The xDAPT External Validation Study

Overview

Dual antiplatelet therapy (DAPT) is routinely recommended after percutaneous coronary intervention (PCI) with drug-eluting stent (DES) implantation to prevent thrombotic complications. However, DAPT is also associated with an increased risk of bleeding, which may have a similar or even greater impact on prognosis compared to recurrent ischemic events. To balance these risks, individualized risk stratification at the time of PCI is crucial for determining the optimal DAPT composition and duration, aiming to reduce thrombotic risk while minimizing bleeding complications. For this purpose, an artificial intelligence-based risk stratification tool (xDAPT, Abbott) was introduced and demonstrated strong clinical performance in its development study (ClinicalTrials.gov identifier: NCT06089304). This analysis aims to evaluate the performance of xDAPT in a real-world cohort of patients who underwent PCI over the past decade at a large urban center (Mount Sinai Hospital, New York).

While dual antiplatelet therapy (DAPT) is recommended after percutaneous coronary intervention (PCI) with drug-eluting stent (DES) implantation to prevent thrombotic complications, it is notably associated with an increased risk of bleeding. Recent evidence suggests that bleeding events occurring early after PCI have a prognostic impact comparable to or even greater than that of recurrent ischemic events. Currently, decisions regarding DAPT duration and composition after PCI are guided by several risk scores that classify patients as having a high bleeding and/or high ischemic risk based on predefined clinical or angiographic factors. However, the predictive performance of these scores is suboptimal, primarily due to the limitations of the analytical approaches used in their development, which typically rely on linear models incapable of capturing the complex interplay of multiple clinical variables. Machine learning (ML) methods offer the potential to address these limitations by leveraging algorithms to analyze large datasets and identify high-dimensional, non-linear relationships among variables. The xDAPT (Abbott), is a recently developed ML-based tool consisting of two separate random forest survival models for predicting ischemic and bleeding risks, respectively (ClinicalTrials.gov identifier: NCT06089304). Each model incorporates 11 clinical variables identified as the most relevant predictors for ischemic and bleeding events. The xDAPT model was developed and internally validated using a pooled dataset of 11 clinical trials on the XIENCE stent, including approximately 19,000 patients who underwent PCI with an everolimus-eluting stent (XIENCE, Abbott) across 21 countries between 2008 and 2020. Within the test cohort of this dataset, both ischemic and bleeding risk models demonstrated good discriminatory ability, achieving a C-index of ≥0.65 for the prediction of their respective outcomes. However, the generalizability of the xDAPT tool for routine clinical practice remains to be established, as it has not yet been validated in an independent real-world population of patients receiving PCI with various DES types. The present study aims to externally validate the ischemic and bleeding risk models of xDAPT using data from consecutive patients who underwent PCI at a large urban hospital (Mount Sinai, New York, US) between 2012 and 2022. Consistent with the internal validation analysis, the performance goal for the model will be defined as achieving a C-index of ≥0.65 at the lower 97.5% confidence interval of the bootstrap C-index distribution.