Changes in Biochemical Markers of Bone Turnover in Women Treated with Raloxifene: Influence of Regression to the Mean

Abstract:

Measures with extreme magnitude are most likely to be the result of measurement variability. Repeated measurements genuinely lessen such variability, leading to a phenomenon known as regression to the mean (RTM), which may affect biochemical markers of bone turnover. We therefore studied four markers of bone turnover in the Multiple Outcomes of Raloxifene Evaluation (MORE) trial – serum procollagen type I C-propeptide (PICP), osteocalcin (OC), bone-specific alkaline phosphatase (BAP) and urinary type I collagen breakdown product (CTX) – among the 1704 women treated with raloxifene who had marker measurements and were at least 70% adherent, and among 915 control group patients. We examined the existence of RTM, and applied a method of adjustment for RTM of both baseline and follow-up results. We found that women who had the most extreme values tended to go in the opposite direction with the subsequent measurement, i.e., exhibited a pattern of RTM. For example, among women whose urinary CTX decreased at least 60% in the first 6 months, 61% had an increase in the next 6 months; and among those who had an increase in the first 6 months, 81% had a decrease in the next 6 months. We found a similar pattern for each of the four markers. When adjusting for RTM we obtained estimated true values of both baseline values and change in markers. These estimated true values were substantially different from the observed value when the latter was further from the mean. For example, for a 10% increase in urinary CTX in the first 6 months, after accounting for RTM we estimate that there was in fact a 3% decrease (80% confidence interval: –38% to 53%). We conclude that the few initial extreme marker responses observed in women treated with raloxifene represent RTM, and that one practical consequence is that patients with an increase in markers during the first 6 months should be continued on raloxifene therapy, because the values usually decrease later on.