Rational use of inhaled corticosteroids for the treatment of COPD

ICS monotherapy vs. placebo

Although ICS use has been studied more extensively in asthma, many RCTs have evaluated the efficacy and safety of ICS in COPD (Table 2)⁷. A Cochrane review of treatment with ICS alone (55 primary studies, including >16,000 patients) found that ICS did not modify the long-term decline of FEV₁ or mortality in patients with COPD⁴³. This review noted a relatively small but statistically significant reduction in the mean exacerbation rate with ICS. However, the rate of pneumonia increased by >50% in the ICS group. Furthermore, in the TORCH trial, a trend toward higher mortality was observed for patients treated with ICS (fluticasone propionate) alone⁴⁴. Consequently, the GOLD 2011 report recommended against using ICS monotherapy in COPD⁴⁵.

LABA/ICS vs. placebo

From 2000 onwards, many landmark trials have evaluated the efficacy of ICS in combination and comparison with long-acting bronchodilators⁷. In 2007, TORCH demonstrated a significant reduction in exacerbations and improvement in health status and lung function with LABA/ICS vs. placebo in patients with COPD—a finding that was replicated for exacerbations and lung function in the SUMMIT trial in 2016^44,46. However, both of these large-scale trials failed to achieve their primary objective: a statistically significant, ICS-related reduction in mortality. Moreover, a post hoc factorial analysis of the TORCH trial showed a survival benefit associated with the LABA, but not the ICS, component (Table 2)⁴⁷. Like many COPD studies, both trials were confounded by ICS withdrawal prior to randomisation and incomplete post-discontinuation follow-up for all their secondary end points, including exacerbations^47,48.

LABA/ICS vs. LAMA/LABA/ICS

The Canadian OPTIMAL trial, an independent, non-industry-sponsored trial, notable for its complete exacerbation follow-up, investigated the efficacy of adding LABA/ICS to LAMA in patients with moderate-to-severe COPD, and demonstrated a reduction in exacerbations that did not reach statistical significance⁴⁹. A post hoc analysis of OPTIMAL demonstrated that the apparent decrease in exacerbations was limited to pre-study ICS users and thus largely attributable to ICS discontinuation on randomisation⁴⁷.

LABA/ICS vs. LAMA/LABA

In 2016, the FLAME study highlighted the benefits of LAMA/LABA, finding a reduced annual rate of moderate or severe exacerbations and lower incidence of pneumonia vs. LABA/ICS⁵⁰. However, the superiority of LAMA/LABA over LABA/ICS demonstrated in FLAME was probably exaggerated by its inclusion criteria and 4-week run-in with LAMA and withdrawal of ICS, which is likely to have preserved more LAMA responders relative to ICS responders (Table 2)⁵¹.

LAMA/LABA/ICS vs. LAMA/LABA

Following the shift towards more LAMA/LABA use instead of LABA/ICS between 2013 and 2018, the last 5 years have seen a resurgence in ICS use after the approval of triple-therapy inhalers based on the results of pivotal studies such as TRIBUTE, IMPACT, KRONOS and ETHOS^5,6,52,53. IMPACT and ETHOS have reported the benefits of using LAMA/LABA/ICS vs. LAMA/LABA in patients with COPD and a high exacerbation risk in terms of reducing exacerbations and mortality, albeit with a higher incidence of pneumonia with triple therapy^5,6. However, various publications have commented extensively on methodological issues with the design and analysis of these studies in terms of the populations studied, the confounding effect of ICS withdrawal prior to randomisation, and the inclusion of patients with a history of asthma (Table 3)^54–58. While all three trials are confounded by ICS withdrawal on randomisation, this effect is further magnified in IMPACT and ETHOS (vs. TRIBUTE) by the selective inclusion of frequent exacerbators with non-severe airflow limitation (FEV₁ > 50%, GOLD 2D), thereby selecting an unusual asthma-like, “ICS-sensitive” study population compared with the general COPD population⁵⁹. Additionally, in both KRONOS and ETHOS, the trial design included a 1–4-week screening period in which withdrawal of all long-acting bronchodilators induced a very large pre-randomisation dropout rate (38% and 46%, respectively), thus favouring the selective inclusion of “ICS-sensitive” subjects^6,53. Thus, for IMPACT and ETHOS, the reported survival benefit in the ICS arms is largely attributable to the transient effect of ICS withdrawal on randomisation in a selected “ICS-sensitive” cohort^56,60.

The confounding effect of ICS withdrawal is a recurrent limitation in the RCTs of triple therapy. In an optimally designed trial, patients on triple therapy at screening would be excluded outright, and other patients would be randomly allocated to specific treatment arms based on their current therapy^54,55. However, in the absence of such trials, observational studies from routine clinical practice can provide valuable evidence by investigating the long-term effect of ICS in larger cohorts of “new users”, avoiding the effect of medication switching/withdrawal on randomisation and better representing the overall population of COPD patients.

