Using models to identify the causes of pre-symptomatic transmission from human infection data

Abstract

When disease transmission can precede symptom onset, containing outbreaks requires distinct strategies, like active surveillance. Yet it is rarely clear in advance when such interventions are needed, especially for emerging pathogens. Predicting pre-symptomatic transmission would be easier with knowledge of the within-host dynamics that enable presymptomatic transmission. To investigate those dynamics, we survey controlled human infection (CHI) trials with viral agents, which contain data on incubation times, infection duration, and transmission potential following inoculation with a known dose. We find that all studies report information on the duration of viral shedding, but few report the timing of symptoms. Only one study provided data on the timing of shedding and symptoms for individual participants, following norovirus inoculation. We apply a statistical model to individual time series to show significantly greater potential for pre-symptomatic transmission with faster viral replication, but no evidence for a tradeoff between transmission rate and duration during the pre-symptomatic phase. We then compare within-host models of pathogen replication, immune clearance, and symptom onset to identify plausible assumptions about the causes of pre-symptomatic transmission. We recover the pattern that peak shedding can precede symptom onset if we assume that symptoms are triggered by immune responses rather than pathogen abundance. Only by relaxing the standard assumption of exponential growth can we recover the pattern that faster viral replication enables pre-symptomatic transmission. Thus, data on symptom onset in CHI trials, paired with models, can illuminate the within-host dynamics underpinning pre-symptomatic transmission, guiding efforts to improve control strategies.