SARS-CoV-2 detection in the lower respiratory tract of invasively ventilated ARDS patients

All data of COVID-19 patients were prospectively collected in two specific standardized institutional databases. We collected demographic characteristics, medical history, characteristics at hospital and intensive care unit (ICU) admissions, organ support procedures, administration of antivirals and other agents, and biological and virological results. Of note, prescription of antivirals and anti-inflammatory drugs varied among countries, availability, and choice of physician.

All patients had a SARS-CoV-2 positive nasopharyngeal swab before ICU admission. During the ICU stay, all patients underwent a regularly monitoring of LRT samples (LRTS) until the detection of two consecutive negative real-time polymerase chain reaction (RT-PCR) results [8]. All patients underwent regularly to LRT RT-PCR re-evaluation until removal of the endotracheal tube or death. Endotracheal aspirates, bronchoalveolar leakage fluid, and plugged telescoping catheter were all considered as LRTS. The sampling was performed according to a predefined protocol. All specimens were sent to two accredited reference virology laboratories and used for RNA extraction and amplification by RT-PCR techniques utilizing validated commercial kits that were previously described [9]. A cycle threshold (Ct) value of < 40 was defined as positive for SARS-CoV-2 RNA and ≥ 40 was defined as negative (see supplementary material). Then, RT-PCR values were transformed in viral load as previously published [10]. The French and the regional Swiss Ethics Committees approved this study.

The primary objective was the description of viral shedding duration (i.e., time to negativity) in LRTS and was defined as the time between onset of symptoms and viral clearance process (i.e., first negative detection of viral RNA in LRT). Of note, we previously showed that the association of the first negative RT-PCR with a second negative result was very high (i.e., 97%) in LRTS [11]. Secondary objectives were (1) the analysis of RT-PCR viral load in LRTS and (2) the assessment of 6-week mortality. Patients discharged before 6-weeks were considered alive.

The statistical plan had three steps. First, we explored the duration of viral shedding in LRT using Kaplan-Meier curves for different clinical relevant patient populations (i.e., age, diabetes, cardiac comorbidities, immunosuppression, and corticosteroids during the ICU stay) which are thought to influence viral shedding in upper respiratory tract samples [2, 12‐15]. A log rank test was performed; a specific subgroup had prolonged viral shedding if the log rank test between the two groups (e.g., diabetic versus non-diabetic) was statistically significant. Second, we evaluated viral load in LRT during the ICU stay (i.e., quantitative RT-PCT) using graphical and descriptive statistics (i.e., Wilcoxon test) between survivors and non-survivors. Third, we determined the association between viral presence in LRT and mortality using mixed-effect logistic models for clustered data (PROC GLIMMIX of SAS) and adjusting for the time between symptoms’ onset and date of sampling. All collected samples were considered for this analysis. These models take into account the clustering effect of multiple sampling per patient and the center effect (i.e., random effect). We performed two sensitivity analyses: the first using only endotracheal aspirates and the second adjusting for sex and age. Tests were two-tailed, with p < 0.05 being considered significant. All analyses were performed using SAS (version 9.4) and R (version 3.5.3). The current analysis complied with the STROBE guidelines for observational studies [16].