Every guideline recommends exercise for knee osteoarthritis and none says how much. Patients are told to strengthen the quadriceps and left to guess. This dose-response meta-analysis puts numbers on it.
It pooled 20 randomised trials of isolated isotonic lower-limb resistance training in 791 people with knee osteoarthritis, excluding trials that combined it with other exercise. Dose-response relationships were modelled using restricted cubic splines, with risk of bias by Cochrane RoB 2.0 and certainty by GRADE.
The relationships were non-linear, with identifiable peaks. Pain reduction was greatest at around 6,000 total repetitions (mean difference -3.52, 95% CI -4.08 to -2.96). Functional improvement peaked at around 2,600 repetitions at moderate intensity (-16.66, -18.80 to -14.51). Strength gains peaked at around 1,600 repetitions (standardised mean difference 0.12, 0.01 to 0.22). Very light to moderate intensity gave the best pain reduction, moderate intensity the best function, and moderate to vigorous similar strength gains. Both home and centre-based programmes worked, with centre-based more effective.
The practical translation is a prescription rather than an exhortation. Roughly 2,600 to 6,000 repetitions accumulated over a programme means something like three sessions a week of three exercises at three sets of ten to twelve, sustained for a few months — a number a patient can be given and can track.
The honest caveat is the certainty rating, which is low to very low, and it applies to the whole finding. These are estimated optima from spline models across heterogeneous trials, not thresholds established by trial.
- Prescribe resistance exercise in repetitions and sessions, not as generic advice to strengthen the quadriceps
- Aim for moderate intensity, which gave the best functional gains and near-best pain reduction
- Expect pain benefit to accumulate over thousands of repetitions — this is a months-long programme, not weeks
- Offer centre-based supervision where available; home programmes worked but less well
- Treat the specific repetition figures as estimates from low-certainty evidence, not as thresholds
The statistics, in plain English
Restricted cubic splines model the relationship between dose and effect as a smooth curve rather than assuming it is a straight line, which is why the analysis can identify a peak rather than just a direction. The peaks are estimates from the curve, and a curve fitted across 20 heterogeneous trials will place its peak with much less precision than the narrow confidence intervals suggest. The GRADE rating of low to very low certainty is the number to take seriously: it means the true effect may be substantially different from the estimate. The strength result, a standardised mean difference of 0.12 with an interval reaching 0.01, is trivially small and should not be quoted as a benefit.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for orthopaedics, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free