Predicting fractures in an international cohort using risk factor algorithms without BMD
Philip N Sambrook,Julie Flahive,Fred H Hooven,Steven Boonen,Roland Chapurlat,Robert Lindsay,Tuan V Nguyen,Adolfo Díez‐Perez,Johannes Pfeilschifter,Susan L Greenspan,David Hosmer,J Coen Netelenbos,Jonathan D Adachi,Nelson B Watts,Cyrus Cooper,Christian Roux,Maurizio Rossini,Ethel S Siris,Stuart Silverman,Kenneth G Saag,Juliet E Compston,Andrea LaCroix,Stephen Gehlbach,Adolfo Díez-Perez
DOI: https://doi.org/10.1002/jbmr.503
IF: 6.39
2011-10-20
Journal of Bone and Mineral Research
Abstract:Clinical risk factors are associated with increased probability of fracture in postmenopausal women. We sought to compare prediction models using self-reported clinical risk factors, excluding BMD, to predict incident fracture among postmenopausal women. The GLOW study enrolled women aged 55 years or older from 723 primary-care practices in 10 countries. The population comprised 19,586 women aged 60 years or older who were not receiving antiosteoporosis medication and were followed annually for 2 years. Self-administered questionnaires were used to collect data on characteristics, fracture risk factors, previous fractures, and health status. The main outcome measure compares the C index for models using the WHO Fracture Risk (FRAX), the Garvan Fracture Risk Calculator (FRC), and a simple model using age and prior fracture. Over 2 years, 880 women reported incident fractures including 69 hip fractures, 468 "major fractures" (as defined by FRAX), and 583 "osteoporotic fractures" (as defined by FRC). Using baseline clinical risk factors, both FRAX and FRC showed a moderate ability to correctly order hip fracture times (C index for hip fracture 0.78 and 0.76, respectively). C indices for "major" and "osteoporotic" fractures showed lower values, at 0.61 and 0.64. Neither algorithm was better than the model based on age + fracture history alone (C index for hip fracture 0.78). In conclusion, estimation of fracture risk in an international primary-care population of postmenopausal women can be made using clinical risk factors alone without BMD. However, more sophisticated models incorporating multiple clinical risk factors including falls were not superior to more parsimonious models in predicting future fracture in this population.
endocrinology & metabolism