Abstract
Eosinophilic inflammation and airway remodelling are characteristic features of asthma, but the association between them is unclear. We assessed associations between blood eosinophils and lung function decline in a population-based cohort of young adults.
We used linear mixed models to analyse associations between blood eosinophils and spirometry at 21, 26, 32 and 38 years adjusting for sex, smoking, asthma and spirometry at age 18 years. We further analysed associations between mean eosinophil counts and changes in spirometry from ages 21 to 38 years.
Higher eosinophils were associated with lower forced expiratory volume in 1 s (FEV1)/forced vital capacity (FVC) ratios and lower FEV1 % predicted values for both pre- and post-bronchodilator spirometry (all p-values ≤0.048). Although eosinophil counts were higher in participants with asthma, the associations between eosinophils and spirometry were similar among participants without asthma or wheeze. Participants with mean eosinophil counts >0.4×109 cells·L−1 between 21 and 38 years had greater declines in FEV1/FVC ratios (difference 1.8%, 95% CI 0.7–2.9%; p=0.001) and FEV1 values (difference 3.4% pred, 95% CI 1.5–5.4% pred); p=0.001) than those with lower counts.
Blood eosinophils are associated with airflow obstruction and enhanced decline in lung function, independently of asthma and smoking. Eosinophilia is a risk factor for airflow obstruction even in those without symptoms.
Abstract
Blood eosinophils are associated with lung function decline even in people without asthma or wheeze http://ow.ly/AOEn30iXyIJ
Introduction
Eosinophilic airway inflammation and airway remodelling leading to persistent airflow obstruction are characteristic features of asthma, but the link between them is unclear [1]. Although tissue eosinophils are associated with remodelling of the airway wall, it has not been established whether this is a causal association [2–4]. Controlling eosinophilic inflammation with inhaled corticosteroids reduces exacerbations [5–7], but has not yet been shown to prevent the development of fixed airflow obstruction [1].
One reason for the uncertainty is that measurement of eosinophilic airway inflammation using induced sputum is unsuitable for routine clinical practice or large-scale epidemiological studies [7, 8]. Peripheral blood eosinophil counts have emerged as a promising and easily measured marker of eosinophilic airway inflammation [9–12]. High blood eosinophils are associated with poor asthma control and risk of exacerbations [13–15]. In addition, blood eosinophil counts predict the response to inhaled corticosteroids in patients with chronic obstructive pulmonary disease (COPD) [16–18], and the response to anti-interleukin-5 therapy in asthma [19]. Blood eosinophil counts have been associated with lower forced expiratory volume in 1 s (FEV1) values in participants with and without asthma [13, 20–23], but not all studies have found this [24]. Blood eosinophils did not predict an enhanced decline in FEV1 in asthmatic adults [22, 25, 26], but a greater decline in FEV1 was observed in COPD patients with high blood eosinophil counts who were not treated with inhaled corticosteroids [18].
We investigated associations between blood eosinophil counts and lung function in a population-based birth cohort of young adults. We hypothesised that eosinophilic inflammation would be associated with airflow obstruction and a decline in lung function among participants with asthma.
Methods
The Dunedin Multidisciplinary Health and Development Study is a longitudinal investigation of health and behaviour in an unselected population-based cohort of individuals born in the only maternity hospital in Dunedin in 1972/1973 (http://dunedinstudy.otago.ac.nz) [27, 28]. The cohort was formed when 1037 children living in the greater Dunedin area (91% of eligible births) were assessed at age 3 years. Study members represent the full range of socioeconomic status in the South Island of New Zealand and are primarily of New Zealand/European ethnicity. The study has a high rate of follow-up: 95% of living study members participated in the most recent assessment at age 38 years. Written informed consent was obtained for each assessment. The Otago ethics committee approved the study.
Childhood asthma was defined as a parent-reported diagnosis of asthma with compatible symptoms or asthma medication within the previous year at 9, 11 or 13 years [28]. Adult asthma was defined as a self-reported diagnosis with compatible symptoms or medication within the previous year at ages 21, 26, 32 or 38 years. Wheeze included all episodes of reported wheeze in the previous year, excluding only one or two episodes lasting <1 h [27]. Inhaled corticosteroid use was recorded at each age. Current smoking was defined as daily smoking for ≥1 month in the previous year. Cumulative tobacco exposure was quantified in pack-years (equivalent to 20 cigarettes·day−1 for 1 year) [28].
