Cryotherapy following total knee replacement

Abstract

Background

Total knee replacement (TKR) is a common intervention for people with end‐stage symptomatic knee osteoarthritis, resulting in significant improvements in pain, function and quality of life within three to six months. It is, however, acutely associated with pain, local oedema and blood loss. Post‐operative management may include cryotherapy. This is the application of low temperatures to the skin surrounding the surgical site, through ice or cooled water, often delivered using specialised devices. This is an update of a review published in 2012.

Objectives

To evaluate the effect of cryotherapy in the acute phase after TKR (within 48 hours after surgery) on blood loss, pain, transfusion rate, range of motion, knee function, adverse events and withdrawals due to adverse events.

Search methods

We searched CENTRAL, MEDLINE, Embase, six other databases and two trials registers, as well as reference lists, related links and conference proceedings on 27 May 2022.

Selection criteria

We included randomised controlled trials or controlled clinical trials comparing cryotherapy with or without other treatments (such as compression, regional nerve block or continuous passive motion) to no treatment, or the other treatment alone, following TKR for osteoarthritis.

Data collection and analysis

Two review authors independently selected studies for inclusion, extracted data and assessed risk of bias and certainty of evidence using GRADE. We discussed any disagreements and consulted another review author to resolve them, if required. Major outcomes were blood loss, pain, transfusion rate, knee range of motion, knee function, total adverse events and withdrawals from adverse events. Minor outcomes were analgesia use, knee swelling, length of stay, quality of life, activity level and participant‐reported global assessment of success.

Main results

We included 22 trials (20 randomised trials and two controlled clinical trials), with 1839 total participants. The mean ages reflected the TKR population, ranging from 64 to 74 years.

Cryotherapy with compression was compared to no treatment in four studies, and to compression alone in nine studies. Cryotherapy without compression was compared to no treatment in eight studies. One study compared cryotherapy without compression to control with compression alone. We combined all control interventions in the primary analysis.

Certainty of evidence was low for blood loss (downgraded for bias and inconsistency), pain (downgraded twice for bias) and range of motion (downgraded for bias and indirectness). It was very low for transfusion rate (downgraded for bias, inconsistency and imprecision), function (downgraded twice for bias and once for inconsistency), total adverse events (downgraded for bias, indirectness and imprecision) and withdrawals from adverse events (downgraded for bias, indirectness and imprecision). The nature of cryotherapy made blinding difficult and most studies had a high risk of performance and detection bias.

Low‐certainty evidence from 12 trials (956 participants) shows that cryotherapy may reduce blood loss at one to 13 days after surgery. Blood loss was 825 mL with no cryotherapy and 561 mL with cryotherapy: mean difference (MD) 264 mL less (95% confidence interval (CI) 7 mL less to 516 mL less).

Low‐certainty evidence from six trials (530 participants) shows that cryotherapy may slightly improve pain at 48 hours on a 0‐ to 10‐point visual analogue scale (lower scores indicate less pain). Pain was 4.8 points with no cryotherapy and 3.16 points with cryotherapy: MD 1.6 points lower (95% CI 2.3 lower to 1.0 lower).

We are uncertain whether cryotherapy improves transfusion rate at zero to 13 days after surgery. The transfusion rate was 37% with no cryotherapy and 79% with cryotherapy (risk ratio (RR) 2.13, 95% CI 0.04 to 109.63; 2 trials, 91 participants; very low‐certainty evidence).

Low‐certainty evidence from three trials (174 participants) indicates cryotherapy may improve range of motion at discharge: it was 62.9 degrees with no cryotherapy and 71.2 degrees with cryotherapy: MD 8.3 degrees greater (95% CI 3.6 degrees more to 13.1 degrees more).

We are uncertain whether cryotherapy improves function two weeks after surgery. Function was 75.4 points on the 0‐ to 100‐point Dutch Western Ontario and McMaster Universities Arthritis Index (WOMAC) scale (lower score indicates worse function) in the control group and 88.6 points with cryotherapy (MD 13.2 points better, 95% CI 0.5 worse to 27.1 improved; 4 trials, 296 participants; very low‐certainty evidence).

We are uncertain whether cryotherapy reduces total adverse events: the risk ratio was 1.30 (95% CI 0.53 to 3.20; 16 trials, 1199 participants; very low‐certainty evidence). Adverse events included discomfort, local skin reactions, superficial infections, cold‐induced injuries and thrombolytic events.

We are uncertain whether cryotherapy reduces withdrawals from adverse events (RR 2.71, 95% CI 0.42 to 17.38; 19 trials, 1347 participants; very low‐certainty evidence).

No significant benefit was found for secondary outcomes of analgesia use, length of stay, activity level or quality of life. Evidence from seven studies (403 participants) showed improved mid‐patella swelling between two and six days after surgery (MD 7.32 mm less, 95% CI 11.79 to 2.84 lower), though not at six weeks and three months after surgery. The included studies did not assess participant‐reported global assessment of success.

Authors' conclusions

The certainty of evidence was low for blood loss, pain and range of motion, and very low for transfusion rate, function, total adverse events and withdrawals from adverse events. We are uncertain whether cryotherapy improves transfusion rate, function, total adverse events or withdrawals from adverse events. We downgraded evidence for bias, indirectness, imprecision and inconsistency. Hence, the potential benefits of cryotherapy on blood loss, pain and range of motion may be too small to justify its use. More well‐designed randomised controlled trials focusing especially on clinically meaningful outcomes, such as blood transfusion, and patient‐reported outcomes, such as knee function, quality of life, activity level and participant‐reported global assessment of success, are required.

Plain language summary

Cold therapy following total knee replacement surgery

What are the benefits and risks of cold therapy after total knee replacement?

Key messages

Compared to placebo, cold therapy may improve blood loss, pain, knee range of motion and short‐term swelling after total knee replacement (TKR). We are less certain of its effect on blood transfusions, knee function, pain relief, length of hospital stay, quality of life or activity level. Though evidence was limited, there was little concern for serious adverse events with cold therapy.

What is osteoarthritis, and how is it treated?

Osteoarthritis is a degenerative disease of the joints, such as the knee. Osteoarthritis of the knee can cause pain, limit function and worsen quality of life. TKR can help this condition in the long term, but the effects of surgery during the recovery period (up to 6 months after surgery) can leave people weakened and impaired. Cold therapy (cryotherapy) involves the application of low temperatures to the skin surrounding an injury or surgical site. This can be done using bags of ice or specialised devices that deliver cooled water to the area.

What did we want to find out?

We wanted to find out if cryotherapy has effects on blood loss, pain and knee function within 48 hours after TKR.

What did we do?

We searched for studies that investigated cryotherapy compared with placebo in people after TKR. We compared and summarised the results of the studies and rated our confidence in the evidence, based on factors such as study methods and sizes.

What did we find?

We included 22 trials where people undergoing TKR received any form of cold therapy (with or without other treatments) and were compared with those not receiving any cold therapy. There were a total of 1839 people, aged between 64 and 74 years old. The outcomes of interest were in the acute phase (within 48 hours of surgery), but some studies included up to 12 weeks of follow‐up.

Main results

Blood loss

Blood loss was 264 mL less with cryotherapy at up to 13 days after surgery.

• People lost 561 mL of blood with cryotherapy.

• People lost 825 mL of blood without cryotherapy.

Pain (lower scores mean less pain)

Pain was better by 1.6 points on a 0‐ to 10‐point scale with cryotherapy at 2 days after surgery.

• People who had cryotherapy rated their pain as 3.2 points.

• People who had no cryotherapy rated their pain as 4.8 points.

Blood transfusion

42% more people had a blood transfusion with cryotherapy, or 42 more out of 100, at up to 13 days after surgery.

• 79 out of 100 people had a transfusion with cryotherapy.

• 37 out of 100 people had a transfusion without cryotherapy.

Knee range of motion

Flexion (bending of the knee joint) was 8.3 degrees greater with cryotherapy when people left hospital.

• People who had cryotherapy had 71.2 degrees of flexion.

• People who had no cryotherapy had 62.9 degrees of flexion.

Knee function

Knee function was 13.2 points better on a 0‐ to 100‐point scale with cryotherapy at 2 weeks after surgery.

• People who had cryotherapy had a function score of 88.6.

• People who had no cryotherapy had a function score of 75.4

Total adverse events

0% more people reported adverse events with cryotherapy, or 0 more out of 100, up to 30 days after surgery.

• 2.7 out of 100 people reported adverse events with cryotherapy.

• 2.1 out of 100 people reported adverse events without cryotherapy.

Withdrawals due to adverse events

0% more people withdrew from the study due to adverse events with cryotherapy, or 0 more out of 100, up to 30 days after surgery.

• 0.4 out of 100 people withdrew due to adverse events with cryotherapy.

• 0.2 out of 100 people withdrew due to adverse events without cryotherapy.

What are the limitations of the evidence?

We have little confidence in the evidence showing that cold therapy may slightly improve blood loss, pain and range of motion after surgery. We are uncertain if it lowers the risk of a blood transfusion, improves knee function, increases the risk of adverse events or contributed to withdrawals due to adverse events. Factors that decreased our confidence include flaws in the study design (participants were not assigned to treatments randomly; some participants dropped out of the study; participants could tell what treatment they were receiving), not having enough studies or participants to be certain about the results, and variations between studies in results and methods.

How up to date is the evidence?

The evidence is current to 27 May 2022.

Summary of findings

Summary of findings 1. Any cold therapy compared with no cold therapy following total knee replacement.

Any cold therapy compared with no cold therapy following total knee replacement
Patient or population: people undergoing total knee replacement Settings: post‐operative, either in hospital or in the community Intervention: any cold therapy with or without other treatment for pain or swelling (compression, regional nerve block or continuous passive motion) Comparison: no treatment; or the other treatment alone
Outcomes	Illustrative comparative risks* (95% CI)		Difference	Relative effect (95% CI)	No. of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	No cold therapy	Any cold therapy
Blood loss Time point: mean 48 hours, range 1 to 13 days. Higher blood loss indicates a worse outcome.	The mean blood loss from the control group in Smith 2002 was 824.7 mL.	The mean blood loss in the intervention groups was 560.7 mL.	264 mL less (7 to 516 mL less)	‐	956 (12 studies)	⊕⊕⊝⊝ Lowa,b	Cryotherapy may improve blood loss after TKR. This may or may not be clinically significant.c The SMD was 0.72 standard deviations lower in the cryotherapy group (1.41 to 0.02 lower).d
Pain at post‐operative day 2 Scale from 0 (best) to 10 (worst). Follow‐up: 2 days	The mean pain at post‐operative day 2 in the control group in Smith 2002 was 4.8 points.	The mean pain at post‐operative day 2 in the intervention groups was 3.16 points.	1.64 points lower (2.28 to 1.00 lower)	‐	530 (6 studies)	⊕⊕⊝⊝ Lowa	Cryotherapy may improve pain two days after TKR. This may or may not be clinically significant.e
Transfusion rate Follow‐up: 0 to 13 days	370 per 1000	788 per 1000 (15 to 1000)	418 more per 1000 (355 fewer to 40,146 more)	RR 2.13 (0.04 to 109.63)	91 (2 studies)	⊕⊝⊝⊝ Very lowa,b,f	We are uncertain whether cryotherapy improves transfusion rate after TKR.
Range of motion at discharge Scale from 0 (worst) to 125 (best). Follow‐up: mean 9 days	The range of motion at discharge in the control group in Kullenberg 2006 was 62.9 degrees.	The mean range of motion at discharge in the intervention groups was 71.24 points.	8.34 degrees higher (3.57 higher to 13.12 higher)	‐	174 (3 studies)	⊕⊕⊝⊝ Lowa,g	Cryotherapy may improve range of motion at discharge after TKR. This may or may not be clinically significant.h
Function Standardised scale with positive score indicating a better outcome. Scale from 0 (worst) to 100 (best). Follow‐up: 2 weeks	The function score in the control group in Thijs 2019 was 75.4.	The mean function score in the intervention group was 88.58 points.	13.18 points better (0.55 worse to 27.08 better)	‐	296 (4 studies)	⊕⊝⊝⊝ Very lowa,b	We are uncertain whether cryotherapy improves function after TKR. The SMD was 0.72 standard deviations higher in the cryotherapy group (0.03 lower to 1.48 higher).i
Total adverse events Adverse events included discomfort, local skin reactions, superficial infections, cold‐induced injuries and thrombolytic events. Follow‐up: 0 to 30 days	21 per 1000	27 per 1000 (11 to 66)	6 more per 1000 (10 fewer to 45 more)	RR 1.30 (0.53 to 3.20)	1199 (16 studies)	⊕⊝⊝⊝ Very lowa,f,g	We are uncertain whether cryotherapy improves total adverse events after TKR.
Withdrawals due to adverse events Follow‐up: 0 to 30 days	2 per 1000	4 per 1000 (1 to 26)	2 more per 1000 (1 fewer to 25 more)	RR 2.71 (0.42 to 17.38)	1347 (19 studies)	⊕⊝⊝⊝ Very lowa,f,g	We are uncertain whether cryotherapy improves withdrawals due to adverse events after TKR.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (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; TKR: total knee replacement; RR: risk ratio; SMD: standardised mean difference
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.

^aDowngraded for high risk of bias (two levels if a subjective outcome): selection, performance, detection, reporting or other bias
^bDowngraded for inconsistency: significant unexplained heterogeneity (96% for blood loss, 87% for transfusion rate, 89% for function)
^cThe clinically significant value for blood loss is 300 mL.
^dWe converted the SMD to a clinically meaningful measure using Smith 2002 (SD 366 mL).
^eThe clinically significant value for pain is 1.5 points on a visual analogue scale.
^fDowngraded for imprecision: wide 95% CI for intervention
^gDowngraded for indirectness: limitations in outcome measurement (diversity in definition and timing of measurement)
^hThe clinically significant value for range of motion is 7 degrees.
ⁱWe converted the SMD to a clinically meaningful measure using Thijs 2019 (SD 18.3).

Background

Description of the condition

Total knee replacement (TKR) is a common intervention for people with end‐stage symptomatic osteoarthritis of the knee, and results in significant improvements in pain, function and quality of life within three to six months of surgery. In the acute phase, however, TKR is associated with pain, local oedema and blood loss. Since the inception of the knee prosthesis in 1938, total knee replacement (TKR) has become the mainstay for surgical management of end‐stage symptomatic arthritis of the knee (Felson 2000). Large improvements in pain, mobility, function and health‐related quality of life are reported within three to six months after surgery, and may be evident for seven to fifteen years (Bourne 2007; Harris 1990; Hawker 1998; Sorrells 2004). Despite the significant benefits in the sub‐acute phase, TKR is associated with pain (Jamison 1997), local oedema and blood loss in the acute phase (Berman 1988; Sehat 2004). These outcomes are important to manage as they have implications for analgesia requirements, joint mobility, transfusion requirements, length of stay and early functional recovery.

Description of the intervention

Post‐operative management of TKR may include cryotherapy. This involves the application of low temperatures to the skin surrounding the surgical site, by means of ice or cooled water, often delivered using specialised devices.

Recent systematic reviews on cryotherapy after acute soft tissue injury or surgical procedure found evidence for a reduction in pain and swelling in some circumstances (Bleakley 2004), but these findings were limited by the quality of included studies. A meta‐analysis of cryotherapy after anterior cruciate ligament reconstruction found small reductions in post‐operative pain (Raynor 2005). While cold discomfort and wound problems have been reported, cryotherapy is a relatively safe and cheap intervention. Commercially‐available devices such as the Cryo/Cuff (Aircast, Vista, California (CA) USA), which combine cooling and compression, cost around USD (US dollars) 100, but may be labour‐intensive, as ice exchange is required regularly. More sophisticated automated devices are available but are substantially more expensive. There is also evidence that patient perception of cryotherapy is positive, although this may not be shared by health professionals (Lombardi 1993).

How the intervention might work

Cryotherapy involves the application of low temperatures to the skin surrounding an injury or surgical site, by means of bags of ice or cooled water, with the aim of minimising the extent of tissue trauma. The cold penetrates the soft tissues and, when applied over a joint, reduces the internal temperature of the joint (intra‐articular temperature) (Martin 2002), thereby slowing the conduction of nerve signals (Abramson 1966). A relatively small reduction in tissue temperature of 3 to 4 degrees Celsius substantially reduces inflammatory enzyme activity (Harris 1974), thereby reducing the transmission of noxious signals and the inflammatory response, and theoretically affecting perceived pain and local swelling (Matsen 1975). Cold temperatures also induce vasoconstriction, reducing blood flow, and, theoretically, blood loss following injury or surgery (Knight 1995).

Why it is important to do this review

There is considerable practice variation in the use of cryotherapy following TKR. This may be due to conflicting evidence from existing clinical trials. Cryotherapy after TKR is part of standard care in some healthcare facilities in the United Kingdom and Australia yet is infrequently used in others (Barry 2003; Naylor 2006). While this practice variation may be affected by the local needs of patient populations and resource disparities between countries, it is likely underpinned by conflicting evidence from several randomised controlled trials regarding the value of cryotherapy. This is an update of a review published in 2012 (Adie 2012).

Objectives

To evaluate the effect of cryotherapy in the acute phase after TKR (within 48 hours after surgery) on blood loss, pain, transfusion rate, range of motion, knee function, adverse events and withdrawals due to adverse events.

Methods

Criteria for considering studies for this review

Types of studies

Studies were eligible if they were full‐text, published, randomised controlled trials or controlled clinical trials (including those that used pseudo‐randomised methods of treatment allocation). There were no restrictions on date or language of publication.

Types of participants

We included adults who underwent elective primary TKR of any type (i.e. independent of prosthesis type, patella replacement or releases performed). We included revision TKR and unicondylar knee replacement only if clinical trials listed these as distinct events (i.e. data from these patients were provided separately to data from primary TKR patients). We included data from TKRs indicated for osteoarthritis.

Types of interventions

We defined cryotherapy as the application of cold temperature to the skin around the knee following surgery. This was by means of ice packs, cooling pads or other commercial devices, within the peak inflammatory period of 48 hours after surgery (Hughes 2010; Wirtz 2000).

We included trials comparing:

• cryotherapy to no treatment;

• cryotherapy to other treatments that aim to reduce pain and swelling (e.g. compression bandaging, regional nerve block or continuous passive motion);

• cryotherapy plus another therapy that aims to reduce pain and swelling compared with the other treatment alone.

We combined all three control interventions in the primary analysis.

Types of outcome measures

Major outcomes

• Blood loss: measured post‐operatively as the output volume from the surgical drain, haemoglobin drop from pre‐operative to post‐operative values, or by other validated methods that estimate total body blood loss

• Pain: measured post‐operatively using a subjective pain scale such as the visual analogue scale

• Transfusion rate: measured as the proportion of participants requiring a blood transfusion during the acute stay in hospital following surgery

• Range of motion: measured as the number of degrees of flexion and extension of the knee*

• Function: measured using a validated knee score

• Total adverse events

• Withdrawals due to adverse events

Minor outcomes

• Analgesia use: measured as the total dose of analgesia medication received following surgery

• Knee swelling: measured as the circumference around the knee*

• Length of stay: measured as the number of days the participant spent in an acute health service following surgery

• Quality of life: measured using a validated quality of life score

• Activity level: measured using a pedometer, accelerometer or patient self‐reported diary/questionnaire

• Participant‐reported global assessment of success

* We included any method of measurement in this review. We also recorded any methods trialists used to reduce bias (e.g. tests of inter‐observer reliability, or use of unmarked goniometers/measuring tape).

Timing of outcome assessment

We extracted blood loss at 24 hours, 48 hours and combined; pain at post‐operative day 1, day 2, day 3, up to 6 weeks and up to 12 weeks; transfusion rate at 24 hours; range of motion at week 1, week 2, discharge and 3 months; function at 2 weeks, 6 weeks and 3 months; adverse events and withdrawals from adverse events at end of study.

Search methods for identification of studies

Electronic searches

We searched the following databases on 27 May 2022: Cochrane Central Register of Controlled Trials (CENTRAL); MEDLINE via Ovid; Embase via Ovid; the Cochrane Library which included the Cochrane Database of Systematic Reviews (CDSR); Database of Abstracts of Reviews of Effects (DARE); the Health Technology Assessments (HTA) database; Cumulative Index to Nursing and Allied Health Literature (CINAHL) via EBSCO; Physiotherapy Evidence Database (PEDro); and Web of Science. We also searched reference lists, related links, conference proceedings and trial registers, including ClinicalTrials.gov (https://clinicaltrials.gov/) and the World Health Organization (WHO) International Clinical Trials Registry Platform (www.who.int/trialsearch/Default.aspx), for published protocols of clinical trials. The text of the electronic search strategy is available in Appendix 1. For this review update, we searched the same sources and used the same search strategy as in Adie 2012.

Searching other resources

We reviewed the reference lists of all included articles for eligible studies. We used the 'related links' feature in PubMed to check for eligible studies.

We reviewed the proceedings of the following meetings for eligible studies: American Academy of Orthopedic Surgeons Annual Meeting, British Orthopaedic Association Annual Congress, American Association of Hip and Knee Surgeons Annual Meeting, American Physical Therapy Association Annual Conference and the World Confederation of Physical Therapy International Congress. If we found an eligible abstract, we contacted the authors to determine whether the study had been published as a full‐text article (including in press, in a non‐indexed journal, or as a thesis).

Data collection and analysis

Selection of studies

Two review authors (SA, AA) independently assessed the titles, abstracts and keywords of identified studies for eligibility. Differences were discussed and disagreements were resolved by another review author (IAH, JN), if required. We reviewed the full texts of all suitable studies to determine eligibility for inclusion, and we noted the reasons for excluding studies.

Data extraction and management

Two review authors (SA, AA) independently extracted data from included studies using a data extraction form. Differences were discussed and disagreements were arbitrated by another review author (IAH, JN), if required.

We extracted the following study characteristics.

• Methods: study design, total duration of study, number of study centres and location, study setting, withdrawals and date of study.

• Participants: N, mean age, age range, sex, disease duration, severity of condition, diagnostic criteria, important baseline data; inclusion and exclusion criteria.

• Interventions: indications for TKR, the nature of the intervention and the control.

• Outcomes: major and minor outcomes specified and collected, and time points reported.

• Characteristics of the design of the trial, as outlined below in the Assessment of risk of bias in included studies section.

One review author (AA) transferred data into the Review Manager 5 file (RevMan Web 2020). We checked that data were entered correctly by comparing the data presented in the systematic review with the study reports.

For efficacy outcomes, we extracted data from the end of the intervention time point. If multiple time points were given, we presented the data closest to the end of the intervention. We extracted adverse event outcomes at the last time point (i.e. proportion who had an event by the end of the trial). We extracted the final values if both final values and change‐from‐baseline values were reported for the same outcome. If unadjusted and adjusted values for the same outcome were reported, we extracted unadjusted values for data collection. We extracted intention‐to‐treat samples for all outcomes for the meta‐analysis.

Assessment of risk of bias in included studies

Two review authors (SA, AA) independently assessed the risk of bias in included studies. Differences were discussed and disagreements were arbitrated by another review author (IAH, JN), if required. We performed risk of bias assessment in accordance with the guidelines set out in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). We assessed the risk of bias according to the following domains:

• random sequence generation;

• allocation concealment;

• blinding of participants and personnel;

• blinding of outcome assessment (self‐reported and objective outcomes);

• incomplete outcome data;

• selective outcome reporting;

• other bias: studies which included data from bilateral TKRs without adjusting for lack of independence.

