Risk factors for mortality in a multicenter cohort of mechanically ventilated COVID-19 patients in Belgium.

Background: Considering the high mortality rate of severe Covid-19 patients, it is necessary to identify prognostic factors and therapies which could be valuable in this setting. Methods: The method consisted in a multicentric retrospective analysis in all consecutive Covid-19 patients admitted to intensive care unit (ICU) and mechanically ventilated for more than 24 hours from March 1 to April 25, 2020. Admission date, age, sex, body mass index, underlying conditions, treatments, physiological values, use of vasopressors, renal replacement therapy and extracorporeal membrane oxygenation, duration of mechanical ventilation, length of ICU stay, ICU and ventilator-free days at day 42 were collected. Primary outcome was survival. Simple and multiple time-dependent Cox regression models were used to assess the effects of factors on survival. Results: Out of 2003 patients hospitalized for SARS-CoV-2, 361 were admitted to the participating ICUs, 257 were ventilated for more than 24 hours and 247 were included in the study. The length of stay in ICU was 21 (12-32) days and the mortality rate was 45%. Using multiple regression, risk factors for mortality were age, high serum creatinine value, low mean arterial pressure, low lymphocytes count on day 0 and the absence of corticosteroid therapy during the rst week of mechanical ventilation. The mortality rate of the patients who received corticosteroids was 34% and 48% for patients who did not (p = 0.01). Conclusion: In this multicenter cohort, the mortality of patients with SARS-CoV-2 pneumonia treated with mechanical ventilation was high. The risk factors for mortality included age, renal and circulatory dysfunction, lymphopenia and the absence of corticosteroid therapy during the rst week of mechanical ventilation. for least 24 The patients were studied for their entire hospital stay or for a minimum of 42 days in case of prolonged hospital stay. The following data were retrospectively collected and entered in a clinical report form transmitted to the participating ICUs. On ICU admission, the admission date, age, gender, Body Mass Index (BMI), SARS-CoV-2 viral load (Cycles Threshold, Gene E), quantitative chest CT-scan analysis and underlying conditions (smoking, chronic kidney disease, diabetes, hypertension) were collected. During the ICU stay, diuresis, mean arterial pressure, PaO 2 , FiO 2 , PaO 2 /FiO 2 ratio, and laboratory blood values (creatinine, bilirubin, ferritin, C-reactive protein (CRP), D-dimer, platelets count and lymphocytes count) were collected on days 0 and 7. Sequential organ failure assessment (SOFA) score and Glasgow coma score were calculated on days 0 and 7 after ICU admission. Use of drugs oriented against COVID-19 (hydroxychloroquine, azithromycin, corticosteroid and monoclonal antibodies directed against interleukin (IL)-1 or 6), use of vasopressors, renal replacement therapy (RRT), extra-corporeal membrane oxygenation (ECMO), duration of mechanical ventilation, length of ICU stay, and ICU and ventilator-free days at day 42 were collected. Patients were with corticosteroid for COVID-19 if the corticosteroid therapy (methylprednisolone started between day 0 and day 7 after ICU In corticosteroids