At ages 21, 26, 32 and 38 years, blood was drawn at the end of the assessment day. Eosinophil counts were obtained from automated complete blood counts and were reported to two decimal places at ages 21, 26 and 32 years, but to one decimal place at age 38 years. Eosinophil counts >0.4×109 cells·L−1 were regarded as high [15].
Spirometry was performed at ages 18, 21, 26, 32 and 38 years. This was repeated after inhaled salbutamol at 18, 26, 32 and 38 years [28]. Height was measured at each assessment. Skin-prick tests for 11 common aeroallergens were performed at ages 21 and 32 years [29]. Participants were considered to have atopic sensitisation if they had at least one positive test (weal diameter ≥2 mm greater than negative control) at either age.
Statistical analyses
We compared sex, asthma, atopy, smoking and lung function at age 38 years among those with and without complete eosinophil data. Non-parametric (Spearman) correlations, intraclass correlations and the persistence of high (>0.4×109 cells·L−1) counts between ages were used to assess the consistency of eosinophil counts between ages.
Eosinophil counts were transformed with natural logarithms to approximate normal distributions. To allow transformation of zero values, 0.1 was added before transformation and later deducted from the geometric mean values. Linear mixed models were used to assess predictors of log-eosinophil counts with fixed effects for age and a random effect for participants to accommodate repeated measurements. These models included current smoking, childhood asthma, adult asthma, atopic sensitisation, inhaled corticosteroid use and sex as predictors.
Linear mixed models with fixed effects for age and a random effect for participants were also used to assess whether log-eosinophil counts were associated with airflow obstruction. The primary outcome measure was the pre-bronchodilator FEV1/forced vital capacity (FVC) ratio from ages 21 to 38 years. Secondary outcome measures were FEV1 % predicted and FVC % pred and post-bronchodilator spirometry [30]. Analyses adjusted for spirometry measurements at age 18 years, sex, childhood and adult asthma, inhaled corticosteroids and smoking pack-years. Analyses tested for interactions between eosinophils and asthma diagnoses. Additional separate analyses were performed for those with and without adult asthma. Body mass index (BMI) is associated with lower FEV1/FVC ratios among females in this cohort [31], and because atopic sensitisation was associated with higher eosinophil counts, further analyses adjusted for these factors. In addition, we tested for interactions between eosinophils and smoking. The models were repeated with adjustment for childhood and adult wheeze instead of diagnosed asthma. Spirometry data were approximately normally distributed and were not transformed before analysis. Models were checked by inspection of histograms of residuals and scatterplots of residuals against fitted values.
To further assess whether eosinophil counts were associated with decline in lung function, changes in FEV1/FVC ratio and FEV1 % pred from age 21 to 38 years were plotted with respect to the mean eosinophil counts across these ages. Changes in lung function were compared between those with and without mean eosinophil counts >0.4×109 cells·L−1using t-tests.
Analyses used Stata 13 (StataCorp, College Station, TX, USA). Spirometry from pregnant participants and one implausible measurement were excluded. Otherwise, analyses used all available data. Two-sided p-values <0.05 were considered statistically significant.
Results
Characteristics of the participants are shown in table 1. Those who missed at least one blood sample had similar rates of asthma, atopy, and FEV1/FVC ratios, but were more likely to smoke and had lower FEV1 values (online supplementary table S1). Geometric mean eosinophil counts at each age are shown in table 2.
Participant characteristics
Eosinophil counts at each age and correlations between ages
Eosinophil counts differed between ages (Wald p<0.001): counts were higher at age 21 years than older ages (all pairwise p≤0.001), but were not significantly different between ages 26, 32 and 38 years (all pairwise p≥0.054). 21-year-olds did not have higher eosinophils when these were expressed as a percentage of total leukocytes. Among 691 participants who had eosinophils measured at all four assessments, geometric mean (95% CI) eosinophil counts were 0.25 (0.23–0.27)×109 cells·L−1 for those with adult asthma and 0.17 (0.17–0.18)×109 cells·L−1 for those without (t-test p<0.001). Spearman correlations between counts at each age are shown in table 2. Correlations were similar among those with and without asthma (online supplementary table S2). The interclass correlation coefficient for individual eosinophil counts was 0.58 (p<0.001) over the 17-year follow-up. High eosinophils (>0.4×109 cells·L−1) were found in 193 (20%) out of 961 measurements among 269 participants with adult asthma and 200 (8%) out of 2424 measurements among 701 participants without adult asthma (Chi-squared p<0.001). Among those with high eosinophils, the count remained high at the next measurement for 52% of those with asthma and 42% of those without. For participants with complete eosinophil data, 81 (40%) out of 201 of those with asthma and 96 (20%) out of 490 without asthma had at least one count >0.4×109 cells·L−1 (Chi-squared p<0.001), but only 13 (6%) of those with asthma and five (1%) of those without had high counts at every assessment (Chi-squared p<0.001). 35 (17%) of those with asthma and 30 (6%) of those without had mean eosinophil counts >0.4×109 cells·L-1 across the four tests (Chi-squared p<0.001).
