Introduction

The novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) can lead to a severe respiratory tract infection (coronavirus disease 2019 (COVID-19)) and hit the world with multiple pandemic waves from December 2020. During the first surge of COVID-19, studies reported that around 80% of patients admitted to hospital for COVID-19 would require oxygen support [1, 2] and a high rate of them an invasive mechanical ventilation (IMV) due to acute respiratory distress syndrome (ARDS) leading to high mortality rates [3, 4]. IMV during ARDS exposes patients to severe complications, such as ventilator-associated pneumonia (VAP).

Several studies have recently described the high incidence of ventilator-associated lower respiratory tract infection (VA-LRTI) in COVID-19 patients ranging from 30 to 84% [4,5,6,7]. Previous studies demonstrated a significantly higher incidence of VA-LRTI and notably VAP in SARS-CoV-2 pneumonia patients versus non-SARS-CoV-2 pneumonia patients [5, 8,9,10]. High rates of ARDS, alveolar inflammation, prolonged IMV, lung microbiota alteration, COVID-19-related specific lesions, neuromuscular blocking and immunosuppressive agent use could explain this high rate of VAP in SARS-CoV-2 pneumonia patients [11, 12].

The positive results of the randomized controlled multicenter trial RECOVERY [13] and its further confirmation in large meta-analysis [14] have placed dexamethasone as the first line agent for treating hospitalized COVID-19 patients with a significantly improved 28-day survival, especially in the subgroup of patients invasively ventilated. Yet, the impact of such treatment on the incidence of VAP in COVID-19 patients is still a matter for debate, as available data are scarce and conflicting [3, 15,16,17,18,19].

Hence, we sought to determine the relationship between adjuvant corticosteroid treatment and the incidence of VAP in a large cohort of COVID-19 patients invasively ventilated for more than 48 h, during the first surge of the SARS-CoV-2 pandemic. Our hypothesis was that VAP incidence would be higher in corticosteroid-treated versus non-treated COVID-19 patients.

The primary objective was to compare the incidence of VAP in patients receiving or not adjuvant corticosteroid treatment. Secondary objectives were to determine the relationship between adjuvant corticosteroid treatment and 28-day mortality, duration of mechanical ventilation and ICU length of stay, based on the occurrence of VAP. We also evaluated microbiology of VAP in both groups of patients.

Methods

Study design and population

This is a planned ancillary analysis of the coVAPid study, which design was previously described [8]. Briefly, it was a multicenter retrospective observational cohort conducted in 36 ICUs in Europe (28 centers in France, 3 in Spain, 3 in Greece, 1 in Portugal and 1 in Ireland) aimed to determine the relationship between SARS-CoV-2 pneumonia as compared to influenza pneumonia or no viral infection and the incidence of VA-LRTI. The present analysis was restricted to the population of SARS-CoV-2 pneumonia. Patients were 18-year-old or above and received IMV for more than 48 h. Patients were excluded if data related to corticosteroid treatment were not available.

Patients were consecutively included in each center starting from the beginning of the COVID-19 pandemic if they had a positive polymerase chain reaction (PCR) test for the SARS-CoV-2 virus on nasopharyngeal swab or respiratory secretions.

The Ethics Committee and Institutional Review Boards approved the coVAPid study protocol (Comité de Protection des Personnes Ouest VI; approved by April 14, 2020; registration number RIPH:20.04.09.60039) as minimal-risk research using data collected for routine clinical practice. The requirement for informed consent was waived, and patients or their relative received information about the study and were given the possibility to refuse the use of their personal data. The coVAPid database was registered into the “Commission Nationale de l’Informatique et des Libertés” in accordance with the French law. The coVAPid study was registered at ClinicalTrials.gov, number NCT04359793.

Demographic characteristics, severity scores and comorbidities, past medical history and ongoing treatments, prior hospitalization and exposition to antibiotics (during the past three months) were collected for each patient at baseline. During hospitalization, clinical, biological and radiological findings were recorded as well as the different treatments received during ICU stay (i.e., prone positioning, extra corporeal membrane oxygenation (ECMO)). Antibiotic and corticosteroid use was recorded as type, time of initiation, duration of exposure before occurrence of VAP, maximum daily dose expressed as prednisone equivalent dose (for corticosteroid treatment).

