Background
Despite the approval of new antibiotics and the implementation of ventilator-associated pneumonia (VAP) prevention bundles, VAP remains a major concern and burden to the healthcare system, being associated with increased mortality, morbidity, hospital length of stay (LoS), and healthcare costs [
1‐
4]. Several processes are implicated in VAP pathogenesis: aspiration of oropharyngeal secretions resulting from an altered state of consciousness, loss of natural protective mechanisms of the airways, and direct pathogen inoculation at the time of intubation [
5]. Colonization of the airways by pathogens detected by routine bacterial cultures from endotracheal aspirate (ETA) may be useful to identify patients at increased risk for development of VAP [
6].
In the case of VAP radiological and clinical signs, an ETA sample demonstrating high bacterial burden is considered confirmatory of microbial etiology according to recently issued guidelines from the Infectious Diseases Society of America [
7].
Staphylococcus aureus,
Klebsiella pneumoniae,
Pseudomonas aeruginosa,
Acinetobacter baumannii, and
Escherichia coli are the most common pathogens associated with VAP [
5]. These microorganisms are often multi-drug resistant (MDR) [
8] and/or highly virulent, making early detection particularly important for initiating the appropriate treatment regimen [
9‐
11].
The majority of reports focuses on the analysis of microbiologically confirmed VAP cases; less data are available on frequency and timing of airway colonization, and semi-quantitative culture methods, as well as their value for identification of patients at increased risk of developing VAP. The purpose of this study was to analyze semi-quantitatively cultured, serially collected ETA samples to better understand the association of bacterial colonization with progression to VAP in mechanically ventilated patients. We hypothesized that patients heavily colonized with pathogenic bacterial species in the trachea were more likely to develop culture-positive VAP during their ICU stay, and that heavier burden of colonization is associated with a higher risk of developing a bacterial VAP. VAP incidence in relation to bacterial airway colonization was examined as the primary outcome of the study. Secondary outcomes that were evaluated included the duration of mechanical ventilation (MV), ICU and hospital LoS, as well as all-cause mortality in patients with and without VAP.
Methods
Study population and data collection
A total of 250 mechanically ventilated patients admitted to two medical ICUs (MICU) and one surgical ICU (SICU) at the Lahey Hospital & Medical Center (Burlington, MA, USA) between June 2014 and June 2015 were included in this prospective observational study. Exclusion criteria included age below 18 years, and an expected length of MV of less than 48 h as judged by the treating physician (e.g., admission to the ICU from the operating theatre just for weaning, or moribund condition of the patient). Subjects exited the study upon extubation, tracheostomy, transfer to another facility, death, or initiation of “comfort measures only” protocol. Basic demographic data and the medical history were collected upon ICU admission from the patient directly, from the relatives with the help of collateral history taking, and/or from the available medical records. Hospital and ICU LoS, length of MV was recorded at the end of hospitalization, and diagnostic parameters, such as body temperature, white blood-cell count, and chest X-ray (CXR) readings were recorded daily during the ICU stay. Additionally, local epidemiology data on antibiotic susceptibility of the bacterial isolates, as well as empiric and prescribed antibiotic regimens were recorded.
All patient-related data were received in de-identified form. Informed consent was waived, since only leftover materials (otherwise discarded) were used, and no additional intervention or change in treatment plan was implemented. CXR was assessed by two independent, trained observers; in the case of non-unanimous judgement, the opinion of a senior radiologist was used as decisive. ETA samples were to be collected daily as part of the standard of care.
ETA samples were cultured and analyzed in the clinical microbiology laboratories of the Lahey Hospital and Medical Center. Bacterial species were determined by standard methods. Local rate of MDR pathogen isolation was not assessed separately for the ICUs included in the study. However, upon the commencement of the study global susceptibility rates of all isolates in the Lahey Hospital and Medical Center for the year preceding the study was provided. Multi-drug resistance was determined by demonstrating resistance to multiple antibiotic classes, with the conventional microbiological methods and Microscan® panels (latter used for all organisms except
P. aeruginosa, where a disc diffusion test was employed), and/or detection of ESBL or carbapenemase activity. Alpha-hemolytic
Streptococcus spp., apathogenic
Neisseria spp.,
Bacteroides spp.,
Fusobacterium spp.,
Spirochaetes, and
Candida spp. were considered as normal respiratory flora (NRF) [
12]. Semi-quantitative microbiological analysis of ETA samples (SQ-ETA) was performed; samples were streaked onto appropriate agar culture plates in four consecutive quadrants. Presence of 5 or less colonies in the first quadrant corresponded to 1+ (rare), of 6 and more colonies in the first quadrant to 2+ (few), any amount of colonies in the first and second quadrants to 3+ (moderate), and in three consecutive or all 4 quadrants corresponded to 4+ (many). SQ-ETA data were used to categorize bacterial burden as light (1+ and 2+) or heavy (3+ and 4+). Heavy colonization corresponded to approximately ≥10
5 CFU/mL of ETA [
13]. Purulent ETA samples were detected by light microscopy and defined as > 10 polymorphonuclear cells per low power field. Pathogenic bacterial species recovered from the ETA were shipped to the sponsor, and additional analyses (species confirmation and MRSA status assignment by genetic tests) were performed as described previously [
14].
