RESUMO
Probabilistic models including clinical risk factors with or without bone mineral density (BMD) have been developed to estimate the 5- or 10-year absolute fracture risk. We investigated the performance of the FRAX and Garvan tools in a well-characterized population-based cohort of 3560 postmenopausal, volunteer women, aged 60 to 85 years at baseline, included in the Fracture Risk Brussels Epidemiological Enquiry (FRISBEE) cohort, during 5 years of follow-up. Baseline data were used to calculate the estimated 10-year risk of hip and major osteoporotic fractures (MOFs) for each participant using FRAX (Belgium). We computed the 5-year risk according to the Garvan model with BMD. For calibration, the predicted risk of fracture was compared with fracture incidence across a large range of estimated fracture risks. The accuracy of the calculators to predict fractures was assessed using the area under the receiver operating characteristic curves (AUC). The FRAX tool was well calibrated for hip fractures (slope 1.09, p < 0.001; intercept -0.001, p = 0.46), but it consistently underestimated the incidence of major osteoporotic fractures (MOFs) (slope 2.12, p < 0.001; intercept -0.02, p = 0.06). The Garvan tool was well calibrated for "any Garvan" fractures (slope 1.05, p < 0.001; intercept 0.01, p = 0.37) but largely overestimated the observed hip fracture rate (slope 0.32, p < 0.001; intercept 0.006, p = 0.05). The predictive value for hip fractures was better for FRAX (AUC: 0.841, 95% confidence interval [CI] 0.795-0.887) than for Garvan (AUC: 0.769, 95% CI 0.702-0.836, p = 0.01). The Garvan AUC for "any Garvan" fractures was 0.721 (95% CI 0.693-0.749) and FRAX AUC for MOFs was 0.708 (95% CI 0.675-0.741). In conclusion, in our Belgian cohort, FRAX estimated quite well hip fractures but underestimated MOFs, while Garvan overestimated hip fracture risk but showed a good estimation of "any Garvan" fractures. Both models had a good discriminatory value for hip fractures but only a moderate discriminatory ability for MOFs or "any Garvan" fractures. © 2021 The Authors. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.
RESUMO
We assessed the rate of non-reported fractures in the FRISBEE cohort. Over a median follow-up period of 9.2 years, we registered 992 fractures. The global percentage of non-reported fractures was 21.3%. Underreporting of fracture event might influence any model of fracture risk prediction. INTRODUCTION: Most fracture cohort studies rely on participant self-report of fracture event. This approach may lead to fracture underreporting. The purpose of the study was to assess the rate of non-reported fractures in a well-characterized population-based cohort of 3560 postmenopausal women, aged 60-85 years, included in the Fracture Risk Brussels Epidemiological Enquiry (FRISBEE) study. METHODS: Incident low-traumatic or non-traumatic fractures were registered annually during phone calls. In 2018, we reviewed the medical files of 67.9% of our study participants and identified non-reported fractures ("false negatives fractures (FN)"). We also evaluated whether the rate of FN was influenced by baseline patients' characteristics and fracture risk factors. Generalized estimating equation (GEE) was used to calculate odds ratio (OR) and 95% CI. RESULTS: Over a median follow-up period of 9.2 years, we registered 992 fractures (781 by self-report, confirmed by a radiological report and 211 unreported). The global false negative rate for all fractures was 21.3%, including 22% for MOFs (major osteoporotic fractures), 13.1% for other major fractures, and 25.8% for minor fractures. The rate of non-reported fractures varied by fracture site: for MOFs, it was 2.7% (n = 2/73) at the hip, 5.3% at the proximal humerus (n = 5/94), 7.1% at the wrist (n = 11/154), and 46.5% at the spine (n = 100/215). For "other major" fractures, the highest rate of false negatives fractures was found at the pelvic bone (21%, n = 13/62), followed by the elbow (17.9%, n = 5/28), long bones (10.5%, n = 2/19), ankle (6.2%, n = 4/65), and knee (5.9%, n = 1/17). Older subjects (OR 1.7; 95% CI, 1.2-2.4; P = 0.003), subjects with early non-substituted menopause (OR 1.8; 95% CI, 1.0-3.3; P = 0.04), with a lower education level (OR 1.5; 95%CI, 1.1-2.2; P = 0.01), and those under drug therapy for osteoporosis (OR 1.5; 95% CI, 1.0-2.2; P = 0.05) were associated with a higher rate of FN. CONCLUSIONS: In conclusion, underreporting of a substantial proportion of fracture events will influence any model of fracture risk prediction and induce bias when estimating the associations between candidate risk factors and incident fractures.