Risks of ventilator-associated pneumonia and invasive pulmonary aspergillosis in patients with viral acute respiratory distress syndrome related or not to Coronavirus 19 disease

We conducted a cohort study which retrospectively enrolled, between October 1, 2009, and April 29, 2020, all patients referred to the medical ICU of a French tertiary hospital for viral ARDS and who required mechanical ventilation for more than 48 h. ARDS was defined according to the Berlin definition [6]. C-ARDS were patients with ARDS and a positive polymerase chain reaction (PCR) test for SARS-COV2. NC-ARDS patients were those having ARDS and a positive polymerase chain reaction (PCR) test for influenza, human metapneumovirus, respiratory syncytial virus, parainfluenza virus, other endemic human coronaviruses (OC43, NL63, HK-U1, 229E), and adenovirus. RT-PCR duplex targeting influenza A/B, respiratory syncytial virus, metapneumovirus, coronavirus, adenovirus, and parainfluenza were performed using r-gene™ (Argene, Biomérieux S.A.) according to the manufacturer’s instructions. We carefully identified all patients with viral ARDS using a triple check involving the ICU medical reports, the medical information system database, and the virology department registry. Of note, respiratory viruses were systematically searched in all patients admitted in our ICU for pneumonia. This observational study was approved by the Institutional Review Board of the French intensive care medicine society (CE SRLF 20-45) and informed consent was waived.

The mechanical ventilation of ARDS patients followed a standardized protective ventilation strategy [7]. Other treatments, including neuromuscular blocking agents, inhaled nitric oxide, prone positioning, and veno-venous extra-corporeal membrane oxygenation were administered with respect to national guidelines [8]. Sedation protocol was based on the adaptation of the dose of sedative (midazolam or propofol) on the Richmond Agitation Sedation Scale by the nurse every three hours and did not significantly changed during the study period. Preventing VAP followed an educational program, regular reminders and feedback with a prevention care bundle in accordance with guidelines [9], based on hand hygiene with alcohol-based sanitizer, inclining patients in a semi-recumbent position (30°–45°), oral chlorhexidine mouth washing at least four times a day, tracheal cuff pressure maintenance between 20 and 30 cm H₂O, orogastric rather than nasogastric tubes, and daily chlorhexidine body washing. No routine antibiotic prophylaxis or decontamination antibiotic regimens were prescribed and if stress ulcer prophylaxis was needed, proton pump inhibitor was preferred. All patients had a closed tracheal suction system without daily change.

Demographic characteristics, comorbidities, Charlson comorbidity index [10], clinical, biological, and imaging features at ICU admission, and consumption of alcohol-based handrub liquid (retrieved from hospital pharmacy) was collected, then analyzed on May 25, 2020, after a minimal follow-up period of 28 days for the most recent patients. Respiratory tract secretions were cultured for VAP diagnosis purposes, and susceptibility profiles of recovered microorganisms were recorded.

The primary endpoint was the difference in incidence of first VAP between C-ARDS and NC-ARDS patients. VAP was clinically suspected if any of its classical criteria happened 48 h or more after mechanical ventilation initiation: new or worsening infiltrates on chest roentgenogram, systemic signs of infection (new-onset fever, leukocytosis or leucopenia, increased need for vasopressors to maintain blood pressure), purulent secretions, and impaired oxygenation [11]. All suspected VAP were confirmed from quantitative cultures of lower respiratory tract secretions sampled before administering new antibiotics using a blinded protected telescope catheter [12] or bronchoscopy (10³ and 10⁴ colony forming units/mL for protected telescope catheter and bronchoalveolar lavage, respectively). VAP onset was defined as the day on which the lung sample tested positive. The secondary endpoints were the difference in occurrence of bacterial coinfection at ICU admission, putative invasive pulmonary aspergillosis, and multi-drug-resistant bacteria (MDR) VAP. MDR pathogens included methicillin-resistant Staphylococcus aureus, extended-spectrum β-lactamase-producing Enterobacteriaceae (ESBL-PE), and carbapenem-resistant Enterobacteriaceae (CRE). During the study period, our management of VAP/hospital-acquired pneumonia (HAP) was derived from that recommended by the French guidelines to treat adult ICU-acquired pneumonia [13]. Bacterial coinfection at ICU admission was evidenced by the detection of bacteria in the sputum or in blood samples, in the absence of other sources of infection, or by a positive pneumococcal or L. pneumophila serotype 1 urinary antigen test.

