Performance of the American Heart Association/American College of Cardiology Guideline-Recommended Pretest Probability Model for the Diagnosis of Obstructive Coronary Artery Disease

Background Substantial differences exist between different guideline-recommended pretest probability (PTP) models for the detection of obstructive coronary artery disease (CAD). This study was performed to study the performance of the 2021 American Heart Association/American College of Cardiology (AHA/ACC) guideline-recommended PTP (AHA/ACC-PTP) model in assessing the likelihood of obstructive CAD compared with previously proposed models. Methods and Results Symptomatic patients (N=50 561) referred for coronary computed tomography angiography were included. The reference standard was invasive coronary angiography with optional fractional flow reserve measurements. The AHA/ACC-PTP values based on sex and age were calculated and compared with the 2019 European Society of Cardiology guideline PTP values based on sex, age, and symptoms as well as the risk factor-weighted clinical likelihood values based on sex, age, symptoms, and risk factors. The AHA/ACC-PTP maximum values overestimated by a factor of 2.6 the actual prevalence of CAD. Compared with the AHA/ACC-PTP model (area under the receiver-operating curve, 71.5 [95% CI, 70.7-72.2]), inclusion of typicality of symptoms in the European Society of Cardiology guideline PTP improved discrimination of CAD (area under the receiver-operating curve, 75.5 [95% CI, 74.7-76.3]). Inclusion of both symptoms and risk factors in the risk factor-weighted clinical likelihood model further improved discrimination (area under the receiver-operating curve, 77.7 [95% CI, 77.0-78.5]). The proportion of patients classified as very low PTP was lower using the AHA/ACC-PTP (5%) compared with the European Society of Cardiology guideline PTP (19%) and the risk factor-weighted clinical likelihood (49%) models. Conclusions The new AHA/ACC-PTP model overestimates the prevalence of obstructive CAD substantially if type of symptoms and risk factors are not taken into account. Inclusion of both symptoms and risk factors improves model performance and identifies more patients with very low likelihood of CAD in whom further testing can be deferred.