Hip fracture is rare before age 65

The purpose of measuring bone density is to direct preventive treatments towards those who would most benefit. Drug treatment is principally antiresorptive and it clearly reduces fractures. The 10 year chance of hip fracture in women age 50 who are already at twice the random risk is reportedly 1.4%.3 Antiresorptive drugs can halve the risk of fracture, but the cost-benefit of treating the group at risk, of whom only 0.7% will benefit from 10 years' treatment, is questionable. Furthermore, the cost of detecting everyone who is at twice the random risk of fracture by bone densitometry would be prohibitive. There is no evidence that measurement of bone density in perimenopausal women can direct antiresorptive treatment cost effectively. Benefits are inevitably limited when large sums of money are targeted at the prevention of a rare event.

Fracture protection and bone mass gained are poorly related

Antiresorptive drugs have long been marketed for their ability to restore lost bone. The argument linking bone gain to protection from fracture is intuitive. Bone density rises in people taking these drugs, and treatment reduces fractures. In reality, however, cause and effect are far from established. The three most recently published multicentre fracture prevention trials—for alendronate,4 risedronate,5 and raloxifene6—report similar rates of fracture prevention (40-50%) for very different gains in bone density (2.4-8.2%). What characterises all three trials, however, is a reduction in the markers of resorptive activity by 30% or more within six months. Furthermore, 12 month data, which were available only for risedronate,4 showed that the risk of fracture had fallen well before bone density peaked.

The pathology in osteoporosis is one of high bone resorption, not one of low bone density, which is a variable consequence. The evidence suggests that the strength of bone lies with the integrity of its architecture and with its level of turnover, not merely with its mass. High turnover of bone seems to be intrinsically unstable, whereas low density bone need be weak only if its low mineral content results from chronically high bone turnover. Postmenopausal Chinese women, for example, have significantly lower hip bone mineral density than white women and are classified at higher risk, but in fact they have fewer fractures. 7 8 Their rate of bone loss is reportedly slower, suggesting a lower rate of turnover.

The prevention trials suggest that antiresorptive drugs can halve the risk of fracture in the most osteoporotic bones without restoring significant density. Although change in bone mineral density may relate to reduction in risk of vertebral fracture, 9 10 the evidence is weak for the hip. In many instances, the modest gain in bone density after antiresorptive treatment is insufficient even to reclassify the patient, whose bones remain osteoporotic despite a significant reduction in their risk of fracture.

Fall in bone density with age contributes little to rise in fracture risk

A Dutch study of 5800 healthy men and women over the age of 55 years observed a 13-fold increase in the risk of femoral neck fracture between the ages of 60 and 80 years.11 Multiple regression analysis was used to determine the contribution to the rise in risk of fracture made by the corresponding fall in density of bone. In women, the loss of bone density associated with age contributed only 1.9 (95% confidence interval 1.5 to 2.4) and in men 1.6 (1.3 to 1.8). Thus, 85% of the rise in risk of fracture in ageing women is attributable to something other than the loss of bone density. This study was cross sectional and should be evaluated with this in mind. It was, nevertheless, statistically powerful and involved merely observation of an unselected general population.

The same group compared the predictive value for hip fracture of combining anthropometry (a measure of thinness) and a risk score with and without bone densitometry.12 The area under the receiver operating characteristic curve, indicating discriminatory power, was good in both cases, but not significantly different (0.83 and 0.88 respectively). Bone densitometry added little to a questionnaire in the prediction of hip fracture.

Relation between bone density and fracture risk is age sensitive

The widespread use of T scores to infer risk of fracture from bone density is based on the concept of fracture threshold. Postmenopausal bone density measurements are referred to the young adult norm in the densitometer's software and expressed as standard deviation scores of the young adult mean. The fracture threshold, the WHO definition of osteopenia, is set arbitrarily at −1 standard deviation score, which reflects an absolute bone density. There is no evidence, however, that the risk of fracture relates to the same bone density throughout life. Indeed, Hui and colleagues cross classified incident fractures with age and bone mass and found age to be a better predictor of hip fracture than radial bone density.13 De Laet et al confirmed the independent contribution that age makes to risk of hip fracture using dual energy x ray absorptiometry.11 A fracture threshold has little meaning when the risk of fracture to hips of the same bone density varies several fold over the postmenopausal age range.

