A Tale of a Two Waves Epidemic: Characteristics and Mortality Risk Factors for COVID-19 ICU Patients in the French West Indies

We the two in of characteristics and and to evaluate risk factors associated with COVID-19 and requiring oxygen support were were treated during the rst wave with a combination of Hydroxychloroquine and Azithromycin and during the with dexamethasone and reinforced anticoagulation.


Introduction
Coronavirus disease 2019 (COVID- 19) pandemic has been, since it was rst described 1 , a eld of investigation for physicians worldwide. Since March 2020, Guadeloupe, a French Caribbean island has been affected by two large waves of COVID-19 cases. During the rst wave, (March to May 2020), the disease was poorly understood and had no effective treatment. Since then, large studies have been published 2 and have identi ed prognostic factors related to the host and the level of organ dysfunction.
Since the beginning of the outbreak, speci c SARSCOV2 therapy has been intensively searched [2][3] .
During the rst wave, we used hydroxychloroquine with azythromycin, as no therapy was clearly identi ed at that time. Standard of care has greatly evolved since and now includes steroids [3][4][5] due to preliminary data suggesting reduced 30-day mortality 2,3 . COVID-19 has also been shown to be associated with coagulopathy and a high risk of thrombosis, and several sets of guidelines suggest to keep upkeeping a su cient level of anticoagulation among SARSCOV2 infected patients [6][7][8][9][10] . New drugs are still under evaluation 6 though there is no great consensus on their use at this time. Currently, whether any therapeutic approaches have an impact on morbi-mortality remains unclear 11 . We report here our cohort of patients admitted to an intensive care unit (ICU) with COVID-19 laboratory con rmed severe pneumonia during the two main waves in 2020. We compared the two groups in terms of clinical characteristics and outcomes. We also evaluated risk factors associated with 60-day mortality in these critically ill patients.

Methods
Population selection and study design COROCARA is a single center prospective cohort study conducted in the ICU of the University Hospital in Guadeloupe. All patients over 18 years of age with COVID-19 pneumonia and hospitalized in the ICU during the two COVID-19 outbreaks from March-May ( rst wave) and August-October (second wave) were included in the study. COROCARA received approval from the ethics committee of the University Hospital in Guadeloupe. Laboratory con rmation was de ned as a positive result by real-time reverse transcriptase-polymerase chain reaction (RT-PCR) assay from either nasopharyngeal swabs or lower respiratory tract aspirates. All included patients had a laboratory con rmed diagnosis.

Data collection
All clinical and biological data were collected within the rst twenty-four hours after ICU admission. For each patient we collected the following clinical data: age, sex, body mass index (BMI), comorbidities, immunode ciency, treatment received, the Simpli ed Acute Physiology Score (SAPS) II score 12 , the Sequential Organ Failure Assessment (SOFA) 13 score, date of rst symptoms, dates of hospital and ICU admission, support devices (oxygen mask, high ow nasal cannula, or non-invasive ventilation (NIV)) at admission, or IMV. Routine laboratory data included blood cell count, electrolytes dosages, liver enzymes, blood cultures, blood gas, creatine phosphokinase (CPK), D-dimeres and troponin. When possible, each patient underwent a chest computed tomography before admission.
Syndrome and outcome ARDS was graded based on the Berlin De nition 14 for patients undergoing mechanical ventilation (invasive or non-invasive). To be comparable to other previously published studies, ARDS was only graded in patients receiving mechanical ventilation on ICU day 1. In this study, ICU-complications and organ dysfunction included acute kidney failure requiring renal replacement therapy, pulmonary embolism (proven by pulmonary CT angiography), ventilator-associated pneumonia, and cardiac arrest.
Clinical suspicion of ventilator-associated pneumonia was con rmed before antibiotics either by blind protected specimen brush growing ≥103 cfu/mL, or endotracheal aspirates growing ≥106 cfu/mL. Patient outcome was recorded at the ICU or at hospital discharge. A favourable outcome was de ned as a patient who was alive at day 60 after admission.