Objectives and outcomes

Our primary objective was to determine the impact of corticosteroid use on the incidence of VAP in SARS-CoV-2 pneumonia patients. The primary outcome was the occurrence of VAP [8, 20]. Briefly, new or progressive infiltrates on chest X-rays and two of the following criteria were needed for the diagnosis of VAP: (1) leucocyte count greater than 12,000 cells per μL or less than 4000 cells per μL, (2) hyperthermia above 38.5 °C or hypothermia under 36.5 °C, (3) purulent tracheal secretions. Microbiological confirmation was required to confirm the diagnosis, with at least 105 colony-forming units (CFUs) per mL for endotracheal aspirates or 104 CFU per mL for bronchoalveolar lavage. VAP episodes were prospectively recorded and confirmed by two distinct physicians as well as the chest X-rays readings. Only first VAP episodes were taken into account.

The secondary objective was to determine the impact of adjuvant corticosteroid treatment on 28-day mortality, MV duration and ICU length of stay.

Statistical analysis

Categorical variables are reported as absolute number and percentage, whereas continuous variables are expressed as median with interquartile range (25th–75th percentile). Patient characteristics at ICU admission and during ICU stay were described according to the use or not of corticosteroids without statistical comparisons. The 28-day cumulative incidence of VAP, extubation alive and ICU discharge alive were estimated using the Kalbfleisch and Prentice method [21], considering extubation within 28-day (dead or alive), death under MV, or death during ICU as a competing events. The 28-day cumulative incidence of all-cause mortality was estimated using the Kaplan–Meier method.

The association between corticosteroids and VAP was assessed using Cox regression with cause-specific hazard function considering initiation of corticosteroid treatment as a time-varying covariate. In this model, extubation within 28 days (dead or alive) without VAP was considered as a competing event. The corticosteroid time-varying variable was coded 0 during the period before the start of the treatment and 1 from the day of initiation of corticosteroids until the occurrence of VAP or MV withdrawal or death within the 28-day period of follow-up. The cause-specific hazard ratios (cHR) for corticosteroid vs. non-corticosteroid exposure were calculated as effect size. Hazard assumption proportionality of corticosteroids during follow-up was assessed by using the scaled Schoenfeld residuals plots. Since the assumption was not satisfied, the corticosteroids effect was modeled using time-dependent coefficient by including corticosteroids and corticosteroids*time interaction terms as covariates into cause-specific Cox’s models and the overall corticosteroids effect was assessed by the likelihood ratio test.

Similarly, we investigated the association of corticosteroids with patient’s outcomes censored at day-28 (overall survival, MV duration, length of ICU stay) by using a Cox’s regression model with cause-specific hazard for MV duration (considering as event of interest extubation alive and death under MV as competing event) and for length of ICU stay (considering as event of interest ICU discharge alive, and death in ICU as competing event) [22]. Since there was no deviation on proportionality assumption, the corticosteroid effect was only modeled as a time-varying covariate without time-dependent coefficients.

In addition, we investigated the effect of the occurrence of VAP on the association between corticosteroid use and these outcomes by including in the models the VAP covariate and the corticosteroids x VAP interaction (both treated as time-varying covariates).

All associations were further investigated after adjustment for pre-specified confounders already known to be associated with VAP and patients’ outcomes (age, gender, BMI, SAPS II, MacCabe classification, immunosuppression, recent hospitalization, recent antibiotics, shock, ARDS and cardiac arrest). To avoid case-deletion in multivariate analyses due to presence of missing data in covariates, multivariable Cox’s models were performed after handling missing data by using a multiple imputation procedure [23]. Specifically, imputation of missing values was performed using a regression-switching approach (chained equations with m = 20 obtained) under the “missing at random” assumption, using all baseline characteristics (see Table 1) and outcomes, with a predictive mean matching method for quantitative variables and a logistic regression model (binary, ordinal or multinomial) for categorical variables. Estimates obtained in the different imputed data sets were combined using Rubin’s rules [24]. Sensitivity analysis on complete cases (patients without missing data on covariates) was also performed.

Table 1 Patient characteristics at ICU admission
Full size table

Statistical testing was performed with a two-tailed α level of 0.05. Data were analyzed using the SAS software package, release 9.4 (SAS Institute, Cary, NC). Statistical analysis is fully detailed in the online data supplement.

Results

From March 2020 through May 2020, 568 patients receiving IMV > 48 h for SARS-CoV-2 pneumonia were eligible for this study. Twenty-three patients (4%) were excluded due to missing data. Among the 545 included patients, 191 (35%) received corticosteroids (Fig. 1).

Fig. 1
figure 1

Study flowchart

Full size image

Patient characteristics at ICU admission

At ICU admission, the majority of patients were men with a median age of 64 year-old in corticosteroid and no-corticosteroid groups. Body mass index (BMI), SAPS II and SOFA scores, recent hospitalization (< 3 months) and recent antibiotic use (< 3 months) were comparable between both groups. In the corticosteroid group, the percentage of patients with immunosuppression was higher than in patients who did not receive corticosteroids. The main causes for ICU admission were acute respiratory failure and ARDS in both groups (Table 1).