We graded each potential source of bias as high, low or unclear risk, and provided a quote from the study report, together with a justification for our judgement, in the risk of bias table. We summarised the risk of bias judgements across different studies for each of the domains listed. We considered blinding separately for different key outcomes where necessary (e.g. for unblinded outcome assessment, risk of bias for blood loss may be different from a self‐reported pain scale). We also considered the impact of missing data by key outcomes.

We presented the figures generated by the risk of bias (RoB 1) tool to provide summary assessments of the risk of bias.

Measures of treatment effect

For continuous outcomes, we calculated the mean and standard deviation (SD) of the intervention and control groups. Where studies used identical instruments and units of measurement to measure an outcome, we calculated the mean difference (MD). Where studies used different instruments to measure an outcome, we calculated the standardised mean difference (SMD). We then back‐translated SMDs to standard units for blood loss (mL), function (Dutch Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)) and quality of life (SF‐12) by multiplying the SMD by the baseline SD in the control group from Smith 2002, Thijs 2019 and Yuksel 2022, respectively. We calculated 95% confidence intervals (CIs) for all outcomes. We used the minimal clinically important difference (MCID) of 1.5 points on a 10‐point scale for pain, and 10 points on a 100‐point scale for function or disability (Tubach 2012). We calculated the absolute percent change from the difference in the risks between the intervention and control group using GRADEpro (GRADEpro GDT).

Unit of analysis issues

When a single trial reported multiple arms, we included only the relevant arms. For studies including data from bilateral TKRs without adjusting for a lack of independence, we conducted a sensitivity analysis to explore the effect of these studies on the primary analysis. When comparing more than two intervention groups to the same control group, we halved the number in the control group to prevent duplicate examination of control participants (Albrecht 1997; Ivey 1994). When a study was performed across two phases, each with the same intervention but with a different set and number of participants and results, we analysed each phase separately (Healy 1994).

Dealing with missing data

We attempted to contact study authors for additional data when required. We contacted (or attempted to contact) all authors twice, but none provided additional information. Where studies reported outcomes with incomplete follow‐up, we extracted data using the total number of participants initially assigned to that group.

Where standard deviations were not reported for continuous outcomes, we calculated them using an available standard error, P value or 95% CI, according to the methods recommended in Chapter 6 of the Cochrane Handbook for Systematic Reviews of Intervention (Higgins 2020). If measures of uncertainty were not accessible in full‐text publications, and study authors did not respond to contact, we imputed the missing standard deviations using available data.

Assessment of heterogeneity

We assessed clinical and methodological diversity in terms of participants, interventions, outcomes and study characteristics (e.g. study design, outcome measurement tools, etc.) for the included studies to determine whether a meta‐analysis was appropriate. We conducted this assessment by observing these data from the Characteristics of included studies table. We assessed statistical heterogeneity by visually inspecting forest plots to assess the direction and magnitude of effects and the degree of overlap between CIs.

We used the I² statistic to quantify inconsistency amongst the trials in each analysis. We also considered the P value from the Chi² test.

We used this approximate guide for the interpretation of the I² value (Deeks 2020):

• 0% to 40% might 'not be important';

• 30% to 60% may represent 'moderate' heterogeneity;

• 50% to 90% may represent 'substantial' heterogeneity;

• 75% to 100% represents 'considerable' heterogeneity.

We bore in mind that the observed value of I² depends on: (i) magnitude and direction of effects, and (ii) strength of evidence for heterogeneity.

When there was a P value of 0.10 or less, we interpreted this to mean that the Chi² test indicated evidence of statistical heterogeneity.

If we identified substantial heterogeneity, we reported it and investigated possible causes by following the recommendations in section 10.10 of the Cochrane Handbook (Deeks 2020).

Assessment of reporting biases

We planned to create and examine a funnel plot to explore possible small study biases only when 10 or more studies reported on the same outcome and comparison. In interpreting funnel plots, we planned to examine the different possible reasons for funnel plot asymmetry, and relate this to the results of the review. If we were able to pool more than 10 trials, we planned to undertake formal statistical tests to investigate funnel plot asymmetry, and to follow the recommendations in section 10.4 of the Cochrane Handbook for Systematic Reviews of Interventions (Page 2020).

To assess outcome reporting bias, we checked trial protocols against published reports. For studies published after 1 July 2005, we screened the WHO International Clinical Trials Registry Platform for the prospective trial protocol. We evaluated whether selective reporting of outcomes was present.

Data synthesis

We pooled results when there were at least two studies for each outcome and performed meta‐analysis comparing cryotherapy against each control intervention, with all control interventions combined. We only combined data from participants undergoing primary TKR. We used the random‐effects model as the default method for pooling data. We created forest plots for each outcome at different time points, organised by subgroups.

Subgroup analysis and investigation of heterogeneity

We investigated heterogeneity by performing the following subgroup analyses.

• Different types of intervention and control. We divided studies into intervention with cryotherapy alone, or combined with a form of compression bandaging, versus a control with compression bandaging, or no compression. Compression dressings (often known as the Robert Jones bandage) have long been used after TKR (Brodell 1986), and are thought to reduce post‐operative blood loss and local swelling.

• Different doses of the intervention, measured as different temperatures and frequencies of cryotherapy/control. A previous systematic review (MacAuley 2001), and a randomised controlled trial of cryotherapy for soft tissue injuries, found that different doses of cryotherapy may affect outcomes (Bleakley 2006).

• Use of cemented versus uncemented knee prostheses on blood loss. The use of cement during surgery is thought to reduce blood loss (Porteous 2003).

• Use of continuous passive motion (CPM) versus no CPM on knee range of motion. CPM is widely used as a rehabilitation strategy following TKR. However, a previous systematic review found only marginal benefits for post‐operative knee range of motion (Harvey 2010).

We used the formal test for subgroup interactions in Review Manager 5, and exercised caution in the interpretation of subgroup analyses. We compared the magnitude of the effects between the subgroups by assessing the overlap of the CIs of the summary estimate.

Sensitivity analysis

We planned sensitivity analyses using a fixed‐effect model to assess:

• whether effect estimates were substantially different when missing standard deviations were not imputed; and

• whether effect estimates were different if studies with unclear or inadequate allocation concealment were removed.

Summary of findings and assessment of the certainty of the evidence

At least two review authors assessed the certainty of the evidence behind each estimate of treatment effect, using the GRADE approach. We used methods and recommendations described in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2020b; Schünemann 2020a). We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of the evidence as it relates to the studies that contributed data to the meta‐analyses for the prespecified outcomes, and reported the certainty of the evidence as high, moderate, low or very low.

We presented results comparing cryotherapy to any control in the summary of findings table for the following major outcomes: blood loss, pain, transfusion rate, range of motion, function, total adverse events and withdrawals due to adverse events. We used GRADEpro GDT software to prepare and display the summary of findings tables (GRADEpro GDT). We justified all decisions to downgrade the certainty of the evidence for each outcome in footnotes, and provided comments to aid the reader's understanding of the review where necessary.

Results

Description of studies

Results of the search

The Adie 2012 version of this review identified 12 studies eligible for inclusion, from searches performed on 15 March 2012 (Albrecht 1997; Gibbons 2001; Healy 1994; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Scarcella 1995; Smith 2002; Walker 1991; Webb 1998). Our updated searches on 27 May 2022 identified 4484 potentially eligible records (4206 records retrieved through database searching and 278 records from trial registers (Figure 1)). After de‐duplication, we screened 3083 records as titles/abstracts, and discarded 3002 records as irrelevant. We retrieved the remaining 81 articles for full‐text screening. Of these, we included 10 new studies (Bao 2017; Brouwers 2022; Deng 2015; Kang 2014; Stocker 2016; Thijs 2019; Wang 2017; Wittig‐Wells 2015; Xu 2015; Yuksel 2022), excluded 67 articles, listed one study as 'awaiting classification' (Siva Subramanian 1996), and identified three ongoing studies (ISRCTN12615549; NCT05395273; NCT05572359). We identified no additional eligible studies from searching reference lists, related links, conference proceedings and published protocols. Thus, this update review now includes a total of 22 included studies.

1.

PRISMA study flow diagram

Included studies

Study design and setting

Of the 22 included studies, 20 were randomised controlled trials (Albrecht 1997; Bao 2017; Deng 2015; Gibbons 2001; Healy 1994; Ivey 1994; Kang 2014; Kullenberg 2006; Levy 1993; Radkowski 2007; Scarcella 1995; Smith 2002; Stocker 2016; Thijs 2019; Walker 1991; Wang 2017; Webb 1998; Wittig‐Wells 2015; Xu 2015; Yuksel 2022), and two were controlled clinical trials (Brouwers 2022; Morsi 2002). All studies were parallel‐group trials except for Morsi 2002, which was a split‐body trial, and Wittig‐Wells 2015, which was a randomised trial with cross‐over. Two studies compared two intervention groups to the same control group (Albrecht 1997; Ivey 1994), and one study was performed across two phases, each with the same intervention but a different group and number of participants (Healy 1994).

Trials were conducted in the USA (Healy 1994; Ivey 1994; Levy 1993; Radkowski 2007; Scarcella 1995; Walker 1991; Wittig‐Wells 2015), China (Bao 2017; Deng 2015; Wang 2017; Xu 2015), the United Kingdom (Gibbons 2001; Webb 1998), the Netherlands (Brouwers 2022; Thijs 2019), Germany (Albrecht 1997), South Korea (Kang 2014), Sweden (Kullenberg 2006), Egypt (Morsi 2002), Australia (Smith 2002), Switzerland (Stocker 2016), and Turkey (Yuksel 2022).

Four studies received funding from manufacturers of cryotherapy machines (Healy 1994; Levy 1993; Radkowski 2007; Thijs 2019), three studies were funded by research foundations or universities (Brouwers 2022; Walker 1991; Webb 1998), and seven studies reported no funding (Kullenberg 2006; Morsi 2002; Stocker 2016; Wang 2017; Wittig‐Wells 2015; Xu 2015; Yuksel 2022). Eight studies did not report funding details (Albrecht 1997; Bao 2017; Deng 2015; Gibbons 2001; Ivey 1994; Kang 2014; Scarcella 1995; Smith 2002).

Participant characteristics

The 22 trials included 1839 participants, and the number of participants per trial ranged from 16 to 281. Of the nineteen studies that reported mean age, the mean age in study groups reflected the TKR population and varied from 64 to 74 years of age. Only two studies reported age ranges (Ivey 1994; Scarcella 1995), with the youngest participant being 36 years old and the oldest 88 years old. Fifteen studies reported gender numbers, showing that 54% of participants were female. The studies did not describe disease duration, severity of disease and diagnostic criteria.

Inclusion criteria were consistent between studies. All studies included participants with primary total knee replacement, though four studies included participants with bilateral total knee replacements (Healy 1994; Levy 1993; Morsi 2002; Webb 1998). Four studies explicitly stated that the ability to communicate with the investigators was part of the inclusion criteria (Deng 2015; Stocker 2016; Thijs 2019; Yuksel 2022). Sixteen studies described various medical conditions or operative complications (such as rheumatoid arthritis, vascular disease, opioid dependence or intraoperative fracture) which precluded participation (Albrecht 1997; Bao 2017; Brouwers 2022; Deng 2015; Kang 2014; Kullenberg 2006; Levy 1993; Radkowski 2007; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Webb 1998; Wittig‐Wells 2015; Xu 2015; Yuksel 2022).

Further information can be found in the Characteristics of included studies tables.

Interventions

The 22 included studies demonstrated clinical diversity in interventions and controls. In nine studies, cryotherapy was combined with compression and compared to compression alone (Deng 2015; Gibbons 2001; Healy 1994; Levy 1993; Morsi 2002; Radkowski 2007; Smith 2002; Thijs 2019; Xu 2015), although two of these studies combined compression with short periods of ice application in the control group (Healy 1994; Smith 2002). One study compared cryotherapy alone with compression (Stocker 2016). Eight studies compared cryotherapy without compression to a control with no treatment (Albrecht 1997; Bao 2017; Ivey 1994; Kang 2014; Scarcella 1995; Walker 1991; Wittig‐Wells 2015; Yuksel 2022). In five studies, cryotherapy was combined with compression and compared to no treatment (Albrecht 1997; Brouwers 2022; Kullenberg 2006; Wang 2017; Webb 1998). One study compared different frequencies of cryotherapy (ice exchanged every two to four hours) (Healy 1994), and one study compared different temperature settings on an electronic cryotherapy device (Ivey 1994). Eight studies specified the use of cemented TKRs (Albrecht 1997; Bao 2017; Healy 1994; Ivey 1994; Morsi 2002; Thijs 2019; Walker 1991; Webb 1998).

The doses of cooling applied varied amongst the included clinical trials. Twelve studies applied cryotherapy continuously with a consistent supply of ice, cooled water or both to the applied device (Albrecht 1997; Bao 2017; Gibbons 2001; Ivey 1994; Morsi 2002; Radkowski 2007; Scarcella 1995; Smith 2002; Walker 1991; Wang 2017; Webb 1998; Wittig‐Wells 2015). Five studies had one intervention arm where ice, cooled water or both was exchanged regularly rather than applied continuously. Levy 1993 exchanged ice, cooled water or both every 90 minutes, Healy 1994 every two hours, and Albrecht 1997, Healy 1994 and Kullenberg 2006 every four hours.

Seven studies applied cryotherapy for short intervals each day. Kang 2014 applied cryotherapy for five minutes, Stocker 2016 for 10 minutes three times a day, Xu 2015 for three to five hours, Yuksel 2022 for 12 to 15 minutes every two hours and Deng 2015 continuously for 24 hours then intermittently for one hour five times a day. The cryotherapy timing used by Brouwers 2022 and Thijs 2019 changed at different points after surgery. Cryotherapy was applied for six hours immediately after surgery, then for four hours three times a day (post‐operative day one) and finally for two hours twice a day (post‐operative days two to seven) in the Brouwers 2022 study. Thijs 2019 used cryotherapy for six hours immediately after surgery and then for three sessions of two to four hours on post‐operative days two to seven.

Four studies measured skin temperatures to monitor the doses of cooling administered (Albrecht 1997; Bao 2017; Morsi 2002; Radkowski 2007), but only one study specified a target skin temperature (7 degrees Celsius then 12 degrees Celsius) (Morsi 2002).

Outcomes

Major outcomes

Twelve studies measured blood loss (Albrecht 1997; Deng 2015; Gibbons 2001; Healy 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Smith 2002; Walker 1991; Webb 1998; Xu 2015). The studies used a variety of methods, including output volume from the surgical drain, haemoglobin drop and the Gross formula (Lopez‐Picado 2017).

Fifteen studies measured pain using a visual analogue scale (Boeckstyns 1989) at different time points (Albrecht 1997; Bao 2017; Brouwers 2022; Gibbons 2001; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Wittig‐Wells 2015; Xu 2015; Yuksel 2022). These included one, two and three days after surgery, as well as at six and twelve weeks.

Two studies provided data on blood transfusion rate (Morsi 2002; Webb 1998).

Twelve studies measured knee range of motion (either visually or with a goniometer) at discharge, one week post‐operatively, two weeks post‐operatively and three months post‐operatively (Albrecht 1997; Healy 1994; Kang 2014; Kullenberg 2006; Levy 1993; Morsi 2002; Scarcella 1995; Smith 2002; Stocker 2016; Walker 1991; Webb 1998; Yuksel 2022).

Seven studies measured knee function at two, six and twelve weeks after surgery (Brouwers 2022; Kang 2014; Radkowski 2007; Thijs 2019; Wang 2017; Xu 2015; Yuksel 2022). Tools used included the Knee injury and Osteoarthritis Outcome Score (Roos 2003), the Hospital for Special Surgery knee score (Insall 1976), the Korean Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (Bae 2001). Thijs 2019 used a Dutch translation of the Western Ontario and McMaster Universities Osteoarthritis Index (Dutch WOMAC) with a 0 to 100 scale as per the recommendation from Roorda 2004.

Sixteen studies measured total adverse events, including discomfort, local skin reactions, superficial infections, cold‐induced injuries and thrombolytic events (Albrecht 1997; Brouwers 2022; Gibbons 2001; Healy 1994; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Scarcella 1995; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Walker 1991; Xu 2015). Nineteen studies measured withdrawals due to adverse events (Albrecht 1997; Brouwers 2022; Deng 2015; Gibbons 2001; Healy 1994; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Scarcella 1995; Smith 2002; Stocker 2016; Thijs 2019; Walker 1991; Wang 2017; Webb 1998; Xu 2015; Yuksel 2022).

Minor outcomes

Of the minor outcomes, eight studies measured analgesia use at different time points (Gibbons 2001; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Smith 2002; Walker 1991; Webb 1998), and ten studies measured knee swelling (Brouwers 2022; Deng 2015; Healy 1994; Levy 1993; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Webb 1998; Yuksel 2022). Six studies provided data on length of stay (Gibbons 2001; Kullenberg 2006; Scarcella 1995; Smith 2002; Stocker 2016; Walker 1991), two studies on quality of life (Thijs 2019; Yuksel 2022), and three studies on activity level (Brouwers 2022; Stocker 2016; Yuksel 2022).

Two studies included data from bilateral TKRs without adjusting for a lack of independence (Healy 1994; Webb 1998), although bilateral TKRs contributed a minority of their total participants (29/76 and 9/40 participants, respectively). Default analyses included data from these studies, but we performed a sensitivity analysis excluding these studies from primary outcomes (Analysis 2.1; Analysis 2.2; Analysis 2.3).

2.1. Analysis.

Comparison 2: Any cold therapy versus any control (Sensitivity analyses excluding studies with bilateral TKRs), Outcome 1: COMBINED Blood loss

2.2. Analysis.

Comparison 2: Any cold therapy versus any control (Sensitivity analyses excluding studies with bilateral TKRs), Outcome 2: Transfusion rate

2.3. Analysis.

Comparison 2: Any cold therapy versus any control (Sensitivity analyses excluding studies with bilateral TKRs), Outcome 3: Adverse events total

We imputed standard deviations in Analysis 1.8, Analysis 1.24 and Analysis 1.31 for Gibbons 2001; Analysis 1.3 and Analysis 1.14 for Albrecht 1997; Analysis 1.4, Analysis 1.6, Analysis 1.14, Analysis 1.28 and Analysis 1.29 for Healy 1994; Analysis 1.16 and Analysis 1.31 for Walker 1991; Analysis 1.14, Analysis 1.28 and Analysis 1.29 for Webb 1998; Analysis 1.9 for Morsi 2002, and Analysis 1.11 for Brouwers 2022.

1.8. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 8: Pain at post‐operative day 1 ‐ visual analogue scale (VAS)

1.24. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 24: Analgesic use mg/kg/48 hours of morphine equivalent

1.31. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 31: Length of hospital stay ‐ days

1.3. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 3: Blood loss ‐ wound drainage mL/48 hours

1.14. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 14: Range of motion ‐ degrees of knee flexion post‐operative days 1 to 6

1.4. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 4: Blood loss ‐ total wound drainage mL

1.6. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 6: COMBINED Blood loss

1.28. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 28: Change in swelling at mid‐patella in mm ‐ post‐operative days 2 to 6

1.29. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 29: Change in swelling at mid‐patella in mm ‐ 6 weeks

1.16. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 16: Range of motion ‐ degrees of flexion at discharge

1.9. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 9: Pain at post‐operative day 2 ‐ visual analogue scale (VAS)

1.11. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 11: Pain at 6 weeks ‐ visual analogue scale (VAS)

Excluded studies

We excluded 67 studies. The most common reasons for exclusion were that studies were not randomised controlled trials or controlled clinical trials, or that studies only compared cryotherapy against another form of cryotherapy. See Characteristics of excluded studies and Figure 1.

Studies waiting classification

One study is awaiting classification (Siva Subramanian 1996). The study was presented at a meeting in 1996, but we could not locate the abstract, nor could we contact the authors (see Characteristics of studies awaiting classification).

Ongoing studies

We identified three ongoing studies that did not have study results available at the time of submission of this review (ISRCTN12615549; NCT05395273; NCT05572359). The intervention group in all three studies involves the use of cryotherapy with compression in patients following total knee replacement. We provide a description of these trials in the Characteristics of ongoing studies table.

Risk of bias in included studies

We present the summary of risk of bias assessment in Figure 2. Of the 22 trials, we did not judge any to have a low risk of bias across all domains. Most studies were at risk of selection, performance, detection and selective reporting biases.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Seven studies used an adequate method to generate a random sequence and were therefore at low risk of selection bias (Bao 2017; Ivey 1994; Radkowski 2007; Smith 2002; Thijs 2019; Wittig‐Wells 2015; Yuksel 2022). Four studies were at high risk of selection bias due to quasi‐randomisation (Brouwers 2022; Wang 2017) or non‐randomisation of participants (Morsi 2002; Scarcella 1995). All other studies were at an unclear risk of selection bias for random sequence generation (Albrecht 1997; Deng 2015; Gibbons 2001; Healy 1994; Kang 2014; Kullenberg 2006; Levy 1993; Stocker 2016; Walker 1991; Webb 1998; Xu 2015).

Only one study adequately concealed allocation (with sealed opaque envelopes given by someone unaffiliated with the study) and was at low risk of selection bias due to allocation concealment (Stocker 2016). Two studies were at high risk of selection bias because they did not conceal patient allocation (Brouwers 2022; Morsi 2002). All other studies were at an unclear risk of selection bias for allocation concealment (Albrecht 1997; Bao 2017; Deng 2015; Gibbons 2001; Healy 1994; Ivey 1994; Kang 2014; Kullenberg 2006; Levy 1993; Radkowski 2007; Scarcella 1995; Smith 2002; Thijs 2019; Walker 1991; Wang 2017; Webb 1998; Wittig‐Wells 2015; Xu 2015; Yuksel 2022).

Blinding

Blinding was subdivided into blinding of participants and personnel (performance bias) and blinding of outcomes for self‐reported and objective outcomes (detection bias).

The nature of cryotherapy does not lend itself easily to blinding of participants. Two studies blinded participants with cryotherapy devices containing room‐temperature water (Ivey 1994; Radkowski 2007). In 17 studies, participants would have been aware of the intervention they were receiving and were thus at high risk of performance bias (Albrecht 1997; Bao 2017; Brouwers 2022; Deng 2015; Gibbons 2001; Healy 1994; Kang 2014; Kullenberg 2006; Levy 1993; Morsi 2002; Smith 2002; Stocker 2016; Wang 2017; Webb 1998; Wittig‐Wells 2015; Xu 2015; Yuksel 2022). There was an unclear risk of performance bias in the remaining studies (Scarcella 1995; Thijs 2019; Walker 1991).

Only one study adequately blinded for self‐reported outcomes (Radkowski 2007), whereas 15 studies had a high risk of detection bias from inadequately blinding participants (Albrecht 1997; Bao 2017; Brouwers 2022; Gibbons 2001; Kang 2014; Kullenberg 2006; Levy 1993; Morsi 2002; Smith 2002; Stocker 2016; Wang 2017; Webb 1998; Wittig‐Wells 2015; Xu 2015; Yuksel 2022). The risk of bias for self‐reported outcomes was unclear in six studies (Deng 2015; Healy 1994; Ivey 1994; Scarcella 1995; Thijs 2019; Walker 1991).

