Prevalence and Baseline Clinical Characteristics of Eosinophilic Chronic Obstructive Pulmonary Disease: A Meta-Analysis and Systematic Review

Background: Chronic obstructive pulmonary disease (COPD) is a heterogeneous disease with different clinical and pathophysiological characteristics. Cumulative evidence shows that eosinophil levels may be connected to the therapeutic effects and phenotype of COPD. However, the prevalence of eosinophilic inflammation in COPD and the baseline characteristics of eosinophilic COPD remain unknown. Our study investigated the prevalence of COPD with eosinophil levels of >2% and the characteristics of eosinophilic COPD. Methods: We searched the Cochrane Central Library, Medline, Embase, and the Web of Science for trials of eosinophil and COPD published from database inception to May 1, 2019. Results: In total, 40,112 COPD patients that were involved in 19 trials were included in the final analysis. The prevalence of eosinophilic COPD ranged from 18.84 to 66.88%, with an average prevalence of 54.95% across all studies. We found that men, ex-smokers, individuals with a history of ischemic heart disease, and individuals with a higher body mass index (BMI) were at higher risk of eosinophilic COPD (OR 1.36, 95% CI 1.26–1.46, P < 0.00001; OR 1.23, 1.12–1.34, P < 0.0001; OR 1.31, 1.14–1.50, P = 0.001; MD 0.70, 0.27–1.12, P = 0.001). There was, however, a lower proportion of GOLD stage I patients among those with eosinophilic COPD (OR 0.84, 0.73–0.96, P = 0.01). No significant differences were found in terms of age, current smoker status, pack-years smoked, percent of predicted forced expiratory volume in 1 s, hypertension, diabetes, or other GOLD stages between the two groups (P > 0.05). Conclusions: Our analysis suggests that eosinophilic inflammation is prevalent in COPD. Eosinophilic COPD was more likely to occur in men, ex-smokers, those with a higher BMI, and those with a high risk of some comorbidity; however, a lower proportion of patients with eosinophilic COPD experienced mild airflow limitations.


INTRODUCTION
Chronic obstructive pulmonary disease (COPD) is a heterogeneous disease with a variety of features and characteristics. Identification of COPD phenotypes may allow targeted therapeutic strategies. Eosinophilic inflammation is generally believed to be characteristic of asthma, whereas neutrophilic inflammation is considered to be a typical sign of COPD. However, recent reports have shown that eosinophilic inflammation occurs in COPD, in both the exacerbation and stable phases (1,2). Growing evidence suggests that eosinophil levels may be related to the therapeutic effect and phenotypes of COPD, even after asthma patients are carefully excluded (3)(4)(5)(6).
A sputum eosinophil level of >3% is a recognized sign of airway eosinophilic inflammation (4,6). It was reported that blood eosinophil levels of >2% are indicative of a higher sensitivity in identifying airway eosinophil levels of >3% during COPD exacerbation (1). An alternative cut-off level (≥200 cells per µL or 300 cells per µL) has been used in some studies in addition to the 2% cut-off (7-10). Research has shown that blood eosinophil is a clinically reliable predictor of the inflammatory phenotype. We conclude from these studies that the blood eosinophil level is of reasonable importance in patients with COPD and, as such, is a promising biomarker to guide disease management.
A number of studies have investigated the prevalence and baseline clinical characteristics of patients with eosinophilic COPD. The prevalence of eosinophilic COPD, however, has differed wildly between studies. In one study, 2,083 patients (66%) had eosinophil levels of ≥2% in a post-hoc analysis that included 3,177 patients (11). In a retrospective multicenter study enrolling 605 hospitalized patients, 177 patients (29%) had blood eosinophil levels of >2% (12). In a retrospective analysis of a randomized clinical trial, 18.8% of patients had eosinophil levels of >2% (7).
Nonetheless, the baseline clinical characteristics of eosinophilic COPD remain unclear. An analysis of the ECLIPSE cohort study showed that COPD patients with eosinophil levels that were persistently >2% were older, were more likely to be male, were less likely to be a current smoker, had a lower fat-free mass index, and had a higher percent of predicted forced expiratory volume in 1 s (ppFEV 1 ) compared with the other COPD groups (2). An observational cohort study suggested that significantly higher numbers of male and young patients were found in the eosinophilic COPD group (13). In a national survey, being male and older in age and having congestive heart failure were significantly associated with eosinophil levels of >2% in COPD (14). In an analysis of the SPIROMICS study, significant differences were found in terms of age, sex, genus, body mass index (BMI), smoking history (pack-years), and current smoker status, but there was no evidence of a difference in the Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage between patients with lower eosinophil (<200 cells per µL) Abbreviations: COPD, chronic obstructive pulmonary disease; BMI, body-mass index; ppFEV 1 , percent of predicted forced expiratory volume in 1 s; GOLD, global initiative for chronic obstructive lung disease. and higher eosinophil (≥200 cells per µL) levels. A significantly lower ppFEV 1 and FEV 1 : FVC percentage were found in the higher eosinophilic group (8). Couillard et al. (9), however, reported that there was no significant difference between the two phenotypes of COPD in sex, age, smoking status, home oxygen use, comorbidity, lung function, GOLD stage, or hospitalization for COPD in the previous year.
The aim of this study was to evaluate published studies that investigated the prevalence and baseline characteristics of eosinophilic COPD and apply standard meta-analysis methods to gain a more precise result.

