Abstract
Recently it has been suggested that there is a causal association between the use of inhaled corticosteroids (ICSs) and the risk of developing pneumonia in patients with chronic obstructive pulmonary disease (COPD). An increased risk of pneumonia associated with ICS use has been seen in trials with different design, different study populations and with evidence of a dose–response relationship. However, as none of these clinical trials were originally designed to assess pneumonia risk, radiographic confirmation of pneumonia was not always obtained. The extent to which pneumonia events have been confounded with acute exacerbations of COPD is unclear. As increased pneumonia events were not associated with increased mortality it remains unclear what the clinical significance of these findings are. Further complicating the association between ICSs and pneumonia is that meta-analyses restricted to budesonide trials have not shown an increased risk of pneumonia, and no association has been seen in patients with asthma. A number of mechanisms by which ICSs could increase the risk of pneumonia have been proposed, principally related to their immunosuppressive effect. Well-designed clinical trials with predefined endpoints and objective pneumonia definitions are needed before the real risk of pneumonia conferred by ICSs can be established. In the meantime, it seems reasonable to reduce ICSs given to COPD patients to the lowest effective doses, reduce the risk in individual patients by ensuring appropriate vaccination and to be vigilant for the possibility of pneumonia in patients with COPD on ICSs as they largely overlap with those of an acute exacerbation.
Introduction
Chronic obstructive pulmonary disease (COPD) is a known risk factor for the development of pneumonia [Almirall et al. 1999; Farr et al. 2000] and for hospitalization due to pneumonia [Lange et al. 1995]. COPD patients who require hospitalization for pneumonia may have greater mortality than non-COPD patients [LaCroix et al. 1989; Restrepo et al. 2006], although this has not been a universal finding [Fine et al. 1996, 1997]. Given the great impact that COPD exacerbations have on patients’ quality of life and survival, their prevention and treatment are key in the management of the disease [Fine et al. 1996, 1997].
Although the role of inhaled corticosteroids (ICSs) in COPD is not as well established as in asthma, current COPD management guidelines [Celli and MacNee, 2004; Global Initiative for Chronic Obstructive Lung Disease (GOLD), 2011] recommend adding ICSs to long acting β2-agonists (LABAs) or long-acting muscarinic antagonists (LAMAs) in patients with severe disease and/or frequent exacerbations. This recommendation is based in recent publications which demonstrate that these combinations are superior to monotherapy in reducing COPD exacerbations [Calverley et al. 2007; Tashkin et al. 2008; Wedzicha et al. 2008]. Whether reducing exacerbations has any effect on mortality remains unproven [Nannini et al. 2007], but it clearly results in an improvement in quality of life and lower healthcare costs.
The most clinically relevant reported adverse effects associated with the use of ICSs include the development of oropharyngeal candidiasis, dysphonia and an increased risk of cataracts. Whether ICSs contribute to increased osteoporosis remains controversial.
The TORCH (Towards a Revolution in COPD Health) study published in 2007 was the first large prospective study to raise concern about the possibility of ICSs increasing the risk of pneumonia [Calverley et al. 2007]. Retrospective analysis of several studies of similar design found the same association [Wedzicha et al. 2008; Kardos et al. 2007; Ferguson et al. 2008; Anzueto et al. 2009]. During the same period, the publication of a large database review in Canada [Ernst et al. 2007] supported a cause–effect relationship between ICSs and the risk of hospitalization for pneumonia in COPD patients over the age of 65.
In contrast to COPD, no association between the risk of pneumonia and the use of ICS could be demonstrated in asthma patients [O’Byrne et al. 2011].
In this review, we summarize the results of recent investigations that may have important implications for clinicians in the understanding of this previously unsuspected adverse effect of ICSs and interpreting the relevance of these studies in managing patients with COPD.
