Advanced ventilatory support and mortality in hospitalized patients with COVID-19 caused by Gamma (P.1) variant of concern compared to other lineages: cohort study at a reference center in Brazil

The Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) variant of concern (VOC) Gamma (P.1) has increased transmissibility and resulted in elevated hospitalization, intensive care unit occupancy and mortality rates in Brazil. It is not known whether this VOC is also associated with more severe clinical course of disease. This was a retrospective cohort study with non-elderly patients hospitalized for COVID-19 from June to December/2020 (rst period) and February to May/2021 (second period) at a reference hospital in Brazil. Two cohorts were performed: the main cohort, composed by patients with SARS-CoV-2 lineages conrmed by whole genome sequencing; and the sensitivity cohort, composed by all eligible patients admitted before and after the emergence of Gamma. The primary outcome was the incidence rate of need of advanced ventilatory support.

Some studies performed at population-level suggest that Alpha [5][6][7][8][9] infections are associated with higher hospitalization rates when compared to non-VOC lineages. However, the absence of a more detailed analysis of the clinical course at a patient level may prevent de nitive conclusions on the clinical course of COVID-19 caused by different variants [10]. Two studies evaluating covariates at an individual level found distinct results [11,12]. An increased risk of death of the Alpha variant was not found in hospitalized patients [11], while in a community-based study, a higher risk of critical care admission and 28-day mortality was observed [12]. Nevertheless, these studies have neither evaluated the ventilatory support progression during hospitalization nor other variables that might indicate a more frequent or earlier need of more critical life support.
The Gamma variant has caused a dramatic surge of COVID-19 cases, hospitalizations, ICU demand and deaths in the Amazon region, Brazil [3,13]. Gamma lineage has further spread to other Brazilian regions, determining huge hospital system overload, and further spread to other South and North American countries [1,[14][15][16] and Europe [17]. Despite this large impact over hospitalizations and mortality rates, so far, there is no study addressing whether this VOC is associated with more severe disease. In this study, we investigate the requirement of ventilatory support and mortality rates in non-elderly adult patients hospitalized due to COVID-19 caused by Gamma and non-Gamma SARS-CoV-2 lineages.

Study design, setting and participants
This is a retrospective cohort study carried out at Hospital de Clínicas de Porto Alegre (HCPA), a tertiarycare, COVID-19 reference hospital, located at the city of Porto Alegre (1,488,000 in-habitants), the capital of Rio Grande do Sul State, Brazil.
The study population was composed of patients ≥ 18 and ≤ 65 years admitted at the emergency department of HCPA in two periods, classi ed according to the rst detection of Gamma variant in the State: rst period, before the detection of Gamma (1st June to 31st December/2020) and second period, after the detection of Gamma (1st February to 31st May/2021). Since the rst detection of Gamma in our State was in January/2021 [14], it was considered a month of transition from the predominance of other lineages to predominance of Gamma, and patients admitted in this month were not considered for inclusion. Only patients whose real-time reverse transcriptase polymerase chain reaction (RT-qPCR) cycle threshold (Ct) value for both SARS-CoV-2 nucleocapsid protein N1 and N2 gene targets were < 26, allowing a higher quality whole genome sequencing (WGS), were considered for the study. Patients were excluded if they were transferred from another hospital or if they were admitted because of non-COVID-19 related diseases and had a positive screening test.
Two cohorts were designed for eligible patients. The main cohort was composed by sample of randomly selected individuals who had their RT-qPCR samples sequenced. In this cohort, patients were further excluded if WGS generated a low-quality sequence. The sensitivity cohort was composed by all eligible patients: those from the second period were considered as having Gamma infection and those from the rst period were considered non-Gamma infections.

Ethical aspects
This study is part of a research project on the epidemiology of SARS-CoV-2. It has been approved by the Institutional Ethics Committee (Project no. 2020 − 0163), which waived informed consent. A speci c amendment to this study was submitted and approved by this committee. This study was registered at ClinicalTrials.gov under the number NCT04928469 (https://clinicaltrials.gov/ct2/show/NCT04928469).

Baseline characteristics
The following variables were assessed at hospital admission: age; sex; body mass index; Charlson's Comorbidity Score and speci c comorbidities; National Early Warning Score 2 (only in the main cohort); number of days from onset of symptoms to hospital admission; 6-point ordinal scale classi cation at admission; partial pressure of arterial oxygen/ fraction of inspired oxygen (PaO2/FiO2) ratio; and laboratory values.

