Before the break: a near-term osteoporotic fracture risk score from opportunistic CT (RAVEN)

PURPOSE: Existing fracture-prediction tools primarily estimate long-term fracture probability and may have limited utility for identifying patients at high near-term risk. Opportunistic computed tomography (CT) provides routinely available Hounsfield unit (HU) attenuation measurements associated with bone mineral density and fracture risk. We aimed to develop and evaluate an exploratory risk-stratification score for first documented subsequent fragility fracture over 1-3 years. METHODS: We conducted a retrospective cohort study of 240 consecutive patients identified through a dedicated osteoporosis clinic between 2021 and 2023. Time to event was calculated from the index CT to the first documented subsequent fragility-fracture episode. Patients without a subsequent fracture were censored at their last osteoporosis-clinic or imaging assessment documenting no new fracture. Multiple fractures arising from the same fall or first fracture episode were counted as one patient-level event. RAVEN (Radiodensity, Age, Vertebral fracture history, Extravertebral fracture history, Nottingham) was derived using subsequent fracture occurrence during observed follow-up as the original binary outcome. The completed score was subsequently evaluated for time to first documented subsequent fracture using Cox regression, Kaplan-Meier estimates, Harrell's C-index, time-dependent AUC and IPCW Brier scores. RESULTS: During 760.1 person-years at risk, 91/240 patients (37.9%) sustained a first documented subsequent fragility fracture. Each one-point increase in RAVEN was associated with a higher fracture hazard (HR 1.14, 95% CI 1.08-1.20; p < 0.001), with no evidence of non-proportional hazards (p = 0.188). Harrell's C-index was 0.636 (bootstrap 95% CI 0.564-0.709). Kaplan-Meier curves differed significantly across RAVEN categories (log-rank p < 0.001). Estimated cumulative fracture risks at 1, 2 and 3 years were 6.8%, 9.7% and 9.7% for scores < 4; 9.9%, 24.3% and 32.5% for scores 4-11; and 32.0%, 41.7% and 47.5% for scores ≥ 12. Time-dependent AUCs at these horizons were 0.636, 0.620 and 0.642, respectively. CONCLUSION: RAVEN demonstrated graded stratification of time to first documented subsequent fragility fracture using opportunistic CT attenuation measurements and clinical variables. These findings support its potential for 1- to 3-year fracture-risk stratification, although external validation and prospective clinical-impact assessment are required before clinical use.