Randomized, controlled trials
Probably due to its size and length, the TORCH study [Calverley et al. 2007] was the first to identify an increased risk of pneumonia in patients who were randomized to any of the fluticasone–propionate (FLP)-containing arms. This multicenter, randomized, double-blind study compared the combination of FLP plus salmeterol (SAL) (500/50 μg bid) versus placebo, versus SAL alone, versus FLP alone, during a follow-up period of 3 years. In addition to the beneficial trend observed with combination therapy over placebo and over either FLP or SAL alone in terms of mortality, number of exacerbations, health status and spirometric values, the chance of having pneumonia was greater in those patients in the corticosteroid-containing arms (19.6% for FLP + SAL versus 18.3% for FLP alone versus 12.3% for placebo; p < 0.001) (Table 1). Regardless of these findings, mortality was not increased in the groups with the highest incidence of pneumonia, and in fact there was a borderline trend to lower mortality in the combination ICS/LABA group that had the highest pneumonia rate (Table 1). TORCH was not however designed or powered to analyze specific causes of mortality. Moreover, no set of predefined radiologic or bacteriologic parameters for the diagnosis of pneumonia where uniformly employed among the subjects.
RCTs that showed an increased risk of pneumonia in patients treated with ICSs for COPD.
RCT, randomized controlled trial; ICS, inhaled corticosteroid; FLP, fluticasone propionate; SAL, salmeterol; NR, not reported.
Later, a post hoc analysis of the TORCH study attempted to provide a deeper insight of the excess of pneumonia observed in the ICS-containing groups [Crim et al. 2009]. Adjustments were made based on the increased dropout rate in the placebo group, reporting the incidence of pneumonia as cases per 1000 treatments per year, after which the increased number of pneumonias in both ICS-containing arms still persisted. Similarly, no increase in mortality caused by pneumonia was detected in those patients treated with combination therapy, with less encouraging results in the FLP alone group. Finally, the authors identified by means of multivariate analysis several risk factors associated with the onset of pneumonia under any of the treatments: age ≥ 55, forced expiratory volume in 1 second (FEV1) <50% predicted, worse Medical Research Council (MRC) dyspnea score, body mass index (BMI) <25 kg m−2 and exacerbations in the last year.
Almost simultaneously, in relatively smaller study in Germany, Kardos and colleagues investigated the impact of the combination of FLP + SAL (500/50 μg bid) compared with SAL monotherapy (50 μg bid) in patients with severe COPD (stages 3 and 4 of the Global Initiative for Chronic Obstructive Lung Disease, i.e. GOLD) and history of frequent exacerbations (two or more in the last year) [Kardos et al. 2007]. This study found a 35% reduction in moderate and severe exacerbations in the combined treatment group (p < 0.0001). An increase in the number of reported pneumonia cases was noted, with more than threefold more events in the combined treatment arm (23 versus 7; p = 0.005). Only half the pneumonia cases in both groups were deemed to be related to the medication, although the criteria for determining this were not clear. Two cases of pneumonia (one in each group) were interpreted as a secondary cause of death.
In another large multicentric trial, INSPIRE (Investigating New Standards for Prophylaxis in Reducing Exacerbations), Wedzicha and colleagues compared the efficacy of the FLP + SAL combination (500/50 μg bid) versus tiotropium (18 μg a day) to prevent exacerbations in severe and very severe COPD patients [Wedzicha et al. 2008]. It recruited 1323 patients with a follow-up period of 2 years. Although the dropout rate in the tiotropium arm was considerable (29% more than in the FLP + SAL arm; p = 0.005), mortality, which was not the primary endpoint of the study, was significantly lower in the FLP + SAL group (3% versus 6%; p = 0.032). There was, however, a two-fold increase in pneumonia events in the group treated with ICSs compared with the group treated with tiotropium (7% versus 4%; p = 0.008).
