Abstract Aims The prognostic value of genetic variants for predicting lethal arrhythmic events (LAEs) in Brugada syndrome (BrS) remains controversial. We investigated whether the functional curation of SCN5A variations improves prognostic predictability. Methods and results Using a heterologous expression system and whole-cell patch clamping, we functionally characterized 22 variants of unknown significance (VUSs) among 55 SCN5A mutations previously curated using in silico prediction algorithms in the Japanese BrS registry (n = 415). According to the loss-of-function (LOF) properties, SCN5A mutation carriers (n = 60) were divided into two groups: LOF-SCN5A mutations and non-LOF SCN5A variations. Functionally proven LOF-SCN5A mutation carriers (n = 45) showed significantly severer electrocardiographic conduction abnormalities and worse prognosis associated with earlier manifestations of LAEs (7.9%/year) than in silico algorithm-predicted SCN5A carriers (5.1%/year) or all BrS probands (2.5%/year). Notably, non-LOF SCN5A variation carriers (n = 15) exhibited no LAEs during the follow-up period. Multivariate analysis demonstrated that only LOF-SCN5A mutations and a history of aborted cardiac arrest were significant predictors of LAEs. Gene-based association studies using whole-exome sequencing data on another independent SCN5A mutation-negative BrS cohort (n = 288) showed no significant enrichment of rare variants in 16 985 genes including 22 non-SCN5A BrS-associated genes as compared with controls (n = 372). Furthermore, rare variations of non-SCN5A BrS-associated genes did not affect LAE-free survival curves. Conclusion In vitro functional validation is key to classifying the pathogenicity of SCN5A VUSs and for risk stratification of genetic predictors of LAEs. Functionally proven LOF-SCN5A mutations are genetic burdens of sudden death in BrS, but evidence for other BrS-associated genes is elusive.