Heterogeneity In The Effects of Weight Reduction on Pain And Physical Function In Knee Osteoarthritis: A causal forest analysis of the Osteoarthritis Initiative.

Abstract

Objectives: Despite the emphasis on weight control, heterogeneity in response to weight loss among individuals with or at high risk for knee osteoarthritis (OA) remains unclear. This study examined heterogeneity in the associations of weight reduction with pain among overweight and obese individuals using a machine learning approach.

Methods: We analyzed longitudinal data from the Osteoarthritis Initiative, comprising community-dwelling adults aged 45-79 years with a body mass index ≥25 who completed baseline assessments between 2008 and 2010, with a 24-month follow-up (n=3,644). The exposure was a ≥5% reduction in body weight from baseline to 24 months. The primary outcome of this study was the pain subscale of the Western Ontario and McMaster Universities (WOMAC) Osteoarthritis Index.

Results: The average treatment effect (ATE) was estimated using a doubly robust estimator, and treatment effect heterogeneity was assessed by estimating conditional average treatment effects (CATEs) using a causal forest approach. A reduction of ≥5% in body weight was not significantly associated with a lower WOMAC pain score (ATE=-0.25; 95% confidence interval [CI], -0.51 to 0.01). The causal forest analysis demonstrated substantial heterogeneity, with CATE estimates of -0.77 (95% CI, -1.24 to -0.30) for the high-benefit subgroup (top 20% of estimated CATEs) and 0.41 (95% CI, -0.35 to 1.17) for the low-benefit subgroup (bottom 20%).

Discussion: Our findings highlight the potential value of individualized weight-loss strategies and suggest that patient characteristics may help identify those most likely to benefit from weight reduction.