Computed Tomography-Based Body Composition Assessment for Preoperative Cardiovascular Risk Prediction: A Prospective Cohort Study

BACKGROUND: Current approaches for preoperative cardiovascular risk prediction remain suboptimal. CT-derived body composition metrics may provide objective markers of cardiometabolic health, yet their predictive value for postoperative cardiovascular events remains unclear. METHODS: We included patients with cardiovascular disease or risk factors undergoing major noncardiac surgery in the prospective, multicenter PREVENGE-CB (PREdiction of Vascular Events after Noncardiac surGEry with Cardiac Biomarkers) cohort and its Nanfang extension. Preoperative abdominal CT scans were analyzed to quantify the area and radiodensity of skeletal muscle and adipose tissues at the third lumbar vertebral level. The primary outcome was composite cardiovascular events within 30 days after surgery. We used logistic regression models to evaluate the added predictive value of body composition metrics beyond guideline-recommended predictors. In the Nanfang cohort, the optimal subset of body composition metrics was selected by minimizing the Akaike Information Criterion for the primary outcome. Nested models were compared using measures of model fit, discrimination, risk reclassification, and net benefit. The findings were validated in the PREVENGE-CB cohort. RESULTS: Among 1594 patients, 211 (13.2%) had the primary outcome. Larger skeletal muscle and adipose areas were generally associated with lower risk, whereas higher adipose radiodensity and lower muscle radiodensity indicated higher risk. In the Nanfang cohort, an optimal subset of three body composition metrics-skeletal muscle area, muscle radiodensity and subcutaneous fat radiodensity-improved discrimination of the primary outcome over the Revised Cardiac Risk Index (increase in area under the curve [ΔAUC] = 0.136; 95% CI, 0.083 to 0.188), the Gupta Myocardial Infarction and Cardiac Arrest risk calculator (ΔAUC = 0.032; -0.002 to 0.065), and a refitted clinical model (ΔAUC = 0.035; 0.008 to 0.062). These findings were validated in the PREVENGE-CB cohort. CONCLUSIONS: CT-derived body composition metrics improved prediction of postoperative cardiovascular events beyond conventional clinical predictors.