FormalPara Key Summary Points

Why carry out the study?

Multiple factors contribute to the elevated pneumonia risk in patients with COPD.

ICS-containing triple therapy improves outcomes in patients with COPD but raises concerns of pneumonia risk.

This post hoc analysis analysed the impact of factors such as patient’s age, BMI, FEV1 and blood eosinophil counts on the pneumonia risk in patients treated with triple therapy in the phase 3 IMPACT trial.

What was learned from the study?

Older age, lower BMI and lower FEV1 were associated with increased pneumonia risk, however, blood eosinophil count did not have an impact on the pneumonia risk in patients receiving triple therapy.

There is a need for a balanced clinical approach of risk versus benefit in the use of ICS-containing treatments in COPD.

Introduction

Risk of pneumonia remains a concern despite the well-defined benefits of treating appropriate patients with chronic obstructive pulmonary disease (COPD) using inhaled corticosteroid (ICS)-containing therapies. Several studies have documented an incidence of pneumonia between 2% and 8% in patients with COPD receiving ICS-containing triple therapy [1,2,3]. Previous studies have shown that multiple patient factors can potentially increase the risk of developing pneumonia. These include advanced age, lower body mass index (BMI), worse lung function, presence of asthma along with COPD, previous history of pneumonia and/or exacerbations and smoking status [3, 4]. Williams et al. in 2017 identified various predictors of pneumonia in patients with COPD over 5 years of follow-up, including more severe disease, gender and ICS use [4]. A post hoc analysis of the TORCH trial identified advanced age, low BMI and low forced expiratory volume in 1 s (FEV1), along with previous exacerbation history, to be contributing factors for the elevated risk of developing pneumonia in patients with COPD receiving ICS-containing therapy [5]. A pooled individual patient data analysis of five clinical trials reported similar observations, in addition to ICS-containing therapy, low BMI, previous history of exacerbations, worsening lung function and more severe lung disease as pneumonia risk factors [6]. However, these trials varied in size and duration, and none included ICS as a component of triple therapy. This analysis was thus performed using data from a single large study, the IMPACT trial, of more than 10,000 patients known to be at significant risk of exacerbations.

Methods

A post hoc analysis was conducted utilizing the data from the IMPACT study to assess the association between the annual rate of pneumonia and demographic factors, analysed as continuous variables, including age, BMI, FEV1 and blood eosinophil count (BEC).

Study Design

The IMPACT study (GSK study CTT116855; NCT02164513) was a phase 3, 52-week, randomized, double-blind, parallel-group, multicentre study comparing single-inhaler triple therapy with fluticasone furoate (FF)/umeclidinium (UMEC)/vilanterol (VI) (FF/UMEC/VI) with once-daily dual therapy with FF/VI or UMEC/VI dual therapy. Patients were randomized (2:2:1) to receive FF/UMEC/VI 100/62.5/25 mcg, FF/VI 100/25 mcg or UMEC/VI 62.5/25 mcg, all administered once daily via the ELLIPTA dry-powder inhaler. The study design and primary results have been previously published [3].

Study Population

Inclusion and exclusion criteria have been described previously [3]. Briefly, eligible patients were ≥ 40 years of age with symptomatic COPD [COPD Assessment Test (CAT) score ≥ 10 at screening], and either a FEV1 < 50% of predicted normal values and ≥ 1 moderate or severe exacerbation in the previous year, or FEV1 50% to < 80% of predicted normal values and ≥ 2 moderate or ≥ 1 severe exacerbation in the previous year. Patients with significant bronchiectasis were excluded. The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice Guidelines and received approval from local institutional review boards or independent ethics committees. All patients in that study provided written informed consent.

Endpoints

This post hoc analysis evaluated the annual rate of on-treatment pneumonias in the intent-to-treat (ITT) population for each of the following baseline characteristics that could be analysed as a continuous variable: age, BMI, FEV1% predicted and BEC.

Statistical Analyses

Fractional polynomials (FP) were used to model the relationship between the baseline characteristic of interest as a continuous variable and the treatment outcomes. This method preserves the continuous nature of the covariates in a regression model and allows nonlinear associations to be tested. The selected best fitting model was plotted as the baseline characteristic versus the annual rate of on-treatment pneumonias in each treatment group. The covariates included in each of the models mirror those included in the analysis of the primary endpoint (moderate plus severe exacerbations) in the primary study and were defined a priori. We used modelling of continuous variables of age, BMI, FEV1 and BEC by treatment group in IMPACT (FF/UMEC/VI, FF/VI, UMEC/VI) to evaluate the annual rate of pneumonia. Models were fitted by using the continuous variables as FPs in negative binomial models with covariates of treatment group, geographical region, age, BMI, smoking status (screening), sex, pneumonia history, BEC, FP1, FP2, FP1*treatment and FP2*treatment. FP1 and FP2 represent continuous transformations of the variables of interest (i.e. age, BMI, FEV1, BEC) in each respective model.