We based our definition of invasive pulmonary aspergillosis on one of the following: (1) the recently published influenza‑associated pulmonary aspergillosis definition by expert panel (Influenza-Associated Pulmonary Aspergillosis—IAPA case definition) [14]; (2) the crude AspICU algorithm, as proposed by Blot et al. to distinguish putative invasive pulmonary aspergillosis from respiratory tract Aspergillus spp. colonization in critically-ill patients, relying on clinical, radiological, and mycological criteria [15]; (3) the modified AspICU algorithm of invasive pulmonary aspergillosis, which includes the positivity of serum or bronchoalveolar lavage galactomannan, as proposed by the Dutch-Belgian Mycosis study group to define invasive pulmonary aspergillosis in critically ill patients with severe influenza [16].

No statistical sample size calculation was performed a priori, and sample size was equal to the number of patients treated during the study period. The results are reported as median and interquartile range (25th–75th percentiles) or numbers with percentages. Initial bivariate statistical comparisons were conducted using χ² or Fisher’s exact tests for categorical data and Mann–Whitney U test for continuous data.

As the risk of VAP cumulatively increases over time of mechanical ventilation, death and ventilator weaning are competing risks for VAP occurrence [17]. Patients are no longer at risk for VAP after death or ventilator weaning; conversely, weaning may be prolonged because of VAP. In this context, standard survival methods (Kaplan–Meier method and Cox model) are inappropriate because they assume that censoring is non-informative [18], hence the need to consider specific competing risk methods. We therefore used a competing risk model (cumulative incidence function of the Gray model) [19, 20] to properly estimate the effect of COVID-19 on VAP development, while considering death and ventilator weaning as competing events. The strength of the association between each variable and the outcome was assessed with the sub-hazard ratio and the cumulative incidence function, estimated using cmprsk package developed by Gray in R software (http://​biowww.​dfci.​harvard.​edu/​~gray/​cmprsk_​2.​1-4.​tar.​gz). Because there were differences between the two groups, resulting from the particular profile of patients prone to have severe forms of COVID-19 (e.g., cardiovascular comorbidities) or non-COVID-19 viral pneumonia (e.g., immunosuppression), we further matched C-ARDS and NC-ARDS and performed a sensitivity analysis as follows. First, we screened all differences between groups. Second, we used a seminal review summarizing risk factors for VAP [21], to select among variables that were different between the two groups, those that could explain more VAP in the C-ARDS group as compared to the NC-ARDS group. Following this process, 68 C-ARDS patients could be matched 1:1 to 68 NC-ARDS for ARDS severity (mild, moderate, or severe) and diabetes mellitus. Of note, these matched pairs were comparable regarding age and gender. A sensitivity analysis with competing risk was performed in this matched cohort. Two-sided p values < 0.05 were considered significant. The other analyses were conducted using SPSS Base 21.0 statistics software package (SPSS Inc., Chicago, IL).

Between October 1, 2009, and April 29, 2020, 3821 consecutive patients were mechanically ventilated in our ICU. Among them, 199 had a viral positive PCR, including 172 pneumonia with ARDS criteria that were mechanically ventilated for more than 48 h. Ninety patients had C-ARDS with positive real-time reverse transcriptase PCR tests for COVID-19, while 82 patients had NC-ARDS, including 50 with severe influenza (with two respiratory syncytial virus coinfections), six with endemic human coronavirus (with one respiratory syncytial virus coinfection), 14 with respiratory syncytial virus alone, five with human metapneumovirus, five with parainfluenza, and two with adenovirus. Thus, the present study comprises 90 patients with C-ARDS ad 82 with NC-ARDS (Additional file 1: Figure S1).

