Introduction

SARS-CoV-2 infection (COVID-19) is currently the leading cause of acute respiratory failure [1]. About 5% of patients develop acute respiratory distress syndrome (ARDS) requiring admission to the intensive care unit (ICU) and invasive mechanical ventilation (MV) [2]. Depending on ARDS severity, ICU mortality has ranged from 30 to 50% [3].

Ventilator-associated pneumonia (VAP) is a complication of prolonged MV whose adverse consequences include worsening hypoxemia, increases in MV duration and ICU stay length, antibiotic overuse, and higher mortality. In a prospective observational study, VAP caused 19.6% of all ICU deaths and was a contributory factor in 43.9% of deaths [4]. VAP has been reported in 28% of patients with bacterial pneumonia [5] and 43% with viral pneumonia (including cases due to H1N1) [6] requiring MV. The proportion of VAP in patients with COVID-19 ARDS was 52%, and VAP was more common in patients with COVID-19 than in those with other causes of pneumonia [7].

In severe COVID-19, a massive inflammatory response with a cytokine storm causes diffuse alveolar damage and inflammatory infiltrates [8]. Corticosteroid therapy, which suppresses inflammation, decreases 28-day mortality. Thus, in the open-label randomized RECOVERY trial, among patients receiving MV, significantly fewer died with vs. without dexamethasone (29.3% vs. 41.4%; rate ratio 0.64; 95% confidence interval [95% CI] 0.51–0.81) [9]. The double-blind CODEX trial in patients with moderate or severe COVID-19 ARDS showed mean numbers of MV-free days of 6.6 with dexamethasone and 4.0 with the placebo [10]. Dexamethasone in a dosage of 6 mg/day for 5–10 days has been recommended, with no details on the optimal time of initiation [11]. In a study of patients with H1N1 ARDS, by multivariate analysis, early corticosteroid therapy increased the risk of VAP but not the risk of death [12]. In COVID-19 ARDS, data on potential associations between early corticosteroid therapy and the risk of VAP are conflicting [13, 14].

The objective of this retrospective case–control study was to determine whether early corticosteroid therapy was associated with the incidence of VAP in patients receiving MV in the ICU for COVID-19 ARDS.

Methods

Study design and participants

We retrospectively identified adults who were admitted to any of 15 French ICUs in 2020 and required MV for COVID-19 ARDS. COVID-19 was confirmed by a positive reverse transcriptase-polymerase chain reaction for SARS-CoV-2 in upper and/or lower respiratory tract samples. We defined early corticosteroid therapy as the administration of systemic corticosteroids before or within 24 h after ICU admission.

Outcomes

The main study outcome was the cumulative VAP incidence in groups with (cases) vs. without (controls) early corticosteroid therapy. Secondary outcomes were 90-day mortality, MV duration, other organ dysfunctions, and VAP characteristics.

Data collection

At each center, the study investigator collected the data in Table 1 on standardized forms. Details are available in Additional file 1.

Table 1 Demographics and clinical features at ICU admission and during the ICU stay of patients with COVID-19 requiring invasive mechanical ventilation
Full size table

Diagnosis and management of ventilator-associated pneumonia (VAP)

The criteria used to diagnose VAP are detailed in Additional file 1. Local protocols to minimize the risk of VAP included oral rather than nasal intubation whenever possible; head-of-bed elevation to 30°–45°; periodic suctioning system drainage; use of a new ventilator circuit for each patient; circuit changes only when the circuit was soiled or damaged, not routinely; replacement of the heat–moisture exchanger every 5–7 days or when soiled or malfunctioning; and daily oral hygiene [15].

Statistical analysis

Categorical variables were described as counts (percentages) and continuous variables as mean ± SD. Normality was checked by visual inspection. For baseline characteristics, percentages were compared by applying the Chi-square test or Fisher’s exact test, as appropriate. Means were compared between groups using Student’s t-test or the Mann–Whitney U test. Two-tailed P values smaller than 0.05 were considered significant.

The incidence of VAP was compared between the groups with vs. without early corticosteroid therapy using the Fine and Gray competitive risk survival model, with death and extubation as the competing events [16]. To account for baseline differences in covariates, we then repeated the Fine and Gray analysis with adjustments on age, body mass index, and Charlson’s Comorbidity Index. To compare 90-day mortality between the groups with vs. without VAP, we chose the frailty model for survival data to account for the time dependency of VAP occurrence, with adjustment on age and Sequential Organ Failure Assessment (SOFA) score [17]. Antibiotic resistance of microorganisms causing VAP was rated from 0 to 3 (no resistance, MDR, XDR, PDR) and compared between groups with vs. without early corticosteroid therapy using ordinal logistic regression.

The statistical analysis was performed using SAS software (version 9.4, Cary, NC).

Results

Between February 1 and December 31, 2020, 1124 patients were admitted to one of the 15 participating ICUs for COVID-19 ARDS, including 670 patients who required MV for longer than 48 h. Of these 670 patients, 369 did and 301 did not receive early corticosteroid therapy (Fig. 1).

