Association of sputum and blood eosinophil concentrations with clinical measures of COPD severity: an analysis of the SPIROMICS cohort.

BACKGROUND: Increased concentrations of eosinophils in blood and sputum in chronic obstructive pulmonary disease (COPD) have been associated with increased frequency of exacerbations, reduced lung function, and corticosteroid responsiveness. We aimed to assess whether high eosinophil concentrations in either sputum or blood are associated with a severe COPD phenotype, including greater exacerbation frequency, and whether blood eosinophils are predictive of sputum eosinophils.
METHODS: We did a multicentre observational study analysing comprehensive baseline data from SPIROMICS in patients with COPD aged 40-80 years who had a smoking history of at least 20 pack-years, recruited from six clinical sites and additional subsites in the USA between Nov 12, 2010, and April 21, 2015. Inclusion criteria for this analysis were SPIROMICS baseline visit data with complete blood cell counts and, in a subset, acceptable sputum counts. We stratified patients on the basis of blood and sputum eosinophil concentrations and compared their demographic characteristics, as well as results from questionnaires, clinical assessments, and quantitative CT (QCT). We also analysed whether blood eosinophil concentrations reliably predicted sputum eosinophil concentrations. This study is registered with ClinicalTrials.gov (NCT01969344).
FINDINGS: Of the 2737 patients recruited to SPIROMICS, 2499 patients were smokers and had available blood counts, and so were stratified by mean blood eosinophil count: 1262 patients with low (<200 cells per μL) and 1237 with high (≥200 cells per μL) blood eosinophil counts. 827 patients were eligible for stratification by mean sputum eosinophil percentage: 656 with low (<1·25%) and 171 with high (≥1·25%) sputum eosinophil percentages. The high sputum eosinophil group had significantly lower median FEV1 percentage predicted than the low sputum eosinophil group both before (65·7% [IQR 51·8-81·3] vs 75·7% [59·3-90·2], p<0·0001) and after (77·3% [63·1-88·5] vs 82·9% [67·8-95·9], p=0·001) bronchodilation. QCT density measures for emphysema and air trapping were significantly higher in the high sputum eosinophil group than the low sputum eosinophil group. Exacerbations requiring corticosteroids treatment were more common in the high versus low sputum eosinophil group (p=0·002). FEV1 percentage predicted was significantly different between low and high blood eosinophil groups, but differences were less than those observed between the sputum groups. The high blood eosinophil group had slightly increased airway wall thickness (0·02 mm difference, p=0·032), higher St George Respiratory Questionnaire symptom scores (p=0·037), and increased wheezing (p=0·018), but no evidence of an association with COPD exacerbations (p=0·35) or the other indices of COPD severity, such as emphysema measured by CT density, COPD assessment test scores, Body-mass index, airflow Obstruction, Dyspnea, and Exercise index, or Global Initiative for Chronic Obstructive Lung Disease stage. Blood eosinophil counts showed a weak but significant association with sputum eosinophil counts (receiver operating characteristic area under the curve of 0·64, p<0·0001), but with a high false-discovery rate of 72%.
INTERPRETATION: In a large, well characterised cohort of former and current smoking patients with a broad range of COPD severity, high concentrations of sputum eosinophils were a better biomarker than high concentrations of blood eosinophils to identify a patient subgroup with more severe disease, more frequent exacerbations, and increased emphysema by QCT. Blood eosinophils alone were not a reliable biomarker for COPD severity or exacerbations, or for sputum eosinophils. Clinical trials targeting eosinophilic inflammation in COPD should consider assessing sputum eosinophils. FUNDING: National Institutes of Health, and National Heart, Lung, and Blood Institute.

INTRODUCTION

Airways inflammation in chronic obstructive pulmonary disease (COPD) is thought to be characterized by increased neutrophils,[1] macrophages,[2] proteases, IL-6, IL-8, and Th1 cytokines[3] while airways inflammation in asthma is traditionally characterized by increased eosinophils, and Th2 cytokines.[4] However, reports challenge these presumptive differences between asthma and COPD. The ECLIPSE study reported that in COPD, sputum neutrophils are weakly associated with lung function and health status, and not associated with exacerbations, emphysema or systemic inflammation.[1] ECLIPSE also reported a mean 1·3% sputum eosinophil level in 359 subjects with COPD,[1] but did not observe blood eosinophil associations with radiologic measure of emphysema or with COPD exacerbations and hospitalizations. ECLIPSE reported ≥2% (150/µL) blood eosinophils associated with evidence of higher FEV1, lower St. Georges Respiratory Questionnaire (SGRQ) and modified Medical Research Council scores.[5] Other COPD studies have reported increased eosinophils associated with exacerbations and greater hyperinflation on QCT,[6,7] suggesting Th2 inflammation may contribute to disease progression. Moreover, increased epithelial Th2 signature gene expression has been associated in two COPD cohorts with more severe airflow obstruction.[8] Eosinophils may represent a potential biomarker in COPD since eosinophilia is related to corticosteroid responsiveness.[1,9-11] In a phase II clinical trial, anti-IL-5 receptor therapy reduced the COPD exacerbation rate in a subgroup of patients with elevated blood and sputum eosinophilia.[12]

Determination of disease severity in COPD is complex and involves more than lung function assessments; additional clinical characteristics have been incorporated in successive revisions of the GOLD severity stages.[13] Current classification includes lung function, symptom scores and exacerbation frequency. Thus, severity of COPD is dependent on multiple characteristics; eosinophilic inflammation may contribute.

