Quantitative analysis of massive SARS-CoV-2 testing in the community in France in 2021-2022 reveals the associations of variant, vaccination, and age with viral dynamics in symptomatic individuals.

The COVID-19 pandemic has shown the value of large-scale community PCR tests for epidemic surveillance, but the viral load measurements they provide have seldom been exploited to reconstruct within-host trajectories. Because these data are collected for diagnostic rather than research purposes, they are characterized by sparse longitudinal follow-up, heterogeneous sampling, and missing metadata, raising uncertainty about their usefulness to reconstruct with-host viral dynamics. We first conducted a simulation study to assess the feasibility of estimating the viral dynamics patterns from such datasets. Across multiple scenarios replicating realistic sampling patterns, we found that peak viral load and clearance time could be estimated with good accuracy, although uncertainty tended to be underestimated. Parameters driving early viral kinetics, e.g., incubation and proliferation time, were mostly estimated with poor precision, as most community tests are conducted after symptom onset, later in infection. We then applied this framework to a large dataset of 322,218 PCR tests associated with symptomatic SARS-CoV-2 infections in France between July 2021 and March 2022, encompassing both Delta-variant circulation and the emergence of first Omicron variants. We quantified the associations of age, vaccination status, and variant type with viral load trajectories. Age ≥ 65 years was consistently associated with a longer clearance time (2-6 days), while vaccination was associated with a shorter clearance time (2-4 days). Infections with Omicron variants were associated with lower peak viral load (2-3 Ct) and shorter clearance times (1-2 days) compared with pre-Omicron (Delta) infections. Thus, community PCR tests can be leveraged to identify key parameters of viral dynamics. As multiplex PCR testing becomes increasingly widespread, establishing robust frameworks for data collection, sharing, and privacy protection will be essential to support the use of these data for modelling purposes.