Taking into account the possibility of pneumonia misdiagnosis in the TORCH study, two recent studies evaluated FLP in lower doses and required that the diagnosis of pneumonia should be confirmed radiologically. Ferguson and colleagues compared the efficacy of FLP + SAL in lower doses (250/50 μg bid) with SAL monotherapy (50 μg bid) in COPD exacerbations in 782 severe COPD patients [Ferguson et al. 2008]. In the follow-up period of 1 year there was a significant reduction in the rate of exacerbations in the combined therapy group. Once again, the incidence of pneumonia was greater with the use of ICSs (7%) compared with controls (4%). With a similar design, the study conducted by Anzueto and colleagues yielded very similar outcomes [Anzueto et al. 2009]. Whether there is a difference in risk between the 250 and 500 μg doses of FLP remains unknown, since there are no trials comparing both regimens head-to-head, and due to the methodological differences among the mentioned studies in terms of pneumonia definition.
Whether the association between ICS is a class effect or limited to FLP was partially addressed by Tashkin and colleagues and Rennard and coworkers with two consecutive studies evaluating the efficacy and safety of budesonide plus formoterol in different doses, combined into the same inhaler or administered independently, versus each of the components alone versus placebo, with an initial follow-up period of 6 months and then extended to 1 year [Tashkin et al. 2008; Rennard et al. 2009]. Neither study found any significant difference in the occurrence of pneumonia between treatment groups. Later, Welte and colleagues assessed the efficacy of budesonide plus formoterol with or without tiotropium in COPD patients and found only three cases of pneumonia in each treatment group (<1%) [Welte, 2008].
Finally, Calverley and colleagues have recently published a post hoc analysis of the INSPIRE study population [Calverley et al. 2011] in which they investigated patient characteristics and symptoms occurring before the 87 pneumonia events that were reported and found that, although the number of events that were not preceded by symptoms of an exacerbation were similar between the treatment groups, pneumonia diagnosis was more likely after a prolonged, unresolved exacerbation in patients receiving ICSs (32 in the FLP + SAL group versus seven in the tiotropium group). This finding was confirmed even when the analysis was restricted to radiologically confirmed events only. The baseline demographics were similar in both treatment groups; however, by means of covariate analyses, the risk of pneumonia was found to be significantly higher in those patients with baseline C -reactive Protein (CPR) >10 mg/l and those with increased baseline breathlessness during daily activity. It remains unclear whether the diagnosis was pneumonia from the outset in each case or if secondary pneumonia developed following the initial exacerbation.
Meta-analyses
Several systematic reviews and meta-analyses have been carried out based on the information provided by randomized, controlled trials (RCTs) evaluating the benefits of ICSs combined with LABAs in COPD patients. Some of these trials included pneumonia as an adverse effect of ICS use as a secondary endpoint, and found a consistent increase in the risk of pneumonia with the use of these compounds [Sobieraj et al. 2008; Rodrigo et al. 2009]. Drummond and colleagues designed a meta-analysis in which the development of pneumonia as a consequence of ICS use was specifically addressed, making it a predefined primary endpoint [Drummond et al. 2008]. Their analysis was limited to 11 published clinical trials. They reported a significant increment in the risk of pneumonia (relative risk [RR] 1.34, 95% confidence interval [CI] 1.03–1.75; p = 0.03) but with a considerable statistical heterogeneity (I2 = 72%). Data was later re-analyzed by Loke and Singh to overcome the heterogeneity bias, finding that the increment in the risk of pneumonia still persisted after the adjustment [Loke and Singh, 2009].
Another robust meta-analysis by Singh and colleagues included 18 trials and 16,996 patients [Singh et al. 2009]. They concluded that the use of ICSs by COPD patients for more than 24 weeks is associated with a significant increment (near 60%) of the risk of pneumonia, and of severe pneumonia (near 70%), without a concomitant significant increment in mortality. According to the authors, these findings must be interpreted bearing in mind the inherent limitations of the available data: (1) the included RCTs did not use an objective definition of pneumonia, nor did they require radiological confirmation; (2) most trials were insufficiently powered to detect significant differences in global mortality or pneumonia-related mortality; (3) there is no patient-level data available to adjust for potential confounding variables; (4) the relatively small amount of trials with budesonide (only two in this meta-analysis) makes it difficult to draw conclusions about any possible intraclass differences between this compounds.