Reports suggest blood eosinophil counts may represent a useful surrogate measure of airway eosinophils in COPD,[11,14] although blood eosinophils appear to correlate poorly with sputum eosinophils in asthma,[15,16] and do not distinguish between asthma-dominant, COPD-dominant or asthma/COPD overlap populations.[17] However, larger studies of comprehensively phenotyped COPD patients often lack robust sputum eosinophil data.[18-21] Thus it is uncertain whether peripheral eosinophils accurately predict airway eosinophils.

We investigated the hypotheses that blood and sputum eosinophils in subjects with a history of tobacco use were associated with a more severe COPD phenotype identified by diminished lung function, QCT measurements of emphysema or air-trapping, clinical COPD characteristics, and exacerbations. We also investigated relationships of blood and sputum eosinophils to determine whether blood eosinophils reliably predicted sputum eosinophils. These hypotheses were evaluated in the comprehensively characterized SPIROMICS cohort.[22] A portion of these studies were presented as an abstract at the 2016 American Thoracic Society meeting.[23]

METHODS

Subjects

Subjects with current or former history of tobacco use (≥20 pack-year), recruited to include specific groups of smokers with preserved lung function (31%), GOLD stages 1 and 2 (41%), and GOLD stages 3 and 4 (21%) and a control group of nonsmokers, age 40–80 (N=2737), were enrolled in the Subpopulations and Intermediate Outcome Measures in COPD Study (SPIROMICS) at six clinical sites and additional subsites.[22] COPD was defined in long-term smokers as a post-bronchodilator FEV1/FVC ratio <0·7. Subjects underwent extensive baseline phenotypic characterization including lung function assessment pre- and post-bronchodilator with albuterol and ipratropium, CT total lung capacity and residual volume (TLC and RV) using QCT indicators for emphysema (% voxels less than −950 Hounsfield Units [HU]) at TLC, and air-trapping (% voxels less than −856 HU) at RV, airway metrics (VIDA Diagnostics, Iowa)[24] and parametric response mapping (PRM) for functional small airways disease (fSAD),[25] collection of blood (for DNA, RNA, plasma, sera, IgE and complete blood cell counts [CBC]), urine, 6 minute walk distance, GOLD stage, BODE index, COPD Assessment Score (CAT), St. Georges Respiratory Questionnaire (SGRQ), and administered questionnaires for medical history, exacerbations (retrospective from past year), hospitalizations, respiratory exposures, and medications.[22] In a subset (see online supplement Figure S1 detailing subset of SPIROMICS with induced sputum cell counts), induced sputum was performed. Subjects with a primary asthma diagnosis were excluded, but all subjects were asked if they had ever had a health care professional say that they had asthma (“prior asthma label”). In the N=2499 enrolled ever smokers, the mean ± standard deviation for blood eosinophil count was 200 ± 240/µL (median=190/µL; full range 0–8300/µL); a higher eosinophil cutpoint of 300/µL was also examined. See supplement for additional information on the selection of stratification cutpoints, reproducibility, and other details.

Sputum Induction and Processing

SPIROMICS subjects with post-bronchodilator FEV1 % predicted ≥35% were eligible for sputum induction with saline solutions, nebulized for three 7 min intervals each. Expectorated sputum samples were processed as described in detail in the supplement. Cytospin slides were read by the central reading center. Slides were available for 1001 subjects. Differential counts (500–600 total cells) which had ≤100 leukocytes total (N=179) or ≥80% squamous (N=11) were excluded as unacceptable. The mean sputum eosinophil % (± standard deviation) for the subgroup with acceptable sputum (N=827) was 1·25 ± 4·25% (median=0·3%; full range 0–75%); a higher cutpoint of 2% for sputum eosinophils was also examined. (See supplement on subject distribution with induced sputum.)

Statistical Analyses

Subjects were stratified by eosinophil mean blood counts or mean sputum %. Demographic and biomarker data are presented as means ± standard deviations, or medians (25%–75% interquartile range) for continuous variables, and as % positive for categorical variables. Measures not meeting Kolmogorov-Smirnov test for normal distribution, were transformed by log, or square root values. Continuous variables were tested by parametric or non-parametric tests (see supplement; SAS 9·2, or Sigmastat 12·5). Categorical variables were analyzed using Chi-square or Fisher Exact tests. Correlations were examined by Pearson Correlation test or linear regression. Receiver Operating Curve (ROC) analysis was performed for blood eosinophil prediction of sputum eosinophils. The False Discovery Rate (= False Positives / False Positives + True Positives) was examined.[26] Classification tree analysis examining sputum and blood eosinophils to model exacerbations was performed using R part routines in R software package. Variables with a p value <0·05 were accepted as significant.[27]

RESULTS

Subject Demographics

Demographic characteristics of SPIROMICS subjects stratified by mean blood eosinophils (< or ≥200/µL) and by mean sputum eosinophils (< or ≥1·25%) are shown in Table 1. For those subjects with low compared to high blood eosinophils, age, gender, race, BMI, cigarette smoking pack-years, % current smoker, and % use of inhaled corticosteroids (ICS) statistically differed, although differences between the groups were small (<10%). The proportion of current smokers was less in the ≥200/µL eosinophil group, but the number of cigarettes smoked each day was the same. Medications (supplement Table S1) did not differ, except for ICS treatment, higher in the ≥200/µL eosinophil group (Table 1). Total serum IgE levels (range 5·5–1660) and sputum eosinophils % differed in blood eosinophil groups, but those with a “prior asthma label” or childhood asthma did not differ (Table 1). Median sputum eosinophil % in low and high blood eosinophil groups differed (p<0.001), but were lower than the overall sputum eosinophil mean of 1·25%.