The study was conducted in accordance with the ethical principles of the Declaration of Helsinki and the local regulations; and was approved by the Lahey Hospital and Medical Center Institutional Review Board.
VAP diagnosis
Modified criteria outlined by Johanson et al. [
15] were used for retrospective assignment of VAP by the sponsor and was defined by the presence of a new or progressive infiltrate determined by CXR, and at least two of the following clinical findings: (1) fever or hypothermia, (2) leukocytosis or leukopenia, and (3) presence of purulent respiratory secretions, all appearing > 2 days after initiation of MV. For this study, patients were defined as “progressing” to VAP if they had bacterial colonization and one or more episodes of VAP at any time during the study period. A bacterial pathogen was considered a potential causative agent of VAP if isolated from the ETA on the same day or within the 2 days preceding or following the detected of VAP clinical signs. In this study, both heavy and light ETA bacterial burden were accepted for determination of VAP bacterial etiology for the purpose of a more comprehensive overview. In selected patients, upon orders of the treating physicians, bronchoalveolar lavage (BAL) and/or non-bronchoscopic bronchoalveolar lavage (NBAL) was performed during the ICU stay. In order to avoid potential bias in this study, however, results of these tests were not included in the analysis, as they were not carried out routinely, were not limited to cases of suspected pneumonia (e.g. lung malignancy, interstitial lung disease) and were performed in less than 10% of the study population. No other nosocomial infections other than VAP was systematically assessed or recorded in this study.
Ventilator-associated tracheobronchitis (VAT) was not assessed as part of this study, as the authors believe that the retrospective diagnosis of VAT, in absence of documented radiological findings as in case of VAP, is most accurate when corroborated by the treating clinicians, which conflicted with the design of the study.
Antibiotic use in the studied ICUs
Selective digestive decontamination (SDD) was not administered as part of routine patient care. Upon suspected VAP, cefepime and vancomycin was administered empirically; if a patient was found to be negative for S. aureus, vancomycin was removed from the regimen, and, depending on recovered bacterial flora, cefepime was continued or changed to ceftriaxone or cefazolin. In case of cephalosporin allergy levofloxacin was administered.
Statistical analysis
Reported variables were grouped into continuous (e.g. age, MV duration, hospital LoS), and categorical values (e.g., gender, underlying disorders, all-cause mortality). Continuous variables were represented as mean with standard deviation, whereas categorical variables as percentage. Continuous variables were tested with Student’s two-tailed unpaired t-test; categorical ones were organized into contingency tables and tested with Fisher’s exact test. Statistical tests were performed with the Prism® 6.07 (GraphPad) software package. Level of statistical significance was set at p value ≤0.05.
Discussion
In this study, we investigated the microbiology of upper airway colonization in ICU-admitted mechanically ventilated patients. We noted significant differences in MV duration, hospital LoS and mortality between the patients admitted to the surgical or medical ICUs. Similar observations on mortality rates have been reported [
16]. Importantly, there were no significant differences in basic demographic parameters (age, gender, BMI, etc.) and in the clinical variables (comorbidities, smoking and alcohol history, etc.). This supports the generalization of our findings on the airway bacterial colonization and VAP incidence to the studied patient population. Approximately one third of all patients developed VAP. When assigning a pathogen to a VAP event, a two-day period prior to the clinical diagnosis has been suggested as clinically appropriate [
17]. Patients categorized by ETA bacteriological analysis results as
S. aureus-positive, yielding Gram-negative organisms, or no bacterial growth/NRF, all developed VAP at about the same rate, 25, 24, and 31%. The principal pathogens implicated in VAP as shown in this study are consistent with universally reported species [
9,
10,
18], as is the proportion of VAP without bacterial association [
2].
S. aureus was the most prominent individual pathogenic species isolated. Higher prevalence of
S. aureus VAP in the United States compared to Europe has been widely published [
19]. The
S. aureus tracheal colonization rate (~ 30%) from this study was similar to that reported 5 years earlier in the same ICUs [
14]. Data on the temporal pattern and onset of
S. aureus tracheal colonization is limited; however, available studies report both MSSA and MRSA colonization appearing early during the MV period [
20‐
22]. There was a tendency for a higher MSSA-to-MRSA ratio in VAP patients when compared to the ratio of patients colonized without pneumonia, which is consistent with our previous observation, and is in line with reports of MRSA isolates characterized with higher persistence, but lower virulence [
14,
23]. Nearly half of the
S. aureus-positive ETA samples were detected within the first two days of mechanical ventilation, irrespective of methicillin resistance.
