise, were included in this review.

There is also a lack of evidence regarding the effective intervention components, as listed in the Template for Intervention Description and Replication (TIDieR) framework, and BCTs that should be included in prehabilitation interventions. This information can help design effective prehabilitation interventions. This systematic review and meta-analysis aimed to synthesize data from randomized controlled trials (RCTs) and observational studies to answer the following review questions: (a1) Are physical activity interventions before TKR and THR effective at improving patient and surgical outcomes? (a2) To help guide future interventions, what are the specific components of these interventions? (b) Is physical activity adherence before hip or knee arthroplasty associated with improved surgical outcomes?

2. Methods

The findings of this review are reported in line with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines.²⁶ The review was registered on International Prospective Register of Systematic Reviews (PROSPERO; CRD42023429030).

2.1. Study inclusion criteria

There were originally 2 review questions, as stated above. However, no eligible observational studies were identified for the second question to examine physical activity levels before joint arthroplasty on surgical outcomes. Thus, the following methodology and results relate to review question one, which included RCTs only. Table 1 details the inclusion/exclusion criteria for review Question 1, meanwhile Supplementary Table 1 outlines the eligibility criteria for review Question 2.

Table 1.

Eligibility criteria.

Item	Eligibility criteria
Participants	Inclusion:
	● Adults (≥18 years) waiting for partial or total hip or knee arthroplasty (THA or TKA)
	Exclusion:
	● Pregnant women
Interventions	Inclusion:
	● Any intervention involving a physical activity (including exercise) component
	● Any method of delivery
Control/comparator	Inclusion:
	● No treatment (e.g., usual care or wait-list control)
	● Attention control (e.g., intervention of different content)
	● Minimal intervention comparable to usual care (e.g., providing generic infographics or electronic communication about physical activity)
Outcome	Inclusion:
	● Any pre-operative and post-operative patient-reported outcomes measuring physical activity, physical function, pain, quality of life and/or emotional wellbeing
	● Any pre-operative and post-operative performance-based outcomes objectively measuring physical function and ability
	● Any of the following objective surgical outcomes: hospital length of stay, readmission, adverse events and complications, and discharge destination

Abbreviations: THA = total hip arthroplasty; TKA = total knee arthroplasty.

2.2. Search strategy

A systematic search of the following databases was conducted on August 8, 2024: MEDLINE, PsychINFO, SPORTDiscus, PubMed, PEDro, Embase, CENTRAL, and Scopus. No restrictions were implemented; trials published from inception to the search date were eligible. Both English and non-English studies were included. Data from the non-English textswere extracted using translations generated through the Google Translate document upload function.²⁷ The main search terms included prehabilitation, pre-operative care, exercise, physical activity, orthopedic, and arthroplasty (full search strategy: Supplementary Table 2). Reference lists of included studies and systematic reviews were also manually searched.

2.3. Study selection

Search results were uploaded to Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). Duplicates were automatically removed. Two independent reviewers (from among NAH, CDM, BB, JJCT, and JPS) screened study titles, abstracts, and full texts using the inclusion/exclusion criteria. Any disagreements were discussed between reviewers and resolved by consensus or with a third reviewer.

2.4. Outcomes

The main outcome intended to be explored was change in physical activity at the end of the intervention and post-surgery; however, this was not often reported (n = 3). As the outcomes measured and reported by studies varied, only the most reported outcomes were analyzed. Supplementary Tables 3–6 list all the outcomes reported by the eligible studies; Supplementary Table 7 specifies which patient-reported outcome measures were used. The outcomes that were most reported and able to be investigated were: patient-reported quality of life, pain, and function; objective performance-based physical function (including joint range of motion (ROM), timed up-and-go (TUG), and 6-min walk test (6MWT)); and, an objective surgical outcome, hospital length of stay (LOS). The mean difference of the change in these outcomes was measured pre-surgery (from baseline to post-intervention, prior to surgery) and post-surgery (from baseline to after surgery), with the obvious exception of LOS.

2.5. Data extraction

Data were extracted and summarized by one reviewer (NAH) and checked by a second reviewer (BB). Extracted data included the study date and country, study design, methodology, sample size, participant characteristics, intervention details, planned surgical procedure (hip, knee, or both), control/comparator group, follow-up time, outcome measures, results, and BCTs. Study investigators were contacted via email if additional details were required.

Some studies did not define data collection points with time-specific details in relation to surgery.28, 29, 30, 31, 32, 33, 34, 35, 36, 37 The authors of these studies were contacted to determine the data collection time intervals. Studies also measured post-surgery outcomes at different time points. To include as many studies as possible, but to decrease heterogeneity, studies were synthesized if they reported data between 6 and 12 weeks post-surgery due to most studies reporting post-operative data within this timeframe.

Intervention components were extracted in line with the TIDieR.³⁸ Meanwhile, BCTs were identified following the Behavior Change Technique Taxonomy v1 (BCTTv1).³⁹ For TIDieR Item 12 (how well), a satisfactory compliance rate was defined as completing ≥80% of intervention sessions based upon a systematic review defining adherence on therapeutic exercise.^40,41 BCTs were summarized individually (e.g., BCT 1.1 goal setting (behavior)) and within the categorized groups (e.g., Group 1 goals and planning).³⁹ BCTs were coded and included if they were explicitly stated in the intervention and absent in the control group.

