Arthroscopic debridement for knee osteoarthritis

Abstract

Background

Knee osteoarthritis (OA) is a progressive disease that initially affects the articular cartilage. Observational studies have shown benefits for arthroscopic debridement (AD) on the osteoarthritic knee, but other recent studies have yielded conflicting results that suggest AD may not be effective.

Objectives

To identify the effectiveness of AD in knee OA on pain and function.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library Issue 2, 2006); MEDLINE (1966 to August, 2006); CINAHL (1982 to 2006); EMBASE (1988 to 2006) and Web of Science (1900 to 2006) and screened the bibliographies, reference lists and cited web sites of papers.

Selection criteria

We included randomised controlled trials (RCT) or controlled clinical trials (CCT) assessing effectiveness of AD compared to another surgical procedure, including sham or placebo surgery and other non‐surgical interventions, in patients with a diagnosis of primary or secondary OA of the knees, who did not have other joint involvement or conditions requiring long term use of non‐steroidal anti‐inflammatory drugs (NSAIDs). The main outcomes were pain relief and improved function of the knee.

Data collection and analysis

Two review authors independently selected trials for inclusion, assessed trial quality and extracted the data. Results are presented using weighted mean difference (WMD) for continuous data and relative risk (RR) for dichotomous data, and the number needed to treat to benefit (NNTB) or harm (NNTH).

Main results

Three RCTs were included with a total of 271 patients. They had different comparison groups and a moderate risk of bias. One study compared AD with lavage and with sham surgery. Compared to lavage the study found no significant difference. Compared to sham surgery placebo, the study found worse outcomes for AD at two weeks (WMD for pain 8.7, 95% CI 1.7 to 15.8, and function 7.7, 95% CI 1.1 to 14.3; NNTH=5) and no significant difference at two years. The second trial, at higher risk of bias, compared AD and arthroscopic washout, and found that AD significantly reduced knee pain compared to washout at five years (RR 5.5, 95% CI 1.7 to 15.5; NNTB=3). The third trial, also at higher risk of bias, compared AD to closed‐needle lavage, and found no significant difference.

Authors' conclusions

There is 'gold' level evidence that AD has no benefit for undiscriminated OA (mechanical or inflammatory causes).

Plain language summary

Arthroscopic debridement for osteoarthritis of the knee

This summary of a Cochrane review presents what we know from research about the effect of arthroscopic debridement (AD) for osteoarthritis (OA) of the knee.

The review shows that in people with OA, arthroscopic debridement:

‐ Probably does not improve pain or ability to function compared to placebo (sham surgery) 
 ‐ Probably leads to little or no difference in pain or ability to function compared to lavage 
 ‐ May improve pain compared to washout 
 ‐ May not lead to any difference in pain or ability to function compared to closed needle joint lavage

We often do not have precise information about side effects and complications. This is particularly true for rare but serious side effects. Possible side effects may include a small risk of infection and of venous thromboembolism.

What is osteoarthritis and what is arthroscopic debridement? 
 Osteoarthritis (OA) is the most common form of arthritis that can affect the hands, hips, shoulders and knees. In OA, the cartilage that protects the ends of the bones breaks down and causes pain and swelling. OA can occur in different areas of the knee or the whole knee. When the cartilage breaks down, bits of tissue are left around the joint which can add to the inflammation and prevent the joint from working properly.

Arthroscopic debridement (AD) involves using instruments to remove damaged cartilage or bone. Often the doctor will start the procedure by using a tool to spray jets of fluid to wash and suck out all debris around the joint. This is called lavage or washout. Then, the parts of the joint bone that are loose or misshapen are removed.

Best estimate of what happens to people with OA who have arthroscopic debridement compared with washout:

Pain: 66 more people out of 100 reported being pain free after 1 year and 48 more people out of 100 reported being pain free after 2 years. These results are based on low quality evidence.

Best estimate of what happens to people with OA who have arthroscopic debridement compared with placebo:

Pain two weeks after treatment: Pain scores increased by 9 more points on a scale of 0‐100.

Physical function two weeks after treatment: The ability to function improved 8 more points on a scale of 0‐100 for the placebo group. These results are based on moderate quality evidence.

Physical function 12 months after treatment: The ability to function improved 7 more points on a 0‐100 scale for the placebo group, indicating that the AD group experienced significantly more limited function. These results are based on low quality evidence.

The numbers given are our best estimate. When possible, we have also presented a range because there is a 95 percent chance that the true effect of the treatment lies somewhere between that range.

Background

Knee osteoarthritis (OA) is a progressive disease that originally affects the articular cartilage. Certain mechanical and biological events may destabilise the normal degradation and repair processes of chondrocytes and extracellular matrix, causing deterioration of the articular cartilage (Dabov 2003). The cartilage breaks down resulting in fibrillation, fissures, ulceration and then full thickness loss of the joint surface. Ultimately, the subchondral bone and almost the entire joint become damaged with disabling deformities.

Knee OA has a worldwide distribution, though there is variation in the prevalence among different ethnic or cultural groups and genders (Zhang 2003). The elderly population has a higher risk of developing this condition. OA can be classified according to its aetiological factors as primary and secondary (Altman 2004). Although the end stage of both types may be the same, the progression of primary OA is usually slower and less relentless (Dabov 2003).

Patients with knee OA may seek medical treatment at different stages of the disease. Common complaints are pain exacerbated by knee motion or weight bearing, stiffness, swelling and deformity (genu varum, genu valgum or flexion contracture), and decreased walking distance. The objectives of management are to relieve pain, maintain or improve mobility, and minimise disability. Initial management of most patients is usually nonoperative (Pendleton 2000). This may combine analgesics with physical therapy, bracing, orthoses, ambulatory aids, nonsteroidal anti‐inflammatory drugs (NSAIDs) and other novel medications (Lequesne 1994; Pelletier 2001), intra‐articular injections of corticosteroids or chondroprotective agents. Changes in daily work and recreational activities may also be necessary. Obesity is a known risk factor for knee OA and weight loss has been shown to slow the progression of the disease (Messier 2004).

Because of the progressive nature of the condition, many patients with knee OA are eventually offered operative treatment. A variety of procedures have been described, ranging from arthroscopic lavage or debridement to corrective osteotomy or total knee arthroplasty. The choice of procedure depends on the severity of the disease and the patient's individual condition. For more localised articular lesion, current practices include microfracture, osteo‐articular transplantation (Makino 2001; Nakaji 2006) and autologous chondrocyte implantation (Bentley 2003; Kish 2004; Knutsen 2004). Debridement for knee OA using an arthroscopic technique produces less postoperative pain and shorter rehabilitation time than the older open procedure (Dabov 2003). Arthroscopic debridement (AD) consists of tidal irrigation to wash out all debris. Unstable chondral flaps, redundant synovia, degenerated menisci and ligaments, loose bodies and osteophytes are shaved away or burred down by using mechanical instruments. AD can by no means stop the degenerative process inherent in the disease, and the full thickness chondral defect is not healed. AD is expected to remove chemical and mechanical components that contribute to the symptoms of OA (Smith 1997; Cameron 2004). Although pain and functions might be improved for a certain postoperative period, it is expected that the symptoms will return over time. Other techniques such as electrocautery, lasers or radiofrequency are also available for debridement.

