Between October 1, 2020, and May 30, 2021, 845 patients were admitted in one of the 11 participating ICUs. Among them, 737 had at least one positive SARS-CoV-2 RNA RT-PCR performed in a nasopharyngeal swab sample in the hospital; 413 patients with a Ct ≤32 and a nasopharyngeal sample conserved at -80°C that could be used for full-length viral genome sequence analysis were included in this study. Among these 413 patients, 183 (44.3%) had been infected with so-called “pre-existing” variants, i.e., variants circulating before the emergence of variant α, 197 (47.7%) with variant α (B.1.1.7), and 33 (8.0%) with other variants, including β (B.1.351) (n=19, 4.6%), γ (P.1) (n=2, 0.5%) and other variants of interest (n=12, 2.9%). Figure 1 illustrates the time course of emerging SARS-CoV-2 variants during the study period. Variant α (B.1.1.7) was first detected during the last week of 2020 in 11.1% of patients, became predominant during the fourth week of January 2021 (51.9% of patients) and remained so until the last week of May (100% of patients) that corresponded to the end of the inclusion period.
SARS-CoV-2 variant association with clinical phenotypes at ICU admission
Patients infected with pre-existing variants were older and had significantly more frequent chronic kidney disease and COPD than those infected with other variants (Table 1). Other comorbidities, including diabetes, peripheral vascular disease, solid cancer, and hypertension followed the same trend, although differences were not statistically different.
There were also marked differences between variant groups regarding severity of disease at ICU admission. Indeed, patients infected with preexisting variants exhibited significantly higher severity of illness scores, with not only more severe respiratory disease, as assessed by the respiratory component of the SOFA score, but also more frequent extra-pulmonary organ failures, as compared with patients infected with variant α (B.1.1.7) and other emerging variants (Table 1 and Figure 2). As a result, patients infected with pre-existing variants received less frequent high-flow oxygen therapy and more frequent extracorporeal membrane oxygenation (ECMO) support than other patients within the first 24 hours of ICU admission, while they more frequently met the ARDS definition criteria.
Relationship between SARS-CoV-2 variants and patient outcomes
Although patients infected with pre-existing variants were older, had more comorbidities and had more severe diseases at ICU admission, they did not show different outcomes from those infected with other variants. Overall, the need for invasive mechanical ventilation or ECMO support did not significantly differ between the different variant groups, nor did the duration of these supports. There were also no significant differences between groups regarding extra-pulmonary organ supports (i.e., vasopressors and renal replacement therapy) during ICU stay (Table 2). Patients with pre-existing variants received less dexamethasone and tocilizumab than others, likely reflecting the changes in practice during the study period. As shown in Table 2, day-28 and day-90 mortality rates were not statistically different between groups.
Relationship between SARS-CoV-2 Spike substitutions and deletions and mortality
Pre-existing and emerging variants (VOC or VOI) are characterized by multiple lineage-specific deletions and amino-acid substitutions. Spike mutations undergo evolutive convergence at several signature sites. Spike mutations at sites identified to undergo convergent mutational evolution were selected a priori. There was no significant relationship between any of these mutations (n=17) and day-28 mortality in univariate analysis (Figure 3). The selected mutations were also included in multivariate logistic regression models to determine their relationship with day-28 (Table 3) and day-90 (Supplemental Table 1) mortality, adjusting on age, gender, SOFA score at ICU admission, and dexamethasone treatment. None of them was significantly associated with mortality at either time point. The variant status was not either associated with mortality.
Relationship between SARS-CoV-2 spike and other gene mutation hotspots and mortality
Overall, 1017 non-synonymous mutations (including 953 amino acid substitutions, 52 deletions and 11 insertions) were detected in full-length viral genomes in at least one variant. There was no significant relationship between any of these mutations and day-28 mortality in univariate analysis (Supplemental Table 2). Although 11 mutations were found to be associated with mortality (p<0.05), the number of patients harboring these mutations was considered too small (range: 2-6) to drive conclusions.
We then focused on mutations previously reported to correlate with worse clinical outcomes [2,4,17]. We found that Q57H in Orf3a (but not other mutations, including P25L in Orf3a and S194L, R203K or G204R in N) was highly prevalent in our cohort (31.8%). Q57H was more prevalent in patients who were dead at day-28 (35.6%) than in those who were still alive, but the difference did not reach significance (29.0%; p=0.196).