Patient characteristics during ICU stay

Patient characteristics during ICU stay are presented in Table 2. Over 90% of patients in both groups had an antibiotic treatment during their ICU stay with a slightly longer duration of treatment in the corticosteroid group. Prone positioning was more frequent in patients who received corticosteroids than in those who did not. ECMO use was comparable in the two groups. 28-day mortality was higher in patients who received corticosteroids than in those who did not. Duration of MV and ICU length of stay were longer in the corticosteroid group than in the no-corticosteroid group.

Table 2 Patient characteristics during ICU stay
Full size table

In the corticosteroid group, patients mainly received methylprednisolone (48.1%) for a median time of 6 days (3–8 days) before occurrence of VAP. The highest daily dose of prednisone equivalent was 100 mg (50–133 mg).

Incidence of VAP and corticosteroid use

In the corticosteroid group, 74 (38.7%) patients developed at least one episode of VAP versus 138 (38.9%) patients in the no-corticosteroid group. As shown in Additional file 1: e-Fig. 1, the proportional hazard assumption for the effect of corticosteroids on the incidence of VAP, assessed by the scaled Schoenfeld residuals, revealed a time tendency indicating that this effect varied during ICU stay. By modeling the effect of corticosteroids with time-dependent coefficients, the association of corticosteroids with incidence of VAP was not significant before and after pre-specified adjustment (Table 3), with time-dependent adjusted hazard ratios (95% confidence interval) of 0.47 (0.17–1.31) at day 2, 0.95 (0.63–1.42) at day 7, 1.48 (1.01–2.16) at day 14 and 1.94 (1.09–3.46) at day 21. Similar results were found in sensitivity complete case analysis. Additional file 1: e-Table 1 depicts the association of different corticosteroids with VAP in all study patients.

Table 3 Association between corticosteroid treatment use and outcomes in all study patients
Full size table

Patient characteristics at the day of VAP and microbiological data

VAP episodes were mainly diagnosed with endotracheal aspirates in both groups. SOFA score, modified CPIS and PaO2/FiO2 ratio did not differ at the day of VAP between the two groups. Thirty-five (25%) patients had at least one VAP recurrence in the no-corticosteroids group versus 8 (11%) in the corticosteroids group. A higher percentage of patients received antibiotic treatment at the day of VAP in the corticosteroid group versus no-corticosteroid group (Table 4).

Table 4 Description of patients at the day of VAP diagnosis
Full size table

Gram-negative bacilli were the main bacteria responsible for VAP in the 2 groups (60%). Pseudomonas aeruginosa, Enterobacter spp. and Escherichia coli were the most frequently identified bacteria. A Higher proportion of VAP due to multidrug-resistant bacteria (MDR) was found in the corticosteroid-treated group (Table 5).

Table 5 Microorganisms responsible for VAP
Full size table

Impact of corticosteroid use on patient outcomes

Twenty-eight-day mortality was significantly associated with corticosteroid exposure in unadjusted analysis and after adjusting for pre-specified confounders in complete case analysis, and after multiple imputation analysis. In adjusted analysis, corticosteroid use was associated with prolonged duration of MV after multiple imputation analysis, but not with ICU length of stay (Table 3). Associations between different corticosteroids and VAP, 28-day mortality, MV duration and ICU length of stay are presented in e-Table 1

VAP occurrence did not significantly modify the relationship between corticosteroids and mortality, duration of MV, or length of ICU stay (Table 6). However, a significant association was found between corticosteroids exposure and ICU mortality in the absence of exposure to VAP. Twenty-eight-day cumulative incidence of VAP, all-cause mortality, extubation alive and ICU discharge alive are presented in Additional file 1: e-Fig. 2.

Table 6 Association between corticosteroid use and mortality, MV duration and ICU length of stay according to VAP
Full size table

Discussion

In this ancillary study of the coVAPid study, we found no significant association between adjuvant corticosteroid treatment and the incidence of VAP in a population of SARS-CoV-2 patients invasively ventilated for over 48 h. However, we found a time-varying effect of corticosteroid adjuvant therapy on the incidence of VAP over the 28-day period.

The incidence of VAP in the SARS-CoV-2 population was high, nearly half of the patients presenting at least one VAP episode during ICU stay, irrespective of corticosteroid use. This result is in line with several previous studies that reported high incidence of VAP in critically ill COVID-19 patients [3, 8,9,10, 17].