Numerous retrospective and some prospective studies have suggested benefits of AD on the osteoarthritic knee at different stages of disease severity, although these studies are not as reliable as randomised controlled trials. This literature reports success rates of about 40% to 75% and favourable outcomes in reducing pain and improving function of the knee (Sprague 1981; Baumgaertner 1990; Timoney 1990; McLaren 1991; Hubbard 1996; Harwin 1999; McGinley 1999; Shannon 2001). These palliative effects were maintained for mostly two to five years (Baumgaertner 1990; Timoney 1990; McLaren 1991; Hubbard 1996; Shannon 2001), though possibly as long as 7 to 13 years (Harwin 1999; McGinley 1999). Most reports propose AD as a temporary treatment for knee OA, which is more effective in the early stages of the disease if malalignment of the joint has not developed. Moreover, patients with mechanical symptoms and symptoms of short duration tended to do well with this intervention.

However, results from recent studies suggest that AD may not be effective. Some studies have reported AD as having no clinically meaningful difference from placebo surgery (Moseley 1996; Moseley 2002). In comparison to arthroscopic lavage, some improvement in quadriceps isokinetic torque at 6 and 12 weeks was observed after joint lavage but not after AD (Gibson 1992). A systematic review was therefore needed to evaluate the effectiveness of this procedure.

Objectives

The main objective of this review was to estimate the effectiveness of AD on knee OA on pain reduction (reduced use of relevant medications) and/or functional improvement.

The secondary objectives were to observe: 
 1. The type or stage of severity of the OA in which AD is most effective. 
 2. The expected length of effectiveness until the patients need further intervention.

Methods

Criteria for considering studies for this review

Types of studies

Any randomised controlled trials (RCT) or controlled clinical trials (CCT, trials using quasi‐ or pseudo‐random process) which assess the effectiveness and persistence of the effects of the AD in osteoarthritic knees in reducing symptoms and/or reducing needs of non‐steroidal anti‐inflammatory drugs (NSAIDs) and/or analgesics and/or improving knee functions were considered.

Types of participants

Patients with diagnosis of primary or secondary OA of the knees, who did not have other joint involvement or conditions requiring long term use of NSAIDs, were included. Primary OA was any OA where a specific cause for the condition was not found. Secondary OA was where a definite cause could be found, such as trauma, joint instability, a metabolic disorder or other rheumatic disorder affecting the joint. The diagnosis should be established by pertinent history taking, physical examination and appropriate imaging.

Trials studying the following conditions were excluded from the review: 
 1. Other conditions in which prolonged use of NSAIDs was required. 
 2. Bed‐ridden or wheel‐chair‐ridden conditions from any cause. 
 3. Combined surgery, such as combining AD with corrective osteotomy or AD with simultaneous operation on the other limb or the other joint(s) of the same limb. 
 4. Post‐operative knee immobilisation (comparable with casting) for more than 2 weeks, which might compromise joint motion.

Types of interventions

The intervention for treated cases was arthroscopic, NOT open, debridement performed on the osteoarthritic knee. It was assumed that the procedure may also have included shaving, lavage, drilling, microfracture technique or abrasion arthroplasty, unless the study specifically stated that they were not used.

We recorded the method or modalities other than the common mechanical instruments (e.g., electrocautery, lasers, radio‐frequency or coblation) used in the AD process if these were described in the studies.

The control could be any non‐surgical intervention or comparative operation such as chondrocytes implantation, corrective osteotomy and replacement arthroplasty, including sham or placebo surgery.

Types of outcome measures

Primary outcomes included: 
 1. Reduction of knee pain. 
 2. Improvement of knee functions.

Knee pain and functions could be assessed and recorded as continuous data in scores. The scores could be measured directly (e.g. using visual analogue scales for pain), or using validated functional rating systems (e.g. Lysholm Knee Scores, International Knee Documentation Committee (IKDC) Scores, Hospital for Special Surgery (HSS) Scores or Western Ontario and McMaster Universities (WOMAC) OA Index). These outcomes could be affected by patients' use of NSAIDs and/or analgesics, which will also be measured. 
 
 Secondary outcomes included: 
 1. Time to next major intervention (e.g., TKA) indicating failure of the treatment or censoring due to end of the study or dropout. 
 2. Amount (doses, frequencies and types) of NSAIDs and/or analgesics used as rescue therapies in parallel with the treatment and control. 
 3. Post‐operative morbidities or complications. 
 4. Other outcomes according to the authors' reports.

We recorded subsequent interventions. We counted use of topical drugs, injection of corticosteroids or any chondroprotective agents and taking of novel medications such as glucosamine sulfate and diacerein as co‐interventions. When available, we also considered data on confounding factors (litigation, body weight, co‐morbidity, etc) in the analysis.

Search methods for identification of studies

Studies were identified from the following electronic databases: The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library Issue 2, 2006); MEDLINE (1966 to August, 2006); CINAHL (1982 ‐ 2006); EMBASE (1988 to 2006) and Web of Science (1900‐2006). No language limitation was applied. In combination with search terms to identify randomised controlled trials, as defined by the Cochrane Collaboration and detailed in the Cochrane Reviewers' Handbook (Higgins 2005), we used the following search strategy to search MEDLINE:

1. exp osteoarthritis/ 
 2. osteoarthrit$.tw. 
 3. osteoarthro$.tw. 
 4. oa.tw. 
 5. degenerative joint disease.tw. 
 6. degenerative arthritis.tw. 
 7. djd.tw. 
 8. gonarthro$.tw. 
 9. or/1‐8 
 10. exp knee/ 
 11. knee$.tw. 
 12. femorotibia$.tw. 
 13. or/10‐12 
 14. exp Osteoarthritis, Knee/ 
 15. 9 and 13 
 16. 14 or 15 
 17. debridement/ and Arthroscopy/ 
 18. arthroscop$ debride$.tw. 
 19. 17 or 18 
 20. 16 and 19

The strategy was adapted for each database (see details in Additional Table 1). The bibliographies, reference lists and cited web sites of all papers identified by these strategies were searched. We also contacted the Cochrane Musculoskeletal Group for handsearching, but this could not be completed due to resource limitations.