Childhood and adult asthma, atopic sensitisation, current smoking and inhaled corticosteroid use were associated with higher eosinophil counts (table 3). Smokers did not have higher eosinophils when these were expressed as a percentage of total leukocytes (online supplementary table S3).
Predictors of blood eosinophil counts
Eosinophil counts were associated with lower pre- and post-bronchodilator FEV1/FVC ratios and FEV1 values (table 4). There were no interactions between adult asthma and eosinophil counts for any spirometry measure (all interactions p≥0.359): when analysed separately, eosinophil counts were associated with lower FEV1/FVC ratios and FEV1 values among those with and without asthma, although these associations were not statistically significant for post-bronchodilator FEV1. Adjustments for atopic sensitisation and BMI made no material difference to the analyses (not shown). There were no interactions between current or cumulative smoking and eosinophil counts for any spirometry measure (all interactions p≥0.166).
Associations between eosinophil counts and pre- and post-bronchodilator spirometry in participants with and without asthma
Eosinophil counts were associated with lower pre- and post-bronchodilator FEV1/FVC ratios and FEV1 values in analyses adjusting for wheeze and for pre-bronchodilator values in the subgroups with and without wheeze. Post-bronchodilator FEV1/FVC ratios and FEV1 values were only significantly associated with eosinophils among those reporting wheeze although there was a tendency (p=0.081) for an association with lower FEV1/FVC ratios among those without wheeze (table 5).
Associations between eosinophil counts and pre- and post-bronchodilator spirometry in participants with and without wheeze
Scatterplots of mean eosinophil counts and changes in FEV1/FVC ratios and FEV1 values between ages 21 and 38 years are shown in figure 1. The coefficients of the fitted regression lines were similar for those with and without asthma (interaction p-values ≥0.781). Higher eosinophils were associated with greater declines in FEV1/FVC and FEV1 in both groups, although the association with FEV1 was not statistically significant among asthmatics (p=0.052). After adjustment for smoking pack-years, all coefficients were statistically significant (p-values ≤0.033). Participants with mean eosinophil counts >0.4×109 cells·L−1 had mean excess declines of 1.8% in FEV1/FVC (95% CI 0.7–2.9%, p=0.002) and FEV1 3.3% pred (95% CI 1.3–5.3% pred, p=0.001) between ages 21 and 38 years compared to those with lower eosinophil counts. In comparison, those who smoked ≥10 pack-years had mean excess declines of 2.4% in FEV1/FVC (95% CI 1.6–3.2%, p<0.001) and FEV1 3.5% pred (95% CI 2.0–4.9, p<0.001).
Scatter plots of changes in forced expiratory volume in 1 s (FEV1)/forced vital capacity (FVC) ratios and FEV1 between the ages of 21 years and 38 years and mean eosinophil counts between the ages of 21 years and 38 years. a) The coefficient for change in FEV1/FVC ratio for participants without asthma is −5.4% per 1×109 cells·L−1 eosinophils (p<0.001). For those with asthma the coefficient is −4.7% per 1×109 cells·L−1 eosinophils (p=0.028). b) The coefficient for change in FEV1 for participants without asthma is −7.8% pred per 1×109 cells·L−1 eosinophils (p=0.004). For those with asthma the coefficient is −7.0% pred per 1×109 cells·L−1 eosinophils (p=0.052).
Discussion
In this population-based cohort of young adults, blood eosinophil counts were associated with lower FEV1/FVC ratios and lower FEV1 values. These associations were found for both pre- and post-bronchodilator measurements and were independent of childhood or adult asthma diagnoses, cumulative smoking exposure and lung function measured at the beginning of adulthood. In addition, higher eosinophil counts were associated with a greater decline in FEV1/FVC ratios and FEV1 values. The declines in FEV1 and FEV1/FVC ratio among those with high mean eosinophil counts across the four assessments were of a similar magnitude to the declines observed among those who smoked =10 pack-years over the same time. The findings demonstrate that blood eosinophilia is associated with airflow obstruction and indicate that it may be an important risk factor for lung function decline.
