Development and external validation of predictive models for prevalent and recurrent atrial fibrillation: a protocol for the analysis of the CATCH ME combined dataset

Abstract

Background: Atrial fibrillation (AF) is caused by different mechanisms but current treatment strategies do not target these mechanisms. Stratified therapy based on mechanistic drivers and biomarkers of AF have the potential to improve AF prevention and management outcomes. We will integrate mechanistic insights with known pathophysiological drivers of AF in models predicting recurrent AF and prevalent AF to test hypotheses related to AF mechanisms and response to rhythm control therapy. Methods: We will harmonise and combine baseline and outcome data from 12 studies collected by six centres from the United Kingdom, Germany, France, Spain, and the Netherlands which assess prevalent AF or recurrent AF. A Delphi process and statistical selection will be used to identify candidate clinical predictors. Prediction models will be developed in patients with AF for AF recurrence and AF-related outcomes, and in patients with or without AF at baseline for prevalent AF. Models will be used to test mechanistic hypotheses and investigate the predictive value of plasma biomarkers. Discussion: This retrospective, harmonised, individual patient data analysis will use information from 12 datasets collected in five European countries. It is envisioned that the outcome of this analysis would provide a greater understanding of the factors associated with recurrent and prevalent AF, potentially allowing development of stratified approaches to prevention and therapy management.