Four studies were adequately blinded for objective outcomes (Ivey 1994; Radkowski 2007; Thijs 2019; Xu 2015). Fourteen studies had a high risk of detection bias for objective outcomes due to insufficient blinding (Albrecht 1997; Brouwers 2022; Deng 2015; Gibbons 2001; Healy 1994; Kang 2014; Kullenberg 2006; Levy 1993; Morsi 2002; Smith 2002; Stocker 2016; Wang 2017; Webb 1998; Yuksel 2022). There was an unclear risk in four studies (Bao 2017; Scarcella 1995; Walker 1991; Wittig‐Wells 2015).

Incomplete outcome data

Fifteen studies addressed incomplete outcome data (Albrecht 1997; Brouwers 2022; Gibbons 2001; Healy 1994; Kullenberg 2006; Morsi 2002; Radkowski 2007; Smith 2002; Stocker 2016; Thijs 2019; Walker 1991; Wang 2017; Webb 1998; Xu 2015; Yuksel 2022), and thus had a low risk of attrition bias. Two studies did not state from which allocation group participant withdrawals occurred, and thus were at high risk of attrition bias (Levy 1993; Scarcella 1995). There was an unclear risk in five studies (Bao 2017; Deng 2015; Ivey 1994; Kang 2014; Wittig‐Wells 2015).

Selective reporting

Selective reporting was potentially problematic in seven studies (Gibbons 2001; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Thijs 2019; Yuksel 2022). Only one study had a low risk of selective reporting: it provided sufficient information and reported all outcomes prespecified in the trial protocol (Stocker 2016). There was an unclear risk of reporting bias in fourteen studies (Albrecht 1997; Bao 2017; Brouwers 2022; Deng 2015; Healy 1994; Ivey 1994; Kang 2014; Scarcella 1995; Smith 2002; Walker 1991; Wang 2017; Webb 1998; Wittig‐Wells 2015; Xu 2015).

Other potential sources of bias

Two studies included data from bilateral TKR patients without adjusting for lack of independence, and thus were at high risk of bias in this domain (Healy 1994; Webb 1998). Figure 3 depicts one forest plot for the most frequently measured primary outcome (standardised blood loss). There was no evidence of publication bias. No study reported assessing reliability when measuring outcomes.

3.

Funnel plot of comparison: 1 Any cold versus any control, outcome: 1.6 COMBINED Blood loss.

Effects of interventions

See: Table 1

Any cold therapy compared to any control

The results are summarised in Table 1.

Major outcomes

Blood loss

Twelve of the 22 included studies, with a total of 956 patients, measured blood loss (Albrecht 1997; Deng 2015; Gibbons 2001; Healy 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Smith 2002; Walker 1991; Webb 1998; Xu 2015). The clinically important difference for blood loss has previously been defined as 300 mL (Kalairajah 2005).

Studies used different methods and time points to measure blood loss: wound drainage at 24 hours (Radkowski 2007), 48 hours (Gibbons 2001; Smith 2002; Webb 1998), or no time specified (Healy 1994; Walker 1991; Xu 2015), as well as haemoglobin drop at 24 hours (Kullenberg 2006), 48 hours (Albrecht 1997), 72 hours (Levy 1993), or on discharge (Morsi 2002). One study used a blood loss equation combining haemoglobin drop and wound drainage at 48 hours (Deng 2015). We conducted a combined analysis of these twelve studies (Analysis 1.6; Figure 4). The mean blood loss in the control group of Smith 2002 was 824.7 mL. The SMD was 0.72 units lower in the cryotherapy group (95% CI 1.41 lower to 0.02 lower; I² = 96%), indicating less blood loss in those with cryotherapy. This was converted to a clinically meaningful measure using the SD (366) from Smith 2002. The mean difference was 264 mL less blood loss (95% CI 7 mL to 516 mL less blood loss). Cryotherapy may reduce blood loss after TKR. This may or may not be a clinical important difference.

4.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.6 COMBINED Blood loss.

Acute time points: one study measured wound drainage at 24 hours (Radkowski 2007). The MD was 71.00 mL less blood loss in the cryotherapy group (95% CI 234.51 less to 92.51 more; P = 0.39; Analysis 1.2). One study measured haemoglobin drop from pre‐operative levels to 24 hours post‐operatively (Kullenberg 2006). The MD was 11.20 mmol/L less (95% CI 21.81 less to 0.59 less; P = 0.04), indicating less blood loss in those with cryotherapy (Analysis 1.5).

1.2. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 2: Blood loss ‐ wound drainage mL/24 hours

1.5. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 5: Blood loss ‐ haemoglobin drop mmol/L/24 hours

One study with 240 participants measured total blood loss (Xu 2015). The MD was 118.76 mL less blood loss in those with cryotherapy (95% CI 126.17 less to 111.35 less; P < 0.05). This result could not be pooled because the time point at which the data were recorded was reported as 'perioperative'.

Sub‐acute time points: four studies, with a total of 273 participants, measured wound drainage at 48 hours (Albrecht 1997; Gibbons 2001; Smith 2002; Webb 1998). The mean blood loss in the control group of Smith 2002 was 824.7 mL. The MD in the cryotherapy group was 142.14 mL less (95% CI 300.85 less to 16.57 more; P = 0.08, I² = 51%; Analysis 1.3). Two studies, with a total of 66 participants, measured total wound drainage (Healy 1994; Walker 1991). The MD was 96.04 mL less (95% CI 256.86 less to 64.77 more; P = 0.24, I² = 0%). Two studies, with a total of 140 participants, measured total body blood loss (Levy 1993; Morsi 2002). The MD was 637.06 mL less (95% CI 753.61 less to 520.50 less; P < 0.001, I² = 0%).

Subgroup analysis:Analysis 1.7 compares standardised blood loss in studies which specified use of a cemented prosthesis and those which did not. The pooled effect in studies using a cemented prosthesis was 0.34 standardised units lower (95% CI 0.92 lower to 0.23 higher). The pooled effect in studies using an uncemented prosthesis was 0.82 standardised units lower (95% CI 1.53 lower to 0.10 lower). There was a significant difference in this subgroup effect, indicating less blood loss in those with cemented prostheses (P = 0.05, I² = 74%). Insufficient data were available for subgroup analysis of different doses of the intervention and the use of continuous passive motion (CPM).

1.7. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 7: COMBINED Blood loss ‐ cement and no cement

Pain

Fifteen studies measured pain at different time points (Albrecht 1997; Bao 2017; Brouwers 2022; Gibbons 2001; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Wittig‐Wells 2015; Xu 2015; Yuksel 2022). We pooled outcomes from studies using a 10‐point visual analogue scale (VAS), with lower scores indicating less pain and higher scores indicating greater pain. Negative scores indicate a beneficial effect in those with cryotherapy.

Post‐operative day two: six studies, with a total of 530 participants, measured pain at the primary time point of post‐operative day two (Analysis 1.9). Smith 2002 was used as a reference study, with the mean score in the control group being 4.8 points. The MD was 1.64 points lower in the cryotherapy group (95% CI 2.28 lower to 1.00 lower; P < 0.001, I² = 62%), indicating less pain in those with cryotherapy (Figure 5). Cryotherapy may improve pain two days after TKR. With a minimal clinically important difference of 1.5 points on the visual analogue scale for this outcome, this difference may or may not be of clinical significance.

5.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.9 Pain at POD 2 VAS.

Post‐operative day one: eight studies, with a total of 735 participants, measured pain at post‐operative day one (Analysis 1.8). The MD was 0.78 points lower (95% CI 1.78 lower to 0.22 higher; P = 0.13, I² = 87%).

One study with 29 participants measured their pain at 30 minutes and 60 minutes after receiving analgesia with or without cryotherapy (Wittig‐Wells 2015). The MD was 0.11 points higher and 0.54 lower for 30 minutes and 60 minutes, respectively. The study reported no statistical difference in pain score change at any time point, indicating no difference between analgesia alone and analgesia with cryotherapy. We did not pool these results because outcome data were missing.

Post‐operative day three: nine studies, with a total of 789 participants, measured pain at post‐operative day three (Analysis 1.10). The MD was 0.80 lower points (95% CI 1.45 lower to 0.15 lower; P = 0.02, I² = 77%), indicating less pain in those with cryotherapy.

1.10. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 10: Pain at post‐operative day 3 ‐ visual analogue scale (VAS)

Six weeks: five studies, with a total of 522 participants, measured pain at six weeks after surgery (Analysis 1.11). The MD was 0.55 points lower (95% CI 1.01 lower to 0.09 lower; P = 0.02, I² = 73%), indicating less pain in those with cryotherapy.

Three months: two studies, with a total of 341 participants, measured pain at three months after surgery (Analysis 1.12). The MD was 0.43 points lower (95% CI 0.94 lower to 0.08 higher; P = 0.10, I² = 31%).

1.12. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 12: Pain at 12 weeks ‐ visual analogue scale (VAS)

Transfusion rate

Evidence from two studies (Morsi 2002; Webb 1998), involving 91 participants, indicates that 370 of 1000 people in the control group (no cryotherapy) would require a transfusion, compared to 788 of 1000 people in the cryotherapy group (RR 2.13, 95% CI 0.04 to 109.63; P = 0.71, I² = 87%; Figure 6). We are uncertain whether cryotherapy improves transfusion rate after TKR.

6.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.13 Transfusion rate.

One study measured the perioperative transfusion rate (Xu 2015). The risk ratio for transfusion was 0.5 in those with cryotherapy (95% CI 0.34 to 0.73; P < 0.001). We did not pool this result because it was unclear whether the outcome was measured before or after the intervention.

Range of motion

Twelve studies provided data on knee range of motion at different time points (Albrecht 1997; Healy 1994; Kang 2014; Kullenberg 2006; Levy 1993; Morsi 2002; Scarcella 1995; Smith 2002; Stocker 2016; Walker 1991; Webb 1998; Yuksel 2022). Outcomes were pooled, with a positive MD favouring those with cryotherapy. The mean knee range of motion at discharge in the control group of Kullenberg 2006 was 62.9 degrees.

At discharge: three studies, with a total of 174 participants, provided data on knee flexion at the primary time point time of discharge (Analysis 1.16). The MD in the cryotherapy group was 8.34 degrees more flexion (95% CI 3.57 more to 13.12 more; P < 0.001, I² = 3%), indicating increased range of motion in those with cryotherapy (Figure 7). Cryotherapy may improve knee range of motion at discharge after TKR. This may or may not be clinically significant.

7.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.16 Range of motion ‐ degrees of flexion at discharge.

Post‐operative week one: eight studies, with a total of 478 participants, provided data on knee flexion during post‐operative week one (Analysis 1.14). The MD was 5.57 degrees more (95% CI 0.16 more to 10.97 more; P = 0.04, I² = 73%).

Post‐operative week two: three studies, with a total of 170 participants, provided data on knee flexion during post‐operative week two (Analysis 1.15). The MD was 8.56 degrees more (95% CI 0.77 less to 17.89 more; P = 0.07, I² = 90%).

1.15. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 15: Range of motion ‐ degrees of knee flexion post‐operative days 7 to 14

Three months: three studies, with a total of 128 participants (Walker 1991; Webb 1998; Yuksel 2022), measured knee flexion at three months after surgery (Analysis 1.17). The MD was 2.00 degrees less (95% CI 6.61 less to 2.61 more; P = 0.40) based on precision data from Yuksel 2022. Neither Walker 1991 nor Webb 1998 provided sufficient data for meta‐analysis: the Walker 1991 study reported a favourable effect in those with cryotherapy, while the Webb 1998 study reported an unfavourable effect in those with cryotherapy.

1.17. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 17: Range of motion ‐ degrees of flexion at 3 months

One study reported an improvement in total arc range of motion at one and two weeks post‐operative in those with cryotherapy (Levy 1993). Only one study measured knee extension range of motion (Stocker 2016). They found no between‐group differences at six weeks post‐operative (MD = 0.4 degrees more, 95% CI 1.93 less to 2.73 more; P = 0.741).

Function

Seven studies measured knee function at different time points (Brouwers 2022; Kang 2014; Radkowski 2007; Thijs 2019; Wang 2017; Xu 2015; Yuksel 2022). Outcomes were pooled, with a positive score favouring those using cryotherapy. The mean function in the control group in Thijs 2019 was 75.4 on the Dutch WOMAC 0 to 100 scale (lower scores indicate worse function).

Two weeks: four studies (Brouwers 2022; Kang 2014; Thijs 2019; Wang 2017), with a total of 297 participants, provided data on knee function for the primary time point of two weeks after surgery (Analysis 1.18). The SMD in the cryotherapy group was 0.72 (95% CI 0.03 worse to 1.48 better; P = 0.06, I² = 89%). Using the SD (18.3) from Thijs 2019, this was equivalent to a mean difference of 13.18 (95% CI 0.55 worse to 27.08 improved) on the Dutch WOMAC. We are uncertain whether cryotherapy improves function after TKR (Figure 8).

1.18. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 18: Function ‐ post‐operative days 8 to 14

8.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.18 Function at POD 8‐14.

Six weeks: two studies (Brouwers 2022; Thijs 2019), with a total of 160 participants, measured knee function at six weeks after surgery (Analysis 1.19). The SMD was 0.21 better (95% CI 0.10 worse to 0.53 better; P = 0.18, I² = 0%), equivalent to a mean difference of 3.84 (95% CI 1.83 worse to 9.70 better).

1.19. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 19: Function ‐ 6 weeks

Three months: two studies (Thijs 2019; Yuksel 2022), with a total of 127 participants, measured knee function at three months after surgery (Analysis 1.20). The SMD was 0.21 (95% CI ‐0.14 to 0.56; P = 0.23, I² = 0%), equivalent to an improvement of 3.84 (95% CI 2.56 worse to 10.25 better).

1.20. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 20: Function ‐ 3 months

One study with 64 participants planned to measure knee function (Radkowski 2007), but did not report the outcome.

Two studies measured knee function with a 0 to 100 function score (higher score indicates better function) (Wang 2017; Xu 2015): participants scoring higher than 85 were graded 'excellent', 70 to 84 'good', 60 to 69 'reasonable' and lower than 60 as 'poor'. Xu 2015 assessed the knee function of 240 participants at three months after surgery with an unspecified knee scoring system. The cryotherapy group had 90 excellent, 26 good, four reasonable and zero poor gradings, compared to 85 excellent, 30 good, five reasonable and zero poor gradings in the compression group. The study reported no difference in function three months after surgery in those with cryotherapy. Wang 2017 assessed the knee function of 106 participants at six months after surgery with the Hospital for Special Surgery Score. The cryotherapy group had 38 excellent, eight good, five reasonable and two poor gradings, compared to 30 excellent, seven good, 11 reasonable and five poor gradings in the control group. The study reported a significant difference in the proportion of excellent or good gradings, with 87% in the cryotherapy group compared to 69% in the control group (P = 0.032).

Total adverse events

Sixteen studies, with a total of 1199 participants (Albrecht 1997; Brouwers 2022; Gibbons 2001; Healy 1994; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Scarcella 1995; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Walker 1991; Xu 2015), measured the occurrence of adverse events (Analysis 1.21). However, we pooled data from only seven studies (Albrecht 1997; Brouwers 2022; Gibbons 2001; Kullenberg 2006; Levy 1993; Scarcella 1995; Stocker 2016; Thijs 2019); the other studies reported no events. The most commonly reported adverse event was withdrawal due to cold discomfort. Except for Kullenberg 2006 and Levy 1993, which each reported two events of deep venous thrombosis (one in the intervention group and one in the control group), and Brouwers 2022, which reported one event of infection in the intervention group, no serious adverse effects were reported. Meta‐analysis showed that 21/1000 participants in the control group had adverse events, and 27/1000 in the cryotherapy group, indicating that cryotherapy is safe (RR 1.30, 95% CI 0.53 to 3.20; P = 0.56, I² = 14%; Figure 9). We are uncertain whether cryotherapy improves adverse effects after TKR.

1.21. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 21: Total adverse events

9.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.21 Adverse events.

Withdrawals due to adverse events

Nineteen studies, with a total of 1347 participants (Albrecht 1997; Brouwers 2022; Deng 2015; Gibbons 2001; Healy 1994; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Radkowski 2007; Scarcella 1995; Smith 2002; Stocker 2016; Thijs 2019; Walker 1991; Wang 2017; Webb 1998; Xu 2015; Yuksel 2022), measured withdrawals due to adverse events (Analysis 1.22). We pooled data from only two studies (Brouwers 2022; Kullenberg 2006); no other studies reported withdrawals from adverse events. Brouwers 2022 reported two withdrawals in the intervention group: one due to confusion after surgery and one due to infection. There were three withdrawals in Kullenberg 2006, two in the intervention group (one from deep vein thrombosis and one from superficial soft tissue infection) and one in the control group (from deep vein thrombosis). Meta‐analysis showed that 2/1000 participants withdrew in the control group and 4/1000 in the cryotherapy group (RR 2.71, 95% CI 0.42 to 17.38; P = 0.29, I² = 0%). We are uncertain whether cryotherapy affects withdrawals due to adverse events after TKR (Figure 10).

1.22. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 22: Withdrawals due to adverse events

10.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.22 Withdrawals due to adverse events.

Minor outcomes

Analgesia use

Eight studies provided data on analgesia use at different time points (Gibbons 2001; Ivey 1994; Kullenberg 2006; Levy 1993; Morsi 2002; Smith 2002; Walker 1991; Webb 1998). Outcomes were pooled, with negative scores indicating less analgesia use in those with cryotherapy.

We conducted a combined analysis of the eight studies, with a total of 516 participants: the SMD was 0.12 units less (95% CI 0.39 less to 0.14 more; P = 0.37, I² = 55%; Analysis 1.27, Figure 11).

1.27. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 27: COMBINED Analgesic use

11.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.27 COMBINED Analgesic use.

Acute time points: one study with 83 participants measured analgesia use at 24 hours (Kullenberg 2006). The MD was 0.06 mg/kg less units of morphine equivalent in the cryotherapy group (95% CI 0.30 less to 0.18 more; P = 0.62; Analysis 1.23).

1.23. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 23: Analgesic use mg/kg/24 hours

Sub‐acute time points: four studies, with a total of 255 participants, measured analgesia use at 48 hours (Analysis 1.24). The MD was 0.04 mg/kg less units of morphine equivalent (95% CI 0.18 less to 0.10 more; P = 0.59, I² = 75%). One study with 88 participants provided data on analgesia use at 72 hours (Ivey 1994). The MD was 0.19 mg/kg more units of morphine equivalent (95% CI 0.10 less to 0.48 more; P = 0.19, I² = 0%; Analysis 1.25).

1.25. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 25: Analgesic use mg/kg/72 hours

Length of stay: two studies, with a total of 90 participants, provided data on analgesia use over the participants' length of stay (Analysis 1.26). The MD was 213.80 mg less units of morphine equivalent (95% CI 604.00 less to 176.39 more; P = 0.28, I² = 82%).

1.26. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 26: Analgesia use total mg/kg

Knee swelling

Ten studies provided data on knee swelling at different time points (Brouwers 2022; Deng 2015; Healy 1994; Levy 1993; Smith 2002; Stocker 2016; Thijs 2019; Wang 2017; Webb 1998; Yuksel 2022). Outcomes were pooled, with negative scores indicating less swelling in those with cryotherapy.

Post‐operative days two to six: seven studies, with a total of 403 participants, measured change in knee swelling at the mid‐patella between post‐operative days two and six (Analysis 1.28). The MD was 7.32 mm less in the cryotherapy group (95% CI 11.79 less to 2.84 less; P = 0.001, I² = 73%), indicating less swelling in those with cryotherapy (Figure 12).

12.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.28 change in swelling at mid‐patella.

Six weeks: five studies, with a total of 283 participants, measured change in knee swelling at the mid‐patella at post‐operative week six (Analysis 1.29). The MD was 1.89 mm less (95% CI 13.66 less to 9.88 more; P = 0.75, I² = 64%).

Three months: two studies, with a total of 98 participants, measured change in knee swelling at three months after surgery (Webb 1998; Yuksel 2022; Analysis 1.30). Yuksel 2022 assessed swelling through a frustum formula, calculated using the circumference of the operated limb at different points. The SMD was 0.15 units less (95% CI 0.63 less to 0.33 more; P = 0.55). Webb 1998 measured swelling two centimetres proximal to the mid‐patella. They did not provide sufficient data for meta‐analysis, though reported no significant difference between groups.

1.30. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 30: Change in swelling at 12 weeks

Length of stay

Six studies, with a total of 297 participants (Gibbons 2001; Kullenberg 2006; Scarcella 1995; Smith 2002; Stocker 2016; Walker 1991), measured participants' length of stay (Analysis 1.31). There were no significant differences between the two groups (MD = 0.54 days less in the cryotherapy group, 95% CI 1.80 less to 0.72 more; P = 0.40, I² = 85%; Figure 13).

13.

Forest plot of comparison: 1 Any cold versus any control, outcome: 1.31 Length of hospital stay ‐ days.

Quality of life

Two studies, with a total of 127 participants (Thijs 2019; Yuksel 2022), measured patient quality of life after three months (Analysis 1.32). Outcomes were pooled, with a positive score favouring those using cryotherapy. The SMD was 0.14 units higher in the cryotherapy group (95% CI 0.21 lower to 0.49 higher; P = 0.42, I² = 0%). Using the SD (15.28) from Yuksel 2022, this was equivalent to 2.14 points higher (95% CI 3.41 lower to 7.49 higher) on the SF‐12 0 to 100 scale (higher score indicates increased quality of life).

1.32. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 32: Quality of life ‐ 3 months

Thijs 2019 also showed that differences in EQ‐5D Index score were insignificant after two weeks (MD = 0.003 more, 95% CI 0.04 less to 0.05 more; P = 0.89) and after six weeks (MD = 0.012 less, 95% CI 0.05 less to 0.03 more; P = 0.55).

Activity level

Three studies, with a total of 183 participants, measured activity level at different time points (Brouwers 2022; Stocker 2016; Yuksel 2022). A negative score favours those using cryotherapy.

One underpowered study with 16 participants measured activity level with the ‘Fast‐paced walking test’ by recording the time taken for the participant to walk 20 metres as fast as possible, turn around (time was not measured when turning around) and walk back as fast as possible (Stocker 2016). The MD was 12.50 seconds less in the cryotherapy group (95% CI 53.39 less to 28.39 more; P = 0.55) during week one and 2.30 seconds less (95% CI 16.00 less to 11.40 more; P = 0.74) after six weeks.

Two studies measured activity level using the Timed Up and Go (TUG) test at different time points (Brouwers 2022; Yuksel 2022). This is the time taken for a person to stand from a chair, walk 10 metres, turn around, walk back and then sit back down in the chair. Yuksel 2022 found no significant differences between the cryotherapy and control groups at discharge (MD = 4.84 seconds less, 95% CI 15.49 less to 5.81 more; P = 0.37) or after three months (MD = 0.02 seconds more, 95% CI 1.56 less to 1.60 more; P = 0.98). Brouwers 2022 assessed TUG scores in 100 participants. They did not provide sufficient data for meta‐analysis, though reported no significant differences between groups after six weeks.

