BACKGROUND: Guidelines recommend fracture risk assessment in postmenopausal women aged 50-64, but the optimal method is unknown.

OBJECTIVES: To compare discrimination and calibration of the Fracture Risk Assessment Tool (FRAX) and Garvan fracture risk calculator for predicting fractures in postmenopausal women aged 50-64 at baseline.

DESIGN: Prospective observational study.

PARTICIPANTS: Sixty-three thousand seven hundred twenty-three postmenopausal women aged 50-64 years participating in the Women's Health Initiative Observational Study and Clinical Trials.

MAIN MEASURES: Incident hip fractures and major osteoporotic fractures (MOF) during 10-year follow-up. Calculated FRAX- and Garvan-predicted hip fracture and MOF fracture probabilities.

KEY RESULTS: The observed 10-year hip fracture probability was 0.3% for women aged 50-54 years (n = 14,768), 0.6% for women aged 55-59 years (n = 22,442), and 1.1% for women aged 60-64 years (n = 25,513). At sensitivity thresholds > /= 80%, specificity of both tools for detecting incident hip fracture during 10 years of follow-up was low: Garvan 30.6% (95% confidence interval [CI] 30.3-31.0%) and FRAX 43.1% (95% CI 42.7-43.5%). At maximal area under the receiver operating characteristic curve (AUC(c), 0.58 for Garvan, 0.65 for FRAX), sensitivity was 16.0% (95% CI 12.7-19.4%) for Garvan and 59.2% (95% CI 54.7-63.7%) for FRAX. At AUC(c) values, sensitivity was lower in African American and Hispanic women than among white women and lower in women aged 50-54 than those 60-64 years old. Observed hip fracture probabilities were similar to FRAX-predicted probabilities but greater than Garvan-predicted probabilities. At AUC(c) values (0.56 for both tools), sensitivity for identifying MOF was also low (range 26.7-46.8%). At AUC(c) values (0.55 for both tools), sensitivity for identifying any clinical fracture ranged from 18.1 to 34.0%.

CONCLUSIONS: In postmenopausal women aged 50-64 years, the FRAX and Garvan fracture risk calculator discriminate poorly between women who do and do not experience fracture during 10-year follow-up. There is no useful threshold for either tool.