Age is itself a robust predictor of risk of fracture, irrespective of previous treatment with antiresorptive drugs. The belief that taking antiresorptive drugs in middle age is an investment against fracture risk later on, seems ill founded. Protection against fracture is demonstrable within a few months of starting treatment but is lost again within five years of stopping it.14 There seems to be no age beyond which antiresorptive drugs are ineffective but no point in starting them earlier than the emergence of fracture risk.1

Epidemiological argument is flawed

Randomised controlled trials show that antiresorptive drugs reduce the risk of fracture. The rate of fracture could be reduced by either treating the whole population or targeting those at highest risk. Universal treatment, even if workable, is wasteful and expensive. Use of bone density to target treatment at high risk patients has therefore been widely advocated. Even though the cost of bone densitometry represents a tiny fraction (perhaps 0.1%) of the treatment that may ensue, screening with bone densitometry has long been deemed uneconomic in the United Kingdom, and use of clinical guidelines to select appropriate cases is encouraged.15

All of this, however, neglects flaws in the way epidemiological evidence is applied to support bone densitometry. Firstly, the evidence that bone density predicts fracture is based on studies of large populations. Clinicians, on the other hand, treat individuals and, although each person must by the nature of randomisation own some part of the population risk, the confidence with which bone density can be used to predict fracture in the individual is arguably negligible. Only if treatment is offered to a population can the outcome be expected to reflect the generality of population based data.

Screen shot of densitometry scan of osteoporotic hip

The statistic so often quoted in support of bone densitometry comes from epidemiology. It suggests that for each standard deviation decrease in femoral neck bone density, there is a 2.6 times increase in the age adjusted risk of hip fracture (95% confidence interval 1.9 to 3.6).16 Women with bone density in the lowest quarter had an 8.5-fold greater risk of hip fracture than those in the highest quarter. The data seem impressive but are based on just 65 women who sustained a fracture out of 8134 observed. Treatment of the entire 8000 with an antiresorptive drug (assuming acceptance, compliance, and a budget) could be expected to save a maximum 33 fractures (a generous estimate as the 50% reduction in fractures claimed for bisphosphonates is based on populations already selected for low bone density or existing fracture). Alternatively, 8000 scans to classify the 2000 women in the lowest quarter of bone density would, according to the trial data, identify only 34 of the 65 who suffered fractures,16 and only half (or 17) of these fractures could be prevented by drugs. Bone densitometry can tell us about populations, but the chances of predicting a preventable fracture by bone densitometry in an osteoporosis clinic of largely self referred individuals must be close to random.

Secondly, the epidemiological data relating bone density to fracture risk is derived mainly from short term studies, which are by their nature limited to association, not prediction. Individuals lose bone density at different rates throughout later life, so that even the epidemiological relation between current bone mineral density and future fracture risk breaks up progressively with time.17

Finally, epidemiological studies tend to compare the bone densities of people in a large cohort who have had fractures with the densities of people who have not. In order to do so, they correct the bone density measurements for age with a z score but do not correct the other age related factors that contribute to risk of fracture. In the study by Cummings et al, the mean age of people who had a fracture (76.4 years) was significantly greater than the age of people who did not (73.2 years).16 Bone density was corrected for the age difference, but nothing else was. The omission is important because some 85% of the contribution to the rise in fracture risk with age is unrelated to bone density. It will overestimate the contribution which bone density makes to risk of fracture, perhaps seriously, by attributing to bone density other (more powerful) factors that are independent of it. Epidemiology provides data on the burden of disease. It is for the medical community to apply the data appropriately, but in the case of bone densitometry it clearly is not doing so.