Participant‐reported global assessment of success

No studies measured participant‐reported global assessment of success.

Subgroup analysis

We investigated different comparisons of cryotherapy and controls (with and without compression) by subgrouping studies in each analysis. We found no consistent benefit for any variation of the intervention and control. However, we noted a trend in favour of interventions combining cold and compression for blood loss, pain, analgesia use and swelling.

For combined blood loss (Analysis 1.6; Figure 4), the MD was 0.84 mL less blood loss (95% CI 1.61 less to 0.06 more; P = 0.03) in the combined cold and compression group (contained 10 studies) and 0.08 mL more (95% CI 0.28 less to 0.45 more; P = 0.65) in the other groups (contained two studies).

The MD for pain at post‐operative day two was 1.88 points less (95% CI 2.52 less to 1.24 less; P < 0.001) in the combined group (contained five studies) and 0.50 points less (95% CI 1.33 less to 0.33 more; P = 0.24) in the other groups (contained one study) (Analysis 1.9; Figure 5).

For combined analgesic use (Analysis 1.27; Figure 11), the SMD was 0.30 units lower (95% CI 0.59 lower to 0.00; P = 0.05) within the cold and compression group (contained three studies) and 0.14 units higher (95% CI 0.22 lower to 0.51 higher; P = 0.44) within the other groups (contained five studies).

The MD for change in knee swelling at the mid‐patella between post‐operative days two and six was 7.63 mm less (95% CI 12.52 to 2.75 mm less; P = 0.002) in the combined group (contained five studies) and 5.36 mm less (95% CI 18.02 less to 7.31 more; P = 0.41) in the other group (contained two studies) (Analysis 1.28; Figure 12).

Sensitivity analyses

Results of analyses utilising fixed‐effect meta‐analysis did not substantially change any results, and are not reported in this review. Excluding studies with imputed data also did not substantially change any effect estimates. We were unable to perform a sensitivity analysis based on the presence of adequate allocation concealment because only one study had adequate concealment.

Discussion

Summary of main results

This systematic review and meta‐analysis compares different forms of cryotherapy (ice packs, cooling pads, other commercial devices) to different controls (compression bandaging, nothing) applied to the knee after total knee replacement in adults. It provides low‐certainty evidence that cryotherapy may reduce blood loss (0.72 standardised units less, equivalent to approximately 264 mL less blood loss using reported standard deviation data). It also provides low‐certainty evidence that knee range of motion during week one after surgery and on discharge may be improved, with an additional 6 and 8 degrees of flexion, respectively (Table 1). No differences in knee flexion were found two weeks and three months after surgery. There was very low‐certainty evidence for total adverse events (discomfort, local skin reactions, superficial infections, cold‐induced injuries and thrombolytic events) and withdrawals due to adverse events following cryotherapy. The risk ratios had wide confidence intervals that included both no harm and harm. Despite this, both the theoretical and reported adverse events of cryotherapy were not serious (Nadler 2004), and it is likely cryotherapy is safe.

There was low‐certainty evidence that cryotherapy may improve pain at 48 hours, though it is uncertain if this benefit is of clinical importance (minimal clinically important difference is 1.5 points on the visual analogue scale (VAS)). This was further seen at 72 hours and six weeks after surgery, though not at 24 hours or three months after surgery. Because of very low‐certainty evidence, we are uncertain whether cryotherapy improves transfusion rate or knee function.

Whilst the change in swelling at the mid‐patella improved between post‐operative days two to six, these improvements were not maintained after six or twelve weeks. Cryotherapy may have little to no effect at any time point for analgesia use, length of stay, quality of life or activity level either in the short‐term or in the longer‐term results. We planned to measure participant‐reported global assessment of success, but this outcome was not measured by any study.

The benefits of cryotherapy for blood loss and knee range of motion may or may not be clinically meaningful. A clinically important difference for blood loss has previously been defined as 300 mL (Kalairajah 2005). However, a more clinically meaningful outcome would have been blood transfusion (preferably using a transfusion protocol). Only two studies measured transfusion rates (neither of which used a transfusion protocol), showing no benefit for those with cryotherapy. The evidence was also of very low certainty, meaning we are uncertain whether cryotherapy improves blood transfusions. While knee range of motion is an important outcome and may correlate with knee function and activity (Rowe 2000), the patient‐centred outcomes of knee function and activity level, in which no differences were found between groups, are arguably more important. In addition, studies which reported knee range of motion at three months after surgery did not find any benefit in those with cryotherapy, suggesting that downstream factors, such as physiotherapy and rehabilitation, may override the early benefits from cryotherapy. It is also possible that the outcomes measured lacked responsiveness to the intervention. In particular, transfusion rate, analgesia use and length of stay may have been driven by standard protocols which did not best account for individual variance or may have been guided by unrelated patient or clinician factors, such as level of support or insurance coverage.

Overall completeness and applicability of evidence

Overall, evidence identified in this review was of low‐ to very low‐certainty and there was a lack of standardised outcome measurement across studies. For instance, no studies measured the effect of cryotherapy on participant‐reported global assessment of success. Most studies measured blood loss, but used different methods and time points, including wound drainage at 24 hours (Radkowski 2007), 48 hours (Gibbons 2001; Smith 2002; Webb 1998), or no time specified (Healy 1994; Walker 1991; Xu 2015); or haemoglobin drop at 24 hours (Kullenberg 2006), 48 hours (Albrecht 1997), 72 hours (Levy 1993) or on discharge (Morsi 2002). One study used a blood loss equation (Deng 2015). A more clinically meaningful outcome regarding blood management is blood transfusion, but only two studies measured this outcome (Morsi 2002; Webb 1998), and neither study specified a transfusion threshold. Pain and knee range of motion were measured more uniformly across studies, using a VAS scale and range of flexion, respectively. Seven studies measured knee function using different scales and time points. In Radkowski 2007, the outcome of knee function was not reported. Two studies categorised continuous function scales into arbitrary groups (Wang 2017; Xu 2015). A standardised function score was pooled from five studies (Brouwers 2022; Kang 2014; Thijs 2019; Wang 2017; Yuksel 2022), although different validated outcomes were used – the Knee injury and Osteoarthritis Outcome Score (KOOS)‐pain score in Brouwers 2022, the Korean WOMAC in Kang 2014, the Dutch WOMAC in Thijs 2019 and Hospital for Special Surgery Score in Wang 2017 and Yuksel 2022.

Primary authors investigated intervention subgroups with varying temperatures, frequencies, pressures and durations of cryotherapy. Two studies had controls which also used small doses of ice application (Healy 1994; Smith 2002), which could have confounded results. We compared and measured these intervention subgroups against varying control subgroups. There was no evidence of any consistent benefit to any cryotherapy subgroup (with or without compression, and varying temperatures or ice exchange), although we noted a trend in favour of interventions combining cold and compression for blood loss, pain and analgesia use. We also noted that most studies applied cryotherapy immediately post‐operatively, continuously and for varying periods. To derive any benefit, it is likely similar continuous cryotherapy must be applied in routine clinical practice, which may be a challenge for existing rehabilitation protocols.

Certainty of the evidence

We extracted results from 20 randomised controlled trials and two controlled clinical trials. We did not judge any study to have low risk of bias across all domains.

Due to the nature of cryotherapy, it was difficult to blind participants and most studies had a high risk of performance and detection bias. Only two studies blinded participants and personnel. Just one study adequately concealed allocation and only seven studies used an adequate method to generate a random sequence. Seven studies were selective in their reporting of results and data.

Consequently, we graded evidence certainty as low for blood loss, pain and knee range of motion and very low for transfusion rate, function, total adverse events and withdrawals due to adverse events (Table 1).

We downgraded the certainty of the evidence for total blood loss by two levels due to the high risk of bias and inconsistency. Of the 12 studies assessed, most had limitations across various items. There was high statistical heterogeneity across studies (I² = 95%) which could not be explained by exploring different subgroups or other factors. We also graded the evidence for pain at post‐operative day two as low due to the high risk of bias across multiple items. We downgraded the evidence by two levels since it is a self‐reported outcome which would be strongly influenced by the lack of blinding in studies. The three studies used to assess range of motion at discharge suffered from selection, performance and detection bias; we therefore downgraded the evidence due to high risk of bias. We also downgraded for indirectness since a validated measuring instrument was not used, and it was unclear who performed the measurements in some studies.

We graded the certainty of the evidence for transfusion rate as very low. There was selection, performance, detection and reporting bias for the two studies, meaning there was a high risk of bias. Inconsistency was due to the high statistical heterogeneity (I² = 87%) across the two studies, and imprecision was due to the wide 95% confidence interval for the intervention. The evidence for function was also of very low certainty due to the high risk of bias and inconsistency. Because this is a subjective, self‐reported outcome, the lack of blinding of both participants and personnel meant that we downgraded the evidence by two levels. Inconsistency was because of the high unexplained heterogeneity (I² = 89%). We graded the evidence for adverse events and withdrawals due to adverse events as very low certainty because of the high risk of bias across studies, indirectness and imprecision. Indirectness was due to the differing definitions of adverse effects across studies, which meant that many studies did not account for the full array of potential side effects associated with both cryotherapy and total knee replacement. The outcome also had a wide 95% confidence interval which included both an increase and decrease in the risk ratio, and we therefore downgraded the evidence for imprecision.

Potential biases in the review process

The strengths of this review include the detailed electronic search strategy and data extraction process. We included non‐English language articles and translated seven articles: two German language articles and five Chinese language articles. We attempted to examine different forms of the intervention and control using a detailed subgroup analysis, and planned to measure a broad range of outcomes. A weakness of this review is insufficient data, particularly measures of precision, from many studies. We attempted to address this by contacting authors for additional data, and by imputing precision data in select circumstances.

Agreements and disagreements with other studies or reviews

A systematic review on cryotherapy following TKR included only randomised controlled trials, had different definitions of primary outcomes (knee extension not considered), and did not plan to measure as many outcomes as this review (Adie 2010). The Adie 2010 review also had a slightly different approach to meta‐analysis, using the fixed‐effect method when heterogeneity was not substantial. Nevertheless, similar results to ours were found for blood loss (MD = 117.40 mL less, 95% CI 215.69 mL less to 19.11 mL less; P = 0.02, I² = 58%) and knee range of motion at discharge (MD = 6.18 degrees more flexion, 95% CI 1.55 more to 10.80 more; P = 0.009, I² = 47%).

A systematic review on post‐operative analgesia after TKR found that cooling and compression techniques were effective in reducing pain and analgesic use and recommended its use in clinical practice (Fischer 2008). Whilst our review also found evidence to suggest that cryotherapy reduces pain, it did not find similar evidence to indicate any effect on analgesia use.

A more recent systematic review and meta‐analysis examined the effect of cryotherapy after TKR and total hip replacement (Ni 2015). The review did not measure as many outcomes as this review (it measured blood loss, adverse effects and pain) and also used a fixed‐effect method when heterogeneity was substantial. Subgroup analysis of the TKR group found similar results to ours for blood loss and adverse events. Similar results were found for pain after post‐operative day one and day two, but not after day three.

A systematic review on the use of cryotherapy after acute soft tissue injury found marginal evidence that its use with exercise reduced pain and swelling after ankle sprain and after surgery (anterior cruciate ligament reconstruction, total knee replacement, total hip replacement, knee arthroscopy, lateral retinacular release, carpal tunnel release) (Bleakley 2004). Our review did not find any significant benefits of cryotherapy on knee swelling.

A meta‐analysis on the use of cryotherapy after anterior cruciate ligament reconstruction in the knee concluded that cryotherapy significantly reduced post‐operative pain. However, there was no significant improvement in post‐operative knee range of motion (Raynor 2005). This differed from our review in that knee range of motion improved with cryotherapy. This may be due to the variation between operations.

All five reviews noted that the limited amount of evidence was of low quality and agreed that more high‐quality studies were needed to evaluate the use of cryotherapy in routine clinical use.

Authors' conclusions

Implications for practice.

There is low‐certainty evidence that cryotherapy may improve blood loss after total knee replacement (TKR), as well as pain, knee range of motion and knee swelling in the short term (within the first week) but not the long term (after twelve weeks). The evidence is uncertain regarding cryotherapy's impact on transfusion rate, knee function, analgesia use, length of stay, quality of life or activity level. Whilst the certainty of the evidence was limited, there was little concern for any serious adverse effects with the use of cryotherapy. With low‐ and very low‐certainty evidence, it is difficult to draw conclusions regarding the effect of cryotherapy on the outcomes measured in this review, and its use in clinical practice. This needs to be taken into account when considering the use of cryotherapy against its potential inconvenience and expense, particularly when using more expensive automated forms of cryotherapy.

Implications for research.

This review included only randomised controlled trials and controlled clinical trials. While these are considered to produce the highest level of evidence for clinical studies, the quality of many included studies was low due to high risk of bias.

More high‐quality clinical trials are needed, focusing especially on clinically meaningful outcomes such as blood transfusion, and patient‐reported outcomes such as knee function, quality of life, activity level and participant‐reported global assessment of success. Due to the nature of cryotherapy, blinding is difficult, particularly for patients and caregivers. Blinding is still possible, however, if automated electronic cryotherapy devices are used and temperature settings are concealed. It is always possible to blind outcome assessors. Allocation concealment methods should be described in detail. It is important that studies adequately report their results, and refer to their protocol to ensure that all stated outcomes have been addressed.

Most studies included in this review clearly described the duration and frequency of cryotherapy and the presence of concurrent compression or continuous passive motion, and often described the temperature of cryotherapy. Future studies should clearly document the units, time and direction of measurements. Future studies should also explicitly define outcomes, especially blood loss and adverse effects. If including participants undergoing unilateral and bilateral knee replacement, future investigators should distinguish between these two groups so as to delineate their results. In addition, future studies should plan controls to account for co‐interventions, such as compression or continuous passive motion. As the aim of cryotherapy is to reduce deep tissue and intra‐articular temperatures, methods should be adopted to ensure delivery of an optimum dose of cooling, particularly in people with thicker surrounding soft tissues (Bleakley 2011), who may be more prevalent in this population.

This review was updated in 2023, but the inclusion of recent clinical trials has not substantially changed our conclusions. Future research should have a robust design to minimise risk of bias, be adequately powered to detect between‐group differences, and focus on patient‐centred outcomes (knee function, quality of life, activity level and participant‐reported global assessment of success). This would provide further information and improved quality of evidence for outcomes, therefore potentially changing our conclusions on the use of cryotherapy after total knee replacement. Thus, the presence of better blinded studies, focusing on clinically significant outcomes, would meaningfully contribute to the literature and support a future update of the review.

What's new

Date	Event	Description
30 October 2023	Amended	Typographical error corrected in the plain language summary.

History

Protocol first published: Issue 3, 2009
Review first published: Issue 9, 2012

Date	Event	Description
14 September 2023	New search has been performed	Search for trials was updated in May 2022. An additional ten trials (Bao 2017; Brouwers 2022; Deng 2015; Kang 2014; Stocker 2016; Thijs 2019; Wang 2017; Wittig‐Wells 2015; Xu 2015; Yuksel 2022) were found and incorporated into the review.
14 September 2023	New citation required but conclusions have not changed	All additional trials were small and of low quality, and no substantive changes were made to the conclusions of the review.
30 July 2008	Amended	CMSG ID A042

Appendices

Appendix 1. Electronic search strategy

Cochrane Database of Systematic Reviews, CENTRAL, DARE, HTA database and MEDLINE via Ovid

1	cryotherapy/
2	cryo$.mp.
3	exp cold/
4	cold$.mp.
5	ice/
6	(ice$ or icing).mp.
7	cool$.mp.
8	temperature/
9	temperature$.mp.
10	thermal$.mp.
11	or/1‐10
12	Arthroplasty, Replacement, Knee/
13	Knee Prosthesis/
14	tkr.tw.
15	exp Knee/
16	knee$.tw.
17	15 or 16
18	exp arthroplasty/
19	Joint Prosthesis/
20	(arthroplast$ or prosthe$ or replac$).tw.
21	or/18‐20
22	17 and 21
23	or/12‐14, 22
24	11 and 23

Embase via Ovid

1	cryotherapy/
2	cryo$.mp.
3	cold/
4	cold treatment/
5	cold$.mp.
6	ice/
7	(ice$ or icing).mp.
8	cooling/
9	cooling water/
10	cool$.mp.
11	temperature/
12	temperature$.mp.
13	thermal$.mp.
14	or/1‐13
15	knee arthroplasty/
16	knee prosthesis/
17	total knee replacement/
18	(knee adj3 (arthroplast$ or replac$ or prosthe$)).mp.
19	or/15‐18
20	14 and 19

CINAHL

1	exp CRYOTHERAPY/
2	cryo*.tw.
3	exp COLD/
4	cold*.tw.
5	(ice* or icing).tw.
6	exp HEAT/
7	cool*.tw.
8	exp TEMPERATURE/
9	temperature*.tw.
10	thermal*.tw.
11	or/1‐10
12	exp Arthroplasty, Replacement, Knee/
13	exp KNEE/ or exp KNEE JOINT/
14	exp Joint Prosthesis/
15	and/13‐14
16	(knee adj3 arthroplast*).tw.
17	(knee adj3 replac*).tw.
18	(knee adj3 prosthe*).tw.
19	or/16‐18
20	12 or 15 or 19
21	20 and 11

PEDro

Therapy: electrotherapies, heat and cold

AND

Body part: lower leg or knee

AND

Method: clinical trial

Web of Science

1	TS=(cryo* OR cold* OR ice* OR icing OR cool* OR temperature* OR thermal*)
2	TI=(cryo* OR cold* OR ice* OR icing OR cool* OR temperature* OR thermal*)
3	2 OR 1
4	TS=knee
5	TI=knee
6	5 OR 4
7	TS=(arthroplast* OR replac* OR prosthe*)
8	TI=(arthroplast* OR replac* OR prosthe*)
9	8 OR 7
10	9 AND 6 AND 3

ClinicalTrials.gov

cryo* OR cold OR ice OR icing OR cool OR temperature* OR thermal* | knee* OR arthroplasty OR tkr OR tka OR prosthe* OR replac*

World Health Organization: International Clinical Trials Registry Platform Search Portal

cryo* OR cold OR ice OR temperature* OR thermal*) AND (knee* OR arthroplasty OR tkr OR tka OR prosthe* OR replac*)

