CURRENTLY, THE MOST widely used endpoint in clinical trials for new osteoporosis drugs is vertebral fracture. This is the most common fragility-related fracture in postmenopausal women, and because hip fractures occur at about one-third the annual rate of vertebral fractures,1 most trials have difficulty showing antifracture efficacy at the hip. Even for assessing efficacy in reducing the risk of vertebral fractures, however, trials must include large numbers of patients and confirm vertebral fractures by radiological analysis. Use of a surrogate marker for fracture, such as bone mineral density (BMD), would vastly reduce the number of subjects needed for these studies. This would both reduce the cost of these studies and also potentially accelerate the pace at which useful new drugs are brought to the market. Thus, for example, effects of the selective estrogen receptor modulator (SERM), raloxifene, were apparent in a study of 601 postmenopausal women2; the comparable study designed to show efficacy of raloxifene in reducing vertebral fractures required 7705 women.3 There is now little doubt that, in the population, BMD is an excellent predictor of fracture risk. In a meta-analysis of prospective cohort studies published between 1985 and 1994, Marshall et al.4 concluded that, depending on the site measured and the type of osteoporotic fracture being predicted, each SD decrease in BMD resulted in an approximately 1.5- to 2.5-fold increase in fracture risk (Table 1). Moreover, the predictive ability of BMD was roughly similar to or better than that of a 1 SD increase in blood pressure for stroke and better than a 1 SD increase in serum cholesterol level for cardiovascular disease. The main problem with BMD as a surrogate marker for the risk of fracture came from studies with sodium fluoride, in which marked (approximately 33%) increases in spine BMD were not associated with any significant reductions in vertebral fracture risk.5, 6 Lafage et al.7 subsequently demonstrated that, in minipigs, fluoride treatment resulted in the formation of bone with reduced biomechanical competence. This occurred despite the use of relatively modest doses of sodium fluoride in these animals and without the presence of osteomalacia or other histological changes. Thus, fluoride may reduce the strength of bone as a result of its effects on crystal structure or matrix properties. The experience with sodium fluoride clearly led to a pendulum swing in favor of requiring fracture endpoints for new osteoporosis drugs. The question at this point is whether we have gone too far with this stringent requirement. Thus, if in animal models, the relationship between bone mass and strength is preserved for a therapeutic agent, can BMD be used as a surrogate marker for fracture risk? This question needs to be examined separately for antiresorptive and formation-stimulating drugs because the mechanism(s) of the reduction in fracture risk seems to be quite different for the two classes of agents. In both instances, however, the evidence seems to be that changes in BMD do predict (albeit to a variable extent) reductions in fracture risk. Hochberg et al.8 examined this issue for the antiresorptive agent, alendronate, through a re-analysis of data from the Fracture Intervention Trial. They found that women with the largest increases in BMD also had the greatest reductions in vertebral fractures (Fig. 1), indicating that at least among women taking alendronate, greater increases in BMD are associated with lower risk of new vertebral fractures. This issue was examined further in a meta-analysis of data from clinical trials of antiresorptive agents by Wasnich and Miller.9 As shown in Fig. 2, in this analysis, the changes in BMD were generally related to the reduction in vertebral fracture risk seen with these agents, with the drugs resulting in the largest increases in BMD also reducing vertebral fractures to the greatest extent. Similar findings were reported subsequently by Hochberg et al.10 for changes in BMD with antiresorptive agents predicting the reduction in the risk of nonvertebral fractures. Cumulative incidence of vertebral fractures, stratified by tertiles of baseline BMD and categories of change in spine BMD (no gain, <3% gain, ≥3% gain) at 24 months. Adapted from Hochberg et al.8 with permission. Relative risk of new vertebral fracture vs. change in spine BMD (vs. placebo) for randomized controlled trials of antiresorptive agents. A, alendronate; C, calcitonin; E, etidronate; H, hormone replacement (estrogen); R, raloxifene; T, tiludronate. Adapted from Wasnich et al.9 with permission. The major caveat to these findings, however, is that the magnitude of the reduction in vertebral fracture risk with antiresorptive drugs tends to be considerably greater than what would be predicted by the relatively modest changes in BMD induced by many of these agents. Black et al.11 for example, analyzed the results of 13 randomized trials of antiresorptive drugs between 1966 and 1999 and found that (on average) each 1% improvement in spine BMD corresponded to a 3.3% (95% CI, 1.0–5.6%, p = 0.01) decrease in vertebral fracture risk. However, the magnitudes of the observed fracture risk reductions were at least twice as large as would be expected (based on observational data) from the changes in BMD. Even more surprising were the findings of Sarkar et al.12 in their re-analysis of data from the Multiple Outcomes of Raloxifene Evaluation (MORE) trial. In this analysis, for any percentage change (either increase or decrease) in femoral neck or lumbar spine BMD, raloxifene-treated patients had a statistically significantly lower vertebral fracture risk compared with placebo-treated patients. Collectively, these findings suggest that while changes in BMD with antiresorptive drugs are, on average, related to decreases in fracture risk, clearly these agents are having effects on fracture risk through a mechanism(s) independent of BMD. While there are several possibilities for these effects, perhaps most convincing are the data indicating that the reduction in bone turnover by these agents is decreasing fracture-risk-independent of changes in BMD. There is accumulating observational data indicating that, in the population, increased bone turnover is an independent predictor of the