Development of a model for identification of individuals with high risk of osteoporosis

Many developing countries, including Vietnam, lack DXA resources for the diagnosis of osteoporosis, which poses difficulties in the treatment and prevention of osteoporosis at the individual level. We have developed and validated a prediction model for individualized assessment of osteoporosis based on age and body weight for men and women.

To estimate the prevalence of osteoporosis and to develop and validate a prediction model for estimating the absolute risk of osteoporosis in the Vietnamese population.

The study involved 1477 women and 669 men aged 50 years and older, who were recruited from the general population in Ho Chi Minh City (Vietnam). Bone mineral density (BMD) at the femoral neck, total hip, and lumbar spine was measured by DXA (Hologic Horizon). The diagnosis of osteoporosis was based on BMD T-score (T-score ≤ − 2.5) at the femoral neck or lumbar spine which was derived from a published reference range for the Vietnamese population. The logistic regression model was used to develop the prediction model for men and women separately. The bootstrap method was used to evaluate the model performance using 3 indices: the area under the receiver’s operating characteristic curve (AUC), Brier score, and R-squared values.

The prevalence of osteoporosis at any site was 28.3% in women and 15.5% in men. The best predictors of osteoporosis risk were age and body weight. Using these indices, a cut-off of 0.195 for women yielded an AUC of 0.825, Brier score = 0.112, and it explained 33.8% of total variance in risk of osteoporosis between individuals. Similarly, in men, the internal validation with a cut-off of 0.09 yielded good accuracy, with AUC = 0.858, Brier score = 0.040, and R-squared = 30.3%.

We have developed and validated a prediction model for individualized assessment of osteoporosis. In settings without DXA, this model can serve as a useful screening tool to identify high-risk individuals for DXA scan.