Will a knee brace stop runner's knee?
Perhaps in heavy training: in 2 small trials (recruits, young athletes, 6-8 weeks) braces cut new kneecap pain by 60%, low certainty. No runner trial, nothing on racing with pain.
Main caveat: Low to very low certainty throughout. Risk of bias was low in 1 of 13 trials; few cases in the brace and retraining trials make the estimates imprecise.
Culvenor and colleagues pooled every RCT they could find that tested a way to prevent new patellofemoral pain diagnosed on physical assessment (self-report alone excluded), against a control or unexposed group.
Thirteen trials with 8210 participants qualified — mostly military recruits, plus young and elite athletes, school pupils and two trials of runners — testing six kinds of intervention.
Two trials of patellofemoral braces worn during all physical training (a silicone patellar ring in young athletes training 6 h a day for 8 weeks; dynamic neoprene braces in Belgian recruits for 6 weeks; 227 participants) gave RR 0.40 (0.22 to 0.73), low certainty, NNT 4.
One trial of 320 novice runners who had 8 sessions over 2 weeks of treadmill feedback to lower the vertical impact peak ("run softer") gave RR 0.21 (0.07 to 0.60) over 12 months, low certainty, NNT 11.
Multicomponent strengthening and neuromuscular programmes (4 trials, 3364 participants) gave a pooled RR of 0.49 (0.18 to 1.36) with substantial heterogeneity, low certainty; foot orthoses and insoles (4 military trials, 2314) RR 0.63 (0.35 to 1.13), very low certainty; static stretching (1538 recruits) and a running programme that progressed intensity rather than volume (447 recreational runners) did not change risk.
Risk of bias was low in only one trial.
The authors conclude that braces and soft-landing retraining may cut the risk by 60-79%, but also that widespread prophylactic bracing should not be recommended outside military and intensive-training settings, and they back appropriately dosed strengthening and neuromuscular warm-ups on the grounds that they are unlikely to harm and may prevent other injuries.
Our reading: for a runner the usable result is the single soft-landing trial (already in the base as REF-1119); the brace result comes from short bursts of very heavy training in which 37-58% of unbraced controls developed knee pain, so the absolute benefit for a recreational runner would be much smaller, and the review has nothing on braces for existing pain.
Protocol
Search: MEDLINE, EMBASE, CINAHL, Web of Science and SPORTDiscus in April 2020, no restriction on year or language at search; full-text English articles eligible.
Eligibility: RCTs of any intervention to reduce incident patellofemoral pain vs a control or non-exposed group. Excluded: pain or injury at baseline unless adjusted for; patellofemoral pain not diagnosed by physical assessment (self-report only); patellofemoral pain reported only together with other injuries.
Trials and participants: 13 RCTs, n = 8210 — multicomponent exercise 4 (elite female basketballers, Italy, 86/74; Danish recruits 487/490; UK recruits 759/743; Canadian junior high school students 353/372), foot orthoses 4 (all military: Australia 153/153, UK 200/200, Finland 73/147, South Africa 237/1151), patellofemoral braces 2 (Israeli young athletes 27/33; Belgian recruits 54/113), static stretching 1 (Australian recruits 735/803), intensity- vs volume-progressed running 1 (Danish recreational runners 221/226), running technique retraining 1 (Hong Kong novice runners 166/154).Mean age 13 to 39 years; follow-up 6 weeks to 12 months; diagnosis mostly by medical doctors.
Analysis: risk ratios (cumulative incidence intervention / control), log-transformed and pooled by random effects in Stata 14, stratified by intervention; I² for heterogeneity; NNT = 1 / absolute risk reduction; 0.5 continuity correction for zero-event arms.Planned short- vs long-term analysis not feasible (only two trials had long-term follow-up); sex and setting subgroups where possible; publication bias not assessable (no intervention had 10 trials). Certainty by GRADE (risk of bias, consistency, indirectness, imprecision, publication bias).
Funding and interests: AGC holds a National Health and Medical Research Council of Australia Early Career Fellowship (Neil Hamilton Fairley Clinical Fellowship, APP1121173); EMM received a Canadian Institutes of Health Research Banting Postdoctoral Fellowship. Competing interests: none declared.
