Several studies have demonstrated that treatment with inhaled corticosteroids (ICS) in COPD patients with higher blood eosinophil counts results in a greater decrease in the exacerbation rate and greater improvement in lung function and quality of life compared with patients with low blood eosinophil concentrations.1–5 However, the association of high blood eosinophil levels and increased risk of exacerbations is less clear. Some studies have reported a significant risk of exacerbations in patients with high eosinophil counts,6–8 while others have not.9–12 Larger epidemiological studies are required to investigate the relationship between eosinophil levels and the risk of exacerbation in COPD. The use of population-based databases can help to increase the knowledge on real-life patterns of blood eosinophilia and their relationship with exacerbations.

The aims of the study were to quantify blood eosinophil levels, describe the characteristics of patients with different blood eosinophil levels and determine the risk of exacerbations associated with eosinophil blood levels and their variability in patients with COPD.

This was an observational, retrospective, population-based study with longitudinal follow-up. The data used in this study was obtained from the SIDIAP (System for the Development of Research in Primary Care) database, a computerised database containing the anonymized patient records of 5.8 million people registered in one of the 279 primary care centres of the Catalan Health Service (approximately 80% of the population of Catalonia, Spain).13 All the general practitioners in the Catalan Health Service use the same specific software called ECAP to record the clinical information of their patients. Health professionals gather this information using ICD-10 codes and structured forms designed for the collection of variables, such as smoking history and body mass index (BMI), as well as test results such as blood tests and spirometries.

The index date for each patient was defined as the date of the last blood eosinophil measurement in stable state within the baseline period from January 1st, 2014 to December 31st, 2014. Baseline information was collected from the clinical visit closest to the index date and data on health resource use (HRU) was collected from the 12-month period prior to the index date. In order to analyse the risk associated with blood eosinophil levels, information on HRU was collected from the 12 months after the index date

The study was approved by the Research and Ethics Committee of the Jordi Gol Institute of Research In Primary Care (Barcelona, Spain). Since the analysis was performed using anonymised clinical records, informed consent was not necessary.

The prevalence of COPD in the SIDIAP database is 2.70% (confidence interval [CI]95%: 2.68–2.72) in individuals of all ages and 10.5% (CI95%: 10.4–10.6) in individuals>65 years old.13 In 2014, 57,209 COPD patients fulfilling the inclusion and exclusion criteria were identified (Fig. 1), with a mean eosinophil count of 253.9cell/μl (SD=190). The mean age was 70.2 years (SD=10.4), and 74% were men. Only 44.5% had a spirometry with a mean FEV1 (%) of 64.1% (SD=22.3%). Up to 51.6% of patients had at least one exacerbation in the year before the index date (Table 1).

Fig. 1.

Disposition of the population of the study.

(0.24MB).

Characteristics of the patients according to different blood eosinophil levels.

The distribution of blood eosinophils in the 4 subgroups is depicted in Fig. 2a. The mean number of exacerbations in the previous year was higher in patients with the lowest and the highest eosinophil levels (1.03 [SD=1.31] and 0.97 [SD=1.22], respectively) compared with the intermediate groups (0.88 [SD=1.16] and 0.87 [SD1.16], respectively) (Table 1, Fig. 3a). The percentage of patients admitted for an exacerbation followed the same distribution. The treatment administered to the patients in the different groups was very similar (Table 2).

Fig. 2.

Distribution of the population according to the established cut-off of blood eosinophilia: (a) distribution in the global population; (b) distribution in the study population without inhaled corticosteroids.

(0.12MB).

Fig. 3.

Frequency of exacerbations per year during follow-up in patients according to: (a) group of baseline blood eosinophil count; (b) variability in blood eosinophil count during the previous year.

(0.09MB).

In the present study, the mean blood eosinophil concentrations in patients with COPD was 254cells/μl, with concentrations>300cell/μl in 28.4%. Patients with the lowest blood eosinophil levels (<150cells/μl) presented the highest frequency of exacerbations and hospitalisations, both the year before and after eosinophil measurement, followed by patients with the highest eosinophil levels (>500cells/μl). These results remained unchanged on analysing only patients who did not receive ICS in either the year prior to or after eosinophil measurement. The classification of patients in the four categories of blood eosinophil counts was quite stable, with 62% remaining within the same category and only 4.3% showing large variability, defined as two measurements belonging to non-consecutive categories. Interestingly, patients with the greatest variability were those with more frequent exacerbations and hospitalisation during the follow-up year.