Data and analyses

Comparison 1. Any cold therapy versus any control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Blood loss ‐ total body mL	2	140	Mean Difference (IV, Random, 95% CI)	‐637.06 [‐753.61, ‐520.50]
1.1.1 Cold Compression vs. Compression	2	140	Mean Difference (IV, Random, 95% CI)	‐637.06 [‐753.61, ‐520.50]
1.2 Blood loss ‐ wound drainage mL/24 hours	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.2.1 Cold Compression vs. Compression	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.3 Blood loss ‐ wound drainage mL/48 hours	4	273	Mean Difference (IV, Random, 95% CI)	‐142.14 [‐300.85, 16.57]
1.3.1 Cold vs. Compression	1	84	Mean Difference (IV, Random, 95% CI)	75.30 [‐107.23, 257.83]
1.3.2 Cold Compression vs. Nothing	2	129	Mean Difference (IV, Random, 95% CI)	‐187.45 [‐321.12, ‐53.78]
1.3.3 Cold Compression vs. Compression	1	60	Mean Difference (IV, Random, 95% CI)	‐480.00 [‐949.99, ‐10.01]
1.4 Blood loss ‐ total wound drainage mL	2	66	Mean Difference (IV, Random, 95% CI)	‐96.04 [‐256.86, 64.77]
1.4.1 Cold Compression vs. Compression	1	36	Mean Difference (IV, Random, 95% CI)	‐188.00 [‐458.89, 82.89]
1.4.2 Cold and Continuous Passive Motion vs. Continuous Passive Motion	1	30	Mean Difference (IV, Random, 95% CI)	‐46.00 [‐245.84, 153.84]
1.5 Blood loss ‐ haemoglobin drop mmol/L/24 hours	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.5.1 Cold Compression vs. Nothing	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.6 COMBINED Blood loss	12	956	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐1.41, ‐0.02]
1.6.1 Cold vs. Compression	1	84	Std. Mean Difference (IV, Random, 95% CI)	0.17 [‐0.26, 0.60]
1.6.2 Cold Compression vs. Nothing	3	212	Std. Mean Difference (IV, Random, 95% CI)	‐0.29 [‐0.73, 0.16]
1.6.3 Cold Compression vs. Compression	7	630	Std. Mean Difference (IV, Random, 95% CI)	‐1.10 [‐2.18, ‐0.03]
1.6.4 Cold and Continuous Passive Motion vs. Continuous Passive Motion	1	30	Std. Mean Difference (IV, Random, 95% CI)	‐0.16 [‐0.88, 0.56]
1.7 COMBINED Blood loss ‐ cement and no cement	10	662	Std. Mean Difference (IV, Random, 95% CI)	‐0.55 [‐0.99, ‐0.11]
1.7.1 Cement	5	295	Std. Mean Difference (IV, Random, 95% CI)	‐0.34 [‐0.92, 0.23]
1.7.2 No Cement	5	367	Std. Mean Difference (IV, Random, 95% CI)	‐0.82 [‐1.53, ‐0.10]
1.8 Pain at post‐operative day 1 ‐ visual analogue scale (VAS)	9	735	Mean Difference (IV, Random, 95% CI)	‐0.78 [‐1.78, 0.22]
1.8.1 Cold versus compression	1	84	Mean Difference (IV, Random, 95% CI)	0.10 [‐0.72, 0.92]
1.8.2 Cold compression versus nothing	4	389	Mean Difference (IV, Random, 95% CI)	‐1.66 [‐2.59, ‐0.73]
1.8.3 Cold compression versus compression	3	204	Mean Difference (IV, Random, 95% CI)	0.12 [‐0.66, 0.89]
1.8.4 Cold plus analgesia versus analgesia	1	58	Mean Difference (IV, Random, 95% CI)	Not estimable
1.9 Pain at post‐operative day 2 ‐ visual analogue scale (VAS)	6	530	Mean Difference (IV, Random, 95% CI)	‐1.64 [‐2.28, ‐1.00]
1.9.1 Cold versus compression	1	84	Mean Difference (IV, Random, 95% CI)	‐0.50 [‐1.33, 0.33]
1.9.2 Cold compression versus nothing	3	306	Mean Difference (IV, Random, 95% CI)	‐1.58 [‐1.73, ‐1.42]
1.9.3 Cold compression versus compression	2	140	Mean Difference (IV, Random, 95% CI)	‐2.36 [‐4.41, ‐0.31]
1.10 Pain at post‐operative day 3 ‐ visual analogue scale (VAS)	9	789	Mean Difference (IV, Random, 95% CI)	‐0.80 [‐1.45, ‐0.15]
1.10.1 Cold versus compression	2	100	Mean Difference (IV, Random, 95% CI)	‐0.24 [‐2.38, 1.89]
1.10.2 Cold compression versus nothing	2	185	Mean Difference (IV, Random, 95% CI)	‐0.98 [‐2.06, 0.10]
1.10.3 Cold compression versus compression	5	504	Mean Difference (IV, Random, 95% CI)	‐1.13 [‐1.68, ‐0.57]
1.11 Pain at 6 weeks ‐ visual analogue scale (VAS)	5	522	Mean Difference (IV, Random, 95% CI)	‐0.55 [‐1.01, ‐0.09]
1.11.1 Cold versus nothing	1	67	Mean Difference (IV, Random, 95% CI)	‐0.07 [‐0.52, 0.38]
1.11.2 Cold versus compression	1	16	Mean Difference (IV, Random, 95% CI)	0.40 [‐1.17, 1.97]
1.11.3 Cold compression versus nothing	1	100	Mean Difference (IV, Random, 95% CI)	‐1.30 [‐1.80, ‐0.80]
1.11.4 Cold compression versus compression	1	58	Mean Difference (IV, Random, 95% CI)	‐0.40 [‐1.18, 0.38]
1.11.5 Cold and continuous passive motion (CMP) versus CPM	1	281	Mean Difference (IV, Random, 95% CI)	‐0.65 [‐0.89, ‐0.41]
1.12 Pain at 12 weeks ‐ visual analogue scale (VAS)	2	341	Mean Difference (IV, Random, 95% CI)	‐0.43 [‐0.94, 0.08]
1.12.1 Cold compression versus compression	1	60	Mean Difference (IV, Random, 95% CI)	0.10 [‐0.95, 1.15]
1.12.2 Cold and continuous passive motion (CMP) versus CPM	1	281	Mean Difference (IV, Random, 95% CI)	‐0.56 [‐0.82, ‐0.30]
1.13 Transfusion rate	2	91	Risk Ratio (IV, Random, 95% CI)	2.13 [0.04, 109.63]
1.13.1 Cold compression versus nothing	1	31	Risk Ratio (IV, Random, 95% CI)	20.19 [1.28, 319.17]
1.13.2 Cold compression versus compression	1	60	Risk Ratio (IV, Random, 95% CI)	0.35 [0.16, 0.77]
1.14 Range of motion ‐ degrees of knee flexion post‐operative days 1 to 6	8	478	Mean Difference (IV, Random, 95% CI)	5.57 [0.16, 10.97]
1.14.1 Cold versus compression	2	100	Mean Difference (IV, Random, 95% CI)	5.55 [‐11.26, 22.36]
1.14.2 Cold compression versus nothing	3	212	Mean Difference (IV, Random, 95% CI)	7.49 [‐2.51, 17.48]
1.14.3 Cold compression versus compression	2	136	Mean Difference (IV, Random, 95% CI)	4.49 [‐8.74, 17.71]
1.14.4 Cold versus ultrasound	1	30	Mean Difference (IV, Random, 95% CI)	3.20 [‐1.74, 8.14]
1.15 Range of motion ‐ degrees of knee flexion post‐operative days 7 to 14	3	170	Mean Difference (IV, Random, 95% CI)	8.56 [‐0.77, 17.89]
1.15.1 Cold compression versus compression	2	140	Mean Difference (IV, Random, 95% CI)	12.89 [8.48, 17.30]
1.15.2 Cold versus ultrasound	1	30	Mean Difference (IV, Random, 95% CI)	0.70 [‐2.31, 3.71]
1.16 Range of motion ‐ degrees of flexion at discharge	3	174	Mean Difference (IV, Random, 95% CI)	8.34 [3.57, 13.12]
1.16.1 Cold versus nothing	2	91	Mean Difference (IV, Random, 95% CI)	5.91 [0.05, 11.78]
1.16.2 Cold compression versus nothing	1	83	Mean Difference (IV, Random, 95% CI)	12.20 [4.75, 19.65]
1.17 Range of motion ‐ degrees of flexion at 3 months	3	128	Mean Difference (IV, Random, 95% CI)	‐2.00 [‐6.61, 2.61]
1.17.1 Cold compression versus nothing	1	31	Mean Difference (IV, Random, 95% CI)	Not estimable
1.17.2 Cold and continuous passive motion (CPM) versus CPM	1	30	Mean Difference (IV, Random, 95% CI)	Not estimable
1.17.3 Cold versus nothing	1	67	Mean Difference (IV, Random, 95% CI)	‐2.00 [‐6.61, 2.61]
1.18 Function ‐ post‐operative days 8 to 14	4	296	Std. Mean Difference (IV, Random, 95% CI)	0.72 [‐0.03, 1.48]
1.18.1 Cold compression versus nothing	2	206	Std. Mean Difference (IV, Random, 95% CI)	0.98 [0.28, 1.68]
1.18.2 Cold compression versus compression	1	60	Std. Mean Difference (IV, Random, 95% CI)	‐0.34 [‐0.85, 0.17]
1.18.3 Cold versus ultrasound	1	30	Std. Mean Difference (IV, Random, 95% CI)	1.33 [0.52, 2.13]
1.19 Function ‐ 6 weeks	2	160	Std. Mean Difference (IV, Random, 95% CI)	0.21 [‐0.10, 0.53]
1.19.1 Cold Compression vs. Nothing	1	100	Std. Mean Difference (IV, Random, 95% CI)	0.32 [‐0.07, 0.72]
1.19.2 Cold Compression vs. Compression	1	60	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.47, 0.54]
1.20 Function ‐ 3 months	2	127	Std. Mean Difference (IV, Random, 95% CI)	0.21 [‐0.14, 0.56]
1.20.1 Cold vs. Nothing	1	67	Std. Mean Difference (IV, Random, 95% CI)	0.08 [‐0.40, 0.56]
1.20.2 Cold Compression vs. Compression	1	60	Std. Mean Difference (IV, Random, 95% CI)	0.37 [‐0.14, 0.88]
1.21 Total adverse events	16	1199	Risk Ratio (IV, Random, 95% CI)	1.30 [0.53, 3.20]
1.21.1 Cold vs. Nothing	2	112	Risk Ratio (IV, Random, 95% CI)	0.11 [0.01, 1.86]
1.21.2 Cold vs. Compression	2	100	Risk Ratio (IV, Random, 95% CI)	2.00 [0.22, 17.89]
1.21.3 Cold Compression vs. Nothing	4	357	Risk Ratio (IV, Random, 95% CI)	4.39 [0.93, 20.72]
1.21.4 Cold Compression vs. Compression	7	600	Risk Ratio (IV, Random, 95% CI)	0.82 [0.26, 2.60]
1.21.5 Cold and Continuous Passive Motion vs. Continuous Passive Motion	1	30	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.22 Withdrawals due to adverse events	19	1347	Risk Ratio (IV, Random, 95% CI)	2.71 [0.42, 17.38]
1.22.1 Cold vs. Nothing	3	179	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.22.2 Cold vs. Compression	2	100	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.22.3 Cold Compression vs. Nothing	5	388	Risk Ratio (IV, Random, 95% CI)	2.71 [0.42, 17.38]
1.22.4 Cold Compression vs. Compression	8	650	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.22.5 Cold and Continuous Passive Motion vs. Continuous Passive Motion	1	30	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.23 Analgesic use mg/kg/24 hours	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.23.1 Cold compression versus nothing	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.24 Analgesic use mg/kg/48 hours of morphine equivalent	4	255	Mean Difference (IV, Random, 95% CI)	‐0.04 [‐0.18, 0.10]
1.24.1 Cold versus compression	1	84	Mean Difference (IV, Random, 95% CI)	0.10 [‐0.03, 0.23]
1.24.2 Cold compression versus nothing	1	31	Mean Difference (IV, Random, 95% CI)	‐0.14 [‐0.30, 0.02]
1.24.3 Cold compression versus compression	2	140	Mean Difference (IV, Random, 95% CI)	‐0.06 [‐0.27, 0.14]
1.25 Analgesic use mg/kg/72 hours	1	88	Mean Difference (IV, Random, 95% CI)	0.19 [‐0.10, 0.48]
1.25.1 Cold versus nothing	1	88	Mean Difference (IV, Random, 95% CI)	0.19 [‐0.10, 0.48]
1.26 Analgesia use total mg/kg	2	90	Mean Difference (IV, Random, 95% CI)	‐213.80 [‐604.00, 176.39]
1.26.1 Cold Compression vs. Compression	1	60	Mean Difference (IV, Random, 95% CI)	‐450.00 [‐781.16, ‐118.84]
1.26.2 Cold and Continuous Passive Motion vs. Continuous Passive Motion	1	30	Mean Difference (IV, Random, 95% CI)	‐46.00 [‐105.42, 13.42]
1.27 COMBINED Analgesic use	8	516	Std. Mean Difference (IV, Random, 95% CI)	‐0.12 [‐0.39, 0.14]
1.27.1 Cold vs. Nothing	1	88	Std. Mean Difference (IV, Random, 95% CI)	0.27 [‐0.17, 0.71]
1.27.2 Cold vs. Compression	1	84	Std. Mean Difference (IV, Random, 95% CI)	0.34 [‐0.10, 0.77]
1.27.3 Cold Compression vs. Nothing	2	114	Std. Mean Difference (IV, Random, 95% CI)	‐0.26 [‐0.71, 0.19]
1.27.4 Cold Compression vs. Compression	3	200	Std. Mean Difference (IV, Random, 95% CI)	‐0.31 [‐0.77, 0.15]
1.27.5 Cold and Continuous Passive Motion vs. Continuous Passive Motion	1	30	Std. Mean Difference (IV, Random, 95% CI)	‐0.54 [‐1.27, 0.19]
1.28 Change in swelling at mid‐patella in mm ‐ post‐operative days 2 to 6	7	403	Mean Difference (IV, Random, 95% CI)	‐7.32 [‐11.79, ‐2.84]
1.28.1 Cold versus compression	2	100	Mean Difference (IV, Random, 95% CI)	‐5.36 [‐18.02, 7.31]
1.28.2 Cold compression versus nothing	2	137	Mean Difference (IV, Random, 95% CI)	‐4.76 [‐5.84, ‐3.69]
1.28.3 Cold compression versus compression	3	166	Mean Difference (IV, Random, 95% CI)	‐10.83 [‐21.53, ‐0.12]
1.29 Change in swelling at mid‐patella in mm ‐ 6 weeks	5	283	Mean Difference (IV, Random, 95% CI)	‐1.89 [‐13.66, 9.88]
1.29.1 Cold versus compression	1	16	Mean Difference (IV, Random, 95% CI)	‐2.00 [‐28.34, 24.34]
1.29.2 Cold compression versus nothing	2	131	Mean Difference (IV, Random, 95% CI)	1.32 [‐12.11, 14.75]
1.29.3 Cold compression versus compression	2	136	Mean Difference (IV, Random, 95% CI)	‐3.80 [‐26.68, 19.09]
1.30 Change in swelling at 12 weeks	2	98	Std. Mean Difference (IV, Random, 95% CI)	‐0.15 [‐0.63, 0.33]
1.30.1 Cold versus nothing	1	67	Std. Mean Difference (IV, Random, 95% CI)	‐0.15 [‐0.63, 0.33]
1.30.2 Cold compression versus nothing	1	31	Std. Mean Difference (IV, Random, 95% CI)	Not estimable
1.31 Length of hospital stay ‐ days	6	297	Mean Difference (IV, Random, 95% CI)	‐0.54 [‐1.80, 0.72]
1.31.1 Cold versus nothing	1	24	Mean Difference (IV, Random, 95% CI)	‐1.50 [‐3.65, 0.65]
1.31.2 Cold versus compression	2	100	Mean Difference (IV, Random, 95% CI)	‐1.16 [‐3.17, 0.84]
1.31.3 Cold compression versus nothing	1	83	Mean Difference (IV, Random, 95% CI)	‐1.40 [‐2.16, ‐0.64]
1.31.4 Cold compression versus compression	1	60	Mean Difference (IV, Random, 95% CI)	2.00 [0.93, 3.07]
1.31.5 Cold and continuous passive motion (CPM) versus CPM	1	30	Mean Difference (IV, Random, 95% CI)	‐0.20 [‐1.72, 1.32]
1.32 Quality of life ‐ 3 months	2	127	Std. Mean Difference (IV, Random, 95% CI)	0.14 [‐0.21, 0.49]
1.32.1 Cold versus nothing	1	67	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.46, 0.50]
1.32.2 Cold compression versus compression	1	60	Std. Mean Difference (IV, Random, 95% CI)	0.28 [‐0.23, 0.79]
1.33 Activity level ‐ post‐operative days 1 to 7	1	16	Mean Difference (IV, Random, 95% CI)	‐12.50 [‐53.39, 28.39]
1.33.1 Cold versus compression	1	16	Mean Difference (IV, Random, 95% CI)	‐12.50 [‐53.39, 28.39]
1.34 Activity level ‐ 6 weeks	2	116	Mean Difference (IV, Random, 95% CI)	‐2.30 [‐16.00, 11.40]
1.34.1 Cold versus compression	1	16	Mean Difference (IV, Random, 95% CI)	‐2.30 [‐16.00, 11.40]
1.34.2 Cold compression versus nothing	1	100	Mean Difference (IV, Random, 95% CI)	Not estimable

1.1. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 1: Blood loss ‐ total body mL

1.13. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 13: Transfusion rate

1.33. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 33: Activity level ‐ post‐operative days 1 to 7

1.34. Analysis.

Comparison 1: Any cold therapy versus any control, Outcome 34: Activity level ‐ 6 weeks

Comparison 2. Any cold therapy versus any control (Sensitivity analyses excluding studies with bilateral TKRs).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 COMBINED Blood loss	10	849	Std. Mean Difference (IV, Random, 95% CI)	‐0.85 [‐1.67, ‐0.03]
2.1.1 Cold versus compression	1	84	Std. Mean Difference (IV, Random, 95% CI)	0.17 [‐0.26, 0.60]
2.1.2 Cold compression versus nothing	2	181	Std. Mean Difference (IV, Random, 95% CI)	‐0.18 [‐0.70, 0.35]
2.1.3 Cold compression versus compression	6	554	Std. Mean Difference (IV, Random, 95% CI)	‐1.48 [‐2.72, ‐0.23]
2.1.4 Cold and continuous passive motion (CPM) versus CPM	1	30	Std. Mean Difference (IV, Random, 95% CI)	‐0.16 [‐0.88, 0.56]
2.2 Transfusion rate	1	60	Risk Ratio (IV, Random, 95% CI)	0.35 [0.16, 0.77]
2.2.1 Cold compression versus compression	1	60	Risk Ratio (IV, Random, 95% CI)	0.35 [0.16, 0.77]
2.3 Adverse events total	12	833	Risk Ratio (IV, Random, 95% CI)	1.13 [0.44, 2.89]
2.3.1 Cold versus nothing	2	112	Risk Ratio (IV, Random, 95% CI)	0.11 [0.01, 1.86]
2.3.2 Cold versus compression	1	84	Risk Ratio (IV, Random, 95% CI)	Not estimable
2.3.3 Cold compression versus nothing	3	283	Risk Ratio (IV, Random, 95% CI)	3.63 [0.77, 17.11]
2.3.4 Cold compression versus compression	5	324	Risk Ratio (IV, Random, 95% CI)	0.82 [0.26, 2.60]
2.3.5 Cold and continuous passive motion (CPM) versus CPM	1	30	Risk Ratio (IV, Random, 95% CI)	Not estimable

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Albrecht 1997.

Study characteristics
Methods	Parallel‐group prospective randomised trial. No time period stated.
Participants	98 TKRs in 98 participants in Berlin, Germany. Cemented knee prosthesis used. Number and location of study centres not explicitly stated. Sample size Group 1: 31 TKR in 31 participants Group 2: 32 TKR in 32 participants Group 3: 35 TKR in 35 participants Mean age Group 1: 69.6 Group 2: 71.9 Group 3: 73.8 Age range Not discussed Sex (M:F) Not reported ASA score Group 1: 2.43 Group 2: 2.39 Group 3: 2.62 Inclusion criteria Elective total knee replacement Exclusion criteria Coronary heart disease; hypertonic labile heart problems; patients who were > 2 SD than normal on State Trait Anxiety Inventory; blood disturbances
Interventions	Group 1: Knee control No cold therapy. Standard post‐op regimen (also offered to intervention groups). No cooling or compression. CPM was used for all participants. Regional nerve block with bupivacaine was given to all participants. Group 2: Knee intermittent ice blocks Ice pack (35 cm x 28 cm) applied post‐op within 2 layers of a compressive bandage. Changed every 4 hours. Application temperature 4 °C. Group 3: Knee continuous cold therapy Continuous cooling (without compression) via pad placed on the post‐op knee and a closed pump system with temperature at 4 °C. Speed of 300 mL/min applied.
Outcomes	Outcomes included in this review 1. Blood loss (blood loss mL at 48 hrs; Hb drop mg/dL at 48 hrs) 2. Pain (visual analogue scale at POD 1 and 2) 3. Adverse effects 4. Knee range of motion (Active/passive knee extension at POD 1 and 2 in degrees) Other outcomes Skin temperature (°C) Cold sensation. Participants classed the feeling of cold as "very good", "good", "annoying" and "unpleasant".
Notes	7 participants in the continuous cooling group were removed from the study after 6 hours, 5 due to pain and 2 due to local skin reactions. We halved the number on the control group to prevent duplicate examination of control participants in analysis. We imputed standard deviations in Analysis 1.3 and Analysis 1.14 using the average standard deviations from other studies in the meta‐analysis. Conflicts of interest not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Not mentioned (in translation)
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants received intermittent, continuous or no cold therapy, in the form of ice packs. They would have been aware of their intervention. Unclear if personnel blinded.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain is a subjective outcome and influenced by participants who were not blinded
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Knee range of motion may be influenced by unblinded participant.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Insufficient information (from translation)
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included

Bao 2017.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Conducted between February 2015 and July 2016.
Participants	281 TKRs in 281 participants in Nanfang Hospital, Guangzhou, China. Posterior stabilised and conventional bone cement used. Sample size Group 1: 140 Group 2: 141 Information not given on demographic characteristics or baseline data of each group Inclusion criteria Unilateral TKR for knee OA Exclusion criteria Patients with infectious arthritis, diabetes, tumour, immune and blood system diseases, chronic inflammatory disease, severe osteoporosis, knee bone deformities or ligament insufficiency, intraoperative patella replacements, complications (incision infection, prosthesis infection, joint stiffness, bed sores, deep vein thrombosis (DVT) or other complications), body mass index (BMI) ≥ 30 kg/m2, active infection of body, muscle strength of quadriceps femoris muscle < Grade IV, history of open surgery or fracture in knee.
Interventions	Group 1: control group No cold therapy. Standard post‐op regimen (also offered to intervention groups) Group 2: intervention group Continuous ice treatment for 12 weeks after the operation, except during joint mobilisation training (twice/day). No detail given on duration and method of ice treatment. Both groups started walking using a walker after removal of drainage tube and continuous passive motion (CPM) was commenced on the first day after the operation.
Outcomes	Outcomes included in this review Pain scores ‐ visual analogue scale 0 to 10 at 6‐WPO, 12‐WPO Other outcomes Skin temperature of operated side
Notes	Article machine‐translated from Mandarin Chinese Participant withdrawals not discussed Conflicts of interest not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were assigned according to the random number table method
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants either had nothing or cold therapy. They would have been aware of their intervention. Unclear if personnel blinded.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain measurement is subjective and influenced by unblinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	Unclear risk	No self‐reported outcomes included
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Participant withdrawal not reported
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included

Brouwers 2022.

Study characteristics
Methods	Parallel‐group prospective controlled clinical trial. Conducted from August 2019 to March 2020.
Participants	106 TKRs in 106 participants in Martini Hospital, Groningen, Netherlands. Genesis II (Smith & Nephew) posteriorly stabilised prosthesis used. Use of cement not stated. Sample size Group 1: 51 Group 2: 51 Age (SD) Group 1: 69.2 (6.8) Group 2: 69.6 (9.1) Age range Not discussed Sex (M:F) Group 1: (22:29) Group 2: (22:29) BMI (SD) Group 1: 31.1 (5.9) Group 2: 31.5 (5.1) ASA score (1/2/3/4) Group 1: (3/38/10/0) Group 2: (5/38/8/0) TKA in contralateral knee (Y/N) Group 1: (14/37) Group 2: (20/31) Kellgren Lawrence Score (0/1/2/3/4) Group 1: (0/0/5/30/16) Group 2: (0/0/3/30/18) Analgesics schedule (Normal/No non‐steroidal anti‐inflammatory drugs (NSAIDS)/Other) Group 1: (30/10/11) Group 2: (28/9/14) Inclusion criteria Elective TKR for end‐stage OA over the age of 18 Exclusion criteria Rheumatoid arthritis, skin or other infections, vascular disease, other comorbidities on which cooling may have a negative effect, strong preference for one of the treatment options.
Interventions	Group 1: control group No cold therapy. Standard post‐op regimen (also offered to intervention groups) Group 2: cryotherapy group Computer‐assisted cryotherapy used with ZAMAR wrap (cooling brace): ZAMAR ZHC‐MG665A in hospital and ZAMAR ZT Cube with pre‐set program for home. Cryotherapy applied for 6 hrs at 6 °C immediately after surgery, then for 3 x 4 hour intervals at 8 °C on POD1 (morning, afternoon and evening). Application for 2 hours at 8 °C in morning and 2 hours at 10 °C in evening, with optional application at 8 °C in afternoon, on POD 2‐7. Both groups received compressive bandaging for 24 hours after surgery, full weight‐bearing mobilisation and active range of motion exercises starting on the day of surgery. Use of CPM not stated.
Outcomes	Outcomes considered in this review Pain scores ‐ numeric rating pain scale 0 to 10 at POD 0‐7, 2‐WPO, 6‐WPO Analgesia use ‐ number of participants that used 5 and 10 mg oxycodone as escape medication, and total number of units used (POD 0‐7) Knee range of motion ‐ maximal active flexion and extension measured with goniometer at 6‐WPO Knee swelling ‐ circumference measured in relaxed extended knee at mid‐patella, 7 cm proximally and 7 cm distally at 6‐WPO Activity level ‐ Timed Up and Go (TUG) at 6‐WPO Knee function ‐ Knee injury and Osteoarthritis Outcome Score (KOOS) 0 to 100; and Work, Osteoarthritis and joint‐Replacement Questionnaire (WORQ) 0 to 100 at 2‐WPO and 6‐WPO Serious adverse events Other outcomes Patient satisfaction ‐ numeric rating satisfaction scale 0 to 10 Length of stay ‐ cryotherapy administered after outcome measured Blood loss by drain in hospital ‐ cryotherapy administered after outcome measured
Notes	4 participants excluded before study commenced, balanced between groups: 2 excluded as they received another kind of prosthesis, 1 excluded because they had a strong preference for cryotherapy treatment, 1 excluded because surgery was postponed until after study period due to participant preference. Standard deviation was imputed in Analysis 1.11. 2 dropped out during the study, both from cryotherapy group: 1 developed delirium (POD 0‐6), 1 developed infection (after POD 7). Authors declare no conflicts of interest. Study funded by the Martini Hospital research and innovation fund.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Participants were grouped based on whether TKR was performed on an odd (control) or even (cryotherapy) week
Allocation concealment (selection bias)	High risk	No concealment
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants were informed of allocation group two weeks before surgery. All researchers, surgeons and assessors unblinded.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Outcomes are subjective and influenced by unblinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Outcome assessors were aware of group outcome; outcomes likely to be affected.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Small numbers withdrew, balanced between allocation groups.
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Only patients who had unilateral knee replacements were included.

Deng 2015.