The characteristics of C-ARDS and NC-ARDS patients are displayed in Table 1. NC-ARDS patients had worse past history (Mc Cabe classification and Charlson comorbidity index) and were often immunosuppressed, whereas C-ARDS patients were often diabetic and hypertensive males. At ICU admission almost all patients received antibiotics in both groups, but C-ARDS received less steroid had less organ failure (as assessed by Sequential Organ Failure Assessment score), and lower PaO₂/FiO₂ ratio than NC-ARDS patients. During ICU stay, C-ARDS patients more often required neuromuscular blockade, prone positioning, nitric oxide inhalation, extra-corporeal membrane oxygenation support, and longer duration of mechanical ventilation than NC-ARDS patients. In-ICU and day 28 mortalities were similar in both.

There were significantly fewer documented bacterial coinfections in C-ARDS than in NC-ARDS (Table 1). The microorganisms involved in bacterial coinfection at ICU admission are reported in Additional file 2: Table S1. The types of isolated microorganisms differed between the groups with fewer Gram-positive cocci in C-ARDS than in NC-ARDS, 4/14 (29%) vs. 23/39 (59%), p = 0.05, but similar Gram-negative bacilli [9/14 (64%) vs. 19/39 (49%), p = 0.32].

At day 28 of ICU admission, significantly more patients developed at least one VAP episode in C-ARDS group than in NC-ARDS group: 56 (62%) vs. 35 (43%), p = 0.016. The mechanical ventilation lasted longer in C-ARDS than in NC-ARDS group: 16.5 [9.0–28.8] vs 9.0 [6.0–17.3] days, p < 0.0001. Fine and Gray model results showed that VAP probability was significantly higher in C-ARDS group after adjusting for death and ventilator weaning [sub-hazard ratio = 1.72 (1.14–2.57), p < 0.01, Fig. 1]. Conversely, the probability of successful ventilator weaning was significantly reduced in C-ARDS group after adjusting for VAP and death as competing events [sub-hazard ratio = 0.34 (0.19–0.63), p < 0.001]; the probability of death was similar in both groups [sub-hazard ratio = 1.18 (0.58–2.41), p = 0.64, Fig. 1]. These results persisted after matching for ARDS severity (mild, moderate or severe) and diabetes mellitus, with a higher VAP probability (after adjusting for death and ventilator weaning) and a lower weaning probability (after adjusting for VAP and death) in C-ARDS group [sub-hazard ratio of 1.74 (1.09–2.80), p = 0.02, and 0.37 (0.20–0.70), p < 0.01, respectively]. Risk factors for developing VAP were tested by univariate analysis in Additional file 3: Table S2. The factors associated with VAP occurrence, as shown by multivariable analysis (Additional file 4: Table S3), were C-ARDS [OR = 2.1 (1.1–4.0), p = 0.02] and male gender [OR = 2.2 (1.04–4.5), p = 0.04]. These results were similar by competing risk analysis: VAP probability was significantly higher in C-ARDS group as compared to NC-ARDS [sub-hazard ratio = 1.58 (1.05–2.39), p = 0.03] and in males as compared to females [sub-hazard ratio = 1.72 (1.03–2.88), p = 0.04], while adjusting for death and ventilator weaning.

The most commonly isolated microorganisms in both groups with VAP were Enterobacteriaceae, which were more common in C-ARDS than in NC-ARDS group: n = 42 (72%) vs. 17 (47%), p = 0.01 at the first VAP episode (Table 2). MDR bacteria (all were ESBL-PE except one CRE New Delhi metallo-β-lactamase (NDM)) were retrieved in eleven (20%) C-ARDS and seven (19%) NC-ARDS patients during the first VAP episode (p = 0.57). However, the incidence of MDR VAP was significantly higher in C-ARDS than NC-ARDS during the entire ICU stay: 21 (23%) vs. 9 (11%), p = 0.03. Three C-ARDS patients had MDR VAP caused by CRE [two NDM and one oxacillinase-48 (OXA-48) producing Enterobacteriaceae, all without ESBL coproduction], 18 had ESBL-PE, and one had polymicrobial VAP caused by methicillin-resistant Staphylococcus aureus and ESBL-PE. In NC-ARDS patients, nine had their VAP caused by ESBL-PE. All patients received antibiotics during their ICU stay. Rate of administering Carbapenem for the first VAP was 16 (18%) in C-ARDS vs. 9 (11%) in NC-ARDS, p = 0.21. The three mostly used antibiotics were amoxicillin/clavulanic acid, third-generation cephalosporin, and piperacillin/tazobactam in NC-ARDS group, versus third-generation cephalosporin, carbapenem, and aminoglycoside in C-ARDS group. Carbapenem was more used in C-ARDS than in NC-ARDS patients: 48 (53%), vs 21 (26%), p < 0.01 (Additional file 5: Table S4). Consuming alcohol-based handrub liquid in the ICU reached 135 mL per patient-day for the NC-ARDS study period versus 522 mL per patient-day for the C-ARDS study period.