Background
In late 2019, the virus responsible for COVID-19 was identi ed and called SARS-CoV-2. In China, 5% of COVID-19 patients were admitted in intensive care unit (ICU), 2,3% were ventilated and 1,4% died [1]. In early 2020, Covid-19 quickly spread in Europe and the rst 9 cases were described in Belgium on February 4, 2020. In Belgium, the highest hospitalization rate was 629 patients a day on March 28 while the highest number of patients (n = 1285) in ICU was reached on April 8. The highest mortality rate was 340 patients a day on April 12 and on April 25, 4355 patients were still hospitalized including 970 in ICU.
Considering the high mortality rate of severe COVID-19 patients, it is necessary to identify prognostic factors and therapies which could be valuable in this setting [1]. Since the start of the SARS-CoV-2 pandemic, the need for trials to assess the bene t of antiviral treatment, anti-cytokine drugs, convalescent plasma and hydroxychloroquine was advocated by the world health organization. However, evidence of the e cacy of such strategies is still lacking. Hydroxychloroquine was tested because of its capacity to increase endosomial pH and interfere with the glycosylation of cellular receptor of SARS-CoV-2 [2,3] but no clinical bene t could be shown in a randomized controlled trial [4]. The main hypothesis supporting the use of anti-in ammatory drugs is the concept of "cytokine storm", referring to an exacerbation of the host response mimicking what is observed after CAR-T cells administration for hematologic malignancies, and supposedly responsible for mortality of Covid-19 despite apparent clearance of the virus [5,6]. Among anti-in ammatory drugs, steroids are frequently used because they are cheap and easily available around the world, and recommended use in both septic shock [7] and acute respiratory distress syndrome (ARDS) [8]. The Surviving Sepsis Campaign guidelines were recently adapted to COVID-19, despite the absence of evidence in this particular setting [9]. Eventually, a prospective beforeafter study suggested a bene cial effect of methylprednisolone use in patients with moderate to severe COVID-19 [10]. Other treatments, either antivirals, convalescent plasma or anti-cytokines are still marginally used because of their cost and/or low availability.
Until now, the COVID-19 patients treated with tracheal intubation and mechanical ventilation have been reported to have short-term mortality rates between 28 and 81%, depending on the patients characteristics [11][12][13][14][15]. Two months after the dissemination of COVID-19 in Belgium, we set up a clinical database of the patients admitted to 12 ICUs in Wallonia and requiring mechanical ventilation to describe these patients'course and the factors associated with their outcomes.

Study design and participants
This retrospective observational cohort study was performed in the following 12  The creation of the database was planned during the study period and collection started after approval of the Ethics Committee.
The study protocol was approved by the Ethics Committee of the University Hospital of Liege on May 29, 2020. Due to the retrospective nature of the data collected, in accordance with Belgian law, no consent from the patient was required.

Procedures
We identi ed all consecutive adult patients admitted to the participating ICUs for acute respiratory failure due to SARS-CoV-2 pneumonia (diagnosed with a chest tomodensitometry suggestive of COVID-19 and with a positive polymerase chain reaction (PCR) for SARS-CoV-2 in nasal swab), and mechanically ventilated for at least 24 hours from March 1 to April 25, 2020. The patients were studied for their entire hospital stay or for a minimum of 42 days in case of prolonged hospital stay. The following data were retrospectively collected and entered in a clinical report form transmitted to the participating ICUs. On ICU admission, the admission date, age, gender, Body Mass Index (BMI), SARS-CoV-2 viral load (Cycles Threshold, Gene E), quantitative chest CT-scan analysis and underlying conditions (smoking, chronic kidney disease, diabetes, hypertension) were collected. During the ICU stay, diuresis, mean arterial pressure, PaO 2 , FiO 2 , PaO 2 /FiO 2 ratio, and laboratory blood values (creatinine, bilirubin, ferritin, C-reactive protein (CRP), D-dimer, platelets count and lymphocytes count) were collected on days 0 and 7. Sequential organ failure assessment (SOFA) score and Glasgow coma score were calculated on days 0 and 7 after ICU admission. Use of drugs oriented against COVID-19 (hydroxychloroquine, azithromycin, corticosteroid and monoclonal antibodies directed against interleukin (IL)-1 or 6), use of vasopressors, renal replacement therapy (RRT), extra-corporeal membrane oxygenation (ECMO), duration of mechanical ventilation, length of ICU stay, and ICU and ventilator-free days at day 42 were collected. Patients were de ned as treated with corticosteroid for COVID-19 if the corticosteroid therapy (methylprednisolone or dexamethasone) was started between day 0 and day 7 after ICU admission. In particular, corticosteroids use at a later stage of the stay, either for rescue therapy of ARDS or for prevention of extubation stridor was not considered.

