If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Little Belt Hospital, Department of Medicine, Beriderbakken 4, 7100, Vejle, DK, DenmarkUniversity of Southern Denmark, Department of Regional Health Research, Campusvej 55, 5230, Odense, DK, Denmark
Hvidovre Hospital, Department of Respiratory Diseases, Kettegård Alle 30, 2650, Hvidovre, DK, DenmarkUniversity of Copenhagen, Institute of Clinical Medicine, Blegdamsvej 3B, 2200, Copenhagen N, DK, Denmark
Little Belt Hospital, Department of Medicine, Beriderbakken 4, 7100, Vejle, DK, DenmarkUniversity of Southern Denmark, Department of Regional Health Research, Campusvej 55, 5230, Odense, DK, Denmark
Bispebjerg and Frederiksberg Hospital, Department of Clinical Pharmacology, Bispebjerg Bakke 23, 2400, Copenhagen NV, DK, DenmarkUniversity of Copenhagen School of Pharmaceutical Sciences, Department of Drug Design and Pharmacology, Universitetsparken 2, 2100, Copenhagen, DK, Denmark
In COPD patients exists a correlation between effect of ICS and eosinophil.
•
ICS might not be beneficial for COPD patients with eosinophil count <150 cells/μl.
•
Exacerbation risk increased in patients with high eosinophil count when stopping ICS.
Abstract
Inhaled corticosteroid (ICS) in patients with chronic obstructive pulmonary disease (COPD) has been debated for 20 years. In our systematic literature review and meta-analysis, we addressed the following: Should patients with COPD and a blood eosinophil count (EOS) of, respectively, a) < 150 cells/μl, b) 150–300 cells/μl, and c) > 300 cells/μl continue treatment with ICS? Protocol registered in PROSPERO (CRD42020178110) and funded by the Danish Health Authority.
We searched Medline, Embase, CINAHL and Cochrane Central on 22nd July 2020 for randomized controlled trials (RCT) of ICS treatment in patients with COPD (≥40 years, no current asthma), which analyzed outcomes by EOS count and where >50% of patients used ICS prior. We used the GRADE method. Meta-analyzes for the outcomes were divided into EOS subgroups and analyzed for differences.
We identified 11 RCTs with a total of 29,654 patients. A significant difference (p < 0.00001) between the three subgroups’ reduction of risk of moderate to severe exacerbation was found. Rate ratios for EOS counts: <150 cells/μL was 0.88 (95%CI: 0.83, 0.94); 150–300 cells/μL was 0.80 (95%CI: 0.69, 0.94); >300 cells/μL was 0.57 (95%CI: 0.49, 0.66). Overall, the certainty of the effect estimates was low to very low due to risk of bias, unexplained heterogeneity, few RCTs, and wide confidence intervals.
A clear correlation was demonstrated between effect of continued ICS treatment (number of exacerbations, lung function, and quality of life) and increasing EOS count. Our meta-analyses suggested that treatment with ICS seemed beneficial for everyone except patients with EOS count below 150 cells/μl.
The use of inhaled corticosteroid (ICS) as a routine treatment in patients with chronic obstructive pulmonary disease (COPD) has been debated back and forth for more than 20 years.
The positive effects of ICS in preventing exacerbations in COPD and in slowing health status decline in patients with COPD was first established in 2000, in the ISOLDE trial [
Randomised, double blind, placebo controlled study of fluticasone propionate in patients with moderate to severe chronic obstructive pulmonary disease: the ISOLDE trial.
]. Later, the TORCH trial in 2007 surprisingly failed to show reduction in all-cause mortality; however, it did manage to demonstrate significant benefits in all other outcomes [
] showed the superiority in preventing exacerbation in patients with COPD of a long-acting beta-agonist (LABA) plus a long-acting muscarinic antagonist (LAMA), as compared to an ICS-LABA combination. This seriously questioned the role of ICS as a first-line treatment in patients with COPD. Furthermore, the European Medicines Agency (EMA) in 2016 included pneumonia in all ICS product summaries as a common side effect (1–10%) [
], and blood eosinophils may therefore be a relevant biomarker in the decision-making process on whether ICS treatment should be continued or withdrawn in patients with COPD. The level of eosinophils in sputum may be predictive for the effect of both systemic and inhaled steroid [
]. Studies have suggested that patients with COPD with blood eosinophils above 2% respond better to treatment with systemic steroids than patients with blood eosinophils below 2% [
], using the relative measurement of blood eosinophils as a percentage of the total count of leukocytes. However, blood eosinophils are mostly measured in absolute numbers, i.e. cells per microliter, as recommended by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) [
Stability of blood eosinophils in patients with chronic obstructive pulmonary disease and in control subjects, and the impact of sex, age, smoking, and baseline counts.