Key Results
Prevention of new patellofemoral pain (Results, Tables 1-2, Figure 2; RR < 1 favours the intervention):
| Intervention (control) | Trials, n, who | RR (95% CI), I² | NNT | Certainty |
|---|---|---|---|---|
| Patellofemoral brace worn for all physical training, 6-8 weeks (silicone ring brace; bilateral dynamic neoprene brace) (no brace) | 2, n = 227; young athletes in 6 h/day training, military recruits | 0.40 (0.22 to 0.73), 24.0% | 4 | low (high risk of bias, few events) |
| Running technique retraining: 8 sessions over 2 weeks, treadmill, instantaneous visual feedback of the vertical impact peak, cue "run softer" (same treadmill running without feedback) | 1, n = 320; novice runners (<2 years), 12-month follow-up | 0.21 (0.07 to 0.60) | 11 | low (high risk of bias, few events) |
| Multicomponent exercise: strengthening and neuromuscular warm-ups, 15-30 min, 2-7 times a week for 12-34 weeks (stretching or light aerobic warm-up) | 4, n = 3364; basketballers, recruits, school pupils | 0.49 (0.18 to 1.36), 64.9% | 60 | low (risk of bias, inconsistency) |
| Foot orthoses or insoles worn 7-26 weeks: three prefabricated (two heat-moulded, one arch-profiled) and one shock-absorbing flat insole (flat insoles or none) | 4, n = 2314; all military | 0.63 (0.35 to 1.13), 0.0% | 14 | very low |
| Static stretching before every training session, 1 × 20 s per muscle (no stretching) | 1, n = 1538; military recruits | no significant influence | 76 | very low |
| Running programme progressing intensity (more time at ≥88% VO2max after each 4-week block) (progressing volume by 23% per block) | 1, n = 447; recreational runners, 24 weeks | no significant influence; about 36% higher risk with intensity | number needed to harm 207 | low |
Subgroups and sensitivity: foot orthoses in the two male-only trials RR 0.74 (0.33 to 1.66); multicomponent exercise in the two military trials RR 0.54 (0.14 to 2.07; I² 87.1%); orthoses without the flat-insole trial RR 0.58 (0.26 to 1.32). Subgroup effects were generally consistent with the overall estimates.
Base rates behind the brace result: patellofemoral pain developed in 37% of unbraced recruits and 58% of unbraced young athletes training 6 h a day; the authors expect a larger NNT where the base rate is lower.
Exercise dose: the two trials with the largest effects gave more than double the weekly minutes and 2-3 quadriceps strengthening exercises; the two without an effect gave 5-10 min of strengthening with one quadriceps exercise.
Orthoses: prefabricated orthoses and shock-absorbing insoles can cause adverse events (mostly minor blisters and arch irritation) in up to 20% of wearers (from one included trial).
Risk of bias: low in 1 of 13 trials (8%); participants and personnel clearly blinded in 2 (15%); outcome assessors blinded in 7 (54%).
Not tested: education ("No study evaluated the effect of education"); braces or any strategy in people who already have knee pain; long-term effects (two trials only).
Mechanisms
Measured: clinically diagnosed new patellofemoral pain only; the review measured no mechanism.
Authors' interpretation: braces may change patellar mechanics, proprioceptive feedback and muscle recruitment, but participants and assessors could not be blinded (skin shaving, brace marks), so a placebo or contextual effect is possible. Soft-landing retraining lowered the vertical loading rate at the end of 2 weeks; whether that lasted through 12 months is unknown.
Authors' interpretation: the authors also write that other overuse injuries (e.g. Achilles tendinitis, shin splints) rose with the new running technique, and cite a mixed-methods review of retraining for the risk of changing technique too quickly; this review gives no numbers for it, and the base's entry for the trial reports fewer running injuries overall with retraining (16% vs 38%, REF-1119, B+).
Inference: NNT 4 for braces is the arithmetic of a 37-58% control incidence over 6-8 weeks of very heavy training; at a recreational runner's much lower base rate the same relative effect would prevent far fewer cases per brace worn.
Practical Application
Will a knee brace stop runner's knee? Perhaps in heavy training: in 2 small trials (recruits, young athletes, 6-8 weeks) braces cut new kneecap pain by 60%, low certainty. No runner trial, nothing on racing with pain.
"A brace for my marathon?" This review is about preventing new pain in people without it; it holds no data on wearing a brace with existing pain or in a race. The authors themselves say widespread prophylactic bracing outside military and intensive-training settings should not be recommended, given low certainty, cost, inconvenience and non-adherence (A-, low certainty).For treating existing kneecap pain, see REF-1381 (taping as a supporting option, graded low; braces not part of that synthesis).
