Many response criteria are poor predictors of outcomes after cardiac resynchronization therapy: validation using data from the randomized trial

Abstract

The aim of the study was to assess the predictive value for outcomes of various response criteria currently used in patients undergoing cardiac resynchronization therapy (CRT). Data from TRUST CRT randomized trial in patients with New York Heart Association (NYHA) III-IV class, QRS ≥ 120 ms, ejection fraction ≤ 35%, and mechanical dyssynchrony was analysed. Ninety-seven subjects who survived 6 months after implantation of CRT-defibrillator were classified as responders or non-responders depending on 15 criteria used in most of the previous trials. Blindly adjudicated data on major adverse cardiac events (MACEs) within 1 year after classification were used to calculate the predictive value of response criteria. After adjustment for baseline confounding variables only eight criteria were significantly predictive for future MACEs. Sensitivity and specificity ranged substantially for clinical (32-94% and 26-63%) and echocardiographic criteria (40-93% and 22-70%, respectively). The most powerful clinical predictor was >a NYHA class reduction ≥ 1 [adjusted relative risk (RR) 4.41 for non-responders; 95% confidence interval (CI) 1.75-11.04, P = 0.002], while the strongest echocardiographic predictor was a reduction in the left ventricular end-systolic index by > 15% (RR 3.49; 95% CI 1.59-7.64, P = 0.002). A combination of these two criteria did not improve the predictive value of a single parameter. Both criteria showed multiple significant interactions with baseline patients' characteristics. Only some of the commonly used response criteria predict outcome in patients undergoing CRT. The predictive value varies substantially across different criteria, with a higher sensitivity observed for the clinical parameters and a higher specificity observed for echocardiographic parameters. Combining various criteria adds little to their prognostic value. The predictive accuracy of various criteria can be different in various subgroups due to multiple interactions with baseline characteristics. CLINICALTRIALS. GOV IDENTIFIER: NCT00814840.