]. Therefore, it may be relevant to consider when to measure EOS. Furthermore, it appears that, in both healthy individuals and patients with COPD, variation in EOS can occur throughout life [
Stability of blood eosinophils in patients with chronic obstructive pulmonary disease and in control subjects, and the impact of sex, age, smoking, and baseline counts.
A recent study indicated that patients with an EOS count of more than 300 cells/μL have positive effect of treatment with ICS, whereas patients with EOS count below 100 cells/μL only have modest effect [
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
]. There is no consensus on EOS limit values. A Danish study in a representative population sample found that the mean EOS count was 180 cells/μL in patients with COPD [
In the light of this, the aim of this systematic literature review and meta-analysis is to summarize the evidence regarding benefits and harms of continued treatment with ICS in patients with COPD in relation to EOS count and, based on this, determine the EOS subgroup that would benefit the most from either withdrawal or continued treatment with ICS. This summary of all published evidence is needed, as the previous systematic reviews in this area have had a low number of included studies (e.g. Chalmers J.D. et al. from 2020 [
Blood eosinophil count-guided corticosteroid therapy and as a prognostic biomarker of exacerbations of chronic obstructive pulmonary disease: a systematic review and meta-analysis.
The following research question was addressed: Should patients with COPD and an EOS count of, respectively: a) below 150 cells/μL, b) between 150 and 300 cells/μL, and c) above 300 cells/μl continue treatment with ICS?
2. Study design and methods
Prior to conducting our systematic review and meta-analysis, a protocol was submitted to The International Prospective Register of Systematic Review (PROSPERO) on 30th June 2020 and registered on 31st July 2020 (CRD42020178110), following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [
]. The protocol included the review question, a search strategy, inclusion and exclusion criteria, a strategy for risk of bias assessment, and a data synthesis plan, including a suggested approach, should substantial heterogeneity be found.
We followed the Cochrane Collaboration established methods, using the Cochrane Handbook for Systematic Reviews of Interventions [
]. The transparent framework for developing and presenting summaries of evidence suggested by GRADE (Grading of Recommendations, Assessment, Development and Evaluations) was rigorously followed [
The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration.
] (PRISMA checklist is provided in the Supplementary Material).
2.1 PICO
The research question and inclusion/exclusion criteria for the review included the components of PICO (Population, Intervention, Comparison, and Outcome), and the timeframe for follow-up was defined [
Studies were included if the patients were 40 years or older and had COPD and an EOS count of either: a) below 150 cells/μL, b) between 150 and 300 cells/μL or c) more than 300 cells/μL. The review authors chose to also include studies that had EOS limits other than 150, and 300, if the limit used was obviously close to one limit and far from another (pragmatically, we used ± 50 from a limit). Studies including patients with current asthma were excluded. The patients should have undergone treatment with ICS for at least three months either before and/or during the trial. The review authors chose to include studies where more than 50% of included patients were treated with ICS before the start of the study, and where possible ICS free run-in periods were of a maximum of 2 weeks.
We included randomized controlled trials (RCTs) with placebo investigating the effect of ICS either as continued treatment or new treatment, and placebo-controlled RCTs investigating the withdrawal of ICS either immediately or by tapering down to zero.
The comparator group received no treatment with ICS. In both the intervention and control groups, other inhaled medication during the study (apart from ICS) was allowed.
The definitions of outcomes are specified in Table 1.
Table 1Overview and definition of outcomes, including timeframe and priority.
Outcomes
Timeframe
Critical/Important
Clinically relevant difference
Exacerbations per year (moderate or serious) (all definitions will be included)
Longest follow-up (min 3 mths from start)
Critical
25%
Number of patients with serious adverse events per year
Longest follow-up (min 3 mths from start)
Critical
10%
Dropout rate (any reason)
Longest follow-up (min 3 mths from start)
Critical
10%
Pneumonia per year (all definitions will be included)
Longest follow-up (min 3 mths from start n)
Important
Lung function (FEV1)
Longest follow-up (min 3 mths from start)
Important
Number of patients with improvement in lung function (FEV1)
Longest follow-up (min 3 mths from start)
Important
Dyspnoea
Longest follow-up (min 3 mths from start)
Important
Quality of Life
Longest follow-up (min 3 mths from start)
Important
In the included studies, moderate exacerbation was defined as the need for oral steroids or antibiotics. Serious exacerbation was defined as need for contact with hospital (either as emergency room visit or hospitalization). Serious adverse event was defined as an event which either: was life-threatening; resulted in death, hospitalization, prolonged hospitalization, disability or permanent damage; or required intervention to prevent permanent impairment or damage. Clinically relevant difference was only decided for critical outcomes. FEV1: Forced expiratory volume in 1 s.
