One thousand and three women aged 45-64 from the Chingford general population survey were studied cross sectionally to find the effect of quantity and distribution of body fat on the prevalence of radiologically confirmed osteoarthritis (OA) in the knee, carpometacarpal (CMC), distal interphalangeal (DIP), and proximal interphalangeal (PIP) joints. Obesity was classified as the upper tertile of body mass index (BMI kg/m2); the boundaries of the middle tertile were 23.4 and 26.4 kg/m2. The age adjusted odds ratio (OR) [and 95% confidence interval (CI)] of radiographic OA at the knee comparing the high and low tertile of BMI was 6.17 (3.26-11.71) and for bilateral knee radiographic OA was 17.99 (6.25-51.73). Comparing the middle and low tertile of BMI, the odds ratio for radiographic OA knee was 2.86 (1.44-5.68). For other joints the association between BMI and radiographic OA was less strong; the OR at CMC was 1.71 (1.05-2.78), at DIP was 1.52 (0.90-2.57), and at PIP was 1.23 (0.52-2.91). For all joints except PIP these OR increased if the diagnostic criteria included knee pain for at least a month, clinically evident swelling at the DIP or PIP, and pain or tenderness at the CMC. Recalled weight at age 20 years, or recalled maximum weight improved prediction of radiographic OA from current BMI, but measurement of fat distribution from circumference of waist, hip and thigh did not. Our results confirm that excess body weight is a powerful predictor of OA of the knee in middle aged women, and a modest predictor of DIP and CMC OA.(ABSTRACT TRUNCATED AT 250 WORDS)