1. Search strategy ‐ additional detail.

CINAHL	EMBASE	CENTRAL	Web of Science
1. exp osteoarthritis/ 2. osteoarthrit$.tw. 3. osteoarthro$.tw. 4. oa.tw. 5. degenerative joint disease.tw. 6. degenerative arthritis.tw. 7. djd.tw. 8. gonarthro$.tw. 9. or/1‐8 10. exp KNEE/ 11. knee$.tw. 12. femorotibia$.tw. 13. or/10‐12 14. 9 and 13 15. DEBRIDEMENT/ 16. ARTHROSCOPY/ 17. 15 and 16 18. arthroscop$ debride$.tw. 19. 17 or 18 20. 14 and 19	1. exp osteoarthritis/ 2. osteoarthrit$.tw. 3. osteoarthro$.tw. 4. oa.tw. 5. degenerative joint disease.tw. 6. degenerative arthritis.tw. 7. djd.tw. 8. gonarthro$.tw. 9. or/1‐8 10. exp KNEE/ 11. knee$.tw. 12. femorotibia$.tw. 13. or/10‐12 14. 9 and 13 15. exp Knee Osteoarthritis/ 16. 14 or 15 17. exp DEBRIDEMENT/ 18. exp ARTHROSCOPY/ 19. 17 and 18 20. arthroscop$ debride$.tw. 21.19 or 20 22.16 and 21	1. MeSH descriptor Osteoarthritis explode all trees in MeSH products 2. osteoarthrit* in All Fields, from 1800 to 2005 in all products 3. osteoarthro* in All Fields, from 1800 to 2005 in all products 4. oa in All Fields, from 1800 to 2005 in all products 5. "degenerative joint disease" in All Fields, from 1800 to 2005 in all products 6. degenerative arthritis in All Fields in all products 7. djd in All Fields in all products 8. gonarthro* in All Fields in all products 9. (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8) 10. MeSH descriptor Knee explode all trees in MeSH products 11. knee* in All Fields, from 1800 to 2005 in all products 12. femorotibia* in All Fields, from 1800 to 2005 in all products 13. (#10 OR #11 OR #12) 14. (#9 AND #13) 15. MeSH descriptor Osteoarthritis, Knee explode all trees in MeSH products 16. (#14 OR #15) 17. MeSH descriptor Debridement explode all trees in MeSH products 18. MeSH descriptor Arthroscopy explode all trees in MeSH products 19. (#17 AND #18) 20. arthroscop* next debride* in All Fields in all products 21. (#19 OR #20) 22. (#16 AND #21)	1 (arthroscop* debride*) and (osteoarthr* and knee*)

Data collection and analysis

Selection of trials 
 Two review authors (WL and ML) independently selected the trials, initially based on title, keywords and abstract of references identified by the search strategy. We assessed whether the study met the inclusion criteria regarding diagnosis, participants and intervention. We retrieved the full article of the selected trials, and trials that raised disagreement or doubt during the selection, for final assessment. Disagreements on inclusion were resolved by discussion and decided by consensus.

Quality assessment 
 We assessed the methodological quality of the selected trials using the criteria described in the Cochrane Handbook (Higgins 2005). 
 Methods used for generation of the randomisation sequence were described for each trial.

(1) Selection bias (randomisation and allocation concealment) 
 We assessed the possibility of selection bias for each trial, using the following criteria: 
 (A) adequate concealment of allocation: such as telephone randomisation, consecutively numbered, sealed opaque envelopes; 
 (B) unclear whether adequate concealment of allocation: such as list or table used, sealed envelopes, or study does not report any concealment approach; 
 (C) inadequate concealment of allocation: such as open list of random‐number tables, use of case record numbers, dates of birth or days of the week; 
 (D) concealment of allocation not used.

(2) Performance bias (blinding of participants, researchers and outcome assessment) 
 We assessed performance bias for each trial, using the following criteria: 
 (2.1) blinding of participants 
 yes: such as patients did not know which procedure they received 
 no: such as patients knew which procedure they received 
 unclear: no information 
 (2.2) blinding of outcome assessment 
 yes: such as investigators measured pain among the patients without awareness of the interventions they received; 
 no: such as pain was measured from the patients among the treatment groups 
 unclear: investigators measured pain among the patients similarly

(3) Attrition bias (loss of participants, for example, withdrawals, dropouts, protocol deviations) 
 We assessed completeness to follow up using the following criteria: 
 (A) less than 5% loss of participants; 
 (B) 5% to 9.9% loss of participants; 
 (C) 10% to 19.9% loss of participants; 
 (D) more than 20% loss of participants.

(4) Sample size calculation 
 (A) adequate explanation of sample size calculation: such as all information related to sample size calculation were available 
 (B) unclear whether the sample size was calculated or no available information 
 (C) inadequate explanation of sample size calculation: such as some information related to sample size calculation were available 
 (D) not calculated

Low risk of bias was defined as those receiving an 'A' rating for selection bias, attrition bias and sample size calculation, and 'yes' for blinding of participants and outcome assessment. 
 Moderate risk of bias was defined as those receiving at least one 'B' or 'C' rating for selection bias, attrition bias, sample size calculation , or 'unclear' for blinding of participants or outcome assessment. 
 High risk of bias was defined as those receiving at least one ' D' or 'No rating' for selection bias, attrition bias, sample size calculation, and blinding of participants or outcome assessment. 
 
 The evidence of review was graded according to the Cochrane Musculoskeletal Group Method Guidelines (Maxwell 2006) as the following:

Platinum level 
 The Platinum ranking is given to evidence that comprises a published systematic review that has at least two individual controlled trials each satisfying the following:

• Sample sizes of at least 50 per group. If they do not find a statistically significant difference, they are adequately powered for a 20% relative difference in the relevant outcome.

• Blinding of patients and assessors for outcomes.

• Handling of withdrawals >80% follow up (imputations based on methods such as Last Observation Carried Forward (LOCF) acceptable).

• Concealment of treatment allocation.

Gold level 
 The Gold ranking is given to evidence if at least one randomised clinical trial meets all of the following criteria for the major outcome(s) as reported:

• Sample sizes of at least 50 per group. If they do not find a statistically significant difference, they are adequately powered for a 20% relative difference in the relevant outcome.

• Blinding of patients and assessors for outcomes.

• Handling of withdrawals > 80% follow up (imputations based on methods such as Last Observation Carried Forward (LOCF) acceptable).

• Concealment of treatment allocation.

Silver level 
 The Silver ranking is given to evidence if randomised trial does not meet the above criteria. Silver ranking would also include evidence from at least one study of non‐randomised cohorts who did and did not receive the therapy or evidence from at least one high quality case‐control study. A randomised trial with a 'head‐to‐head' comparison of agents is considered Silver level ranking unless a reference is provided to a comparison of one of the agents to placebo showing at least a 20% relative difference.

Bronze level 
 The bronze ranking is given to evidence if there is at least one high quality case series without controls (including simple before/after studies in which the patient acts as their own control) or if it is derived from expert opinion based on clinical experience without reference to any of the foregoing (for example, argument from physiology, bench research or first principles).

Data extraction 
 We modified data collection forms developed by the Cochrane Musculoskeletal Group for data inclusion or exclusion. Two review authors (WL and ML) independently extracted the data. Extracted items included: 
 1. General information: publication, title, authors, contact address, country, resource, publication year, duplication of the publishing, sponsor. 
 2. Characteristics of the study: design, sample size calculation, research setting, inclusion/exclusion criteria, randomisation method, concealment, allocation procedure, blinding (patients, caregivers and outcome appraisers). 
 3. Intervention: treatment (AD) versus comparisons (placebo or sham operation or other types of therapy, e.g., NSAIDs, analgesics, non‐pharmacological treatments and other surgeries). 
 4. Patients: characteristics (sex, age, ethnic group, side, location of lesions in the knee, diagnostic criteria, staging criteria of the OA, duration of disease, co‐morbidities), total number or number in each study group. 
 5. Outcomes: Level of pain and functions measured at multiple points as continuous data. Length of the study and other outcomes reported in papers were also extracted. 
 6. Results: We paid particular attention to 'intention‐to‐treat' and dropout rate.

