Study population and setting
This study systematically reviewed, collated, and analyzed unique individual patient data (IPD) from published studies which described temporal viral load and shedding data from coronaviruses, including SARS-CoV, SARS-CoV-2 and MERS. Following PRISMA-IPD guidelines for systematic reviews of individual IPD, this study searched for papers on PubMed through June 8, 2020. In total, data from 43 studies including 932 unique individuals with 5,328 observations were utilized for the analysis of viral load distribution, and 66 studies including 1,198 individuals and 7,240 observations were used for the viral shedding analysis. Viral load trajectories were generated using a series of generalized additive models and the duration of viral shedding was analyzed by fitting log-normal models to the data accounting for interval censoring.
Summary of Main Findings
This study observed that SARS-CoV-2 viral load peaked prior to symptom onset and remained elevated for up to three weeks, while MERS-CoV and SARS-CoV viral loads peaked after symptom onset. Severe patients had significantly higher viral loads than mild patients (9.9% higher after converting from log10 copies to raw copy number). SARS-CoV-2, MERS-CoV, and SARS-CoV had median viral shedding durations in respiratory samples of 4.8, 4.2, and 1.2 days after symptom onset. In gastrointestinal samples, SARS-CoV-2 and MERS-CoV had median viral shedding of 4.9 days, and 1.9 days, respectively. No differences in duration of SARS-CoV-2 viral shedding were seen by sex or with age.
Combing data from across studies allows for a large sample of individual-level data to support statistical modeling to quantify temporal distributions with adjustment for factors such as clinical and demographic characteristics. Comparison of SARS-CoV-2 with other coronaviruses allowed for better understanding of why the transmission dynamics of SARS-CoV-2 may differ from other coronaviruses.
Testing frequency was not consistent across specific study samples or disease severity, which may introduce selection bias. The pooled data did not have did not have complete data for age, sex, severity, and hospitalization status for all individuals. Copy number values were not reported consistently, as 65% of the studies included reported only cycle threshold (Ct) values from the qPCR samples instead of copy number, and therefore Ct values were converted to log10 copy number using an average standard curve which may have introduced error. There was limited uniformity in time measurement for the timing of symptom onset. Finally not all eligible studies contributed data leaving the possibility of selection bias.
This study leveraged pooled individual-level to assess temporal viral load dynamics and shedding duration across studies for SARS-CoV, MERS-COV, and SARS-CoV-2.
This review was posted on: 19 November 2020