Table 1

Demographics for subjects stratified by mean blood or sputum eosinophils (Eos; cutoff 200/µL or 1·25%, respectively).

Variable	BloodEos<200/µL	BloodEos≥200/µL	PValue*	SputumEos<1.25%	SputumEos≥1.25%	PValue*
Number	1262	1237		656	171
Age (yr)	65 (56 – 70)	65 (59 – 71)	0·001	65 (57–71)	64 (57–71)	0·87
Male Sex, N (%)	631 (50)	730 (59)	<0·001	377 (57)	99 (58)	0·99
RACE N (%) Cau/N (%) AA/N (%) other	934 (74)/ 285 (23)/ 43 (3)	1004 (81)/ 173 (14) / 60 (5)	<0·001	511 (78)/ 110 (17)/ 35 (5)	132 (77)/ 29 (17)/ 10 (6)	0·90
BMI	26·8 (24 – 31)	28·2 (25 – 32)	<0·001	28·3 (25–32)	28·1 (25–32)	0·92
Smoking (pack-years)	41 (30 – 60)	45 (34 – 60)	0·008	43 (32–60)	44 (33–60)	0·70
Cigarettes/day	15 (9 – 20)	15 (8 – 20)	0·97	15 (10–20)	15 (6–20)	0·52
Current Smoker, N (% positive)	522 (42)	451 (37)	0·003	293 (45)	70 (41)	0·39
ICS, N (% positive)	404 (32)	470 (38)	0·002	169 (26)	66 (39)	0·002
IgE (geometric mean)	34 (14–93)	49 (19–166)	<0·001	41 (16–104)	56 (15–203)	0·32
Sputum Eos %	0·23 (0·0–0·76)	0·65 (0·12–2·42)	<0·001
Blood Eos count/µL				150 (100–200)	230 (160–350)	<0·001
“Prior asthma”, N (% positive)	249 (20)	255 (21)	0·48	122 (19)	48 (29)	0·003
Childhood asthma, N (% positive)	97 (7·7)	118 (9·5)	0·10	52 (7·9)	20 (11·7)	0·07

Mann-Whitney rank sum test for continuous variables, results as median (25–75% interquartile range); Chi-square for categorical variables, results as N (% positive response).

In subjects stratified by mean sputum eosinophils at 1·25%, age, gender, race, BMI, smoking pack-years, cigarettes/day, and % current smokers did not differ (Table 1. Differential counts for leukocytes in sputum eosinophil groups are presented in supplement Table S2). Medication use in the sputum cohort did not differ, except increased use of ICS, and inhaled or nebulized bronchodilators in the ≥1·25% sputum eosinophils group (supplement Table S1). Higher sputum eosinophils, unlike higher blood eosinophils, did not have significantly increased IgE levels compared to low sputum eosinophil group. Nevertheless, IgE levels in low and high sputum eosinophil groups were similar to IgE levels in low and high blood eosinophil groups, respectively. In addition, the sputum eosinophil ≥1·25% group had elevated blood eosinophils (230/µL, p<0·001) and a greater proportion of subjects reporting a “prior asthma label” (p=0·003). Similar results for blood or sputum eosinophil stratification were obtained with higher cutpoints (≥300/mL blood eosinophils levels, or ≥2% sputum eosinophils, respectively; supplement Table S3).

Spirometry

Dividing by blood eosinophils, the ≥200/µL high eosinophil group had marginally lower values for pre-bronchodilator FEV1 % predicted (4·2%), and no difference post-bronchodilator (0·6%), compared to the <200/µL low eosinophil group (Table 2). The sputum eosinophils ≥1·25% group had greater differences between pre bronchodilator and post-bronchodilator FEV1 % predicted (10 and 5·7%, respectively), compared to <1.25% group. Due to safety exclusion of subjects with post-bronchodilator FEV1 % predicted <35% from sputum induction, fewer GOLD Stages 3 and 4 subjects were included in the sputum cohort. However, baseline and post-bronchodilator FEV1 % predicted showed larger differences between low and high sputum eosinophil groups than observed between low and high blood eosinophil groups. Reversibility of baseline FEV1 % predicted, was larger in the elevated sputum eosinophil group (p<0·001), but did not differ for the blood eosinophil groups. Similar observations were found between subgroups stratified by ≥300/µL blood and ≥2% sputum eosinophil cutpoints (supplement Table S4).

Table 2

Lung function for subjects stratified by mean blood or sputum eosinophils (Eos; 200/µL cutoff or 1·25% cutoff, respectively).