2.2 Search strategy, data collection and quality assessment
We sought relevant articles reporting completed RCTs by searching electronic databases including Medline via OVID (including the Cochrane Database of Systematic Reviews), Embase via OVID (including the Cochrane Database of Systematic Reviews), CINAHL via EBSCO and Cochrane Central (including ClinicalTrials.gov, and WHO's and the European Union's trial databases). First, we searched for existing systematic reviews from the period 2010–2020, to identify eligible RCTs from the references (included and excluded), and secondly, we searched for primary RCTs, either from the period 2000–2020, or using the latest search date of any eligible existing systematic review of high quality that covered our research questions (based on an AMSTAR (A MeaSurement Tool to Assess systematic Reviews) evaluation) [
]. The searches for systematic reviews were conducted on 1st July 2020 by KB and the searches for primary RCTs were conducted on 22nd July 2020 also by KB. The search strategy, including keywords and justification of publication restrictions, is available in supplementary material S1. Manual searches of reference lists/bibliographies of included secondary RCTs were performed. We consulted content experts in the field from the guideline group (consisting of AL, PK, DAD, CJ, KB, MNH) and reference group (consisting of OH, NSG, HRC, CV) to ascertain whether any relevant RCTs were missing. Conference abstracts were considered if data were not published elsewhere [
]. Study authors were not contacted to identify additional RCTs or unpublished data.
References identified in the defined search strategy on both systematic reviews and individual primary RCTs were imported to the bibliography manager RefWorks. After de-duplication, the remaining references were imported into Covidence software for literature screening and data management [
]. Titles and abstracts of potentially eligible RCTs were screened independently by two out of a group of review authors (DAD, PK, TM), who were blinded to each other's decisions. Subsequently, the full text of potential RCTs was screened independently by two out of a group of review authors (DAD, PK, TM) for eligibility. Discrepancies were resolved through discussion and when necessary with the entire guideline group. Neither of the review authors were blinded to the journal titles, study authors/institutions or year of publication. Assessment of the RCTs included study settings, population demographics and baseline characteristics, details on intervention and control conditions, study design, outcome and time of measurement. Two out of the group of review authors (AU, JFR, IC, EMG and MNH) independently extracted data. Discrepancies were identified and resolved through discussion in the review team and when necessary with a third reviewer (DAD). Covidence [
], the methodological quality of systematic reviews was assessed independently by two review authors (DAD, MNH), who were blinded to each other's decisions. The tool includes the following 11 questions: 1. Was an ‘a priori’ design provided? 2. Was there duplicate study selection and data extraction? 3. Was a comprehensive literature search performed? 4. Was the status of publication used as an inclusion criterion? 5. Was a list of studies (included and excluded) provided? 6. Were the characteristics of the included studies provided? 7. Was the scientific quality of the included studies assessed and documented? 8. Was the scientific quality of the included studies used appropriately in formulating conclusions? 9. Were the methods used to combine the findings of studies appropriate? 10. Was the likelihood of publication bias assessed? 11. Were potential conflicts of interest included? Discrepancies were identified and discussed between the two review authors. The risk of bias in the primary studies was independently assessed, using the Cochrane Risk of Bias Tool [
], by two out of a group of review authors (AU, JFR, IC, EMG, DAD, and MNH). The tool includes the following domains: Randomization sequence generation; Treatment allocation concealment; Blinding of patients and personnel; Blinding of outcome assessors; Completeness of outcome data; Selective outcome reporting; Other sources of bias. Discrepancies were identified and discussed between the two review authors and, where necessary, discussed with the entire guideline group.
The certainty of the effect estimates of each outcome was evaluated using the GRADE method [
], with four possible rating levels: high, moderate, low, and very low. In randomized trials, the certainty of the evidence starts at high. If there are issues regarding risk of bias, inconsistency, indirectness, imprecision, and/or risk of publication bias, the certainty can be downgraded one or two levels within each domain. The overall certainty of the evidence was determined by the lowest certainty of the critical outcomes.
