Prediction of acute kidney injury after left ventricular assist device implantation: Evaluation of clinical risk scores

Acute kidney injury (AKI) is frequent in patients scheduled for implantation of a left ventricular assist device (LVAD) and associated with increased mortality. Although several risk models for the prediction of postoperative renal replacement therapy (RRT) have been developed for cardiothoracic patients, none of these scoring systems have been validated in LVAD patients. A retrospective, single center analysis of all patients undergoing LVAD implantation between September 2013 and July 2016 was performed. Primary outcome was AKI requiring RRT within 14 days after surgery. The predictive capacity of the Cleveland Clinic Score (CCS), the Society of Thoracic Surgeons Score (STS), and the Simplified Renal Index Score (SRI) were evaluated. 76 patients underwent LVAD implantation, 19 patients were excluded due to preoperative RRT. RRT was associated with a prolonged ventilation time, length of stay on the ICU and 180 day mortality (14(60.9%) vs 6(17.6%), P < .01). Whereas the Thakar Score (7.43 +/- 1.75 vs 6.44 +/- 1.44, P = .02) and the Mehta Score (28.12 +/- 15.08 vs 21.53 +/- 5.43, P = .02) were significantly higher in patients with RRT than in those without RRT, the SRI did not differ between these groups (3.96 +/- 1.15 vs 3.44 +/- 1.05, P = .08). Using ROC analyses, CCS, STS, and SRI showed moderate predictive capacity for RRT with an AUC of 0.661 +/- 0.073 (P = .040), 0.637 +/- 0.079 (P = .792), and 0.618 +/- 0.075 (P = .764), respectively, with comparable accuracy in the Delong test. Using univariate logistic regression analysis, only the De Ritis Ratio (OR 2.67, P = .034) and MELD (OR 1.11, P = .028) were identified as predictors of postoperative RRT. Risk scores which are predictive in general cardiac surgery cannot predict RRT in patients after LVAD implantation. Therefore, it seems to be necessary to develop a specific risk score for this patient population.