Changes in Blood Eosinophil Counts Predict the Death of Patients With Myocardial Infarction After Hospital Discharge.

Background: Preclinical studies demonstrate a cardioprotective role of eosinophils in acute myocardial infarction. Yet clinical studies show conflicting correlations between blood eosinophil counts and acute myocardial infarction risk and mortality. This study evaluates blood eosinophil counts of patients with acute myocardial infarction at hospital admission (EOS) and discharge (EOS) on all-cause and cardiac mortalities.

Methods And Results: Of 2681 consecutive patients with a median follow-up of 2.55 years, 45 patients died within 30 days, 28 died within 30 to 150 days, and 92 died within 150 days or later postdischarge. Cardiac death occurred in 108 patients. According to the receiver operating characteristic analyses, the best cutoffs of EOS and EOS were 0.02×10/L and 0.03×10/L, respectively, to predict 30-day all-cause death (area under the curve [AUC], 0.60; AUC, 0.67). The optimal cutoffs of EOS and EOS were 0.20×10/L and 0.14×10/L to predict long-term all-cause mortality (1-year AUC, 0.59; 1-year AUC, 0.61). Multivariate logistic analysis showed that low EOS (<0.02×10/L) or low EOS (<0.03×10/L) predicted the 30-day all-cause (odds ratio [OR], 2.56; =0.005; OR, 8.14; <0.001) and cardiac (OR, 2.16; =0.025; OR, 7.89; <0.001) mortalities. Patients with combined low EOS (<0.02×10/L) and low EOS (<0.03×10/L) displayed synergistic risk of 30-day all-cause (OR, 13.93; <0.001) and cardiac (OR, 11.38; <0.001) deaths. In contrast, adjusted Cox proportional hazard test indicated that high EOS (≥0.14×10/L) was an independent risk for long-term all-cause mortality (hazard ratio, 1.84; =0.010).

Conclusions: High and low blood eosinophil counts both predict the risk of all-cause and cardiac deaths in patients with acute myocardial infarction depending on the time of data collection. Dynamic changes of blood eosinophil counts offer a more accurate prediction of post-myocardial infarction deaths than a single time point data analysis.