2.6. Data synthesis

2.6.1. Quality assessment

Risk of bias (RoB) was assessed by 2 independent reviewers (NAH and BB) using RoB 2 (Version 2, August 2019, The Cochrane Collaboration).⁴² RCTs were categorized as low, some concerns, or high overall bias.⁴³ Certainty of evidence was rated by 2 independent reviewers (NAH and CDM) following the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach. Evidence was graded based upon RoB, inconsistency, indirectness, imprecision, and publication bias. Ratings were categorized as very low, low, moderate, and high.⁴⁴ Any disagreements between reviewers for both risk of bias and GRADE were discussed and resolved by discussion or by a third reviewer (JPS).

2.6.2. Outcomes

Meta-analyses were conducted using Review Manager (RevMan, Version 5.4, The Cochrane Collaboration; revman.cochrane.org) based on intention to treat principle. All outcome data were continuous, and random effects models were used as the diversity of the intervention components and comparator conditions meant that treatment effects were expected to differ. Prior to conducting the meta-analyses, normality was assessed by using absolute values to calculate the mean to standard deviation (SD) ratio of all studies for all outcome measures. A ratio of less than two was used to identify skewness.⁴⁵ A sensitivity analysis that excluded skewed data was conducted to assess the impact on the overall, primary meta-analysis. Robustness between the primary and sensitivity analysis was determined by comparing the effect sizes, confidence intervals, and heterogeneity.⁴⁵ Where substantial differences were identified, the sensitivity analyses were prioritized. Where no substantial differences were observed, the primary analyses were retained to maximize statistical power by including a greater number of studies.

For the health-related quality of life (HRQOL), pain, and function outcomes, standardized mean differences (SMDs) were used, as these outcomes were measured using different assessment tools (Western Ontario and McMaster Universities Arthritis Index (WOMAC), 36-item Short Form Health Survey questionnaire (SF-36), Hip Disability and Osteoarthritis Outcome Score (HOOS), and Knee Injury and Osteoarthritis Outcome Score (KOOS))46, 47, 48, 49, 50, 51, 52, 53 (Supplementary Table 7).⁵⁴ These were interpreted using Cohen’s guidelines for interpreting magnitude of SMD.⁵⁵ Subgroup analyses comparing the different assessment tools for HRQoL, pain, and function outcomes were conducted to investigate heterogeneity. To further enhance clinical relevance, the SMDs were re-expressed by calculating a pooled sample standard SD, which was then used to establish a mean difference and scale-specific pooled effect. The pooled sample SD was calculated using the studies included within each meta-analysis, whereas mean differences were calculated for outcomes TUG, 6MWT, ROM, and LOS.

If there were more than 2 intervention groups, the number of participants in the comparator group was divided by the number of intervention groups, and each was analyzed individually.⁵⁶ RCTs that reported mean changes and SD were included in the meta-analysis, and a pooled mean difference of the change was calculated. Similarly, if a different measure of variability was reported, the RevMan calculator (Version 5.4, The Cochrane Collaboration) was used to calculate the SD.^57,58 I² was reported to quantify heterogeneity and Τ² to report between study variances.⁴⁵ For studies that only reported scores at follow-up and not a change, these were included if baseline characteristics were similar.⁵⁶ Funnel plots were generated to evaluate small study effects. We conducted an unplanned post hoc analysis comparing outcomes by TKA and THA separately to determine whether pooling TKA and THA analyses together created substantial differences.

2.6.3. Intervention details

Intervention characteristics of each study were summarized using the TIDieR framework. The frequency of items reported was calculated to identify the most common components. Similarly, BCTs of each study were coded using a BCTTv1 table. Each row represented a study, and individual BCTs and their categorized groups were totaled to establish the most commonly used techniques. Assumptions used for coding the TIDieR and BCTTv1 tables are outlined in Supplementary Tables 8 and 9. In a non-prespecified post hoc analysis, a random-effects model meta-regression was performed using R Studio (Version 4.3.2, Posit, Boston, MA, USA) to determine whether the effectiveness of physical activity interventions was influenced by the number and type of BCTs present. The R packages used were meta (Version 7.0.0), metafor (Version 4.6.0), and dmetar (Version 0.1.0). Interventions that included the most common BCTs were compared to those without them on the following outcomes: HRQOL, function, pain, 6MWT, and TUG pre-surgery. The most common BCTs were: instruction on how to perform the behavior, self-monitoring of behavior, and demonstration of behavior.

3. Results

The search identified 8644 unique study titles and abstracts after removing duplicates. Following screening, 101 full-text articles were assessed for eligibility, with 33 meeting the inclusion criteria. Seven studies were included through hand-searching references and existing systematic reviews, resulting in 40 included studies (Fig. 1).

Fig. 1.

PRISMA diagram displaying the flow of study selection process. PRISMA = Preferred Reporting Items for Systematic reviews and Meta-Analyses.

