Peripheral blood eosinophils: a surrogate marker for airway eosinophilia in stable COPD

Abstract

Introduction: Sputum eosinophilia occurs in approximately one-third of stable chronic obstructive pulmonary disease (COPD) patients and can predict exacerbation risk and response to corticosteroid treatments. Sputum induction, however, requires expertise, may not always be successful, and does not provide point-of-care results. Easily applicable diagnostic markers that can predict sputum eosinophilia in stable COPD patients have the potential to progress COPD management. This study investigated the correlation and predictive relationship between peripheral blood and sputum eosinophils. It also examined the repeatability of blood eosinophil counts.

Methods: Stable COPD patients (n=141) were classified as eosinophilic or noneosinophilic based on their sputum cell counts (≥3%), and a cross-sectional analysis was conducted comparing their demographics, clinical characteristics, and blood cell counts. Receiver operating characteristic curve analysis was used to assess the predictive ability of blood eosinophils for sputum eosinophilia. Intraclass correlation coefficient was used to examine the repeatability of blood eosinophil counts.

Results: Blood eosinophil counts were significantly higher in patients with sputum eosinophilia (n=45) compared to those without (0.3×10(9)/L vs 0.15×10(9)/L; P<0.0001). Blood eosinophils correlated with both the percentage (ρ=0.535; P<0.0001) and number of sputum eosinophils (ρ=0.473; P<0.0001). Absolute blood eosinophil count was predictive of sputum eosinophilia (area under the curve =0.76, 95% confidence interval [CI] =0.67-0.84; P<0.0001). At a threshold of ≥0.3×10(9)/L (specificity =76%, sensitivity =60%, and positive likelihood ratio =2.5), peripheral blood eosinophil counts enabled identification of the presence or absence of sputum eosinophilia in 71% of the cases. A threshold of ≥0.4×10(9)/L had similar classifying ability but better specificity (91.7%) and higher positive likelihood ratio (3.7). In contrast, ≥0.2×10(9)/L offered a better sensitivity (91.1%) for ruling out sputum eosinophilia. There was a good agreement between two measurements of blood eosinophil count over a median of 28 days (intraclass correlation coefficient =0.8; 95% CI =0.66-0.88; P<0.0001).

Conclusion: Peripheral blood eosinophil counts can help identify the presence or absence of sputum eosinophilia in stable COPD patients with a reasonable degree of accuracy.

Keywords: chronic obstructive pulmonary disease; diagnostic accuracy; sputum eosinophilia; stability of eosinophil counts.

Figure 1

Study flowchart.

Figure 2

Scatter dot plot comparing. Notes: (A) Blood eosinophil count, (B) eosinophil/neutrophil ratio, (C) eosinophil/lymphocyte ratio, and (D) neutrophil/lymphocyte ratio between eosinophilic (≥3% sputum eosinophils) and noneosinophilic COPD (<3% sputum eosinophils). Graphs represent individual data points and a median as a bar (red). Abbreviations: ENR, eosinophil/neutrophil ratio; ELR, eosinophil/lymphocyte ratio; NLR, neutrophil/lymphocyte ratio.

Figure 3

Scatter plots for correlations between sputum and blood eosinophil counts. Notes: (A) Correlation between percentage sputum eosinophils and absolute blood eosinophil count (×10⁹/L). Vertical dotted line represents upper limit of normal for percentage sputum eosinophils. (B) Correlation between absolute sputum eosinophil counts (×10⁴/L) and absolute blood eosinophil count (×10⁹/L).

Figure 4

Receiver operating characteristic (ROC) curve for absolute blood eosinophil count to predict sputum eosinophilia (≥3%). Note: Points on the ROC curve are presented as (specificity %, sensitivity %).

Figure 5

Bland–Altman plot showing the difference between the absolute blood eosinophil counts of two measurements against the mean of the absolute blood eosinophil counts of the two measurements. Notes: The blue line represents the mean of differences (bias line). Horizontal dotted lines represent the 95% limits of agreement (mean difference ±1.96 SD). Abbreviation: SD, standard deviation.