Summary of findings 1. Summary of findings.

Arthroscopic surgery compared to placebo for degenerative knee disease
Patient or population: people with degenerative knee disease (osteoarthritis including degenerative meniscal tears) Setting: surgical Intervention: arthroscopic surgery Comparison: placebo for benefits and all control groups for adverse events including knee replacement
Outcomes	Anticipated absolute effects* (95% CI)			Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Placebo	Arthroscopic surgery	Difference
Paina Scale: 0 to 100, 0 is no pain Follow‐up: 3 months	The mean pain in the placebo group was 40.1 pointsb	The mean pain in the arthroscopic surgery group was 35.5 points	4.6 points better (0.02 better to 9 better)c		309 (4 studies)	⊕⊕⊕⊕ Highd	SMD ‐0.23 (‐0.45 to ‐0.001). Knee arthroscopic surgery results in little or no clinically important improvement in pain. Absolute change 5% better (0.02% better to 9% better) Relative change 8% better (0.03% better to 15% better)e
Knee functiona Scale: 0 to 100, 100 is best function Follow‐up: 3 months	The mean knee function in the placebo group was 75.9 pointsb	The mean knee function in the arthroscopic surgery group was 76 points	0.1 points better (3.2 worse to 3.4 better)		302 (3 studies)	⊕⊕⊕⊕ Highd	SMD 0.01 (‐0.22 to 0.23). Knee arthroscopic surgery results in little or no improvement in function. Absolute change 0.1% better (3% worse to 3% better) Relative change 0.2% better (5% worse to 6% better)e
Knee‐specific health‐related quality of lifea Scale: 0 to 100, 100 is best quality of life Follow‐up: 3 months	The mean quality of life in the placebo group was 69.7 pointsb	The mean quality of life in the arthroscopic surgery group was 75.3 points	5.6 points better (0.4 better to 10.7 better)		188 (2 studies)	⊕⊕⊕⊝ Moderatef	SMD 0.31 (0.02 to 0.59). Knee arthroscopic surgery probably provides little or no clinically important improvement in knee‐specific quality of life. Absolute change 6% better (0.4% to 11% better). Relative change 11% better (0.8% better to 20% better)
Participant‐reported success Last follow‐up	74%	82% (49% to 100%)	8% more (25% fewer to 63% more)	RR 1.11 (0.66 to 1.86)	189 (3 studies)	⊕⊕⊝⊝ Lowf,g	Knee arthroscopic surgery may result in little or no improvement in the number of people reporting success. Relative change 11% more reported success (34% fewer to 86% more)
Serious adverse events Last follow‐upi Events include repeat arthroscopy, pulmonary embolism, deep vein thrombosis, heart attack, death, knee surgery, post‐operative knee infection, anterior cruciate ligament reconstruction	5.6%	7.6% (3.6% to 15.8%)	2% more (2% fewer to 10% more)	RR 1.35 (0.64 to 2.83)	1206 (8 studies)j	⊕⊕⊝⊝ Lowf,h	Knee arthroscopy may or may not lead to more serious adverse events. Relative change 35% more (36% fewer to 183% more)
Total adverse events Last follow‐upi Events include serious events and less serious transient pain in the back, hip, foot, tendonitis, syncope, rupture of Baker's cyst, pain and swelling in index knee after surgery, superficial infection, haemarthrosis, cutaneous nerve lesion, nausea, dizziness	15.0%	17.2% (11.7% to 25.5%)	2% more (3% fewer to 11% more)	RR 1.15 (0.78 to 1.70)	1326 (9 studies)j	⊕⊕⊝⊝ Lowf,h	Knee arthroscopy may or may not slightly increase total adverse events. Relative change 15% more (22% fewer to 70% more)
Knee surgery (replacement or osteotomy) Last follow‐upi	1.5%	5% (1.4% to 10.8%)	2% more (0.1% fewer to 9% more)	RR 2.63 (0.94 to 7.34)	864 (4 studies)j	⊕⊕⊕⊝ Lowk	Knee arthroscopy may or may not lead to slightly more knee surgery. Relative change 163% more (6% fewer to 634% more)
*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; RR: risk ratio
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.