Variable	BloodEos<200/µL	BloodEos≥200/µL	PValue*	SputumEos<1.25%	SputumEos≥1.25%	PValue*
Number	1262	1237		656	171
Pre-bronchodilator:
FEV1 (L)	1.86 (1.22–2.54)	1.81 (1.16–2.55)	0.38	2.15 (1.57–2.77)	1.83 (1.38–2.32)	<0.001
FEV1 % predicted	70·5 (46·6 – 88·2)	66·3 (42 – 85·6)	0·006	75·7 (59·3–90·2)	65·7 (52–81·3)	<0·001
FVC % predicted	87.3 (73–99)	85.5 (72–97)	<0·001	90.9 (79–100)	87.1 (77–97)	0·06
FEV1/FVC	0·64 (0·49 – 0·73)	0·61 (0·47 – 0·72)	0·016	0·66 (0·58 – 0·74)	0·61 (0·52 – 0·69)	<0·001
Post-bronchodilator:
FEV1 (L)	2.05 (1.43–2.72)	2.03 (1.39–2.75)	0.62	2.34 (1.78–2.59)	2.11 (1.69–2.59)	0.003
FEV1 % predicted	75.9 (53–94)	76.5 (53–92)	0·85	82.9 (68–96)	77.2 (63–88)	0·001
FVC % predicted	92·9 (81– 104)	90·5 (79 – 102)	0·001	94·5 (85 – 105)	94·2 (86 – 104)	0·84
FEV1/FVC	0·66 (0·5 – 0·76)	0·63 (0·49 – 0·74)	0·004	0·68 (0·59 – 0·76)	0·64 (0·55 – 0·72)	<0·001
% FEV1 reversibility	9·3 (4·2 – 17·7)	9·8 (4·5 – 19)	0·46	8 (3·7 – 15·4)	11·6 (6 – 21·7)	<0·001

Mann-Whitney rank sum test for continuous variables, results as median (25–75% interquartile range).

Imaging

Indices of emphysema and air trapping at TLC and RV[24], respectively, did not differ between blood eosinophil groups (Table 3). In contrast, significantly higher emphysema indices (% voxels <−950 HU) in left upper and lower lobes, and right upper lobe were observed in sputum eosinophil ≥1·25%. In addition, air trapping (% voxels <−856 HU) and functional small airways disease assessed by parametric response mapping (PRM fSAD,[25]) were higher in subjects with ≥1·25% sputum eosinophils.

Table 3

Imaging parameters for subjects stratified by mean blood or sputum eosinophils (Eos; cutoff 200/µL, or cutoff 1·25%, respectively).

Variable	BloodEos<200/µL(N=1262)	BloodEos≥200/µL(N=1237)	Pvalue*	SputumEos<1.25%(N=656)	SputumEos≥1.25%(N=171)	Pvalue*
DENSITY MEASURES
TLC Left Upper Lobe %<−950 HU	3·22 (1·11 –11·84)	3·77 (1·20 –11·50)	0·66	2·24 (0·89 – 5·74)	2·88 (1·09 – 7·65)	0·046
TLC Right Upper Lobe %<−950 HU	2·78 (0·72 –12·97)	2·87 (0·73 –11·96)	0 76	1·73 (0·6 – 5·58)	2·43 (0·9 – 7·24)	0·011
TLC Left Lower Lobe %<–950 HU	2·09 (0·81–7·09)	2·42 (0·88–7·5)	0·23	1·60 (0·72 – 3·78)	1·98 (0·76 – 5·32)	0·044
RV BOTH Lungs %<−856 HU	17·4 (6·71–39·38)	18·91 (7·39–40·71)	0·28	12·52 (5·34–25·27)	17·20 (8·57–33·13)	0·001
PRM functional small airway disease	14 (4–33)	15 (4–34)	0·21	9 (3–22)	13 (6–26)	0·011
AIRWAY MEASURES
AVG WALL THICK apical right upper lobe	1·26 (1·15–1·37)	1·28 (1·16–1·39)	0·032	1·28 (1·18–1·38)	1·29 (1·19–1·42)	0·08
Taper Ratio - apical right upper lobe	0·038 (−0·01–0·086)	0·039 (−0·01–0·09)	0·87	0·03 (−0·01–0·08)	0·03 (−0·01–0·08)	0·93

Mann-Whitney rank sum test for continuous variables, results as median (25–75% interquartile range).

There was a small, 0·02 mm increase in average airway wall thickness at RB1 (prespecified pathway in apical segment of right upper lobe) for elevated blood eosinophils, but not for elevated sputum eosinophils. Neither blood nor sputum stratification showed any difference in airway tapering (an index of bronchiectasis).

The higher ≥300/µL blood eosinophil cutpoint did not alter density measures for emphysema or air trapping, but reduced significance for RB1 airway wall thickness. The higher ≥2% sputum eosinophil cutpoint maintained significance for both emphysema and air trapping indices (supplement Table S5).

Clinical Characteristics

Among subjects with ≥1·25% sputum eosinophils, there were fewer GOLD Stage 0 and increased GOLD Stage 2 subjects compared to subjects with <1·25% sputum eosinophils (p=0·0006, Table 4). The 6 min walk distance, BODE Index and COPD Assessment Score did not differ for either blood or sputum eosinophil stratifications. The blood eosinophil ≥200/µL group showed significantly higher frequency of self-reported wheezing (Table 4). St. George Respiratory Questionnaire (SGRQ) symptom score was also higher in the ≥200/µL blood eosinophil subgroup; both SGRQ total and symptom scores were significantly higher in the ≥1·25% sputum eosinophils subgroup.

Table 4

Clinical characteristics for subjects stratified by mean blood or sputum eosinophils (Eos; cutoff 200/µL or 1·25%, respectively).

