Multiparametric models for predicting major arrhythmic events in Brugada syndrome: a systematic review and critical appraisal

Despite several risk models to predict major arrhythmic events (MAE) in Brugada syndrome (BrS) having been developed, reproducibility and methodology remain a concern. Our aim was to assess the quality of model development and validation, and determine the discriminative performance of available models.

Electronic databases (Medline, Embase, and Central) were searched through September/2024 for studies developing or validating multivariable prediction models for MAE in BrS. Methodological quality and risk of bias (RoB) were assessed using the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist and the Prediction Model Risk of Bias Assessment (PROBAST) Tool. Pooled random-effects c-statistics were obtained for each model. A total of 16 studies, including 11 unique multivariable scores, were included. All models had domains classified as high RoB. Common sources of bias were inappropriate inclusion/exclusion criteria, predictor selection, low number of events and underreporting of performance measures. Pooled c-statistics among patients without previous MAE showed good performance for Brugada-Risk [AUC 0.81, 95% confidence interval (CI) 0.71–0.91; I² 64%; three studies], fair for PAT (AUC 0.79, 95% CI 0.45–1.12; I² 95%; two studies), Delise (AUC 0.77, 95% CI 0.72–0.81, I² 39%, three studies), and Sieira (AUC 0.73, 95% CI 0.64–0.82; I² 64%; five studies), and moderate for Shanghai (AUC 0.69, 95% CI 0.61–0,76; I² 13%; three studies).

Currently available multiparametric models for prediction of MAE in BrS have important shortcomings in model development and inadequate evaluation. Further validation of current models in external cohorts is required before safe transition to clinical practice.