From genes to outcomes: the prognostic role of genetic mutations in dilated cardiomyopathy

Abstract Background Dilated cardiomyopathy (DCM) is a heterogeneous condition with varying disease progression rates. While imaging metrics like left ventricular ejection fraction (LVEF) are established prognostic tools, the role of genetic testing in predicting outcomes remains unclear. Variants of uncertain significance (VUS) and pathogenic/likely pathogenic (P/LP) mutations are found in a significant proportion of DCM patients, but their prognostic value is not well-defined. Objective To evaluate prognostic value of genetic findings, clinical, and imaging characteristics in predicting adverse cardiovascular (CV) outcomes in DCM patients. Methods We conducted a single-center, retrospective study of 137 DCM patients who underwent genetic testing between 2018 and 2024. Data were collected on demographics, clinical history, imaging parameters (echocardiogram and cardiac MRI), and genetic testing (gene negative, VUS, or P/LP variant). The primary endpoint was a composite of CV events, including heart failure admission, malignant arrhythmia (ventricular tachycardia/fibrillation), cardiac syncope, cardiovascular death, myocardial infarction, and/or ischemic stroke. Two predictive models were employed to assess the impact of clinical and genetic variables on events: (1) logistic regression and (2) Random Forest. Performance metrics, including accuracy and area under the receiver operating characteristic curve (AUC), were calculated for both models using a train-test split (80% training, 20% testing). Results A total of 119 patients were suitable for analysis (mean age 60±13 years, 65% male); 55.5% were gene positive – 46.2% had at least one VUS, and 9.3% had P/LP (TTN was the most common gene). The primary outcome was met in 68.1%. In logistic regression, age (coefficient: -0.03, p = 0.09) and imaging variables, including LVEF (coefficient: -0.05, p = 0.01), were the strongest predictors of adverse outcomes. Genetic category was not statistically significant (coefficient: -0.63, p = 0.06). The rate of adverse events increased with more VUSs present: no VUS 69%, 1 VUS 65%, 2 VUS 62%, 3 VUS 86%, 5 VUS 100%, but this trend did not achieve statistical significance (coefficient: 0.31, p = 0.20). In the Random Forest analysis, LVEF accounted for 40% of total feature importance, followed by LV end-diastolic volumes (25%) and age (21%). Genetic data, including VUS count (6%) and genetic category (5%), contributed minimally to the model’s predictive performance. The Random Forest model significantly outperformed logistic regression, with an accuracy of 77.8% and an AUC score of 69.5%, compared to logistic regression’s accuracy of 61.1% and AUC score of 55.8%. Conclusion In this cohort of DCM patients, genetic variables such as P/LP variants and the number of VUS had limited prognostic value. Data limitations prevented assessing specific high-risk genotypes, which could impact outcomes.