Ventilator-Associated Pneumonia in Patients with COVID-19: A Systematic Review and Meta-Analysis

The aim of this systematic review and meta-analysis was to estimate the pooled occurrence of ventilator-associated pneumonia (VAP) among patients admitted to an intensive care unit with COVID-19 and mortality of those who developed VAP. We performed a systematic search on PubMed, EMBASE and Web of Science from inception to 2nd March 2021 for nonrandomized studies specifically addressing VAP in adult patients with COVID-19 and reporting data on at least one primary outcome of interest. Random effect single-arm meta-analysis was performed for the occurrence of VAP and mortality (at the longest follow up) and ICU length of stay. Twenty studies were included in the systematic review and meta-analysis, for a total of 2611 patients with at least one episode of VAP. The pooled estimated occurrence of VAP was of 45.4% (95% C.I. 37.8–53.2%; 2611/5593 patients; I2 = 96%). The pooled estimated occurrence of mortality was 42.7% (95% C.I. 34–51.7%; 371/946 patients; I2 = 82%). The estimated summary estimated metric mean ICU LOS was 28.58 days (95% C.I. 21.4–35.8; I2 = 98%). Sensitivity analysis showed that patients with COVID-19 may have a higher risk of developing VAP than patients without COVID-19 (OR 3.24; 95% C.I. 2.2–4.7; P = 0.015; I2 = 67.7%; five studies with a comparison group).


Introduction
During the Coronavirus disease 2019 (COVID- 19) pandemic, an unprecedented number of patients were admitted to intensive care units (ICUs) with COVID-19-related severe respiratory failure and underwent invasive mechanical ventilation (IMV) [1]. Invasive mechanical ventilation, especially if prolonged, is a risk factor for the occurrence of ventilatorassociated pneumonia (VAP) [2,3]. Though there is no univocal definition of VAP, the most recent definitions suggest its identification by both radiological and clinical criteria, often combined with microbiological criteria [4][5][6][7]. The time span from the beginning of IMV to the fulfilment of the criteria is also an important component of the definitions [4][5][6], to exclude previously acquired pulmonary infections, thus not directly related to IMV [4]. Duration of IMV in patients with COVID-19 and admitted to ICU is often relatively long [1]. The clinical course of patients with the most severe form of COVID-19 in ICU may be complicated by the need for extracorporeal membrane oxygenation [8]. Other factors, such as the lung damage itself and the use of immunomodulant therapies (e.g., corticosteroids, anti-il-6 drugs) may also increase the risk of VAP in these cohorts of patients [9].
To date, the reported proportion of patients with COVID-19 who developed VAP COVID-19 has been variably reported. The main aim of this systematic review and meta-analysis was to estimate the pooled occurrence of VAP among patients with COVID-19 admitted to ICU and mortality of this patient population, to provide reliable data to clinicians caring for patients with COVID-19 undergoing IMV.

Materials and Methods
The protocol of this systematic review and meta-analysis was registered in Open Science Framework (https://osf.io/32mva).
We performed a systematic search of PubMed, EMBASE and Web of Science from inception to 2 March 2021 for nonrandomized studies, both prospective and retrospective, specifically addressing VAP in adult patients with COVID-19 and reporting data on at least one primary outcome of interest. Further surveillance searches were performed using the 'related articles' feature [10].
Primary outcomes were the occurrence of VAP and mortality at the longest available follow-up. Length of ICU stay (LOS) was an additional outcome. No language restrictions were applied to the search. Studies including less than ten patients, case reports, abstracts and not peer-reviewed articles were excluded. The search strategy included keywords as exact phrases and subject headings, according to databases syntaxes and is provided in Supplementary Material S1. All the retrieved records were independently screened from title and abstract by two authors (M.I., C.M.). The selected records were then independently reviewed from full text by the same two authors, to verify the fulfilment of the inclusion criteria. Studies were included if the screening authors agreed regarding eligibility. Disagreements at any stage were adjudicated by a third author (A.C.). The corresponding authors of the screened articles were contacted by two authors (M.I., A.C.) when questions arose regarding eligibility or data presentation at any time during this process. The reference lists of relevant articles were also searched for additional potentially relevant papers (i.e., snowballing method) by the same authors. Data extraction was performed independently by three authors (M.I., G.M., G.C.). No a priori definition of VAP was adopted, and data were extracted according to authors' definitions. When disaggregated data were presented on multiple episodes of VAP, we extracted data on patients who had developed at least one episode of VAP, to achieve the most comprehensive estimate of the outcomes. The final version of the tabulated data was validated by all the authors involved in data collection (M.I., G.M., G.C., C.M., A.C.). The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) [11] checklists are provided as Tables S1 and S2 in Supplementary Material S2.

