Abstract 12641: Performance of Mortality Prediction Models in Dilated Cardiomyopathy Patients and the Incremental Value of Late Gadolinium Enhancement

Abstract

Introduction: Mortality prediction models assist decision-making for patients with dilated cardiomyopathy (DCM). Published heart failure (HF) models have not been validated in large DCM cohorts; whether late gadolinium enhancement (LGE) provides incremental predictive power remains unclear. Hypothesis: Incorporating LGE location improves the performance of existing HF mortality prediction models. Methods: We used data from a prospective registry enrolling DCM patient who underwent clinical CMR examination between June 2012 and May 2020. Five recent or classic models estimating the risk of all-cause death or heart transplantation in chronic HF patients were examined: SHFM, GISSI-HF, MAGGIC, BIOSTAT-CHF, and PARADIGM-HF score. We added LGE location to the risk scores generated by the original models. Results: Of 524 DCM patients with available data required for all five models (age 48.7 ± 15.1, 71% male), 100 patients (19.1%) died or underwent heart transplantation (median follow-up of 26.3 months). PARADIGM-HF showed the best overall discrimination [Harrell’s C index and 95% CI: 0.75, 0.70-0.80], similar to SHFM and MAGGIC [0.72, 0.67-0.77 and 0.71, 0.66-0.76, respectively; all comparisons P >0.05], and better than GISSI-HF and BIOSTAT-CHF [0.66, 0.60-0.71 and 0.66, 0.60-0.72, respectively]. PARADIGM-HF exhibited decreased discrimination for 2-year and 3-year mortality/heart transplantation risk. The net reclassification by adding LGE location to the models improved the 2-year and 3-year risk prediction in 25-29% of patients, except for MAGGIC; however, 1-year risk prediction remained unaltered. Conclusions: Existing HF mortality prediction models made moderately accurate risk estimates. LGE assisted long-term risk stratification but may have less additive value for the short-term risk prediction.