A number of recent studies have suggested that non-invasive measures of bone turnover are associated with bone loss at the forearm in postmenopausal women. Whether bone turnover markers are predictive of bone loss from the clinically important sites of lumbar spine and femoral neck remain unclear, and was the aim of this 4-year prospective study. One hundred and forty-one normal, postmenopausal women (mean age 52.0 +/- 3.3 years, mean menopause duration 20.4 +/- 5.7 months) were recruited for the study in 1988. Fasting early morning samples of blood and urine were collected at the baseline visit and stored at -20 degrees C prior to analysis. Serum was assayed for osteocalcin, oestradiol, oestrone, oestrone sulphate, testosterone, sex hormone binding globulin, dehydroepiandrosterone sulphate and total alkaline phosphatase. Urine was assayed for calcium, hydroxyproline, oestrone glucuronide and the collagen cross-links pyridinoline and deoxypyridinoline using high-performance liquid chromatography. Bone density was measured at the lumbar spine and femoral neck using dual photon absorptiometry at time 0, 12, 24 and 48 months. The mean annual percentage change in bone density (SE) was -1.41% (0.18) at the lumbar spine and -0.86% (0.22) at the femoral neck. There was no evidence of bimodality or a fast loser subgroup as the rates of change were normally distributed. Both simple and multiple stepwise regression analyses revealed no significant correlation between the rates of change in bone density with any biochemical marker, either individually or in combination, despite the study having sufficient power (80%) to detect a correlation of 0.5 between any biochemical marker levels and bone loss. We conclude that single measurements of these markers of bone turnover and endogenous sex hormones appear unlikely to be clinically useful in predicting early postmenopausal bone loss from either the spine or the hip.