3.1. Study characteristics

Forty RCTs were included (Supplementary Table 10), of which 12 were pilot/feasibility studies.^{28,29,31,33,34,}59, 60, 61, 62, 63, 64, 65 Twenty-four^{32,36,37,59,60,62,}64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81 studied total knee arthroplasty (TKA), 9 studies^28,29,82, 83, 84, 85, 86, 87, 88 investigated total hip arthroplasty (THA), and 7 studies^30,31,33, 34, 35^,61,63 reported both TKA and THA. The number of participants ranged from 10 to 262 (median n = 60). RCTs were conducted across 21 countries, with USA contributing the most studies (n = 6)^{30,60,62,67,74,75} followed by Australia (n = 4).^34,35,82,84 On average, the pre-operative exercise interventions were implemented 3 days a week for 6 weeks.

Participants (n = 2877) were on average 66.0 ± 7.7 (mean ± SD) years old. One RCT⁷⁸ recruited only female participants, otherwise the average percentage of females was 62%. The average body mass index (BMI) was 30.4 ± 5.2 kg/m².

3.2. Quality assessment

RoB was considered low in 10 RCTs,^29,31,33, 34, 35^{,59,63,73,80,83} there were some concerns in 23 studies,^{30,32,37,60,62,}64, 65, 66, 67, 68, 69, 70, 71, 72^,74, 75, 76, 77, 78, 79^,84,85,88 and high concern in 7 studies^{28,36,61,81,82,86,87} (Supplementary Fig. 1). The certainty of evidence varied across all outcomes. HRQOL pre-surgery and 6MWT post-surgery were rated as very low certainty. Meanwhile, pre-surgery pain, function, 6MWT, ROM, and post-surgery HRQOL and ROM were graded as low certainty. Post-surgery pain, function, and LOS were deemed to have moderate certainty. Notably, both pre- and post-surgery TUG was ranked high certainty (Supplementary Table 11). Visual interpretation of the funnel plots suggested that publication bias was present in both pre- and post-surgery 6MWT and ROM (Supplementary Figs. 2–17).

3.3. Meta-analysis

Overall, 29 RCTs were included in the meta-analyses. Six studies^28,29,82, 83, 84^,88 were THA, 18 studies^32,37,59,60,64, 65, 66, 67, 68, 69^,72,74, 75, 76^,78, 79, 80, 81 TKA, and 5 studies^30,31,33, 34, 35 included both. Eleven trials^36,61, 62, 63^{,70,71,73,77,}85, 86, 87 could not be included in the meta-analysis because they did not report data between 6 and 12 weeks post-surgery (Supplementary Table 12).

To increase statistical power, we combined TKA and THA in the meta-analyses for each outcome. Table 2 provides an overview summary of the pooled TKA and THA meta-analyses.

Table 2.

Summary of meta-analysis results, certainty of evidence, and publication bias for pre-operative exercise interventions in TKA and THA populations.

Outcome	n	SMD	95%CI	MD	95%CI	p	Τ2	I2(%)	Cohen's	GRADE	Publication bias
QOL PRE	20	–0.43	–0.75 to –0.11	–9.79	–17.07 to –2.50	0.009*	0.43	85	Small	Very low	No
QOL POST	19	–0.38	–0.61 to –0.14	–8.54	–13.71 to –3.15	0.002*	0.17	65	Small	Low	No
Pain PRE	23	–0.37	–0.63 to –0.11	–7.51	–12.79 to –2.23	0.005*	0.27	73	Small	Low	No
Pain POST	22	–0.25	–0.43 to –0.07	–6.68	–11.49 to –1.87	0.006*	0.08	44	Small	Moderate	No
Function PRE	23	–0.51	–0.77 to –0.24	–10.23	–15.45 to –4.82	0.0002*	0.29	74	Moderate	Low	No
Function POST	22	–0.49	–0.75 to –0.22	–14.42	–22.08 to –6.48	0.0003*	0.28	74	Small	Moderate	No
6MWT PRE	5	–	–	38.97	5.64 to 72.29	0.002*	1004.7	80	–	Low	Yesa
6MWT POST	4	–	–	–1.78	–47.60 to 44.05	0.94	1656.4	78	–	Very low	Yesa
TUG PRE	14	–	–	–1.16	–2.00 to -0.32	0.007*	1.25	69	–	High	No
TUG POST	11	–	–	–1.27	–1.71 to -0.82	<0.00001*	0.04	6	–	High	No
ROM Flexion PRE	12	–	–	3.52	0.60 to 6.45	0.02*	13.3	60	–	Low	Yes
ROM Flexion POST	11	–	–	3.11	–0.05 to 6.26	0.05*	17.98	66	–	Low	Yes
ROM Extension PRE	9	–	–	0.8	–2.17 to 3.77	0.6	18.52	93	–	Low	Yesa
ROM Extension POST	9	–	–	1.3	–0.40 to 3.00	0.13	4.98	79	–	Low	Yesa
LOS	6	–	–	–0.41	–0.89 to 0.07	0.09	0.01	3	–	Moderate	Noa

Note: MD and 95%CI for QoL, pain, and function are calculated using scale-specific pooled effect. Publication bias was assessed by visually interpreting funnel plots.

^⁎Significant difference.

^aLess than 10 studies.

