Trajectory of Viral RNA Load Among Persons With Incident SARS-CoV-2 G614 Infection (Wuhan Strain) in Association With COVID-19 Symptom Onset and Severity

Key Points Question What are the characteristics of SARS-CoV-2 G614 viral shedding in incident infections in association with COVID-19 symptom onset and severity? Findings In a cohort study of persons who tested positive for SARS-CoV-2 after recent exposure, viral RNA trajectory was characterized by a rapid peak followed by slower decay. Peak viral load correlated positively with symptom severity and generally occurred within 1 day of symptom onset if the patient was symptomatic. Meaning A detailed description of the SARS-CoV-2 G614 viral shedding trajectory serves as a baseline for comparison with new viral variants of concern and informs models to plan clinical trials to end the pandemic.

eTable 1. COVID-19 Symptoms Severity Definition COVID-19 symptoms severity levels were defined in reference to the Centers for Disease Control and Prevention (CDC) clinical criteria. 1 A person was mildly/moderately/severely symptomatic if he/she reached mild/moderate/severe levels in at least two of the following symptoms: fever chills, myalgia, headache, sore throat, nausea or vomiting, diarrhea, fatigue, congestion or runny nose; or any one of the following symptoms: cough, shortness of breath, new olfactory disorder, new taste disorder.

Symptom
Categorization of symptom severity provided by participants using daily REDCap  *Participants were asked to rate each symptom on a 5-point ordinal scale: 0 ("none"), 1 ("mild -does not interfere with daily activities"), 2 ("moderate -interferes with daily activity"), 3 ("severeprevents daily activities"), 4 ("Emergency room or hospitalization"  (22) 17 (13) a Index case(s) is a person with a laboratory-confirmed diagnosis of SARS-CoV-2 infection. All participants in this analysis were exposed to at least one lab-confirmed index case but could be exposed to more persons with presumptive diagnosis of SARS-CoV-2 infection by CDC criteria (meets clinical and epidemiologic evidence with no confirmatory laboratory testing performed for COVID-19).

Lineage
Number  Since we only collected nasal swabs from each participant for 14 consecutive days, many (108 out of 180) participants who were infected with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) during the follow-up period had censored shedding, i.e., their first and/or last collected swabs tested positive for SARS-CoV-2. We developed a Bayesian piece-wise linear mixed-effect model to estimate individual-level viral load trajectories and population-level viral load trajectory charactersitcs, including peak viral load, time from shedding onset to peak and time from peak to shedding cessation. We also summarized population-level timing of symptom onset in relation to peak viral load based on the modelestimated viral RNA load trajectories and the observed symptom onset times. The analysis was stratigied by participants' reported COVID-19 symptom severity.

Data
We conducted model-based analysis using data from 129 participants who had 2 or more positive swabs during the 14 days of follow-up in the parent study to maximize data utilization from participants with sustained viral detection. The participants who only had 1 observed day of shedding were excluded from this model because we were unable to ascertain whether the single positive swab represented the tail of an infection acquired before study entry, a re-infection, or a false-positive result. The participants who were SARS-CoV-2 positive at baseline were included to enrich the observed viral load trajectory data, especially for estimating the rate of decay of viral load.

Model description
We fitted a piece-wise linear mixed-effect model within each symptom severity group to estimate the peak viral load measured in cycle threshold (Ct) value, the time from shedding onset to peak viral load, and the time from peak to viral clearance. We assumed that the viral load trajectories follow a trend that consists of a proliferation phase with linear growth of viral load on the Ct scale, followed by a clearance phase with linear decay of viral load on the Ct scale. This corresponds to exponential growth and decay in viral RNA concentration in the respective phases. This idealized trajectory is depicted in eFigure2 of this supplement and represented by the following equations: Here, (⋅) represents an indicator function. represents time since the peak viral load of the idealized viral trajectory, so that = 0 at the peak of the trajectory. [ ( )] represents the expected value of viral load in Ct at time . μ ( ) represents the difference between the level of detection (LOD, equals 40) and the expected viral load at time . represents the absolute difference between LOD and the peak viral load in Ct. represents time in days between shedding onset and peak viral load, and represents time in days between peak viral load and viral clearance. represents the time difference between the latent peak viral load and the observed peak viral load. We assumed that the difference between the observed viral load and the LOD, ( ), has the following distribution: This model assumes that the observed Ct values are normally distributed around the expected trajectory with standard error σ during the viral shedding, and with standard error 0.05 before and after viral shedding to allow for deviation from LOD due to potential errors from various sources such as misplaced swabs. We used random effects to capture individual-level variations from the population-level means (μ) and specified the following distributions for the random effects: We ran an MCMC chain for 10,000 iterations, with the first 5,000 iterations discarded as burn-in. Using a thinning interval of 5, we used 1,000 of the second 5,000 iterations for inference.

Summary results
The posterior distributions of key population-level parameters are shown in eFigure6. Specifically, we created the density plots of 1,000 posterior samples of the population-level mean (a) peak viral load, 40 -, (b) time from shedding onset to peak viral load, ,(c) time from peak viral load to viral clearance, , and (d) total duration of shedding, + . Specifically, for each quantity, we first obtain the average value across individuals in each posterior iteration. We then plot the density of the average values across the 1000 posterior iterations.
To estimate the population-level time from peak viral RNA load to mild symptom onset among the mildly symptomatic participants, we use the following procedure: First, for each posterior iteration and each individual, we obtained the latent time of peak viral load. Second, for each posterior iteration and individual, we obtained the time from the latent peak to the observed mild symptom onset. Next, for each posterior iteration, we calculated the average time from the latent peak to the observed mild symptom onset across individuals. Last, we summarized the distribution of the average time across 1000 posterior iterations. Similar procedures were used to estimate the population-level time from peak viral RNA load to mild and moderate or severe symptom onset among the moderately or severely symptomatic participants.