The aim of this retrospective cohort study was to estimate the changes in bone mineral density (BMD) as a consequence of exercise in female ex-athletes and age-matched controls. Eighty-three ex-elite female athletes (67 middle and long distance runners, 16 tennis players, currently aged 40-65) were recruited from the original records of their sporting associations. Controls were 585 age-matched females. The main outcome measures were BMD of lumbar spine (LS), femoral neck (FN), and forearm, estimated by dual-energy X-ray absorptiometry (DXA) scan. Levels of physical activity were assessed using a modified Allied Dunbar Fitness Survey scale and classified as (a) ex-athletes, (b) active controls (> or = 1 h of vigorous physical activity currently and in the past), (c) low activity controls with inconsistent or intermediate levels of activity, and (d) inactive controls (< 15 minutes of exercise per week). After adjustment for differences in age, weight, height, and smoking, athletes had greater BMDs than controls: 8.7% at the LS (95% confidence interval [CI] 5.4-12.0; p < 0.001) and 12.1% at FN (CI 9.0-15.3; p < 0.001). The benefits of exercise appeared to persist after cessation of sporting activity. Active controls (n = 22) had greater BMDs than the inactive group (n = 347): 7.9% LS (CI 2.0-13.8; p = 0.009) and 8.3% FN (CI 2.7-13.8; p = 0.004). The low activity controls (n = 216) had an intermediate BMD. Tennis players had greater BMDs compared with runners: 12.0% LS (CI 5.7-18.2; p = 0.0004) and 6.5% FN (CI -0.2-13.2; p = 0.066). The BMD of tennis players' dominant forearms were greater than their nondominant forearms. In conclusion, regular vigorous weight-bearing exercise of 1 h or more per week is associated with an increase in BMD within a normal population. This study confirms long-term weight-bearing exercise as an important factor in the regulation of bone mass and fracture prevention.