Abbreviations: 6MWT = 6-min walk test; 95%CI = 95% confidence interval; GRADE = grading of recommendations assessment, development and evaluation; LOS = length of stay; MA = meta-analysis; MD = mean difference; POST = post-surgery; PRE = pre-surgery; QOL = quality of life; ROM = range of motion; SMD = standardized mean difference; THA = total hip arthroplasty; TKA = total knee arthroplasty; TUG = timed-up-and-go.

We conducted an unplanned post hoc analysis comparing outcomes by TKA and THA separately to determine whether pooling TKA and THA analyses together created substantial differences. Substantial differences were defined as meeting one of the following: change in SMD greater than 0.2; change in Cohen’s d threshold (i.e., from small to moderate); and 95% confidence intervals (95%CIs) do not overlap. Based on this criteria, QOL pre-surgery for both TKA and THA, pain post-surgery for THA and function post-surgery for both TKA and THA had substantial differences (Supplementary Table 13). Consequently, for these 5 analyses, results are reported for TKA and THA separately, while all the remaining outcomes are reported with TKA and THA pooled together.

The sensitivity analyses, which excluded skewed data, showed no substantial differences compared to the primary analyses reported below, regardless of whether the outcomes were reported as TKA and THA pooled, TKA-only, or THA-only. Therefore, the primary meta-analyses including all relevant studies are reported for all outcomes.

For transparency, Supplementary Tables 14–16 summarize the primary and sensitivity analyses for TKA and THA pooled and separately for all outcomes.

For the patient-reported outcomes (HRQOL, pain, and function), a summary table of the scale-specific pooled mean differences can be found in Supplementary Table 17.

3.3.1. 1. HRQOL

The SMD for TKA revealed statistically significant improvements in pre-surgery HRQOL favoring the intervention group (15 trials, SMD = −0.5, 95%CI: −1.0 to −0.1, I² = 87%, Τ² = 0.55, p = 0.01, moderate effect; pooled MD (mean difference) = −11.9, 95%CI: −21 to −2.7). Subgroup differences, which examined whether HRQOL outcomes varied depending on the measurement tool used, also indicated significant differences pre-surgery (p = 0.02). (Supplementary Fig. 18).

The SMD across three trials for THA favored the intervention group for HRQOL pre-surgery (SMD = −0.1, 95%CI: −1.1 to 0.9, I² = 89%, Τ² = 0.68, p = 0.89; pooled MD = −1.5, 95%CI: −22.7 to 19.8). Subgroup differences by HRQOL assessment tool were non-significant pre-surgery (p = 0.17) (Supplementary Fig. 19).

When pooled together, post-surgery HRQOL significantly improved, with small effect sizes favoring the intervention group (19 trials, SMD = −0.4, 95%CI: −0.6 to −0.1, I² = 65%, Τ² = 0.2, p = 0.002; pooled MD = −8.5, 95%CI: −13.7 to −3.1). Subgroup differences, examining whether HRQOL outcomes varied depending on the measurement tool used, indicated significant differences post-surgery (p = 0.04) (Supplementary Fig. 20).

3.3.2. Pain

When pooled, TKA and THA had significant improvements, with a small effect size for pain pre-surgery in favor of the intervention group (23 trials, SMD = −0.4, 95%CI: −0.6 to −0.1, I² = 73%, Τ² = 0.27, p = 0.005; pooled MD = −7.5, 95%CI: −12.8 to −2.2). No differences between subgroups were found (p = 0.38); however, some subgroups only included one study and therefore should be interpreted with caution (Supplementary Fig. 21).

Pain post-surgery for TKA had statistically significant improvements, with small effect sizes favoring the intervention group (16 trials, SMD = −0.3, 95%CI: −0.6 to −0.1, I² = 57%, Τ² = 0.15, p = 0.02; pooled MD = −5.8, 95%CI: −10.9 to −1.0). Subgroup differences did not signify significant differences post-surgery (p = 0.09), although one subgroup only included 1 study and therefore should be interpreted with caution (Supplementary Fig. 22).

Pain post-surgery among THA favored the intervention group (5 trials, SMD = −0.2, 95%CI: −0.4 to 0.1, I² = 0%, Τ² = 0, p = 0.31; pooled MD = −2.8, 95%CI: −7.9 to 2.6). Subgroup differences by pain assessment tool were not statistically significant post-surgery (p = 0.78) (Supplementary Fig. 23).

3.3.3. Function

Pooled TKA and THA function significantly improved, with moderate effect sizes in favor of the intervention group pre-surgery (23 trials, SMD = −0.5, 95%CI: −0.8 to −0.2, I² = 74%, Τ² = 0.29, p = 0.0002; pooled MD = −10.2, 95%CI: −15.4 to −4.8). No subgroup differences were found at pre-surgery (p = 0.32), although one subgroup only included one study and therefore should be interpreted with caution (Supplementary Fig. 24).

Post-surgery function significantly improved, with moderate effect sizes in favor of the intervention group for TKA (16 trials, SMD = −0.6, 95%CI: −1.0 to −0.2, I² = 79%, Τ² = 0.44, p = 0.003; pooled MD = −11.3, 95%CI: −18.7 to −4.2), and subgroup differences were statistically significant (p = 0.02). However, one subgroup included only one study and should be interpreted with caution (Supplementary Fig. 25).