Variable	BloodEos<200/µL(N=1262)	BloodEos≥200/µL(N=1237)	Pvalue*	SputumEos<1.25%(N=656)	SputumEos≥1.25%(N=171)	Pvalue*
GOLD Stage 0 N(%)/ 1 N (%)/ 2 N(%)/ 3 N(%)/ 4 N(%)	505 (40)/ 150 (12)/ 323 (26)/ 190 (15)/ 79 (6)	425 (34)/ 153 (12)/ 359 (29)/ 200 (16)/ 86 (7)	0·10	295 (45)/ 106 (16)/ 200 (31)/ 47 (7)/ 0 (0)	51 (30)/ 31 (18)/ 76 (44)/ 11 (6)/ 1 (1)	0·001
6 Minute Walk Distance (m)	418 (354 – 482)	410 (341 – 471)	0·12	426 (372 – 482)	426 (363 – 478)	0·40
BODE Index	1 (0 – 2)	1 (0 – 2)	0·29	0 (0 – 1)	1 (0 – 1.5)	0·09
COPD Score (CAT)	13 (7–20)	13 (7–19.5)	0·45	12 (7–19)	13 (8–20)	0·18
SGRQ (Total)	31·5 (14·9 – 48·2)	31·2 (16·4 – 47·1)	0·81	26·2 (14·0 – 43·6)	31·8 (17·2 – 47·1)	0·05
SGRQ (Symptoms)	45·3 (22·9 – 66·3)	48·8 (27·0 – 66·3)	0·037	45·2 (23·5 – 65·1)	53·6 (34 – 70)	0·004
Symptoms:
Wheezing, N (% positive)	741 (59)	788 (64)	0·018	389 (60)	116 (68)	0·07

Mann-Whitney rank sum test for continuous variables, results as median (25–75% interquartile range); Chi-square for categorical variables, results as N (% positive response).

The higher ≥300/µL blood eosinophils cutpoint showed a significant difference for GOLD Stages (supplement Table S6). The higher ≥2% sputum eosinophil cutpoint maintained significance for GOLD Stages, SGRQ total and symptoms scores, and became significant for BODE Index, SGRQ Impact and self-reported wheezing.

Exacerbations

Blood eosinophils ≥200/µL and sputum eosinophils ≥1·25% were tested for association with exacerbations (Table 5). Elevated blood eosinophils were not associated with any of the different categories of reported exacerbations. In contrast, elevated sputum eosinophils were associated with increased proportions of subjects with exacerbations requiring corticosteroids, exacerbations requiring treatment with any drug, and severe exacerbations requiring emergency department visit. The higher ≥300/µL blood eosinophil cutpoint did not show any association with exacerbations, but the higher ≥2% sputum eosinophil cutpoint demonstrated significance for all categories of reported COPD exacerbations (supplement Table S7). Tree classification of sputum and blood eosinophil association with exacerbations selected sputum eosinophils before blood eosinophils and showed similar cutpoints, sequentially < or ≥1·9% for sputum eosinophils and < or ≥ 176/µL for blood eosinophils, supporting the 2% and 200/µL cutpoints investigated for sputum and blood eosinophils in this study (supplement Figure S2). Sputum eosinophils >1·9% identified a subgroup of subjects with exacerbations (27 of 119 or 23%). In contrast, those with <1.9% sputum eosinophils showed a lower proportion of subjects with exacerbation (65 of 692 or 9%).

Table 5

Comparison of exacerbations ≥1 (in the previous year) for subjects stratified by mean blood or sputum eosinophils (Eos; cutoff 200/µL or 1·25%, respectively). All values are percentage positive.

Variable	BLOODEos<200	BLOODEos≥200	Pvalue*	SPUTUMEos<1.25%	SPUTUMEos≥1.25%	Pvalue*
N	1262	1237		656	171
Definition of exacerbation:
Total, N (%)	311 (25)	309 (25)	0·35	125 (19)	44 (26)	0·05
Healthcare Utilization, N (%)	294 (23)	291 (24)	0·36	125 (19)	43 (25)	0·07
Antibiotic treatment, N (%)	232 (18)	240 (19)	0·29	92 (14)	34 (20)	0·09
Corticosteroid treatment, N %)	199 (16)	209 (17)	0·27	66 (10)	32 (19)	0·002
Any drug-treatment, N (%)	265 (21)	273 (22)	0·29	105 (16)	39 (23)	0·033
Severe, N (%)	137 (11)	162 (13)	0·15	52 (8)	22 (13)	0·044

Chi-square, results as % positive response.

Blood Eosinophil Prediction of Sputum Eosinophils

ROC analyses demonstrated a relatively weak, although significant relationship for blood eosinophils to predict sputum eosinophils ≥1·25% (Figure 1, AUC=0·63, p<0·0001); ROC to predict ≥2% sputum eosinophils was similar (supplement Figure S3). Highest sensitivity and specificity for predicting sputum eosinophils ≥1·25% and ≥2% were found at 150/µL and 250/µL blood eosinophils, respectively, with equivalent, significant AUCs observed at adjacent cutpoints (supplement Table S8). Nevertheless, both associations had very large false discovery rates; 72% for blood eosinophils ≥150/µL to predict sputum eosinophils ≥1·25% (false negative rate of 22%), and 74% for blood eosinophils ≥250/µL to predict sputum eosinophils ≥2% (false negative rate of 50%).

