Utility of Eosinophil Count as Predictor of Bacteremia

Sir—The utility of the neutrophil count as a clinical indicator of bacterial infection is widely accepted, and, in the appropriate clinical context, the presence of leukocytosis, with or without a left shift, will often prompt investigation and empirical treatment of presumptive acute bacterial infection. Analogously, the association between the eosinophil count and parasitic infection is well known [1]. However, the possible usefulness of the eosinophil count as an indicator of bacteremia has not been well studied. A marked diminution in the number of circulating eosinophils is generally regarded as a physiologic consequence of acute bacterial infection. Peripheral blood eosinopenia was documented by Bass [2] in experimentally induced bacterial infection in mice. Xanthou [3] and Weinberg et al. [4] studied blood counts in newborn infants and found that eosinophil counts decreased to as low as 0 in subjects who developed acute bacterial infection. Peripheral blood eosinopenia is well described in adult and pediatric patients with typhoid fever [5] and has been observed in association with other agents of sepsis [6]. The potential utility of eosinophil counts as predictors of bacteremia was investigated by Lipkin [7] in a study of 75 adult inpatients with positive results of blood cultures that showed that, as the number of positive blood culture results per patient increased, the percentage of eosinophils in the peripheral blood smear decreased [7]. No patient with ⩾2 positive blood culture results had a peripheral blood smear with >1% eosinophils, demonstrating a possible correlation between eosinopenia and bacteremia. We sought to confirm this relationship between decreased eosinophil count and bacteremia. We reviewed the inpatient medical records of 171 patients with blood cultures positive for gram-negative bacilli, Staphylococcus aureus, and Streptococcus species during a 6-month period at the University of North Dakota School of Medicine and Health Sciences (Fargo). Immunosuppressed patients with bacteremia were excluded from analysis, as were patients with neutropenia (i.e., those with <1000 WBCs per high-power field). Case patients were age- and sex-matched with control patients with negative results of blood cultures during the same study period. Total leukocyte count and relative and absolute eosinophil counts were determined for both case and control patients, and χ2 analysis was used to compare results (table 1). Sensitivity, specificity, and positive and negative predictive values were calculated in the usual manner. We found no utility associated with eosinopenia (defined as a relative eosinophil level of 0%) for predicting bacteremia and conclude that the absence of peripheral blood eosinophils cannot be used as a clinically reliable marker of bacteremic infection. Relative eosinophil counts in and results of statistical analysis for 171 patients with or without bacteremia.