A clinical decision rule to enhance targeted bone mineral density testing in healthy mid-life women

The rates of bone mineral density testing for osteoporosis among healthy mid-life women are high, although their osteoporosis or fracture risk is low. To reduce unnecessary testing, we created and evaluated a tool to guide bone density testing based on the woman's age, weight, fracture history, and menopausal status.

This study aims to improve case finding of mid-life women with low bone mass on bone mineral density (BMD) assessment.

Among healthy women aged 40-60 years having their first BMD test, osteoporosis risk factors were assessed by questionnaire and BMD by dual-energy X-ray absorptiometry. The combination of risk factors that best discriminated women with/without low bone mass (T-score ≤ -2.0) was determined from the logistic regression model area under the curve (AUC) and internally validated using bootstrapping. Using the model odds ratios, a clinical prediction rule was created and its discriminative properties assessed and compared with that of the osteoporosis self-assessment tool (OST). Sensitivity analyses examined results for pre-/peri- and post-menopausal women, separately.

Of 1,664 women referred for baseline BMD testing, 433 with conditions known to be associated with bone loss were excluded. Of 1,231 eligible women, 944 (77%) participated and 87 (9.2%) had low bone mass (35 pre-/peri- and 52 post-menopausal). Four risk factors for low bone mass were identified and incorporated into a clinical prediction rule. Selecting women for BMD testing with weight of ≤70 kg or any two of age >51, years' post-menopause of ≥1, and history of fragility fracture after age 40 was associated with 93% sensitivity to identify women with low bone mass, compared with 47% sensitivity for an OST score of ≤1 (AUC 0.75 versus OST AUC 0.69, p = 0.04). Results restricted to post-menopausal women were similar.

Among healthy mid-life women receiving a baseline BMD test, few had low bone mass, supporting the need for guidance about testing. A prediction rule with four risk factors had improved sensitivity over the OST. Further validation is warranted.