Analysis 
 Dichotomous outcomes were expressed as the relative risk (RR). Continuous outcomes were expressed as the weighted mean difference (WMD). We planned to meta‐analyse the data using the statistical package in Review Manager software (Review Manager 2005). However, since the included trials measured different comparison groups, this was not appropriate. We will conduct a meta‐analysis if more trials are found when updating our review in the future.

Clinical relevance tables 
 We compiled clinical relevance tables for primary outcomes under Additional Tables to improve the readability of the review. For dichotomous outcomes, we calculated the weighted absolute risk difference using the risk difference (RD) statistic in RevMan. RR‐1 calculates the weighted relative percent change. We calculated the number needed to treat (NNT) to benefit (NNTB) or harm (NNTH) from the control group event rate (unless the population event rate was known) and the relative risk using the Visual Rx NNT calculator (Cates 2004). This was done for the primary outcomes measured.

For continuous outcome tables, see Additional Tables. We calculated weighted absolute change from the weighted mean difference (WMD) statistic in RevMan when trials using the same scale were pooled. For outcomes pooled on different scales, we multiplied the standardised mean difference (SMD) by the baseline standard deviation in the control group to obtain the weighted absolute change. We calculated relative percent change from baseline as the absolute benefit divided by the baseline mean of the control group. We calculated NNT using the Wells calculator available at the Cochrane Musculoskeletal Group editorial office. We determined the minimal clinically important difference (MCID) for each outcome for input into the calculator.

Results

Description of studies

We initially identified 18 studies from multiple database searches. All were in English. After screening their titles and most available abstracts, we retrieved the full text of four studies (Chang 1993; Hubbard 1996; Moseley 2002; Forster 2003) for considering their potential eligibility. We finally excluded 15 studies for various reasons. Three of them were trials, twelve were not. Three studies were commentaries and one was a pilot study. Reasons for excluding individual studies are shown in the Table of Characteristics of Excluded Studies.

We included three trials (Chang 1993; Hubbard 1996; Moseley 2002). Details of each trial are shown in the Table of Characteristics of Included Studies.

Chang 1993 randomised 32 of 34 eligible patients stratified by each of the two study sites to receive either arthroscopic surgery (18) or closed‐needle joint lavage (14). Two eligible patients were withdrawn before giving interventions due to concurrent medical problems. A single assessor for each site was blinded to the patient's treatment allocation when measuring all outcomes at 3 and 12 months follow‐up. Pain and functional status were measured by the Arthritis Impact Measurement Scales (AIMS), from zero (best) to ten (worst), according to the patient's responses to the self‐administered questionnaire. Baseline characteristics were similar in the two groups, except the mean initial AIMS Physical Activity score, which the authors attributed to random variation.

Hubbard 1996 randomised 76 knees of eligible patients to receive either AD (40 knees) or washout (36 knees). Pain was measured as 'success' for its absence and 'failure' for its presence. Physical function was measured as a modified Lysholm score with a maximum of 70 points. Outcome assessors were neither independent nor blinded. The outcomes were measured at three months, 12 months and every year until five years after the intervention, but were only reported at one and five years. Baseline characteristics of the two groups were not reported.

In Moseley 2002, there were three treatment groups: AD, lavage and placebo surgery. Three hundred and twenty‐four patients were eligible. Among them, 180 (56%) agreed to participate to the trial. Participants were younger than those who refused to participate (mean age 52.3 ± SD 11.3 years for participants versus 55.3 ± SD 12.4 years for those who refused), were more likely to be caucasian (62.2% versus 50.7%), and had more severe arthritis (25.0% versus 12.5 % with grade seven or eight arthritis). The 180 participants were randomly assigned to receive AD (59), arthroscopic lavage (61), or placebo surgery (60). Patients and assessors were blinded to the treatment group assignment. Pain and physical function were measured at two weeks, six weeks, three months, six months, 12 months, 18 months and 24 months after the intervention. Pain was the primary outcome, assessed by a 12‐item, self‐reported Knee‐Specific Pain Scale (KSPS, 0 to 100 scale, higher scores indicating more severe pain). Physical function was measured by two self‐reported scales: the five‐item walking‐bending subscale from the AIMS2 tool (0 to 100 scale, higher scores indicating more limited function) and the ten‐item physical‐function subscale from the SF‐36 tool (0 to 100 scale, higher scores indicating better function). Baseline characteristics were similar in the three study groups.

Risk of bias in included studies

Allocation was adequately concealed in Hubbard 1996 and Moseley 2002. In Hubbard 1996, the sequence of numbers was computer‐generated and kept in sealed envelopes. In Moseley 2002, a stratified randomisation process with fixed blocks of six was used. Sealed, sequentially numbered, stratum‐specific envelopes containing treatment assignments were prepared and given to the research assistant. The allocation concealment was unclear in Chang 1993.

Participants were blinded to treatment allocation in two trials (Hubbard 1996; Moseley 2002), but not in Chang 1993. Outcome assessors were blinded to treatment allocation in Chang 1993 and Moseley 2002, but not blinded in Hubbard 1996.

In Chang 1993, 22% of participants withdrew from the arthroscopy group and 7% from the control group at 12 month follow‐up. Sensitivity analysis was conducted for the missing data and no effect on the summary measures were found. In Hubbard 1996, 20% of patients were lost to follow‐up in the debridement group and 28% in the wash‐out group. Around 10% of patients in each group were lost to follow‐up in Moseley 2002.

Sample size calculation was clearly explained only in Moseley 2002. However, this trial was originally designed to test for superiority of the arthroscopic procedures over the sham surgery, but the authors then changed their analysis to test for equivalence, after evidence of the superiority was not found. They also set the minimal important difference based on the trial data and available literature. These decisions may have led to bias in the interpretation of results.

Overall, Moseley 2002 was assessed to be at moderate risk of bias while the other two trials (Chang 1993; Hubbard 1996) were at high risk of bias.

Effects of interventions

Results of the three trials are individually described due to differences of the comparison groups and heterogeneity of the clinical and methodological aspects, which precluded meta‐analysis of results.

Two trials (Chang 1993; Moseley 2002) compared AD to lavage. Their results were presented separately because the scales of pain scores and physical function were different.

AD versus closed‐needle joint lavage 
 Chang 1993 found that after controlling for baseline differences, the adjusted mean AIMS pain scores were 5.0 in the AD group and 5.4 in the lavage group with no statistically significant difference (WMD ‐0.4, 95% CI ‐1.6 to 0.8) at three months of follow‐up (see 'Analyses: Comparison 01, Outcome 01' and 'Additional Table 2'). The adjusted mean AIMS pain scores at 12 months of follow‐up were 5.3 in the AD group and 5.0 in the lavage group with no statistically significant difference (WMD 0.3, 95% CI ‐1.1 to 1.8)(see 'Analyses: Comparison 01, Outcome 01' and 'Additional Table 2').