Outcomes
The primary outcome was incidence of need for advanced respiratory support within 28 days from the onset of symptoms. Advanced respiratory support was de ned as the requirement of supplementary oxygen by non-invasive ventilation, high-ow nasal cannula, mechanical ventilation, or extracorporeal membrane oxygenation. The main prespeci ed secondary outcomes, which were evaluated in both main and sensitivity cohorts, were incidence of need of invasive ventilatory support (requirement of mechanical ventilation or extracorporeal membrane oxygenation) from the onset of symptoms; 28-day mortality from onset of symptoms and from hospital admission; Other secondary outcomes, evaluated only in the main cohort, were progression in an ordinal clinical scale of COVID-19 severity during the rst 28 days and PaO2/FiO2 ratio during hospitalization. De nitions and other secondary outcomes are described in the Additional le 1.

Whole Genome Sequencing
The procedure is described in the Additional le 1.

Statistical Analysis
The sample size was estimated in 86 patients (43 Gamma and 43 non-Gamma patients) as detailed in Additional le 1.
Bivariate analysis of Gamma and non-Gamma-infected patients baseline characteristics was performed using Student's T or Mann-Whitney tests for continuous variables, and chi-square test or Fisher's exact test for categorical variables.
Kaplan-Meier estimates for the need for advanced respiratory support were calculated, and the difference between groups was compared using the log-rank test.
Cox proportional hazard models were performed to evaluate the effect of Gamma infection on the need of advanced respiratory support, need of invasive ventilatory support and 28-day mortality from onset of symptoms including the following variables age, sex, body mass index and Charlson's score. Other baseline variables were not considered in this model since they are supposed to represent at some degree the severity of illness and, therefore, are not confounding variables for this outcome. For 28-day mortality from hospitalization, other variables at admission were considered as detailed in Additional le 1.
Other secondary outcomes were assessed as described in the statistical analysis plan, which can be found at https://clinicaltrials.gov/ct2/show/NCT04928469, and detailed in the Additional le 1.
The database was double entered, revised and validated in the SPSS program (version 18.0). All tests were two-tailed and a P value ≤ 0.05 was considered statistically signi cant.

Characteristics of patients
A total of 2434 patients had a positive result of SARS-CoV-2 RT-qPCR collected at the emergency department. Exclusions occur as displayed in the study owchart ( Fig. 1). A total of 433 patients eligible for the study and composed the sensitivity cohort: 174 (40.2%) patients from the rst period and 259 (59.8%) from the second period. Ninety-seven randomly selected patients had their samples sequenced, 8 were excluded because the WGS presented low quality (average coverage < 350), and 1 Gamma patient recovered in the rst period and 2 non-Gamma from the second period were excluded because the rst 86 patients planned to compose the main cohort had already been included (Fig. 1).
Four of 43 samples from the Gamma group were Gamma sub-lineages (2 P.
The risk of being dead or under invasive ventilation was signi cantly higher for Gamma than non-Gamma infected patients on days 7, 14, 21 and 28 (Additional le 1: STable 3); and the risk of having severe acute respiratory distress syndrome was signi cantly higher in Gamma patients on days 7 and 14 from hospitalization (Additional le 1: STable 4). The distribution of patients in the ordinal scale in the rst 28 days from hospitalization and the progression of PaO2/FiO2 in Gamma and non-Gamma groups are shown in Fig. 3.
Patients infected by Gamma had signi cantly lower days alive and free of supplemental oxygen support than those infected by non-Gamma lineages (median, 2.0 days; IQR, 0.0-15.5 versus 18.0 days; IQR, 2.5-22.5; P < 0.001). There were no statistically signi cant differences between Gamma and non-Gamma patients in admission to ICU, need of renal replacement therapy, prone positioning, occurrence of thromboembolic event and in-hospital mortality (Additional le 1: STable 5). There was no signi cant difference in leukocyte counts, lymphocyte counts, C reactive protein and creatinine between groups along hospitalization (Additional le 1:SFigure 1).
In the Cox proportional hazard model, Gamma infection was associated with a higher risk of requiring advanced respiratory support ( Table 2). The incidence rate of mechanical ventilation, death after 28 days from onset of symptoms and from hospital admission were also signi cantly higher in Gamma than in non-Gamma-infected patients after adjustment for covariates ( Table 2).

Sensitivity analysis
Patients from the second period (presumably Gamma) had signi cantly lower comorbidity score than patients from the rst period (presumably non-Gamma) in the sensitivity cohort. Gamma patients also had signi cantly more severe acute respiratory distress syndrome and less frequently presented PaO2/FiO2 above 300 at admission (Additional le 1: STable 6); and had more often scores 4 or 5 in the ordinal clinical scale and less often score 2 than non-Gamma patients at admission (Additional le 1: STable 6).
The unadjusted risk for advanced and invasive respiratory support, 28-day mortality from onset of symptoms and from hospital admission were signi cantly higher in patients from the second than the rst period ( Fig. 4 and Additional le 1: STable 7). The adjusted risks for advanced respiratory support (P < 0.001), invasive ventilatory support (P < 0.001), and 28-day mortality from the onset of symptoms (P = 0.001) were signi cantly higher in patients from the second than the rst period (Additional le 1: STable 7). The risk for 28-day mortality from hospital admission P = 0.09) was not statistically signi cant after adjustment (Supplemental Digital Table 8).

