Background Fragility hip fractures are associated with increased mortality, especially in men1. As bone mineral density (BMD) determination is the single most important determinant of bone strength, and mass dual-energy X-ray absorptiometry (DXA) testing is not cost-effective, several clinical decision rules have been developed to identify individuals with higher probability of low BMD. However, their validation across different populations has led to heterogeneity in the proposed cut-offs, making its applicability unclear and calling for calibration. Finally, when similar validity can be ensured, the simpler the clinical decision tool the more likely it is to be applied in clinical practice.
Objectives To develop and validate a simple clinical screening tool (EPIPOST) able to identify men with higher probability of having low bone mineral density (BMD) who may benefit from DXA testing and to compare its discriminatory ability with two other osteoporosis screening tools in men (OST and MORES), after calibration for our population.
Methods As part of EPIPorto population-based study among adults, 147 men aged between 40 and 65 years were assessed. Age, height, weight, body mass index and several body circumferences were recorded by trained observers. DXA whole-body scans were performed for BMD assessment. For the calibration of OST and MORES, new regression parameters were estimated for each risk factor included accounting for their prevalence and also for the prevalence of low BMD in our population. For EPIPOST development, the different anthropometric variables were tested using logistic regression models to predict low BMD. EPIPOST validation was done by the leave-one-out cross-validation method. The overall fit and discriminatory capacity of the different models were assessed by direct comparison of the observed and expected prevalences of low BMD by quartiles of each score, Hosmer-Lemeshow “goodness-of-fit” test and area under the receiver operating characteristic (ROC) curve. Finally, likelihood ratios (LR) were calculated to select the ideal cut-off for each model.
Results Calibration maintained the discriminatory capacity of OST and MORES (AUC of 0.73 and 0.75, respectively) and improved the fit. The EPIPOST included only upper arm circumference and showed better discriminatory capacity (AUC 0.76). For predicting low BMD, OST≤2 had a sensitivity of 100% and a specificity of 8.2%; MORES>-2 had a sensitivity of 93.9% and a specificity of 30.6%; EPIPOST>-2 had a sensitivity of 98.0% and a specificity of 18.6%. The LR analysis showed that EPIPOST had higher discriminative ability across different risk levels (LR range of 0.1-18.4, compared to 0.0-2.4 with OST and 0.2-2.8 with MORES).
Conclusions Calibration of OST and MORES improved the fitting of both models to our population data and maintained their discriminatory ability to identify men with low BMD. The newly developed tool, EPIPOST, is easier to execute and performed similarly to OST and MORES in terms of overall accuracy while showing a wider range of discriminatory ability.
Hopkins RB, et al. Estimation of the lifetime risk of hip fracture for women and men in Canada. Osteoporos Int, 2012. 23(3): p 921-7.
Disclosure of Interest None declared