2. Clinical relevance table: AD versus closed‐needle joint lavage.

Outcome	#patients (#trials)	Control baseline m	Wt absolute change	Relative % change	NNT (B) or NNT (H)	Stat. significance	Quality of evidence
Pain at 3 months (AIMS scale, 0‐10)	32 (1)	6.1	‐4% (0 fewer points on a 0‐10 scale)	‐7% (I)	n/a	Not significant	Silver
95% CI			(‐16%, 8%)	(‐26% (I), 13% (W))
Pain at 12 months (AIMS scale, 0‐10)	32 (1)	6.1	3% (0 more points on a 0‐10 scale)	5% (W)	n/a	Not significant	Silver
95% CI			(‐11%, 17%)	(‐19% (I), 28% (W))
Physical function at 3 months (AIMS scale, 0‐10)	32 (1)	1.7	‐5% (0 fewer points on a 0‐10 scale)	‐30% (I)	n/a	Not significant	Silver
95% CI			(‐12%, 2%)	(‐71% (I), 29% (W))
Physical function at 12 months (AIMS scale, 0‐10)	32 (1)	1.7	‐3% (0 fewer points on a 0‐10 scale)	‐18% (I)	n/a	Not significant	Silver
95% CI			(‐11%, 5%)	(‐65% (I), 29% (W))
Legend		m = mean	Wt = weighted	I = improvement W = worsening	NNT = number needed to treat B = benefit H = harm

AD versus lavage 
 Moseley 2002 presented results from a total of 163 patients who completed the trial at 24 months. The pain scores showed quite a big decrease from the baseline of around 10 points in AD and lavage at two weeks after the intervention. The WMD for pain scores was 2.5 (95% CI ‐4.4 to 9.4) (see 'Analyses: Comparison 02, Outcome 01' and 'Additional Table 3'). After that the pain scores fluctuated less than 5 points at each measurement and the WMD was not statistically significant at any of the measurement points (up to 24 months after the intervention) (see 'Analyses: Comparison 02, Outcome 01' and 'Additional Table 3'). A similar pattern but smaller changes were seen in physical function. The WMD difference at 24 months was ‐0.6 (95% CI ‐8.3 to 7.1), with higher scores indicating more limited function (see 'Analyses: Comparison 02, Outcome 02' and 'Additional Table 3'). The authors reported that 79.7% of participants in the AD group and 88.5 % of patients in the lavage group used analgesics (prescribed and non‐prescribed). 
 
 AD versus washout 
 Hubbard 1996 found a significant difference in pain relief with a relative risk of 5.76 (95% CI 2.52 to 13.18) between debridement and washout at one year follow‐up (see 'Analyses: Comparison 03, Outcome 01'). A significant difference in pain relief of 5.15 (95% CI 1.71 to 15.49) between debridement and washout at five years follow‐up was also found (see 'Analyses: Comparison 03, Outcome 01'), and the number needed to treat to benefit was 2 at one year and 3 at five year follow‐up (see 'Additional Table 4'). Physical function measured as mean modified Lysholm scores were presented without standard deviations for each subgroup for pain relief (success or failure). The scores were similar for each comparable pain relief subgroup. The higher mean scores were seen in the success groups with 61 for debridement versus 63 for washout at one year follow‐up, and 58 for debridement versus 59 for washout at five years follow‐up. Lower mean scores were seen in the failure groups, with 33 for debridement versus 35 for washout at one year and five year follow‐up (data not shown). 
 
 AD versus placebo 
 Moseley 2002 found a large decrease of 19 points from the baseline in the placebo group at two weeks after the intervention. The WMD for pain was 8.7 (95% CI 1.7 to 15.8), indicating a statistically significant result in favour of the placebo group (that is, the AD group experienced more pain) (see 'Analyses: Comparison 04, Outcome 01'), and the number needed to treat to harm was 5 (see 'Additional Table 5). After that the pain scores fluctuated, and the WMD at each measurement point was not statistically significant at any of the other measurement points (up to 24 months after the intervention) (see Analyses: Comparison 04, Outcome 01'). A similar pattern of changing scores was seen for physical function. The WMD for function at two weeks was 7.7 (95% CI 1.1 to 14.3), indicating that the AD group experienced significantly more limited function (see 'Analyses: Comparison 04, Outcome 02'), and the number needed to treat to harm was 6 (see 'Additional Table 5'). A second statistically significant result was found at 12 months follow‐up, finding a WMD of 6.9 (95% CI 0.4 to 13.4) (see 'Analyses: Comparison 04, Outcome 01'), and the number needed to treat to harm was 9 (see 'Additional Table 5'). The authors reported that 79.7% of participants in the AD group and 91.7% in the placebo group used analgesics. 
 
 The other outcomes of interest for this review were not measured by the included studies, including time to next major intervention; post‐operative morbidities or complications; subsequent interventions; use of topical drugs; injection of corticosteroids or any chondroprotective agents; taking of novel medications such as glucosamine sulfate and diacerein; and data on confounding factors (litigation, body weight, co‐morbidity, etc.).

3. Clinical relevance table: AD versus lavage.

Outcome (scale)	#patients (#trials)	Control baseline m	Wt absolute change	Relative % change	NNT (B) or NNT (H)	Stat. significance	Quality of evidence
Pain at 2 weeks (KSPS scale, 0‐100)	118 (1)	65.0	3% (3 more points on a 0‐100 scale)	4% (W)	n/a	Not significant	Gold
95% CI			(‐4%, 9%)	(‐7% (I), 14% (W))
Pain at 6 weeks (KSPS scale, 0‐100)	116 (1)	65.0	‐2% (2 fewer points on a 0‐100 scale)	‐3% (I)	n/a	Not significant	Gold
95% CI			(‐10%, 6%)	(‐15% (I), 9% (W))
Pain at 3 months (KSPS scale, 0‐100)	117 (1)	65.0	‐4% (4 fewer points on a 0‐100 scale)	‐6% (I)	n/a	Not significant	Gold
95% CI			(‐12%, 4%)	(‐18% (I), 6% (W))
Pain at 6 months (KSPS scale, 0‐100)	115 (1)	65.0	‐3% (3 fewer points on a 0‐100 scale)	‐5% (I)	n/a	Not significant	Gold
95% CI			(‐11%, 5%)	(‐17% (I), 7% (W))
Pain at 12 months (KSPS scale, 0‐100)	107 (1)	65.0	‐3% (3 fewer points on a 0‐100 scale)	‐5% (I)	n/a	Not significant	Gold
95% CI			(‐11%, 5%)	(‐17% (I), 8% (W))
Pain at 18 months (KSPS scale, 0‐100)	107 (1)	65.0	0% (0 fewer points on a 0‐100 scale)	‐1% (I)	n/a	Not significant	Gold
95% CI			(‐10%, 9%)	(‐15% (I), 13% (W))
Pain at 24 months (KSPS scale, 0‐100)	108 (1)	65.0	‐2% (2 fewer points on a 0‐100 scale)	‐4% (I)	n/a	Not significant	Gold
95% CI			(‐11%, 7%)	(‐17% (I), 10% (W))
Physical function at 2 weeks (AIMS scale, 0‐100)	114 (1)	48.5	3% (3 more points on a 0‐100 scale)	6% (W)	n/a	Not significant	Gold
95% CI			(‐6%, 12%)	(‐12% (I), 24% (W))
Physical function at 6 weeks (AIMS scale, 0‐100)	112 (1)	48.5	2% (2 more points on a 0‐100 scale)	5% (W)	n/a	Not significant	Gold
95% CI			(‐6%, 10%)	(‐12% (I), 21% (W))
Physical function at 3 months (AIMS scale, 0‐100)	111 (1)	48.5	1% (1 more points on a 0‐100 scale)	1% (W)	n/a	Not significant	Gold
95% CI			(‐6%, 8%)	(‐13% (I), 16% (W))
Physical function at 6 months (AIMS scale, 0‐100)	106 (1)	48.5	0% (0 more points on a 0‐100 scale)	1% (W)	n/a	Not significant	Gold
95% CI			(‐7%, 8%)	(‐14% (I), 16% (W))
Physical function at 12 months (AIMS scale, 0‐100)	101 (1)	48.5	2% (2 more points on a 0‐100 scale)	4% (W)	n/a	Not significant	Silver
95% CI			(‐5%, 10%)	(‐11% (I), 20% (W))
Physical function at 18 months (AIMS scale, 0‐100)	93 (1)	48.5	2% (2 more points on a 0‐100 scale)	3% (W)	n/a	Not significant	Silver
95% CI			(‐7%, 10%)	(‐13% (I), 20% (W))
Physical function at 24 months (AIMS scale, 0‐100)	94 (1)	48.5	‐1% (1 fewer points on a 0‐100 scale)	‐1% (I)	n/a	Not significant	Silver
95% CI			(‐8%, 7%)	(‐17% (I), 15% (W))
Legend		m = mean	Wt = weighted	I = improvement W = worsening	NNT = number needed to treat B = benefit H = harm