Study characteristics
Methods	Randomised controlled trial Parallel groups Conducted from January 2010 to April 2012
Participants	50 participants with 50 TKRs performed in Hubei, China. Number and location of study centres not explicitly mentioned. Type of prosthesis not mentioned. Sample size Control: 25 Intervention: 25 Sex: M/F Control: 16/9 Intervention: 15/10 Mean age Control: 68.23 Intervention: 69.15 Age range Not discussed Inclusion criteria Primary unilateral surgery, no diabetes or coagulation dysfunction, no uncontrolled hypertension, can communicate with investigators Exclusion criteria High blood pressure, diabetes, severe cardiovascular or cerebrovascular disease, low leukocyte count, severe anaemia, hypoproteinaemia, coagulopathy, vasospasms, hypothermia, circulatory disorders, reduced sensation in affected limb
Interventions	In both groups, from post‐operative day 2, exercises were started for the legs. Control group Bandages only. One elastic bandage was wrapped from the sole to the proximal thigh. Intervention group Ice plus bandages. As well as the method described in the control group, 3 ice bags were then placed into an ice pack, which was placed on top of the knee and secured. Another bandage was used to compress the knee from 10 cm below to 10 cm above the knee joint. The ice pack was monitored once an hour to check for correct positioning, skin temperature, colour, blood supply or signs of frostbite. After 24 hrs of continuous ice application, ice was then applied intermittently, with 1 hr icing and 4 hrs break in between icings.
Outcomes	Main outcomes Blood loss: as defined by this equation: occult blood loss (g/L) = pre‐op Hb – post‐op Hb (minimum) – post‐op total drainage volume + blood transfusion volume Hb measurement taken once (1‐3 days) pre‐op and once post‐op (1‐3 days) Swelling: change in circumference of leg at tibial plateau The circumference was measured daily from day 1 to day 3 post‐op. The values were averaged then subtracted from the pre‐op circumference.
Notes	Article machine‐translated from Chinese and confirmed with a native speaker. No participant withdrawals. Article did not explain how often ice bags were replaced. Conflicts of interest not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: participants were randomised into 2 groups, but the method of randomisation was not explained
Allocation concealment (selection bias)	Unclear risk	Comment: not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Use of ice packs in intervention groups would be apparent to participant. Unclear if research personnel blinded.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Unclear risk	No self‐reported outcomes
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Assessors would have been aware of allocation based on whether participants had only local bandaging or extended bandaging. This could influence measurement of swelling.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: no participant withdrawals but no indication if participants were following study protocol
Selective reporting (reporting bias)	Unclear risk	Comment: insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Comment: inclusion criteria of only unilateral TKR

Gibbons 2001.

Study characteristics
Methods	Parallel‐group prospective randomised trial. No time period specified.
Participants	60 TKRs in 60 participants in Princess Royal Hospital, West Sussex, UK. Rotaglide Knee prosthesis (Corin) used. Use of cement not stated. Sample size Group 1: 30 Group 2: 30 Age Group 1: 70 Group 2: 71 Variance not reported Age range Not discussed Sex (M:F) Group 1: (11:19) Group 2: (14:16) Osteoarthritis/Rheumatoid arthritis Group 1: 25/5 Group 2: 27/3 Inclusion criteria Total knee replacement with Rotaglide Knee prosthesis (Corin) Exclusion criteria Not specified
Interventions	Group 1: cold compression dressing (CryoCuff) + patient‐controlled analgesia Cold compressive dressing (Cryo/Cuff, Aircast, UK) placed over knee immediately after procedure. Inflatable cuff was filled with ice‐cold water, which lasts up to 60 minutes. CryoCuff was used for a minimum of 6 hrs per day throughout hospital stay. Group 2: modified Robert Jones bandage + patient‐controlled analgesia Three layers of Velband and crepe placed over knee immediately after procedure. Bandage remained in place for 48 hrs. Neither group received continuous passive motion and the post‐operative physiotherapy regimen was the same.
Outcomes	Outcomes included in this review Blood loss (suction drainage in mL at 48 hrs) Pain scores (visual analogue scale 0 to 100 at POD 1, 3, 5, 7 and 9) Adverse effects Transfusion rate Analgesia use (mg/kg of morphine received via pump at 48 hrs; doses of co‐dydramol oral analgesia in mg at 10 days) Knee range of motion (degrees of flexion at POD 10 measured with goniometer) Length of hospital stay (days) Other outcomes None
Notes	Conflicts of interest not stated. Standard deviations were imputed in Analysis 1.8; Analysis 1.24; Analysis 1.31 No participant withdrawals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants had either a CryoCuff or a Robert Jones bandage: both participants and assessors would have been aware of this.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Outcomes (e.g. pain) subjective and may be influenced by unblinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Narcotics use, length of stay and knee range of motion may be influenced by unblinded patient.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals
Selective reporting (reporting bias)	High risk	Length of hospital stay and complications do not seem to be part of the original methods and materials.
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included

Healy 1994.

Study characteristics
Methods	Parallel‐group prospective randomised trial. No time period specified.
Participants	Sample size: 105 knees in 76 participants (29 participants had bilateral TKRs) in Lahey Medical Centre, Massachusetts, USA. Porous coated anatomic modular or a Duracon prosthesis used (both cemented) Phase 1: 48 knees in 36 participants Group 1a: 21 knees in 17 participants (4 bilateral TKRs) Group 1b: 27 knees in 19 participants (8 bilateral TKRs) Phase 2: 57 knees in 40 participants Group 2a: 34 knees in 23 participants (11 bilateral TKRs) Group 2b: 23 knees in 17 participants (6 bilateral TKRs) Indication: 75 participants had osteoarthrosis, 1 had osteonecrosis Inclusion criteria: primary total knee replacement Demographic data not discussed. Exclusion criteria not specified
Interventions	Groups A: control ACE wrap and ice packs, made up of a plastic sac containing crushed ice, were applied to the knee after operation. CPM was used in both intervention and control groups. Groups B: cold compressive dressing Cold compressive dressing (Cryo/Cuff, Aircast, USA) applied "after operation" (exact time of application not specified). The vinyl bladder was attached by Velcro straps to the leg so that it draped over the suprapatellar pouch and the medial and lateral gutters. The cuff is filled with ice water, and delivers both compression and cooling to the knee. Phase 1: ice water in CryoCuff dressing exchanged every four hours Phase 2: ice water changed every one to two hours
Outcomes	Outcomes included in this review Blood loss (wound drainage mL) ‐ no method of measurement stated Adverse effects Analgesia use (morphine equivalents in mg at POD 1‐3, and POD 4‐7) Knee range of motion (active flexion in degrees; POD 2‐4; POD 7‐14 ; WPO 4‐6) Knee swelling (mid‐patella in cm; distal thigh 2.54 cm proximal to patella in cm; POD 2‐4; POD 7‐14; WPO 4‐6) Other outcomes None
Notes	Outcomes were measured on TKRs, however, randomisation was done by participant, although many participants had bilateral TKRs. The intervention took place in two consecutive "phases" in the trial, designed to examine any differences in effect in the timing of cryotherapy. The "phases" included a different set of participants and does not imply the same participants had the intervention twice. Bilateral TKR patients had the same treatment on both knees. No adjustment was made for lack of independence. The treatment for participants in each phase was the same and the phase just reflected the different times in which participants were recruited. Participants were thus not randomised to phase. Since results were reported separately across phases, we included each phase as a separate analysis using the participants numbers and results reported for each phase in the article. Standard deviations were imputed in Analysis 1.4; Analysis 1.6; Analysis 1.14; Analysis 1.28; Analysis 1.29 No participant withdrawals. Supported by a research grant from Aircast (New Jersey), manufacturers of the Cryo/Cuff.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants had either a CryoCuff or ice pack plus compression bandage applied.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Unclear risk	No self‐reported outcomes
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Narcotic use and knee range of motion may be influenced by participants who were not blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	High risk	Data were collected from individual knees in 29 bilateral TKR patients and no adjustment for lack of independence.

Ivey 1994.

Study characteristics
Methods	Parallel‐group prospective randomised trial. No time period stated.
Participants	90 TKRs in 90 participants in Texas, USA. 57 female; 31 male. Cemented cruciate sacrificing prosthesis (Zimmer IB‐II) used. Number and location of study centres not explicitly stated. Sample size Group 1: 28 Group 2: 30 Group 3: 30 Age Group 1 ‐ Age bracket: 53 to 85 years Mean age: 64.5 years SD: 8.1 years Group 2 ‐ Age bracket: 36 to 85 years Mean age: 64.2 years SD: 10.3 years Group 3 ‐ Age bracket: 36 to 88 years Mean age: 66.9 years SD: 36 years Sex (M:F) Group 1: (12:16) Group 2: (11:19) Group 3: (8:22) Knee side (L:R) Group 1: (13:15) Group 2: (17:13) Group 3: (15:15) Indication (Osteoarthritis: rheumatoid arthritis) Group 1: (27:1) Group 2: (28:2) Group 3: (28:2) Inclusion criteria Primary unilateral TKR Exclusion criteria Not discussed
Interventions	Following surgery, a single layer of gauze/cast padding dressing was applied. Hot Ice (Proaction Medical, Texas) thermal pads (no compression) were applied to either side of the incision and were attached to the Hot Ice machine. They were held in place with three additional layers of cast padding and an elastic bandage. A knee immobiliser was applied and the thermal pads were attached distally to the Hot Ice machine. After 72 hrs, the post‐operative dressings and thermal pads were removed. Use of CPM was not stated. Group 1: machine settings generated a temperature of 50 degrees Fahrenheit (10 °C) Group 2: machine settings generated a temperature of 60 degrees Fahrenheit (15.5 °C) Group 3: machine set at room temperature, 70 degrees Fahrenheit (21.1 °C)
Outcomes	Outcomes included in this review Adverse effects Analgesic use (mg/hr of morphine injected over 72 hrs post‐op; attempts of patient‐controlled analgesia (PCA) use ‐ attempts/hr over 72 hrs post‐op) Other outcomes None
Notes	2 participants were dropped from the 50‐degree Fahrenheit group because post‐op PCA records were lost. Outcomes were adjusted for age and body weight using ANOVA (analysis of variance). We halved the number on the control group to prevent duplicate examination of control participants. Conflicts of interest not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Prior to the study, 30 sets of three random numbers were drawn from the uniform distribution on the interval 0,1 using the function UNIFORM in the SAS System. The three random numbers in each set were then transformed into ranks. This resulted in 30 sets of random sequences each with the integers 1,2,3. Participants entering the study received one of three treatments according to the following assignment: 1 ‐ 50 °F; 2 ‐ 60 °F, 3 ‐ 70 °F.
Allocation concealment (selection bias)	Unclear risk	Machines were programmed to a specific temperature ‐ but study did not describe by whom, or whether this was before participants were assigned.
Blinding of participants and personnel (performance bias) all outcomes	Low risk	All participants had Hot Ice Machine attached ‐ but at one of three different temperatures. Machines were programmed and participants were unaware of which temperature group they were assigned to.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Unclear risk	No self‐reported outcomes
Blinding of outcome assessment (detection bias) objective outcomes	Low risk	Measurement of outcomes unlikely to be affected by assessor knowledge of intervention.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Missing data not balanced between groups. "Two patients in group 1 were dropped from the study group because their postoperative records monitored by the PCA pump were lost."
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included

Kang 2014.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Conducted September to October 2013.
Participants	46 participants with at least 46 TKR (bilateral TKR not specified) from M hospital, Korea. All females. Type of prosthesis not stated. Sample size Group 1: 15 Group 2: 15 Group 3: 16 Age (years) Group 1: 66.6 +/‐ 3.9 Group 2: 68.5 +/‐ 4.7 Group 3: 67.6 +/‐ 3.5 Age range Not discussed Height (cm) Group 1: 155.4 +/‐ 4.7 Group 2: 156.2 +/‐ 4.6 Group 3: 157.1 +/‐ 3.6 Weight (kg) Group 1: 56.6 +/‐ 3.7 Group 2: 59.1 +/‐ 8.2 Group 3: 58.9 +/‐ 5.1 Inclusion criteria Primary TKR Exclusion criteria Issues in other musculoskeletal and nervous system lesions other than the surgical site
Interventions	All groups were given CPM (continuous passive motion) for 30 minutes and TENS (transcutaneous electrical nerve stimulation) for 15 minutes. Interventions given once a day, 5 times a week for 3 weeks. Group 1: low‐intensity pulsed ultrasound therapy using Care Star GM‐002 (Genemedi, Korea) for 1 minute at 1 MHz and 0.4 W/cm2 Group 2: Cryotherapy using Crais (Century, Korea) (i.e. cold air jet machine) for 5 minutes Group 3: both therapies
Outcomes	All outcomes measured 1 day post‐surgery and repeated 4 times a week for 3 weeks Knee function – self‐assessment Korean Western Ontario and McMaster Universities Arthritis Index (K‐WOMAC) questionnaire (score out of 96), with questions about pain, joint stiffness, difficulties in daily life. Higher score indicates greater severity of knee joint arthritis. Knee range of motion – participant lies face down and range of motion is measured with a goniometer using the lateral femoral condyle, greater trochanter and fibular lateral malleolus as reference points C‐reactive protein – 5 cc of participant's blood tested with latex agglutination turbidimetry using monoclonal antibody A15
Notes	Study does not state how long after surgery the intervention was applied; however, the measurement of outcomes started 1 day after surgery. Participant withdrawals not discussed Conflicts of interest not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "participants were randomly allocated" Comment: no other details provided about method of randomisation
Allocation concealment (selection bias)	Unclear risk	Comment: not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Use of ultrasound machine and cryotherapy machine interventions make both participants and assessors aware of allocation.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Measurement of knee function influenced by unblinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Knee range of motion may be influenced by participant.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: participant sample size used for data evaluation not specified. Participant withdrawal not reported.
Selective reporting (reporting bias)	Unclear risk	Comment: insufficient information
Other bias ‐ bilateral knee replacements included	Unclear risk	Comment: whether any participants had bilateral TKR was not stated

Kullenberg 2006.

Study characteristics
Methods	Randomised controlled trial. No time period specified.
Participants	86 TKRs in 86 participants in Karlshamn, Sweden. 50 females, 36 males. Mean age 68 years. Type of prosthesis not stated. Number and location of study centres not explicitly mentioned. Sample size Group 1: 43 Group 2: 40 Age Group 1: 68.1 +/‐ 6 Group 2: 68.9 +/‐ 6.8 Age range Not discussed Sex (M:F) Group 1: (1:1.35) Group 2: (1:1.43) Inclusion criteria Primary unilateral TKR Exclusion criteria Coagulopathies and preclosure tourniquet deflation
Interventions	Group 1: CryoCuff plus standard care Cold compressive dressing (CryoCuff, Aircast, USA) was applied immediately post‐op after tourniquet release. Cuff was inflated with ice water and rechilled every 60 minutes for 3 days. The cuff was connected to canisters containing 4 L of iced water providing 30 mmHg of cold compression. Ice was changed in the canisters every 4 hours. Partcipants also received usual practice of epidural anaesthesia described in the control group. Use of CPM not stated. Group 2: standard care Epidural analgesia with ropivacaine applied immediately until post‐operative day 3. Then intravenous analgesia with non‐steroidal anti‐inflammatory drugs and opioids. Pain was controlled every 2 hours in all participants. Oral tramadol (150 mg x 2) and paracetamol (1 g x 4) was started post‐operatively. Breakthrough pain was treated with ketobemidon (2.5 to 5 mg) intravenously.
Outcomes	Outcomes included in this review Blood loss (Hb drop to POD 1 in mmol/L) Pain (visual analogue scale 0 to 10 on POD 1, 3, at rest and during exercise) Adverse effects Analgesic use (mg/kg/24 hrs of morphine over total stay) Knee range of motion (degrees of flexion and extension at POD 1, at discharge (4 to 5 days) and at 3 weeks post‐op) Length of hospital stay (days) Other outcomes Knee weight bearing (measured using 2 separate bedside scales and recorded as percentage of bodyweight). Time to completed in‐hospital rehab. Assessed by physiotherapists and measured in days.
Notes	2 participants were excluded in intervention group and 1 in the control group due to complications. One participant in the cold compression group and another in the control group developed deep vein thrombosis below the popliteal fossa. One knee in the cold compression group developed a superficial soft tissue infection of unclear origin. These participants were excluded. No benefits or funds were received in support of this study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Sealed envelopes were used to assign participants, but the study did not describe whether the envelopes were sequentially numbered, opaque, tamper‐proof or who handled them.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants were either treated with a CryoCuff or with standard analgesic medication. They would be aware of intervention.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain is a subjective outcome and likely to be affected by unblinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Narcotic use, knee range of motion and length of stay may be influenced by participants who were not blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Small numbers lost to follow‐up, balanced between allocation groups.
Selective reporting (reporting bias)	High risk	Range of movement for flexion and extension was to be recorded, but only one figure was recorded.
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

Levy 1993.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Time period not stated.
Participants	100 TKRs in 80 participants (20 bilateral TKRs) in Pennsylvania, USA. Prosthesis according to surgeon preference, but none were posterior stabilised. Use of cement not stated. Number and location of study centres not explicitly discussed. Sample size Group 1: 40 Group 2: 40 Age (average) Group 1: 74 Group 2: 73 Age range Not discussed Sex (M:F) Group 1: (7:33) Group 2: (8:32) Weight (average) Group 1: 84.4 kg Group 2: 77.9 kg Pre‐operative haemoglobin Group 1: 11.8 mg Group 2: 12.3 mg Inclusion criteria Primary TKR Exclusion criteria Intraoperative fractures, coagulopathies, drain pullout, and preclosure tourniquet deflation
Interventions	Group 1: Aircast CryoCuff After skin closure, incision was covered with sterile gauze pads and two layers of sterile Webril. Cold compressive dressing (CryoCuff, Aircast, USA) applied immediately after surgery and inflated with ice water. An additional ACE wrap was applied from the toes to the inferior edge of the cuff. Tourniquet was then deflated. Rechilling of ice water was performed every 90 minutes for 3 days. Use of CPM not stated. Group 2: control After skin closure, incision was covered with gauze pads, four layers of Webril and an Ace wrap (Robert Jones bandage) from the toe to thigh before the tourniquet was released.
Outcomes	Outcomes included in this review Blood loss ‐ drain output (drain measured over 48 hrs in mL and adjusted for weight); haemoglobin drop (from pre‐op to POD 3 in mg); total body blood loss (in mL, calculated by the Gross method); extravasation of blood into soft tissue (mL) Pain ‐ visual analogue scale 0 to 10 (POD 1; POD 2; POD 3) measured prior to physiotherapy Adverse effects Analgesia use ‐ total injected morphine over 48 hrs (mg; mg/kg/day) Knee range of motion ‐ total arc at POD 7 and POD 14 (degrees); measured by one of two physiotherapists Swelling ‐ measured only on 20 participants undergoing bilateral TKR. Measured in cm as the circumference 2 cm above the patella, pre‐op and days 3 and 7 post‐op. Other outcomes None
Notes	100 TKRs in 80 participants. Results from 80 participants (one knee per participant) were studied. For participants in bilateral group: participants were randomised and second knee was chosen for intervention to prevent contamination. Results were only reported for 80 knees. Two participants who received uncemented knee replacements were excluded from the study. Not stated which allocation groups the participants were from. Study supported by Aircast (manufacturers of CryoCuff).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Sealed envelopes were used to assign participants. These were opened at the time of skin closure to prevent potential surgeon bias. But study did not describe whether the envelopes were sequentially‐numbered, opaque, tamper‐proof or who handled them.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Either a CryoCuff or a compression bandage was applied. Participants would have been aware of intervention.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Subjective measurement of pain would have been influenced by unblinded participants.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Narcotic use and knee range of motion may be influenced by participants who were not blinded. Unclear if physical therapist measuring other objective outcomes (e.g. knee range of motion, swelling) was blinded.
Incomplete outcome data (attrition bias) All outcomes	High risk	Two participants who received uncemented knee replacements were excluded from the study. Not stated which allocation groups those participants were from.
Selective reporting (reporting bias)	High risk	Extravasation of blood into soft tissues was not mentioned in the materials and methods section as a form of blood loss measurement.
Other bias ‐ bilateral knee replacements included	Unclear risk	Bilateral TKRs were included in 20 participants, but outcomes were measured at the participant level.

Morsi 2002.

Study characteristics
Methods	Split‐body controlled clinical trial. Time period not specified.
Participants	30 bilateral TKRs in 30 participants from Menoufyia, Egypt. All components were cemented and all participants were placed in a continuous passive motion machine after the operation. Location and number of study centres not explicitly stated. Sample size: 30 participants Group 1: 30 TKR Group 2: 30 TKR Demographic data Not discussed Inclusion criteria Primary TKR for OA Exclusion criteria Not discussed
Interventions	Group 1: Flow‐flow cooling device plus compression bandage This was the first knee operated on. After skin closure and tourniquet deflation, the knee was covered with a light gauze dressing and 1 layer of sterile Webril. The cooling device was then wrapped, with a final layer of compressive crepe bandage. Cooling flow was applied continuously for 6 days post‐operatively ‐ except during ambulatory and knee mobilisation exercises. Skin temperature was kept at 7 °C ± 2 °C for the first 2 hrs postoperatively. Then skin temperature was maintained at 12 °C ± 3 °C. The cooling device used consisted of 2 glass bottles. 1 bottle was filled with solution such as saline, and the other bottle was left empty. The 2 bottles are connected to each other by 5 m of plastic intravenous infusion tubing. The tubing was applied over the surgical dressing in a coil‐like manner around the knee, which then was wrapped with a compressive crepe bandage. The full bottle is hung over the patient in a stand, and the empty bottle is put into a tank of crushed ice. The cooling solution flows from the bottle on the stand by gravity to the cooling coil around then to the empty bottle. The button on the plastic intravenous infusion tubing can control the flow of the fluid. When the higher bottle empties, and the bottle in the in the ice tank is full, the bottles are swapped. Group 2: control This was the second knee operated on, 6 weeks after the first. The knees were covered as above, with gauze, Webril and a compression bandage, but no cooling device.
Outcomes	Outcomes considered in this review Blood loss ‐ total stay blood loss (mL); total stay haemoglobin decrease (mg); total stay extravasation of blood into soft tissue (mL) Pain ‐ visual analogue scale Adverse effects ‐ cold‐related complications Transfusion rate ‐ units Analgesia use ‐ ratio of participants who received analgesia at POD 1 and POD 6 (intervention: control); total stay (mg); total stay (number of pills) Knee range of motion ‐ flexion and extension every week for 6 weeks (degrees) Other outcomes Wound healing
Notes	30 sequential bilateral TKRs were included, with the second operation always being the control (non‐cryotherapy) group. Bias may have been introduced in that outcomes measured in the control group were always measured after the participants' previous experience with a knee replacement. Measurement of blood loss may have also been biased as 6 weeks is insufficient time for haemoglobin levels to normalise following an initial TKR. Standard deviations were imputed in Analysis 1.9. No benefits or funds were received in support of this study. No participant withdrawals.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Each participant received both treatments ‐ one on either side.
Allocation concealment (selection bias)	High risk	Each participant received both treatments ‐ one on either side. The first TKR was to receive one treatment, and the second TKR to received the other treatment.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants received both treatments ‐ one on either side.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Outcomes measured in the control group were always measured after the participants' previous experience with a knee replacement. This could influence self‐reported outcomes.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Assessors would be aware of group allocation based on previous TKR. Blood loss measurement may have been biased, as 6 weeks is insuffiicent for haemoglobin levels to normalise after an initial TKR.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals.
Selective reporting (reporting bias)	High risk	Incomplete data for analgesic use and range of motion. Extravasation of blood into soft tissue and transfusion rate/units were recorded but not part of original materials and methods section.
Other bias ‐ bilateral knee replacements included	Low risk	All participants included (30) had sequential bilateral TKRs in this study, but the study was non‐randomised and therefore no unit of analysis issues exist.

