key: cord-259368-k8t8brjy
authors: Ren, Xiang; Li, Yu; Yang, Xiaokun; Li, Zhili; Cui, Jinzhao; Zhu, Aiqin; Zhao, Hongting; Yu, Jianxing; Nie, Taoran; Ren, Minrui; Dong, Shuaibing; Cheng, Ying; Chen, Qiulan; Chang, Zhaorui; Sun, Junling; Wang, Liping; Feng, Luzhao; Gao, George F.; Feng, Zijian; Li, Zhongjie
title: Evidence for pre‐symptomatic transmission of coronavirus disease 2019 (COVID‐19) in China
date: 2020-08-07
journal: Influenza Other Respir Viruses
DOI: 10.1111/irv.12787
sha: 
doc_id: 259368
cord_uid: k8t8brjy

BACKGROUND: Between mid‐January and early February, provinces of mainland China outside the epicentre in Hubei province were on high alert for importations and transmission of COVID‐19. Many properties of COVID‐19 infection and transmission were still not yet established. METHODS: We collated and analysed data on 449 of the earliest COVID‐19 cases detected outside Hubei province to make inferences about transmission dynamics and severity of infection. We analysed 64 clusters to make inferences on serial interval and potential role of pre‐symptomatic transmission. RESULTS: We estimated an epidemic doubling time of 5.3 days (95% confidence interval (CI): 4.3, 6.7) and a median incubation period of 4.6 days (95% CI: 4.0, 5.2). We estimated a serial interval distribution with mean 5.7 days (95% CI: 4.7, 6.8) and standard deviation 3.5 days, and effective reproductive number was 1.98 (95% CI: 1.68, 2.35). We estimated that 32/80 (40%) of transmission events were likely to have occurred prior to symptoms onset in primary cases. Secondary cases in clusters had less severe illness on average than cluster primary cases. CONCLUSIONS: The majority of transmissions are occurring around illness onset in an infected person, and pre‐symptomatic transmission does play a role. Detection of milder infections among the secondary cases may be more reflective of true disease severity.

A novel coronavirus named "severe acute respiratory syndrome coronavirus 2" (SARS-CoV-2) was first identified in January 2020 as the pathogen responsible for a cluster of cases of atypical pneumonia in Wuhan, a large city located in Hubei province in central China. 1, 2 Genetic analysis of the virus indicates that it originated from a bat coronavirus. 2 SARS-CoV-2 is considered distinct from SARS-CoV or MERS-CoV, and coronavirus disease 2019 caused by it has rapidly become a global health concern. 3 Incidence of infections slowly increased through January 2020 with a reproductive number estimated to be in the range 2.2-3.6. [4] [5] [6] [7] Starting in mid-January 2020, COVID-19 cases began to be identified in other cities in China and also in other countries. 8, 9 A number of studies suggested the probable phenomenon of COVID-19 transmissions during incubation period, [10] [11] [12] these studies used cluster cases in limited number of families for analysis, and the relative frequency of pre-symptomatic transmission was not quantified. Moreover, there was an analysis on publicly available data indicating the existence of negative serial intervals, also implying pre-symptomatic transmission. 13 Here, we retrospectively analyse data on cases identified outside of Hubei province through the Chinese Public Health Event Surveillance System at the early stage of transmission in China, in order to provide insights on the transmission dynamics of COVID-19.

