Bone Matters: Are Density Increases Necessary to Reduce Fracture Risk?

Decades of published studies have confirmed the relationship between bone mass and the mechanical properties of bone tissue. When bone samples are investigated in a laboratory setting under controlled loading conditions, bone mineral density (BMD) typically explains 60 to 80% of bone strength.1-3 The relationship between BMD and bone strength has been confirmed in human subjects, in which BMD has been shown to predict the risk for various types of fracture in prospective studies.4-6 This relationship is exponential, with relatively small decrements in BMD (10–15%), approximately doubling the risk for fracture.7, 8 As a consequence of the exponential BMD/fracture risk relationship, very small increases in BMD can be expected to have very large effects on reducing fracture rates (Fig. 1). It is not necessary to double BMD to reduce fractures by a factor of 2. On the basis of the established relationships between BMD, bone strength, and fracture risk, current therapies for reducing fractures have been targeted at reversing bone loss to increase bone strength. Several different therapies have been shown to increase bone density and reduce fractures.9, 10 The effects have been somewhat consistent with regard to hip fractures, with the magnitude of BMD increase roughly corresponding to the decrease in risk. For vertebral fractures, the BMD/risk relationships have been less consistent (Table 1). In the case of antiresorptive treatments (including, calcium, estrogens, estrogen analogs, calcitonin, and bisphosphonates), small changes in BMD (a few percent) have been associated with dramatic, but varying, declines in vertebral fracture rates (30 to 90%).11-18 It has been noted that the therapeutic reductions in fracture rates are much larger than expected for the given changes in bone density, based on observations from epidemiological studies (Table 2). Furthermore, published studies have suggested that the relationship between BMD change and fracture risk reduction varies for different pharmaceutical agents. As a result, the importance of BMD as a surrogate for bone strength has been questioned. Nondensity-related effects of therapeutics Technical limitations of measuring BMD changes Differences in skeletal fragility of study populations Hysteresis effects in the BMD/fracture risk relationship Clearly, fracture risk is related to many factors, not just bone density. Age, propensity to fall, skeletal geometry, bone turnover (to name a few), all contribute to the outcome of fracture. The precise relative contribution of each of these factors is not known. It is possible that a change in bone quality caused by the treatment and not reflected by a density measurement is working to reduce fracture risk. However, existing therapies are known to work by reducing resorption, leading to a temporal decoupling of the formation/resorption relationship and resulting in an increase in bone density. This mechanism does not directly stimulate formation. Histomorphometic studies have confirmed that the bone produced as a result of antiresorptive therapy is of normal quality.19-21 Because increased bone resorption is an independent risk factor, suppressing resorption would be a feasible nondensity-related factor for reducing fracture risk. In most cases, changes in bone density and changes in bone resorption are related. The treatments showing the largest increases in density tend to show the largest reductions in the markers of bone resorption. Likewise, the therapeutics with minimal increases in bone density are also associated with minimal reductions in bone resorption. Therefore, if the current antiresorptive compounds are causing a nondensity-related effect, alteration of bone resorption is probably not the mechanism for discrepancies between therapeutic agents. Exponential relationship between bone density and fracture risk. Curves are shown for odd's ratios of 1.5, 2.0, and 3.0 per SD change in bone density. Shown are studies with alendronate (FIT I and II), risedronate (RVE and RVN), raloxifene (MORE) and calcitonin (PROOF). Listed is the 3 year change in spinal BMD for treated subjects, 3 year reduction in vertebral fracture rates compared to calcium controls, and the baseline values for spinal BMD and vertebral fracture prevalence in each study. Other nondensity effects of therapeutics could include alterations to muscular strength, possibly by influencing cellular calcium channels. Effects on balance, vision, or other fall-related influences are also possible. However, studies to date have shown no significant effects of approved antiresorptive therapeutics on these parameters. Thus, if antiresorptive agents are working through nondensity means, the mechanism remains uncertain. Current clinical BMD techniques (except for quantitative computed tomography) measure integral bone: the combination of cortical and trabecular bone. Depending on the skeletal site measured, 50 to 100% of the bone mineral content is contained in the cortical bone.22 Yet it is the trabecular bone that (1) provides weight-bearing capacity and (2) responds to antiresorptive therapy. For example, at the lumbar spine, 50 to 75% of the bone mineral is contained in cortical bone. Thus, when the lumbar spine is measured in the anteroposterior projection using dual X-ray absorptiometry (DXA), 50 to 75% of the signal will be generated from cortical bone. However, it is the trabecular bone of the vertebral body that predominately responds to antiresorptive therapy. As a result, a 10% change in trabecular (weight-bearing) bone of the spine will appear as only a 2.5 to 5% change in spinal BMD. The same will be true at the proximal femur. Thus, the true density effect of antiresorptive therapy is typically underestimated by most clinical BMD assessment techniques. At the spine, BMD increases may be additionally masked due to osteophytes, facet joint sclerosis, arthritis, and other degenerative diseases. This will be more likely in elderly subjects, who comprise the majority of participants in treatment studies. At lower baseline values, relatively smaller percentage increases are required to produce a given reduction in fracture risk. In addition, the estimated density increases needed to reduce fractures are much larger than those seen in published treatment studies. Calculations are based on data from the Study of Osteoporotic Fractures using femoral neck bone density for the prediction of hip fractures.5 These technical limitations cannot fully explain the discrepancy between BMD changes and fracture reduction, because the same BMD assessment technologies have been used for epidemiological investigations and clinical trials. In other words, the same limitation applied to both the observational and the interventional data. Any lack of sensitivity for detecting age-related density loss would also exist for detecting a BMD increase in an interventional study. It is possible that technical limitations of BMD assessment could hinder the ability to distinguish differences in BMD response for different therapies. It has been suggested that different brands of densitometers can yield different percentage BMD responses, even when the same subjects and skeletal sites are measured.23 Technical variations in analysis algorithms