ReviewIs a change in bone mineral density a sensitive and specific surrogate of anti-fracture efficacy?
Introduction
There is no evidence – using anti-fracture efficacy as the endpoint – guiding the choice of one anti-resorptive over another, two anti-resorptives over one, an anti-resorptive versus parathyroid hormone (PTH) or PTH plus an anti-resorptive given before, during or after PTH versus PTH alone. The only evidence available is based on trials comparing a single agent against a placebo [14], [24].
The lack of evidence is the result of the high cost of conducting large randomized comparator trials using anti-fracture efficacy as an endpoint [32]. This formidable challenge would be overcome if a change in BMD was a sensitive and specific predictor of anti-fracture efficacy—a requirement met if (i) a greater increase in BMD with one drug over another predicted a greater fracture risk reduction, (ii) no change in BMD predicted no change in fracture risk and (iii) a decrease in BMD predicted an increased fracture risk.
The observation validating the use of the change in BMD in an individual as a surrogate of anti-fracture efficacy is finding in clinical trials, that the confidence intervals (CI) around the BMD changes do not overlap. This is illustrated in the cartoon (Fig. 1, left panel). The 95% CIs around a net BMD loss of 2%, no net change in BMD, and a net 8% increase in BMD do not overlap, and these respectively predict a 30% fracture risk increase, no increase and 50% fracture risk reduction, with 95% CIs around these risk estimates also not overlapping. Under these circumstances, a net loss of BMD predicts an increased fracture risk, no BMD change predicts no change in fracture risk and a net BMD increase predicts a decrease in fracture risk.
There is little evidence from clinical trials support the above criteria [3], [52]. For example, Cummings et al. [11] report that only 16% of the variation in vertebral fracture risk reduction between clinical trials was explained by the variation in BMD change as reflected in the overlapping 95% confidence intervals added to the published figure (Fig. 1, right panel). Similarly, only 4–30% of the variance in vertebral fracture risk reduction was explained by the change in BMD in other trials [34], [47], [50]. As 70–96% of differences in vertebral fracture risk reduction remain unaccounted for by differences in BMD changes, a greater increase in BMD with one drug over another is no assurance of greater anti-fracture efficacy, and more importantly no change or a fall in BMD during treatment does not mean treatment has failed to reduce fracture risk.
Only one study examined the association between a change in BMD and non-vertebral fracture risk reduction. Hochberg et al. [28] reported that the risk of non-vertebral fractures decreased when an increase in BMD accompanied anti-resorptive treatment. They inferred that the change in BMD can be used as a surrogate of fracture risk reduction and, more importantly, that a lack of change in BMD implied lack of anti-fracture efficacy. However, a reanalysis of the data correcting for methodological issues suggested that the reduction in non-vertebral fracture risk was not associated with changes in BMD [16].
Section snippets
The meaning of ‘change’ in BMD and fracture risk ‘reduction’
The percent ‘change’ in BMD used to predict fracture risk reduction in the above regression analyses is the net result of the change in BMD in the treated group minus the change in controls during a study of usually 3 years duration (Fig. 1, [2], Fig. 2). This net change is the result of a range of scenarios, only three are shown.
A net increase in BMD (of say 4%) in scenario A is the result of a 2% decrease in BMD in the treated group and a 6% decrease in BMD in controls—i.e., treatment slowed
Abnormalities in BMU balance and remodeling rate produce structural decay and bone fragility
There are several cogent reasons why a change in BMD cannot be used to infer anti-fracture efficacy of a drug, or greater anti-fracture efficacy of one drug over another. These concern the influence of pre-treatment rates of remodeling on the BMD response and the effects of therapy on the material and structural determinants of bone strength, not all of which are contained in the BMD measurement or necessarily produce a beneficial effect on bone strength when they are contained in the BMD
Suppressing remodeling rate and filling of the transient remodeling space deficit
During steady state, before treatment is given, there is simultaneous appearance of equal numbers of newly excavated sites and disappearance of previously excavated sites as they fill or partially fill with osteoid that promptly mineralizes. (There is net bone loss from each BMU because each resorption–formation sequence leaves a small deficit.)
