Introduction

Despite the availability of evidence-based guidelines and continuing innovations in prevention, ventilator-associated pneumonia (VAP) remains a significant threat for patients receiving prolonged mechanical ventilation in the intensive care unit (ICU) [1–3]. On average, 10–20 % of patients mechanically ventilated for more than 24 h develop VAP [4]. However, depending on the availability of resources and the specific risk profile of the index population, the overall prevalence may be substantially higher, reaching up to 40 % in particular patient subgroups [5, 6]. This high incidence is particularly significant given the important impact of VAP in terms of morbidity, mortality and healthcare costs [3, 7, 8]. The added morbidity of VAP contributes to an estimated excess length of hospitalisation of 4–6 days [4, 9]. Furthermore, patients with VAP may face a mortality risk twice that of similar ICU patients without VAP [4]. As with the risk of VAP onset, outcomes associated with VAP might vary according to the severity of acute illness and presence of underlying conditions [10, 11]. The clinical relevance of chronic respiratory disease to the occurrence and outcome of VAP warrants particular consideration.

Chronic obstructive pulmonary disease (COPD) is characterised by airflow limitation associated with an abnormal inflammatory response of the lungs to noxious gas particles [12–14]. In patients with advanced disease, emphysema may lead to further airflow limitation [12–14]. COPD is a major risk factor for community-acquired pneumonia (CAP), as the patient’s defences might be impaired due to long-term corticosteroid treatment and reduced microbial clearance [13–18]. Whether COPD increases mortality in CAP remains, however, controversial [19–24].

The relationship between COPD, VAP and mortality has not been thoroughly investigated. A previous single-centre study has reported increased ICU mortality associated with VAP in patients with COPD [25]. Using data from a large multicentre observational cohort, our objectives were to determine whether patients with COPD have a higher rate of development of VAP than those without COPD and whether patients with COPD who develop VAP have different microbiological aetiology and outcomes than patients without underlying COPD or COPD without VAP. Our hypothesis was that COPD patients who develop VAP might represent a unique patient group in terms of incidence, prognosis, aetiology and management.

Materials and methods

This study is a secondary analysis of the European (EU)-VAP/CAP database [26]. The EU-VAP/CAP project was a prospective, observational survey conducted in 27 ICUs from nine European countries: Belgium, France, Germany, Greece, Italy, Ireland, Portugal, Spain and Turkey. All patients requiring admission for a diagnosis of pneumonia or invasive mechanical ventilation for ≥48 h, irrespective of the admission diagnosis, were included. Patients that received only non-invasive mechanical ventilation were excluded. The target was the inclusion of 100 consecutive admissions in each ICU. This study was conducted in accordance with the 1964 declaration of Helsinki and its later amendments. The participating centres either received institutional ethical approval or had the requirement for ethical approval waived. Informed consent was waived due to the observational nature of the study and the collection of anonymised data.

Patient demographics, primary diagnosis, comorbidities, McCabe classification of comorbidities, Simplified Acute Physiology Score (SAPS) II on admission, ICU and hospital length of stay (LOS), duration of mechanical ventilation and ICU mortality were recorded for all patients. For patients with a clinical diagnosis of pneumonia as judged by the attending physicians, data collection included pneumonia type, clinical signs, sepsis severity (sepsis/severe sepsis/septic shock), SAPS II on the previous day and the day of clinical suspicion for VAP, Sequential Organ Failure Assessment (SOFA) score on the day of clinical suspicion of pneumonia, diagnostic procedures performed, microbiologic data, treatment data and outcome data [27–30]. The definition of septic shock followed published definitions from the 1992 American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference on sepsis and organ failure [27]. Organ dysfunction was defined according to SOFA score definitions [30].

Chronic disease status was categorised according to McCabe’s classification: non-fatal underlying disease or no underlying disease, ultimately fatal underlying disease (<5 years expected survival) and rapidly fatal underlying disease (<1 year expected survival) [27]. Each clinical episode of pneumonia was described separately. However, for patients who developed more than one episode of VAP, only variables related to the first episode of VAP were analysed.