Data were extracted from the epidemiological reports of the Chinese Public Health Event Surveillance System, through which the first confirmed case or potential outbreaks leading to clusters of suspected cases for each county were required to be investigated and reported. The extracted variables included demographic data, possible exposure and travel history, and clinical data using a structured form. The events reported by 29 January 2020 were included for data extraction. For events by 23 January 2020, which was the date of "locking down" Wuhan city, all events (168 cases) in the system were extracted, and we used these data to construct an epidemic curve and estimate the incubation period distribution. Of the events (281 cases) reported between 24 January and 29 January 2020, we focused on those events which included probable human-to-human transmission or epidemiologically linked cases, so that we could capture a larger number of these clusters for analyses of serial intervals, transmission events and comparative severity between primary cases and secondary cases. All cases included in our analyses were laboratory-confirmed cases, and the case definitions followed We drew the epidemic curve by illness onset for cases who reported that they had been in Wuhan in the 14 days of preceding onset, and a separate epidemic curve for the other cases. Cases with onset dates closed to the end of epidemic curve may not be reported due to delays in seeking medical attention and consequent delays in laboratory testing. To allow for onset-to-reporting delays, we used the nowcasting approach described by van de Kassteele et al 15 to jointly estimate the augmented case number on the most recent dates and the onset-to-reporting distribution. Based on 1000 augmented epidemic curves, then we fitted exponential growth models to obtain estimates of the growth rate and doubling time. We examine the characteristics of confirmed cases and compared the demographics, onset symptoms and results of some clinical tests between confirmed sporadic and cluster primary cases and cluster secondary cases. We also obtained information on secondary cases that had not been to Wuhan but had close contact with another case that had been to Wuhan and were presumed to be infected by that particular case.

In these pairs of primary and secondary cases, we fitted a normal distribution to the serial intervals between illness onset dates, allowing for negative and zero serial intervals, and correcting for growth rates in the early stage of an epidemic. Specifically, the parameters of the fitted normal distribution were corrected to Normal (µ′,σ) where µ′ = µ + σ 2 ρ for the epidemic growth rate R estimated below (personal communication, Neil Ferguson). Because information was available on the periods during which each secondary case had been exposed to their presumptive infector, we were able to identify cases where exposure is likely to have occurred before or after illness onset in the primary case infector. When the exposure window overlapped the onset date, we used our fitted incubation period distribution to resample 1000 infection times at random, excluding any that fell outside the exposure window, and counted the proportion of resampled times that fell before symptom onset in the infector. 

We collected detailed information on 168 confirmed cases, including 145 and 23 reporting that they had or had not, respectively, been in Wuhan in the past 14 days. The demographics of the 145 and 23 cases were very similar (data not shown). The median age was 48 F I G U R E 1 Panel A: occurrence by date of illness onset of cases identified outside of Hubei province in persons with a history of travel from Wuhan in the 14 d prior to onset. Panel B: occurrence by date of illness onset of cases identified outside of Hubei province in persons without a history of travel from Wuhan in the 14 d prior to onset. Panel C: augmented occurrence (yellow bars) by date of illness onset of cases identified outside of Hubei province in persons with a history of travel from Wuhan in the 14 d prior to onset with back filled cases considering delays between illness onset and seeking care and being tested (range 10-81) and 91 (54%) were male. Only three (2%) of the cases who had visited Wuhan also reported visiting the Huanan Seafood Wholesale Market. Figure 1 shows the epidemic curve by illness onset of the cases in persons that had been to Wuhan compared to cases in other persons that were presumed to represent onwards transmission. We used data augmentation to correct for reporting delays in cases with recent onset ( Figure 1C ) and estimated that the growth rate was 0.14 per day (95% CI: 0.11, 0.17), and the doubling time was 5.3 days (95% CI: 4.3, 6.7). In this analysis, the mean onset to reporting delay was estimated to be 5.7 days (95% CI: 5.3, 6.1).

We obtained data on exposure windows for 98 cases in individuals that had been in Wuhan, by assuming that they had been infected while they were in Wuhan. We fitted a lognormal distribution to the data on exposure periods and onset dates, correcting for epidemic growth, and estimated that the incubation period had mean 5. (14%) occurred in those who had a meal together, 2 (3%) were colleagues working together, 2 (3%) were those who took the same vehicle, and 1 (1%) occurred among neighbours.

