Abstract
3370
Introduction: Myocardial perfusion imaging (MPI) with SPECT or PET is widely utilized to evaluate patients with suspected obstructive coronary artery disease (CAD). The presence of myocardial ischemia on MPI can help physicians target more aggressive interventions; however, the prevalence of ischemia has markedly declined over time. While the likelihood of patients having ischemia on MPI is central to physician decision-making regarding test selection, no contemporary, dedicated prediction scores exist. We developed and validated novel ischemia risk scores to support physician decision making and compared their predictive performance with established risk scores for obstructive CAD.
Methods: The model derivation population included 15,186 and the model validation population included 2,995 patients from a different center. Logistic regression was used to assess associations with the presence of ischemia to derive point-based and calculated ischemia scores, based on standard clinical variables. Predictive performance for the presence of ischemia was assessed using area under the receiver operating characteristic curve (AUC) and calibration assessed with Brier scores. The newly developed ischemia risk scores were compared with established risk scores developed to predict obstructive CAD, the CAD consortium basic and clinical models.
Results: In the derivation population, the AUC for the calculated ischemia risk score (0.80, 95% CI 0.791 – 0.812) was significantly higher compared to the point-based ischemia risk score (0.786, 95% CI 0.775 – 0.798, p<0.001). Both ischemia risk scores had significantly higher predictive performance compared to the basic and clinical CAD models (p<0.001 for all).
The AUC for the calculated ischemia risk score (AUC 0.716, 95% CI 0.684 – 0.748) was significantly higher compared to the point-based ischemia risk score (AUC 0.699, 95% CI 0.666 – 0.732, p=0.016). Both the calculated ischemia risk score and point-based ischemia risk score had significantly higher predictive performance compared to the basic CAD model (AUC 0.628, 95% CI 0.593 – 0.664, p<0.01 for both) and the clinical CAD model (AUC 0.667, 95% CI 0.633 – 0.701, p<0.05 for both). The calculated and point-based ischemia risk scores demonstrated better calibration (Brier score 0.078 and 0.079) compared to the basic and clinical CAD models (Brier scores 0.106 and 0.103)
Conclusions: We developed two novel risk scores for predicting probability of ischemia on MPI which demonstrated high predictive performance during model derivation and in external testing. These scores could support physician decisions regarding diagnostic testing strategies based on likelihood of ischemia.