Objective: To investigate the relationships between body mass index (BMI), physical activity and erectile dysfunction (ED). Design and subjects: A population representative cross-sectional analytic study of ED in Hong Kong, with two-stage stratified random sampling, and face-to-face interviews conducted by trained interviewers with structured questionnaires. Study subjects were 1506 men aged 26-70. Results: Age (odds ratio (OR) ¼ 1.30; 95% confidence interval (CI) 1.20-1.40), physical activity (OR ¼ 0.91 per 500 kcal/week; 95% CI 0.84-0.98), and general psychological distress (OR ¼ 1.03 per GHQ score; 95% CI 1.00-1.06) were independently associated with ED after multivariate adjustments. An U-shaped relationship between BMI and ED was observed only among men with no exercise (oonce/week): BMI o18.5 (OR ¼ 2.99; 95% CI 1.01-8.86), 18.5-19.9 (OR ¼ 2.66; 95% CI 1.04-6.79), 20.0-20.9 (OR ¼ 1.37; 95% CI 0.49-3.79), 22.0-22.9 (OR ¼ 1.36; 95% CI 0.58-3.17), 23.0-24.9 (OR ¼ 1.66; 95% CI 0.70-3.93), X25.0 (OR ¼ 2.47; 95% CI 1.08-5.67) using BMI 21.0-21.9 as reference, adjusted for age, GHQ and smoking status. Being physically active (X1000 kcal/week) only reduced the risk of ED (OR ¼ 0.40, 95% CI 0.16-0.95) in men who were obese, adjusted for age, GHQ, smoking status and BMI. Conclusions: BMI and physical activity independently and differentially affected ED risk. BMI had greatest influence with low physical activity, and physical activity exerted greatest influence when BMI was high. This is the first study to demonstrate an U-shaped relationship between BMI and ED risk, but only in men with no exercise, and to identify underweight as a risk factor for ED. This relationship has clinical implications for obese as well as underweight individuals.