Inhaled corticosteroids and pneumonia mortality in COPD patients

To the Editor:

In a recent article, Suissa [1] proposes several alternative hypotheses to explain the paradoxical effect of inhaled corticosteroids (ICS) in chronic obstructive pulmonary disease (COPD) patients, whereby these drugs can increase the risk of pneumonia and simultaneously reduce pneumonia mortality. In our opinion, the conclusion of the article, that pneumonia mortality is increased in COPD patients treated with ICS, both in observational and in randomised clinical trials (RCTs), is not supported by data and cannot be considered confirmed. Basically, the author maintains that this paradox can be explained by the small number of events in randomised clinical trials (RCTs) and selection bias in observational studies [1].

First, pneumonia risk is increased in patients with COPD, and chronic treatment with ICS further increases this risk [2, 3]. However, this higher incidence of pneumonia is not uniform in different studies, and in recent RCTs conducted in a large number of patients the incidence of pneumonia was not increased in patients randomised to ICS treatment [4–6]. This suggests that other factors, such as the specific drug, dose and population studied, may affect this risk.

Suissa [1] suggests that ICS-treated patients may develop pneumonia for the same reasons as the rest of COPD patients, and additionally due to factors directly attributable to ICS. Two different concepts used by the author should be clarified: 1) pneumonia mortality (pneumonia deaths/total number of patients), and 2) case fatality (pneumonia deaths/number of patients with pneumonia). Following the same example used in the article, if the risk of pneumonia in ICS users is 2, and the prognosis of severe cases of pneumonia (case fatality) is similar for ICS and non-ICS users, global mortality for pneumonia in ICS users should be 2. Conversely, if pneumonia mortality is similar, the prognosis of severe cases of pneumonia should be 0.5. In our opinion, this option is more realistic since in the meta-analysis conducted by Festic et al. [3] case fatality in observational studies was lower in ICS users (RR 0.7, 95% CI 0.59–0.88).

Secondly, Suissa [1] proposes that the lower mortality for pneumonia observed in these studies can be explained by selection bias. The argument is that this lower mortality in ICS patients may be related to fewer comorbidities or having less severe COPD, among other factors. However, observational studies statistically adjust for many of these possible confounders, and we cannot find any plausible reason to explain why patients not treated with ICS should have more comorbidities or more severe COPD than ICS-treated patients [7].

To avoid this bias Suissa [1] analysed a cohort study defined by ICS exposure at inclusion. In this study, performed between 1998 and 2003, pneumonia mortality was increased in ICS patients. However, global mortality was lower with ICS use, which can lead to a bias (only living patients can die of pneumonia). In subsequent publications, S. Suissa and co-workers confirm that ICS patients had an increased risk of severe pneumonias, but regrettably, the authors did not explore the mortality data in these studies [8, 9].

Finally, the meta-analysis of RCTs referenced in the article showed an increase in pneumonia mortality among patients treated with ICS, though this relationship did not reach statistical significance (RR 1.50, 95% CI 0.85–2.67), and the authors concluded that the risk of pneumonia mortality was not increased in ICS users [3]. Suissa [1] suggests that the reason is the limited number of events. This may be a factor, but the total number of patients included in this meta-analysis is considerable (12 958 patients). Of note, since the publication of this meta-analysis, at least 10 RCTs containing 40 424 COPD patients randomised to ICS treatment have been published (table 1). Of these, in seven (24 811 patients) non-fatal pneumonia cases were reported. We performed a joint analysis of the RCTs included in the meta-analysis of Festic et al. [3] and those published subsequently with a total of 53 940 patients (31 396 ICS and 22 544 non-ICS). In this analysis, mortality from pneumonia was 58/31 396 for ICS users and 33/22 544 for non-ICS. (RR 0.97, 95% CI 0.58–1.60; p=0.89, I²=7.8%). We believe that these data reinforce the notion that pneumonia-related deaths are very infrequent in RCTs and, given the number of patients included, the clinical relevance of ICS in pneumonia mortality is doubtful.

Certainly, pneumonia is an undesirable event, but the increased risk of pneumonia in COPD patients should be counter-balanced by a decrease in exacerbations, especially the most severe that require hospitalisation, and that lead to an increase in mortality, as Suissa et al. [23] have previously demonstrated.

Therefore, we believe that the conclusion stated by Suissa [1] that: “the evidence is now clear that ICS use is associated with a higher incidence of pneumonia and of pneumonia-related mortality”, is debatable, given available data. The use of ICS should be in accordance with current recommendations, and clinicians must balance the pneumonia risk with the reduction in both severe exacerbations and the mortality.

• Received April 11, 2019.
• Accepted May 30, 2019.

• Copyright ©ERS 2019