There is increasing interest in the use of blood eosinophil count as a biomarker in COPD.1,17–20 Although no definitive prospective studies have yet been conducted, and there is no agreement on the most effective eosinophil cut-off level to identify responders to ICS, there is increasing recognition that high blood eosinophil counts are associated with better response to ICS in a dose-response fashion.3–5 However, the use of blood eosinophil levels as a biomarker of increased risk of exacerbations is more controversial.1,17–20

Some of the differences observed in demographic or clinical characteristics were statistically significant due to the large sample size, but most were small at less than 10%. Patients with a higher eosinophil count had more leukocytes and more frequently had an ACO phenotype.21–23 The lack of significant or relevant differences in the characteristics of COPD patients with different blood eosinophil levels has consistently been reported in previous studies, despite the different cut-offs used.10–12,24,25 The possible use of eosinophil level as a predictor of risk of exacerbations and hospitalisations is of greater interest. In the present study, patients with the lowest eosinophil levels (<150cells/μl) had the highest number of moderate and severe exacerbations, followed by those with the highest eosinophil concentrations (>500cells/μl). Differences in the frequency of exacerbations in patients in the intermediate groups were significant, albeit of small magnitude. Our results are in contrast with those observed in a study using electronic medical records and insurance claims data in the US, in which patients with ≥150cells/μl had higher all-cause and COPD-related HRU.26 However, the U-shape distribution in the risk of moderate exacerbations observed in our study was also found in a Copenhagen cohort of 7225 patients.8 In that study, patients with less than 130cells/μl and those with more than 340cells/μl showed the highest frequency of moderate exacerbations. However, the risk of severe exacerbations was only increased in those with high eosinophil counts, and the cut-off of 340cells/μl demonstrated an AUC of 0.63 to identify patients with increased risk of exacerbations. Interestingly, this increased risk was relevant for severe exacerbations (OR=1.76), but quite modest for moderate exacerbations (OR=1.15). In contrast, we were unable to identify any cut-off value that could reliably indicate an increased risk for either exacerbations or frequent exacerbations. These results concur with others performed in different COPD populations. In a series of 294 patients, Turato et al.24 did not observe an increased risk of exacerbations in those with eosinophil levels greater than 150cells/μl, and no significant difference was found in blood eosinophil counts in exacerbators compared with non-exacerbators. Similarly, Shin et al.25 did not find any difference in eosinophil counts in exacerbators and non-exacerbators in a sample of 299 COPD patients. The same results were observed in the CHAIN and BODE cohorts,10 as well as in the BPCO French cohort,11 the SPIROMIC cohort,12 a recent study on hospitalised patients,27 and in a large population of 27,557 patients from the UK database.9 In these studies, no differences were found in exacerbation history using different blood eosinophil cut-offs.

The possible increased risk of severe exacerbations with increased eosinophil counts also contradicts the consistent finding in several studies conducted in different populations of a better survival associated with higher blood eosinophil levels.10,24,25,28 These controversial results indicate that blood eosinophils may either be a very poor predictor of exacerbation risk or they may be a reasonable predictor of risk only in a subgroup of patients with COPD, namely the frequent exacerbators; in fact, our population was predominantly made up of non-exacerbators. In a recent study analysing data from the ECLIPSE and the COPDGene cohorts, a blood eosinophil concentration>300cells/μl was a good predictor of future risk of exacerbations only in patients with frequent exacerbations, but not in patients with 0 or 1 exacerbation in the previous year.6 Finally, it is possible that only patients with a persistently high eosinophil count have an increased risk of exacerbations,29 but the need for more than one determination makes the implementation of the use of eosinophils levels difficult in clinical practice.

