Further validation of a questionnaire to identify women likely to have low bone density

A questionnaire instrument called the Simple Calculated Osteoporosis Risk Estimation (SCORE*) stratifies risk for osteoporosis, potentially reducing population-screening costs. SCORE is calculated using race, weight, age, history of estrogen use, fracture history, and presence/absence of rheumatoid arthritis. We tested SCORE in 912 postmenopausal women aged 45 yr or more using a Hologic QDR4500C densitometer for the total hip, femoral neck, and lumbar spine. National Health and Nutrition Examination Survey (NHANES III) norms were used to calculate hip T-scores. Low bone density (T-score < -2.0) was found in 29.6% of patients at the femoral neck, 17.7% at the total hip, 36.1% at the lumbar spine, and 46.1% at one or more sites. The sensitivity/specificity of the SCORE model (using a threshold of 6 points) was 0.97/0.36 for the femoral neck, 0.95/0.30 for the total hip, 0.86/0.35 for the spine, and 0.88/0.41 for any site (total hip, femoral neck, or spine). When used to detect low bone density at any site, SCORE would have deferred 27.6% of women referred for DXA scans, but 20.7% of these (5.7% of the entire population) would have been false-negatives, and thus inappropriately deferred. At a cutpoint of 3 instead of 6, sensitivity/specificity was 0.96/0.16. In those aged 50-60, the group with the greatest need for risk stratification, sensitivity/specificity for low density at any site was 0.72/0.54, and 46.1% would have been deferred, but 18.5% of this group would have been false-negatives. A cutpoint of 1 in this age group yielded sensitivity/specificity of 0.94/0.16. After age 65, few women would be deferred.

When used to detect low bone density at any site with sufficient sensitivity for clinical practice, SCORE did not have sufficient discriminatory power to be broadly applicable.