among manufacturers (particularly edge-detection software) may influence the measured response and contribute to the discrepancy. If two studies are performed using completely different DXA systems, differences in response might be expected unrelated to the type of treatment used. This will also be true if the composition of DXA instruments (i.e., the relative number of DXA systems for different manufacturers) differs significantly between various studies. The relationship between bone density and fracture risk is exponential (Fig. 1). In this type of relationship, the slope of the curve varies with density. As a result, the effect of a change in bone density on fracture risk depends on the starting BMD of the population. As the density of the study population decreases, progressively smaller declines in BMD are required to yield the same increase in fracture risk. Likewise, a treatment study performed in a lower-density population would require a smaller BMD increase to obtain the same reduction in fractures. Table 2 shows the theoretical increase in BMD required to reduce fracture risk by 25, 33, and 50% as a function of the initial BMD T score of the study population. This calculation was performed for femoral neck T score and hip fractures, though similar results can be obtained for other BMD measurements and fracture types. These models suggest that the increase in BMD needed to reduce fractures by a given amount depends on the starting density of the study population. Significant differences in the baseline BMD may explain the observed differences in BMD increases associated with different treatments, despite similar effects on fracture rates. In addition to baseline BMD, differences in skeletal fragility may exist because of other risk factors. Advanced age and existing fractures increase fracture risk dramatically. The presence of additional risk factors will result in an increase in the slope of the fracture relationship shown in Fig. 1. Thus, even if baseline BMD is equal, it is conceivable that high risk individuals will require smaller BMD increases to reduce fractures than less frail subjects. Fracture reductions in high risk subjects would also be expected to be greater than in low risk subjects on the same treatment. It is important to note that there are no direct, head to head studies, comparing the effects of antiresorptive therapies on fracture rates. At this point, the comparative studies are limited to looking at BMD changes. When comparing different treatment studies and their effect on fracture rates, it is important to take into account the baseline risk of the study population due to both skeletal and nonskeletal factors (Table 1). Hysteresis is a physical phenomenon defined by an alteration in response on a reversal of effect. In engineering, certain materials exhibit hysteresis, in that the force/deformation relationship depends on whether the mechanical tests are performed in tension or compression. This same type of phenomenon may also exist in the relationship between bone density and fracture risk (Fig. 2). The BMD/fracture risk relationship defined by observational studies (in which density declines with age) may not necessarily apply in situations where density increases (in response to therapy). For example, if a 10% decrease in BMD increases fracture risk by a factor of 2, it is possible that reducing fracture risk by a factor of 2 requires either more or less than a 10% increase in BMD. The BMD/fracture risk relationship may not be, nor is it required to be, bidirectional. Hysteresis effect for bone density (shown in standard deviation units) and fracture risk. The fracture risk curve based on observational data is shown as the solid line, indicating the increase in risk for a given decrease in bone density. The dashed lines represent potential variations in the bone density/fracture risk relationship as bone density increases (and fracture risk decreases) due to therapeutic intervention. Note that the density/fracture risk curve is not necessarily bidirectional. A hysteresis effect in the BMD/fracture risk relationship could explain why significant fracture risk reductions have been observed with such small BMD increases. It is possible that the magnitude of this effect varies for different therapeutics. However, existing therapies all act via an antiresorptive mechanism, so that a variable effect on the BMD/risk relationship is not a likely explanation for the observed difference. Yet as previously mentioned, variations in skeletal fragility due to additional risk factors might induce a type of hysteresis effect by increasing the slope of the fracture risk curve. The relationship between BMD and fracture risk is well established. The association between bone density (measured at several different skeletal sites) and hip fracture risk is stronger than that between cholesterol levels and heart disease.24 Currently, the best method for assessing fracture risk remains a bone density measurement. However, the importance of increasing bone density for reducing fracture risk has been less clear, because of variations in the effect of existing therapeutics on BMD and fracture rates. Several possibilities exist for explaining the variable influence of different treatments on increasing BMD and reducing fractures. Nonskeletal effects can explain some of the difference, possibly by influencing bone quality independent of density. At this time, an independent effect of antiresorptive therapy on bone quality remains to be shown. Some of the difference may be the result of technical limitations for measuring BMD, although this would be expected to influence different therapies equally. Variations in the densitometers used for assessing BMD may be partially responsible for the discrepancy, particularly if multiple brands of devices are used. It is also possible that differences in the skeletal fragility of the study populations and BMD hysteresis effects are influencing the fracture risk relationships. Comparison of fracture studies using different therapeutic agents is complicated by several factors. Specific details of the study populations are often not stated in the published reports, either because they are in abstract form, or have simply not been described in detail. Furthermore, a lack of standardization for different BMD instruments complicates the comparison of baseline density information.25 Standardized bone density values could be used,25 but these values are only available for the L2–L4 spine and the total femur regions of interest, and they are not typically provided in published reports. Instead, baseline population characteristics are usually reported as T scores to avoid these known calibration differences. Otherwise, it would be necessary to report the baseline BMD for each type of densitometer used in the study. Yet even when T scores are used, variations in normative data can prevent accurate comparisons.26 Even though the precise reasons for the observed variability in the BMD/fracture risk relationship have not been determined, it remains true that bone density is the primary determinant of bone strength. Although studies have shown varying effects of different therapeutics, bone density still represents the best and most readily quantifiable measure of fracture risk and skeletal response.