risk of fracture. In a prospective study of 7598 healthy French women over the age of 75 years, Garnero et al.13 found that the urinary excretion of the bone resorption markers, type I collagen cross-lined C-telopeptide (CTx) and free deoxypyridinoline (free D-Pyr), above the upper limit of the premenopausal range was associated with an increased hip fracture risk with odds ratios (95% CI) of 2.2 (1.3–3.6) and 1.9 (1.1–3.2), respectively. The effects of the bone resorption markers were independent of BMD, and the combination of a femoral BMD value of 2.5 SD or more below the mean of young adults and either high CTx or high free D-Pyr resulted in marked increases in fracture risk, with odds ratios of 4.8 and 4.1, respectively. Similarly, Melton et al.14 found that among 213 postmenopausal women recruited from an age-stratified random sample of Rochester, Minnesota women, a history of osteoporotic fractures of the hip, spine, or distal forearm was associated with reduced hip BMD and with elevated free pyridinoline (Pyd) excretion. Several mechanisms have been proposed for the adverse effects of high bone turnover on the skeleton. These include an increase in the rate of bone loss, as well as accelerated microarchitectural damage to bone, with perforation of plates in cancellous bone and increased porosity of cortical bone due to increased numbers of resorption spaces. Given these independent effects of increased bone turnover on fracture risk, it is not surprising that agents that reduce bone turnover (i.e., the antiresorptive drugs) would reduce fracture risk independent of their effects on BMD. This is perhaps best demonstrated by the analysis of Riggs et al.15 (Fig. 3), which shows vertebral fracture rates as a function of BMD and histomorphometrically determined bone formation rates (as an index of bone turnover) in the placebo and estrogen treatment limbs of a clinical trial of transdermal estrogen.16 As is evident in the placebo group, vertebral fracture rate increased as a function both of decreases in BMD and of increases in bone formation rates. With estrogen treatment, however, the component of vertebral fractures related to high bone turnover was eliminated, whereas the relationship with BMD remained. Computer-generated, three-dimensional surface plot of data from a 1-year randomized controlled trial in postmenopausal women with osteoporosis reported by Lufkin et al.,16 comparing results of placebo treatment and treatment with transdermal estrogen. The two plots show the relationship among vertebral fracture rate (VFR), lumbar spine BMD (LS-BMD), and bone formation rate (BFR) assessed by double tetracycline labeling of an iliac biopsy sample. Note that there are two fracture peaks in the placebo treatment limb: one associated with high bone turnover and one associated with low BMD. By contrast, in the estrogen treatment limb, the fracture peak associated with high bone turnover has been eliminated, whereas the fracture peak associated with the low BMD is still present. Adapted from Riggs et al.15 with permission. Similar findings were reported by Bjarnason et al.17 in their re-analysis of the data from the MORE trial. They found that changes in bone turnover markers after treatment with raloxifene were significantly related to the reduction in vertebral fracture risk, even after adjusting for baseline vertebral fracture status and BMD. In addition, Hochberg et al.10 showed in their meta-analysis of trials of antiresorptive drugs, that the change in resorption markers induced by these agents was related to the reduction in nonvertebral fracture risk. Specifically, they found that, on average, an agent that reduced bone resorption by 70% would decrease the risk of nonvertebral fractures by 40%, independent of effects on BMD. Finally, Eastell et al.18 performed an analysis of the risedronate trials and found that the reduction in the bone resorption markers, urinary N-telopeptide of type I collagen (NTx) and CTx, accounted for up to two-thirds of the reduction in vertebral fracture risk with this agent, with changes in BMD accounting for approximately one-third of the effect. Collectively, therefore, it is clear that changes in bone turnover, in addition to changes in BMD, need to be considered in any estimates of the antifracture efficacy of an antiresorptive agent. There is, at present, considerably less data on the relationship between changes in BMD and fracture risk reduction with formation stimulating agents. The situation with sodium fluoride has been reviewed above. The only other agent that has undergone extensive evaluation is parathyroid hormone (PTH), where approximately 9–13% increases (relative to placebo) in spine BMD resulted in 65–69% reductions in the risk of vertebral fractures.19 Clearly, data from additional studies with anabolic drugs is needed to better define the relationship between changes in BMD and fracture risk reduction with these agents, as has now been done reasonably extensively for antiresorptive drugs. In summary, while vertebral fractures have been the traditional endpoints in clinical trials of new drugs for osteoporosis, it may be time to reconsider this stringent requirement. BMD is a robust predictor of fracture risk in the population, and a case can be made that, assuming animal models indicate that a therapeutic agent is associated with normal bone (histologically and biomechanically), BMD could be used as a surrogate marker for fracture. The caveat to this, however, is that with antiresorptive agents, changes in BMD may significantly underestimate the reduction in fracture risk. With the extensive data on these agents from a number of large clinical trials, it is possible that the prediction of fracture risk reduction with antiresorptive agents can be improved by models that incorporate changes both in BMD and in bone turnover markers, thus providing more robust surrogate markers. This approach should be evaluated with data from these trials. For formation stimulating agents, BMD may well be a valid surrogate for fracture, provided again that the bone formed by these agents is of normal quality, although data from additional trials with these agents is needed. This work was supported by Grant AG04875 and AR27065 from the National Institute on Aging, U.S. Public Health Service.