Key points

• The major clinical benefit of ICS in COPD is a ~25% reduction in exacerbations, observed in RCTs of patients with frequent or severe exacerbations.

• Though some RCTs suggest that adding ICS to LAMA/LABA or LABA therapy increases survival in patients with COPD, methodological flaws in these trials have led regulatory authorities to dismiss claims of survival benefit associated with ICS.

• The study design and populations involved in the pivotal studies of ICS in COPD were very specific and did not represent many patients with COPD, focusing a priori on a small subset of patients who stand to benefit the most from ICS treatment.

Exacerbation frequency/phenotype

Global treatment guidelines for COPD state that patients with frequent or severe exacerbations (≥2 moderate exacerbations per year, or ≥1 exacerbation requiring hospitalisation) and a blood eosinophil count ≥300 cells/μl are more likely to benefit from the addition of ICS⁴. However, when identifying frequent exacerbators, it is not just the frequency of exacerbations that is important, but also the type of exacerbation. Not all exacerbations are the same and they have different underlying triggers, which unlike eosinophilic inflammation, are not steroid responsive. For example, 50–70% of exacerbations are attributed to respiratory infections, 10% to environmental pollution and up to 30% are of unknown aetiology⁷⁵. Notably, hospital admissions for COPD exacerbations nearly halved during the COVID-19 pandemic, likely due to a reduction in respiratory viral infections triggering exacerbations⁷⁶. Comorbidities such as heart failure or gastroesophageal reflux may also be drivers of exacerbations, as well as mental health conditions such as anxiety and depression^75,77,78.

Before considering use of ICS in frequent exacerbators, it is important to consider the phenotype and cause of the exacerbations in order to tailor treatment, rather than taking a “one-size-fits-all” approach to treatment^73,79,80.

Eosinophil threshold

Analysis of several RCTs has shown an association between eosinophil count and ICS responsiveness in terms of preventing future exacerbations when used in combination with long-acting bronchodilators in patients with COPD^5,52,81–84. Some post hoc analyses of RCTs have suggested that the benefit of ICS begins at a blood eosinophil threshold of 100 cells/μl^72,81. However, the ETHOS results bring into question this threshold as no treatment benefit of triple therapy vs. non-ICS dual therapy on the annual rate of moderate or severe exacerbations was observed in a subgroup analysis of patients with eosinophils <150 cells/μl (at either dose of budesonide)⁶. In the IMPACT study, the authors state that the benefits of triple therapy on the annual rate of moderate/severe exacerbations were seen regardless of eosinophil level, though they acknowledge a greater reduction in exacerbation rate in the ≥150 cells/μl subgroup⁵. In a post hoc analysis of data from IMPACT, triple therapy was associated with a lower exacerbation rate vs. LAMA/LABA in patients with eosinophils ≥100 cells/μl, but only in those with a history of frequent moderate or severe exacerbations. In patients with a history of a single moderate exacerbation, a lower exacerbation rate in the triple therapy arm was only observed at higher eosinophil levels (≥200 cells/μl)⁸⁵.

Since findings on eosinophil thresholds can be affected by selection bias for certain patient populations (such as frequent exacerbators) in RCTs, as described earlier in this article, the threshold derived from observational studies may be more reliable. Several observational studies have shown that the optimal blood eosinophil count threshold for ICS response in terms of the ability to reduce exacerbations is considerably higher (300–450 cells/μl)^86–88 than findings from RCTs (100–150 cells/μl)^5,6,72,81. However, other analyses have not replicated these findings. In a systematic review of 11 RCTs and 5 observational studies, an overall association between blood eosinophil count and the effect of ICS in reducing exacerbation risk was found. However, this association was not observed in four of the five observational studies, suggesting that the predictive effect of eosinophils may not apply to the real-world COPD population⁸⁹.

An observational study comparing initial COPD treatment with LABA/ICS or LAMA suggests that while the exacerbation benefit of initiating LABA/ICS is demonstrated only in patients with high blood eosinophil counts (>300 cells/μl), the increased pneumonia risk with LABA/ICS is observed at all eosinophil concentrations⁸⁷. Based on this, the authors conclude that initial treatment with a LAMA should be preferred in patients with blood eosinophil concentrations ≤300 cells/μl due to its superior risk/benefit profile⁸⁷. A blood eosinophil threshold of ≥300 cells/μl forms the basis of the guidance for initial treatment of COPD in the GOLD 2022 strategy report, aiming to identify exacerbating COPD patients that are likely to derive the greatest benefit from ICS⁴.

Key points

• It is important to assess the number and type of exacerbations before prescribing ICS to patients with COPD (ICS reduce eosinophilic exacerbations but not infection-based exacerbations).

• In patients with a history of frequent or severe exacerbations, global treatment guidelines recommend starting with dual bronchodilator therapy (LAMA/LABA).

• Addition of ICS to LAMA/LABA is recommended for frequent/severe exacerbators with blood eosinophil levels ≥300 cells/µl (or ≥100 cells/µl if exacerbations are not well controlled by LAMA/LABA).

