Inhaled Corticosteroids and the Pneumonia Risk in Patients With Chronic Obstructive Pulmonary Disease: A Meta-analysis of Randomized Controlled Trials

Background: Whether all types of inhaled corticosteroids (ICSs) would increase the pneumonia risk in patients with chronic obstructive pulmonary disease (COPD) remains controversial. We aimed to assess the association between ICSs treatment and pneumonia risk in COPD patients, and the impact of medication details and baseline characteristics of patients on the association. Methods: Four databases (PubMed, Embase, Cochrane Library, and Clinical Trials.gov) were searched to identify eligible randomized controlled trials (RCTs) comparing ICSs treatment with non-ICSs treatment on the pneumonia risk in COPD patients. Pooled results were calculated using Peto odds ratios (Peto ORs) with corresponding 95% confidence intervals (CIs). Results: A total of 59 RCTs enrolling 103,477 patients were analyzed. All types of ICSs significantly increased the pneumonia risk (Peto OR, 1.43; 95% CI, 1.34–1.53). Subgroup analysis showed that there was a dose-response relationship between ICSs treatment and pneumonia risk (low-dose: Peto OR, 1.33; 95% CI, 1.22–1.45; medium-dose: Peto OR, 1.50; 95% CI, 1.28–1.76; and high-dose: Peto OR, 1.64; 95% CI, 1.45–1.85). Subgroup analyses based on treatment durations and baseline characteristics (severity, age, and body mass index) of patients were consistant with the above results. Subgroup analysis based on severity of pneumonia showed that fluticasone (Peto OR, 1.75; 95% CI, 1.44–2.14) increased the risk of serious pneumonia, while budesonide and beclomethasone did not. Conclusions: ICSs treatment significantly increased the risk of pneumonia in COPD patients. There was a dose-response relationship between ICSs treatment and pneumonia risk. The pneumonia risk was related with COPD severity.


INTRODUCTION
Chronic obstructive pulmonary disease (COPD) is currently the third leading cause of death in the world, and acute exacerbations contribute substantially to this (GBD 2015Chronic Respiratory Disease Collaborators, 2017Viniol and Vogelmeier, 2018;López-Campos et al., 2019). Treatment and prevention of repeated exacerbations have been identified as a priority by the Global Initiative for Chronic Obstructive Lung Disease (GOLD). Currently, the management of patients with stable COPD mainly relies on inhaled agents such as inhaled corticosteroids (ICSs), long-acting muscarinic antagonist (LAMA), long-acting β-agonist (LABA), etc. Among them, ICSs have been recommended by GOLD as first-line maintenance treatment in patients with repeated exacerbations to relieve the frequency and severity of acute exacerbations of COPD, and improve their quality of life (Yang et al., 2017).
The aim of this meta-analysis was to objectively reappraise the pneumonia risk and serious pneumonia associated with various types of ICSs in COPD patients through all available RCTs. We also aimed to assess the impact of medication details (including dosage level and treatment duration) and demographic characteristics (severity, age, and body mass index) of patients on this association.

Protocol and Guidance
This meta-analysis was carried out according to the Preferred Reporting Items for Systematic review and Meta-Analysis (Moher et al., 2009). Ethics committee approval is not applicable for this meta-analysis. The study was registered with PROSPERO prospectively (#CRD42020213586).

Search Strategy
Two reviewers (Hong Chen and Jian Sun) independently searched the databases of PubMed, Embase, Cochrane Library, and Clinical Trials.gov from inception until February 2021, using the following terms: ("chronic obstructive pulmonary disease" OR "COPD" OR "pulmonary disease, chronic obstructive" or "chronic obstructive airway disease" OR "airflow obstruction, chronic" OR "chronic airflow obstruction" OR "chronic obstructive lung disease" OR "emphysema" OR "Bronchitis") AND ("inhaled corticosteroids" OR "ICS" OR "budesonide" OR "fluticasone" OR "mometasone" OR "beclomethasone" OR "triamcinolone" OR "ciclesonide"). Articles in English were included. Disagreements regarding eligibility were resolved by discussion by two investigators and, if necessary, consultation with a third investigator (Hao Yan).