Treatment
Standard of care was different during the rst and second wave, due to an increased availability of data and results in the literature over time and an improving understanding of the disease. Hydroxychloroquine was initially used during the rst wave in combination with azithromycin in our center, even though its use was debated at the time. Dexamethasone (steroids) was systematically used for severe patients requiring oxygen during the second wave, and at the discretion of the physician during the rst. In the second wave, anticoagulation was reinforced, systematic screening for pulmonary embolism was performed, and NIV and high ow oxygen were used. The use of mechanical ventilation was at the discretion of the physician.

Statistical analysis
All analyses were performed using R 4.0.4 15 . Data are reported as median (interquartile range) or number (percentage). The baseline data are reported from the twenty-four hours period after ICU admission. No sample size calculations were performed. Univariate characteristics of the two cohorts ( rst and second wave) were compared using chi-square or Fisher's exact tests for categorical variables and using Student's t-test or Wilcoxon's rank-sum test for continuous variables. Kaplan-Meier overall survival curves up to day 60 were computed separately for rst wave and second wave patients, and in patients with delayed mechanical ventilation (>4 days after ICU admission) vs those mechanically ventilated starting in the rst four days after treatment administration. No imputation was performed for missing values. We furthermore ran a multivariate analysis to asses risk factors for death in patients requiring IMV. The nal set of variables to be included in the multivariate logistic model were chosen on the basis of pathophysiological interest and the requirement p < 0.2. We performed backward selection on the model, stopping when the Akaike information criterion (AIC) reached its minimum.

Patients Enrolled
Patient characteristics and their day 1 vital status are described in Table 1, Patients were majority male (n=129, 69%) with median age 64 years (54 -71). Within twenty-four rst hours, the median SAPS II was Page 5/15 34 (24 -46) and the median SOFA was 4 (3 -8). Patient characteristics were strikingly similar between the two waves for most of evaluated variables. Median respiratory rate and median Body-Mass Index were higher during the second wave (respectively, 34 vs 31/min, p=0.03, and 30.9 vs 27.2 kg/m 2 , p= 0.01). Few patients had a bacterial coinfection at admission (n=10, 6%).

Ventilatory support, adjunctive therapies, and ARDS severity comparison
Within the rst twenty-four hours, rst wave patients more often received invasive mechanical ventilation (IMV) (45% vs 39%). In the second wave, patients often received high ow oxygen or NIV (51% vs 0%) ( Table 2). However, there was no signi cant difference regarding requiring IMV at some point during ICU hospitalization between the two periods (71% vs 63% of patients (p=0.47)). Mechanical ventilationassociated therapies used for ARDS management including prone positioning (n = 85, 71% in total) and the use of a neuromuscular blockade (n = 106, 89% in total), were not signi cantly different between the two waves ( Table 2).

ICU complications and organ support in patients requiring mechanical ventilation
Ventilator-associated pneumonia was diagnosed in 50% of patients who received IMV while 25% had acute kidney failure requiring renal replacement therapy. No statistical differences were observed between the two waves for these two variables. Median length of IMV was longer among patients during the rst wave rather than in patients during the second wave (respectively 20 days vs 7, p<0.001) Patient outcomes and predictors of 60-day mortality Results of the univariate and multivariate analysis are reported in Table S1, S2 and Table 3. Non-Survivors within 60 days were older, had more comorbidities at admission than survivors (OR 95% CI: 1.65 [1.1;2.7], p=0.04) and had much higher renal and hemodynamic SOFA component scores. Time to mechanical ventilation was also associated with death within 60 days with an OR of 1.6 (95% CI 1.2 -

2.4).
Kaplan-Meier survival curves are presented in Figure 1. There were no statistical differences in mortality rates between the two periods. During the second wave, high ow oxygen and NIV were often used as rst-step therapy (51% of patients within the rst twenty-four hours), thus delaying IMV for 25% of the patients (39/156).
The study of the delay between dexamethasone and IMV revealed a subgroup of patients characterized by high case fatality rate (89%, n=16/18) with a signi cant difference (p=0.04) when comparing patients under mechanical ventilation 4 days after dexamethasone onset versus patients under mechanical ventilation less than or equal 4 days after dexamethasone onset (see Figure 1B)