Radkowski 2007.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Trial completed December 2002.
Participants	67 TKRs in 67 participants in North Carolina, USA. 3 participants were excluded for various reasons. Prosthesis type not stated. Number and location of study centres not explicitly specified. Sample size Group 1: 28 Group 2: 36 Age (mean) Group 1: 63.7 Group 2: 66.9 Age range Not discussed Sex (M:F) Group 1: (15:13) Group 2: (13:23) BMI (mean) Group 1: 32.6 Group 2: 30.7 Race/ethnicity Group 1: African American (3.6%); Caucasian (understood to be White) (89.3%) Group 2: African American (16.7%); Caucasian (understood to be White) (80.6%) Inclusion criteria Undergoing a primary TKA procedure performed for osteoarthritis by one of the participating surgeons. Exclusion criteria Concomitant diabetes mellitus, opioid dependency or intolerance, skin diseases affecting the surgical knee, previous pyogenic arthritis or osteomyelitis in the ipsilateral knee, bilateral procedures, rheumatoid arthritis, generalised neurological conditions, or those without a residential phone line.
Interventions	The cuff from the cryotherapy device (Thermo‐Tek Solid State Recirculating Chiller, Thermo‐Tek Inc.) was placed by the surgeon around the knee over the operative dressing immediately after surgery. This device consists of a cuff filled with circulating cold water at the desired temperature via a tube attached to the cooling device. A 3‐layer dressing consisting of 4 x 4‐inch gauze, cast‐padding, and ACE wrap was placed in the interface between the cryotherapy device and the skin. The consistency of temperature over time was monitored by temperature probes placed on the skin surface. The cryotherapy device was used continuously during the participants' hospital stay. CPM not used. Group 1: cryotherapy at 45 degrees Fahrenheit (7.2 °C) and continued until discharge and ACE wrap Group 2: cryotherapy (cuff padding at room temperature) at 75 degrees Fahrenheit (23.8 °C) and ACE wrap
Outcomes	Outcomes considered in this review Blood loss ‐ wound drainage (drain output) at POD 1 (mL) Pain ‐ verbal analogue scale (0 to 10) measured daily at POD 1, 2 and 3 during inpatient stay. Then through patient‐recorded diary entries measured daily for 7 days after discharge. Then through telephone interview at POD 30. Adverse events Analgesia use ‐ dichotomised into group who required additional narcotic analgesia and those who did not (regional nerve block used as standard and morphine patient‐controlled analgesia (PCA) for additional analgesia Knee range of motion ‐ passive knee flexion (degrees); range (arc) of motion; measured with hand‐held goniometer Knee function ‐ self‐reported knee function using International Knee Documentation Committee (IKDC) questionnaire (24) measured daily at POD 1, 2 and 3 during inpatient stay. Then through patient‐recorded diary entries measured daily for 7 days after discharge. Then through telephone interview at POD 30. Other outcomes Efficacy of cryotherapy device. Skin temperatures were recorded in degrees using probes
Notes	Study supported by DeRoyal Industries. No participant withdrawals.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were assigned by using a computer‐generated blocked randomisation.
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	Low risk	An administrative assistant set and concealed the assigned interventions in each cryotherapy device. The participants, participating surgeons, research nurse and clinical epidemiologist performing the statistical analysis were all blinded to the treatment assigned to each participant.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Low risk	Pain and function measured by blinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	Low risk	Research nurse performing assessments was blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals.
Selective reporting (reporting bias)	High risk	Participant diary results were not recorded. Self‐reported knee function was not recorded. Passive knee range of motion was not recorded.
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

Scarcella 1995.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Conducted between 1990 and 1991.
Participants	24 TKRs in 24 participants in Ohio, USA. 4 participants were later excluded for various reasons. Prosthesis type not stated. Location and number of study centres not explicitly stated. Sample size Group 1: 12 Group 2: 12 Age Group 1: 69 Group 2: 67 Age range 47 to 79 Sex Not reported Inclusion criteria Primary total knee replacement Exclusion criteria Not discussed
Interventions	Immediately following the procedure, the cold therapy unit was applied to the operative site, usually over a light dressing consisting of two 4x8‐inch pads and two 8x10‐cm gauze pads. The cold therapy unit employed in this study was a thermal blanket (Hot/Ice Blanket, Thermo Temp, USA) The unit consists of a rubber pad connected by a rubber hose to the main cooling unit. A special cloverleaf‐shaped pad was available to better encompass the midline knee incision. CPM use not stated. Drains not used. The units were checked on a daily basis. Participants wore the cold therapy pads continuously until the time of discharge. Group 1: cryotherapy at 50 degrees Fahrenheit (10 degrees Celsius) until discharge Group 2: control at 70 degrees Fahrenheit (21.1 degrees Celsius), until discharge
Outcomes	Outcomes considered in this review Adverse effects Analgesic use ‐ total stay (mg/kg of meperidine) of total parenteral narcotics used Knee range of motion ‐ daily progression (degrees); at discharge (degrees) Length of hospital stay ‐ days Activity level ‐ arbitrarily picked at the day the participant received a walker for use in the room (time to independent ambulation) Other outcomes None
Notes	3 participants excluded as they had patient‐controlled analgesia (PCA), and 1 because of unrelated complications. Excluded participants remained on results tables. Conflicts of interest not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Participants were assigned to their intervention based on whether it was part of their post‐operative regimen.
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	Participants were not informed that a study was in progress. Participants assigned to the control group were not aware they were in a placebo group. They all were supplied with cold therapy units but at one of two different temperatures. Study did not mention whether assessors were blinded.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Unclear risk	No self‐reported outcomes
Blinding of outcome assessment (detection bias) objective outcomes	Unclear risk	Unclear if person assessing outcomes was aware of group assignment.
Incomplete outcome data (attrition bias) All outcomes	High risk	3 participants who received patient‐controlled analgesia were withdrawn from the study. Not stated which allocation group they were from.
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

Smith 2002.

Study characteristics
Methods	Randomised controlled trial. No time period stated.
Participants	84 TKRs in 84 participants in Lismore, Australia. Prosthesis not stated. Location and number of study centres not stated. Sample size Group 1: 40 Group 2: 44 Age (SD) Group 1: 72 (7.1) Group 2: 72.1 (7.8) Age range Not discussed Sex (M:F) Group 1: (21:19) Group 2: (21:23) Inclusion criteria Unilateral primary TKR Exclusion criteria Haematological or coagulation disorder
Interventions	A Robert Jones compression bandage was applied after the procedure for all participants. Group 1: compression bandage Robert Jones bandage applied immediately after surgery and removed after 24 hours. Ice bags were then applied for 15 minutes three time a day for the next 48 hours. Group 2: cold therapy Robert Jones bandage applied immediately after surgery as per control group. Robert Jones bandage removed after 6 hrs and cryotherapy delivered via cryo‐pad machine with the pad placed directly over but not completely covering incision site. Cryo‐pad infused with water between 2 °C and 5 °C for 15 minutes, then deflated for 15 minutes. The cryo‐pad was removed after 24 hrs. Following each intervention, both groups received an ice pack wrapped in a linen towel, which was applied to the wound site of all participants for 15 minutes 3 times a day for the next 24 to 48 hours. Use of CPM not stated.
Outcomes	Outcomes included in this review Blood loss ‐ wound drainage in 48 hrs (mL); transfusion volume (mL); haemoglobin (g/L); Pain ‐ visual analogue scale (0 to 10) at POD 1, POD 2 and POD 3 Analgesic use ‐ total opiate use (morphine) in 48 hrs post‐surgery (mg/kg) Knee range of motion ‐ degrees (pre‐operative flexion; flexion at 24 hrs; flexion at 48 hrs) Knee swelling ‐ mm (pre‐operative; 24 hrs; 48 hrs). Specific measurement not stated. Length of hospital stay ‐ days Other outcomes Blood transfusion requirements in mL
Notes	Conflicts of interest not stated No participant withdrawals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were assigned using a coin toss at their pre‐admission visit.
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants either had ice in plastic bags plus compression bandage or a cryo‐pad (no compression) applied. They would therefore be aware of allocation.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain is subjective and likely to be influenced by unblinded participant.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Analgesic use, knee range of motion and length of hospital stay could have been affected by unblinded participants.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

Stocker 2016.

Study characteristics
Methods	Randomised trial, parallel groups September 2014 to March 2015
Participants	16 participants had 16 TKRs in a single Swiss hospital. PFC Sigma PePuy (Johnson & Johnson) prosthesis used. Use of cement not stated. Sample size Group 1: 8 Group 2: 8 Sex: M/F Group 1: 4/4 Group 2: 5/3 Age Group 1: 68 +/‐ 8.6 Group 2: 73 +/‐ 6.3 Age range Not stated Height (m) Group 1: 1.7 +/‐ 0.04 Group 2: 1.7 +/‐ 0.07 Weight (kg) Group 1: 78.3 +/‐ 6.7 Group 2: 85.3 +/‐ 9.6 BMI Group 1: 27.9 +/‐ 2.3 Group 2: 31.1 +/‐ 3.4 Inclusion criteria 45 to 90 years old, participant consent, knowledge of German, could walk 10 minutes with/without walking stick, undergoing single TKR for degenerative or post‐traumatic osteoarthritis, planned anaesthesia with femoral nerve block Exclusion criteria BMI > 40, relevant cardiac or neurological deficits, pre‐existing lymphoedema, uncontrolled diabetes mellitus, renal failure stage 3, major osteoporosis, rheumatoid arthritis in knee, additional tuberositas osteotomy, peripheral arterial occlusive disease from stage 2b, change in diuretics use after surgery, post‐thrombotic syndrome, tumours, alcohol or drug abuse, psychological disorders (e.g. depression)
Interventions	Post‐operatively, all participants had femoral nerve block (1 to 3 days) and physiotherapy (days 1 to 5 daily) Interventions applied from the first to the fifth post‐operative day. Group 1: multi‐layer compression bandage (gauze, padding, bandage) applied by physiotherapists after surgery with participant supine, knee at 0‐20°, from mid‐lower leg to groin. The bandage was worn for 22 hours per day and removed for the remaining 2 hours to perform measurements, wound control, reapplication after discomfort. Group 2: cold pack (Physiopack stored at ‐19°) wrapped in a towel and applied for 10 minutes, 3 times a day. In addition, for mobilisation, bandages were used on day 1, 2 then anti‐thrombosis T.E.D. socks from day 3 onwards.
Outcomes	Outcomes measured pre‐operatively and post‐operatively on days 1, 3, 6, 42: Swelling: measurement of circumference at 5 points on the operated leg: around the knee joint, 5 cm, 10 cm and 15 cm proximal to the knee and 15 cm distal to the knee. Knee range of motion: measured on supine participant with goniometer Pain: measured with 11‐point numerical rating scale Activity/function: fast‐paced walking test: the participant walked as fast as possible a distance of 20 meters without excessive effort, turned around calmly (the time was not measured when turning) and then ran back the 20 meters as fast as possible, again without excessive effort. Also measured: Length of hospital stay (days) Adverse effects
Notes	Article machine‐translated from German Authors declared no conflicts of interest No participant withdrawals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: participants randomised, but method of randomisation not stated
Allocation concealment (selection bias)	Low risk	Quote: "closed, non‐translucent envelopes and with the help of a person not included in the study"
Blinding of participants and personnel (performance bias) all outcomes	High risk	Use of ice packs in intervention groups would be apparent to participants. Assessors were blinded by removing all material before the measurement and preventing participants from commenting on the intervention to the assessor.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Self‐reported pain and activity level measurements may have been influenced by unblinded participants.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Length of stay and knee range of motion could have been influenced by unblinded participants.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no participant withdrawals
Selective reporting (reporting bias)	Low risk	Comment: all outcomes in trial protocol reported
Other bias ‐ bilateral knee replacements included	Low risk	Comment: inclusion criteria of only unilateral TKR

Thijs 2019.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Conducted between 2015 and 2016.
Participants	60 TKRs in 60 participants in Zuyderland Medical Center, Sittard‐Geleen, the Netherlands. Cemented cruciate retaining prosthesis (Vanguard CR) used. Sample size Group 1: 30 Group 2: 30 Age (SD) Group 1: 65.5 (6.2) Group 2: 64.7 (6.8) Age range Not discussed Sex (M:F) Group 1: (17:13) Group 2: (15:15) BMI (SD) Group 1: 27.7 (3.6) Group 2: 27.4 (3.6) Knee side (L:R) Group 1: (11:19) Group 2: (13:17) ASA classification (1/2/3/4) Group 1: (17/13/0/0) Group 2: (15/15/0/0) Inclusion criteria TKR from osteoarthritis Exclusion criteria General or active knee infection, revision TKR, extension deficit of the knee of more than 15° and/or a flexion less than 100°, rheumatoid arthritis, pregnancy, patients unable to understand and complete procedure due to cognitive dysfunction or language barrier.
Interventions	Group 1: cold group Computer‐assisted cryotherapy (ZAMAR Therapy Cube) at home. Cryotherapy applied for 6 hrs at 10 °C immediately after surgery then an additional 4 hrs at 10 °C at night on POD 0. Application for 2 hrs at 10 °C in morning, 2 hrs at 10 °C in afternoon, and 4 hrs at 12 °C at night (with option to continue for longer if desired) on POD 1‐6. Group 2: warm group In the warm group, participants followed exactly the same cooling protocol but with another temperature setting of 21 °C for all cooling sessions. Both groups received compressive dressing in operating theatre, replaced by a thin elastic compressive tubular dressing (Tubigrip) before leaving the hospital to allow cryotherapy to be applied. They also received full weightbearing mobilisation and active range of motion exercises starting on the day of surgery. CPM was not used.
Outcomes	Outcomes considered in this review Pain scores ‐ numeric rating pain scale (NRS) 0 to 10: difference in rating before and after each time computer‐assisted cryotherapy was used and at 2‐WPO, 6‐WPO, 12‐WPO; EuroQol Visual Analogue Scale (EQ‐VAS) 0 to 100 at 2‐WPO, 6‐WPO, 12‐WPO Analgesia use ‐ number of tablets and equivalent dosage of tramadol used as escape medication at POD 0‐4 and in total Knee swelling ‐ circumference measured at mid‐patella, 7 cm proximally and 7 cm distally at 1‐WPO, 2‐WPO and 6‐WPO Adverse events ‐ patient‐related (e.g. nausea, vomiting, thromboembolic events, persistent wound leakage), surgical related (e.g. infection), prosthesis related (e.g. loosening) Knee function ‐ Dutch Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) and Oxford Knee Score (OKS, 12‐60) at 2‐WPO, 6‐WPO, 12‐WPO Quality of life ‐ EQ‐5D index (0‐1) at 2‐WPO, 6‐WPO, 12‐WPO Other outcomes Presence of visual haematoma ‐ measured at 1‐WPO, 2‐WPO, 6‐WPO
Notes	3 participants from cold group not discharged at POD 0 as planned (all discharged at POD 1) 4 participants from warm group not discharged at POD 0 as planned (all discharged at POD 1) One participant had an "unsatisfied feeling" and decided to stop the therapy. One author was a paid consultant for Zimmer‐Biomet (manufacturers of cryotherapy device used). Dutch WOMAC 0‐100 used in this paper as per recommendation from Roorda 2004
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were assigned using computer, web‐based generated randomised numbers.
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	Randomisation and preparation of computer‐assisted cryotherapy system done by independent trial nurse. However, computer‐assisted cryotherapy performed at home so unclear if participants could just see screen.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Unclear risk	Unclear if self‐reported outcomes influenced by participants.
Blinding of outcome assessment (detection bias) objective outcomes	Low risk	Outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals related to cryotherapy. "One patient had an unsatisfied feeling and decided to stop the therapy"?
Selective reporting (reporting bias)	High risk	Range of motion noted in protocol but not included. Adverse events and follow‐up beyond 7 days included but not mentioned in protocol.
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

Walker 1991.

Study characteristics
Methods	Randomised controlled trial conducted between 1984 and 1988 in three phases. Phase 3, which investigated cryotherapy, was conducted between 1987 and 1988 and was included in this review.
Participants	30 TKRs in 30 participants in California, USA. Sex not stated. Cemented total condylar semi‐constrained prostheses used. Location and number of study centres not explicitly discussed. Sample size Group 1: 15 Group 2: 15 Age Group 1: 75.0 (58 to 87) Group 2: 70.0 (56 to 82) Sex Not discussed Diagnosis (osteoarthritis/rheumatoid arthritis/avascular necrosis) Group 1: (13/0/2) Group 2: (12/0/3) Inclusion criteria Unilateral TKR Exclusion criteria Not discussed
Interventions	For both groups, the knee was initially splinted except for during physical therapy, a standardised post‐unilateral TKR physical therapy was begun on POD 1, with range of motion exercises also begun on POD 1. CPM was initiated on POD 3 with range 0° to 45°. CPM was continued throughout hospitalisation with daily advancements of flexion by 10° to tolerance. Group 1: CPM plus continuous cooling pad (CCP) Continuous cooling pad (without compression) applied post‐op in the recovery room. The pad was applied over a light dressing and under a knee immobiliser. It was worn continuously for a minimum of 3 post‐operative days. The CCP consisted of a flexible thermal blanket held in place at the knee over a light sterile dressing by elastic mesh. Two tubes allowed continuous circulation of water between the pad and a bedside cooling unit. The water bath was maintained at 50 °F to 55 °F. 15 TKRs in 15 participants. Mean age 75. CPM was used in both intervention and control groups Group 2: CPM Standard post‐op care. This included CPM for all participants, started at day 3 post‐op and continued until discharge.
Outcomes	Outcomes considered in this review Blood loss ‐ wound drainage (cc). No time period specified Adverse effects Analgesia use ‐ total intramuscular and oral morphine (mg); total analgesia in morphine equivalents (mg) Knee range of motion ‐ pre‐operative flexion/discharge flexion/flexion arc (degrees); 3‐month flexion/flexion arc (degrees) Length of hospital stay ‐ days Other outcomes None
Notes	Supported by a grant from the National Institutes of Health Standard deviations were imputed in Analysis 1.16; Analysis 1.31 No participant withdrawals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not stated whether study was randomised
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	Unclear risk	For "all phases, patients, nursing staff, and physical therapists involved were blinded as to parameters under study". However, what blinding participants received is not described.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	Unclear risk	No self‐reported outcomes
Blinding of outcome assessment (detection bias) objective outcomes	Unclear risk	Knee range of motion, analgesic use and length of hospital stay may have been influenced by participants (blinding unclear).
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

Wang 2017.

Study characteristics
Methods	Randomised controlled trial, parallel groups Conducted February 2013 to February 2015
Participants	106 participants had 106 TKRs from Zhongshan Hospital, Hubei Province, China. Type of prosthesis not stated. Sample size Cryo: 53 Control: 53 Sex: M/F Cryo: 20/33 Control: 22/31 Age Cryo: 65.23 +/‐ 5.41 Control: 64.97 +/‐ 5.36 Age range Not discussed Years with disease Cryo: 21.02 +/‐ 2.34 Control: 20.98 +/‐2.26 Knee operated: L/R Cryo: 30/23 Control: 28/25 60 participants with osteoarthritis, 37 with rheumatoid arthritis, 9 with traumatic arthritis Inclusion criteria Aged 50 to 80, confirmed diagnosis of knee osteoarthritis, consented, unilateral TKR Exclusion criteria Severe underlying systemic disease, diabetes mellitus, ankylosing spondylitis, infection of joints, persistent analgesics within 14 days prior to surgery, simultaneous bilateral TKR, immune diseases, disease affecting ipsilateral lower extremity strength or lower limb pain, did not consent to or unable to complete treatment
Interventions	Both groups received routine anti‐inflammatories and anticoagulants post‐surgery and active and passive rehabilitation exercises were performed. Control 48 hrs post‐surgery or 1 day after drainage tubes removed, the CPM machine was used for exercise for 30 to 60 minutes, 2 to 3 times a day for 2 weeks. Cryo: local compression cryotherapy Same as control, combined with EVERYCRYO automatic pressurised cold water circulation system, with temperature set to 11 to 13 degrees Celsius. Cryotherapy was continuously applied when participants were first returned to the wards post‐surgery and stopped after 48 hrs.
Outcomes	Main outcomes Measured 24 hr (at rest) and 48 hr (while moving?) post‐surgery: Pain: assessed with visual analogue scale 0 to 10 Swelling: pre‐op/post‐op difference in average circumference of thigh 2 cm above patella, midpoint of patella and the thickest part of the gastrocnemius Measured pre‐operatively and 2‐weeks post‐operatively and followed up 6 months later: Knee function: Hospital for Special Surgery (HSS) knee scoring system. Score out of 100, higher score is better function. 6 months later, the HSS score is graded into: > 85 is excellent; 70 to 84 is good; 60 to 69 is reasonable; < 59 is poor. The proportion of participants with excellent or good responses were noted. Adverse effects
Notes	Article machine‐translated from Chinese and confirmed with a native speaker. Authors declare no conflicts of interest. No participant withdrawals.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Comment: participants were randomised based on admission time and split into 2 groups based on odd or even numbers allocated
Allocation concealment (selection bias)	Unclear risk	Comment: not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Comment: use of the cryotherapy machine in only the intervention group would be apparent to participants and investigators.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Comment: pain and function are subjective outcomes which may have been influenced by unblinded participants.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Comment: unclear if person assessing swelling was aware of group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Comment: insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Comment: bilateral TKRs were only included if the surgeries were done one knee at a time, with at least 6 months between surgeries for each knee

Webb 1998.