The SMD for function post-surgery in THA had a small, non-statistically significant effect in favor of the intervention group (5 trials, SMD = −0.3, 95%CI: −0.6 to 0.1, I² = 21%, Τ² = 0.03, p = 0.13; pooled MD = −4.3, 95%CI: −9.8 to 1.2), and subgroup differences were not significant (p = 0.79) (Supplementary Fig. 26).

3.3.4. 6MWT

A mean difference of 39 m (95%CI: 5.6 to 72.3, I² = 80%, Τ² = 1004.7, p = 0.02) in favor of the intervention group pre-surgery was found; however, there was no difference post-surgery (MD = −1.8 m, 95%CI: −47.6 to 44.1, I² = 78%, Τ² = 1656.4, p = 0.94) (Supplementary Figs. 27 and 28).

3.3.5. TUG

There was a mean difference favoring the intervention group at both pre- (MD = −1.2 s, 95%CI: −2.0 to −0.3 s, I² = 69%, Τ² = 1.3, p = 0.007) and post-surgery (MD = −1.3 s, 95%CI: −1.7 to −0.8 s, I² = 6%, Τ² = 0.04, p < 0.00001) (Supplementary Figs. 29 and 30).

3.3.6. ROM

There were mean differences in favor of the intervention group for knee flexion pre-surgery (3.5°, 95%CI: 0.6°–6.5°, I² = 60%, Τ² = 13.3, p = 0.02) and post-surgery (MD = 3.1°, 95%CI: −0.1° to 6.3°, I² = 66%, Τ² = 18, p = 0.05) (Supplementary Figs. 31 and 32). However, the mean differences in knee extension did not indicate significant improvements pre- (MD = 0.8°, 95%CI: −2.2° to 3.8°, I² = 93%, Τ² = 18.5, p = 0.6) and post-surgery (MD = 1.3°, 95%CI: −0.4° to 3.0°, I² = 79%, Τ² = 5, p = 0.13) (Supplementary Figs. 33 and 34).

3.3.7 LOS

Six studies were included in the meta-analysis for LOS as they reported mean and standard deviation data in number of days whereas others provided odds ratios, log transformed data, a narrative sentence, or involved rehabilitation LOS not hospital.^{28,34,59,60,85,87} There was no significant mean difference in LOS (MD = −0.4, 95%CI: −0.9 to 0.1, I² = 3%, Τ² = 0.01, p = 0.09) (Supplementary Fig. 35).

3.4. Intervention characteristics

A summary of the intervention components of the included studies can be found in Supplementary Table 19, with full details in Supplementary Table 20. Items 10 (modifications) and 11 (how well planned) were not reported by the included studies, therefore there are no results on these 2 items. There were 44 interventions in the 40 studies as 4 RCTs tested 2 interventions.^33,78, 79, 80 The training types for item 4 (procedures) were categorized as aerobic, strength/resistance, mobility and flexibility, and balance. Interventions were coded into the categories based upon the terminology reported in the intervention description (i.e., resistance exercises, strengthening exercises, and aerobic) or exercise name (i.e., stationary bike, walking). One study⁶⁰ was aquatic, thus an aqua therapist and swimming pool were Items 5 (who) and 7 (where).

3.4.1. Why?

Only 6 interventions (14%) applied a theoretical framework to their exercise interventions. These were the Social Cognitive Theory (SCT),⁶⁷ DeLorme and Watkins training system,⁸⁰ Self-efficacy Theory,³⁶ and Theoretical Model of Prehabilitation.^74,75

3.4.2. What?

3.4.2.1. Materials

Four intervention materials were reported: information booklet, exercise program details, logbook, and instructor training. Thirty (68%) interventions^{28,29,31,33,35,36,}59, 60, 61^,63,64,66, 67, 68^,71,73, 74, 75, 76, 77, 78, 79, 80^,82,83,86,87 provided clear information for at least one of the four materials, while 6 interventions (14%)^{32,34,72,79,81,85} were unclear only, and 8 interventions (18%)^{30,37,62,65,69,70,84,88} did not specify any. Eight interventions (18%)^31,61,71,73,76, 77, 78 provided detailed interventions that were replicable. Sixteen (36%)32, 33, 34^{,63,64,66,67,72,74,75,79,80,82} provided some details but did not have explicit instructions on how to perform the exercises, classifying them as mostly replicable. Twenty (45%)28, 29, 30^,35, 36, 37^{,59,60,62,65,}68, 69, 70^,81,83, 84, 85, 86, 87, 88 were not replicable.

3.4.2.2. Procedures

With the exception of a single intervention,²⁸ type of training was specified. Most of the interventions (91%, n = 40) comprised a strength element. Six (14%)^{36,62,68,80,83,87} were strength only; 34 (77%)30, 31, 32, 33, 34, 35^,37,59, 60, 61^,63, 64, 65, 66, 67^,69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80^,82,84,85,88 were a combination of strength and another training type. The remaining three (7%)^29,81,86 that did not include strength specified flexibility/mobility training only. Intensity was explicitly reported by 22 (50%) interventions, 9 interventions^{30,60,63,67,72,74,78,88} were qualitative, 4 interventions^29,35,59,69 based upon percentage heart rate, 7 interventions^{35,37,62,64,80,83} calculated in relation to repetition maximum, and 2 interventions^29,61 using the rate of perceived exertion scale.