VAP was defined as a pulmonary infection arising ≥48 h after endotracheal intubation with no evidence of pneumonia at the time of intubation or the diagnosis of a new pulmonary infection if the initial admission to ICU was for pneumonia [8]. Early-onset VAP was defined as VAP developing 2–4 days after intubation, while late-onset was defined as VAP developing ≥5 days after intubation. Microbiologically confirmed VAP (VAP with definite aetiology) was defined as VAP with a microorganism isolated from respiratory samples or blood in a patient with suspicion of pneumonia and characterised as definite aetiology of VAP according to clinical judgment and interpretation of specimen culture results [26]. COPD was defined as a pre-existing disease state characterised by the presence of airflow limitation due to chronic bronchitis or emphysema [19, 20]. The presence of COPD was recorded by the attending physician based on pulmonary function tests before ICU admission where available, or on clinical criteria, medical history (including significant smoking exposure, previous use of respiratory medications) and evidence of hyperinflation on chest radiograph [19, 20].

The study is reported in accordance with observational study guidelines [31]. Further information on the methods, variables and definitions of the study has been reported previously [26].

Trauma patients were excluded from the current analysis as they represent a group with distinct characteristics, including a significantly lower prevalence of COPD [32, 33]. Patients with and without COPD were compared. Also, patients with COPD that developed VAP were compared to those that did not develop VAP. Outcome variables in this analysis were length of ICU stay and duration of mechanical ventilation, mortality and time to event (ICU death) calculated from the date of VAP diagnosis.

Statistical analysis

VAP prevalence was compared between patients with and without COPD using Pearson’s Chi-squared test. VAP incidence was expressed as first episodes of VAP per 1000 ventilator days at risk across 26 out of 27 sites, given non-consecutive enrolment at one site. Incidence calculations were performed for mechanical ventilation days to VAP onset, with censoring at the cessation of mechanical ventilation in patients who did not develop VAP (mechanical ventilation days at risk). Categorical variables are expressed as frequency and percentage. Continuous variables are expressed as mean and standard deviation (SD) or median and inter-quartile range (IQR), depending on whether the data followed a normal or non-normal distribution, respectively. A Student’s t-test for normal data and a Mann–Whitney U-test or Kruskal–Wallis H-test for non-normal data were used to compare continuous variables between groups. A Pearson Chi-squared test or Fisher’s exact test was used to compare differences in categorical variables between groups. Logistic regression was used to identify independent predictors of ICU death in patients with COPD; variables identified in the bivariate analysis (p < 0.20; event rate >5 %) were included in the model; site was entered as an independent factor into the model to control for site differences in mortality. Odds ratios (ORs) and 95 % confidence intervals (CIs) are reported. Manual and backward stepwise techniques were used to identify the model with best fit. A two-sided p-value <0.05 was considered statistically significant. Statistical analysis was conducted using IBM SPSS 21 (IBM Corporation, Armonk, New York).

Results

A total of 2082 patients were included in this analysis, of which 397 (19.1 %) had COPD; 79 COPD patients developed VAP (Fig. 1). Demographic and outcome characteristics of COPD patients compared to non-COPD patients are shown in Table 1. ICU mortality rates in COPD and non-COPD patients in our cohort were equivalent to predicted mortality based on SAPS II at ICU admission (the observed and predicted mortalities were 35 and 37 %, respectively).

Fig. 1
figure 1

Study profile

Full size image
Table 1 Demographic, clinical and outcome characteristics of mechanically ventilated patients with and without COPD
Full size table

The prevalence of VAP in 26 study centres with consecutive enrolment was 18.3 %; there was no significant difference between patients with or without COPD (18.6 % vs. 18.2 %, p = 0.89). The incidence of VAP was 18.2/1000 ventilator days at risk, with no significant difference between patients with or without COPD (16.7 vs. 18.6 per 1000 ventilator days at risk, p = 0.41). There was no statistically significant difference between the incidence of VAP in COPD and non-COPD patients when those with neurological major organ failure at ICU admission were excluded (15.5 vs. 17.5 per 1000 ventilation days at risk, p = 0.40) and when patients with CAP on admission were excluded (17.9 vs. 19.6 per 1000 ventilation days at risk, p = 0.53). The median onset of VAP was not significantly different between patients with and without COPD (6 [4–15] vs. 6 [4–10.8] days; p = 0.48).

