Prediction of osteoporotic fractures in the general population by a fracture risk score. A 9-year follow-up among middle-aged women.

The possibility of predicting the occurrence of fractures on the basis of risk factors for osteoporosis was evaluated between 1975 and 1986 in a follow-up study of 1,014 women aged 45-64 years from a Dutch population sample. During the 9-year period of follow-up, 16% of the women experienced one or more fractures. Of 12 historical and radiologic risk factors for osteoporosis which are considered important in the biomedical literature, none were found to be strong indicators of future fractures. Complete information on risk factors was available for 742 women. A fracture risk score was calculated for each person by combining the simultaneous influence of several risk factors in a multivariate analysis. The risk score discriminated relatively well between women with high and low risks of fractures. The risk ratio between the highest and the lowest risk score quintiles was 6.4 for all fractures and 7.0 for type I osteoporotic fractures (fractures of the vertebral bodies and distal forearm). However, if belonging to the highest quintile was considered as a screening test for fracture prediction, the sensitivity and specificity were poor: 0.37 and 0.84, respectively, for all fractures and 0.48 and 0.83, respectively, for type I osteoporotic fractures. These results indicate that it might not be efficient to use risk factor status to select women for fracture prevention programs.