3.4.3. Who?

Most of the interventions (55%, n = 24) were delivered by a physiotherapist, 3 studies (7%)^36,67,71 were delivered by project staff/researchers, and 2 studies (5%)^35,73 were delivered by exercise physiologists. One intervention³² did not clearly state the deliverer, while 14 studies (32%)^{28,29,62,69,70,72,}74, 75, 76, 77^,80,84,87 did not specify the deliverers at all.

3.4.4. How?

Thirty-seven (84%) interventions provided details on how the intervention was delivered. All but 6 interventions (14%)^{36,70,71,78,79,85} involved supervision. Most interventions (57%, n = 25) involved in-person supervision, with 6 (14%)^{33,34,67,73,75,84} of those including additional unsupervised sessions. The remaining 6 interventions (14%)^{29,33,61,68,76,78} were remotely supervised at home using live phone/video calls.

3.4.5. Where?

Nineteen interventions (43%)33, 34, 35^{,59,63,64,66,67,73,74,}79, 80, 81, 82, 83, 84^,88 were delivered at a hospital or clinic, six (14%)^{34,67,73,81,82,84} of which included additional home-based exercise. Two interventions (5%)^30,62 were community-based, and 11 interventions (25%)^{29,33,36,61,68,70,71,75,76,78,85} were home-based only. Eleven interventions (25%)^{28,31,32,37,65,69,70,72,77,86,87} provided no information on where the interventions were held.

3.4.6. How much?

Thirty-eight interventions (86%) provided frequency information: 13 interventions (30%)^31,33, 34, 35^{,59,62,63,65,69,72,81,83} were one to two times a week, 17 interventions (39%)^{30,32,36,37,60,64,}66, 67, 68^{,70,75,79,80,82,84,85} were three to 4 times a week, and 8 interventions (18%)^{29,61,70,71,78,85,88} were 5–6 times a week.

Thirty-eight interventions (86%) provided information on how long the intervention occurred: 15 interventions (34%)^{28,32,34,59,62,64,66,68,76,}78, 79, 80 were less than or equal to 4 weeks; 1 intervention²⁹ was 3–6 weeks, and 1 intervention⁶⁰ was 4–8 weeks. Sixteen (36%)^{30,31,37,61,63,65,67,}69, 70, 71, 72^{,74,81,82,84,85} were 5–8 weeks, and 5 interventions (11%)^33,35,36,83 were 9–12 weeks.

3.4.7. Tailoring

Most of the interventions (59%, n = 26) involved some tailoring of the intervention. The remaining 18 interventions^{28,35,36,62,65,}70, 71, 72, 73, 74, 75, 76, 77^,79,85, 86, 87, 88 did not specify any tailoring.

3.4.8. How well?

Sixteen (36%) of the interventions provided information on intervention compliance. Two interventions (5%)^33,71 reported an unsatisfactory (<80%) compliance rate, and 14 interventions (32%)^{30,33,59,64,66,67,74,79,80,}82, 83, 84 reported a satisfactory (≥80%) compliance rate.

3.5. Behavior change techniques (BCTs)

Out of a possible 93 BCTs, 38 BCTs (40.9%) were used across interventions. Only 32 RCTs28, 29, 30, 31, 32, 33^,35,36,59, 60, 61, 62, 63^,65, 66, 67, 68^,70,71,73, 74, 75, 76^,78, 79, 80, 81, 82, 83^,85, 86, 87 had clearly coded at least 1 BCT, while 3 RCTs^77,84,88 were unclear and 5 RCTs^{34,37,64,69,72} did not report any BCTs (Supplementary Table 21 Parts 1 and 2).

The BCTs belonged to 14 of 16 (87.5%) possible groups. The most common group was “Shaping knowledge” (n = 20 clear, n = 6 unclear), followed by “Feedback and monitoring” (n = 20 clear, n = 4 unclear) and “Goals and planning” (n = 14 clear, n = 2 unclear). The most frequently coded BCTs were 4.1 “Instruction on how to perform the behavior” (n = 20 clear, n = 6 unclear), followed by 6.1 “Demonstration of behavior” (n = 9 clear, n = 3 unclear) and 2.3 “Self-monitoring of behavior” (n = 10 clear, n = 1 unclear). The presence and absence of these 3 BCTs within interventions were examined to assess their impact on HRQOL, pain, function, TUG, 6MWT, and ROM outcomes.

3.6. Meta-regression

The meta-regression models indicated that the number of BCTs and their effect on the interventions were not associated with each other (Supplementary Figs. 44–49). When we explored the most common BCTs and their effect on outcomes, the only significant BCT was “Instruction on how to perform the behavior” for function (p = 0.0067). Having this BCT as part of the intervention improved functional outcomes.