Fig. 1
figure 1

Patient flowchart

Full size image

Demographics and characteristics

Table 1 reports the main patient features. The median time from COVID-19 symptom onset to ICU admission was 8 ± 4 days. Overall, the patients had severe hypoxemia, with a mean PaO2/FiO2 of 143 ± 67 mmHg at ICU admission.

Details of corticosteroid treatment

In the early corticosteroid group, dexamethasone was the most frequently used form (336/369, 91%), the mean prednisone-equivalent dosage was 60 ± 48 mg, and the median corticosteroid treatment duration was 10 ± 4 days. Treatment was initiated 0 (IQR: [− 1;0]) days after ICU admission. Of the 301 other patients, 53 (17.6%) received delayed corticosteroid therapy, with initiation at a mean of 13 ± 9 days after ICU admission; methylprednisolone was the most commonly used corticosteroid (19/53, 36%), the mean prednisone-equivalent dose was 171 ± 207 mg, and the mean duration was 9 ± 11 days. Adjuvant immunomodulatory drugs such as IL-6 antagonists were used in 10 (3%) patients with vs. none of the patients without early corticosteroid therapy.

Ventilator-associated pneumonia (VAP) (Table 2)

Table 2 Characteristics of ventilator-associated pneumonia (VAP in the groups with vs. without early corticosteroid therapy
Full size table

Of 660 respiratory tract samples, 466 produced positive bacterial cultures. The mean number of samples per patient was 2.3 ± 2.5 in the group with vs. 1.9 ± 1.9 in the group without early corticosteroids (P = 0.017). Of the 670 patients, 349 (52%) experienced at least one VAP episode. Mean time from intubation to the first episode was 9.9 ± 7.7 days. The proportions of patients with a first VAP episode who experienced a second episode were similar in the two groups. The bacterial antibiotic resistance profiles for the first VAP episode were also similar, with an odds ratio of 0.942 (95% confidence interval [95% CI] 0.58–1.52; P = 0.80) (Additional file 1).

Bloodstream infection (BSI) diagnosed during VAP was not significantly more common in the early corticosteroid group, and neither were significant differences observed for the occurrence of abscess or pleural effusion (Table 2).

The Fine and Gray competitive risk model with death and extubation as the competing events showed that early corticosteroid treatment was significantly associated with VAP (hazard ratio [HR] 1.29; 95% CI 1.05–1.58; P = 0.016) (Fig. 2). The results were similar after adjustment on age, body mass index, and Charlson’s Comorbidity Index (HR 1.28; 95% CI 1.03–1.58; P = 0.026).

Fig. 2
figure 2

VAP probability according to time

Full size image

Bloodstream infection (BSI)

BSI diagnosed in the absence of VAP was significantly more common in the group with vs. without early corticosteroid therapy (18.7% vs. 12.7%, P = 0.03) (Additional file 1). The antibiotic resistance profiles of bacteria from blood cultures were similar in the two groups, as detailed in Additional file 1.

Mortality

Overall, 187 (27.9%) patients died in the ICU or hospital before day 90 (Table 3, Additional file 1). The frailty model with adjustment on age and SOFA score indicated that VAP was associated with higher 90-day mortality (HR 1.86; 95% CI 1.33–2.61; P = 0.0003). Each 1-year increase in age and each 1-point increase in the SOFA score was significantly associated with higher 90-day mortality (HR 1.05; 95% CI 1.03–1.07; P < 0.0001; and HR 1.15; 95% CI 1.09–1.22; P < 0.0001; respectively (Fig. 3).

Table 3 Outcomes in the groups with and without early corticosteroid therapy
Full size table
Fig. 3
figure 3

Cumulative incidence of death according to the occurrence of VAP with extubation as a competing event

Full size image

Discussion

In this large retrospective cohort, both VAP and BSI were more common in patients with vs. without early corticosteroid treatment given to treat severe COVID-19 ARDS. The risk of death by day 90 was higher in patients with VAP and was not lower in those given early corticosteroid treatment. Bacterial resistance profiles did not differ between the groups with vs. without early corticosteroid therapy.

The incidence of VAP in patients with COVID-19 has varied from 40 to 60% [13, 14, 18]. This range is higher than in patients with ARDS not related to COVID-19 [19]. Possible explanations to this high VAP incidence may include insufficient human and material resources to deal with surges of patients requiring MV for COVID-19 [20]. However, during the first COVID-19 wave in France, in an area where ICU bed numbers were sufficient and nurse–patient ratios unaltered, VAP developed at least once in 50% of patients with COVID-19-ARDS [13]. Moreover, the lymphopenia and cytokine storm caused by the SARS-CoV-2 virus may induce lung–parenchyma lesions conducive to bacterial growth [21]. The reported association linking VAP to higher 28-day mortality in COVID-19-ARDS but not in other causes of ARDS supports a role for specific SARS-CoV-2-generated lung lesions in inducing greater VAP severity [22]. We assessed infections and mortality until day 90, whereas other studies usually stopped data collection on day 28 [23, 24], notably a randomized clinical trial on the effect of corticosteroid therapy [10].