There were no significant differences in the demographics, clinical characteristics, comorbidities and severity of illness at ICU admission (as assessed by SAPS II) between ICU patients with COPD who developed VAP and those who did not develop VAP (Table 2). VAP development was associated with a longer total duration of mechanical ventilation compared to those that did not develop VAP, including in the subgroup of survivors. Also, VAP development was associated with a longer ICU stay (13.5 days) compared to those without VAP, including among survivors (13 days). Patients with COPD and VAP had higher mortality than COPD patients without VAP (48.1 % vs. 31.1 %, p = 0.005). In patients with COPD, VAP (OR 2.28; 95 % CI 1.35–3.87) and SAPS II (OR 1.03; 95 % CI 1.02–1.05) were independent predictors of ICU mortality (Table 3). When patients with CAP on admission were excluded, VAP (OR 3.19; 95 % CI 1.56–6.51) and SAPS II (OR 1.06; 95 % CI 1.04–1.08) remained independent predictors of ICU mortality.

Table 2 Characteristics of COPD patients who developed VAP versus those who did not develop VAP
Full size table
Table 3 Logistic regression model indicating independent predictors of ICU mortality in patients with COPD
Full size table

Physical and laboratory characteristics on VAP onset by COPD status in patients with VAP were similar, but with a higher prevalence of hypothermia and pleural effusion when COPD was present (data not shown). There were no significant differences in the frequency of septic shock at VAP onset between patients with and without COPD (38.4 % vs. 33.9 %, p = 0.47).

There was no difference in the prevalence of microbiologically confirmed VAP between patients with and without COPD (Table 4). The type of diagnostic technique used (bronchoscopic vs. non-bronchoscopic) did not affect the rate of microbiologic confirmation (74.7 % % vs. 76.5 %, p = 0.41).

Table 4 Aetiology of VAP in patients with and without COPD
Full size table

In COPD patients, there was a higher prevalence of Pseudomonas aeruginosa VAP (overall and with early-onset) compared to those without COPD (Table 4). When patients admitted with CAP were excluded, the prevalence of P. aeruginosa VAP remained higher in patients with COPD than those without (26.4 % vs. 15.8 %, respectively; p = 0.037). Overall, the prevalence of non-fermenting Gram-negative bacilli was higher in COPD patients (p = 0.056), especially in those with early-onset VAP (54.1 % vs. 20.0 % in patients with and without COPD, respectively; p < 0.001). This difference between COPD and non-COPD patients remained significant when patients with CAP were excluded (49.9 % vs. 18.5 %, respectively; p = 0.020). There was no difference in the mean onset of P. aeruginosa VAP in patients with and without COPD (6 [3–16] vs. 8.5 [4.0–16.0] days, p = 0.38).

Discussion

This is the largest study examining the relationships between COPD and VAP from many different European centres. Although COPD was not associated with a higher incidence of VAP than in non-COPD patients, we found that VAP development in COPD patients was associated with a 17 % higher mortality than in those not developing VAP, as well as with increased duration of mechanical ventilation and length of ICU stay. In addition, COPD was associated with a higher prevalence of VAP due to non-fermenting Gram-negative bacilli, notably in early-onset episodes. This association has important potential implications for the selection of empirical antibiotic therapy.

The prevalence of VAP in our cohort falls within the wide range of 6–27 % previously reported [8]. This high variability appears attributable to differences in case-mix and definitions for VAP. For example, the reported VAP prevalence of 6 % was limited to microbiologically confirmed VAP with quantitative tracheal aspirates cultures ≥106 cfu/mL and high rates of prior antibiotic use [25]. Regarding COPD status, we found that VAP prevalence and incidence were not different in patients with and without COPD. This contradicts a number of earlier studies that identified COPD as an independent risk factor for VAP development with a reported OR or relative risk varying from 1.4 to 3.9 [34–36]. Similar to our findings, a large prospective observational study found no significant difference in the incidence of VAP between patients with non-exacerbated COPD and those without COPD (11.9 vs. 16.0 per 1000 mechanically ventilated days, p = 0.40) [37].