4. Clinical relevance table: AD versus washout.

Outcome	#patients (#trials)	Control event rate	Wt Absolute RD	Wt Rel % change	NNT(B) or NNT(H)	Stat. significance	Quality of evidence
Pain free at 1 year	76 (1)	13.9% 14 out of 100	66% 66 more out of 100	476% (I)	NNT (B) = 2	Significant	Silver
95% CI			(49, 83)	(152% (I), 1218% (I))	(2,5)
Pain free at 5 years	58 (1)	11.5% 12 out of 100	48% 48 more out of 100	415% (I)	NNT (B) = 3	Significant	Silver
95% CI			(27, 69)	(71% (I), 1449% (I))	(2, 13)
Legend			Wt = weighted RD = risk difference	Wt Rel = weight relative I = improvement W = worsening	NNT = number needed to treat B = benefit H = harm

5. Clinical relevance table: AD versus placebo.

Outcome (scale)	#patients (#trials)	Control baseline m	Wt absolute change	Relative % change	NNT(B) or NNT(H)	Stat. significance	Quality of evidence
Pain at 2 weeks (KSPS scale, 0‐100)	118 (1)	55.0	9% (9 more points on a 0 to 100 scale)	16% (W)	NNT(H) = 5	Significant	Gold
95% CI			(2%, 16%)	(3%(W), 29% (W))	(3, 27)
Pain at 6 weeks (KSPS scale, 0‐100)	116 (1)	55.0	4% (4 more points on a 0 to 100 scale)	7% (W)	n/a	Not significant	Gold
95% CI			(‐5%, 12%)	(‐8% (I), 21% (W))
Pain at 3 months (KSPS scale, 0‐100)	114 (1)	55.0	1% (1 more point on a 0 to 100 scale)	1% (W)	n/a	Not significant	Gold
95% CI			(‐7%, 8%)	(‐14% (I), 15% (W))
Pain at 6 months (KSPS scale, 0‐100)	113 (1)	55.0	2% (2 more points on a 0 to 100 scale)	4% (W)	n/a	Not significant	Gold
95% CI			(‐5%, 10%)	(‐10% (I), 18% (W))
Pain at 12 months (KSPS scale, 0‐100)	103 (1)	55.0	3% (3 more points on a 0 to 100 scale)	5% (W)	n/a	Not significant	Gold
95% CI			(‐6%, 11%)	(‐10% (I), 21% (W))
Pain at 18 months (KSPS scale, 0‐100)	103 (1)	55.0	‐2% (2 fewer points on a 0 to 100 scale)	‐3% (I)	n/a	Not significant	Gold
95% CI			(‐11%, 8%)	(‐20% (I), 14% (W))
Pain at 24 months (KSPS scale, 0‐100)	108 (1)	55.0	0% (0 fewer points on a 0 to 100 scale)	0%	n/a	Not significant	Gold
95% CI			(‐9%, 9%)	(‐16% (I), 16% (W))
Physical function at 2 weeks (AIMS scale, 0‐100)	116 (1)	48.5	8% (8 more points on a 0‐100 scale)	16% (W)	NNT(H) = 6	Significant	Gold
95% CI			(1%, 14%)	(2%, 30%)	(3, 45)
Physical function at 6 weeks (AIMS scale, 0‐100)	114 (1)	48.5	6% (6 more points on a 0‐100 scale)	12% (W)	n/a	Not significant	Gold
95% CI			(‐1%, 13%)	(‐3% (I), 27% (W))
Physical function at 3 months (AIMS scale, 0‐100)	110 (1)	48.5	2% (2 more points on a 0‐100 scale)	5% (W)	n/a	Not significant	Gold
95% CI			(‐4%, 8%)	(‐8% (I), 17% (W))
Physical function at 6 months (AIMS scale, 0‐100)	108 (1)	48.5	3% (3 more points on a 0‐100 scale)	6% (W)	n/a	Not significant	Gold
95% CI			(‐3%, 9%)	(‐6% (I), 18% (W))
Physical function at 12 months (AIMS scale, 0‐100)	96 (1)	48.5	7% (7 more points on a 0‐100 scale)	14% (W)	NNT(H) = 9	Significant	Silver
95% CI			(0%, 13%)	(1% (W), 28% (W))	(4, 199)
Physical function at 18 months (AIMS scale, 0‐100)	90 (1)	48.5	4% (4 more points on a 0‐100 scale)	9% (W)	n/a	Not significant	Silver
95% CI			(‐3%, 11%)	(‐6% (I), 24% (W))
Physical function at 24 months (AIMS scale, 0‐100)	88 (1)	48.5	5% (5 more points on a 0‐100 scale)	10% (W)	n/a	Not significant	Silver
95% CI			(‐1%, 11%)	(‐2% (I), 22% (W))
Legend		m = mean	Wt = weighted	I = improvement W = worsening	NNT = number needed to treat B = benefit H = harm

Discussion

Our systematic review includes three studies with different comparison groups. There is only one RCT of moderate quality (Moseley 2002) that shows the effect of AD in comparison to placebo. In this study, AD did not differ significantly from lavage or placebo surgery for pain and physical functions at two years, although results indicated that AD may in fact be worse over the short term. Of the two low quality trials, Chang 1993 found no significant difference between AD and closed‐needle joint lavage, but Hubbard 1996 found that AD was significantly superior to arthroscopic washout. Hubbard 1996 included participants with degenerative lesions of grade three or four on the Outerbridge classification (Outerbridge 1961) and confined at the medial femoral condyle. This study gives some information on our secondary objective regarding the type or stage of severity of the OA in which AD is most effective. However, there is no study giving a solution to the expected length of effectiveness until the patients need further intervention. 
 