In an attempt to overcome these limitations, Sin and colleagues published a meta-analysis that included patient-level data from more than 7000 subjects enrolled in seven RCTs with budesonide [Sin et al. 2009]. They arrived at the conclusion that the use of budesonide, alone or combined with LABAs, was not associated with the risk of pneumonia, even after adjustment for possible confounding variables.
A year later, Singh and Loke updated their original meta-analysis to include the information contributed by Sin and colleagues’ meta-analysis of the budesonide trials [Singh and Loke, 2010b]. With a total of 24 trials (16 with FLP, seven with budesonide and one with mometasone), they confirmed a significant increase in the risk of pneumonia for any of the ICSs (RR 1.56, 95% CI 1.40–1.74; p < 0.0001). In a subgroup analysis, with special attention to budesonide exposure, the direction of effect was consistent with an increased risk for pneumonia; however, this did not meet the statistical significance threshold. The results of this analysis suggest that the apparent absence of an excess pneumonia rate with budesonide, as compared with FLP, is probably due to the greater number of patient-years of follow up in studies of the latter.
Observational studies
There are four observational studies on the risk of pneumonia in COPD patients treated with ICSs. The first was conducted by Ernst and colleagues in Quebec, Canada, using a large database to identify COPD patients and asses the frequency of ICS utilization in those who required hospital admission because of pneumonia [Ernst et al. 2007]. Designed as a nested case–control study, it included data from 175,906 COPD patients of advanced age. When comparing the number of admissions with pneumonia between users and nonusers of ICSs, they found a dose-dependent increment in the relative risk of pneumonia of nearly 70% with current ICS use (in the last 60 days) (Figure 1). Interestingly, past use of ICSs was also associated with an increase in risk (RR 1.20, 95% CI 1.11–1.29) even after a washout period of up to 12 months, which suggests the presence a small but significant residual effect (Figure 2). However, adjustment for the disease severity was based on medication prescription registries, and this could lead to bias since there’s no guarantee that the patients were actually taking these medications.
Adjusted odds ratios for pneumonia according to daily dose of fluticasone equivalent. A dose–response relationship can be observed with higher risk for pneumonia with increasing inhaled corticosteroid doses. (Adapted from Ernst et al. [2007].)
Adjusted odds ratios for pneumonia according to history of inhaled corticosteroid (ICS) use in the last year. A significant residual or ‘wash-out’ effect can be observed, with an increase in risk of 20% even after a period of 9–12 months without ICS use. (Adapted from Ernst et al. [2007].)
Another case–control study conducted by Joo and colleagues with 145,586 recently diagnosed COPD patients from the US Veterans Affairs database showed an increment in the incidence of pneumonia with current (in the last 90 days) use of ICSs (odds ratio [OR] 1.38, 95% CI 1.31–1.45) [Joo et al. 2010]. Although a dose–response relationship that was seen in the study by Ernst and colleagues could not be verified (Figure 3), the presence of a residual effect with a persistently increased risk of pneumonia even after up to 12 months of washout (OR 1.31, 95% CI 1.14–1.50) could be observed (Figure 4). Since the association was assessed in patients recently diagnosed with COPD, it provided the chance to evaluate outcomes in patients with a limited long-term exposure to ICSs and with lower doses (equivalent to FLP <500 μg/day), compared with Ernst and colleagues’ study population. In terms of mortality, neither study showed significant differences.
Adjusted odds ratios for pneumonia according to daily dose of beclometasone equivalent. No dose–response relationship can be observed in comparison with Ernst and colleagues’ study (see Figure 1). (Adapted from Joo et al. [2010].)
Adjusted odds ratios for pneumonia according to history of inhaled corticosteroid (ICS) use in the last year. A significant residual or ‘wash-out’ effect can be observed, with an increase in risk of 31% even after a period of 9–12 months without ICS use. (Adapted from Joo et al. [2010].)
