Time-dependent prognostic improvement by late gadolinium enhancement in dilated cardiomyopathy.

BACKGROUND: Dilated cardiomyopathy (DCM) represents a major cause of heart failure (HF), but current HF prediction models lack validation in DCM cohorts. Late gadolinium enhancement (LGE) predicts mortality in DCM patients. The incremental value of LGE to existing models warrants exploration. METHODS: In this single-center cohort study, hospitalized patients with DCM who underwent cardiac magnetic resonance (CMR) between June 2012 and August 2020 were included. We externally validated five published HF models as follows: SHFM, GISSI-HF, MAGGIC, BIOSTAT-CHF, and PREDICT-HF. The composite endpoints were all-cause mortality and heart transplantation. We then developed three LGE-enhanced models by integrating LGE presence, location, and patterns with risk scores from the published models, respectively. RESULTS: Of 524 DCM patients (age 48.7±15.1, 71% male), 154 patients (29.4%) reached the composite endpoint (median follow-up of 47.6 months). PREDICT-HF showed the best overall discrimination (Harrell's C index: 0.73, 95% CI: 0.69-0.77), similar to SHFM (0.71, 95% CI： 0.67-0.75), better than MAGGIC (0.68, 95% CI: 0.64-0.73), GISSI-HF (0.65, 95% CI: 0.60-0.70), and BIOSTAT-CHF (0.66, 95% CI: 0.61-0.71). Using time-dependent C index, incorporating LGE presence, location, or pattern improved overall discrimination in all LGE-enhanced models, and for 3- and 4-year predictions (all P<0.05), but not in 1- and 2-year predictions. LGE-enhanced models with LGE presence and location yielded similar findings. CONCLUSION: Existing HF models, primarily utilizing clinical variables, moderately predict outcomes in DCM. Adding LGE improves long-term mortality prediction accuracy but not short-term efficacy. REGISTRATION: URL: https://www. CLINICALTRIALS: gov; Unique identifier: ChiCTR1800017058.