Our findings confirm that although there is a strong relation between low bone mineral density and increased fracture risk, with higher fracture rates observed among women with osteoporosis (as defined on the basis of bone mineral density), fractures occurred more commonly in women with normal or osteopenic values, since more women fall into these categories. Suprisingly, although we found that, among those with osteoporotic T-scores, fracture rates were higher among women at least 65 years of age than among those 50–64 years of age, over 60% of fractures occurred in younger women with T-scores greater than –2.5. We also found that excess fracture rates by T-score category were highest when using the total hip bone data, which suggests that T-scores for this site might afford greater stratification of fracture risk than the lumbar spine T-score or the minimum T-score. This finding differs from what is currently recommended in the Canadian reporting guidelines for bone mineral density.14
Our findings are consistent with those of other authors, who have reported that the majority of fractures occur in postmenopausal women with T-scores above –2.5.6,7,15
Compared with these other studies, the strengths of our study include use of multiple central sites for dual x-ray absorptiometry and confirmation of fractures by radiography.7
Our study had a number of limitations. The study group was selected on the basis of physician referral for bone mineral densitometry and was not a random sample of the general population in Manitoba. However, our use of a regionally based clinical cohort reflects the population in which such testing is typically used in clinical practice. Ascertainment of fractures from administrative databases has limitations, particularly the potential misclassification of traumatic and nontraumatic fractures and the underreporting of minor fractures. Many vertebral fractures go undiagnosed, since more than 65% of spine fractures do not cause obvious symptoms.16
Underrepresentation of vertebral fractures as a percentage of all fractures could partly explain why the total hip T-score captured a higher percentage of osteoporotic fractures than did the lumbar spine T-score.
This study highlights the fact that most fractures occur in women with normal or osteopenic bone mineral density. If treatment decisions are based on bone mineral density values alone, many of these women would not receive treatment.4,5
This is concerning, because a history of fracture indicates an increase in risk of future fractures (by 1.5 to 2.0 times), regardless of bone density.17
As such, treatment decisions should be based not only on the T-score but also on overall fracture risk.
Until recently, clinical guidelines have made little distinction between osteoporosis as defined by the T-score and fracture risk. As a result, older patients with fragility fractures who did not have osteoporosis, as defined by the World Health Organization criteria, may not have received treatment.18
Our study supports the concept that bone mineral density captures only one aspect of fracture risk, namely the quantity of bone, and that other factors that may influence bone strength or the risk of falls are also contributors to fracture risk. For example, prospective studies have consistently demonstrated that clinical risk factors such as age, prior fracture, a family history of osteoporosis and corticosteroid use contribute to fracture risk, independent of bone density as determined by absorptiometry.19–21
The assessment of a patient's fracture risk could be enhanced by considering the clinical risks, especially for people who are at moderate risk on the basis of bone mineral density alone.22–24
Our findings suggest that osteoporosis treatment guidelines should be modified to allow for targeted case-finding, whereby a combination of clinical risk factors (e.g., prior fracture and age) and bone mineral density could be used to derive an individual's 10-year probability of fracture and hence the overall risk of fracture.17,25
Clinicians could then use the absolute fracture risk to guide treatment decisions, in a manner similar to the way in which clinical risk factors are combined with laboratory data to stratify risk and guide therapy in cardiovascular disease.
The World Health Organization is currently developing algorithms that will incorporate clinical risk factors into the calculation of 10-year fracture risk. The Scientific Advisory Board of Osteoporosis Canada and the Canadian Association of Radiologists both recommend that bone mineral density results be reported in the context of clinical risk factors, such as age, sex, prior fracture history and steroid use, which can then be used to generate a 10-year prediction of fracture risk.14
This approach will increase our ability to accurately identify men and women at high risk for fracture and to appropriately guide treatment decisions.