In an age-matched, case-control study, we investigated the association between endogenous sex steroid hormones and incident vertebral fractures in both elderly men and women (aged 67.7 +/- 6.8 yr). Drawn from the Rotterdam Study, participants required radiographs of the lumbar spine at both baseline and follow-up (average time of follow-up, 6.5 yr) and frozen blood samples, taken at baseline. One hundred and seventy-eight men (45 cases) and 454 women (115 cases) were thus selected. Serum estradiol, SHBG, testosterone, and insulin were measured, along with bone mineral density at both spine and hip. Women in the lowest tertile of serum estradiol (< or =15.5 pmol/liter) had a 2.1 times increased risk (95% confidence interval, 1.3-3.5) of incident vertebral fractures, independently of bone mineral density measured at either site. SHBG levels in the lowest two tertiles were associated with a 50% reduction in incident vertebral fracture risk. Women with a combination of both low estradiol and high SHBG had a 7.8 times higher risk of an incident vertebral fracture (95% confidence interval, 2.7-22.5; P < 0.001), adjusted for age and weight. This increased risk did not change when non-SHBG-bound estradiol was used instead of total estradiol. For men, no clear association was found, possibly due to insufficient power. No clear association between testosterone and incident vertebral fractures was observed in either men or women.