A major focus of this study was to evaluate whether differences in semi-quantitative bacterial ETA colonization could support the identification of patients who would more likely progress to VAP with the colonizing bacteria. We found that high ETA bacterial burden increases the likelihood of progression to bacterial VAP, as demonstrated by a significantly higher proportion of patients heavily colonized by Gram-negative bacteria progressing to VAP (30.0%) compared to those lightly colonized (0%). For
S. aureus this difference did not reach statistical significance, although heavily colonized patients still progressed to
S. aureus VAP more frequently (32.4%) than lightly colonized (17.6%). The detection of
S. aureus in ETA preceded the onset of VAP by approximately 4 days, offering clinicians time to identify and characterize the bacterium and apply prophylactic measures. For Gram-negative bacteria, however, this period was shorter, only 2.5 days. Although antibiotic intervention is not recommended as prophylaxis of VAP and has shown to be ineffective [
24], the early choice of appropriate antibiotic therapy is expected to reduce LoS, MV duration, and mortality [
25,
26]. Moreover, several pathogen-specific non-antibiotic approaches (such as monoclonal antibodies) are being evaluated in preemptive settings, mainly targeting
S. aureus, but also MDR Gram-negative pathogens [
27,
28]. Identifying patients at increased risk of pneumonia allows to focus clinical trial study populations on those patients that would most likely benefit from preemptive intervention.
We observed that heavy colonization by
S. aureus based on the bacterial load in the ETA was associated with numerically higher VAP rate (32.4%) compared to light colonization (17.6%), when the ETA was also positive during the VAP event. However, similarly high rate (31.3%) of VAP was detected among those patients whose ETA samples were negative for potential bacterial species before VAP diagnosis. This was a surprising finding, and calls for possible explanations. For this study, only ETA samples were considered in assigning a bacterial pathogen to a VAP episode, and the information on deep respiratory sample (such as bronchoalveolar lavage or a protected specimen brush) microbiology was not included. This constitutes one of the limitations of the study, and may explain some of VAP episodes with no causative bacterial species identified. At the same time, apart from technical reasons (e.g., low bacterial inoculum size due to successful antibiotic therapy, suboptimal sampling), this may also be linked to non-bacterial pathogens such as viruses or fungi, not routinely cultured organisms (such as
Legionella pneumophila), as well as pathogens not covered by the standard microbiological culture methods [
5]. Some reports state that as many as 56% of the pathogens causing VAP were not identified by the standard microbiological methods, and suggested a higher complexity of VAP microbiology than currently thought [
29]. Additionally, non-infectious lung diseases, and diseases with secondary lung involvement mimicking pneumonia need to be considered. A critically ill, multimorbid patient with signs of infection and chest X-ray infiltrate, but negative bacterial culture of airway secretions presents a diagnostic challenge, and requires a careful diagnostic workup by the treating physician to consider other conditions with clinical presentation similar to VAP [
30]. Microbiologically non-confirmed VAP remains under-represented in the literature and needs to be addressed by further research, including.
In one fifth of all monomicrobial VAP cases, pathogens were present at low abundance (1+ or 2+ SQ-ETA) during the VAP-relevant period. Along with the scarce presence of bacterial cells in the respiratory tract, such results can also be explained by suboptimal or difficult sampling, or by the low number of viable bacterial cells as a result of antibiotic therapy, which is frequent in the ICU population. There is no consensus on how to interpret such cases. While targeted antibiotic therapy for VAP is only recommended after confirmatory qualitative respiratory sample with bacterial counts at or above diagnostic threshold [
7], lower bacterial burden associated with VAP in certain patients (high clinical suspicion, deteriorating patient, absence of other infection source) cannot be ignored.
Other significant observations included hospital length of stay, time on mechanical ventilation, mortality, and time to onset of VAP. Importantly, significant differences in hospital LoS and MV duration were observed between patients with and without VAP, regardless of the detectability of bacterial pathogens in ETA samples. No significant differences were seen in all-cause mortality as a consequence of VAP, although there was a trend for approximately 10% higher mortality when an ETA bacterial pathogen was detected, and a 10% lower mortality when no bacterial cause was identified. Previous studies on bacterial VAP also reported approximately two-fold increase in MV duration and hospital LoS and significant mortality difference of ~ 10% (attributable mortality) [
1,
3,
31]. We observed a temporal difference between the bacterial pathogen-positive and -negative VAP groups, with the former being detected in autumn, and the latter peaking in the winter period, indicating possible seasonality of both groups.
The limitations of the present study need to be acknowledged. First, although data and samples were prospectively collected, diagnosis of VAP was assigned retrospectively by the sponsor based on generally accepted criteria and not by the treating physicians’ diagnosis. Second, the relatively small sample size, confined to a single medical center, limits the generalization of our data to other centers and ICUs. Moreover, analysis of the contribution of individual bacterial species other than S. aureus to VAP is limited due to low case number.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.