Qualitative Analysis
Two investigators (M.I., G.M.) assessed the methodological quality of the included studies independently and in duplicate. Disagreements over the assessment were resolved by a third author (A.C.). The Methodological Index for non-Randomized Studies (MI-NORS) [12] tool was used for the qualitative assessment, due to its ability to evaluate the methodological quality of single-arm studies. The items were scored as 0 (not reported), 1 (reported but inadequate) or 2 (reported and adequate). The global ideal score was 16 for non-comparative studies and 24 for comparative studies.

Quantitative Analysis
A meta-analysis was performed in case of two or more included studies reporting data on the outcomes of interest. For the dichotomous outcomes, the summary estimates were derived from logit transformation of individual study proportions and presented along with the corresponding 95% confidence interval (C.I.), calculated using random effect. For the continuous outcome, the summary estimate was derived from one-arm metric mean and standard deviation (SD) of individual study outcome and presented along with the corresponding 95% confidence interval (C.I.), calculated using random effect. Mean and SD were calculated with appropriate formulae when not available [13]. A sensitivity analysis including articles comparing the occurrence of VAP in patients with COVID-19 to those of patients admitted to ICU without COVID-19 was performed, providing an odds ratio as a measure of risk. A sensitivity analysis including articles comparing the risk of death in COVID-19 patients with VAP to those of patients without VAP was also performed, providing an odds ratio as a measure of risk. A subgroup analysis was performed based on the number of centers (e.g., single or multicenter studies). An I-squared (I 2 ) statistical model was used to evaluate heterogeneity among the included studies. All the analyses were performed by A.C. and M.I., using Open Meta-Analyst 8 [14].

Characteristics of Included Studies and Patients
A total of 1555 records were retrieved in the comprehensive search. The full search output is available as Supplementary Material S3. After the exclusion of duplicates and not relevant records, twenty studies were included in the systematic review and metaanalysis, for a total of 2611 patients with COVID-19 who developed at least one episode of VAP [9,[15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33]. The process of inclusion and exclusion is detailed in the PRISMA flow diagram, provided in Figure 1. The characteristics of the included studies are provided in Table 1. Eighteen studies had an observational retrospective design (18/20, 90%), of which five were multicenter [9,17,22,27,32] and thirteen were performed in a single center [15,16,[19][20][21][24][25][26][28][29][30][31][32][33]. Two studies had a prospective observational design (2/20, 10%) and were multicenter. All the studies were conducted in the European Union (E.U.), except one in China [27] and one in Russia [24]. The patients evaluated in the included studies had a mean or median age ranging from 49 to 69.5 years, with a percentage of female gender ranging from 18% to 55%. All the studies were single-arm studies, except five also including patients admitted to ICU without COVID-19 as a comparison group [17,19,21,22,28]. The median duration of mechanical ventilation prior to VAP in the included patients ranged from 7 to 13 days. Details on isolated microorganisms are provided in Table 2. The detailed qualitative assessment with individual domain and overall MINORS score per study is provided as Tables S3 and S4, Supplementary Material S2. Only three of the studies reported protocol registration and only one reported information on sample calculation. These two were the most frequently downgraded domains at the qualitative assessment of the included studies.