Following the RECOVERY trial, corticosteroid adjuvant use dramatically increased in critically ill COVID-19 patients, although only 15% patients were invasively ventilated in this trial. In invasively ventilated COVID-19 patients, high rates of ARDS, prolonged ICU stay and IMV duration along with immunologic disorders and neuromuscular blocking agent use expose these patients to higher risk of developing VAP and one could question the impact of wide use of corticosteroids on the incidence of hospital-acquired infections (HAI) and VAP in this population, already at higher risk for these infections.

In COVID-19 patients, VAP is a serious complication, as our group and others demonstrated a significant association between VAP and mortality in this population [25, 26]. To date, there is no prospective interventional study that evaluated the impact of corticosteroid use on the incidence of VAP, as the RECOVERY study itself did not include it as an outcome. However, there is a growing body of evidence in the literature pointing out the increased risk of developing VAP in COVID-19 patients treated with corticosteroids [16,17,18,19, 27, 28].

A recent study by Mesland et al. [17]. reported, among 322 COVID-19 patients invasively ventilated, a significantly higher incidence for VA-LRTI in patients receiving corticosteroid adjuvant therapy compared with those who did not. Corticosteroids were found to be an independent factor associated with VA-LRTI in a multivariable adjusted analysis. Another recent study found the same association between corticosteroid use and VA-LRTI [18]. However, other studies suggested no association between corticosteroid exposure and the incidence of VA-LRTI in COVID-19 [15, 29].

In our study, we did not find a significant association between corticosteroid exposure and VAP incidence during the 28-day period of follow-up. However, our results suggest a trend toward a time-varying effect of corticosteroid exposure on the incidence of VAP. This time-varying effect increased all along the 28-day follow-up, and the association between corticosteroid use and VAP became statistically significant after the 14th day of mechanical ventilation. Several explanations could be provided for this time-varying effect. First, corticosteroid-immunosuppressive effect is a time-dependent but also a cumulative-dose-dependent mechanism [30] that may explain the late occurrence of hospital-acquired infections in treated patients. Second, sepsis-associated immunosuppression is a time-dependent process occurring in the late course of ICU stay, resulting in immunoparalysis and ICU-acquired infection [31, 32]. Finally, SARS-CoV-2-mediated immune dysfunction is associated with a progressive lymphopenia [33, 34], and a multifactorial immunosuppression culminating after day 14 [35] that exposes patients to late-onset ICU-acquired infection.

Our study reports higher 28-day mortality and longer MV duration in the group of patients who received adjuvant corticosteroid treatment, as compared to those who did not. The main reason for this discrepancy with results of randomized controlled trials [13, 36] might be the observational design of this study taking place before the RECOVERY era when corticosteroid prescriptions were at the discretion of attending physicians who selected the sickest patients to receive these treatments. Septic shock, non-resolving ARDS and intense systemic inflammation were the main reasons for corticosteroid adjuvant therapy, and this led de facto to a poorer prognosis in the corticosteroid-treated patients. Several other observational studies also failed to demonstrate this benefit from corticosteroids [17, 36].

Microorganisms responsible for VAP in our study did not differ between the corticosteroid treated and non-treated patients. In line with previous results, Gram-negative bacilli were the main bacteria found in respiratory tract samples with a high rate of Enterobacteriaceae and Pseudomonas aeruginosa. Methicillin-sensitive Staphylococcus aureus was the main Gram-positive cocci, and the rate of multidrug-resistant bacteria was also high [3, 15, 17].

To our knowledge, our study is the largest multicenter study to examine the impact of corticosteroid adjuvant therapy on the incidence of VAP in COVID-19 patients, and the first to show a varying association between corticosteroid use and the incidence of VAP. Additional strengths of our study are the large number of patients, the multicenter design and the strict definition of VAP with quantitative microbiological documentation for each event. However, some limitations should be mentioned. First, the observational retrospective design of our study taking place before the RECOVERY era is responsible for possible selection bias with disparate use of corticosteroids among centers that intentionally selected the sickest patients to receive corticosteroids, leading to poorer outcomes. Second, a small proportion of included patients (30%) received corticosteroids, which could also introduce a selection bias. Third, in our study we collected all steroid prescriptions irrespective of variations in molecule, duration and dosage. The impact of this discrepancy on patient outcomes might be real, yet no data support this in the literature.

Conclusions

We found no significant association between corticosteroid adjuvant therapy and the incidence of VAP in the population of invasively ventilated patients for a SARS-CoV-2 pneumonia. However, we found a time-varying effect of corticosteroid adjuvant therapy on the cause-specific estimated risk of developing VAP throughout patient course in ICU. Further studies are needed to clarify the role and effects of adjuvant corticosteroid treatment in the sickest COVID-19 patients. Future research should also investigate the interest of a corticosteroid-use strategy based on patient inflammatory phenotypes.