Incremental prognostic value of left atrial and biventricular feature tracking in dilated cardiomyopathy: a long-term study

Background Despite the use of cardiovascular magnetic resonance (CMR) feature tracking (FT) imaging to detect myocardial deformation, the optimal strain index in dilated cardiomyopathy (DCM) is unclear. This study aimed to determine whether atrial and biventricular strains can provide the greatest or joint incremental prognostic value in patients with DCM over a long follow-up period.Methods Four hundred-twelve DCM patients were included retrospectively. Comprehensive clinical evaluation and imaging investigations were obtained, including measurements of CMR-FT derived left atrial (LA) reservoir, conduit, booster strain (epsilon s, epsilon e, epsilon a); left ventricular (LV) and right ventricular (RV) global longitudinal, radial, circumferential strain (GLS, GRS, GCS). All patients were followed up for major adverse cardiac events (MACE) including all-cause mortality, heart transplantation, and implantable cardioverter defibrillator discharge. The predictors of MACE were examined with univariable and multivariable Cox regression analysis. Subsequently, nested Cox regression models were built to evaluate the incremental prognostic value of strain parameters. The incremental predictive power of strain parameters was assessed by Omnibus tests, and the model performance and discrimination were evaluated by Harrell C-index and integrated discrimination improvement (IDI) analysis. Patient survival was illustrated by Kaplan-Meier curves and differences were evaluated by log-rank test.Results During a median follow-up of 5.0 years, MACE were identified in 149 (36%) patients. LA epsilon e, LVGLS, and RVGLS were the most predictive strain parameters for MACE (AUC: 0.854, 0.733, 0.733, respectively). Cox regression models showed that the predictive value of LA epsilon e was independent from and incremental to LVGLS, RVGLS, and baseline variables (HR 0.74, 95% CI 0.68-0.81, P < 0.001). In reclassification analysis, the addition of LA epsilon e provided the best discrimination of the model (chi 2 223.34, P < 0.001; C-index 0.833; IDI 0.090, P < 0.001) compared with LVGLS and RVGLS models. Moreover, LA epsilon e with a cutoff of 5.3% further discriminated the survival probability in subgroups of patients with positive LGE or reduced LVEF (all log-rank P < 0.001).Conclusion LA epsilon e provided the best prognostic value over biventricular strains and added incremental value to conventional clinical predictors for patients with DCM.