As far as we are aware, this is the first study to explore longitudinal associations between blood eosinophils and lung function in a general population sample. Although participants with asthma had higher mean eosinophil counts, blood eosinophil counts were associated with worse lung function and similar declines in FEV1 and FEV1/FVC among participants with and without asthma and even among those who did not report wheezing symptoms. This suggests that eosinophilic inflammation contributes to lung function decline and airflow obstruction even among those without diagnosed airway disease or relevant symptoms. This is consistent with evidence from some cross-sectional studies that high blood eosinophil counts are associated with lower FEV1 values in nonasthmatics [20, 21, 23], although this was not observed in the large National Health and Nutrition Examination Survey (NHANES) III study [24]. Findings from previous longitudinal research on eosinophils and lung function have been inconsistent [22, 23, 26, 32]. The largest study found that although blood eosinophils were associated with lower FEV1 values at baseline among adults with asthma, they did not predict lung function decline over 12 years [22], while a post hoc analysis of the placebo arm of an inhaled corticosteroid trial found that COPD patients with high blood eosinophils had a more rapid decline in FEV1 [18]. Differences in the characteristics of the populations studied, cut-offs for defining high blood eosinophils and potential treatment responses may explain some of the inconsistencies between these studies.
We used blood eosinophils as a marker of airway eosinophilic inflammation. Studies of lung function decline and direct markers of airway eosinophilia are limited to smaller asthmatic cohorts and the findings are inconsistent. One study found that neither blood nor sputum eosinophils predicted FEV1 decline over 5 years in difficult-to-treat asthma [25], whereas a study of asthmatics with fixed airflow obstruction found that sputum eosinophils predicted 5-year FEV1 decline [33]. Another study found that patients who had experienced a rapid decline in FEV1 over 8 years had higher blood and sputum eosinophils [34]. The variability in sputum eosinophils counts, rather than persistently high levels, was found to be associated with a greater decline in FEV1 over 6 years in patients with refractory asthma [35]. Bronchial biopsy studies have found that lung function decline is associated with inflammatory changes, including the numbers of CD4- and CD8-positive cells, but not with eosinophils [34, 36, 37]. These inconsistencies may be due to differences in the characteristics of the asthmatic subgroups studied and potential treatment responses, as well as small sample sizes and the difficulties in directly measuring airway inflammation.
As this is an observational study, we cannot prove that eosinophilic inflammation causes airflow obstruction or whether this could be prevented with treatment. Sputum eosinophil counts are associated with asthma exacerbations and treatment with inhaled corticosteroids reduces this risk [5–7], but the finding that blood eosinophils were associated with lower lung function among participants without either asthma or wheeze makes it highly unlikely that the association can be explained by recurrent asthma exacerbations among those with high blood eosinophils.
Our findings suggest that long-term eosinophilic airway inflammation may worsen lung function decline and increase the risk of COPD. This is important because it may be possible to prevent this with treatment. A post hoc analysis of the Inhaled Steroids in Obstructive Lung Disease in Europe (ISOLDE) study found that inhaled fluticasone appeared to reduce the decline in FEV1 in COPD patients with high blood eosinophils [18]. We need information from prospective randomised controlled trials to establish whether treating eosinophilic inflammation with inhaled corticosteroids would improve long-term outcomes for lung function in either COPD or asthma [1, 38].
Few factors are known to cause lung function decline among healthy young adults: only exposure to pollutants (including tobacco smoke) has been found to lead consistently to airflow obstruction at a population level. Our findings suggest that eosinophilic inflammation may be another factor, and this effect appears to be independent of asthma. The drivers of this inflammation are unknown, but it is plausible that exposure to environmental or occupational allergens contribute to this. However, atopic sensitisation was not an independent predictor of lung function, and adjusting the analyses for atopic sensitisation made no material difference to the findings. Higher BMI has been found to be associated with lower FEV1/FVC ratios in females in this cohort [31], but was not associated with eosinophil counts and adjusting the analyses for BMI did not alter the findings for either sex.
Blood eosinophils were moderately correlated between ages. This is consistent with previous reports [10, 22], and extends the period of observation to 17 years and to young adults with and without asthma. There was considerable variability around our prespecified cut point of 0.4×109 cells·L−1. Variability around this and other cut points has been noted in other studies, indicating that a single test is insufficient to characterise long-term eosinophilia [39]. As anticipated, atopic sensitisation and asthma were associated with higher counts. In keeping with other observations [32], we found that smokers had higher absolute eosinophils counts than nonsmokers, but this could be explained by higher total leukocyte counts. Total leukocyte and eosinophil counts were higher at age 21 years than older ages, but percentage eosinophil counts were similar. We analysed absolute eosinophil counts as recommended [40], but repeating the analyses using percentage eosinophils made no material differences to the findings.