Two unpublished observational studies from the clinical registry of GlaxoSmith Kline have assessed the risk of pneumonia with the use of ICSs [Singh and Loke, 2010a]. One of them is a retrospective study of 10,918 patients with recently diagnosed COPD that did not find an increment in the risk of pneumonia during the first 3 years of treatment, but reported a twofold increase in the risk at 4 years from the initiation of treatment with fluticasone (OR 2.08, 95% CI 1.01–7.77). The second study included 15,614 adult patients with recently diagnosed COPD. Both studies extracted data from the GeneralPractice Research Database in the United Kingdom, and none of them could demonstrate a significant association between pneumonia and the use of ICSs in the 24-month follow-up period. However, the greatest risk was observed in the 18–24 months subcategory (OR 1.82, 95% CI 0.95–3.27; p = 0.07).
A meta-analysis of the four studies by Singh and Loke showed a significant increase in the risk of pneumonia in ICS users (RR 1.44, 95% CI 1.20–1.75; p < 0.0001) [Singh and Loke, 2010a].
Recently, O’Byrne and colleagues published the results of a retrospective, industry-sponsored analysis of a dataset of asthma patients that had been included in double-blind, randomized trials of at least 3 months duration [O’Byrne et al. 2011]. With 86 trials included and over 50,000 patients, they compared the effects of budesonide versus controls and budesonide versus FLP. In their analysis, the authors found no differences in the risk of pneumonia, the dose–response relationship for the development of pneumonia with budesonide, or any intraclass difference between budesonide and FLP.
Discussion
One observational study, many prospective RCTs, and a few meta-analyses have concluded that ICSs, alone or in combination with LABAs, increase the risk of having pneumonia in COPD patients. This was reported as either a severe or very severe adverse event. Most of these studies had an important limitation in that pneumonia was not an anticipated adverse event, so no radiological confirmation was required, raising the possibility of misdiagnosis owing to the similarities in clinical presentation with COPD acute exacerbations. It also remains unclear whether or not there is a difference in mortality attributable to these pneumonia events, and certainly the overall mortality rate suggests that pneumonia in this setting has a much lower risk of death than typical cohort studies of community-acquired pneumonia.
The exact mechanism by which ICSs could be responsible for an increased risk of pneumonia in COPD patients is not fully understood. ICSs reach the lung in high concentrations [Johnson, 1996] and could contribute to the development of pneumonia due to their immunosuppressive effect. Particularly, the inhibition of nuclear factor kappa B (NF-κB) by ICSs in COPD, one of the proposed mechanisms for their therapeutic effect, could also lead to the suppression of normal host responses to bacterial infection [Singanayagam et al. 2010]. Inhaled FLP in doses of 1000 μg daily raises serum cortisol levels in the same way as with 10 mg a daily of oral prednisone, a dose that has proved to double the risk of pneumonia in rheumatoid arthritis patients [Wolfe et al. 2006].
The paradox of increased episodes of suspected pneumonia despite the reduction in the total number of exacerbations is another area of uncertainty. It could be argued that instead of an actual increase in the occurrence of pulmonary infiltrates, we might be in the presence of a distinct clinical presentation of an acute exacerbation, altered by the local effects of ICSs. The fact that no simultaneous increase in mortality was found supports this possibility.
Another explanation for an increased risk of pneumonia but lower mortality is that ICSs may have a beneficial modulating effect on the local anti-inflammatory response in patients with pneumonia, and in particular on macrophage activation [Gutierrez et al. 2010]. Reduction in the proinflammatory response in pneumonia could lead to less subsequent organ dysfunction [Welte, 2009; Martinez et al. 2011; Annane and Meduri, 2008] and therefore better clinical outcomes. Against this hypothesis, however, a recent observational study on the impact of ICSs on the outcomes of pneumonia in patients with COPD found no significant difference in the levels of markers of systemic inflammation, such as C-reactive protein and white cell count [Singanayagam et al. 2011]. Clinical studies have also failed to find a consistent benefit of systemic corticosteroid administration in CAP [Salluh et al. 2008], which similarly does not support a beneficial effect of ICSs on local inflammatory response.