 It is interesting that there is only one RCT showing the effect of AD on knee OA in comparison with placebo. The limitation of such trials could be due to difficulty in conducting research on placebo effects for surgical intervention. This requires comparison with a sham operation, which is subject to comprehensive criticism about medical ethics. Although direct comparison with placebo is important, we feel that researchers investigating a similar research question should compare alternative treatment options to increase the number of options for people with knee OA who have not responded to conservative treatments. 
 
 The finding reported by Moseley et al (Moseley 2002) is striking because it is contrary to most of the previous literature, which indicated AD as the treatment of choice after failure of conservative therapies in controlling osteoarthritic symptoms of the knee, especially when there are intra‐articular mechanical derangements (Sprague 1981; Baumgaertner 1990; McLaren 1991; Harwin 1999; McGinley 1999; Shannon 2001). However, these conclusions are mainly based on case series or observational studies, which do not provide strong evidence.

We agree with the other commentaries (Bernstein 2003; Gillespie 2003) that Moseley's trial (Moseley 2002) was well planned and the design was robust. The authors had properly described processes of the trial and the validity was strengthened by the concealed randomisation.

The authors reported that the study was based on 56% of eligible subjects who were younger, more likely to be white and had more severe arthritis than the 44% eligible subjects who refused to participate in the study. The findings are therefore limited in their generalisability to common clinical practice, such as for people with earlier stages of OA arising from a specific cause. In addition, those who did participate were more likely to trust and expect benefits from the arthroscopic interventions, which may have positively influenced the results for the placebo group. Future research should compare AD, placebo surgery and no intervention to evaluate this effect. 
 
 Although pain was one of the primary outcome measures for this review, it is a subjective outcome and tends to be modified by various confounding factors such as the use of rescue analgesics, NSAIDs and/or other concomitant therapies. The analgesic use reported in Moseley 2002 provided information on other agents that could be important confounders of the results. It would be important for future research in this area to measure these outcomes, as well as other objective end points such as length of time until subsequent interventions are required (e.g. osteotomy or replacement arthroplasty).

Although neither Moseley 2002 nor Chang 1993 identified a benefit for AD, debates continue about the procedure (Bernstein 2003; Day 2005). The studies included in this review did not examine whether there are specific indications or levels of disease at which AD is more effective. For example, if the OA stage is too advanced (e.g., with significant deformities), then the patient might not be a good responder; while too early (e.g., with only minimal chondral lesions), the procedure may not be justifiable. Since the procedures affect only superficial structures, it is reasonable to believe that they might not be able to alleviate pain caused by the deeper subchondral bone.

The studies also did not compare the different techniques and components of AD. Chang 1993 compared AD with lavage, and found that the removal of soft tissue did not contribute to a better result over lavage alone, with the exception of a subgroup with particular meniscal tears. More evidence is required to identify the mechanical effects that are most important, and also to explain the positive effects found in the placebo participants in this review.

Authors' conclusions

Implications for practice.

Based on the results of this review, we conclude that there is gold level evidence (Moseley 2002) that AD has no significant benefit for knee OA of undiscriminated cause. Debatable areas remain to be addressed, for example, there may be groups of patients or levels of severity of disease for which the intervention may be effective. Hubbard 1996 found that AD provides more successful results for localised lesion on the medial femoral condyle than arthroscopic washout, but the study was of lower methodological quality.

Implications for research.

New, high quality research on larger numbers of participants should be conducted to investigate the effects of AD, in particular comparing groups of people with different levels of disease severity and other disease characteristics. Outcomes measured should include survival data on the time to subsequent interventions such as rescue NSAIDs or analgesics or other surgical interventions. Different techniques for AD should be compared. It would also be interesting to investigate the strength of placebo effects of sham surgery over no intervention or conservative treatments on pain and dysfunction of the knee.

What's new

Date	Event	Description
19 September 2008	Amended	Converted to new review format. A004‐R

Data and analyses

Comparison 1. Debridement versus closed‐needle joint lavage.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Pain in AIMS scale (scores from 0 (best) to 10 (worst))	1		Mean difference (Fixed, 95% CI)	Totals not selected
1.1 at 3 months	1		Mean difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 at 12 months	1		Mean difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
2 Physical function in AIMS scale (scores from 0 (best) to 10 (worst))	1		Mean difference (Fixed, 95% CI)	Totals not selected
2.1 at 3 months	1		Mean difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 at 12 months	1		Mean difference (Fixed, 95% CI)	0.0 [0.0, 0.0]

1.1. Analysis.

Comparison 1 Debridement versus closed‐needle joint lavage, Outcome 1 Pain in AIMS scale (scores from 0 (best) to 10 (worst)).

1.2. Analysis.

Comparison 1 Debridement versus closed‐needle joint lavage, Outcome 2 Physical function in AIMS scale (scores from 0 (best) to 10 (worst)).

Comparison 2. Debridement versus lavage.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Pain in KSPS scale (scores from 0 (best) to 100 (worst))	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.1 At 2 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 At 6 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.3 At 3 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.4 At 6 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.5 At 12 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.6 At 18 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.7 At 24 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2 Physical function in AIMS scale (scores from 0 (best) to 100 (worst))	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.1 At 2 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 At 6 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.3 At 3 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.4 At 6 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.5 At 12 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.6 At 18 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.7 At 24 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

2.1. Analysis.

Comparison 2 Debridement versus lavage, Outcome 1 Pain in KSPS scale (scores from 0 (best) to 100 (worst)).

2.2. Analysis.

Comparison 2 Debridement versus lavage, Outcome 2 Physical function in AIMS scale (scores from 0 (best) to 100 (worst)).

Comparison 3. Debridement versus washout.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Pain free	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected
1.1 at 1 year	1		Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 at 5 years	1		Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]

3.1. Analysis.

Comparison 3 Debridement versus washout, Outcome 1 Pain free.

Comparison 4. Debridement versus placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Pain in KSPS scale (scores from 0 (best) to 100 (worst))	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.1 At 2 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 At 6 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.3 At 3 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.4 At 6 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.5 At 12 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.6 At 18 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.7 At 24 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2 Physical function in AIMS scale (scores from 0 (best) to 100 (worst))	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.1 At 2 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 At 6 weeks	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.3 At 3 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.4 At 6 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.5 At 12 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.6 At 18 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.7 At 24 months	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

4.1. Analysis.

Comparison 4 Debridement versus placebo, Outcome 1 Pain in KSPS scale (scores from 0 (best) to 100 (worst)).

4.2. Analysis.

Comparison 4 Debridement versus placebo, Outcome 2 Physical function in AIMS scale (scores from 0 (best) to 100 (worst)).

