Regression to the mean for physical function and quality of life in clinical trials for symptomatic knee osteoarthritis.

Abstract

Objective: To estimate the size of regression to the mean (RTM) for common patient-reported outcomes in trials for knee osteoarthritis (OA).

Design: Longitudinal cohort study; we included participants of the Osteoarthritis Initiative who fulfilled typical inclusion criteria for enrolment in a trial. These included: age 40-79 years, symptomatic knee OA, Kellgren-Lawrence grade 2-3, use of pain medication more than half the days of a month in past 12 months, numerical rating scale pain of 4 to 9. We studied observed changes in WOMAC physical function and KOOS quality of life (QOL).

Results: We identified 547 subjects who fulfilled inclusion criteria on at least one annual follow-up between year 1 and year 8. The mean level of physical function and QOL at each follow-up time point was similar, about 18 and about 51, respectively. However, at the time of theoretical inclusion in a trial, the mean levels in the same subjects were 23 and 43, respectively (both worse scores). The mean improvement in physical function between inclusion and 1 and 2 years later, respectively, was 2.5 (95% confidence interval 1.7 to 3.2) and 3.1 (2.3 to 3.8). The corresponding improvement in QOL was 2.7 (1.7 to 3.7) and 4.2 (3.1 to 5.3).

Conclusion: RTM in trials for knee OA is likely to explain improvement in physical function and QOL, not only in knee pain. RTM often misleads investigators to overinterpret effectiveness as RTM neither represents improvement from the intervention nor placebo effect from the intervention and its context.