Discussion
The collective ndings of this retrospective cohort with non-elderly adult patients attending at the emergency department of a COVID-19 reference hospital suggest that infections caused by the Gamma VOC present a more severe course of disease, indicated by higher incidence rates of advanced and invasive respiratory support, death in the rst 28 days from the onset of symptoms compared to patients with infections by other lineages, mostly B.1.1.28 and B.1.1.161, which are not VOCs.
The results were similar in the sensitivity cohort indicate that the sequenced sample is representative of the eligible population. Furthermore, the sensitivity cohort corroborates the hypothesis that Gamma infections are associated with an unfavorable course of COVID-19, since patients from the second period presented at emergency department in worse clinical status, despite the remarkably similar time from onset of symptoms to admission in both groups, as evidenced by higher scores in ordinal clinical scale and lower PaO2/FiO2 ratios.
Differently from previous studies with Alpha versus non-VOC [5][6][7][8][9]12], and with Delta compared to Alpha [18], which mainly analyzed the risk of hospitalization or in-hospital death, our study provided more detail in the ventilatory demand along the clinical course of COVID-19 caused by the Gamma VOC and other SARS-CoV-2 lineages. Advanced respiratory support was chosen as the primary outcome because during the Gamma surge, there was a restriction in the number of ICU beds available [14,16]; thus, we a priori considered that differences in the incidence of mechanical ventilation, for example, could not be detected since invasive ventilatory support might be postponed due to limited access to ICU during this period. For instance, Bastos et al.[16] have shown that the second wave in Brazil was associated with higher mortality among hospitalized patients but not with higher ICU admission, which might suggest a potential limitation in access to critical care. Surprisingly, the incidence rate of invasive respiratory support from the onset of symptoms was signi cantly higher in patients infected by Gamma, even in the context of restricted access to ICU, corroborating to the hypothesis that infections by Gamma resulted in more severe disease.
We included only non-elderly patients to minimize the effect of age on the course of the disease. Along with adjustment for age in the multivariable model, this was important to minimize the effect of this extremely relevant confounding factor. Additionally, despite the effect of aging oh the course of COVID-19, including higher age groups VOC will introduce vaccination as another confounding factors. The vaccination of individuals who were 65 years started on early April in our State. Since we could not have these data standardly registered in medical records, it is possible that some patients of our study have received the rst shot when admitted. However, vaccinated patients in Gamma or second period groups favor the null hypothesis, since it is a factor of protection for more severe disease.
In addition to those inherent to the retrospective design, our study has some limitations that must be acknowledged. First, it must be considered that it is a single center experience of a large tertiary-care COVID-19 reference hospital, which may restrict generalizability. Second, since Gamma and non-Gammainfected patients were hospitalized in different periods, they might be affected by distinct practices in COVID-19 management. However, it should be noted that more than 90% of patients have been treated with corticosteroids and there was no other pharmacological intervention used in the second period. Furthermore, if better practices in COVID-19 patients care had been incorporated in more recent periods, it would favor the null hypothesis. Third, the magnitude of the effect of Gamma on 28-day mortality in the main cohort must be interpreted with caution because of the low number of this events, which resulted in a large 95% CI in the model. The lower adjusted risk in the sensitivity cohort more likely re ects the magnitude of the effect of Gamma infections. Fourth, the study might be affected by a selection bias because the higher hospital occupancy in the second period may have delayed the admission of less severe cases. Nonetheless, it must be pointed out that time from onset of symptoms to hospital admission were essentially the same in both main and sensitivity cohorts. Finally, our study addressed clinical evolution of patients with severe COVID-19 that required hospital admission and the nding cannot be extrapolated to outpatient population.

Conclusions
Our study suggests that in non-elderly adult patients who require hospital admission, COVID-19 caused by Gamma present a more severe clinical course, as evidenced by the higher incidence of advanced respiratory support and 28-day mortality in patients infected by the Gamma VOC. In addition to the Availability of data and materials The datasets analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
This study is part of a research project on the epidemiology of SARS-CoV-2. It has been approved by the Institutional Ethics Committee (Project no. 2020-0163), which waived informed consent. A speci c amendment to this study was submitted and approved by this committee.

Consent for publication
Not applicable.