This longitudinal study included 288 postmenopausal women without estrogen use (median age, 72 yr) and 352 men (median age, 66 yr). All were community-dwelling, ambulatory, and Caucasian. Blood for hormone assays (total and bioavailable estradiol and testosterone, estrone, androstenedione, dihydrotestosterone, dehydroepiandrosterone, and dehydroepiandrosterone sulfate) was obtained in 1984-1987, and vertebral fractures were diagnosed from lateral spine radiographs obtained in 1992-1996. At least one vertebral fracture was found in 21% of women and 8% of men. Among men, age-adjusted hormone levels differed by fracture status only for total (64.1 vs. 75.4 pmol/L, P = 0.012) and bioavailable (43.0 vs. 51.4 pmol/L, P = 0.008) estradiol. There was a graded association between higher concentrations of total and bioavailable estradiol and lower fracture prevalence (trend P<0.01 for both hormones). Men with total testosterone levels compatible with hypogonadism (<7 nmol/L) were not more likely to have vertebral fractures. In women, none of the measured sex hormones was associated with vertebral fractures. There was also no increased prevalence of fractures in women with estradiol levels below the assay sensitivity (<11 pmol/L). These data suggest that estrogen plays a critical role in the skeletal health of older men and confirm other studies showing no association of postmenopausal endogenous estrogen levels with vertebral fractures in older women.