Echocardiographic predictors of right ventricular failure following left ventricular assist device implantation: Systematic review and meta-analysis.

Abstract

Introduction:  Right ventricular failure (RVF) is a significant complication following left ventricular assist device (LVAD) implantation, with no universally accepted predictors. This meta-analysis identifies the most reliable echocardiographic predictors.

Methods: OVID Medline was systematically searched for observational studies reporting ten preoperative echocardiographic parameters in patients who did and did not develop RVF post-LVAD. Random-effects meta-analyses were performed. Subgroup analyses and meta-regression assessed the influence of RVF definitions and clinical characteristics on predictive capacity. Logistic regression modeling identified key cutoffs.

Results: Thirty-nine studies involving 2,975 patients were included, with pooled RVF prevalence of 0.30 (0.26-0.34). Higher right ventricular end-diastolic diameter (RVEDD) (SMD: 0.368, p < 0.0001, I²: 1.73%) and less negative right ventricular free wall strain (RVFWS) (SMD: 0.931, p < 0.0001, I²: 82.9%) were significant predictors. Higher right ventricular end-diastolic area (RVEDA) was also reliable but weaker (SMD: 0.224, p = 0.0282, I²: 0.00%). Lower tricuspid annular plane systolic excursion (TAPSE) was a strong predictor (SMD: -0.512, p < 0.0001,) but less reliable due to high heterogeneity (I²: 86.1%). Subgroup analyses of TAPSE by RVF definition showed modestly reduced heterogeneity (43.7%). Predictive capacity was significantly better in more stable patients (i.e., no inotropes/IABP, higher INTERMACS status). Logistic regression identified increased RVF risk at RVEDD ≥41.5 mm (sensitivity: 50.0%, specificity: 83.3%) and RVFWS ≥ -11.3% (sensitivity: 88.9%, specificity: 88.9%).  Conclusion: RVEDD ≥ 41.5 mm and RVFWS ≥ -11.3% are the strongest, most reliable predictors of RVF post-LVAD. Variations in RVF endpoint definitions only partially explain the observed heterogeneity, while patient characteristics significantly influence predictive accuracy. Future studies should explore subgroup-specific cutoffs.