When an anti-resorptive agent is given, this steady state is perturbed. The rate of appearance of new BMUs decreases precipitously while bone formation
Do anti-resorptives differ in potency in suppressing remodeling?
The dependence of the initial increase in BMD on baseline remodeling makes assessment of effects of anti-resorptives on BMD difficult because slight differences in baseline remodeling influence the BMD response independent of any differences in the inhibitory potency of the drugs. Bisphosphonates suppress remodeling more than raloxifene [1], [31], but whether one bisphosphonate suppresses remodeling more than another in the doses used clinically is not clear.
Rosen et al. [44] report that
The effect of potency on long-term changes in BMD
The continued increase in BMD after 12–18 months has a different mechanism. As the birth rate of new remodeling sites is suppressed, existing osteons (that would otherwise have been removed by high remodeling) have more time to undergo secondary mineralization—a physicochemical change produced by enlargement and proliferation of crystals deposited during primary mineralization [5], [45]. The process of secondary mineralization is also part of the remodeling ‘transient’. That is, it is a
Effects of drug therapy on BMU balance
There is no histomorphometric evidence in human subjects that bisphosphonates reduce the volume of bone resorbed or increase the volume deposited in each BMU (by altering the production, work or life span of the osteoclast and osteoblast cells). There is evidence for a reduction in erosion depth using estrogen [17]. If one drug reduced the volume of bone resorbed and increased the volume of bone formed in the BMU more greatly than another, then structural decay will occur less with that drug.
Fracture risk reduction is independent of baseline remodeling rate, gain or loss of BMD
Whatever the mechanisms involved in producing the increase in BMD, the question is whether the greater increase in BMD predicts a greater vertebral or non-vertebral fracture risk reduction. This does not appear to be the case. Although patients with higher baseline remodeling have a higher absolute risk for fracture and a greater increase in BMD with anti-resorptive treatment than patients with lower baseline remodeling, the fracture risk reduction, relative to controls with the same baseline
Dissociation between BMD change and fragility
A change in BMD is unlikely to predict fracture risk because the morphological changes responsible for reversing or slowing progression of bone fragility, although not yet well defined, are not necessarily captured in the BMD change. For example, completion of bone formation in resorptive cavities that are stress concentrators reduces bone fragility [26], but this structural change that is unlikely to be sensitively captured in the BMD increase which is driven by the baseline remodeling rate. A
Where to next?
There are more questions than answers. If no change in BMD or a fall in BMD does not necessarily mean treatment failure, if fracture incidence is reduced but not abolished during treatment because drugs reduce but do not abolish fracture risk, then what is treatment failure? If there have been no comparator trials showing that one drug is more efficacious than another, no trials showing that two drugs are better than one, or anabolics are better than anti-resorptives in preventing fractures,
References (53)
- et al.
Raloxifene enhances vertebral mechanical properties independent of bone density
Bone
(2006) - et al.
Alendronate increases bone strength by increasing the mean degree of mineralization of bone tissue in osteoporotic women
Bone
(2000) - et al.
Improvement in spine bone density and reduction in risk of vertebral fractures during treatment with anti-resorptive drugs
Am. J. Med.
(2002) Treatment of postmenopausal osteoporosis
Lancet
(2002)- et al.
Fracture risk reduction during treatment with teriparatide is independent of pre-treatment bone turnover
Bone
(2006) - et al.
Changes in bone mineral density explain little of the reduction in vertebral for non-vertebral fracture risk with anti-resorptive therapy
Bone
(2004) - et al.
Bisphosphonates suppress periosteal osteoblast activity independently of resorption in rat femur and tibia
Bone
(2006) - et al.
Long term effects of calcium supplementation on bone loss and fractures in postmenopausal women
Am. J. Med.
(1995) - et al.
Alendronate increases degree and uniformity of mineralization in cancellous bone and decreases the porosity in cortical bone of osteoporotic women
Bone
(2001) - et al.
The influence of bone morphology on fracture toughness of the human femur and tibia
Bone
(1997)