2.3 Strategy for data synthesis
Data were combined in a standard meta-analysis to generate a pooled measure of effect estimates, using either: a) rate ratio and 95% confidence interval (CI), b) risk ratio 95% CI in dichotomized outcomes, or c) mean difference (MD) 95% CI in continuous outcomes.
To prepare data for synthesis, some data conversions were needed. If data was reported in mL [
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
], we calculated a mean based on the data provided in their supplemental material, i.e. EOS estimates ranging from 5 to 140, 150 to 300 and 310 to 2510. When the studies included more than one intervention group, the intervention groups were combined.
Random effect models with inverse variance weights were used to account for expected heterogeneity of the effect. Statistical heterogeneity was quantified using I2 statistic [
], with an I2 value greater of 50% considered to be substantial heterogeneity. Sensitivity analyses were performed to investigate causes of heterogeneity and the robustness of including non-peer reviewed conference abstracts. Where applicable, data were divided into subgroups according to the predefined groups of EOS count: a) below 150 cells/μL, b) between 150 and 300 cells/μL or c) above 300 cells/μL. Test for subgroups were tested using Chi-square, and p-value below 0.05 was considered statistically significant. When more than 10 RCTs were included, funnel plot was produced to assess the risk of publication bias across the RCTs.
A sensitivity analysis was conducted to investigate if the effect estimates of the critical outcomes differed between RCTs that included only patients with prior ICS treatment, and RCTs where more than 50% but less than 100% of patients used ICS prior to inclusion.
Review Manager Software (version 5.3) (Nordic Cochrane Centre, Cochrane Collaboration, Copenhagen, Denmark) was used to produce the analyses, together with forest and funnel plots.
Blood eosinophil count, a marker of inhaled corticosteroid effectiveness in preventing COPD exacerbations in post-hoc RCT and observational studies: systematic review and meta-analysis.
] was identified that had slightly different inclusion criteria than our research question. From this systematic review, nine RCTs (eight publications) [
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
Blood eosinophil counts, exacerbations, and response to the addition of inhaled fluticasone furoate to vilanterol in patients with chronic obstructive pulmonary disease: a secondary analysis of data from two parallel randomised controlled trials.
Blood eosinophil count and exacerbations in severe chronic obstructive pulmonary disease after withdrawal of inhaled corticosteroids: a post-hoc analysis of the WISDOM trial.
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
Blood eosinophil count, a marker of inhaled corticosteroid effectiveness in preventing COPD exacerbations in post-hoc RCT and observational studies: systematic review and meta-analysis.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
Blood eosinophil counts, exacerbations, and response to the addition of inhaled fluticasone furoate to vilanterol in patients with chronic obstructive pulmonary disease: a secondary analysis of data from two parallel randomised controlled trials.
Blood eosinophil count and exacerbations in severe chronic obstructive pulmonary disease after withdrawal of inhaled corticosteroids: a post-hoc analysis of the WISDOM trial.
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
Blood eosinophil count and exacerbations in severe chronic obstructive pulmonary disease after withdrawal of inhaled corticosteroids: a post-hoc analysis of the WISDOM trial.
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
], also addressed patient populations with EOS count above 150 cells/μl. The 11 RCTs included a total of 29,654 patients. Flow charts can be found in supplementary material S2.
3.2 Review of the evidence
The interventions consisted of ICS in combination with LABA or both LABA and LAMA, and comparison groups were either LABA alone or in combination with LAMA (Table 2).
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
Blood eosinophil counts, exacerbations, and response to the addition of inhaled fluticasone furoate to vilanterol in patients with chronic obstructive pulmonary disease: a secondary analysis of data from two parallel randomised controlled trials.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
Blood eosinophil count and exacerbations in severe chronic obstructive pulmonary disease after withdrawal of inhaled corticosteroids: a post-hoc analysis of the WISDOM trial.
The duration of the interventions varied between 24 and 52 weeks. Among the RCTs, the patients were primarily male (60–80%), with an average age of 63–65.3 years, average body mass index (BMI) between 25.11 and 27.98 kg/m2, between 40 and 70% were current smokers and the patients’ COPD severity, assessed by pulmonary impairment measured by FEV1 (GOLD classification), were: 0% mild, 32% moderate, 52% severe and 16% very severe. Two RCTs [
Blood eosinophil counts, exacerbations, and response to the addition of inhaled fluticasone furoate to vilanterol in patients with chronic obstructive pulmonary disease: a secondary analysis of data from two parallel randomised controlled trials.
