Physical activity decreases the risk of osteoarthritis (OA)-related disability; however, pain and lack of confidence represent barriers for older adults with knee OA. The purpose of this study was to examine (a) the baseline associations among self-efficacy and physical activity, function, and pain; (b) longitudinal changes in self-efficacy; and (c) whether self-efficacy mediates treatment effects on clinical outcomes. The Intensive Diet and Exercise for Arthritis (IDEA) trial was a single-blind, randomized controlled 18-month study including 454 overweight/obese older adults (M age = 66 years) with knee OA. Participants were randomized to one of three interventions: exercise (E), diet-induced weight loss (D), or both (D+E). Self-efficacy for gait, balance, and walking duration were assessed at baseline, 6 months, and 18 months. Baseline associations were tested using Pearson correlations, and group least squares means were compared using mixed linear models at follow-up. Participants with higher self-efficacy reported significantly better physical function and less knee pain at baseline, walked farther (6-min walk), and were more physically active (all |r| > 0.12, all p < .01). Significant differences between groups were detected for all self-efficacy measures at 18 months; the D+E group reported significantly (all p < .005) higher self-efficacy for gait, walking duration, and balance compared with the D- or E-only groups. Self-efficacy significantly (p < .05) mediated treatment effects on physical function and pain at 18 months. A combined intervention of diet-induced weight loss and exercise is the treatment of choice to maximize self-efficacy, improve physical function, and reduce pain in overweight/obese adults with knee OA.