Discussion
The main finding of this systematic review and meta-analysis was that nearly one patient with COVID-19 out of two may develop VAP during ICU stay. Previous reports showed that VAP occurs in up to 23-40% of patients admitted to ICU [2,34], with variability usually attributable to differences regarding the clinical setting (e.g., countries, staffing), the population of patients (e.g., the reasons for admission, the severity of underlying disease) [35] or the criteria used to define VAP [7]. A higher occurrence in COVID-19 patients in comparison with patients without COVID-19 was also observed in our sensitivity analysis (OR 3.24; 95% C.I. 2.2-4.7; P=0.015), although performed on unadjusted data. We can speculate that the high occurrence of VAP in patients with COVID-19 may be explained by several factors, both disease and not disease related. Patients with COVID-19 admitted to ICU are generally severely hypoxemic, with both parenchymal and microvascular lung damage [36]. Thus, they are reasonably at a high risk of developing VAP, especially if compared with cohorts with a low proportion of patients with ARDS [28] or large mixed cohorts of patients admitted to ICU [37]. The severity of lung damage in patients with COVID-19 pneumonia may lead to an increase in the rate of infections. Furthermore, patients with COVID-19 frequently needed prolonged mechanical ventilation, prone positioning [1] and received immunomodulant therapies, that may have increased their risk of developing VAP [9]. On the other hand, the ICUs may have been overcrowded in the period of peak of COVID-19 pandemic, with a possibly inadequate staffing (e.g., nurse to patient ratio) and more episodes of cross-contaminations. Healthcare workers may have also encountered difficulties related to the wearing of personal protective equipment in the setting of COVID-19-dedicated ICUs [38]. When considered together, these issues may have reduced the adherence to infection control standard and VAP prevention bundles [16]. Although in most centers broad-spectrum antibiotic therapy is administered concomitantly with the rest of the treatment received by these patients, the occurrence of VAP should be specifically considered by the clinicians, for a timely diagnosis and a targeted therapy, especially in the case of atypical presentation of illness.
The evaluation of mortality as the outcome of patients with VAP has been widely discussed as controversial, mainly due to the difference between crude and attributable mortality, relevant to distinguish the increase in risk attributable to a complication (e.g., VAP) of a primary condition (e.g., acute respiratory failure and mechanical ventilation in COVID-19) [39]. Our analysis estimated a pooled occurrence of death in patients with COVID-19 and VAP of 42.7%. These data may be relevant on their own, but no information can be derived on the impact of VAP on mortality from our data. None of the studies reported attributable mortality due to VAP. We explored this issue in a sensitivity analysis showing no significantly different risk of death in COVID-19 patients with VAP versus non VAP, although the estimate was imprecise.
Our study has strengths, such as the comprehensive search and the methodology and reporting according to PRISMA 2020 [11]. Indeed, we were able to retrieve data from corresponding authors of the included studies that were not primarily reported in the published manuscripts. This increased the comprehensiveness and internal validity of the data collection and analysis.
Nevertheless, our analysis has several limitations. First, the overall sample size used to estimate the occurrence of VAP was modest (5593 patients, of whom 2611 developed VAP) if considering the large amount of mechanically ventilated patients with COVID-19 worldwide. Second, no studies were conducted in the United States (U.S.), thus limiting the external validity of our findings, especially considering that the study event definition may vary from E.U. to U.S.-based centers. Our analysis may also have underestimated the occurrence of VAP, as the diagnosis is based on clinical suspicion, and some clinical features of a VAP are common to the condition of COVID-19 itself. Less frequent use of traditional imaging (X-ray, computerized tomography) may have also hindered the identification of such events as VAP. In addition, our aim was mainly descriptive, with no further analysis conducted on associations with potential risk factors. Moreover, we did not collect data on the adequacy of antimicrobial therapy performed or on any factor associated with the outcomes of interest (e.g., use of protocolized preventive strategies for VAP, study period coincidence to peak of infection, etc.). Our analysis also had a high statistical heterogeneity, probably due to a high variability among the studies in microbiological epidemiology, severity of patients' clinical condition, observation of VAP prevention bundles and ICU capacity in relationship with peaks of COVID-19 outbreaks. However, a strength of this analysis was to provide a reliable estimate of the occurrence of VAP in patients with COVID-19 and of their mortality.

Conclusions
Nearly half of patients with COVID-19 admitted to ICU may develop VAP, with a pooled estimate of mortality of 42.7% for COVID-19 patients who developed VAP. Data are needed from the U.S., and further studies should evaluate the attributable risk of death of VAP in patients with COVID-19.
Supplementary Materials: The following are available online: https://www.mdpi.com/article/10.3 390/antibiotics10050545/s1, Supplementary Material S1: Search strategy; Supplementary Material S2: Figure S1: Forest plots with the results of cumulative single-arm meta-analysis for the occurrence of ventilator-associated pneumonia in patients with COVID-19; Figure S2: Forest plots with the results of cumulative single-arm meta-analysis for the mortality of patients with COVID-19 and ventilator-associated pneumonia; Figure S3: Forest plot with the results of single-arm meta-analysis for intensive care unit length of stay of patients with COVID-19 and ventilator-associated pneumonia; Figure S4: Forest plot with the results of the sensitivity analysis on the mortality of patients with COVID-19 and ventilator-associated pneumonia compared to patients with COVID-19 and admitted to ICU who did not develop ventilator-associated pneumonia; Figure S5: Forest plot with the results of the subgroup analysis conducted on the occurrence of ventilator-associated pneumonia in patients with COVID-19 in multicenter studies; Figure S6: Forest plot with the results of the subgroup analysis conducted on the occurrence of ventilator-associated pneumonia in patients with COVID-19 in single center studies; Table S1: PRISMA Main Checklist; Table S2: PRISMA Abstract Checklist; Table S3: Quality assessment of studies according to MINORS score for non-comparative studies; Table S4: Quality assessment of studies according to MINORS additional score for comparative studies; Supplementary Material S3: Full search output. All the authors gave the final approval of the version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All authors have read and agreed to the published version of the manuscript.