Risk factor algorithms enable fracture prediction without BMD
MedWire News: Postmenopausal fracture risk can be predicted without taking bone mineral density (BMD) measurements, according to a large international study published in the Journal of Bone and Mineral Research.
Philip Sambrook (University of Sydney, Australia) and colleagues used data from the Global Longitudinal Study of Osteoporosis in Women (GLOW), which involved women from primary care practices in 10 countries, to compare two fracture risk models that use clinical risk factors, excluding BMD, for risk assessment.
The study involved 19,586 women over 60 years of age who were not receiving any medication for osteoporosis. These participants completed questionnaires regarding fracture risk factors, previous fracture, and health status.
Two models were tested - the World Health Organization Fracture Risk (FRAX) and the Garvan Fracture Risk Calculator (FRC) - and compared against a simple model that only used age and prior fracture.
Over the course of the 2-year study period, 880 women reported incident fractures, 468 sustained a FRAX-defined major fracture, and 583 sustained an FRC-defined osteoporotic fracture.
Both models showed similar success in predicting hip fracture - the C-index (derived from Cox regression models) was 0.78 for FRAX and 0.76 for FRC. The C-indices for "major" and "osteoporotic" fractures were 0.61 and 0.64, respectively.
However, neither model was any better than the simple age and fracture history based model (C-index of 0.78).
"These results are of interest given that the two algorithms use quite different clinical risk factors to estimate risk," say Sambrook et al. The FRAX algorithm comprises eleven variables whereas FRC uses only five. One difference is that the FRC algorithm includes the number of falls, whereas the FRAX algorithm does not include falls and previous fracture is simply catagorized 'yes' or 'no.'
"In the GLOW cohort, in the absence of BMD data, the inclusion of additional risk factors does not appear to alter fracture risk estimation substantially," write the researchers, which suggests simple models are no less effective than complex algorithms when BMD is unknown.
They conclude that although these models do not provide a significant basis for treatment decisions, the results do suggest that estimates of fracture risk can be made in everyday clinical practice.
MedWire (www.medwire-news.md) is an independent clinical news service provided by Current Medicine Group, a trading division of Springer Healthcare Limited. © Springer Healthcare Ltd; 2010
By Chloe McIvor