Pneumonia

The higher risk of pneumonia in COPD patients treated with ICS is well documented and is acknowledged in the GOLD 2023 report⁴. In a systematic review of 19 RCTs by Miravitlles et al., exposure to ICS for ≥1 year increased the risk of pneumonia by 41% (risk ratio 1.41, 95% confidence interval 1.23–1.61)⁸. The risk varied according to the type of ICS used, with fluticasone propionate or furoate having the highest risk (10 studies: n = 45,870)⁸. Conversely, exposure to budesonide (six studies: n = 13,479) was not associated with an increased risk of pneumonia, although a high degree of heterogeneity was observed due to one large study that reported an increased risk⁸. These findings are in contrast to a European Medicines Agency Pharmacovigilance Risk Assessment Committee review of the known risk of pneumonia in patients with COPD receiving ICS, which reported no conclusive evidence of differences in the risk of pneumonia between different products⁹⁰. However, in support of the Miravitlles systematic review, a large observational analysis of 39,362 new users of triple therapy (either fluticasone- or budesonide-based) showed a lower incidence of severe pneumonia in patients on budesonide-containing regimens compared with fluticasone⁹¹.

As shown by one US database study of 135,445 patients with COPD, the use of ICS in newly diagnosed patients was associated with a dose-related increase in the risk of pneumonia³⁰. In an observational study of patients recruited from routine clinical practice in the UK (the Salford Lung Study), mortality after admission with pneumonia was higher than after admission with an exacerbation, suggesting that pneumonia may have a greater impact on survival than exacerbations⁹². Real-world data such as these are important in the interpretation of fatal pneumonia risk associated with ICS use in COPD, as data from randomised controlled trials are often confounded by the exclusion of patients at highest risk of pneumonia, e.g. those with low lung function, very low BMI or significant comorbidities³⁰. However, it should also be acknowledged that definitions of pneumonia and acute exacerbations of COPD often overlap within electronic health records. As such, they may not always be as rigorously differentiated from each other as they should be, compared with clinical trials, in which an X-ray diagnosis of pneumonia is often a requirement.

Variations in the risk of pneumonia and other adverse events associated with ICS use in patients with COPD may possibly be explained by their effects on the composition of the lung microbiome^93–98. Several studies have reported changes in airway microbiome composition following ICS treatment, including reductions in α-diversity, increases in sputum bacterial load/modification of sputum microbial composition and increased airway load of potentially pathogenic bacteria, e.g., increased risk of acquiring the respiratory pathogen Pseudomonas aeruginosa^93,94,97,98. However, further studies are needed to clarify the effects of ICS on the lung microbiome. Most recently, ICS withdrawal in the INCOGNITO study was associated with potentially beneficial changes in microbiome composition and altered the exacerbation endotype, with a reduction in bacterial-associated exacerbations^99,100. The increased risk of respiratory infections with ICS treatment may be linked with their immunosuppressive effects, including reductions in T-cell, macrophage and neutrophil function in the lung^101,102.

Other adverse events

In addition to the increased risk of pneumonia, both cohort and nested case–control studies show an association between ICS use and the risk of tuberculosis and mycobacterial disease⁸. A strong association has also been reported between ICS use and local disorders such as oral candidiasis and dysphonia, although the association with diabetes and bone fractures is less clear and appears significant only at high doses of ICS^8,9. Some studies have found a significantly increased risk of cataracts associated with cumulative ICS exposure but results from other studies did not find a significant association⁸. Of note, elderly patients with COPD have an increased risk of osteoporosis and cataracts¹¹ and non-elderly patients with COPD have an increased risk of osteoporosis^101,103,104. ICS also carries an increased risk of developing type 2 diabetes in patients with COPD, particularly at high doses¹⁰⁵.

Cost implications

Triple therapy containing ICS may not be the most cost-effective approach, especially as first-line treatment and in patients with less severe COPD. Compared with dual therapy, triple therapy is associated with significant increases in hospitalisation rates and cost of care¹⁰⁶. Inappropriate prescription of ICS is associated with poorer physical health status as well as higher costs of COPD management¹⁰⁷. Several studies have shown that reducing inappropriate ICS use and increasing use of LAMA/LABA resulted in better outcomes, including a reduction in exacerbations and pneumonia cases, accompanied by lower total COPD costs^108,109. As first-line treatment, LAMA/LABA is also associated with significantly lower pharmacy costs attributable to COPD or pneumonia vs. triple therapy and is more cost-effective than triple therapy in patients in GOLD groups A/B vs. GOLD groups C/D⁶⁷.

Key points

• ICS-containing therapies increase the risk of many unwanted side effects, in particular, pneumonia, in patients with COPD.

• Because of this, it is important to consider the benefits vs. the risks of treatment when prescribing ICS.

• Unnecessary use of ICS has cost implications for primary care providers.

• We have included a practical guide to the appropriate prescription of ICS, to help identify patients for whom the benefits of ICS are likely to outweigh the risks (Fig. 3).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.