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Chang 1993.

Methods	Method of allocation/randomisation: stratified randomisation by study site Setting: 2 sites, the Rheumatology‐Orthopedic Knee Clinic of the Northwestern Medical Faculty Foundation and the Division of Rheumatology of the Lutheran General Medical Group, USA. Design: Randomised, controlled trial. Power of study: No information. Number of patients randomised: 32. Number of patients analysed: 32. Concealment of allocation: no information. Outcome assessor blinding: Clear; assessors of outcome were blinded to the treatment‐group assignment. Dropout: No. Source of funding: No information.
Participants	Patients > 20 years old and 1) persistent knee pain for longer than 3 months, 2) weight bearing knee radiographs showing grade 1,2 or 3 changes described by Kellgren and Lawrence and 3) willingness to attend followup visits at 3 and 12 months, and give written informed consent. For patients with bilateral disease, the more symptomatic knee was included. Exclusions: the patients with knee surgery within 6 months of study entry, total knee replacement, any concurrent illness having effect on knee functional assessments or precluding arthroscopic surgery.
Interventions	Arthroscopic surgery: 1) debridement of torn meniscus and removal of maniscal and cruciate ligament fragments, 2) removal of proliferative synovia, or 3) excision of loose articular cartilage fragments. During the procedure, the patients received continuous saline lavage. Closed‐needle joint lavage: Giving non‐narcotic analgesia and physical therapy identical to the arthroscopy group. Tidal knee lavage procedure using a total of 1 liter of saline injected into and aspirated from the knee in aliquots of 40‐120 cc.
Outcomes	A single assessor at each site was blinded to the patient's treatment and assessed the outcomes at 3 and 12 months followup . Pain and functional status measured by the AIMS from 0 (best) to 10 (worst) according to the patient's responses to the self‐administered questionnaire. A decrease of at least 1 point from the baseline was assessed for improvement. Clinical outcome of active and passive range of knee motion, knee joint swelling and tenderness. Global well being outcome measured as 10 cm visual analog scale. Economic outcome based on cost of all arthroscopic surgery.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Hubbard 1996.

Methods	Method of allocation/randomisation: Computer‐generated random numbers. Setting: No information. Design: Randomised controlled trial. Power of study: No information. Number of knees randomised: 76. Number of patients analysed: 76. Concealment of allocation: Sealed, numbered, enveloped. Outcome assessor blinding: Patients and assessors of outcome were unblinded to the treatment group. Dropout: No. Source of funding: Research Committee of the Clwyd Area Health Authority.
Participants	Patients suffering unremitted symptoms in the knee for one year, no previous surgery to the knee, no laxity, no deformity, single medial femoral condyl degenerative lesion grade 3 or 4, no other intra‐articular pathology, normal plain radiograph and modified Lysholm score < 38/70. Exclusions: knees with their radiographs showing a loss of joint space and all which had had a previous operation or steroid injection with any reason. Source of patients: No information. Location: No information.
Interventions	Debridement group: Resection of loose cartilage using a 4.5 mm, 90 degree angled punch through an anteromedial portal with the arthroscope in an anterolateral portal and was completed using straight and curved 3 mm punches with 3 litres of saline running through the knee afterward. Washout group: 3 litres of saline running through the knee in similar manner.
Outcomes	Pain was measured as 'success' for absence of pain and 'failure' for presence of pain. Physical function was measured as a modified Lysholm score with a maximum of 70 points. They were measured at three months, 12 months and every year until 5 years.
Notes	The authors reported without explanation that it had not been possible to use independent observers or blinded observation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Moseley 2002.

Methods	Method of allocation/randomisation: Stratified randomisation process with fixed blocks of six. Setting: 1 centre at the Houston Veterans Affairs Medical Center, USA. Design: Randomised blinded placebo controlled trial. Power of study: 90%. Number of patients randomised: 180. Number of patients analysed: 163. Concealment of allocation: Sealed, sequentially numbered, stratum‐specific envelopes containing treatment assignments were prepared and given to the research assistant. Outcome assessor blinding: Clear; patients and assessors of outcome were blinded to the treatment‐group assignment. Dropout: 17; 5 placebo, 6 lavage and 6 debridement, without detail of dropout reasons. Source of funding: Department of Veterans Affairs.
Participants	Patients 75 years old or younger, osteoarthritis of the knee as defined by the American College of Rheumatology, reported at least moderate knee pain on average ( > 4 on a visual‐analogue scale ranging from 0 to 10) despite maximal medical treatment for at least six months, and had not undergone arthroscopy of the knee during the previous two years. Exclusions: the patients with a severity grade of 9 or higher, severe deformity, and serious medical problems. Source of patients: Houston Veterans Affairs Medical Center Location: Texas, USA.
Interventions	Debridement: lavage with at least 10 litres of fluid, shaving of rough articular cartilage, removal of loose debris, and trimming of torn or degenerated meniscal fragments. No abrasion arthroplasty or microfracture performed. Lavage: the joint was lavaged with at least 10 litres of fluid. Anything that could be flushed out through arthroscopic cannulas was removed. Placebo surgery: received three 1 cm skin incisions under a short acting tranquiliser and an opioid and spontaneously breathed oxygen‐enriched air. simulation of the operating atmospheres but no instrument was admitted into the knee joint.
Outcomes	Pain in the studied knees at 24 months (Knee‐Specific Pain Scale, scores 0‐100 [most severe]) Secondary outcomes: General arthritis pain (Arthritis Impact Measurement Scales [AIMS2]), body pain (pain subscale of the 36‐item Short Form General Health Survey [SF‐36]), and physical function (5‐item walking bending subscale from the AIMS2 and the 10‐item physical function score from the SF‐36).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Dervin 2003	Cohort study
Felson 2002	Commentary
Forster 2003	Interventions were not associated with inclusion criteria
Gillespie 2003	Commentary
Goldman 1997	Review
Grifka 1994	Interventions were not associated with inclusion criteria
Gunther 2001	Review
Hanssen 2001	Review
Knutsen 2004	Interventions were not associated with inclusion criteria
Lubowitz 1993	Descriptive study
Merchan 1993	Prospective studies
Mohtadi 2003	Commentary
Moseley 1996	Pilot study
Stein 2003	Diagnostic Images
Wai 2002	Descriptive study

Contributions of authors

Wiroon Laupattarakasem (WL) proposed the review, developed the protocol and search strategy. Pisamai Laupattarakasem (PL) helped collecting searched results and primarily selected relevant studies from the results. WL and Malinee laopaiboon (ML) selected the studies based on the abstracts and made the final selection after reading the full articles, performed data‐extraction, assessed the methodological quality and performed data‐extraction. WL and ML prepared the full review. (Chut Sumananont, CS, left away for studying during the review period.)

Sources of support

Internal sources

• Khon Kaen University, Thailand.

External sources

• Thailand Research Fund (Senior Research Scholar), Thailand.

Declarations of interest

The review authors do not have any potential conflict of interest regarding internal or external financial support about this review or clinical practices.

Edited (no change to conclusions)