How to Interpret Surrogate Markers of Efficacy in Osteoporosis


We appreciate the opportunity to respond to the letter of Drs Eastell and Delmas regarding our recent paper.(1) We agree that it is important to compare the efficacy of osteoporosis treatments in head to head studies such as the one we recently conducted.(1,2) We also agree that new information regarding the relationship between changes in BMD and bone turnover with fracture risk reduction has emerged in the last few years,(3–5) leading some to conclude that “changes in BMD or markers of bone turnover can be used (in the conduct of clinical trials) as supportive evidence of the effectiveness of treatment.(6)

The first issue raised is that changes in bone turnover are at least as informative as changes in BMD. We concur, and that is why changes in biochemical markers of bone turnover were an important secondary efficacy endpoint. Whereas the primary efficacy endpoint was the change in trochanter BMD at 12 months (for reasons noted in the paper), changes in four different biochemical markers of bone turnover measured at three different time-points (3, 6, and 12 months) were reported as secondary efficacy endpoints and discussed with equal prominence to the BMD changes. For all four markers, at all time-points measured, the reduction in the level of the marker from baseline was statistically greater with alendronate than with risedronate.

The second point raised is that the relationship between changes in BMD and markers and fracture risk reduction may be nonlinear. Eastell et al.(7) reported that reductions in turnover during treatment with risedronate were associated with reductions in the risk of vertebral fractures, but appeared to reach a plateau beyond which fracture risk did not further decline. Data from a much larger study with alendronate reported no evidence of such a plateau over a wider range of reductions in turnover markers.(8) Moreover, in both publications, there was no evidence of a plateau or nonlinearity in the relationship between reduction in turnover markers and reduction in risk of nonvertebral fractures. It was also suggested that we did not evaluate cut-points for changes in bone turnover markers. These results are presented in a separate publication.(9)

Analyses of data from the FIT and VERT studies are not comparable in design and did not reach the same conclusion as suggested by Drs Eastell and Delmas.(3,10) Although both studies addressed the issue of the relationship between incidence of spine fracture and changes in spine BMD on treatment, neither addressed the relationship between nonspine or hip fracture risk and changes in BMD. Thus, whereas we have acquired new information about the correlates of fracture reduction with antiresorptive treatments, the exact nature of those relationships are not yet clearly understood.

In response to the comment regarding the number of fractures, it is important to recall that fractures were captured as adverse events. The fractures were self-reported, included any fracture (fingers, toes, face, etc.) regardless of location or cause, and were not verified by review of radiographic reports. In addition, this represents the total number of fractures, not the number of patients with fracture. As stated in the paper, there was no significant difference in the percent of patients reporting fractures as adverse events between treatment groups. We are certain that Drs Delmas and Eastell would agree that it is not prudent to make inferences from unadjudicated and underpowered fracture studies. Furthermore, in the recently completed FACTs-International study of identical design, 18 fractures occurred in the alendronate treatment group and 20 fractures occurred in the risedronate treatment group. These results will be presented later this year, but serve to underscore the difficulty in interpreting fracture data from clinical trials not designed to detect differences in fracture incidence.

There was also a question raised as to whether some of the women in the FACT trial may have had “oversuppression” of bone turnover. Although this is an important theoretical concern, it is not supported by scientific or clinical evidence presented in recent scientific meetings.(11–13) For example, long-term safety data with alendronate treatment for up to 10 years of continuous use has shown no evidence of an adverse cumulative effect or loss of antifracture efficacy, even among the women with the lowest indices of bone turnover.(11,12) Similarly, double tetracycline label was found in all bone biopsies obtained after 10 years of therapy, with no histomorphometric evidence of adynamic bone.(13) Additionally, in preliminary comparisons of the bone marker changes to the 95% reference range for urinary NTx in 107 healthy premenopausal women (at the same central laboratory used in our study), there was no significant difference in the proportion of patients who fell below the lower limit of the premenopausal range with alendronate (3.6%) compared with risedronate (3.3%) after 3 months (unpublished data, Merck & Co).

We acknowledge that calcium and vitamin D administration can affect markers of bone turnover. However, calcium and vitamin D supplementation was started at the randomization visit, after the baseline markers had been obtained.

We hope these comments help clarify the issues raised by Drs Eastell and Delmas and further illustrate the possible implications of the differences in reductions in bone turnover markers and increases in BMD seen between women treated with alendronate and risedronate.