Figure 1

ROC analysis for blood eosinophil prediction of sputum eosinophil. Blood eosinophils at cutpoints from 50/µL (highest sensitivity) to 500/µL (lowest sensitivity) were examined for correct prediction of sputum eosinophils < or ≥1·25%. Although significant (p<0·001), the area under the curve (AUC) was only 0·63, demonstrating a lack of strength for the prediction. Maximum sensitivity and specificity were observed at a blood eosinophil cutpoint of 150/µL.

Combined Blood and Sputum Eosinophil Phenotypes

The correlation between sputum eosinophils and blood eosinophils was poor, but significant (Figure 2; correlation coefficient r=0·178, p<0·001). However, numerous subjects (42%) had discordant blood and sputum eosinophil levels; either high in blood or in sputum, but not in the other compartment. Lung function and reported exacerbations for subjects in the two discordant quadrants and two concordant quadrants were compared (Table 6). High sputum eosinophil groups, without or with high blood eosinophils, had the lowest lung function. Lung function for the high blood eosinophil group without high sputum eosinophils did not differ from the group with low eosinophils in both blood and sputum. However, high blood eosinophil groups only in combination with elevated sputum eosinophils had more COPD exacerbations; exacerbations treated with corticosteroids (p=0·006) or severe (p=0·013) were significant.

Figure 2

Distribution of subject’s blood eosinophils with respect to sputum eosinophils. Although there is a significant association between blood and sputum eosinophils (EOS) (Pearson correlation coefficient r=0·178, p<0·001), use of the cutpoint ≥200/µL blood eosinophils (vertical red line) to predict sputum eosinophils ≥1·25% (horizontal red line) will mistakenly identify many subjects with lower sputum eosinophil% (lower right quadrant, pink shade) and miss many subjects with actual sputum eosinophils ≥1·25% (upper left quadrant, blue shade).

Table 6

Spirometry and exacerbations for subjects in sputum (spt) versus blood (cbc) eosinophil (Eos) distribution stratified by quadrant (blood Eos cutpoint ≥200/µL; sputum Eos cutpoint ≥1·25%):

	Low spt Eos+Low cbc Eos	Low spt Eos+High cbc Eos	High spt Eos+Low cbc Eos	High spt Eos+High cbc Eos	P value*
N	355	290	50	116
Pre-bronchodilator:
FEV1 % pred	77 (62–91)	74 (57–89)	62 (51–81)	66 (50–79)	<0·001
FVC % pred	93 (81–101)	88 (77–100)	89 (76–99)	87 (76–96)	0·033
ratio FEV1/FVC	0·67 (0·58–0·74)	0·66‡ (0·57–0·73)	0·61‡ (0·47–0·66)	0·61 (0·52–0·69)	<0·001
Post-bronchodilator:
FEV1 % pred	84 (70–97)	81 (66–95)	76 (63–87)	77 (63–90)	0·005
FVC % pred	96 (86–106)	93 (84–104)	98 (88–104)	93 (85–104)	0·24
ratio FEV1/FVC	0·69 (0·59–0·77)	0·67‡ (0·59–0·76)	0·63‡ (0·49–0·72)	0·64 (0·58–0·70)	<0·001
% reversibility	7·8 (3·6–15·3)	8·5‡ (4·0–16·6)	14·5‡ (6·8–22·6)	11·4 (5·8–20·5)	<0·001
Definition of exacerbation:
Total N (%)	76 (21)	47 (16)	10 (20)	32 (28)	0·07
Requiring HCU N (%)	74 (21)	45 (16)	9 (18)	31 (27)	0·07
Antibiotics treatment N (%)	56 (16)	36 (12)	10 (20)	23 (20)	0·21
Corticosteroid treatment N (%)	37 (10)	25 (9)	6 (12)	24 (21)	0·006
Any Drug treatment N (%)	61 (17)	38 (13)	10 (20)	27 (23)	0·08
Severe (ED or hospital) N (%)	35 (10)	15 (5)	3 (6)	17 (15)	0·013

Kruskal-Wallis for continuous variables, results as median (interquartile range); Chi-square for exacerbation categories, results as N (% positive response).

Indicates High sputum Eos+ Low cbc Eos quadrant was significantly different from Low spt Eos+ High cbc Eos quadrant by Dunn’s method.

Additional Stratifications for Blood and Sputum Eosinophil Subgroups

Subjects were stratified by “prior asthma label” or ICS use and examined for interaction with high eosinophils (blood eosinophils ≥200/mL or sputum eosinophils ≥1·25%) on lung function and exacerbations. No interactions were significant (supplement Tables S9, and S10). Subjects who did not have acceptable sputum slides for various reasons (N=1498) were stratified by blood eosinophil counts to determine whether these subjects represented a phenotype with different characteristics (supplement Table S11). There was slightly higher proportion of subjects using ICS, lower lung function and increased proportions of GOLD Stage 3 and 4 subjects as would be expected in these groups which contained subjects with lower lung function and therefore ineligible for sputum induction, but otherwise resembled the larger cohort of smokers.

Another stratification examined whether blood eosinophil groups < or ≥200/µL showed differences when restricted to just those subjects in the sputum cohort (supplement Table S12). The sputum cohort divided into blood eosinophil subgroups had slightly greater proportion of current smokers, less ICS use and slightly better lung function, but did not show the same radiologic, clinical or exacerbation diffierences observed for sputum eosinophil stratification.

