Controversy exists about rates of bone mineral loss from the lumbar spine at the menopause. Cross-sectional studies of lumbar bone mineral against age have been interpreted as evidence for or against a menopause-related change in lumbar bone loss, depending on their goodness of fit to either linear or nonlinear equations. To investigate the ability of cross-sectional studies to detect accelerated lumbar bone loss at the menopause, we constructed models of different patterns of loss and evaluated the findings in cross-sectional analyses of computer-simulated populations of normal women. A sample size of greater than 300 was needed to distinguish between linear and nonlinear patterns of bone loss with a power of 90% when the data was analyzed throughout life. Examining for differences between the slopes of the regressions of pre- and postmenopausal women was more useful for detecting nonlinear bone loss under some circumstances. A difference between slopes was apparent in studies containing 100 subjects if lumbar bone density (BMD) was assumed to be unchanged prior to the menopause, but a larger study size was needed (greater than 1000) if BMD was assumed to fall before the menopause. We also measured lumbar BMD by dual photon absorptiometry in 141 normal females. When the data was analyzed against age throughout life, a nonlinear pattern of bone loss was found, and comparison of pre- and postmenopausal subjects showed a significant difference in the slopes of the linear regressions. Our patient data support the concept of a relatively stable lumbar BMD prior to the menopause with a rapid but transient decline postmenopausally.(ABSTRACT TRUNCATED AT 250 WORDS)