Low eosinophils and their dynamic as a predictor of death in patients with infections: a systematic review and meta-analysis of cohort studies

Abstract

Background: Eosinophils prognostic significance in predicting mortality has become particularly notable during the COVID-19 pandemic. We aimed to evaluate the prognostic value of peripheral low eosinophil (eosinopenia) with focus on their dynamics (eosinophil recovery) in patients with infections.

Methods: We searched databases (MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Library) and additional sources from inception to 1 December 2023. Cohort studies involving adult patients hospitalized with infections were evaluated using dual-reviewer methodology.

Results: Out of 15,066 screened papers, 151 studies met the inclusion criteria, with 107 focused on COVID-19, 14 on sepsis, 9 on Clostridioides difficile, 6 on acute COPD exacerbations, and 17 on other infections. The majority of studies reported significantly lower admission eosinophil levels in non-survivors compared to survivors. Random-effects model meta-analysis showed mean eosinophil difference between deceased and survivors -15.31 (95% CI: -18.72 to -11.90) for COVID-19 and -44.6 (95% CI: -95.62 to 6.41) for sepsis (I² > 90%). Mortality with eosinopenia showed low certainty of evidence for C. difficile (0 cells/mm³ cut-off: RR 2.35; 95% CI: 1.84-2.99) and very low for COVID-19 when considering different cut-offs (0 cells/mm³: RR 2.37; 95% CI: 1.86-3.01; 20 cells/mm³: RR 2.90; 95% CI: 1.59-5.28; 50 cells/mm³: RR 2.70; 95% CI: 1.33-5.49). Survivors generally showed an increase in eosinophil counts within the first 2-5 days of hospitalization, while the deceased had persistently low levels.

Conclusions: Eosinopenia, particularly the trajectory of eosinophil recovery, may serve as a cost-effective and accessible prognostic marker for mortality in patients with infections.

Keywords: COVID-19; Clostridioides difficile; dynamic; eosinopenia; eosinophil; infection; mortality; recovery; sepsis; systematic review.

Figure 1.

The PRISMA flowchart of the study. *Some studies reported multiple types of infections.

Figure 2.

Distribution of eosinophil cut-off values used to define eosinopenia and assess mortality across different disease contexts.

Figure 3.

Qualitative synthesis results of the included studies. The assessment was made across different infection types: COVID-19, sepsis, C. difficile, AECOPD, and other infections. Numbers in parentheses indicate the number of studies included per category. Top row pie charts represent the proportion of studies reporting the statistical significance of admission mean/median difference of eosinophil values between survivors and deceased. The p values for these differences are: ▀ lower eosinophil values in deceased (p < 0.05), ▀ higher (p < 0.05), ▀ lower (p > 0.05), ▀ higher (p > 0.05), ▀ equal (p > 0.05). Middle row represents unadjusted mortality: ▀ higher mortality with eosinopenia (p < 0.05), ▀ lower mortality with eosinopenia (p < 0.05), ▀ higher mortality with eosinopenia (p > 0.05), ▀ lower mortality with eosinopenia (p > 0.05). The bottom row shows the adjusted mortality (only the studies with statistically significant findings). Compared to unadjusted results, we used gray to indicate studies that did not adjust for confounders, solid green for those that maintained statistical significance after adjustment, and green stripes for those that lost statistical significance after adjustment.

Figure 4.

Meta-analysis of eosinophil count differences and mortality risk associated with eosinopenia in infectious diseases based on admission eosinophil values. The top panels show pooled mean differences in absolute eosinophil counts and eosinophil ratios between deceased and surviving patients, stratified by infection type (COVID-19 and sepsis), COVID-19 severity subgroups (as defined by the authors according to WHO-subgroup analysis was conducted based on the proportion of severe patients included in each study), and clinical settings (ED, ICU, ward) [13]. The bottom panel presents pooled risk ratios for mortality in patients with and without eosinopenia, using three different eosinophil cut-offs (0, 20, and 50 cells/mm³).

Figure 5.

Eosinophil dynamic changes in hospitalized patients with infections across 36 studies. This visualization summarizes eosinophil trends during hospitalization in COVID-19, sepsis, and other infections. Each line represents a study, with circles indicating eosinophil value comparisons between deceased and survivors at admission (A), discharge or death (D) and at specified time points during hospitalization. Circle colors represent differences in the degree of statistical significance (green: p < 0.05; red: p > 0.05; grey: not compared), diamonds denoting the reported day of eosinophil recovery, and lines reflecting the frequency of assessment (daily vs. specific time points). Three main eosinophil recovery patterns were identified: (1) persistent eosinopenia in both groups, (2) eosinophil recovery in survivors only, and (3) progressive increases in both groups with persistently lower values in non-survivors.