In contrast, there is greater agreement regarding the possible role of blood eosinophil counts as a predictor of response to ICS in COPD. Different studies have consistently shown that increased blood eosinophil concentrations are a reliable biomarker for better clinical response to ICS in COPD.3–5 In order to evaluate the possible confounding effect of the use of ICS, we replicated the analysis in a subgroup of more than 24,000 patients who had not received any ICS in the year prior to and after the baseline date. The results obtained were practically identical to those found with the total sample, thereby ruling out the influence of ICS use on the results, as suggested in previous studies.27,30

Biological measurements are subject to variability, but in the case of biomarkers, large variability may jeopardise their usefulness in clinical practice. Interestingly, different studies have suggested good reproducibility of blood eosinophil measurements, with interclass correlation coefficient (ICC) of measurements of blood eosinophils of around 0.8.31,32 In the ECLIPSE and COPDGene, the ICC of repeated blood eosinophil measurements in stable state was 0.57.6 However, other studies have highlighted the possible switching of categories (high/low eosinophil counts) in between 13% and 43% of patients.8,10,25,32 In our study, blood eosinophil measurements were quite stable, with repeated measurements belonging to the same category in nearly two thirds of the patients, and only 4.3% of measurements jumped at least two categories. This is important because a switch in category may not always represent a biological change; i.e. a change from 295 to 305cells/μl is defined as a switch in category if the threshold is 300cells/μl, but this does not represent a significant biological change. Our results concur with those of Southworth et al.32 who observed a change between the lowest and highest categories of blood eosinophils in only 1.7% of their patients, and Greulich et al.33 found that eosinophil counts were quite stable especially when the counts were low (<150cells/μl). Similarly, our results also concur with those obtained in large databases, such as the Clinical Practice Research Datalink in the Netherlands, which includes the data of +39,000 COPD patients. In this latter study, stability, defined as values persistently above or below 340cells/μl, was reported in 85% of patients at 6 months and in 62% at 2 years.34

A novel observation of our study was the increased risk of HRU in the small subgroup of patients with large variability in blood eosinophil counts. This “instability” in blood eosinophil counts may be a marker of clinical instability; nonetheless, further studies are needed to confirm these results.

It was interesting to observe that approximately 30% of patients with a diagnosis of COPD were not receiving regular treatment. This high percentage of patients without treatment was also observed in previous studies14 in primary care and is very different from the data obtained in respiratory departments22,35 and indicated that guidelines are not well implemented in primary care.

The current study has some limitations. As in most database studies, full categorisation of COPD patients is not always possible, as spirometry was available in less than half of the patients; however, a validated diagnostic algorithm was used to try to minimise the risk of misclassification. Although we attempted to capture all the episodes of exacerbations, there is a risk of underreporting. Notwithstanding, this underreporting should affect all the categories of blood eosinophils equally, and therefore, should not significantly affect the results of the study. The population was predominantly of non-exacerbators, and therefore, specific studies in a population with a high frequency of exacerbations are required. Variability in eosinophil counts was only analysed in patients in whom more than one measurement was available, and therefore, it is possible that these patients are the most severe and require greater medical attention. However, the clinical characteristics of this subgroup were not significantly different from the remaining patients; nevertheless, this circumstance would have biased the results towards more HRU in this subgroup, which was not the case. In contrast, the strength of this study is its representativeness, since more than 80% of the general population is included in this database, and blood eosinophil values were collected from lab reports, thereby excluding any possible transcription errors. Eosinophils were only analysed in absolute numbers, but the quartiles described correspond to a distribution of mean percentages of approximately 1%, 2.5%, 4.5% and 7%.

In summary, our results do not support the use of blood eosinophil counts as a marker of risk for HRU in the general COPD population at low risk of exacerbations, although an increased risk was observed in association with large variability of blood eosinophil levels. Nonetheless, this should be confirmed in other studies. These results extend our understanding of the role of eosinophils as a biomarker in COPD.

This study was funded by GlaxoSmithKline (study HO-16-6446).

Marc Miravitlles has received personal fees from GlaxoSmithKline during the conduct of the study and has received speaker or consulting fees from AstraZeneca, Bial, Boehringer Ingelheim, Chiesi, Cipla, CSL Behring, Laboratorios Esteve, Ferrer, Gebro Pharma, GlaxoSmithKline, Grifols, Menarini, Mereo Biopharma, Novartis, pH Pharma, Rovi, TEVA, Verona Pharma and Zambon, and research grants from GlaxoSmithKline and Grifols.

Bernardino Alcázar-Navarrete reports personal fees from GSK, grants, personal fees and non-financial support from Novartis AG, personal fees and non-financial support from Boehringer Ingelheim, personal fees and non-financial support from Chiesi, grants, personal fees and non-financial support from Laboratorios Menarini, personal fees from Gebro, personal fees from AstraZeneca, outside the submitted work. The remaining authors have no relevant competing interests to disclose.