] had not reported the severity of the lung impairment. To be included in the RCTs, all patients had to be stable (no exacerbation or pneumonia) for four weeks before inclusion.
The risk of bias can be seen in the forest plot of each outcome (Fig. 1, Fig. 2, Fig. 3, Fig. 4 and supplementary material S4–S7).
Fig. 1Meta-analysis as a forest plot for the outcome risk of moderate to severe exacerbations. Sub-grouped by blood eosinophil (EOS) count and tested for subgroup differences.
Fig. 4Meta-analysis as a forest plot of the outcome lung function. Mean difference is shown in liters. Sub-grouped by blood eosinophil (EOS) count and tested for subgroup differences.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
] differed from the other RCTs, and found no difference in tests for the subgroups (INSTEAD + SUNSET versus other RCTs: exacerbations p = 0.41; lung function p = 0.66; serious adverse events p = 0.30; dropout p = 0.98), as seen in supplementary material S4.
Of the 11 RCTs we used for our first question, three had a different EOS limit than below 150 cells/μL. FORWARD (<181.6) [
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
]. The RCTs measured EOS at baseline, where patients had been stable for at least four weeks.
3.3 Results for critical outcomes
Continued treatment with ICS was likely to reduce the risk of moderate to severe exacerbations, regardless of EOS, as seen in Fig. 1. However, it was uncertain whether the effect estimate was clinically relevant (minimum clinically relevant difference: 25%). For patients with EOS counts below 150 cells/μL, the reduction was 12% (rate ratio: 0.88; 95% CI: 0.83–0.94). For patients with EOS counts between 150 and 300 cells/μL, the reduction was 20% (rate ratio: 0.80; 95% CI: 0.69–0.94). For patients with EOS counts above 300 cells/μL, there was a clinically relevant reduction of 43% (rate ratio: 0.57; 95% CI: 0.49–0.66). Testing of differences in subgroups between EOS counts below 150 cells/μL, between 150 and 300 cells/μL and above 300 cells/μL, respectively, showed that there was a significant difference in the reduction of the risk of moderate to severe exacerbation in the three groups (p < 0.00001). When we performed sensitivity analyses on exacerbation rate with (0.80 [0.74, 0.87]) and without (0.80 [0.73, 0.87]) the results from the conference abstract from the INSTEAD trial, the results were robust. There was moderate confidence in the effect estimate, as the certainty was downgraded due to severe risk of bias due to insufficient allocation concealment, risk of selective outcome reporting (almost all RCTs had subgroup analyzed on EOS counts as a post-hoc analysis), and consistent involvement from the pharmaceutical industry in the experimental design and reporting of results. For EOS counts between 150 and 300 cells/μL there was no downgrading for inconsistency, although I2 was 63%. This was because all RCTs favoured the treatment group to a greater or lesser degree, and the inconsistent results were thus between RCTs that showed moderate to large or very large effects. The observed heterogeneity was therefore not expected to have a significant impact on the assessment parameters.
Continued treatment with ICS was unlikely to affect the number of patients experiencing serious adverse events (risk ratio: 0.96; 95% CI: 0.9–1.03, see Fig. 2). Serious adverse events had not been reported in EOS subgroups, but we assumed that EOS counts had no influence. Confidence in the effect estimate was downgraded to moderate due to severe risk of bias, which is due to insufficient allocation concealment, as well as consistent interference from the pharmaceutical industry in experimental design and reporting of results.
Continued treatment with ICS may reduce dropout rate in the RCTs (risk ratio: 0.78; 95% CI: 0.73–0.84). Dropout rate was not calculated in EOS subgroups, and therefore there may be serious risk of indirectness between the population which were calculated in this outcome and the population defined in the review question. Confidence in the effect estimate was low due to serious risk of bias, which is due to insufficient allocation concealment, and consistent involvement from the pharmaceutical industry in the experimental design and reporting of results.
3.4 Results for important outcomes
Continued treatment with ICS was likely to increase the risk of pneumonia calculated independently of EOS subgroups (risk ratio: 1.39; 95% CI: 1.19–1.63, see supplementary material S5). Confidence in the effect estimate is moderate due to the serious risk of bias. The risk of bias was due to insufficient allocation concealment, as well as consistent interference from the pharmaceutical industry in experimental design and reporting of results.
