key: cord-0857977-y4cbatxe authors: Regan, James; Flynn, James P; Rosenthal, Alexandra; Jordan, Hannah; Li, Yijia; Chishti, Rida; Giguel, Francoise; Corry, Heather; Coxen, Kendyll; Fajnzylber, Jesse; Gillespie, Elizabeth; Kuritzkes, Daniel R; Hacohen, Nir; Goldberg, Marcia B; Filbin, Michael R; Yu, Xu G; Baden, Lindsey; Ribeiro, Ruy M; Perelson, Alan S; Conway, Jessica M; Li, Jonathan Z title: Viral Load Kinetics of SARS-CoV-2 In Hospitalized Individuals with COVID-19 date: 2021-04-17 journal: Open Forum Infect Dis DOI: 10.1093/ofid/ofab153 sha: d4efbd6b1712cf8c8a0f92aff86017103a64ce58 doc_id: 857977 cord_uid: y4cbatxe SARS-CoV-2 kinetics remain understudied, including the impact of remdesivir. In hospitalized individuals, peak sputum viral load occurred in week 2 of symptoms while peak viremia occurred within a week of symptom-onset, suggesting early systemic seeding of SARS-CoV-2. Remdesivir treatment was associated with faster rates of viral decay. Understanding viral load dynamics has provided key insight on viral pathogenesis and treatment effects across the spectrum of viral infections [1] . Study of SARS-CoV-2 kinetics within the respiratory tract has already provided valuable information about disease course, transmission risk, and efficacy of antibody therapeutics [2] [3] [4] [5] . During severe infection, SARS-CoV-2 viral RNA can be detected not only in the upper (URT) and lower respiratory tracts (LRT), but also systemically in plasma [6] . Viral decay kinetics can be influenced by multiple factors, including replication dynamics, host cell turnover, and focal intensity of immune responses. However, little is known about the differences in viral decay between respiratory and non-respiratory compartments [7] , especially as viremia is associated with COVID-19 disease severity and mortality [6, 8] . Whereas viral decay from nasopharyngeal sampling has been valuable in evaluating the efficacy of monoclonal antibody treatments against SARS-CoV-2 [4, 5] , the effect of remdesivir on viral dynamics remains unclear. While remdesivir appears to confer a clinical benefit [9] , its ability to alter respiratory tract SARS-CoV-2 kinetics has not been demonstrated [10] . It is unknown whether remdesivir treatment effects may be more accurately observed by evaluating a range of specimen types. We present an observational study of viral kinetics in patients hospitalized for COVID-19. We quantify SARS-CoV-2 viral load in longitudinal samples from the respiratory tract and plasma, and evaluate the effect of remdesivir on viral load decay. A c c e p t e d M a n u s c r i p t We enrolled hospitalized COVID-19 patients. Longitudinal nasopharyngeal swabs, oropharyngeal swabs, sputum, and blood were collected from some patients. Each participant's medical record was reviewed to determine their oxygenation status, demographics, comorbidities, treatment status, and clinical outcome. Informed written consent was obtained from all patients except for 10 patients who received waivers of informed consent. This study was approved by the Mass General Brigham Institutional Review Board. SARS-CoV-2 viral loads were quantified with an in-house RT-qPCR assay as previously described [6] . Viral load kinetics amongst compartments were compared using all data from 196 participants and also compared using only longitudinal data from participants with samples collected 7-14 days apart. To analyze the effect of remdesivir on SARS-CoV-2 decay rate, we employed mixed-effects modeling on a subset of data comprised of participants who were sampled longitudinally [11] , using Monolix Software 2018R2 (http://www.lixoft.eu). We modeled viral decay using patient-compartment datasets, which consist of the viral load data for a single anatomical compartment in a single patient. Starting from our original 196 patient A c c e p t e d M a n u s c r i p t dataset, patients without longitudinal sampling were excluded along with any patient with unknown treatment status, unknown symptom onset, or remdesivir treatment during a previous hospital admission. We included detectable measurements and one subsequent undetectable measurement if it was within one week of the last detectable measurement. We excluded patient-compartment data from participants for whom there were no detectable measurements and data from before viral load decay. After this post-processing, we excluded any patient-compartment dataset that did not have at least two data points. We were left with data for 51 participants, with 70 distinct patient-compartment datasets. Of these 70 datasets, 18.6%, 15.7%, 45.7% and 20% belonged to the nasopharyngeal, oropharyngeal, plasma, and sputum compartments, respectively. Decaying viral loads were modeled as V(t)=V 0 e -rt , using mixed-effects modeling on the Log 10 -transformed data and treating undetectable measurements as censored at the assay limit of detection. This analysis was performed for all compartment data together. Levels of SARS-CoV-2 RNA were compared to the duration of time between symptom onset and sample collection. All correlation analysis was performed using Spearman rank-based testing. Changes in SARS-CoV-2 viral load were calculated as the change in log 10 copies of RNA per day between sample collections and were treated as a continuous variable. Estimated decay rate r was treated as a continuous variable. All continuous variables were analyzed with non-parametric rank-based testing. Comparison of viral loads (detectable vs undetectable) were treated as categorical variables and analyzed using Fischer's exact tests. A c c e p t e d M a n u s c r i p t We enrolled 196 symptomatic, hospitalized participants with COVID-19 (Supplemental Table 1 ). The proportion of samples with detectable SARS-CoV-2 RNA were highest within the first week of symptoms for samples collected from the nasopharyngeal (57%), oropharyngeal (83%), and plasma (38%) compartments ( Figure 1A ). In contrast, there was a delay in viral seeding of the LRT with significant increases in the proportion of samples with detectable sputum viral loads in the second week after symptom onset (week 1 vs 2: 56% vs 100%, P=0.003) and higher median peak sputum viral loads (week 1 vs 2: 1.8 vs 5.6 log 10 RNA copies/mL, P=0.02, Supplemental Figure 1 ). In the setting of delayed peak sputum viral load, the proportion of samples with detectable sputum viral load was significantly higher in subsequent weeks compared to other compartments. Four weeks after symptom onset, 63% of participants had detectable viral load in sputum compared to 13% by nasopharyngeal swab, A c c e p t e d M a n u s c r i p t We analyzed a subset of participants with longitudinal sample collections to model the distribution of viral decay rates simultaneously in all anatomical compartments (Supplemental Figure 2 ). This subset of participants had more severe illness and longer hospitalizations when compared to all participants (Supplemental Table 1 ). In this model, we found higher median viral decay rates in remdesivir-treated participants (untreated vs treated: r = 0.15 vs 0.31, p<0.0001, Figure 2 ). Note that, in the estimated decay rate r, we found no statistical support for the anatomical compartment being a covariate. Remdesivir-treated and untreated participants in this analysis had comparable estimated initial viral loads across all compartments (treated vs untreated, 5.9 vs 6.0 log 10 RNA copies/mL). In this study, we evaluated SARS-CoV-2 viral load dynamics across multiple anatomic compartments in hospitalized individuals with COVID-19 and assessed the effect of remdesivir treatment on viral kinetics. Our results demonstrate that viral loads in the blood and URT were highest within 1 week of symptom onset, while suggesting that both viral peak and clearance in sputum were delayed compared to that of other sampling locations. Remdesivir treatment was associated with an increased rate of viral decay in a combined viral decay analysis across multiple compartments. Our observation of a delayed viral peak in sputum samples is consistent with the viral dynamics observed in animal models [12] . A study of SARS-CoV-2-infected rhesus macaques suggests that viral load peaks earlier in the URT than in the LRT, and that virus disseminates from the URT to the rest of the body. Our data also shows that viral RNA is detected in the LRT for a longer period than other anatomical compartments, suggesting the A c c e p t e d M a n u s c r i p t importance of LRT testing for the diagnosis and monitoring of patients with severe infection. The detection of SARS-CoV-2 viremia has been attributed to viral extravasation from the pulmonary tract. Unexpectedly, our results show that viral load kinetics may be asynchronous between the LRT and plasma. The early viremia peak suggests that systemic seeding and disseminated infection may be occurring sooner than previously recognized. Efficient viral replication is a major factor in these early SARS-CoV-2 dynamics [13] , but viral decay is also affected by immune responses and cell turnover. Additional studies are needed to determine whether the relatively similar viral decay rates suggest uniform impact of these factors across compartments. Whereas there is in vitro data that remdesivir inhibits SARS-CoV-2 replication [14] , to date no evidence has been published that remdesivir has significant effects on viral load [10] . In our model across sampling compartments, we observed a significant increase in viral decay rate for participants treated with remdesivir. To the best of our knowledge, this is the first in vivo data to suggest that remdesivir affects SARS-CoV-2 kinetics. Limitations of this analysis include the relatively limited sample size of individuals that restricted our ability to compare remdesivir-associated effects on viral decay kinetics between different compartments. Additionally, this is an observational study, but we were able to match the treated and untreated groups based on baseline virological data. Finally, our analysis is based solely on hospitalized individuals and the findings may not be generalizable to patients with asymptomatic or mild disease. In conclusion, we find that SARS-CoV-2 kinetics differed in the sputum compared to other compartments, and that systemic spread of SARS-CoV-2 into the circulatory system occurs early in the disease course. Remdesivir-treated individuals had significantly faster rates of viral decay in a combined analysis across multiple compartments. Larger clinical trials are necessary to further assess the virologic effect of remdesivir treatment. M a n u s c r i p t Mechanistic modelling of SARS-CoV-2 and other infectious diseases and the effects of therapeutics Virological assessment of hospitalized patients with COVID-2019 Temporal dynamics in viral shedding and transmissibility of COVID-19 SARS-CoV-2 Neutralizing Antibody LY-CoV555 in Outpatients with Covid-19 REGN-COV2, a Neutralizing Antibody Cocktail, in Outpatients with Covid-19 SARS-CoV-2 viral load is associated with increased disease severity and mortality Detection of SARS-CoV-2 in Different Types of Clinical Specimens High Frequency of SARS-CoV-2 RNAemia and Association With Severe Disease Remdesivir for the Treatment of Covid-19 -Final Report Remdesivir in adults with severe COVID-19: a randomised, double-blind, placebo-controlled, multicentre trial Mixed Effects Models for the Population Approach: Models, Tasks, Methods, and Tools. Chapman and Hall/CRC Virulence and pathogenesis of SARS-CoV-2 infection in rhesus macaques: A nonhuman primate model of COVID-19 progression Coronavirus biology and replication: implications for SARS-CoV-2 Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro A c c e p t e d M a n u s c r i p t