Validation of Seattle Heart Failure Model for mortality risk prediction in patients treated with cardiac resynchronization therapy.

Aims: Survival prediction by the Seattle Heart Failure Model (SHFM) of patients treated with cardiac resynchronization therapy (CRT) remains ill defined. The performance of the SHFM in this clinical setting was therefore evaluated.
Methods and Results: Data from 1309 consecutive CRT patients (five centres) were collected retrospectively; 1139 of these patients were considered for analysis. Three-hundred and seven deaths occurred over 40.1 months (interquartile range 25.2-60.0 months; mean event rate 9.7%/year; survival of 89, 81, and 64% at 1, 2, and 5 years). Kaplan-Meier event-free survival analysis stratified according to tertile of SHFM score was significant (log rank test P < 0.001). High-risk tertile (T1) survival was 82, 67, and 46% at 1, 2, and 5 years, respectively. Observed compared with SHFM-predicted survival was 0.11 vs. 0.08, 0.19 vs. 0.16, and 0.36 vs. 0.36, at 1, 2, and 5 years. Model discrimination by c-statistic was 0.64; the logistic models' area under the receiver operating characteristic curve (AUC-ROC) of risk tertiles was 0.66, 0.68, and 0.67, at 1, 2, and 5 years. Compared with the other two groups, T1 was globally more compromised. Within the T1 group, independent predictors of death were male gender, ischaemic heart failure aetiology, lower body weight, and CRT pacemaker.
Conclusions: SHFM performance was found to be modest, tending to overestimate survival. However, SHFM identified a high-risk, globally more compromised patient subgroup, hence supporting a comprehensive approach, which should include nutritional, metabolic, and immunological aspects, as well as defibrillator back-up.