Study characteristics
Methods	Parallel‐group prospective randomised trial. No time period stated.
Participants	49 TKRs in 40 participants (9 bilateral TKRs) in Oxford, UK. Sex not stated. Press Fit Condylar prosthesis (posterior cruciate ligament (PCL) retaining and cemented) used. Number and location of study centres not explicitly stated. Sample size Group 1: 15 (24 TKRs in 15 participants) Group 2: 16 (25 TKRs in 16 participants) Age Group 1: 69.0 Group 2: 70.9 Age range Not discussed Sex Not discussed Weight Group 1: 72.6 Group 2: 74.8 Indication (Osteoarthritis: rheumatoid arthritis) Group 1: (10:5) Group 2: (13:3) Inclusion criteria Patients undergoing TKR Exclusion criteria Coagulopathy, intraoperative fracture, premature removal of the suction drainage, non‐consenting patients.
Interventions	Group 1: CryoCuff Cold compression (CryoCuff, Aircast, UK) was applied prior to the release of the thigh tourniquet. Cold compression was achieved with the Aircast CryoCuff, which consists of an inflatable bladder that is strapped to the front of the knee. Chilled water can be infused repeatedly by connection to an insulated container that is raised 40 cm above the knee, thereby providing 30 mmHg or compression. Time period of application not stated. Use of CPM not stated. Group 2: control Compression dressing (wool and crepe) applied immediately after surgery. Time period of application not stated.
Outcomes	Outcomes considered in this review Blood loss ‐ suction drainage in 48 hrs (mL) Pain ‐ visual analogue scale total pain score (0 to 100) taken on POD 1 Transfusion rate ‐ number in each group requiring transfusion Analgesia use ‐ total opiate use over first 48 hrs (mg/kg) Knee range of motion ‐ degrees of flexion (pre‐operative; POD 5; 6 weeks; 3 months) taken by senior physiotherapist blinded to allocation Knee swelling ‐ 2 cm proximal to patella (cm): pre‐operative; POD 5; 6 weeks; 3 months Other outcomes None
Notes	Participants in the bilateral TKR group (n = 9) were not randomised. The intervention was always applied to the second knee. Outcomes were reported on individual knees and no adjustment was made for lack of independence. Standard deviations were imputed in Analysis 1.14; Analysis 1.28; Analysis 1.29. Supported by a grant from the Wishbone trust (British Orthopaedic Association). No participant withdrawals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information
Allocation concealment (selection bias)	Unclear risk	Sealed envelopes were used, but it was not described whether they were sequentially‐numbered, opaque, tamper‐proof, or by whom they were handled.
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants were not blinded ‐ as intervention group received CryoCuffs and the control group received a wool and crepe bandage.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Self‐reported outcome (e.g. pain) may have been influenced by unblinded participants
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Analgesia and knee range of motion may have been affected by unblinded participants
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Incomplete data for pain, knee range of motion and swelling outcomes. Transfusion data included but not mentioned in methods.
Other bias ‐ bilateral knee replacements included	High risk	Data from 9 bilateral TKRs were included. No adjustment for lack of independence.

Wittig‐Wells 2015.

Study characteristics
Methods	Pre‐test, post‐test randomised trial with cross‐over Conducted September to October 2013
Participants	29 participants with 29 TKRs recruited from one American hospital. Type of prosthesis not stated. Sample size Group 1: 15 Group 2: 14 Age Age: 64 +/‐ 9.3 Age range Not discussed Sex Female: 62% Inclusion criteria 30 to 80 years old, elective single TKR, no pre‐operative use of narcotics other than for knee pain, physiologically stable with no post‐operative complications
Interventions	Group 1: analgesic only first Group 2: analgesic plus cryotherapy first Analgesics were given to both groups as per physician orders. Cryotherapy was then applied for 30 minutes in Group 2. A sealed zip‐lock bag half‐filled with crushed ice was wrapped in cloth and secured on the operative knee with tape. Ice was refilled if necessary. At least 3 hrs after the completion of the first treatment episode, the interventions of each group were swapped.
Outcomes	Pain: measured on a 0 to 10 verbal numeric report scale. Measured before analgesia, then 30 minutes and 60 minutes after analgesia. Satisfaction with pain management: measured on the same scale (higher score reflects higher satisfaction). Measured 60 minutes after analgesia.
Notes	Authors declare no conflicts of interest. Participant withdrawals not discussed.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "random assignment to treatment order was done using a computer randomization program"
Allocation concealment (selection bias)	Unclear risk	Quote: "investigators were blinded to treatment order until the first dose of analgesic" Comment: unclear how allocation was concealed
Blinding of participants and personnel (performance bias) all outcomes	High risk	Comment: use of ice pack in the assigned group would be apparent to participants and assessors.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain and satisfaction scores may be influenced by unblinded participants
Blinding of outcome assessment (detection bias) objective outcomes	Unclear risk	No objective outcomes reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: participant withdrawals not discussed in the article. Sample size used in data analysis not stated.
Selective reporting (reporting bias)	Unclear risk	Comment: insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Comment: single TKRs only

Xu 2015.

Study characteristics
Methods	Retrospective study? Parallel groups Conducted from March 2008 to March 2014
Participants	240 participants received 240 TKRs at Chengde Medical College (China). PFC Sigma RPF (Johnson & Johnson) prosthesis used. Sample size Control: 120 Cryo: 120 Sex: M/F Control: 78/42 Cryo: 80/40 Age: mean +/‐ SD Control: 67.34 +/‐ 3.22 Cryo: 67.30 +/‐ 5.49 Age range Not discussed BMI Control: 24.9 +/‐ 4.1 Cryo: 24.78 +/‐ 5.09 Affected side: L/R Control: 65/55 Cryo: 68/52 Inclusion criteria Diagnosed with osteoarthritis, unilateral primary TKR, no hepatic, renal or coagulation abnormalities pre‐operatively, surgery conducted by chief surgeon, normal cardiopulmonary function, pre‐operative haemoglobin > 100 g/L, normal platelets, no history of anticoagulants taken within 6 weeks pre‐operatively, no non‐steroidal anti‐inflammatory drugs (NSAIDs) taken 1 week pre‐operatively Exclusion criteria Allergic (allergen not stated), excess acute blood loss > 1 L, unreliable vital signs after replacement, concurrent systemic disease
Interventions	Post‐surgery, both groups received bandage compression, intravenous and oral analgesics for 2 days and routine antibiotics for 2 to 3 days. Control: received nothing else Cryotherapy: given ice pack for 2 hrs post‐surgery, then for 3 to 5 hrs daily at around noon for 7 days.
Outcomes	Main outcomes Blood loss: difference in haemoglobin 48 hrs before and 48 hrs after the operation, perioperative indicators (blood loss, transfusion rate and transfusion volume) Pain: visual analogue scale measured on post‐operative days 2, 3, 7 Knee function: overall function judged after 3 months. Total score > 85 is excellent; 70 to 84 is good; 60 to 69 is reasonable; < 60 is poor. Adverse effects
Notes	Article machine‐translated from Chinese and confirmed with a native speaker. Article claims to be retrospective, but the methods describe a prospective RCT Scoring tool for knee function assessment not stated ‐ based on similar stratification of knee function in Wang 2017; the scoring system is assumed to be the Hospital for Special Surgery (HSS) knee scoring system. Authors declared no conflicts of interest. No participant withdrawals
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: participants were randomised into 2 groups but the method of randomisation was not explained
Allocation concealment (selection bias)	Unclear risk	Comment: not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Use of ice packs in intervention groups would be apparent to participants and assessors
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain and knee function are subjective outcomes and could be influenced by unblinded participants.
Blinding of outcome assessment (detection bias) objective outcomes	Low risk	Measurement of blood loss unlikely to be affected by knowledge of intervention.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no participant withdrawals
Selective reporting (reporting bias)	Unclear risk	Comment: insufficient information
Other bias ‐ bilateral knee replacements included	Low risk	Comment: inclusion criteria of unilateral TKR only

Yuksel 2022.

Study characteristics
Methods	Parallel‐group prospective randomised trial. Time period not specified.
Participants	100 TKRs in 100 participants in an unspecified university hospital in Turkey. Type of prosthesis and use of cement not stated. Sample size Group 1: 34 Group 2: 33 Group 3: 33 Age (SD) Group 1: 65.4 (8.9) Group 2: 65.4 (6.7) Group 3: 66.8 (9.8) Age range Not specified BMI (SD) Group 1: 30.5 (5.3) Group 2: 31.1 (3.3) Group 3: 29.6 (4.2) Sex (M:F) Group 1: (11:23) Group 2: (8:25) Group 3: (10:23) Inclusion criteria Primary TKR for knee OA over the age of 18 Exclusion criteria Unable to understand verbal and written instructions, previous orthopaedic or neurological disorder that causes gait disturbance
Interventions	Group 1: control group No kinesio taping or cold therapy. Standard post‐op regimen (also offered to intervention groups) Group 2: kinesio taping group Two fan‐shaped kinesio tapings: medially from popliteal fossa to medial malleolus and laterally from anterior superior iliac spine (ASIS) to lateral malleolus. Tapes applied on POD 1 and removed after 72 hours. New tapes applied following removal, with participants receiving tapes throughout inpatient rehabilitation program. Group 3: cryotherapy group Two cold packs, covered with a wet towel, placed on patella from POD 0. Cold packs applied for 12 to 15 minutes every 2 hours throughout inpatient rehabilitation program. All participants received the same standard post‐operative rehabilitation: weight‐bearing as tolerated from the first post‐operative day and range of motion exercises (passive, active‐assistive, active), strengthening exercises and gait training twice/day.
Outcomes	Outcomes included in this review Pain scores ‐ numeric pain rating scale 0 to 10 (at discharge, 12‐WPO) Knee swelling ‐ circumference measured at mid‐patella, 10 cm above patella, 15 cm above patella, 10 cm below patella, 15 cm below patella and oedema measured using formula for frustum (at discharge, 12‐WPO) Knee range of motion ‐ degrees of flexion (at discharge, 12‐WPO) Knee function ‐ Hospital for Special Surgery knee score (HSS, 0 to 100), Iowa Level of Assistance Scale (ILAS, 0 to 36) at discharge, 12‐WPO Activity level ‐ Timed Up and Go (TUG, seconds), 10‐Meter Walk Test (10‐MWT, seconds) at discharge, 12‐WPO Quality of life ‐ Short Form‐12 Health Survey (SF‐12) at discharge, 12‐WPO Other outcomes Maximum knee flexor and extensor strength
Notes	Authors declared no conflicts of interest 11 participant withdrawals (4 from control group, 4 from kinesio taping group, 3 from cryotherapy group)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Online randomisation method used to allocate participants: allocated into three groups according to sequentially‐numbered index cards.
Allocation concealment (selection bias)	Unclear risk	Not mentioned
Blinding of participants and personnel (performance bias) all outcomes	High risk	Participants either had nothing, kinesio taping or cold packs. They would have been aware of their intervention.
Blinding of outcome assessment (detection bias) self‐reported outcomes (e.g. pain, function)	High risk	Pain score, knee function and quality of life scores potentially influenced by unblinded participants.
Blinding of outcome assessment (detection bias) objective outcomes	High risk	Participants not blinded, which may influence knee range of motion. Physiotherapist conducting measurements was not blinded, which could influence measurement of knee swelling, knee function.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Small numbers withdrew, balanced between allocation groups.
Selective reporting (reporting bias)	High risk	Methods stated that SF‐12 quality of life score at discharge and 12 weeks was to be recorded, but only 12 weeks was recorded. Oedema, 10 MWT, TUG, SF‐12 scores included but not mentioned in protocol. Iowa Ambulation Velocity Score noted in protocol but not included.
Other bias ‐ bilateral knee replacements included	Low risk	Only participants who had unilateral knee replacements were included.

°C: degree Celsius; °F: degree Fahrenheit; ASA: American Society of Anesthesiologists; BMI: body mass index; cc: cubic centimetre; CPM: continuous passive motion; Hb: haemoglobin; hr(s): hour(s); KOOS: Knee injury and Osteoarthritis Outcome Score; OA: osteoarthritis; POD: post‐operative day; SD: standard deviation; TKA: total knee arthroplasty; TKR: total knee replacement; WOMAC: Western Ontario and McMaster Universities Osteoarthritis Index; WPO: weeks post‐operative

Characteristics of excluded studies [ordered by year]

Study	Reason for exclusion
Tum Suden 1976	Not a clinical trial
Hecht 1983	Cryotherapy was not given in the acute phase
Ryd 1990	Cryotherapy was not used
Schroder 1994	Did not include total knee replacement surgery
Leutz 1995	Retrospective study
Whitelaw 1995	Total knee replacement surgery was not included
Lessard 1997	Did not study total knee replacement surgery
Flórez García 2001	Not a randomised clinical trial or controlled clinical trial
Martin 2002	Did not study total knee replacement surgery
Hubbard 2004	Systematic review
Husted 2004	Systematic review
Laskin 2004	Not a randomised clinical trial or controlled clinical trial
Holstrom 2005	Participants underwent partial or unicompartmental knee replacement, rather than total knee replacement
Sanchez‐Inchausti 2005	Did not include total knee replacement surgery
Wilk 2005	Not a randomised clinical trial or controlled clinical trial
Buckup 2006	Did not include total knee replacement surgery in study
Dean 2007	Case report
Rostlund 2007	Guest editorial
Zhang 2007	Not a randomised clinical trial or controlled clinical trial
Fischer 2008	Systematic review
Heisel 2008	Not a randomised clinical trial or controlled clinical trial
Simanski 2008	Systematic review
Khan 2009	Systematic review
Muller 2009	Systematic review
Centre for Reviews 2010a	Systematic review
Centre for Reviews 2010b	Systematic review
Demoulin 2012	Did not have a non‐cryotherapy arm
Holm 2012	Cryotherapy was not given during the acute phase
Su 2012	Did not have a non‐cryotherapy arm
Thienpont 2014	Did not have a non‐cryotherapy arm
Bech 2015	Did not have a non‐cryotherapy arm
Desteli 2015	Did not have a non‐cryotherapy arm
Kuyucu 2015	Data from participants with primary total knee replacement could not be extracted
Mumith 2015	Did not have a non‐cryotherapy arm
Pan 2015	Did not have a non‐cryotherapy arm
Kayamori 2016	Did not have a non‐cryotherapy arm
Li 2016	Cryotherapy was applied internally and not to the surface of the knee
Schinsky 2016	Did not have a non‐cryotherapy arm
Murgier 2017	Not a randomised control trial or controlled clinical trial
Ruffilli 2017	Did not have a non‐cryotherapy arm
Zhao 2017	Did not have a non‐cryotherapy arm
Rui 2017	Cryotherapy definition not met: ice packs, cooling pads or other commercial device not used
Sadoghi 2018	Did not have a non‐cryotherapy arm
Ueyama 2018	Cryotherapy was not used
Gan 2018	Did not have a non‐cryotherapy arm
Foster 2019	Not a randomised clinical trial or controlled clinical trial
Karaduman 2019	Did not have a non‐cryotherapy arm
Kim 2019	Cryotherapy was not used
Nishigami 2019	Knee icing offered to both groups in the acute phase
Xuan Neo 2019	Did not have a non‐cryotherapy arm
Borgers 2020	Did not have a non‐cryotherapy arm
Chen 2020	Did not have a non‐cryotherapy arm
Shim 2020	Did not have a non‐cryotherapy arm
Van Ooij 2020	Not a randomised clinical trial or controlled clinical trial
Zhong 2021	Did not have a non‐cryotherapy arm

Characteristics of studies awaiting classification [ordered by study ID]

Siva Subramanian 1996.

Methods	No details.
Participants	No details.
Interventions	No details.
Outcomes	No details.
Notes	We could find no details, apart from the reference information, for this study. The study was presented at a meeting in 1996. Neither we nor the Director of the St George Hospital Library could locate the abstract. Given the length of time elapsed since the meeting, it is very unlikely further details will emerge.

Characteristics of ongoing studies [ordered by study ID]

ISRCTN12615549.

Study name	A randomized controlled trial on effects of Cryo/Cuff in early rehabilitation of total knee arthroplasty Scientific title: A randomized controlled trial to evaluate whether Cryo/Cuff compared to placebo can better management of pain, swelling, blood loss, range of motion in postoperative total knee arthroplasty patients
Methods	Parallel‐group randomised control trial. Anticipated recruitment between July 2016 and December 2016.
Participants	Recruitment from a single centre, Peking Union Medical College Hospital (China). Target number of participants is 98. Inclusion criteria Aged 18 to 80 years Degenerative osteoarthritis need for total knee arthroplasty Initial unilateral total knee arthroplasty patients Agree to participate in the study Exclusion criteria Abnormal blood coagulation Cannot tolerate Cryo/Cuff Diagnosed as rheumatoid arthritis, traumatic osteoarthritis, ankylosing spondylitis, haemophilic arthritis, peripheral vascular disease Cold urticaria Preoperative anticoagulation: patients had preoperative deep vein thrombosis (DVT) Preoperative history of anaemia
Interventions	Control group Participants receive care as usual, including rehabilitation protocol, analgesia and range of motion exercises. It does not involve application of a cooling device. Intervention group Participants receive 3 hours of Cryo/Cuff immediately after surgery and one hour twice daily for consecutive three days.
Outcomes	Primary outcome Pain measured using the visual analogue scale (VAS) at rest and during exercise at baseline (1 day before surgery) and 1 to 5 days after surgery. Secondary outcomes Opioid consumption measured by anaesthesiologist at 72 hours removal of the sufentanil patient‐controlled analgesia (PCA) pump Swelling measured using knee girth in millimetres at mid‐patella with the knee in maximal extension at baseline (1 day before surgery) and 2 to 5 days after surgery Wound drainage assessed by the independent member of the research team at 48 hours Haemoglobin measured using blood test at admission, 1, 3 and 5 days after surgery Blood transfusion rate measured by the independent member of the research team at discharge Range of motion measured according to standard reference points (the greater trochanter, the lateral condyle of femur, and the later malleolus) with a goniometer in intervals of 1° and recorded as maximum active extension and flexion at baseline (one day before surgery), 1 and 2 weeks after surgery Length of post‐operative hospital stay assessed by the independent member of the research team at participants’ discharge
Starting date	21 June 2016
Contact information	Bingdu Tong, Department of Orthopedics, Peking Union Medical College Hospital (China)
Notes	Funding: Peking Union Medical College Hospital Clinical trial registration: ISRCTN12615549 Trial status: no longer recruiting Completion date: 23 January 2017. No results available.

NCT05395273.

Study name	Cryocompression After Total Knee Arthroplasty Scientific title: A Randomized Controlled Trial to Evaluate the Effectiveness of Cryocompression After Total Knee Arthroplasty (TKA)
Methods	Parallel‐group randomised control trial. Anticipated recruitment to commence June 2022.
Participants	Number and location of recruitment sites not stated. Target number of participants is 28. Inclusion criteria Signed informed consent Diagnosed arthritis of the knee Planned implantation of a primary total knee arthroplasty by the Regionalspital Surselva AG Exclusion criteria Decompensated hypertension in the affected area Acute inflammatory phlebitis in the affected area Patients with acute paroxysmal cold haemoglobinuria or cryoglobulinaemia Significant vascular disruption in the affected area History of pulmonary embolism or risk factors for deep vein thrombosis or pulmonary embolism in the affected area No increased venous or lymphatic return in the affected leg is desired due to e.g. carcinoma Raynaud's disease Hypersensitivity to cold Fear of cold/compression
Interventions	Control group The standard therapy includes the common physical therapy interventions, which are normally used in the hospital. Intervention group Participants receive daily cryo‐compression in addition to the standard therapy.
Outcomes	Change from baseline in skin temperature 4 days and 6 weeks after surgery Change from baseline in KOOS (Knee and Osteoarthritis Outcome Score) 4 days and 6 weeks after surgery Change from baseline in range of motion 4 days and 6 weeks after surgery Change from baseline in Timed Up and Go (TUG) test 4 days and 6 weeks after surgery Change from baseline in visual analogue scale (VAS) 4 days and 6 weeks after surgery Change from baseline in patient satisfaction at each day of hospitalisation Length of stay (assessed 6 weeks after surgery) Change from baseline in 10‐metre walk test 4 days and 6 weeks after surgery Consumption of morphine (assessed 6 weeks after surgery) Change from baseline in knee swelling 4 days and 6 weeks after surgery
Starting date	June 2022
Contact information	Ron Clijsen, University of Applied Sciences and Arts of Southern Switzerland
Notes	Funding: University of Applied Sciences and Arts of Southern Switzerland Clinical trial registration: NCT05395273 Trial status: not yet recruiting

NCT05572359.

Study name	Cryo and Compression Therapy After TKA and UKA Scientific title: Effect of Cryo‐ and Compression Therapy After Total Knee and Unicompartmental Arthroplasty, A Randomised Controlled Trial
Methods	Parallel‐group randomised controlled trial. Anticipated recruitment to commence June 2023.
Participants	Recruitment from Martini Hospital (the Netherlands). Target number of participants is 208. Inclusion criteria Patients scheduled for a primary total knee arthroplasty (TKA) or unicompartmental knee arthroplasty (UKA) in the Martini Hospital Age ≥ 18 years Exclusion criteria Pre‐operative switch from UKA to TKA (only applicable for the UKA patients) Revision TKA implant (only applicable for the TKA patients) Rheumatoid arthritis Other co‐morbidities on which cooling may have a negative effect on (based on judgement of the orthopaedic surgeon) Inability to read and understand the Dutch language Because the cool pack needs to be cooled in a freezer, the participant or the nursing home must have a freezer that can be used
Interventions	Control group Regular care during the six post‐operative weeks Intervention group Participants in the intervention groups are instructed to use the cryo‐ and compression brace during the six post‐operative weeks 5 times a day, for a maximum of 20 minutes. The amount of compression is dependent on participants' own preference.
Outcomes	Numeric Rating Scale at rest and during loading: measured at baseline, daily during first 6 weeks after surgery, 6 months after surgery and 12 months after surgery Opioid use: measured daily during first 6 weeks after surgery KOOS (Knee Osteoarthritis Outcome Score): measured at baseline and 6 weeks after surgery WORQ (Work Rehabilitation Questionnaire): measured at baseline and 6 weeks after surgery Active range of motion: measured at baseline and 6 weeks after surgery Knee circumference: measured at baseline and 6 weeks after surgery Timed Up and Go (TUG): measured at baseline and 6 weeks after surgery KOOS‐PS (Knee Osteoarthritis Outcome Score ‐ Physical Function Short‐form): measured at baseline, 6 months, 12 months after surgery Oxford Knee Score: measured at baseline, 6 months, 12 months after surgery EQ‐5D‐5L (EuroQol 5D‐5L): measured at baseline, 6 months, 12 months after surgery Frequency of use of cryo‐ and compression brace: measured daily during first 6 weeks after surgery Numeric Rating Scale satisfaction: measured at 6 weeks, 6 months, 12 months after surgery Patient perceived change in pain (7‐point Likert scale): measured at 6 weeks, 6 months, 12 months after surgery Patient perceived change in function (7‐point Likert scale): measured at 6 weeks, 6 months, 12 months after surgery
Starting date	1 June 2023
Contact information	Reinoud W. Brouwer, Martini Hospital
Notes	Funding: Martini Hospital, Groningen Clinical trial registration: NCT05572359 Trial status: not yet recruiting

Differences between protocol and review

Transfusion rate, knee range of motion and knee function moved from minor outcomes to major outcomes.

Participant‐reported global assessment of success added as minor outcome.

Contributions of authors

Ashwin Aggarwal was involved in collecting, entering and interpreting the updated data, and writing the updated manuscript.

Sam Adie was involved in designing and coordinating the study, as well as collecting, entering and interpreting the data and writing the manuscript. Sam Adie is the guarantor for this review.

Justine Naylor was involved in conceiving and designing the study, interpreting the data and prior work contributing to this review.

Ian Harris was involved in conceiving and designing the study and interpreting the data.

Sources of support

Internal sources

• Sydney South West Area Health Service (Liverpool Hospital), Australia

  Infrastructure, Salary

• South West Sydney Clinical School, University of New South Wales, Australia

  Infrastructure, Salary

External sources

• No sources of support provided

Declarations of interest

The authors declare they have no interests, financial or otherwise, that may impact on the findings of this review.

Edited (no change to conclusions)