One DXA scan limits the apparent accuracy of osteoporosis screening tools but explains at most 13% of what FDA-authorized radiograph devices miss
Year: 2026
Type: preprint
Abstract: Purpose: Opportunistic screening tools, including three radiograph devices authorized by the US Food and Drug Administration (FDA), flag patients with low bone mineral density (BMD) for dual-energy X-ray absorptiometry (DXA), but are judged against one imperfect DXA scan. We asked how accurate a perfect classifier can look against one scan and whether this explains the devices' low sensitivity. Methods: From nationally representative US DXA data we estimated precision-error-free BMD and derived the single-scan frontier: the highest apparent sensitivity any test with errors independent of the scan's can show at each false-positive rate. We validated it in 515 repeat-scan pairs, split device shortfalls into reference and device parts, compared two corrections in 194 simulated scenarios and audited 228 published studies. Results: At national-survey precision, a perfect classifier showed apparent sensitivity of 0.695–0.945 across 24 strata, consistent with all 12 repeat-scan measures. In nine FDA strata, device sensitivity was 0.22–0.79 where the frontier was 0.90–1.00. Per 1,000 people screened, the devices missed 49–420 with DXA-defined low BMD, of whom DXA precision accounted for at most 27 (≤13% of the shortfall at guideline-minimum precision; 17% at national-survey precision). Device results flipped on repetition 2–5 times as often as DXA's would. Conditional-independence correction overstated sensitivity; latent-trait correction was unbiased given known precision and a correct model. Only 13 of 136 full-text studies mentioned DXA precision. Conclusion: One DXA scan makes even a perfect classifier look imperfect but explains little of the authorized devices' low sensitivity, so a negative device result should not replace DXA.
Source: Research Square
Author: Yunhao Jiang
Institution: University of Edinburgh
Language: English
Subfield: Orthopedics and Sports Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7225619958
Cites:
Cited by:
Related to: