Three osteoporosis risk assessment tools underperform in younger postmenopausal women
Key takeaways:
- The USPSTF recommends five tools for assessing osteoporosis risk among postmenopausal women.
- Three tools were graded as poor to fair discriminators of osteoporosis risk for women aged 50 to 64 years.
Three risk assessment tools struggled to detect osteoporosis based on bone mineral density T-scores among women aged 50 to 64 years, researchers reported in a study published in JAMA Network Open.
Researchers assessed the performance of the Osteoporosis Self-Assessment Tool, the Osteoporosis Risk Assessment Instrument and the Osteoporosis Index of Risk, which are three of the five clinical risk assessment tools that the U.S. Preventive Services Task Force recommends for osteoporosis screening, among a group of postmenopausal women younger than aged 65 years. All three tools had an area under the receiver operating curve (AUC) of less than 0.7, which was the threshold established by the researchers for an assessment tool to be an “acceptable” discriminator for osteoporosis.
“Because we found that the three clinical risk assessment tools only had fair to moderate performances in identifying osteoporosis, our results suggest that guidelines that recommend using these tools in clinical practice to identify younger postmenopausal women at risk for osteoporosis should be reassessed,” Henry W. Zheng, BS, a PhD student in medical and imaging informatics at University of California, Los Angeles, told Healio.
Researchers collected data from 6,067 postmenopausal women who participated in the Women’s Health Initiative Bone Density Substudy and were not using osteoporosis medication at baseline (mean age, 57.7 years). BMD was measured during DXA scans. Women with a BMD T-score of –2.5 or lower at the femoral neck, total hip or lumbar spine were defined as having osteoporosis. In order to be an acceptable or better discriminator of osteoporosis, a risk assessment tool needed an AUC of 0.7 or higher.
Of the study group, 14.1% had osteoporosis at any site and 4.9% had osteoporosis at the femoral neck.
All three risk prediction tools had a higher AUC for predicting osteoporosis at the femoral neck vs. any of the three sites. The Osteoporosis Index of Risk had the highest AUC for predicting osteoporosis at the femoral neck (AUC = 0.83; 95% CI, 0.829-0.83) followed by the Osteoporosis Self-Assessment Tool (AUC = 0.818; 95% CI, 0.817-0.819) and the Osteoporosis Risk Assessment Instrument (AUC = 0.805; 95% CI, 0.805-0.806).
The guideline-recommended cutoff for predicting osteoporosis is a score of less than 1 point for the Osteoporosis Index of Risk, a score of more than 8 points for the Osteoporosis Risk Assessment Instrument and a score of less than 2 points for the Osteoporosis Self-Assessment Tool. When those score thresholds were used, the Osteoporosis Index of Risk had a sensitivity of 37.8%, a specificity of 88.8%, a positive predictive value of 35.6% and an AUC of 0.633 (95% CI, 0.633-0.634) for predicting osteoporosis. The Osteoporosis Risk Assessment Instrument had a sensitivity of 53.3%, a specificity of 79.4%, a positive predictive value of 29.8% and an AUC of 0.663 (95% CI, 0.663-0.664). The Osteoporosis Self-Assessment Tool resulted in a sensitivity of 62.4%, a specificity of 68.5%, a positive predictive value of 24.5% and an AUC of 0.654 (95% CI, 0.654-0.655).
All three risk assessment tools were categorized as having poor to fair discrimination with an AUC between 0.5 and 0.7.
Zheng said the findings weren’t surprising because previous research found that another USPSTF-recommended screening tool, the Fracture Risk Assessment Tool, was only modestly better than chance at discriminating between women with and without osteoporosis.
The researchers concluded that the findings “cast some doubt” on the USPSTF’s recommendations for using the three risk assessment tools to predict osteoporosis risk in clinical practice for younger postmenopausal women.
“Because of the disappointing performance of existing recommended clinical risk assessment tools, future studies should examine whether new tools leveraging alternative approaches, such as machine learning, will have advantages for identifying appropriate candidates for osteoporosis screening in younger women,” Zheng said.
For more information:
Henry W. Zheng, BS, can be reached at henryzheng@g.ucla.edu.
Collapse