A single-centre case–control study that examined the impact of VAP on patients with COPD found that VAP was independently associated with ICU mortality (OR 7.7; 95 % CI 3.2–18.6) [25]. We also found that VAP development independently increased the risk of death of COPD patients (OR 2.21), confirming the results of the above study in a multicentre, multinational cohort. Our findings that VAP development in COPD patients increased the median duration of mechanical ventilation by 12 days and the median ICU stay by 13.5 days are also in line with the results of the above study that found a mean increase of 11 days in both the duration of mechanical ventilation and the length of ICU stay [25].

In relation to the microbiological flora causing VAP in COPD patients, we found that P. aeruginosa was more frequently isolated in VAP patients when COPD was present. It has been previously reported that P. aeruginosa (31 %) was the most frequently isolated pathogen in COPD patients with microbiologically confirmed VAP, and that the risk of VAP due to P. aeruginosa was increased in patients with COPD [25, 38, 39]. We found that P. aeruginosa was the second most common pathogen in COPD patients (29 %), behind Enterobacteriaceae. Of note, non-fermenting Gram-negative bacilli were isolated 2.5 times more often in COPD patients with early-onset VAP, with P. aeruginosa being the most frequent isolate (33.3 %). This observation has obvious implications in the selection of the appropriate empirical antibiotic regimen when VAP is suspected in COPD patients [8]. Chronic use of corticosteroids in COPD patients might be the reason for this opportunistic pathogen being more frequently observed as a cause of VAP. Unfortunately, our database did not allow checking for the relationship between corticosteroid exposure and P. aeruginosa. However, chronic steroid use was included in the definition of immunosuppression, and the prevalence of immunosuppression did not differ significantly between VAP patients with and without COPD.

The current study, to the best of our knowledge, is the first to explore in detail the relationship between COPD and VAP in a multicentre cohort and the first that describes differences in the aetiology of VAP by COPD status, separately in early- and late-onset VAP.

Whilst this is a ‘real-world clinical study, it does have several limitations. First, it was an observational study with centres predominantly in central and southern Europe. The non-random selection of sites may mean that the results, including the incidence and microbiology of pneumonia, may not be widely generalisable. Similarly, the prevalence and incidence of VAP cannot be extrapolated to all ICU patients because we focused on non-trauma patients requiring invasive mechanical ventilation. Second, this was a secondary analysis of a study not specifically designed to look at the association of COPD and VAP. Therefore, there may be unmeasured factors that modulate the risk and outcomes of VAP in patients with COPD. Finally, data on the severity of COPD as defined by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) were not recorded and, therefore, we could not explore the potential relationship between the severity of COPD and the occurrence of VAP or mortality [12, 14]. However, including only COPD patients with pulmonary function tests before ICU admission would lead to a selection bias and an underestimation of COPD incidence [40]. Not all patients with COPD are diagnosed before critical illness; on the contrary, it is not uncommon for COPD to be diagnosed for the first time during ICU admission [40]. The probability of COPD is very high in patients with smoking history, signs of hyperinflation and/or emphysema in chest X-ray and/or expiratory flow limitation during control mechanical ventilation [40]. In everyday clinical practice, it is common that ICU physicians need to make decisions regarding empirical antibiotic treatment for COPD patients with a suspicion of VAP without having available information on pulmonary function tests and GOLD classification.

Conclusions

Our findings have several important implications for clinical practice. First, they highlight the significant impact of ventilator-associated pneumonia (VAP) development on the morbidity and mortality of patients with chronic obstructive pulmonary disease (COPD) and the need for rigorous VAP prevention measures and prompt initiation of appropriate empirical treatment when VAP is suspected. Second, our findings are consistent with prior studies on colonisation or risk factors for Pseudomonas aeruginosa in mechanically ventilated patients. They suggest the inclusion of an anti-pseudomonal antibiotic in the initial empirical antibiotic treatment of VAP when COPD is present, regardless of the time of onset. Due to the risk of delaying inappropriate therapy, these findings should be considered in future guidelines. Very limited outcome data are available and the current study may act as a platform for a randomised clinical trial specifically designed to improve outcomes. This study reinforces the diversity of VAP, in contrast with community-acquired pneumonia (CAP), and the need to consider VAP as a heterogeneous entity influenced by patient case-mix.