Results

Of the 10,355 patients randomly assigned in the ITT population, pneumonia was reported in 184 (4%), 152 (4%) and 54 (3%) patients in the FF/UMEC/VI, FF/VI and UMEC/VI arms, respectively. The rate [number of events] of pneumonia in the FF/UMEC/VI, FF/VI and UMEC/VI arms was 53.3 [198], 47.7 [165] and 32.4 [55], respectively.

The incidence of pneumonia was evaluated using the long-acting muscarinic antagonist (LAMA)/long-acting β2-agonist (LABA) arm as a baseline control, and the results show a 1.6-fold higher rate (Table 1). Data for pneumonia serious adverse events (SAEs) showed a 1.7-fold difference versus LAMA/LABA (Table 1).

Table 1 Pneumonia incidence in the IMPACT study
Full size table

Amongst these factors, those that could be analysed as a continuous variable were selected in addition to BEC to better understand their relationship to increased pneumonia risk.

Association with Age

The annual rate of pneumonia began to rise above the age of 50 years. The rate was higher in the ICS-containing treatment arms compared with the non-ICS containing arm, and this difference increased with age (Fig. 1).

Fig. 1
figure 1

Annual rates of pneumonia risk by patient age among different treatments (FF/UMEC/VI versus UMEC/VI versus FF/VI). FF fluticasone furoate; UMEC umeclidinium; VI vilanterol

Full size image

Association with BMI

When fitted against BMI of the patients, the annual rate of pneumonia was found to be higher in patients with a lower BMI and this effect was most pronounced with the FF/VI cohort. The pneumonia risk for the three treatment arms (FF/UMEC/VI, FF/VI and UMEC/VI) increased in patients with BMI < 22.5 kg/m2 and the risk appeared to level out for all treatment cohorts in patients with BMI > 22.5 (Fig. 2).

Fig. 2
figure 2

Annual rates of pneumonia risk by BMI among different treatments (FF/UMEC/VI versus UMEC/VI versus FF/VI). BMI body mass index; FF fluticasone furoate; UMEC umeclidinium; VI vilanterol

Full size image

Association with Lung Function

When the annual pneumonia rate was fitted against lung function, as noted by a lower FEV1, it was observed that worse lung function at the initiation of ICS containing triple therapy is associated with higher risk of developing pneumonia in patients with COPD (Fig. 3).

Fig. 3
figure 3

Annual rates of pneumonia risk by lung function among different treatments (FF/UMEC/VI versus UMEC/VI versus FF/VI). FEV1 forced expiratory volume in 1 second; FF fluticasone furoate; UMEC umeclidinium; VI vilanterol

Full size image

Association with Blood Eosinophils

The relationship between BEC and pneumonia was flat up to > 300/µL cells with both ICS-containing treatments and then increased. In contrast, the rate of pneumonia in the non-ICS containing arm appeared to increase at a lower level of BEC (~ 200/µL) (Fig. 4).

Fig. 4
figure 4

Annual rates of pneumonia risk by BEC among different treatments (FF/UMEC/VI versus UMEC/VI versus FF/VI). BEC Blood eosinophil count; FF fluticasone furoate; UMEC umeclidinium; VI vilanterol

Full size image

Discussion

This post hoc analysis of the annual rate of pneumonia data from the 52-week IMPACT trial confirms the effect of different patient factors on pneumonia risk in COPD and extends them with data on blood eosinophils. This risk was seen irrespective of the ICS-containing or non-ICS containing treatment regimen but was higher in those patients receiving an ICS-containing regimen.

Across all variables assessed in this study, annual rate of pneumonia in patients on ICS-containing triple therapy was lower in case of lower age group patients, in patients with higher BMI and in those with less decline in FEV1. The pneumonia risk was seen to be the least with the cohort receiving UMEC/VI, though similar effects of age, BMI and lung function were still applicable. It is noteworthy that the absolute annual rate of pneumonia was small in all the three treatment cohorts, including the ICS-containing triple therapy cohort.