^aPain measured onnumerical rating scale (Sihvonen 2013), Knee‐Specific Pain Scale (KSPS) (Moseley 2002), questionnaire designed specifically for the trial (Moseley 1996); Knee Injury and Osteoarthritis Outcome Score (KOOS) pain subscale (Roos 2018). Knee function measured on Knee Injury and Osteoarthritis Outcome Score (KOOS) (Roos 2018), Lysholm knee score (Sihvonen 2013), Short Form 36‐item questionnaire (SF‐36) bodily pain (Moseley 2002). Knee‐related quality of life (QoL) measured on the KOOS Knee‐related QoL subscale (Roos 2018), and the Western Ontario Meniscal Evaluation tool (WOMET) (Sihvonen 2013)
^bControl group risk was estimated from the placebo value at follow‐up for pain, knee function and knee‐related quality of life in Sihvonen 2013
^cStandardised mean difference (SMD) back‐translated to typical scales by multiplying the SMD by the standard deviation (SD) at baseline in the placebo group as reported in Sihvonen 2013: mean (SD) for knee pain (0 to 100 scale): 60.1 (20.0); mean (SD) knee function (0 to 100 scale): 60.1 (14.6); mean (SD) generic quality of life (15D): 0.90 (0.06); mean (SD) knee‐specific quality of life (WOMET 0 to 100 scale): 52.8 (18.1).
^dOverall, the certainty of evidence was high at 3‐month follow‐up for pain and function. One trial measuring pain was at potential risk of selection bias, but this probably did not change our confidence in the effect estimates. The 95% confidence intervals exclude a clinically important change (defined as 12 points (minimum, maximum: 2, 30) on a 0 to 100 point pain scale; and 13 (3, 34) on a 0 to 100 point WOMAC function scale). Further research is likely to strengthen the conclusion that there was no important differences in pain and function between groups, rather than change the conclusion
^eRelative change: absolute change (mean difference) divided by mean at baseline in the placebo group (values were: 60.1 points on 0 to 100 point pain scale; 60.1 points on 0 to 100 knee function scale; and 52.8 points on 0 to 100 quality of life scale; from Sihvonen 2013).
^fDowngraded due to imprecision: the 95% confidence intervals do not rule in or rule out a clinically important change (defined as 10 points on the 0 to 100 point quality of life scale); or for dichotomous outcomes the total number of participants was small, or number of events was small (< 200); or data were from a single trial only.
^gDowngraded due to indirectness for participant‐reported success as there was diversity in definition and timing of measurement: reported at 6 months, 24 months and 5 years across trials.
^hDowngraded due to possible reporting bias: incomplete reporting of outcome across studies.
ⁱTotal and serious adverse events were reported at 24 months (Roos 2018; Van de Graaf 2018); 25 months (Merchan 1993); and 5 years (Gauffin 2014; Herrlin 2007; Katz 2013; Kise 2016; Sihvonen 2013). Total adverse events only were reported at 6 months in one study (Saeed 2015).
^jFor serious adverse events, adverse events and subsequent knee surgery (replacement or osteotomy), we included trials that compared arthroscopy to placebo or to non‐surgical interventions. For serious adverse events, comparison groups were placebo in 2 trials, exercise in 5 trials and oral non‐steroidal anti‐inflammatory drugs (NSAIDs) in 1 trial. For adverse events, the comparison groups included placebo in 2 trials, exercise in 5 trials, and oral NSAIDs and hyaluronic acid injections in single trials. For knee surgery, the comparison groups included placebo in 1 trial and exercise in 3 trials.
^kDowngraded twice due to imprecision: the 95% confidence intervals do not rule in or rule out a clinically important change as the total number of events was small (< 200).