The runner-specific signal is soft landing: one trial of 320 novice runners — 2 weeks of treadmill feedback to land more softly — cut new patellofemoral pain over 12 months (RR 0.21, low certainty; the same trial as REF-1119); it needed an instrumented treadmill, and whether a cue alone works was not tested.
Strength and neuromuscular warm-ups: pooled, no evidence of a difference for kneecap pain (RR 0.49, 0.18 to 1.36), with the bigger effects in the better-dosed programmes; the authors still back appropriately dosed programmes because they are unlikely to harm and may prevent other injuries — their judgement, not this pool's result. In runners the base's pooled prevention trials also show no evidence of a difference (REF-1244, A-).
Insoles and stretching: foot orthoses (RR 0.63, 0.35 to 1.13, very low certainty, all military) and static stretching showed no evidence of a difference; the authors say neither should be advocated to prevent kneecap pain without better evidence. Shoes: no type reliably prevents running injury (REF-716, A).
Intensity vs volume: progressing a running programme by intensity rather than volume did not lower kneecap pain (about 36% higher, not significant; one trial of 447 recreational runners).
What it is good for as a source: what has and has not been shown to prevent new patellofemoral pain, with honest certainty, and a ready answer to the brace question. What it does not show: an effect of braces in runners, any effect on existing pain or during a race, the effect of education, anything beyond 12 months, or which runners would benefit.
Limitations
Low to very low certainty throughout. Risk of bias was low in 1 of 13 trials; few cases in the brace and retraining trials make the estimates imprecise; one new trial could move any pooled result.
Who. About 79% of participants (6498 of 8210, our arithmetic from the trial sizes) were military recruits; only two trials were in runners. The authors say it is unclear whether the findings generalise to other groups such as active adolescents.
Short exposures. Interventions lasted 2 weeks to 8 months and follow-up 6 weeks to 12 months; the short- vs long-term analysis was not possible.
Heterogeneous interventions. Different braces, insoles and exercise doses were pooled; the exercise pool's I² was 64.9%.
Outcome definition. Physician- or physiotherapist-diagnosed patellofemoral pain; some trials used time-loss criteria, which may undercount cases (the authors' point).
Mixed messages from the authors. The abstract's "braces and running technique retraining can reduce the risk ... by 60%-79%" sits beside the discussion's advice that widespread bracing should not be recommended, and the practical section recommends widespread strengthening programmes whose pooled effect was not significant.
Funding and interests: NHMRC (Australia) and CIHR (Canada) fellowships; competing interests none declared. No device maker, federation or data platform is named.
How this entry was checked and changed
Why this level
a meta-analysis of RCTs is the rubric's A design; held at A- because the certainty is low or very low for every intervention, risk of bias was low in only 1 of the 13 trials, the two positive findings rest on two small brace trials and one retraining trial with few events, and 8 of the 13 trials were in military recruits. Read from the full text (the publisher's PDF in the Erasmus University repository, listed by OpenAlex)
Full text (checked 2026-10-28)
Thirteen RCTs involving a total of eight thousand two hundred and ten individuals (mostly military recruits and young athletes); mean age of participants in the individual trials ranged from 13 to 39 years (full text)
Patellofemoral braces worn during physical activity for 6–8 weeks (compared with no brace): RR 0.40, 95% CI 0.22 to 0.73; I2=24.0%; two trials, 227 participants; level of evidence low (full text)
Running retraining, single study of 320 participants: RR 0.21, 95% CI 0.07 to 0.60, 12 months following eight sessions of instantaneous visual feedback to reduce the vertical impact peak (run softer) (full text)
Multicomponent exercise, four trials, 3364 participants: RR 0.49, 95% CI 0.18 to 1.36; I2=64.9% (full text)
Foot orthoses/insoles, four trials, 2314 participants: RR 0.63, 95% CI 0.35 to 1.13; I2=0.0%; very low level of evidence (full text)
NNT: patellofemoral bracing=4, running technique modification=11, multicomponent exercise=60, foot orthoses=14 and stretching=76; intensity-based running programme number needed to harm 207 (full text)
Unbraced incidence 37% in the military and 58% in the intensive exercise (6 hours daily) groups; widespread implementation of bracing to prevent patellofemoral pain should not be recommended (full text)
Risk of bias was low in only 1 (8%) of the 13 included trials; no study evaluated the effect of education (full text)
Funding: AGC National Health and Medical Research Council of Australia Early Career Fellowship (APP1121173); EMM Canadian Institutes of Health Research Banting Postdoctoral Fellowship; competing interests none declared (full text)