Diagnostic criteria for invasive pulmonary aspergillosis according to IAPA case definition, crude AspICU definition, and modified AspICU definition are shown in Table 3. There was no proven aspergillosis case in the entire study. Probable aspergillosis (as per IAPA case definition) and putative aspergillosis (as per crude and modified AspICU criteria) were less common in C-ARDS than in NC-ARDS patients (Table 3). According to AspICU algorithm, there was no difference in Aspergillus colonization between C-ARDS and NC-ARDS patients. Univariate analysis of risk factors for developing invasive pulmonary aspergillosis is reported in Additional file 6: Table S5. The factors associated with invasive pulmonary aspergillosis as tested by multivariable analysis (Additional file 7: Table S6) were immunodepression [OR = 3.6 (1.4–9.1), p = 0.01] and having influenza, which fell short of statistical significance [OR = 2.5 (0.88–6.4), p = 0.052). Invasive pulmonary aspergillosis was associated with ICU mortality (Additional file 6: Table S5).

There were significantly fewer documented bacterial coinfections in C-ARDS than in NC-ARDS. Bacterial coinfection rate in NC-CARDS group is in accordance with previous studies on ARDS secondary to influenza. Bacterial coinfection rate at ICU admission is reported in less than 10% in C-ARDS cases [22], except in small series using multiplex PCR assay [23, 24].

The high incidence of VAP found in our study is in accordance with the selected population of ARDS (see flowchart), as reported in previous studies [11, 21, 25, 26]. Using ARDS as a selection criteria allowed inclusion of a relatively homogenous group of patients with reduction of potential bias, given that ARDS is a known major risk factor for VAP. In our cohort, all patients with SARS-CoV-2 pneumonia requiring mechanical ventilation fulfilled ARDS criteria. This finding may be ascribable to the virulence of SARS-CoV-2 and/or to the fact that some intermediate care units have been deployed upstream the ICU for the care of patients not requiring immediate intubation during the pandemic. Invasive mechanical ventilation is a cornerstone in the development of VAP. The duration of mechanical ventilation was twice longer in C-ARDS than in NC-ARDS patients, with more recurrent VAP episodes in the former group. Strategies aimed at avoiding intubation, such as continuous positive airway pressure [27], high-flow nasal oxygen [28], or awake prone position in spontaneously breathing patients [29] should be further explored. Sedation protocols should also be optimized to reduce the duration of mechanical ventilation. SARS-CoV-2 infection was associated with encephalopathy, agitation, and confusion [30]. Our competing risk model yielded a reduced probability of ventilator weaning and a higher probability of VAP in C-ARDS patients whenever adjusted for ventilator weaning. Other factors may influence the risk for VAP in C-ARDS, like infectious process or infection control.

Low dose dexamethasone is reported as the first drug to improve survival in COVID-19 pneumonia [31]. In our work, 36% of NC-ARDS and 12% of C-ARDS patients were on low-dose corticosteroid at ICU admission, yet did not increase their risk for VAP. A recent randomized controlled study on dexamethasone treatment in ARDS did not show more ICU acquired infections [32]. Seven patients in the C-ARDS group received tocilizumab. Anti-inflammatory treatment may be associated with the development of bloodstream infection or late onset infections in recent studies [33, 34].

