Comment on Kanis et al.: The use of clinical risk factors enhances the performance of BMD in the prediction of hip and osteoporotic fractures in men and women

Abstract

Dear Editors, I was delighted to see the results of the large international collaborative study for prediction of osteoporotic fractures in the August issue of Osteoporosis International [1]. This study confirms that adding clinical risk factors to bone mineral density (BMD) measurements improves our predictive power for fracture risk. I want to mention a public health point about this paper and the optimal way of presentation of the results in similar studies in this field. Although well established among fracture risk studies [2], use of gradient of risk (risk ratio/SD change in risk scores) has a limited application in clinical settings, as it does not provide the magnitude of attributable risk for any of the risk factors. In other words, despite our knowledge about independent association of several clinical risk factors with fracture risk, it is not clear how we can incorporate these risk factors into clinical evaluations. Studies aiming to provide clinically applicable results need to focus on absolute risks of fractures as well as relative risks. The other point to mention is the importance of specificity of the tests (or risk factors) from a public health point of view. To find high-risk populations, we need tests with high positive predictive value (PPV), and the most important factor to determine PPV is the specificity of the tests. In their paper [1], Kanis et al. describe a scenario in which incorporation of clinical risk factors with BMD for finding 10% of high-risk women at age 50 increases sensitivity from 26% to 42% while not affecting specificity. This increase in sensitivity was translated to a 14% increase in PPV. However, the same amount of increase in PPV could be achieved by a 5% increase in specificity (from 91% to 96%). In fact, we can reach a PPV of more than 50% by increasing specificity to ≥98% even with the sensitivity of 20%. Moreover, with the same sensitivity and specificity, PPV would be about 70% for finding 20% of high-risk women at age 50. The main advantage of highly specific tests is, although we fail to find the majority of high-risk people (due to low sensitivity), that a lower number of people identified by this method can be considered as high-risk individuals with higher confidence. This lower number of diagnosed patients may have a huge impact on society in terms of the cost of treatment and follow-up of patients.