Samenvatting

Objective To investigate whether duration of knee symptoms influenced the magnitude of the effect of exercise therapy compared to non-exercise control interventions on pain and physical function in people with knee osteoarthritis (OA). Method We undertook an individual participant data (IPD) meta-analysis utilising IPD stored within the OA Trial Bank from randomised controlled trials (RCTs) comparing exercise to non-exercise control interventions among people with knee OA. IPD from RCTs were analysed to determine the treatment effect by considering both study-level and individual-level covariates in the multilevel regression model. To estimate the interaction effect (i.e., treatment x duration of symptoms (dichotomised)), on self-reported pain or physical function (standardised to 0–100 scale), a one-stage multilevel regression model was applied. Results We included IPD from 1767 participants with knee OA from 10 RCTs. Significant interaction effects between the study arm and symptom duration (≤1 year vs >1 year, and ≤2 years vs>2 years) were found for short- (∼3 months) (Mean Difference (MD) −3.57, 95%CI −6.76 to −0.38 and −4.12, 95% CI-6.58 to −1.66, respectively) and long-term (∼12 months) pain outcomes (MD −8.33, 95%CI −12.51 to −4.15 and −8.00, 95%CI −11.21 to −4.80, respectively), and long-term function outcomes (MD −5.46, 95%CI −9.22 to −1.70 and −4.56 95%CI −7.33 to-1.80, respectively). Conclusions This IPD meta-analysis demonstrated that people with a relatively short symptom duration benefit more from therapeutic exercise than those with a longer symptom duration. Therefore, there seems to be a window of opportunity to target therapeutic exercise in knee OA.