4. Discussion

4.1. Summary of main findings

Worldwide, healthcare systems are seeking ways to help people who have long waits for hip and knee surgery improve their HRQOL and surgical outcomes. This review demonstrated that exercise interventions improved HRQOL, pain, function, mobility, and distance walked pre-surgery in addition to HRQOL, pain, function, and mobility up to 12-week post-surgery. However, these findings should be interpreted with caution due to small to moderate effect sizes for HRQOL, pain, and function; varying RoBs and certainty of evidence across all outcomes; and potential publication bias identified for distance walked. Although these effects may appear minor, they are potentially clinically meaningful, especially in patients awaiting surgery. Such improvements could enhance patients’ ability to maintain independence, perform activities of daily living, and prevent weight gain during the waiting period. Moreover, when re-expressing the SMDs to pooled MDs for HRQOL, pain, and function, pre-operative exercise resulted in improvements of between 6.7% and 14.4%. This improvement could foster better recovery times and post-operative outcomes, offering benefits to both patients and healthcare providers. While this review highlights short-term improvements in HRQOL, pain, function, and mobility, the longer-term clinical relevance of these effects remains uncertain.

LOS was the most reported surgical-related measure but was only reported in 12 trials, with only 6 eligible for meta-analysis. Although a reduction of 0.4 days is modest, it may still have favorable effects such as cost savings and increased patient throughput, which can positively impact healthcare services and resources.⁸⁹ However, the worldwide implementation of Enhanced Recovery After Surgery (ERAS) protocols may confound these findings, potentially limiting clinical significance of LOS reductions attributable to prehabilitation.89, 90, 91, 92, 93 Complications were reported in 4 studies;^29,30,66,77 however, these lacked detail about the timing in relation to the post-operative period and did not provide sufficient statistical estimates, limiting comparability. Therefore, it was not possible to perform a meta-analysis or draw conclusive findings on complication rates or types.

Studies reported poorly on physical activity changes and adverse events; thus, no conclusions could be made. Most interventions were not based on theory and focused on strength/resistance training, with 25% including an additional aerobic component. They were primarily supervised and delivered by a physiotherapist in a hospital/clinic setting or at home and lasted up to 12 weeks with 1 to 6 exercise sessions per week. Although only one-third of studies measured and reported compliance, 88% of those studies achieved over 80% compliance. This suggests that patients who enroll in the programs have high attendance rates, which could suggest they see benefits in attending. The main BCTs of the interventions involved instruction, self-monitoring, and demonstration of behavior. This could imply that providing details on how to perform and record the desired behavior may enhance behavior adoption. Self-monitoring has been found to improve self-awareness and knowledge of behavior.⁹⁴ However, the only significant effect established was instruction on how to perform the behavior on functional outcomes, but this could be due to lack of reporting of other components. Additionally, when participants learn how to execute exercises correctly, they avoid compensatory movements and directly improve their function.

4.2. Clinical significance

While there are no established minimal clinically important difference (MCID) values for patient-reported HRQoL, pain, function, and performance-based outcomes TUG, 6MWT, and ROM in TKA and THA prehabilitation, MCIDs and minimal detectable changes (MDCs) of similar populations may be applicable. Given that WOMAC is specifically designed for osteoarthritis, we applied the MCIDs reported by MacKay et al.,⁹⁵ in their systematic review of MCIDs and Clement et al’s⁹⁶ study, to the scale-specific pooled MDs of HRQOL, pain, and function in this review. Function was the only outcome deemed clinically important, as the pre- and post-operative MDs of −10.2, −11.3, and −4.3 fall within the 1.8–34.0 range used by other studies to assess important clinical differences.⁹⁵ This is further supported by our TUG results, which also had favorable changes. Even though this objectively measured functional outcome had a MD of 1.2 s pre-surgery and 1.3 s post-surgery, it did not surpass those recommended by Yuksel et al.^97,98 of 1.62 and 2.27 s for post-operative THA and TKA, respectively.

The pre-operative mean difference of 39.0 m in the 6MWT exceeds the upper MCID range of 30.5 m in patients with pathology,⁹⁹ thus suggesting that exercise before surgery improves the functional aerobic capacity of patients awaiting TKA and THA. These findings are similar to a recent review on frail and high-risk patients that reported an improvement of 40 m in the 6MWT before surgery after engaging in prehabilitation.¹⁰⁰

The MDs in knee flexion (3.5° and 3.1°) and extension (0.8° and 1.3°) at pre- and post-surgery, respectively, did not meet the MDC at 90%CIs reported by Stratford et al.¹⁰¹ of 9.6° for flexion and 6.3° for extension post TKA surgery. This suggests that the change observed in knee range of motion is not clinically relevant.

Despite many of the outcomes not meeting the MCIDs of similar populations, the observed improvements and mean differences at both pre- and post-surgery are still likely to benefit patients and healthcare practices. For example, improved HRQOL, pain, and function may support patients’ tolerance and independence during the waiting period and lead to a favorable recovery trajectory resulting in an earlier discharge. However, it is important to interpret the clinical significance of these findings, given the challenges in defining MCID thresholds; small to moderate effect sizes; and methodological limitations of low to high risk of bias, very low to high certainty of evidence, and potential publication bias.

4.3. Comparison with other studies

Our findings differ from those of previous reviews, likely due to different methodologies. While we did not observe an improvement in LOS following pre-operative exercise, Almeida et al.¹² and Konnyu et al.¹⁶ did. However, both of these reviews narratively reported their results, whereas we meta-analyzed results, which can improve precision.⁴⁵ Furthermore, these reviews included both pre-operative exercise interventions as well as the broad, multimodal type of prehabilitation interventions.^12,16 In contrast, our review focused specifically on exercise, with some studies also incorporating education. By narrowing our inclusion to exercise alone, we may have reduced the potential confounding effects of the other prehabilitation components, which can be hard to isolate from narrative analyses.

