Validation of a simple risk stratification tool for patients implanted with Cardiac Resynchronization Therapy: the VALID-CRT risk score
Résumé
AIMS: Mortality after cardiac resynchronization therapy (CRT) is difficult to predict. We sought to design and validate a simple prognostic score for patients implanted with CRT, based on readily available clinical variables, including age, gender, left ventricular ejection fraction (LVEF), New York Heart Association (NYHA) class, presence/absence of atrial fibrillation, presence/absence of atrioventricular junction ablation, coronary heart disease, diabetes, and implantation of a CRT device with defibrillation. METHODS: For predictive modelling, 5153 consecutive patients enrolled in 72 European centres (79% male; LVEF 25.9 ± 6.85%; NYHA class III-IV 77.5%; QRS 158.4 ± 32.3 ms) were randomly split into derivation (70%) and validation (30%) samples. The primary endpoint was total mortality and the secondary endpoint was cardiovascular mortality. The final predictive model fit was assessed by plotting observed vs. predicted survival. RESULTS: In the entire cohort, 1004 deaths occurred over a follow-up of 14 409 person years. Total mortality ranged from 3.1% to 28.2% at 2 years in the first and fifth quintile of the risk score, respectively. At 5 years, total mortality was 10.3%, 18.6%, 27.6%, 36.1%, and 58.8%, from the first to the fifth quintile. Compared with the lowest quintile (Q), total mortality was significantly higher in the other four quintiles [Q2 hazard ratio (HR) = 1.71; Q3 HR = 2.20; Q4 HR = 4.03; Q5 HR = 8.03; all P \textless 0.001). The final model, which was based on the entire cohort using the above variables, showed a good discrimination (Harrell's c = 0.70) and high explained variation (0.26). The mean predicted survival fitted well with the observed survival for up to 6 years of follow-up. CONCLUSIONS: The VALID-CRT risk score, which is based on routine, readily available clinical variables, reliably predicted the long-term total and cardiovascular mortality in patients undergoing CRT. While this score cannot be used to predict the benefit of CRT, it may be useful for predicting survival after CRT. This may have useful implications for follow-up