Discussion
We report here a cohort of patients corresponding to the two rst waves of the COVID-19 outbreak from the ICU of the University Hospital in Guadeloupe. Despite improvement in terms of ventilatory support and treatment severe COVID-19 pneumonia continued to be grieved with a high mortality rate.
Overall mortality was 44% and was higher in the elderly and those with multiple organ dysfunction, as previously reported 2 . Unexpectedly, mortality rate were similar in both waves, even after dexamethasone became part of standard of care in the second wave for patients requiring oxygen.
These results should however be interpreted with caution due to several potential biases. First, there could be a "magnifying glass" effect for the second wave patients., In our ICU, admission criteria were tightened and only patients requiring high ow oxygen or immediate IMV were admitted at that time.
SAPS II is known to poorly predict severity of ARDS and that could explain the similarity of patients at baseline between the two waves (p = 0.87). Secondly, in our institution, second wave was characterized by > 100% COVID-19 patient bed occupancy, at contrary of the rst wave. During periods of care system overload 16 , mortality tends to be higher compared to less strained ones. This could also be one of the explanations for similar mortality rates.
Strikingly, in our center, need for IMV were similar in both waves (64% of the patients), despite the more systematic use of steroids in the second wave. Rates of IMV was nevertheless lower than previously described in other centers 2,17 .
In univariate analysis, VAP occurrence was associated with higher mortality. Its increased frequency in COVID-19 patients compared to standard ARDS patients has been previously described 18,19 . Consistent with these datas, our incidence of VAP in IMV patients was very high (50%), with no statistical differences between the two periods however the length of IMV was longer during the second wave.
Our multivariate analysis among patients under IMV with respect to survival revealed several factors of interest. As previously described 2 , older age, comorbidities and severity at admission assessed by SOFA score were highly predictive of death. The burden of these factors is well known 2,17 and has been extensively discussed elsewhere.
More interestingly, delay between admission to the ICU and IMV onset was predictive of mortality in patients. In the subgroup whose IMV began more than four days after admission to the ICU (n = 18, 10% of patients), mortality was very high (n = 16, 89%).
Several explanations for this nding are possible. First, like many physicians, we used NIV in COVID-19 patients with hypoxemia. Independent of COVID-19, poor prognosis after failure of NIV has already been described in ARDS patients 20 and with a potentially worsening of lung damage due to self-in icted lung injury 21 . In this pandemic, NIV has been recommended by other authors for COVID-19 pneumonia management 19 , but we believe that this should be done with caution with early reappraisal to avoid late IMV Secondly, besides this physiological explanation, we believe that such grim outcomes could also re ect the pulmonary brotic evolution of COVID-19 pneumonia. This pathological nding has already been described in earlier reports 22,23 . We seriously consider this hypothesis here due to the similarity of mechanical ventilation measurements in this subgroup compared to patients with idiopathic pulmonary brosis (data not shown). Larger studies are needed to con rm this result due to its potential impact.
We acknowledge several limitations to our study. First, it is monocentric. However, high standardization of COVID-19 care in our unit and results in accordance to larger cohorts advocate for its reliability. Second, the two periods were highly dissimilar in terms of strains on resources. Second wave was characterized by intense clinical activity and overloaded care system. We cannot rule out a negative effect on the survival of second wave patients.

Conclusions
In this study of 187 critically ill patients with laboratory-con rmed COVID-19 admitted to our ICU, overall, 60-day mortality was 44% with no signi cant difference between the rst and second waves. Mortality increased with the number of comorbidities, delayed mechanical ventilation and the SOFA score.

Declarations
Our research was approved by our local ethics committee. All patients or families were informed of the data collection.
We give consent for publication