Continued treatment with ICS for patients with EOS counts below 150 cells/μL was unlikely to affect lung function (FEV1 mean difference: 0.02 L; 95% CI: 0.00–0.05 L, see Fig. 4). For patients with EOS counts between 150 and 300 cells/μL and over 300 cells/μL continued treatment with ICS improved lung function with 0.05 L (FEV1 mean difference; 95% CI: 0.03–0.07 L) and 0.06 L (FEV1 mean difference; 95% CI: 0.04–0.09 L), respectively. Test of differences in subgroups between the different concentrations of EOS showed that there was a difference in how continued treatment with ICS affected lung function across the subgroups (p = 0.04), with a greater improvement of continued treatment with ICS in patients with EOS counts above 150 cells/μL, compared to patients with EOS counts below 150 cells/μL. Confidence in the effect estimate is moderate due to the serious risk of bias, due to insufficient allocation concealment, as well as consistent interference from the pharmaceutical industry in experimental design and reporting of results.
We found no RCTs reporting the number of patients with a clinically meaningful improvement in lung function.
Continued treatment with ICS was unlikely to reduce dyspnea (transitional dyspnea index (TDI) mean difference 0.08; 95% CI: −0.21 – 0.36, see Fig. 5). There was no difference between EOS subgroups (p = 0.59, see Fig. 5).
Fig. 5Meta-analysis as a forest plot of the outcome dyspnea (transitional dyspnea index (TDI)). Sub-grouped by blood eosinophil (EOS) count and tested for subgroup differences.
Continued treatment with ICS was overall likely to increase quality of life (st. george's respiratory questionnaire (SGRQ) mean difference −1.99; 95% CI: −3.04 to −0.95, see Fig. 6), though not clinically relevant (minimal clinically relevant difference: 4 [
]) and not significantly for all EOS subgroups (SGRQ mean difference for patients with EOS counts below 150 cells/μL −1.20; 95% CI: −2.87 – 0.46, see Fig. 6).
Fig. 6Meta-analysis as a forest plot of the outcome quality of life (St. George's Respiratory Questionnaire (SGRQ)). Sub-grouped by blood eosinophil (EOS) count and tested for subgroup differences.
Based on the findings from the included RCTs, we found that the overall effect of continued use of ICS in patients with COPD varied according to the EOS count, with increased effect of ICS with increasing EOS count. Our findings are in accordance with previous literature [
Blood eosinophil count, a marker of inhaled corticosteroid effectiveness in preventing COPD exacerbations in post-hoc RCT and observational studies: systematic review and meta-analysis.
], but are supported by more data and outcomes. The relationship between EOS counts and the effect of ICS was prominent for the critical outcome risk of exacerbations, where the certainty in the effect estimate was moderate. However, the clinical relevance of the effect may be limited – at least in patients with a low EOS count. In addition, when offering continued treatment with ICS to patients with COPD, an increased risk of pneumonia is observed, regardless of EOS count. However, it is important to note that the absolute number of patients with pneumonia in the included RCTs was low (32 patients without ICS and 44 patients with ICS, out of 1000 patients). The certainty of the effect estimates for lung function, dyspnea and quality of life were low to very low, and therefore, decisions regarding whether patients with COPD should continue with ICS based on EOS count remain inconclusive for these outcomes. We were unable to locate any RCTs reporting serious adverse events or dropout rate according to EOS count. Yet, based on unstratified populations, it seems that continuing treatment with ICS did not increase risk of serious adverse events, and fewer patients dropped out during the course of the trials if they continued treatment with ICS. The certainty in the effect estimate was moderate for serious adverse events, and low for dropout rate.
We rigorously followed GRADE by assessing risk of bias, inconsistency in the results, indirectness, imprecision, and/or risk of publication bias, by highlighting apprehensions on the certainty of the body of evidence. Specifically, the risk of bias assessment showed that, across the RCTs, there was inadequate description of allocation concealment, selective reporting outcomes as EOS subgroup analysis was done post-hoc and there was other bias in the form of funding and co-authoring by the pharmaceutical industry. Moreover, there were concerns on unexplained heterogeneity, few RCTs and wide CI, but no signals of publication bias. Thus, it can be recommended to conduct withdrawal RCTs with protocolled subgroup analysis of EOS counts, which is preferably not funded by the pharmaceutical industry.