Eligibility Criteria
We included eligible studies based on the PICOS ( (Shamseer et al., 2015): 1) Participants: patients aged 40 yr or over, with stable, moderate (GOLD stage II) to very severe (GOLD stage IV) COPD. Patients with other respiratory diseases, such as asthma, bronchiectasis were excluded. 2) Interventions: various types and doses of ICSs as the intervention treatment. 3) Comparisons: non-ICSs treatment as a control treatment. 4) Outcome: Pneumonia for this meta-analysis was defined as an adverse event based on the Medical Dictionary for Regulatory Activities (MedDRA, version 10.0) pneumonia-related preferred terms, including "pneumonia," "lobar pneumonia," "bronchopneumonia," "pneumonia pneumococcal," or "pneumonia staphylococcal" (Rennard et al., 2009;Sin et al., 2009). One or more of the above MedDRA terms reported in the adverse events list or the safety profiles by the RCTs would be identified as pneumonia data, and included in our analysis. Serious pneumonia was defined as a pneumonia leading to mechanical ventilation or death, or requiring hospital admission (Aaron et al., 2007;Pascoe et al., 2015). 5) Study design: only RCTs were included. Non-RCTs, such as retrospective studies, reviews, case reports and case-control studie, were excluded.

Data Extraction and Quality Assessment
Two reviewers (Qiang Huang and Yongqi Liu) independently identified references and extracted data from eligible RCTs. Any disagreements would be resolved by discussion to reach a consensus, and consulted a third reviewer if necessary. The risk of bias of the included RCTs was assessed by two independent reviewers (Hong Chen and Mengxin Yuan) using the Cochrane risk of bias tool (Higgins et al., 2011). Any disagreements would be resolved by discussion and consultation (Hao Yan).

Subgroup Analyses
Subgroup analyses were conducted based on:

Statistical Analysis
The Review Manager 5.3 software was used to calculate the pooled results. Considering Peto odds ratio (Peto OR) could provide the best confidence interval (CI) when events are rare (Bradburn et al., 2007), the pooled results for the comparison of ICSs treatment vs non-ICSs treatment were calculated using Peto ORs. Sensitivity analysis was performed after excluding those studies with high risk of bias. Subgroup analyes based on the baseline demographic characteristics (severity, age, and body mass index) of the patients were conducted using the individual patient level data, which was extracted from the baseline data of the included RCTs (mean or median for lung function, age and BMI). This method of analyzing the individual patient level data was used by Sobieraj et al. (Sobieraj et al., 2018) previously. A two tailed p-value <0.05 was considered to be statistically significant. Statistical heterogeneity was further measured using the I 2 test, and I 2 ≥50% indicated a substantial heterogeneity (Higgins et al., 2003).

Assessment of Risk of Bias
All included studies were assessed using the Cochrane Collaboration risk of bias assessment tool. The results are presented in Figures 2, 3. Thirty-five RCTs were assessed as being at low risk of bias for all aspects. Four had a high risk of bias for performance bias (blinding of participants and personnel) and detection bias (Blinding of outcome assessment). Twenty-two had an unclear risk for random sequence generation, selective reporting, allocation concealment, or other bias (Figures 2, 3).  Figure 4).

Different Treatment Durations of ICSs and Pneumonia Risk
Of the included trials, 31 RCTs (26,408 patients), and 28 RCTs (76,826 patients) assessed short-term ICSs treatment and longterm ICSs treatment and pneumonia risk. Subgroup analysis showed that both short-term ICSs treatment (Peto OR, 1.30; 95% CI, 1.04-1.63) and long-term ICSs treatment (Peto OR, 1.44; 95% CI, 1.34-1.55) significantly increased the pneumonia risk. Test for subgroup differences (I 2 0%) indicated that there was no significant difference in the pneumonia risk associated with different treatment durations of ICSs (Table 2 and Figure 6).   Figure 7). Test for subgroup differences (I 2 90.1%) indicated that there was a significant difference in the pneumonia risk in patients with different severity.  Figure 9).

Various Types of ICSs and Serious Pneumonia Risk
Of the included trials, 15 RCTs (29,008 patients) offered data on serious pneumonia associated with ICSs treatment. Compared with non-ICSs treatment, ICSs treatment significantly increased the serious pneumonia risk (Peto OR, 1.55; 95% CI, 1.31-1.84).