Overall, the pneumonia rate in the triple therapy arm was 7.6%, compared with 4.7% in the non-ICS arm, and the upper 95% CI for the rate of pneumonia in patients on ICS-containing triple therapy crossed 10% in patients aged > 67 years, BMI < 20, FEV1 < 45% predicted and BEC > 300/µL. It has been established through evidence from multiple studies that there are many factors beyond ICS use that increase the risk of pneumonia in patients receiving treatment for COPD. The factors that may result in higher odds of developing pneumonia include more severe disease (GOLD stage IV versus GOLD stage I), followed by previous history of pneumonia, BMI and age of the patients, amongst others [4, 5, 7,8,9,10]. The most important new observation from this analysis is the absence of any relationship between pneumonia and BEC < 300/µL with no differential effect of ICS-containing versus non-ICS containing treatment. However, the rate of pneumonia with non-ICS containing treatment appeared to increase at a lower level of BEC (~ 200/μL). A previous meta-analysis of ten studies suggested that, overall, there may be an increased risk of pneumonia in patients with BEC > 2%. However, in patients treated with ICS, pneumonia occurred in 4.5% of patients with BEC < 2% and 3.9% of patients with BEC ≥ 2%. Those authors concluded that the magnitude of the increased risk was small and should be further explored in large prospective studies [11].

IMPACT was a large prospective study, and the great majority of the patients’ BEC lay in the range of < 300/µL cells, so confidence can be placed in this observation. One study reported that patients with chronic bronchial infection and BEC < 100 cells/µL were at increased risk of pneumonia but constituted 21% of the study population (42 patients) [12]. Whilst our findings do not refute those observations, the absence of any trend for increased pneumonia risk in this large population of patients at high risk of exacerbation suggests any increased risk is confined to that specific group of patients.

We do not have an adequate explanation for the increased rate of pneumonia in patients with very high BEC. This seems to be a real phenomenon, seen in all three treatment arms. One plausible explanation may be the misdiagnosis of an episode of eosinophilic pneumonia in patients who also have COPD. This is a T-helper 2 cell-mediated disease, so the fact that the increase in pneumonia rate was seen most clearly in patients who are not on ICS provides some support for this suggestion [13]. Our findings have important implications for the use of ICS-containing triple therapy. Since all treatments should be used in the context of risk versus benefit, whilst it is clear that at a population level, the probability of benefit with ICS to reduce exacerbations is lower in patients with lower BEC, this is not accompanied by increased risk. In this context, it is paradoxical and unfortunate that the patients who are at greater risk of an exacerbation, specifically older patients [14], those with lower FEV1 [15] or those with lower BMI [16], also have a greater risk of pneumonia. Risk and benefit are also important when considering ICS-containing treatment for patients at lower risk of exacerbations and earlier in their disease. Such patients are at lower risk of pneumonia by virtue of their age and lung function and are less likely to have emphysema-related loss of BMI. Thus, whilst they may have smaller benefits in terms of exacerbation reduction, they also have a lower risk of pneumonia. This consideration lends support to the design of studies testing triple therapies in younger and milder patients with COPD in an attempt to alter disease progression [17]. Therefore, the potential pneumonia risk needs to be balanced against potential benefits across a range of patient outcomes at the individual patient level [18,19,20]. The strengths of this analysis of the IMPACT study include the large sample size (N = 10,355) and the long duration of the study (52 weeks), which allowed for increased precision in the analysis of trends of patient factors that contribute to the risk of pneumonia in COPD. However, it should be noted that these analyses are descriptive and were conducted post hoc.

In addition to age, BMI, COPD severity and BEC levels, there are several factors that may elevate the risk of pneumonia in patients with COPD. One of the limitations of the current analysis was that it did not assess other known factors implicated in potentially increasing the risk of pneumonia in patients with COPD, including smoking, oral corticosteroid use, history of exacerbations, dosage levels of ICS and presence of comorbidities such as cardiovascular diseases, asthma and bronchiectasis. These factors, and their association with pneumonia and ICS use, should be evaluated in future research.

Conclusions

The increasing annual rate of pneumonia was associated with older age, lower BMI and lower FEV1. The key new observation in this study is the absence of any relationship between blood eosinophils and ICS-associated pneumonia risk. It seems clear that lower benefit versus risk in patients with lower eosinophils is due to lower potential benefit, not increased risk. Paradoxically, the patients at greatest risk of pneumonia are those who may have the greatest need for ICS-containing treatment to reduce the risk of exacerbations. These observations provide clinicians with additional insights to guide decisions concerning ICS-containing treatment for COPD at the individual patient level.