Search Strategies
We searched the Cochrane Central Register of Controlled Trials, Medline, the Web of Science, and Embase for studies with the keywords "Eosinophil" and "Chronic obstructive pulmonary disease, " not limited to any language, publication type, or time. We searched for reports published up to May 1, 2019. In order to minimize bias and errors, we also retrieved the reference articles of all included studies. Keywords in related conference articles were also used to retrieve studies. This study was registered with PROSPERO. The findings are reported in compliance with the PRISMA guidelines.

Inclusion and Exclusion Criteria
Studies matching the following criteria were considered suitable for inclusion: (1) randomized controlled trials (RCTs), as well as observational, cohort, case control, and retrospective studies; (2) trials conducted in patients with COPD aged >40 years; and (3) trials reporting data on the prevalence or baseline clinical characteristics of COPD according to an eosinophil cut-off level of 2% in the blood. Patients admitted due to other medical problems; those with a history of asthma, interstitial pulmonary disease, active pulmonary tuberculosis, or lung cancer; those with other diseases that could influence eosinophil count (eosinophilic pneumonia, allergic diseases, parasitic infections); and individuals with severe dysfunction of other organs or systems or malignant tumors were excluded. Conference articles and trials conducted in pregnant subjects were omitted.

Outcomes
The prevalence and baseline clinical characteristics of COPD according to eosinophil levels were the primary and the secondary outcomes, respectively. The baseline clinical characteristics of COPD included demographic characteristics (sex, age, and BMI), smoking status (current-smoker, ex-smoker, and pack-years smoked), lung function (ppFEV 1 ), comorbidity (ischemic heart disease, hypertension, and diabetes), and GOLD stage.

Study Selection
Two phases were performed by two separate researchers to verify the studies that met the eligibility criteria. Duplicated studies were first discarded by checking titles and abstracts. Suitable studies were then identified by assessing the full text.
Trials reporting data on the prevalence or baseline clinical characteristics of COPD and using an eosinophil cut-off level of 2% in the blood were included.

Data Extraction
Two researchers extracted suitable information from the included studies following the criteria suggested by Cochrane (15). Corresponding authors were emailed for any missing data.

Quality Assessment
The Newcastle-Ottawa Scale (NOS) was utilized to evaluate the quality of non-randomized studies (16). Two investigators conducted the quality assessment. A third investigator was consulted to resolve any discrepancies.

Statistical Analysis
The statistical analysis was performed using the Cochrane systematic review software, Review Manager (RevMan; Version 5.3, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, 2014). The Mantel-Haenszel test was used to adjudicate statistical significance at a z-value and P-value < 0.05, as well as evaluate the hypothesis. The outcomes are shown in forest plots. The outcomes of continuous and dichotomous variables are expressed as mean differences (MD) and odds ratios (OR), respectively. The χ 2 test with P < 0.1 and I 2 > 50% was used to determine significance in the test for heterogeneity. The sensitivity analysis was performed to substitute ranges of values or alternative decisions. A random-effects model was used in case of statistical heterogeneity; otherwise, a fixed-effects model was applied. Any disagreement was resolved by a third investigator reaching a mutual consensus.

