Relationship between bone mass and rates of bone change at appendicular measurement sites.

The rate of bone change among postmenopausal women may vary depending upon the initial bone mass. Examining this possibility is difficult, however, because of a negative statistical bias that occurs when change is regressed against the initial value of the same variable. In this article, four statistical methods were applied to measure the association between bone mass and the rate of bone change. The study population was Japanese-American women, who were monitored for approximately 5 years. Bone changes were determined for the calcaneus and the distal and proximal radius. The results were consistent across the bone sites but differed between statistical methods. Three of the four methods indicated that the women with the greater bone mass had the greater loss rates. The fourth method did not support this association. Possible reasons for the discordant results are discussed. Using the "best" estimate of the relationship, a gradual convergence of bone mass was projected over time toward the population mean. The convergence occurred because women with higher bone mass had a somewhat faster loss rate than women with lower bone mass. Overall, however, the variation in bone mass between individuals was large compared to the rate of convergence.