For a study, researchers sought to find new, repeatable DCM subphenotypes utilizing multiparametric data for better patient classification.

Longitudinal, observational cohorts of DCM patients (enrolled 2009-2016) from the UK (n=426; median age 54 years; 67% men) and the Netherlands (n=239; median age 56 years; 64% men) with clinical, genetic, cardiovascular magnetic resonance, and proteomic analyses. Novel illness subgroups were discovered using machine learning and profile regression. For validation, penalized multinomial logistic regression was utilized. Nested Cox models were used to compare novel groups to traditional risk indicators. Cardiovascular mortality, heart failure, or arrhythmia episodes were the primary composite outcomes (median follow-up 4 years).

Three new DCM subtypes were discovered: profibrotic metabolic, moderate nonfibrotic, and biventricular impairment. The prognosis differed by subtype in both the derivation (P<0.0001) and validation cohorts. Diabetes was more prevalent in the new profibrotic metabolic group, as was global myocardial fibrosis, maintained right ventricular function, and increased creatinine. About 5 factors were adequate for categorization in clinical applications (left and right ventricular end-systolic volumes, left atrial volume, myocardial fibrosis, and creatinine). The addition of the new DCM subtype raised the C-statistic from 0.60 to 0.76. In both the derivation (HR: 3.6; 95% CI: 1.9-6.5; P=0.00002) and validation populations, interleukin-4 receptor-alpha was found as a new prognostic biomarker (HR: 1.94; 95% CI: 1.3-2.8; P=0.00005).

Using publicly accessible clinical and molecular data, three repeatable, mechanistically separate DCM subtypes were discovered, adding predictive significance to existing risk models. They have the potential to enhance patient selection for innovative therapies, enabling precision medicine.