The study has a number of strengths including measurements of eosinophils and spirometry on four occasions over 17 years with a high rate of follow-up. We adjusted for many potential covariates or effect modifiers, including childhood and adult asthma, symptoms of wheeze, smoking exposure and spirometry at age 18 years. However, there are some limitations. Peripheral blood eosinophil counts are only a proxy measure of airway eosinophilia, although obtaining direct measurements of airway eosinophils in this large epidemiological study would be unfeasible with current techniques. Asthma diagnoses were self-reported and this may have led to some misclassification. We do not have post-bronchodilator measures of lung function at age 21 years. Participants who missed one or more blood samples were more likely to smoke and had slightly lower FEV1 values, but restricting the analyses to those who had eosinophil counts at every age made no material difference to the findings. Eosinophils counts were only reported to one decimal place (×109 cells·L−1) at age 38 years, but restriction of the analyses to ages 21–32 years also made little difference to the findings. So far, we have only followed the cohort to early mid-adult life and we cannot extrapolate the findings to older adults or the risk for clinically diagnosed COPD.
The implications for clinical practice are not yet clear. We need to know the long-term risk of developing COPD among people with high blood eosinophil counts and whether anti-inflammatory treatment alters this risk. The finding that blood eosinophils are associated with airflow obstruction in people without asthma or wheeze raises questions about the role of asymptomatic eosinophilia: could eosinophils allow us to identify asymptomatic individuals at high risk of developing chronic lung disease and facilitate targeted prevention strategies? In addition, we need to understand more about the intrinsic and environmental drivers of eosinophilic inflammation and whether these are amenable to intervention at individual or population-based levels.
In summary, we have found that blood eosinophil counts are moderately stable across young adult life and that high counts are associated with airflow obstruction. Although eosinophil counts were higher in participants with asthma, associations between blood eosinophil counts and airflow obstruction were present in those without asthma or symptoms of wheeze. The findings suggest that persistent blood eosinophilia is an independent risk factor for the development of airflow obstruction even in those without respiratory disease.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary tables ERJ-02536-2017_Supplement
Acknowledgements
We are grateful to the study members and their friends and families for their continued support. We thank the respiratory physiologists and interviewers who obtained the measurements and Andrew Gray (University of Otago, Dunedin, New Zealand) who provided statistical advice. We also thank Richie Poulton (University of Otago), the study director, Phil A. Silva, the study founder, and Terrie Moffitt and Avshalom Caspi (both Duke University, Durham, NC, USA) for their support. M.R. Sears holds the AstraZeneca Chair in Respiratory Epidemiology at McMaster University.
Footnotes
This article has supplementary material available from erj.ersjournals.com
Author contributions: R.J. Hancox developed the hypothesis for this analysis with input from M.R. Sears and I.D. Pavord. M.R. Sears and R.J. Hancox collected the data. R.J. Hancox analysed the data and drafted the manuscript. All authors provided critical review of the manuscript and approved its submission.
Conflict of interest: R.J. Hancox received lecture and advisory board honoraria from Astra Zeneca, and lecture honoraria from Menarini, outside the submitted work.
Conflict of interest: I.D. Pavord received medical writing support from GSK; and received speaker's honoraria from AstraZeneca, Boehringer Ingelheim, Aerocrine, Almirall, Novartis and GSK, honoraria for attending advisory board panels from Almirall, AstraZeneca, Boehringer Ingelheim, Dey Pharma, GSK, MSD, Schering-Plough, Novartis, Napp Pharmaceuticals and RespiVert, and sponsorship for attending international scientific meetings from AstraZeneca, Boehringer Ingelheim, GSK and Napp Pharmaceuticals, outside the submitted work.
Conflict of interest: M.R. Sears has received lecture and advisory board honoraria from Astra Zeneca, outside the submitted work.
Support statement: The Dunedin Multidisciplinary Health and Development Research Unit is funded by the Health Research Council of New Zealand. Duke University (Durham, NC, USA) provided additional funding for respiratory data collection at age 38 years. Funding information for this article has been deposited with the Crossref Funder Registry.
- Received December 6, 2017.
- Accepted March 11, 2018.
- Copyright ©ERS 2018
References
