Modes of death and prognostic outliers in chronic heart failure

Background The impact of incident sudden cardiac death (SCD) on the predictive accuracy of prognostic risk scores for patients with chronic heart failure (HF) has rarely been examined. We assessed the relationship between estimated probability of death and modes of death in this population, as well as the predictors of death and survival in prognostic outliers. Methods and Results The MAGGIC 3-year probability of death was estimated in 6,859 participants of the GISSI-HF trial (mean age 67 +/- 11 years, 78% men, 91% with ejection fraction <40%, mean follow-up 3.5 +/- 1.3 years, observed mortality 28.4%). The incidence of SCD progressively decreased with increased probability of death, and occurred in 52.5% of patients estimated at low-risk (N = 61 with probability <14%) vs. in 23.5% of the high-risk ones (N = 375 with probability >56%, P < .0001). On the contrary, death from worsening HF was significantly more frequent in the latter group (19.7% vs. 46.1%, P < .0001). The overall predictive accuracy of the MAGGIC model improved after excluding deaths from SCD (AUC from 0.731 to 0.760, P = .0034). Among patients estimated at low-risk (N = 61 dead, 743 alive), independent predictors of death were older age, longer history of HF, higher serum uric acid and chronic obstructive pulmonary disease. The only predictor of survival in patients estimated at high-risk (N = 210 alive, 375 dead) was higher systolic blood pressure. Conclusions The MAGGIC risk score demonstrated its scarce ability to capture SCD, particularly in chronic HF patients estimated at low risk of death. Newer and better prognostic tools in the evolving horizon of HF are needed.