Although low bone mineral density (BMD) predicts fractures, there are postulated sex differences in the fracture "threshold." Some studies demonstrate a higher mean BMD for men with fractures than for women, whereas others note similar absolute risk at the same level of BMD. Our objective was to test the preceding observations in the population-based Canadian Multicentre Osteoporosis Study (CaMOS). We included participants 50+ years of age at baseline. Mean BMD in men was higher than in women among both fracture cases and noncases. Three methods of BMD normalization were compared in age-adjusted Cox proportional hazards models. In a model using the same reference population mean and standard deviation (SD), there were strong effects of age and total-hip BMD for prediction of fractures but no significant effect of sex [hazard ratio (HR) = 0.97, 95% confidence interval (CI) 0.78-1.20] for men versus women. In a model using sex-specific reference means but a common SD, an apparent sex difference emerged (HR = 0.66, 95% CI 0.54-0.81) for men versus women. The sex term in the second model counterbalanced the higher risk introduced by the lower normalized BMD in men. A third model using sex-specific reference means and SDs gave nearly identical results. Parallel results for the three methods of normalization were seen when adjusting for clinical risk factors, excluding antiresorptive users and considering death as a competing risk. We conclude that no adjustment for sex is necessary when using common reference data for both men and women, whereas using sex-specific reference data requires a substantial secondary adjustment for sex.