We examined whether associations with worse lung function and quality of life, and greater exacerbations, emphysema, and air trapping in the high sputum eosinophil group were due to elevated sputum neutrophils in addition to high sputum eosinophils. There was no difference in mean sputum neutrophil % between high and low sputum eosinophil groups (Table S2). Stratification of the sputum cohort into 4 groups based on < or ≥1·25% eosinophils + < or > 68% neutrophils (mean ± std deviation for sputum neutrophils: 68% ± 21%), confirmed differences across low and high sputum eosinophil groups but did not show significant post-hoc differences between the high eosinophil + high neutrophil and high eosinophil + low neutrophil subgroups (Supplement additional results, Table S13).

DISCUSSION

The results of this study from the SPIROMICS cohort, a smoking cohort that includes a spectrum of COPD severity defined by GOLD stages, confirms that elevated sputum eosinophils, but not blood eosinophils alone, identify a subset of COPD subjects with more severe airflow obstruction, worse quality of life, greater emphysema and air trapping, and exacerbations. Using sputum eosinophil stratification at either the mean, ≥1·25%, or ≥2%, we found significant associations with COPD exacerbations, including severe and those requiring corticosteroid therapy. In addition, significant associations were found for lower lung function, baseline and post-bronchodilation including increased bronchodilator reversibility; respiratory symptoms; emphysema and air trapping by QCT; and COPD severity by GOLD Stage. In contrast, blood eosinophils alone, at ≥200/µL, or the even higher cutpoint at ≥300/µL, showed no association with COPD exacerbations, and associations with other phenotypic markers were smaller or non-significant. Although there was an increase in SGRQ symptom scores for higher eosinophils in both blood and sputum compartments, and wheeze (found only for blood, possibly due to the larger N for that group) the differences between low and high eosinophil groups for these variables were greater in the sputum group. In addition, there was no difference in CAT scores for either blood or sputum, which tends to diminish the validity for this observation. Importantly, although the relationship between blood and sputum eosinophilia was statistically significant, blood eosinophils did not reliably predict sputum eosinophils, showing a 72–74% false-discovery rate and a 50% false negative rate for sputum eosinophils ≥2%. Lung function data stratified by high and low sputum and blood eosinophils showed no relationship with high blood eosinophils unless combined with high sputum eosinophils, while high sputum eosinophils even in the absence of blood eosinophils was associated with lower lung function. However, COPD subjects with both high sputum and blood eosinophils exhibited both decreased lung function and more frequent exacerbations. These findings among current and former smokers in a large multicenter cohort with a specified range of COPD severity have important implications for proposed use of blood eosinophils alone as a predictive biomarker to guide individualized COPD therapies.

Our results extend observations from previous studies in COPD cohorts, including ECLIPSE, which focused primarily on neutrophilic airways inflammation,[1] and, though reporting eosinophil presence,[5] did not address association of eosinophilia with indices of COPD severity.[4,5] The importance of our findings and of Th2 inflammation in COPD are emphasized by the recent report of Th2 gene expression overlap in airway epithelial samples from asthma and COPD cohorts,[8] and by shared clinical and biologic characteristics between asthma and COPD reported in several recent studies.[5,7,17,28-29] However, differences are noted between SPIROMICS and other COPD cohorts. COPD gene enrolled a larger cohort (N=10,000) which was slightly older (minimum 45 yr) and had a lower smoking history (>10 pack year), but phenotyping with induced sputum was not performed.[30] Sputum was also unavailable in Copenhagen General Population Study,[20] WISDOM,[21] INSPIRE, and TRISTAN.[31] Although these studies report exacerbations correlating with blood eosinophils, it is important to note that the entry requirements included past history of COPD exacerbations which can impact the results since a past history of exacerbation is the most important factor predicting future exacerbations.[18]

In addition emphasis on persistent Th2 inflammation in COPD[32], has focused on eosinophils as predictors of exacerbations. Bafadhel and colleagues reported a cluster analysis using blood and sputum biomarkers; peripheral blood eosinophils predicted sputum eosinophil-associated exacerbations of COPD.[6] Sputum airway and peripheral blood eosinophils have been used to direct corticosteroid treatment and reduce occurrence of COPD exacerbations.[9-11] The ECLIPSE study reported that 1483 subjects if stratified by blood eosinophils did not differ for COPD exacerbation rate in the previous year.[5] We confirm that higher blood eosinophils are not associated with COPD exacerbations except combined with elevated sputum eosinophils or with other characteristics such as a previous history of exacerbation.[18] However, in SPIROMICS higher sputum eosinophils alone are associated with exacerbations even in mild to moderate COPD.

Eosinophil levels have been suggested to indicate response to corticosteroids, anti-IL5, or anti-IL5 receptor therapy.[9-12,33] In a retrospective analysis of two COPD exacerbation studies with long-acting beta-agonists and inhaled corticosteroids, Pascoe showed that subjects with higher blood eosinophils had greater reduction in COPD exacerbations.[34] These observations suggest eosinophils may be important in development or potential biomarker of some COPD exacerbations. However, two factors may influence previous observations correlating elevated blood eosinophil groups with greater COPD exacerbations; selection criteria requiring recent exacerbation and perhaps lower lung function are both related to future COPD exacerbations and may be surrogate markers of increased sputum eosinophils. Elevated blood eosinophils, if also combined with elevated sputum eosinophils, associate with COPD exacerbations, as shown for the SPIROMICS cohort in this report. However, blood eosinophils alone were not associated with exacerbations, even when combined with “prior asthma label”. This latter observation contrasts with the association of blood eosinophils ≥275/µL with all cause mortality in 662 subjects, but that study found no change after exclusion of subjects with asthma.[35] Blood eosinophils in our study were associated with COPD exacerbations only in the context of higher sputum eosinophils.

