Does using lower limit of normal values enhance the ability of a single bone mineral density measure to predict fractures?

Abstract

Using a single bone mineral density (BMD) measure, we demonstrated that the lower limit of normal (LLN) method is more consistent in predicting osteoporosis fractures than the T-score in white menopausal women from the Study of Osteoporosis Fracture (SOF). In order to circumvent the inconsistencies and limitations with using the T-score when defining osteoporosis, we propose using 95% LLN values derived from centered polynomial models using the NHANES III BMD measures. The main aim of this study was to compare the two methods in prediction of fracture and agreement in osteoporosis classification using cohort data. We compared the fracture prediction ability of the two methods using a single BMD measurement in 4,948 white women aged 67-74 years in the SOF employing kappa statistics, sensitivity, and specificity. The T-score provided inconsistent osteoporosis classification (46.6%) across the five hip regions of interest (ROIs) and this was significantly (p<0.0001) reduced when using the LLN method (36.5%). Kappa statistics of incident fracture during 12 years of follow-up related to the prevalence of osteoporosis at baseline was significantly improved using the LLN method compared to using T-score. Sensitivity and specificity for fracture based on a single BMD measurement of different hip ROIs were more consistent using the LLN method. The LLN method provides a more consistent and efficient method for osteoporosis fracture prediction than the T-score in 67- to 74-year-old white women.