Outcomes
The primary outcome was survival during the hospital stay. Secondary outcomes included use of vasopressors, RRT or ECMO, ICU and ventilator-free days at day 42, and evolution of the main physiological values between days 0 and 7.

Statistical methods
Quantitative variables are reported as median and interquartile range (Q1-Q3). Categorical variables are expressed as number (%).
A Kaplan-Meier plot was used to describe survival rate.
Simple and multiple time-dependent Cox regression models were used to assess the effects of factors on survival. All the variables which had a p-value lower than the critical level of 0.1 were selected for the multivariate model.
A p-value < 0.05 was considered signi cant. Missing data were not replaced. Calculations were done using SAS (version 9.4) and R (version 3.6.2) softwares.

Results
From March 1 to April 25, 2020, 2003 adult patients diagnosed with SARS-CoV-2 pneumonia were hospitalized and 361 were admitted to the 12 participating ICUs for acute respiratory failure. Of these, 257 patients were mechanically ventilated for more than 24 hours and 247 included in the data base ( Fig. 1). Because of too many missing data, ten patients were not included because they were transferred to another hospital (n = 5) or from another hospital (n = 5) during their ICU stay.
As opposed to survivors, non-survivors were older and suffered more often from chronic kidney disease (Table 1). On Day 0, non-survivors had higher SOFA score and serum creatinine value, lower mean arterial pressure and diuresis. Non-survivors were also more often treated with norepinephrine during the rst day in ICU. Survivors received more frequently corticosteroid and hydroxychloroquine.
Evolution for SOFA score, PaO 2 /FiO 2 ratio, platelets count, bilirubin value, creatinine value, Glasgow coma score, ferritin value, CRP value, D-dimer value, and lymphocytosis during the rst week of ICU stay are displayed on Table 2 (electronic supplement Table 3). In non-survivors, SOFA score, serum creatinine and CRP values increased, and lymphocytes count decreased signi cantly. Platelets count increased signi cantly less in non-survivors than in survivors.
Length of stay in hospital was 26  days, length of stay in ICU was 21    Table 4 and Fig. 2).
Using multiple regression, the predictors of mortality were age, creatinine value, mean arterial pressure lower than 70 mmHg, and lymphocytes count on day 0 and absence of corticosteroid use (Table 1).
Survival probability was 75% at 23 days in patients who received corticosteroids versus at 10 days for those who did not (Fig. 2).

