Inferring Timing of Infection Using Within-host SARS-CoV-2 Infection Dynamics Model: Are “Imported Cases” Truly Imported?

Abstract

Importance: Although the COVID-19 epidemic in some countries such as China are in the last phase by large effort for containment of the disease, another outbreaks can occur because huge susceptible population remains. Further, there remain countries in the early phase of outbreak with zero or limited number of cases in southern hemisphere countries. In those countries at risk of future outbreak, ascertaining whether cases are imported or the result of local secondary transmission is important for government to shape appropriate public health strategies. Objective: To develop a method to estimate timing of infection establishment, which helps differentiate imported and autochthonous cases. Design, Setting and Participants: Of the first 18 cases reported in Singapore, 12 were used in our study (1 case with insufficient data and 5 on anti-viral treatment were excluded from the analysis). The viral load data from these initial cases considered imported due to their travel history to Wuhan were analyzed. Another viral load data from 3 cases reported from Zhuhai, China, for whom exposed day were known, were also analyzed to determine the viral load threshold for infection establishment. Exposures: SARS-CoV-2 infection confirmed by the polymerase-chain-reaction (PCR) test. Main Outcomes and Measures: The timing of infection establishment of each case was assessed by analysing viral load data after symptom onset using a within-host viral dynamics model for SARS-CoV-2. Estimated timing of infection will indicate whether cases are imported or autochthonous transmission within Singapore. Results: Six among the 12 cases were clearly imported cases, whereas we could not rule out the possibility of secondary transmission for the rest of 6 cases, which collectively evidenced ongoing transmission in Singapore. For the 6 cases who could be the results of secondary transmission, further investigation to identify the source of infection within Singapore should be warranted (i.e., contact tracing). Conclusions and Relevance: In an early phase of outbreak due to entrance or re-entrance of the virus to countries/communities, collecting viral load data over time from cases from symptom onset is highly recommended, because viral load data are valuable to infer the timing of infection and distinguish between imported cases and ongoing local transmission.

In-host Mathematical Modelling of COVID-19 in Humans

“…While there are many mathematical models developed at epidemiological level for COVID-19 to discuss the transmission of SARS-CoV-2 and de-confinement strategies ( Anderson, Heesterbeek, Klinkenberg, Hollingsworth, 2020 , Ferretti, Wymant, Kendall, Zhao, Nurtay, Abeler-Dörner, Parker, Bonsall, Fraser, 2020 , Lopez, Rodo, 2020 , Mejia-Hernandez, Hernandez-Vargas, 2020 , Peng, Yang, Zhang, Zhuge, & Hong , Ricardo-Azanza, Vargas-Hernandez, 2020 ), there are too few models at within-host level to understand SARS-CoV-2 replication cycle, interactions with the immune system, and drug effects ( Du, Yuan, 2020 , Ejima, Kim, Ito, Iwanami, Ohashi, Koizumi, Watashi, Bento, Aihara, Iwami, 2020 , Gonçalves, Bertrand, Ke, Comets, de Lamballerie, Malvy, Pizzorno, Terrier, Calatrava, Mentré, Smith, Perelson, Guedj, 2020 , Goyal, Cardozo-Ojeda, Schiffer, 2020 , Su, Ejima, Ito, Iwanami, Ohashi, 2020 , Wang, Heiland, Craig, Davis, Ford Versypt, Jenner, Ozik, Collier, Cockrell, Becker, An, Glazier, Narayanan, Smith, Macklin, 2020 ). Among different model structures to represent viral dynamics, the target cell limited model has served to represent several diseases such as HIV Hernandez-Vargas and Middleton (2013) ; Perelson and Ribeiro (2013) ; Pinkevych et al.…”

COVIDStrategyCalculator: A software to assess testing- and quarantine strategies for incoming travelers, contact person management and de-isolation

“…The software allows full flexibility with regards to parameter choices, that, for example, determine the time-course of infection, the proportion of asymptomatic cases and the test sensitivity, etc., and can thus be tailored to user-specific queries. We have however carefully calibrated the models' default parameters to reproduce published and inhouse clinical data on the incubation time 53 , the off-set of infectiousness after peak virus load/symptom onset [54][55][56]59 , as well as the time-dependent test sensitivities 57,58 . Figure 2A shows the cumulative time-to-symptom-onset (grey shaded area) compiled in a meta-analysis of 56 studies 53 , together with the model-predictions (solid-and dashed lines) using the default parameters.…”

“…As can be seen, the utilized model perfectly reproduces not only the mean duration of incubation, but also the entire waiting time distribution. Figure 2B shows a summary of data sets used to evaluate the duration of infectiousness after peak virus load/symptom onset (shaded areas) [54][55][56]59 , including the analysis of in-house data (Supplementary Note 2-3). The depicted data is scaled to represent the relative reduction in infectiousness assessed by culture-positivity, as well as viral load dynamics (details on the data analysis and parameter fitting procedure are provided in Supplementary Note 2-3).…”

“…While Singanayagam et al and van Kampen et al 55,56 report relative culture-positivity over time, we used the in-house data to derive and fit a mechanistic model Supplementary Note 3 that allows to convert virus load profiles post-symptom onset into infectivity profiles. We used the mechanistic model to derive infectivity profiles from viral loads reported by Ejima et al 54 and from the Charité 59 . The infectiousness profiles show a marked dispersion between different studies, which may be partly due to the investigation of different cohorts (mild-moderately ill 55 vs. hospitalized severely ill patients 56 , differences in the definition of `symptom onset', and methodological differences in the laboratory assays used to assess infectiousness.…”

“…The software allows full flexibility with regards to parameter choices, that, for example, determine the time-course of infection, the proportion of asymptomatic cases and the test sensitivity and much more. However, a set of default parameters is provided, that has been carefully derived by fitting the model's parameters to available clinical and in-house data on the incubation time 53 , the off-set of infectiousness after symptom onset or peak virus load [54][55][56] , as well as the time-dependent test sensitivities 57,58 . Details on the parameter fitting procedure and analysis of infectivity profiles are provided in Supplementary Note 2-3.…”