How to Interpret Surrogate Markers of Efficacy in Osteoporosis


To the Editor:

Rosen et al.(1) have made the important point that it is important to compare the efficacy of treatments for osteoporosis in head-to-head studies. They claim that this cannot be done with fracture as the primary endpoint, because it would require samples sizes of around 50,000. The alternative is to compare changes in BMD and bone turnover markers. This is an attractive alternative, but we have learned a lot about the use of these surrogates in the last few years, and so the interpretation of this paper needs further consideration. The first issue is that the change in bone turnover markers is at least as informative as the change in BMD.(2) The authors have focused their analyses on change in BMD. The second is that the relationship between change in these surrogates and fracture risk is not linear; for BMD, a gain of up to 3% is associated with the same fracture risk reduction as a gain of >3%(3,4); for bone turnover markers, there may be no further benefit from further suppression of NTx/Cr <40%.(2) The authors have evaluated various BMD cut-points but not the cut-points in bone turnover markers. The third issue is that suppression of bone turnover with antiresorptive therapy may result in the accumulation of microcracks and a decrease in the toughness of bone.(5) The fourth issue is that, although alendronate showed significant increases in BMD and significant decreases in bone turnover markers compared with risedronate, the number of fractures observed were greater (26 fractures for alendronate versus 20 fractures for risedronate) and is in contrast with the hypotheses by Hochberg et al.(6)

We examined the baseline and follow-up study of our own publication and expressed the bone turnover marker in relation to the premenopausal reference interval. The latter requires logarithmic transformation of the NTx/Cr results. We noted that, before therapy, the majority of patients from the VERT study had results above this interval or were in the upper one-half of the reference interval (less than T score = −2, between T score = −2 and 0, between T score = 0 and 2, and more than T score = 2 were 0%, 11%, 62%, and 27%, respectively; Fig. 1). The results obtained after 3-6 months of risedronate (5 mg/day) and calcium (1000 mg/day) indicate that the results fall mainly into the lower one-half of the reference interval, although there are a few patients with values below the reference interval and a few with values in the upper one-half of the reference interval (less than T score = −2, between T score = −2 and 0, between T score = 0 and 2, and more than T score = 2 were 7%, 63%, 28%, and 2%, respectively).

Figure FIG. 1..

The levels of the bone resorption marker NTx/Cr after logarithmic transformation and in actual units (nmol BCE/mmol creatinine) in all women (A) at baseline and (B) after 3-6 months. The light shaded area represents the lower one-half and the dark shaded area represents the upper one-half of the premenopausal reference interval.

We would like to know what proportion of women in the FACT study after 3-6 months had values below the reference interval for premenopausal women (17 nmol BCE/mmol creatinine in our study). These subjects may have “oversuppression” of bone turnover. Furthermore, we would like to know whether the calcium and vitamin D was commenced before the initial samples for bone turnover markers. Such a supplementation usually decreases bone resorption markers by 10-20% within a few weeks and, if so, would explain why the baseline levels of urinary NTx are rather low in the FACT study.