Discussion
In this retrospective multicenter study of 247 consecutive patients with SARS-CoV-2 pneumonia mechanically ventilated more than 24 hours, we observed that mortality was as high as 45% and that the factors associated with mortality were age, high creatinine value, low mean arterial pressure, low lymphocytes count and absence of corticosteroid use.
These series assessed 37 to 1150 mechanically ventilated patients and reported 28 to 81% mortality rates. However, these studies had only short-term follow-up, while we could study our patients for their entire hospital stay or for a minimum of 42 days in case of prolonged hospital stay.
The therapies used in our patients were consistent with those used during this phase of the pandemic.
Regular use of hydroxychloroquine was based on initial observational studies [16], but its e cacy was not con rmed in subsequent randomized controlled trials [4,17]. No recommendation prompted adjunctive treatments as no speci c treatment had been proven to decrease mortality of Covid-19 critically ill patients on mechanical ventilation [18]. The "Recovery" trial recently demonstrated a bene cial effect of steroids in the most severe patients, but was going on at that time and is still being under reviewing assessment [19].
The outcomes of our patients indicate that death may occur at a late stage of the ICU stay. In fact, the length of organ support and ICU stay were particularly high in survivors. This may be explained by the therapeutic obstinacy that has been commonly developed because the disease was not known and the potential for new treatments was real [20]. This also invites to coordinate long-term physical and psychological evaluation and rehabilitation in survivors of severe SARS-CoV-2 pneumonia [21].
The risk factors for mortality that we found included age, severity at ICU admission, as assessed by low mean arterial pressure or high creatinine values, lower lymphocyte count, and absence of corticosteroid therapy before day 7. When associated with acute respiratory failure, circulatory shock and acute renal failure are indicative of multiple organ dysfunction due to the severity of the disease [22]. Acute renal failure has been recognized as a major issue in COVID-19, and may re ect direct invasion of SARSCoV-2 into kidney tissue in most severe cases [23]. A reduction and exhaustion of the T lymphocytes has been observed in COVID-19, and particularly marked in patients requiring ICU admission [24]. The bene cial role of corticosteroids is consistent with what was recently anticipated by Fadel et al. in a before-after study [10] and observed in the "Recovery" multicenter randomized controlled trial performed in the United Kingdom [19], indicating that corticosteroid use was associated with a better prognosis mostly in those patients with severe disease. In Fadel's study, a short course of methylprednisolone was associated with a reduction of the primary composite endpoint from 54 to 35%. Early released results of the Recovery trial showed that day 28 mortality decreased from 40 to 28% in patients receiving dexamethasone [19]. We also observed that mortality decreased from 48-34% in ICU patients on mechanical ventilation who received corticosteroid therapy.
Corticosteroids may act by controlling the intensity of the immune response to SARS-CoV-2 infection. At the start of the SARS-CoV-2 pandemic, based on prior literature about other respiratory viruses, several authors argued against the use of steroids outside research trials in COVID-19 patients because their use was associated with delay in the clearance of viral RNA from respiratory tract [25]. On the opposite, steroids have been shown to improve outcome in ARDS secondary to Pneumocystis Carinii pneumonia [26] and recently in ARDS of various causes [8]. Additionally, despite concerns that corticosteroids increase the risk of secondary infections, their use was also shown bene cial as an adjunctive treatment of septic shock [27]. The response to SARS-CoV-2 infection has recently been described as "cytokine storm" [28] which is de ned as an excessive production of immune mediators. This excessive production of cytokines is a rationale to assess the e cacy of IL-6 and/or IL-1 blockade [29,30].
Our study has several limitations. First, it was a retrospective analysis. This was due to the rapidity of occurrence of the COVID-19 in our country. This mode of spread precluded the possibility to homogenize the therapy protocols and data assessment. The retrospective nature of our study was attenuated by the fact that the study was decided while most patients were still in the hospital, resulting in very few missing data, and treated with conventional intensive care guidelines, in accordance with international guidelines. Second, the study was performed in one part of the country and our results may not be applicable to other regions. This was a multicenter study and we estimate that our hospitals did not suffer shortage of human or material resources in signi cantly different proportions than others Western countries, which could largely in uence the mortality rate of the cohort. Third, at the time of the study, there was no The rst draft of the manuscript was written by Bernard Lambermont 1 and all authors commented on previous versions of the manuscript. All authors read and approved the nal manuscript.

Funding None
Availability of data and material: The data used and/or analyzed in the present study are available from the corresponding author on reasonable request.
coercive guidelines for speci c treatments or protocols. This may have led to different practices than in other regions of the world and reduced external validity of our results.

Conclusion
Retrospectively analyzing the data of a multicenter cohort, we observed that mortality of patients with SARS-CoV-2 pneumonia treated with mechanical ventilation was as high as 45% and median survival time was 82 days. In this series, the risk factors for mortality included age, renal and circulatory dysfunction, lymphopenia and the absence of corticosteroid use during the rst week of mechanical ventilation. These ndings support the continuing research for effective treatments in these particularly severe patients, as well as applying effective follow-up programs for those patients surviving intensive care for COVID-19.

Figure 1
Flowchart of the patients admitted to the hospital for SARS-CoV-2 pneumonia. Abbreviation: MV: mechanical ventilation. Ten patients ventilated for more than 24 hours and transferred to another hospital (n = 5) or from another hospital (n=5) during their ICU stay were not included in the study because of too many missing data.