We included RCTs in which not all patients used ICS prior to inclusion. We investigated whether the effect estimate in the RCTs where all patients used ICS prior to inclusion (INSTEAD [
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
], although it is recognized in the 2021 GOLD report that the effect of ICS in this subgroup of patients is low. The withdrawal of ICS in patients with a low EOS count may be hindered by both patient and provider resistance. In a recent real-life study by Nielsen et al., 96 consecutive patients with COPD had their ICS treatment abruptly withdrawn [
]. Although lasting withdrawal was possible in some patients, more than half resumed ICS irrespective of history of exacerbations and eosinophilic count. Many barriers for withdrawing medication have been shown in previous RCTs [
]. A study from the US on attitudes towards withdrawal of ICS in patients with COPD showed that, even though physicians worried about resistance from the patients, most patients were willing to try ICS withdrawal [
]. The patients trusted and relied on their physician's expertise. In addition, the physicians further worried about withdrawal of ICS – especially if it was prescribed by a specialist respiratory doctor.
For patients with neither high nor low EOS counts (between 150 and 300 cells/μL), the risk-benefit ratio of ICS was not obvious, as it can be influenced by lung function, pneumonias and exacerbations.
Future RCTs of the patients with EOS counts between 150 and 300 cells/μL could help identify further subgroups, to guide the decision on continued ICS or ICS withdrawal for these patients. Although the reduction in risk of exacerbations at group level may not be clinically relevant, this should still be assessed individually and in collaboration with the patient. Further questions for future research can be found in supplementary material S11.
Our results suggest that the effect of ICS increased with EOS count and that patients with high EOS count (above 300 cells/μL) had a clinically relevant effect of continuing ICS, seen by a decrease in exacerbations (rate ratio: 0.57; 95% CI: 0.49–0.66) and increase in lung function (FEV1 mean difference: 0.06 L; 95% CI: 0.04–0.09 L). The effect of ICS in patients with high EOS count was also found by Chalmer et al., who found an increase in exacerbation when withdrawing ICS for patients with EOS count above 300 cells/μL (rate ratio 1.63; 95% CI: 1.24–2.14) [
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
] analyzed the continuous effect in three RCTs and found a significant effect on exacerbations at EOS counts above 100 cells/μL and effect on FEV1 (50 mL) at EOS counts above 270 cells/μL. Those findings might suggest a lower cut-off at EOS count of 100 cells/μL, but so far it has not been sufficiently substantiated. Bafadhel et al. analyzed patients with COPD with a median of 180 cells/μL (interquartile range 110–280 cells/μL) and a subgroup of EOS counts below 100 cells/μL would thereby comprise less than 25% of patients, while the subgroup 100–300 cells/μL would contain more than 50% of patients [
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
] found the EOS counts of patients with COPD to have a median of 180 cells/μL (interquartile range 120–270 cells/μL) and a register study of patients with COPD in the US and UK found that approximately one third of patients had an EOS count of below 150 cells/μL; between 150 and 300 cells/μL and above 300 cells/μL, respectively [
In the light of the findings discussed above, 150 cells/μL and 300 cells/μL might be useful, pragmatic cut-off values to ensure sub-groups of approximately the same size, as no scientific or clinically meaningful cut-off values exist. We do not believe there is a clinically meaningful difference between, e.g., an EOS count of 299 and 301 cells/μL, but we do believe that there is a difference between EOS counts of 140 and 290 cells/μL that might be relevant. Furthermore, EOS count is affected by multiple factors [
Stability of blood eosinophils in patients with chronic obstructive pulmonary disease and in control subjects, and the impact of sex, age, smoking, and baseline counts.
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
]. The relationship between EOS counts, the clinical effect of ICS and possible modulators should be examined further, as well as clinically meaningful cut-off-values.
The alteration in EOS count over time and the number of measurements necessary to provide a reliable EOS level in a given individual, together with the timing in measuring EOS, depending on exacerbations and corticosteroids given orally, are among other factors that remain to be explored.
This new use of EOS counts as a biomarker for continued use of ICS might call for several political changes in each country, e.g., reimbursement of general practitioners for testing EOS counts or clauses regarding subsidies for medication.
4.1 Strength and limitations of the systematic review
This systematic literature review has several strengths. The GRADE, Cochrane Collaboration and PRISMA consortium guidelines were followed and a protocol was registered prior to literature search. Strict and high-quality methodological standards and recognized tools were followed, i.e. search strategy, independent study selection, data extraction and risk of bias assessment, by two independent reviewers.
