Invisible spread of SARS-CoV-2.

We read with interest Adam Kucharski and colleagues' mathematical modelling study of the early dynamics of coronavirus disease 2019 (COVID-19). We agree that a stochastic transmission model might best fit with the reality around the Huanan Seafood Wholesale Market, which was the origin of the COVID-19 outbreak and 1 mile away from our hospitals in Wuhan. We appreciate the work making use of different datasets and considering travel. However, we have concerns about the clinical and strategic values of this work.

First, the authors separated exposed (and not yet symptomatic) individuals from infectious (and symptomatic) individuals. Clinically, both groups are contagious. We wonder if they considered separately for these two groups the correlation of variation in the viral genome with speed of spread?

Second, epidemiological modelling depends primarily on the use of a realistic and dynamic basic reproduction number (R 0), such as those in a previous study, in which the reported R 0 varied from greater than 7 before, to less than 1 after, control measures were introduced.

Third, it is unclear to us whether the Wuhan-based stochastic transmission model can accommodate variation in cultures and lifestyles, which often affects adherence to social distancing, which is crucial for prevention of respiratory transmission.

The discussion says “Our results…suggest a decline in transmission in Wuhan in late January, 2020, around the time that control measures were introduced.” The daily number of new cases actually kept climbing for another 29 days after the city was sealed off. Considering that asymptomatic transmission was accounted for but the 5·2 days used as the crucial incubation period was too short—relative to a wide range of 0–24 days or an average of 6·4 days—was this discrepancy attributable to underestimation of the incubation period?

We believe that the modelling would be more instructive if it considered comparisons between absence of, presence of, or delays in lockdown. Such data would benefit timely policy making.