Use of a multistate model in survival predictions in cardiology studies

ABSTRACT Introduction and objectives: Multistate models have proven to be effective tools in survival analyses. We propose modeling disease progression in interventional cardiology studies using a multistate model. Methods: The model was fitted to the PACO-PCI database including a total of 1057 elderly patients with atrial fibrillation revascularized with drug-eluting stents to assess the efficacy profile and prognosis of different antithrombotic therapies. The model defines a total of 4 states: treatment, myocardial infarction and/or revascularization, bleeding, and death, with significant factors for each transition, and was compared using a multivariate Cox model. Results: Survival factors common to both analyses were the PreciseDAPT and HAS-BLED scales, anemia, diabetes mellitus, chronic kidney disease, number of vessels treated, and left ventricular function. The multistate model also shows that after a new hemorrhage the probability of myocardial infarction and/or revascularization is influenced by the treatment of left main coronary artery disease and the transition to death from previous coronary artery bypass graft. Compared with Cox models, multistate models allow us to tell which transition in the model is influenced by each predictor. Conclusions: The results illustrate the additional advantages of multistate models in survival analyses through individual predictions for the patients based on their clinical characteristics and disease progression.