Evaluation of a prediction model for long-term fracture risk

The NOF cost-effectiveness model, based on clinical risk factors and femoral neck aBMD, predicted overall fracture risk in a cohort of postmenopausal women followed for up to 22 years.

To assess the ability of a statistical model to predict long-term fracture risk for a population of postmenopausal women, we compared observed fractures to those predicted by the National Osteoporosis Foundation's (NOF) cost-effectiveness model.

In this population-based study, 393 postmenopausal Rochester, MN, women had baseline measurements of femoral neck areal BMD (aBMD) and assessment of the clinical risk factors (personal fracture history, family history of osteoporotic fracture, low body weight, and smoking status) that were included in the NOF model. They were then followed prospectively for up to 22 years. Fractures were ascertained by periodic interview and review of community medical records. Standardized incidence ratios (SIRs) compared observed fractures to predicted numbers.

During 4782 person-years of follow-up, 212 women experienced 503 fractures, two-thirds of which were caused by moderate trauma. When undiagnosed (incidentally noted) vertebral and rib fractures were excluded, there was general concordance between observed and predicted fractures of the hip (SIR, 0.78; 95% CI, 0.56-1.01), distal forearm (SIR, 1.22; 95% CI, 0.86-1.68), spine (SIR, 0.76; 95% CI, 0.50-1.11), and all other sites combined (SIR, 1.18; 95% CI, 0.97-1.42). Fracture prediction by the NOF model was about as good after 10 years as it was earlier during follow-up.

This study validates the ability of a statistical model based on femoral neck aBMD and common clinical risk factors to predict the actual occurrence of fractures in a cohort of postmenopausal white women.