This review also has several limitations. Despite the fact that several databases were searched, references of included RCTs were hand searched and content experts were consulted, the possibility that some RCTs were missed cannot be excluded. Moreover, due to the language limitations of the members in the review team, an exclusive selection of RCTs published in English and Scandinavian languages was performed, which also may have resulted in not identifying relevant RCTs published in other languages. The results are solely based on data published in peer-reviewed articles, because we did not contact the authors of the included RCTs for further information, and grey literature was not searched. Furthermore, as the patients in the included studies are highly selected and therefore differs somewhat from patients in a real-life setting [
], our findings should be interpreted with caution and may not be fully applicable in an everyday clinical setting.
5. Conclusion
This systematic review, using solid and well-established criteria regarding literature searches and inclusion of relevant research articles, is among the first to evaluate the existing literature regarding EOS count and continuous use of ICS in patients with COPD.
A clear correlation was demonstrated between clinical effect of continued ICS treatment, measured as number of exacerbations, and lung function and increasing EOS count. Furthermore, clinically meaningful cut-off values in EOS count of 150 and 300 cells/μL, respectively, were suggested based on the available evidence. However, the relationship between EOS counts, the clinical effect of ICS and possible modulators needs to be examined further, along with possible benefits of withdrawal of ICS at low EOS counts.
Declaration of competing interest
Funding: This study was supported by the Danish Health Authority [05-0000-29].
AU and JFR are employed by the Danish Health Authority. CJ, CV, DAD, HRC, KB, MNH, NSG, AL, OH and PK have nothing to declare.
Acknowledgements and contributorship
Our thanks go to Isabel Cardoso and Elisabeth Ginnerup for the help with the data extraction. And to Torben Mogensen for the help with data collection and patient perspective.
CV, DAD and OH did the initial conceptualization and applied for funding. DAD administrated the project. AL, CJ, DAD, KB, PK and MNH planned the study. AL, AU, CJ, DAD, JFR, KB, PK and MNH planned the methodology, and conducted the data collection. MNH made the meta-analysis. CV, HRC, NSG, OH and MNH supervised the work. AL and DAD created the original draft of article. AL, AU, CJ, CV, DAD, HRC, JFR, NSG, OH, KB, MNH, and PK reviewed, edited, and agreed upon the article.
Appendix A. Supplementary data
The following are the Supplementary data to this article:
Randomised, double blind, placebo controlled study of fluticasone propionate in patients with moderate to severe chronic obstructive pulmonary disease: the ISOLDE trial.
Stability of blood eosinophils in patients with chronic obstructive pulmonary disease and in control subjects, and the impact of sex, age, smoking, and baseline counts.
Predictors of exacerbation risk and response to budesonide in patients with chronic obstructive pulmonary disease: a post-hoc analysis of three randomised trials.
Blood eosinophil count-guided corticosteroid therapy and as a prognostic biomarker of exacerbations of chronic obstructive pulmonary disease: a systematic review and meta-analysis.
The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration.
Blood eosinophil count, a marker of inhaled corticosteroid effectiveness in preventing COPD exacerbations in post-hoc RCT and observational studies: systematic review and meta-analysis.
Long-Term triple therapy de-escalation to indacaterol/glycopyrronium in patients with chronic obstructive pulmonary disease (SUNSET): a randomized, double-blind, triple-dummy clinical trial.
Blood eosinophil counts, exacerbations, and response to the addition of inhaled fluticasone furoate to vilanterol in patients with chronic obstructive pulmonary disease: a secondary analysis of data from two parallel randomised controlled trials.
Blood eosinophil count and exacerbations in severe chronic obstructive pulmonary disease after withdrawal of inhaled corticosteroids: a post-hoc analysis of the WISDOM trial.
Blood eosinophils and treatment response with triple and dual combination therapy in chronic obstructive pulmonary disease: analysis of the IMPACT trial.
Comparison of effect of indacaterol with salmeterol/fluticasone fixed dose combination on COPD exacerbations based on baseline blood eosinophil counts: post-hoc analysis from the instead study.
00145-7&id=gr1.jpg){kind=link}
00145-7&id=gr2.jpg){kind=link}
00145-7&id=gr3.jpg){kind=link}
00145-7&id=gr4.jpg){kind=link}
00145-7&id=gr5.jpg){kind=link}
00145-7&id=gr6.jpg){kind=link}