Abstract
BACKGROUND: Data on the predictive value of coronary computed tomography angiography-derived fractional flow reserve (CT-FFR) for long-term outcomes are limited. METHODS AND RESULTS: A retrospective pooled analysis of individual patient data was performed. Deep-learning-based CT-FFR was calculated. All patients enrolled were followed-up for at least 5 years. The primary outcome was major adverse cardiovascular events. The secondary outcome was death or nonfatal myocardial infarction. Predictive abilities for outcomes were compared among 3 models (model 1, constructed using clinical variables; model 2, model 1+coronary computed tomography angiography-derived anatomical parameters; and model 3, model 2+CT-FFR). A total of 2566 patients (median age, 60 [53-65] years; 56.0% men) with coronary artery disease were included. During a median follow-up time of 2197 (2127-2386) days, 237 patients (9.2%) experienced major adverse cardiovascular events. In multivariable-adjusted Cox models, CT-FFR≤0.80 (hazard ratio [HR], 5.05 [95% CI, 3.64-7.01]; P<0.001) exhibited robust predictive value. The discriminant ability was higher in model 2 than in model 1 (Harrell's C-statistics, 0.79 versus 0.64; P<0.001) and was further promoted by adding CT-FFR to model 3 (Harrell's C-statistics, 0.83 versus 0.79; P<0.001). Net reclassification improvement was 0.264 (P<0.001) for model 2 beyond model 1. Of note, compared with model 2, model 3 also exhibited improvement (net reclassification improvement=0.085; P=0.001). As for predicting death or nonfatal myocardial infarction, only incorporating CT-FFR into model 3 showed improved reclassification (net reclassification improvement=0.131; P=0.021). CONCLUSIONS: CT-FFR provides strong and incremental prognostic information for predicting long-term outcomes. The combined models incorporating CT-FFR exhibit modest improvement of prediction abilities, which may aid in risk stratification and decision-making.