Right Ventricular Failure Risk Prediction Models in Patients Undergoing Left Ventricular Assist Device Implantation: A Systematic Review

Purpose Right ventricular failure (RVF) remains a frequent major cause of morbidity and mortality after left ventricular assist device (LVAD) implantation. Timely intervention with planned right ventricular assistance has been shown to improve patient outcomes. Adequate prediction of RVF is important for guiding appropriate selection of LVAD patients and risk prognostication. Methods We systematically searched literature databases in February 2018, as well as the references of included publications. Two reviewers selected studies on adults undergoing LVAD implantation that derived and/or validated a risk prediction model to predict postoperative RVF. We summarized the results of discrimination and calibration qualitatively. Results A total of 23 studies were included, testing 22 models. There were 2 models from large multicenter registries: the RVFRS score from INTERMACS (10 909 patients) and EUROMACS score (2 000 patients). The rest were from single centers. Of the 22 models (Table 1), 7 were externally validated (RVFRS, Penn, Utah, CRITT, Kormos, Pittsburgh, EUROMACS). 6 models were derived from cohorts based on a majority (>50%) of pulsatile LVADs and had poor-to-modest discrimination (AUC between 0.49 to 0.69). Only EUROMACS showed adequate calibration in a validation study. The definition of RVF as an outcome was heterogenous among both derivation and validation studies. Conclusion Existing RVF prediction scores exhibit heterogeneous derivation and validation methodologies, with varying definitions of events and types of devices used. Many of the classically validated scores originate from cohorts on pulsatile first-generation LVADs and were derived from single centers. The majority of models for post-LVAD RVF risk prediction perform poorly to modestly in validation studies. The EUROMACS score reported more promising performance, though modest. Further validation of scores developed from third-generation LVAD cohorts is needed.