Cluster analysis of preoperative echocardiographic findings and outcomes following left ventricular device implantation.

Abstract

To investigate whether preoperative echocardiography findings determine postoperative continuous-flow left ventricular assist device outcomes. From January 2003 to June 2017, 490 patients received a durable, continuous-flow left ventricular assist device. Two-step clustering of parameters including heart rate and preoperative echocardiographic findings (ie, left ventricular [LV] ejection fraction, right ventricular [RV] function, aortic insufficiency, mitral regurgitation [MR], tricuspid regurgitation [TR]) was performed and identified 5 distinct clusters associated with LV failure: group 1: moderate right ventricular dysfunction (RVD), severe MR and mild TR (n = 110); group 2: severe RVD, severe MR and TR (n = 64); group 3: moderate RVD and severe aortic insufficiency (n = 16); group 4: mild RVD and mild valvular pathology (n = 163); and group 5: moderate-severe RVD and mild valvular pathology (n = 137). Silhouette measure of cohesion and separation demonstrated satisfactory separation at 0.6. Group 2 had the greatest Interagency Registry for Mechanically Assisted Circulatory Support Level 1 (25%, P = .010), preoperative right atrial pressure (11 ± 5 mm Hg, P < .001), incidence of postoperative right ventricular failure (RVF; 20%, P = .001), delayed closure of the sternum (61%, P = .002), postoperative permanent dialysis (6%, P = .04), rate of tricuspid valve repair (n = 52; 81%, P < .001), and lowest RV stroke work index (489 ± 228 cc mm Hg/m²/beat, P < .001). RVF in groups 1, 3, 4, and 5 was 6%, 0%, 4%, and 9%, respectively. No differences in incidence of heart transplantation (P = .400) or survival (P = .535) were found. Severe TR predicted RVF in those with moderate-severe preoperative RVD (P = .001, odds ratio 3.9). Clustering demonstrated the importance of preoperative TR in predicting RVF. Combined severe LV and RV failure with severe MR and TR portends the worse prognosis.