The objective of this study was to examine the relation of endogenous sex hormones to subsequent height loss in postmenopausal women, in whom height loss is usually a surrogate for osteoporotic vertebral fractures. This was a prospective, community-based study. The site chosen was Rancho Bernardo, an upper middle class community in Southern California. A total of 170 postmenopausal women participated, aged 55-80 years. None of them were taking exogenous estrogen between 1972 and 1974. Plasma was obtained for sex hormone and sex hormone-binding globulin (SHBG) assays. Estradiol/SHBG and testosterone/SHBG ratios were used to estimate biologically available hormone levels; bioavailable (non-SHBG-bound) testosterone was measured directly in 60 women. Height loss was based on height measurements taken 16 years apart. Height loss was strongly correlated with age (p = 0.001). These women lost an average 0.22 cm/year in height. Neither estrone nor estradiol levels were significantly and independently related to height loss. Both estimated bioavailable testosterone (testosterone/SHBG ratio) and measured bioavailable testosterone levels predicted future height loss (p = 0.02 and 0.08, respectively) independent of age, obesity, cigarette smoking, alcohol intake, and use of thiazides and estrogen. We conclude that bioavailable testosterone is an independent predictor of height loss in elderly postmenopausal women. The reduced height loss is compatible with a direct effect of testosterone on bone mineral density or bone remodeling.