Conclusion

The ability of bone densitometry to predict future fracture is overstated, and the data on which such claims are based are overinterpreted. Bone densitometry is a major industry (an estimated 34 000 densitometry machines were in existence worldwide in 2000), and much of the research into osteoporosis depends on it. Clinical trials test efficaciousness (can it work?) in selected groups. The clinician is concerned with effectiveness (does it work?) in unselected individuals. The challenge is to show the latter.—Terence J Wilkin, Devasenan Devendra

Fractures other than hip should be considered

Although hip fractures before the age of 65 are rare, other fractures are not. In the European vertebral osteoporosis population study, 33% of the participants over the age of 50 years had a hip fracture.3 Hip fractures are more related to ageing than vertebral, wrist, and other fragility fractures (rib, humerus, etc). Men have fractures at a younger age than women, and corticosteroid induced fractures are not rare in any age group.

The WHO task force on osteoporosis management agrees that screening by densitometry before the age of 65 is not cost effective. But screening of high risk patients (case finding through education) is thought to be cost effective, in particular at age 70 in people who already have a low bone mass together with other risk factors such as low body weight, previous fracture, or family history.4

Osteoporosis is not always postmenopausal

Osteoporosis is the end stage of a number of pathophysiological processes and is not always postmenopausal. To concentrate only on bone resorption, leaving out bone formation, is shortsighted and without scientific basis.

Bisphosphonates and hormone replacement therapy are not the only treatments for prevention or management of osteoporosis. Much cheaper treatments include thiazides, vitamin D, dairy products such as milk (calcium at night), hip protectors, and physiotherapy. There is evidence based data that these treatments all prevent fracture.5-7 A relatively expensive but effective intervention with substantial non-compliance may prevent fewer fractures than an inexpensive, safe, and somewhat less effective intervention with a higher compliance.

Relation between fracture protection and bone mass gained

The relation between protection against fracture and bone mass gained through treatment is debatable. However, evidence based data indicate that fracture protection is related to the gain in bone mass8 and that bone strength is more strongly correlated with areal and actual bone density than with architectural values.9 But it is still possible that quality differences in bone composition contribute to bone strength and can be altered by drugs such as vitamin D, oestrogens, and fluoride.10

The evidence at present is that bone strength lies in its bone mass, geometry, architecture, and composition. Lowering bone turnover over a long time with strong antiresorptive drugs such as bisphosphonates will induce hypermineralisation of osteons, which may result in microcracks.11 The relative contribution of bone density to mechanical competence is also species dependent and relates to differences in bone composition.12

Effect of age

It cannot be argued that age is the main risk factor for fracture because physiological ageing is not equal for everyone. About 70% of the bone mass is genetically determined,13 and comorbidity, lifestyle, and use of corticosteroids over a lifetime will interfere with the physiological effects of ageing.

Nevertheless, the fall in bone density with age makes an important contribution to risk of fracture for at least some fragility fractures. The study by De Laet et al, which appears to show that this is not true for hip fractures, is a cross sectional study and therefore does not show a real fall in density.14 Cross sectional studies are subject to a cohort effect; participants could have already had a low bone mass from adult age. This together with lifestyle changes and falls might explain why differences in density did not explain the risk of fracture. It is not surprising that thinness and risk score may give important information in line with what bone density can measure. But measurement of bone mineral density can still be useful to select further high risk cases because the best reduction in fracture with drugs occurs in people in the lowest third of bone mineral density.15

The risk gradient for fracture roughly doubles for each decade's decrease in bone mineral density after age 50. The fracture threshold is set so that appropriate prevention is possible, but of course not everyone who is at risk will have a fracture. That ageing is an independent contributor is not surprising, because frailty increases with age, raising the risk of falls and failure of other organ systems. These are important elements for hip fracture. Because antiresorption drugs have to be given lifelong, they are now usually started close to the time when fractures occur or after the first fracture has occurred.

Interpreting the science

Although the chances of predicting a preventable fracture by bone densitometry may be small, it is also important that two thirds of patients can be reassured that fracture is less likely and therefore long term treatment is not needed. That there is overuse, overdiagnosis, and overtreatment for osteoporosis is due not to bone densitometry or to epidemiological reports but to various factors, including pressure of some competing pharmaceutical companies and the uncritical medical society relying more on technical instrumentation than on clinical skills and reasoning. As Wilkin and Devendra assert above, effectiveness is critical in daily clinical practice.—Jan Dequeker, Frank P Luyten