Sensitivity Analysis
After excluding four RCTs (4,414 patients) with high risk of bias, the pooled results were similar in magnitude and direction to those (pooled results of association between various types of ICSs and pneumonia risk) obtained from all included RCTs (Table 2).

DISCUSSION
In this meta-analysis of 59 RCTs (including 103,477 patients), all types of ICSs, not only fluticasone, increased the pneumonia risk in COPD patients in a dose-dependent manner, and the risk was particularly evident in more severe COPD patients. Moreover, fluticasone increased the risk of serious pneumonia, while budesonide and beclomethasone did not. To our knowledge, this study was the first meta-analysis which revealed the pneumonia risk associated ICSs treatment was related with COPD severity. In addition, there was a doseresponse relationship between the pneumonia risk and ICSs treatment.
At present, ICSs are widely used in the maintenance treatment of COPD patients. Since numerous COPD patients use ICSs every day, both its efficacy and safety should be considered. Although some studies have reported that fluticasone increases the pneumonia risk in COPD patients, whether other types of ICSs would increase the pneumonia risk in COPD patients remains controversial (Sin et al., 2009;Singh et al., 2009;Kew and Seniukovich, 2014;Festic et al., 2016;Yang et al., 2019;Zhang et al., 2020). In addition, it is still unclear whether different medication details and baseline characteristics (severity, age, and body mass index) of patients would affect the incidence of pneumonia after ICSs treatment.
Our results first revealed that all types of ICSs significantly increased the pneumonia risk in COPD patients regardless of treatment duration. The dose-response relationship further confirmed the causality of ICSs treatment and increased pneumonia risk in COPD patients. Moreover, our results revealed that the pneumonia risk was related with COPD severity. However, age and BMI may not be the determinants of ICSs associated pneumonia. In addition, we found that COPD patients receiving different types of ICSs may have different risk of serious pneumonia. Only fluticasone increased the risk of serious pneumonia, while other types of ICSs did not. We speculated that this may be due to the different pharmacodynamics and pharmacokinetic characteristics of different types of ICSs. Previous studies reported that fluticasone could exhibit a longer retention in the airway mucosa and thus have a more prolonged suppression of local immunity of patients (Brattsand and Miller-Larsson, 2003;Dalby et al., 2009).

