A bio-clinical approach for prediction of sudden cardiac death in outpatients with heart failure: The ST2-SCD score

Background Sudden cardiac death (SCD) is one of the main modes of death in heart failure (HF) patients and its prediction remains a real challenge. Our aim was to assess the incidence of SCD at 5 years HF contemporary managed outpatients, and to find a simple prediction model for SCD. Methods SCD was considered any unexpected death, witnessed or not, occurring in a previously stable patient with no evidence of worsening HF or any other cause of death. A competing risk strategy was adopted using the Fine-Gray method of Cox regressions analyses that considered other causes of death as the competing event. Results The derivation cohort included 744 consecutive outpatients (72% men, age 67.9 ± 12.2 years, left ventricular ejection fraction [LVEF] 36% ± 14). During follow-up, 312 deaths occurred, 40 SCDs (5.4%). Age, haemoglobin, eGFR, HF duration, high-sensitivity troponin T, NTproBNP, and ST2 were associated with SCD in univariate analyses; HF duration (p = 0.006), eGFR (p < 0.001), LVEF <45% (p = 0.03), and ST2 (p = 0.006) remained in multivariable analysis. A predictive score (ST2-SCD) including dichotomous variables (ST2 > 45, LVEF <45%, HF duration >3 years, eGFR < 55, age ≥ 60 years and male sex) provided a Harrell's C-statistic of 0.82 (0.76-0.89)), reaching 0.87 (0.80–0.95) in the validation cohort (n = 149). Conclusions In contemporary managed HF, SCD occurred in 5.4% of outpatients, accounting for 12.8% of all deaths at 5 years. Of the 3 studied biomarkers, only ST2 remained independently associated with SCD. A model containing age, sex, ST2, eGFR, LVEF, and HF duration reasonably predicted 5-years risk of SCD.