Clinical parameters, such as anatomical features, haemodynamic sequelae, or acquired complications, were independent predictors of excess mortality in adults with CHD. Survival of individuals with no risk factors did not differ from the reference population.

No abstract

No abstract

ObjectivesTo develop, calibrate, test and validate a logistic regression model for accurate risk prediction of sudden cardiac death (SCD) and non-fatal sudden cardiac arrest (SCA) in adults with congenital heart disease (ACHD), based on baseline lesion-specific risk stratification and individual’s characteristics, to guide primary prevention strategies.MethodsWe combined data from a single-centre cohort of 3311 consecutive ACHD patients (50% male) at 25-year follow-up with 71 events (53 SCD and 18 non-fatal SCA) and a multicentre case–control group with 207 cases (110 SCD and 97 non-fatal SCA) and 2287 consecutive controls (50% males). Cumulative incidences of events up to 20 years for specific lesions were determined in the prospective cohort. Risk model and its 5-year risk predictions were derived by logistic regression modelling, using separate development (18 centres: 144 cases and 1501 controls) and validation (two centres: 63 cases and 786 controls) datasets.ResultsAccording to the combined SCD/SCA cumulative 20 years incidence, a lesion-specific stratification into four clusters—very-low (<1%), low (1%–4%), moderate (4%–12%) and high (>12%)—was built. Multivariable predictors were lesion-specific cluster, young age, male sex, unexplained syncope, ischaemic heart disease, non-life threatening ventricular arrhythmias, QRS duration and ventricular systolic dysfunction or hypertrophy. The model very accurately discriminated (C-index 0.91; 95% CI 0.88 to 0.94) and calibrated (p=0.3 for observed vs expected proportions) in the validation dataset. Compared with current guidelines approach, sensitivity increases 29% with less than 1% change in specificity.ConclusionsPredicting the risk of SCD/SCA in ACHD can be significantly improved using a baseline lesion-specific stratification and simple clinical variables.