Table 3 shows the results of longitudinal regression analyses. Improvement in upper leg muscle strength over the 38-week study period was significantly associated with improvement in all three outcome measures (P < 0.001 for all): NRS pain {B coefficient −2.5 [95% confidence interval (CI) −3.7 to −1.4], meaning that every change of 1 unit of strength was linked to a change of −2.5 units of pain}, WOMAC physical function (−8.8, 95% CI −13.4 to −4.2) and get-up-and-go test (B −1.7, 95% CI −2.4 to −1.0). Change in muscle strength explained 7%, 6% and 12% of the total variance of change in NRS pain, WOMAC physical function and get-up-and-go test, respectively. In additional analyses, in which quadriceps and hamstring strength were analysed separately, improvement in each muscle group was significantly associated with all three outcome measures. Improvement in knee joint proprioception was not associated with any outcome measures.