Hand hygiene compliance rate was not evaluated by internal audits in our ICU during COVID-19 outbreak. Consumption of alcohol-based handrub liquid was 135 mL per patient-day before the outbreak versus 522 mL per patient-day during COVID 19 period. This increase is probably due to the increased number of health care workers coming to reinforce our team. We do not know if the overall increase in alcohol-based handrub solution consumption was associated with a better hand hygiene compliance. Previous studies showed that during routine clinical care of patients with MDR bacteria, health care workers often contaminate protective gowns and gloves [35]. Moreover, teams dedicated to some procedures like prone positioning were created to decrease nurse work strain, but their transversal nature may have increased the risk of cross contamination. Now more than ever, health systems should continue investing in their infection prevention programs, beyond the current pandemic.

Severe influenza infection has been associated with invasive pulmonary aspergillosis. IAPA case definition and modified AspICU algorithm, specifically designed for severe influenza, were used in this study to assess the role of invasive aspergillosis in C-ARDS. A comprehensive diagnostic approach for invasive aspergillosis was implemented in C-ARDS using a systematic serum galactomannan test, PCR and culture of lower respiratory tract secretions for Aspergillus species, and 1,3-β-D-glucan. Four patients with putative aspergillosis had no previous risk factors, suggesting that C-ARDS is a host factor for invasive aspergillosis. However, the incidence of invasive aspergillosis was significantly lower in C-ARDS than in NC-ARDS and was consistent with previously reported patients with bacterial ARDS [36]. Pre-pandemic environmental air sampling found Aspergillus conidia in our ICU rooms. Laminar air flow unit and high-efficiency particulate air filters were installed during the outbreak which may have decrease invasive Aspergillus nosocomial infection [37] and explain the lower rate of invasive aspergillosis in C-ARDS in our work, but half of invasive aspergillosis cases were diagnosed at ICU admission. The lack of galactomannan in BAL performed in C-ARDS patients may underestimate invasive aspergillosis using IAPA case definition and modified AspICU algorithm, but putative aspergillosis was less common in C-ARDS using crude AspICU criteria. Bartoletti et al. [38] found a higher incidence of invasive pulmonary aspergillosis in COVID-19 patients mostly treated with high doses corticosteroids and tocilizumab. In our cohort, NC-ARDS patients were often immunosuppressed with more blood malignancies and corticosteroid, which are known risk factors for invasive aspergillosis in patients with ARDS or severe influenza [16, 36]. These underlying diseases may at least in part explain the higher risk of invasive aspergillosis in this group.

The strengths of our study come from the detailed description of VAP and the use of competing risk models (cumulative incidence function of Gray model) to properly estimate the effect of COVID-19 on VAP risk, after adjustment on death and ventilator weaning as competing events.

Our study has some limitations. First, due to its monocentric design, our results may not be applicable on other centers. The risks of VAP may vary between centers in parallel with the variation in infection prevention measures and health crisis preparedness strategies. Clinical wards air and contact surfaces, sources of pathogenic fungi, may highly vary between ICUs, especially during a crisis. Second, the study period used to recruit the cohort was long (11 years), which is a major limitation. The lower incidence of ARDS in patients with non-COVID viral pneumonia may be ascribable to a lower virulence and transmissibility of non-SARS-CoV-2 respiratory viruses as compared to SARS-CoV-2 [39]. However, our management protocol for ARDS did not significantly change during the study period. Third, we found different baseline characteristics between groups. C-ARDS patients were mostly males [40, 41], a known risk factor of VAP; however, the latter association with C-ARDS persisted in the multivariable analysis and in matched analysis. Fourth, it might be difficult to interpret chest X-ray because of preexisting parenchymal injury in ARDS patients [42]. The microbiological investigation on lower respiratory tract samples is currently the main diagnostic tool of VAP in C-ARDS and NC-ARDS [43]. The higher respiratory sampling in C-ARDS patients may have theoretically contributed to an overestimation of the VAP frequency in this group, but VAP was first clinically suspected if any of its classical criteria happened and sampling was then performed to confirm VAP. We cannot exclude that respiratory deterioration-labeled VAP was to some extent relative to progression of COVID-19 with a bystander positive bacterial culture.