Sin’s meta-analysis, which was restricted to budesonide trials, failed to show an increased rate of pneumonia [Sin et al. 2009]. These findings differ from those reported in Drummond and coworker’s and Singh and colleagues’ respective meta-analyses [Drummond et al. 2008; Singh et al. 2009], where no distinction in ICS class were made at the time of trial inclusion, the results being more influenced by trials using FLP. The authors concluded that the differences could be related to the fact that none of the previous meta-analyses included patient-level data, thus being unable to adjust for confounding variables. Another reason could be that almost every trial included in those previous meta-analyses utilized FLP, so it is reasonable to assume the existence of an intraclass difference to explain the conflicting results. On the other hand, consistent with our previous comments about the great number of patient-years of follow up with FLP, it has been pointed out that the follow-up period of the trials included in Sin and colleagues’ meta-analysis was significantly shorter (12 months versus 6–40 months compared with Drummond and colleagues) [Sin et al. 2009].
Where all three meta-analyses agree is in the fact that there was no increase in mortality, whether or not there was an increment in the number of pneumonia events, and independently of the compound being tested. To support this hypothesis, a recent observational study on the impact of ICS use on the outcomes in COPD patients admitted with pneumonia found no evidence of increased severity and mortality related to ICS-associated pneumonia [Singanayagam et al. 2011]. Moreover, two recent retrospective studies did even report a decreased risk in short-term mortality and use of mechanical ventilation after hospitalization for pneumonia in COPD patients with prior ICS use [Malo de Molina et al. 2010; Chen et al. 2011].
Despite some skepticism that there are difference in the risk of pneumonia between FLP and budesonide, there are differences in pharmacodynamics and pharmacokinetic factors that could support intraclass effects [Halpin et al. 2011]. FLP, being more lipophilic, dissolves more slowly in the epithelial surface of the airways and is retained for a longer time [Johnson, 1996]. Other mechanisms potentially involved include differences in potency ratio, in the effects on airway clearance and on phagocytes activity, and in inhalers’ output characteristics and drug delivery.
Since ICSs are the mainstream treatment in asthma, the findings in COPD trials of pneumonia as a consequence of ICS use raised concern that this may also affect asthma patients treated with these drugs. Thus, it would be of special interest to be able to evaluate this association in a different population than COPD patients, in which the use of ICSs is even more widespread and where clinical manifestations of the disease and the adverse effects of the medications used to treat it do not overlap as much. Nevertheless, no association between an increased risk of pneumonia and ICS use in asthmatic patients has been reported so far. As it is already well established, different inflammatory patterns are involved in asthma compared with COPD, and this fact could potentially represent one of the underlying reasons for these findings. Also, given that increasing age and comorbid diseases are major risk factors for developing pneumonia, and since COPD patients are older and have more comorbid diseases than asthmatics, it is not surprising that the impact of ICSs on pneumonia rates would be harder to detect in asthmatics.
Conclusions
There is a considerable amount of evidence that supports the concept that prolonged use of ICSs in COPD patients is associated with a significant increment in the risk of pneumonia, which has led to the inclusion of acute bronchitis and pneumonia as side-effects in the labels of most FLP-containing products. As no study has demonstrated an increased risk of death from ICSs, physicians and patients should be relatively reassured that the risk–benefit analysis of ICS use in COPD is still significantly in favor of treatment. Additional cost–benefit and risk–benefit studies will be able to provide more information on the matter [Cave and Hurst, 2011].
Given the increased risk of pneumonia with ICSs, it remains prudent to advise vaccination against pneumococcus and influenza in patients with COPD. Finally, physicians should be alert for the development of signs and symptoms of pneumonia in COPD patients treated with ICSs, which may initially be indistinguishable from those of an exacerbation. Whether different treatment approaches will impact on the outcome of ICS-related pneumonia remains to be determined.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of interest statement
Professor Waterer has been a paid speaker for both GlaxoSmithKline and AstraZeneca in the past 5 years. None of the other authors have any conflict of interest to declare with respect to this manuscript.