After publication of the RECOVERY trial showing lower 28-day mortality in COVID-19 patients receiving respiratory support at randomization and given dexamethasone [9], systemic corticosteroids became a major component of the management of patients with severe COVID-19 ARDS. In the CODeX randomized controlled trial in patients with moderate-to-severe COVID-19, dexamethasone increased the number of days alive without ventilation compared to standard care alone, and the proportions of patients with VAP were similar in the two groups [10]. However, data were collected only until day 28.

Several previous studies did not find a higher risk of VAP in patients given early corticosteroid therapy for COVID-19. In a single-center retrospective study of 135 patients admitted to the ICU for COVID-19, the groups with vs. without corticosteroid therapy did not differ significantly regarding the frequency of respiratory, urinary, or bloodstream infections [18]. Similarly, a retrospective comparison of the first and second COVID-19 waves in France found that, in patients who required MV for at least 48 h, dexamethasone therapy was not independently associated with VAP or BSI [25]. However, the time from intubation to VAP was shorter in the corticosteroid group, and the sample (n = 151) was five times smaller than ours. In a retrospective analysis of data collected prospectively in 70 ICUs, corticosteroid therapy within one day after ICU admission was not associated with VAP by multivariable analysis (adjusted odds ratio [aOR] 1.05; 95% CI 0.83–1.34] [23]. Other studies did find that corticosteroids were associated with VAP. A single-center retrospective study that included first-wave patients in Spain demonstrated a significant independent association of corticosteroids with VAP (aOR 3.23; 95% CI 1.78–5.97) [14]. In a prospective international study of 3777 patients, propensity-score analysis demonstrated that early dexamethasone was significantly associated with VAP (17.1% vs. 13.2%; P = 0.014) [26]. The best-designed studies with the largest sample sizes would thus seem to support an association linking corticosteroid therapy to VAP. The apparent discrepancies across studies may be related to differences in the timing and dosage of corticosteroid therapy and in the mix of COVID-19 phenotypes as defined, for instance, by age and inflammation severity. A study of a prospective database showed that early corticosteroid therapy was associated with lower mortality in patients aged 60 years or older but with higher mortality in those who were younger than 60 years and had laboratory evidence of inflammation [27]. A cluster analysis of prospective data from 63 ICUs in Spain identified three phenotypes that differed regarding variables including age, inflammation severity, and ICU mortality [28]. Another cohort study, done in France and Belgium, also identified three clinical phenotypes associated with different outcomes [29]. Finally, in a prospective cohort investigated using latent class analysis of respiratory parameters, no phenotypes were differentiated at baseline, whereas two phenotypes were identified when data over the first 4 days of MV were considered [30]. Further work to identify phenotypes and the optimal treatments for each is needed [31].

In our study, ICU mortality was not significantly different with vs. without corticosteroid therapy. In a retrospective multicenter study of prospectively collected data in patients similar to ours, mortality was lower with early corticosteroid therapy but higher with corticosteroids started after 17 ICU days [23]. Similarly, corticosteroid therapy started at least 13 days after symptom onset (a median of 11 [8,9,10,11,12,13,14,15,16] days after intubation) was not associated with lower mortality in a multicenter retrospective study [32]. However, both studies were done early in the pandemic, at a time when corticosteroids were used in higher dosages in compliance with recommendations for the treatment of ARDS [33].

The main limitation of our study is the retrospective design. However, we used a strong statistical strategy to compare patients with vs. without corticosteroid therapy. The diagnosis of VAP, known to be challenging given the nonspecific manifestations, was not validated by an adjudication committee. Nonetheless, European Centre for Disease Prevention and Control criteria for VAP were applied [34]. Although these criteria are not fully sensitive or specific, no better diagnostic method exists to date [35]. Importantly, corticosteroid treatment was not allocated at random, and significant baseline differences existed between the two groups. Thus, the group given early corticosteroid therapy was characterized by greater disease severity, a higher frequency of BSI, and a higher proportion of patients managed with high-flow nasal oxygen and delayed MV. These differences suggest that corticosteroids were given to the sickest patients, resulting in selection and misclassification bias. Finally, our study took place before the circulation of variants of the Wuhan-Hu-1 virus. The response to corticosteroids and other treatments may have changed with the emergence of variants.

Conclusion

VAP, as well as BSI, was more common among patients given early corticosteroid therapy to treat COVID-19 ARDS requiring MV. 90-day mortality was not associated with early corticosteroid therapy but was higher in patients with VAP. Further studies should seek to identify patient subgroups likely to benefit from early corticosteroid therapy in the current era of increasing IL-6 antagonist use and emerging variants. Also, the optimal time of corticosteroid initiation should be determined.