We also examined other characteristics in the blood and sputum eosinophil subgroups of SPIROMICS subjects that might suggest overlap with asthma: bronchodilator reversibility, IgE levels, and childhood asthma. Blood eosinophils at two different cutpoints (either ≥200 or ≥300/µL) did not have higher levels of acute bronchodilator reversibility, while elevated sputum eosinophils showed greater reversibility. IgE levels were significantly higher in the elevated blood eosinophil group, although not in sputum, but IgE levels in SPIROMICS blood and sputum eosinophil subgroups were well below median (91 IU/ml) and high IgE cutpoint (173 IU/ml) reported in a recent study of asthma-COPD overlap syndrome.[36] Only ≥2% sputum eosinophils showed increased report of childhood asthma, which represent a small subgroup (12·5%).

There was greater use of prescribed inhaled corticosteroids in both high blood and sputum eosinophil groups. This was observed despite an expected reduction in eosinophils with corticosteroid therapy. Use of corticosteroids in the higher eosinophil groups potentially reflects individuals more likely to have had exacerbations, consistent with GOLD guideline recommendations for corticosteroids in COPD patients with frequent exacerbations.[13] Limitations of this report include somewhat milder COPD in the group who were able to successfully perform induced sputum. For safety reasons, SPIROMICS subjects with post-bronchodilator FEV1 % predicted <35% did not have sputum induction, limiting the sputum subgroup to GOLD Stages 0–3. Although our cohorts (both for blood and sputum eosinophil analyses) included smoking subjects (>40 pack years) with preserved lung function, these subjects were included because they have been shown to have greater symptoms, exacerbations, activity limitations and radiologic evidence of airway disease.[37] These findings are consistent with early COPD in this subgroup.[37] Although SPIROMICS exacerbations data was retrospective, validity of retrospective data for future risk of COPD exacerbation has been shown in the ECLIPSE where self-reported exacerbation from the previous year had predicted exacerbations during the first year of follow up, more accurately than all other variables examined.[18] An additional limitation, at least in clinical settings, is the difficulty in performing accurate sputum analysis. Even in the SPIROMICS network with centralized training for sputum induction and processing there were still reasons preventing sputum analysis on all eligible subjects as indicated in the supplemental methods. However analysis of those who did not have sputum analysis, stratified by blood eosinophils did not differ substantially from the larger cohort.

Of interest, longitudinal follow-up of the SPIROMICS cohort may be used to confirm the observations of Hospers and colleagues that peripheral eosinophils are associated with all cause mortality over a period of 30 years.[35] Alternatively, the differences in decline of lung function associated with blood eosinophils < or ≥2% observed in the much smaller study over 9 years by Rogliani and colleagues should be examined in the larger SPIROMICS cohort longitudinally.[38] In summary, using the larger and comprehensive phenotypic characterization of the SPIROMICS cohort, we show that stratification by elevated sputum eosinophil inflammation identified a subgroup with more severe COPD, having decreased lung function, greater emphysema and air trapping, and greater COPD exacerbations. Peripheral blood eosinophils identified a subgroup with decreased lung function without other indices of more severe COPD, specifically exacerbations unless examined on the background of elevated sputum eosinophils. Moreover, blood eosinophils did not accurately predict sputum eosinophils. These observations confirm the importance of assessing eosinophils in the airways. In the future, it will be important to follow these subjects with higher sputum eosinophilia longitudinally to determine whether long term effects on the progression of COPD.

Supplement Figure S1. Consort Diagram for subjects recruited and reason for removal from analysis. All normal, never smokers were removed. Although 90% of subjects (including normal never smokers) may have been eligible for sputum induction, there were several reasons that reduced the actual number of sputum slide samples available for analysis: no sputum produced upon completion of induction, no sputum processing form entered, removal of aliquots for mucus analysis and microbiome prior to processing leaving too little remaining sample for cytospin slide preparation, slides not sent to central slide reading center, and finally slide counts that were deemed unacceptable (leukocyte cell count <100 or too high, >80% squamous epithelial cells). There were 16 subjects with acceptable sputum differential count but without blood counts who were added to those with both acceptable sputum counts and blood counts (N=811).

Supplement Figure S2. Classification tree diagram for model of exacerbations by sputum and blood eosinophils. The Root has 811 subjects with 92 exacerbations in previous year. The first number in each node is the number of subjects without exacerbations; the second number is the number with exacerbations. The model first divides the subjects based on sputum eosinophils < and ≥1.9% and secondly divides the subjects by both sputum and blood eosinophils (< or >176/µL).

Supplement Figure S3. ROC analysis for blood Eos prediction of sputum Eos. Blood Eos at cutpoints from 50/µL (highest sensitivity) to 500/µL (lowest sensitivity) were examined for correct prediction of sputum Eos < or ≥2%. Although significant (p<0·001), the area under the curve (AUC) was only 0·64, demonstrating a lack of strength for the prediction. Maximum sensitivity and specificity were observed at a blood Eos cutpoint of 250/µL, but with very large false discovery rate (74%) and false negative rate (50%).