Punnoose et al.¹⁵ conducted a systematic review and meta-analysis on prehabilitation interventions of patients undergoing orthopedic surgery (including spine surgery). They reported similar findings to our review despite not focusing specifically on exercise. However, some key differences were statistical discrepancies between pre-operative HRQoL and post-operative pain and function. We identified significant improvements in HRQOL for TKA pre-surgery but not for THA, whereas Punnoose et al.¹⁵ observed the opposite. Additionally, we found significant improvements in post-operative pain for TKA and non-significant changes in post-operative function for THA, while Punnoose et al.¹⁵ reported the contrary. The differences could be because we combined the two surgical procedures to increase statistical power by having greater sample sizes, and we included more than 1 measurement tool (i.e., WOMAC, SF-36, and HOOS/KOOS), whereas Punnoose et al.¹⁵ only analyzed SF-36. Notably, all our 95%CIs, except pain pre-surgery, fell within the ranges of Punnoose and colleagues.¹⁵ This could suggest that we had greater precision for these outcomes due to focusing specifically on exercise-based interventions.

Although there are no known meta-regression studies using BCTTv1 to enhance exercise in TKA and THA patients, two studies^102,103 identified significant BCTs for increasing physical activity. Michie et al.¹⁰² reported self-monitoring of behavior, while Carraça et al.¹⁰³ revealed goal setting (both behavior and outcome), graded tasks, social incentive, behavioral practice, and rehearsal to be significant. One potential explanation for the variation in findings could be that the interventions included in this review do not describe the BCTs within the interventions clearly, leading to subjective BCT coding by the reviewers. These differences, and the paucity of evidence, highlight the need for more research on which BCTs are effective components of exercise interventions in TKA and THA.

4.4. Strength and limitations

This extensive review evaluates the effectiveness of exercise-based interventions and their components, and involves 2877 participants from 44 interventions.28, 29, 30, 31, 32, 33, 34, 35, 36, 37^,59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88 Healthcare organizations can use this information to help plan and design future services and interventions. The data, derived from 21 countries, demonstrated diverse settings. Subgroup analyses were conducted to explore the observed heterogeneity, which revealed significant differences in the WOMAC measurement tool compared to the more general measures (i.e., SF36). This could suggest that WOMAC is more sensitive to change for this population. Another strength of this review is that an attempt to reduce external bias was achieved by calculating the pooled sample standard deviations for HRQoL, pain, and function using the samples analyzed. This enabled a more accurate and contextually relevant standardization of effect for these outcomes.¹⁰⁴

The limited detail of the included interventions restricted a comprehensive analysis of effective intervention components and further understanding of heterogeneity. Similarly, the follow-up period included in this review was only 6–12 weeks post-surgery. Therefore, insight and analysis of how the changes in outcomes continue in the medium- and long-term could not be made. Furthermore, as the interventions did not clearly specify BCTs, the coding of BCTs was subjective, although BCTs were extracted by 2 independent reviewers to reduce bias and improve reliability.

We were unable to synthesize data on physical activity, surgical complications, and adverse events due to data not being reported. While this review focused on physical activity and exercise, almost 40% of the interventions included other components: education (n = 14, 32%), blood-flow restriction (n = 2, 4.5%), and counseling (n = 1, 2%), which may have influenced outcomes. Over 25% of trials were pilot/feasibility trials, and only 10 were deemed a low risk of bias. Additionally, the GRADE certainty ratings varied from very low to high, with most outcomes being low. Both of which imply that there is limited high quality evidence. Lastly, the small sample sizes of the studies included in the meta-analyses could result in sampling error bias and uncertainty in heterogeneity, highlighting the need to interpret the results with caution.^45,105,106

4.5. Unanswered questions and future research

For more robust comparisons and clinical application, pre-operative exercise interventions and the outcome measures used to assess their effectiveness for TKA and THA need to be established. Achieving this requires standardizing interventions and outcomes, identifying relevant BCTs and developing MCIDs specific to this population. The current review found no observational studies on the association of physical activity before surgery and surgical outcomes, meanwhile few RCTs reported any surgical outcomes, preventing any substantial analysis. Consequently, future research trials should consistently investigate surgical outcomes to allow commissioners to understand the impact exercise before THA and TKA has on key surgical outcomes, such as readmission rates, recovery time, and the cost-effectiveness of these interventions. Additionally, more research is needed to determine whether the improvements in HRQOL, pain, function, and mobility were maintained for meaningful long-term benefit.

5. Conclusion

While the direction of effects for all outcomes pre- and post-surgery favored exercise interventions in THA and TKA patients, the findings are limited by some outcomes having a small number of studies, varying levels of RoBs, and certainty of evidence. That said, there is clinically relevant evidence that exercise interventions before THA and TKA can improve HRQOL, pain, function, and distance walked prior to surgery. However, the effect on the surgical procedure and recovery is unknown due to limited measurements of relevant outcomes. Most of the interventions consisted of strength-based exercises, and instruction of how to perform these exercises is important and should be included in future interventions.