Compared With Other Studies
Several previous meta-analyses (Sin et al., 2009;Singh et al., 2009;Kew and Seniukovich, 2014;Festic et al., 2016;Yang et al., 2019;FIGURE 9 | Inhaled corticosteroids associated pneumonia in COPD patients with different body mass index. Frontiers in Pharmacology | www.frontiersin.org June 2021 | Volume 12 | Article 691621 15 Zhang et al., 2020) also assessed the pneumonia risk associated with ICSs treatment. However, there were major differences between our meta-analysis and the previous ones in terms of selected studies, statistical analyses, and outcomes. First, we included data of some recent large-scale RCTs (Bhatt et al., 2017;Papi et al., 2017;Siler et al., 2017;Vestbo et al., 2017;Betsuyaku et al., 2018;Chapman et al., 2018;Ferguson et al., 2018a;Ferguson et al., 2018b;Frith et al., 2018;Lipson et al., 2018;Papi et al., 2018;Ichinose et al., 2019;Kerwin et al., 2019;Rabe et al., 2020) which were published after some of the previous meta-analyses. In addition, varied search strategy may be an important reason for the difference in the number of RCTs included in different meta-analyses. We systematically searched four large databases for relevant RCTs, including PubMed, Embase, Cochrane Library, and Clinical Trials.gov. In particular, we systematically searched the online supplementary documents of relevant RCTs. Indeed, pneumonia risk was not the primary outcome in most RCTs, some researchers provided data on pneumonia risk in the online supplementary documents rather than in the text. Second, compared with the previous meta-analyses, we conducted more subgroup analyses based on the baseline demographic characteristics of the patients (severity, age and BMI) to clarify possible varied pneumonia risk in different patients receiving ICSs treatment. Third, our results indicated that all types of ICSs, not only fluticasone, increase the pneumonia risk in COPD patients in a dose-dependent manner, and the risk is particularly evident in more severe patients.
In 2009, Singh et al. (Singh et al., 2009) performed a metaanalysis (18 RCTs, 16,996 patients) and concluded that ICSs (fluticasone and budesonide) treatment significantly increased the pneumonia risk in COPD patients. However, their study failed to provide some important information on ICSs associated pneumonia due to a lack of subgroup analyses based on medication details of ICSs (including dose, type and treatment duration), and subgroup analyses based on the baseline conducted a meta-analysis of budesonide and pneumonia risk (seven RCTs, 7,042 patients) and found that budesonide treatment for 12 mo did not increase the pneumonia risk in COPD patients. In 2014, a meta-analysis performed by Kew et al. (Kew and Seniukovich, 2014) (43 RCTs, 31,397 patients) suggested that both fluticasone and budesonide increased the serious pneumonia risk in COPD patients. However, that study did not further examine the association between other types of ICSs (momethasone and beclomethasone) and the pneumonia risk, nor conduct subgroup analyses based on the baseline demographic characteristics of patients. In 2016, another meta-analysis (29 RCTs, 33,472 patients) performed by Festic et al. (Festic et al., 2016) also revealed that ICSs increased the pneumonia risk in COPD patients. However, that study also limited by a smaller sample size and absent subgroup analyses. In addition, Yang et al. (Yang et al., 2019) conducted a metaanalysis (25 RCTs, 49,982 patients) and found ICSs significantly increased the pneumonia risk and serious pneumonia risk in COPD patients. However, their study also did not analyse the impact of baseline demographic characteristics of patients on the pneumonia risk. Moreover, in 2020, a meta-analysis (18 RCTs, 49,828 patients) performed by Zhang et al. (Zhang et al., 2020) also investigated the association between different types of ICSs and the pneumonia risk, and suggested that fluticasone increased the pneumonia risk while budesonide or beclomethasone did not. However, their results might be limited by the smaller sample size, since much fewer RCTs (especially RCTs on budesonide and beclomethasone) were included in their meta-analysis. In contrast, we searched more databases, used more search terms, and put less restrictions on literature search, which made more relevant RCTs were identified.

Limitations and Strengths
The major strength of our study was that we conducted a comprehensive literature search including all currently available RCTs, thus ensured the generalizability of the conclusions. Moreover, the multiple subgroup analyses based on the medication details (dose and treatment duration) and baseline of patients (severity, age and BMI of patients) enhanced the reliability of the conclusions, and also provided implications for the clinical practice. As far as we know, our study is the first meta-analysis which systematically assesses the association between various types of ICSs and the pneumonia risk based on baseline characteristics of patients. This meta-analysis had several limitations. First, none of the included RCTs were specifically designed to monitor pneumonia event, therefore, there may be underreporting of pneumonia incidence. However, the underestimate of the pneumonia risk could not substantially impact the pooled results of this metaanalysis, since underreporting of pneumonia incidence might occur equally in ICSs treatment groups and non-ICSs treatment groups. Moreover, in the sensitivity analysis, after removing four non double-blind RCTs, the results were consistent with the previous pooled results. Second, the pooled results of momethasone (four RCTs, 5,413 patients) and beclomethasone (four RCTs, 5,884 patients) may weakened by the relatively small sample size. Third, some studies were excluded because of incomplete data or non-English literature, which may lead to inevitable selection bias.

CONCLUSIONS
ICSs treatment significantly increased the risk of pneumonia in COPD patients. There was a dose-response relationship between ICSs treatment and pneumonia risk. The pneumonia risk was related with COPD severity.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
HC and HY conceived and designed this study. HC and JS searched and selected studies. QH and YL extracted essential information. HC and MY assessed the risk of bias. HC and CM conducted the statistical analysis. HC wrote the original draft. All authors approved the final version to be published.

FUNDING
This study was supported by the Chengdu Science and Technology Project (No. 2015-HM0100621-SF).

ACKNOWLEDGMENTS
We would like to express our appreciation to all authors listed in these all primary studies which were included in the current meta-analysis.