Heterogeneity
No heterogeneity was observed regarding sex, ex-smoker status, ischemic heart disease, or GOLD stage. In contrast, significant statistical heterogeneities were found in the analysis of age, BMI,    (Figures 1, 3, 4; Appendix Figures S1, S3, S4, S6, S7). Sensitivity analysis was performed to assess whether any study biased the overall results. The overall effect and summary MDs or ORs were recalculated after removing each study one at a time. This analysis   Figure S14).

Primary Outcome
The prevalence of eosinophilic COPD ranged from 18.84 to 66.88% and the mean prevalence across all studies was 54.95%.

Lung Function
With regard to lung function, no significant difference was found in the ppFEV 1 between the two groups (MD 0.34, −1.03-1.71, P = 0.62) (Appendix Figure S4).

Subgroup Analysis
When restricted to different disease statuses, a significantly higher proportion of male patients was observed in both the stable and acute exacerbation phases of COPD in the eosinophilic group (n = 34,507, OR 1.36, 95% CI 1.26-1.47, P < 0.00001; n = 5605, OR 1.39, 95% CI 1.11-1.73, P = 0.004) (Appendix Figure S12). The subgroup analysis found that subjects in the eosinophilic group were significantly younger when restricting the analysis to the acute exacerbation phase. No difference was found between groups in the stable phase (n = 5605, MD −1.38, −2.34 to −0.42, P < 0.0001; n= 27,100, MD 0.16, −0.02-0.33, P = 0.08) (Appendix Figure S13).

DISCUSSION
This comprehensive systematic review and meta-analysis investigated the prevalence and baseline clinical characteristics of eosinophilic COPD. The prevalence of eosinophilic COPD ranged from 18.84 to 66.88%, with an average prevalence of 54.95% across all studies. The prevalence of COPD varied greatly owing to differences in diagnostic criteria, as well as survey and analytical methods. Reasons for the large range in the prevalence of eosinophilic COPD may be similar, except for the effect of different races, regions, and countries (28). In Japan, there was a tendency to exclude patients with any feature of asthma from the diagnosis of COPD, especially in younger patients with milder forms of the disease. This then leads to a low diagnostic rate of eosinophilic COPD (29)(30)(31)(32).
In this study, we found that male patients are more at risk for eosinophilic COPD (OR 1.36, 95% CI 1.26-1.46, P < 0.00001) (Figure 2). Sex is one of the most fundamental and defining features of subpopulations in human beings. A higher absolute eosinophil count and eosinophil percentage were observed in men in an observational study; however, the number of participants enrolled was relatively low (476) (33). This may suggest that men are prone to having higher eosinophil levels, and that eosinophilic inflammation increases the risk of progression to COPD in men. More trials are needed, however, to verify this hypothesis. A higher BMI was also observed in the eosinophilic group (MD 0.70, 0.27-1.12, P = 0.001) (Figure 3). Our result was consistent with the result of a longitudinal analysis, which revealed that COPD patients with persistent eosinophil levels of >2% had fat-free mass (2). No significant difference in age was found between the two groups (MD −0.33, −0.73-0.07, P = 0.10) (Figure 4). Considering that the primary analysis in our study was aimed at identifying the characteristics of the subtypes of COPD, there is no prior relevant information that can be referenced. The mechanisms for these differences remain unclear.
Regarding smoking status, we found a significant difference in the proportion of ex-smokers between the two groups (OR 1.23, 1.12-1.34, P < 0.0001) (Appendix Figure S2), but no difference in the proportion of current smokers or in pack-years smoked (OR 0.78, 0.59-1.02, P = 0.07; MD 0.52, −1.62-2.67, P = 0.63) (Appendix Figures S1, S3). Pooled analysis showed that the prevalence of ex-smokers was higher in patients with eosinophilic COPD. The inflammation detected in the respiratory tract may be a modified inflammatory response to chronic irritants, such as cigarette smoke. The presence of persistent lung inflammation after smoking cessation remains unknown; even perturbations and autoantigens in the lung microbiome may play a role (34,35). We hypothesize that smoking may induce eosinophilic inflammation and that the inflammation persists even after smoking cessation, although more research is needed to confirm this hypothesis.
In terms of lung function, no significant difference was found in the ppFEV 1 (MD 0.34, −1.03-1.71, P = 0.62) (Appendix Figure S4). A significantly lower prevalence of GOLD stage I was, however, found in the eosinophilic COPD group (OR 0.84, 0.73-0.96, P = 0.01) (Appendix Figure S8). We believe that the mild severity of airflow limitations is more common in non-eosinophilic COPD and rarer in eosinophilic COPD. Our analysis was not consistent with previous findings. In the ECLIPSE cohort study, patients with COPD with persistent eosinophil levels of >2% had a significantly higher ppFEV 1 (2). In the SPIROMICS study, patients with a lower baseline eosinophil level (<1%) were prone to severe COPD (36). There is no definitive explanation, however, for this problem. Additional studies are needed to further investigate the relationship between eosinophil and FEV 1 in COPD patients.
There was a significantly lower prevalence of chronic heart failure in the eosinophilic COPD group (OR 0.81, 0.68-0.97, P = 0.02) (Appendix Figure S5). No difference in the prevalence of hypertension or diabetes between groups was found. COPD patients often have important concomitant illnesses. The SPIROMICS study suggested a higher incidence of comorbidities (prior heart attack, anemia, diabetes, and chronic heart failure) among COPD patients with eosinophil levels of ≤2% (36). The comorbidities in COPD may be caused by original genetic variances in response to the inhalation of poisonous particles, particularly during smoking (37). More rigorous trials are needed to clarify this issue.
Significant variability in blood eosinophil levels has been shown throughout the course of COPD (38,39). To investigate the stability of blood eosinophilic inflammation (≥2%), subjects were classified into predominantly (PE), intermittently (IE), and rarely (RE) eosinophilic groups in one study (40). The PE group was characterized by an increased risk of eosinophilic inflammation during exacerbation. The PE group at stable visits and eosinophilia during exacerbation were associated with a minor risk of bacterial infection during exacerbation. Bacterial infection during exacerbation was higher in winter in the PE group. Blood eosinophil counts in the stable status could predict the nature of inflammation during future exacerbations. When combined with an understanding of seasonal variation, this may also provide a basis for the development of new therapy. More research, however, is warranted.
Although blood eosinophil is considered to be a promising biomarker, eosinophil-guided treatment of acute exacerbation of COPD remains an issue. Bafadhel et al. (41) showed that systemic corticosteroid use in a low eosinophil (<2%) group was associated with less improvement in chronic respiratory questionnaire scores and higher treatment failure when compared to the placebo group. On the contrary, Sivapalan et al. (42) reported that, when compared to standard therapy in patients hospitalized for COPD, eosinophil-guided therapy did not lead to a difference in the number of days alive, number of patients discharged from the hospital within 14 days of recruitment, or the risk of treatment failure at 30 days. Future studies on eosinophil-guided therapies are needed.
This study has several strengths. First, it is a comprehensive systematic review and meta-analysis to analyze the prevalence and baseline clinical characteristics of eosinophilic COPD. Additionally, the studies that were included were of high quality. All data were collected at the very beginning of each study, protecting against subsequent interference. Our results are thus highly credible. This study has some limitations. First, the studies that were included were not RCTs. Nonetheless, the extracted data were obtained from RCTs that enrolled a large number of patients with COPD and classified according to eosinophilic and non-eosinophilic inflammation status. Second, given that this analysis is the first to verify the baseline clinical characteristics of eosinophilic COPD, the underlying mechanisms remain unclear. Finally, a proportion of patients had already been treated with corticosteroids and antibiotics in the community. It remains unclear whether and to what extent these therapies affect the eosinophil count. Further research is therefore warranted.

CONCLUSIONS
In conclusion, eosinophilic inflammation is prevalent in COPD. Eosinophilic COPD was more common in men, ex-smokers, subjects with higher BMI, and in those with a high risk of some comorbidity. The group also included a low proportion of patients with mild airflow limitations. Future rigorous prospective trials are needed, particularly in basic research, to further identify the relationship between eosinophil levels and COPD. Additional studies should explore the exact mechanisms that are responsible for the characteristics of eosinophilic COPD.

DATA AVAILABILITY STATEMENT
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

AUTHOR CONTRIBUTIONS
H-XW and D-YC initiated and coordinated the study. H-XW and K-QZ were responsible for the data collection and data analysis. Studies were reviewed by D-YC. H-XW wrote the first draft of the manuscript. All the authors were involved in the interpretation of the analyses and gave input to the final manuscript.