The identified 80 observations of transmission events were used for estimation of the serial interval ( Figure 2B ). We fitted a normal distribution to these data, with allowing 4 negative serial intervals and correcting for epidemic growth, and estimated that the serial interval distribution had mean 5.7 days (95% CI: 4.7, 6.8) and standard deviation 3.5 days ( Figure 2B ). In detailed investigations of contact patterns, we obtained information on the period when the secondary infection could have occurred, and related this to the illness onset date of the infector. Figure 3 shows that of 80 pairs identified, pre-symptomatic transmission occurred in 9 pairs, post-symptomatic transmission occurred in 16 pairs, and in the remaining 55 pairs, transmission could have occurred either before or after the corresponding infector's illness onset date. In these 55 pairs, we used Monte Carlo simulations based on the incubation period distribution described above to estimate that 23/55 of transmission events had greater than 50% chance to be pre-symptomatic transmissions.

Thus, in total we infer that 32/80 (40%) of the observed transmission events were likely or very likely to have occurred prior to the onset of symptoms in the infector.

We examined demographic and clinical characteristics of all the confirmed sporadic and clusters cases ( Table 1) . Age distributions were similar between the two groups of patients, while less male cases were identified in cluster secondary cases (43% vs 60%).

Cluster primary and sporadic cases were generally more severe, with 26% and 7% in severe and critical conditions, respectively, compared with 12% and 1% in cluster secondary cases. Nearly all the primary and sporadic cases (96%) had a radiologic indication of pneumonia, compared with 74% in the secondary cases. Among 80 cluster secondary cases, 87% had been classified as mild cases, while 66% (110/166) of the cluster primary and sporadic cases were mild. The primary and sporadic cases showed largely similar onset symptoms as the secondary cases. Fever is the most common symptom, followed by headache, fatigue, dry cough and myalgia. Generally higher proportions of the cluster primary cases and sporadic cases reported an onset of the symptoms than cluster secondary cases (Table 1 ). In addition to systemic and respiratory presentations, a small proportion of the cases also reported gastrointestinal symptoms, including vomiting and diarrhoea. 

In this study, we report estimates of the transmission dynamics of COVID-19 based on cases identified outside of Hubei province.

Importantly, we examine quantitative evidence for pre-symptomatic infectiousness, a feature which complicates control strategies. We

Transmission events used to infer the occurrence of pre-symptomatic transmission. Dots indicate the dates of onset of primary and secondary cases, and the shaded area in each row indicates the period of exposure of the secondary case to the primary case during which the secondary infection is thought to have occurred. Brackets indicate the exposure window when the primary case was thought to have been infected. Data were resolved to the nearest day, and so transmission windows are plotted from the start of the first date to the end of the last date, onset dates are plotted in the middle of the corresponding day, and if the secondary onset date is the same as the primary onset date, then the former is offset slightly so that both can be seen Note: For patients having clinical test results of both lymphocyte count and proportion, a normal lymphocyte count refers to both the lymphocyte count within the range of 1-4 × 10 9 /L and the lymphocyte proportion to be 20%-40%. Patients with either lymphoycyte count or lymphocyte proportion lower or higher than the normal range will be classified as "decreased" or "increased," respectively. a Category of lymphocyte count is determined by both the count and proportion of lymphocyte (the proportion is derived as lymphocyte count divided by white blood cell count) in the blood test.

demonstrate that the phenomenon of pre-symptomatic transmission is not uncommon, having occurred in a minimum of 9/80 transmission events ( Figure 3 ). Using statistical inference based on our estimated incubation period distribution, we estimated that pre-symptomatic transmission was likely to have occurred in up to 40% of the 80 transmission events in our data set. This observation should be interpreted in the context of isolation of some cases after illness onset, reducing the amount of post-symptomatic transmission that might otherwise have occurred. This also implies that social distancing measures may be some of the most important strategies to reduce transmissibility, for example closing schools and encouraging, or facilitating working at home, in addition to control of onwards transmission by sequestering symptomatic persons at home or in isolation facilities.

Our finding was consistent with other reports that the serial interval distribution was similar to, or much lower than the incubation period distribution, 13,18-21 consistent with the occurrence of pre-symptomatic transmission in case reports. [22] [23] [24] [25] [26] Pre-symptomatic infectiousness is generally not thought to occur for most respiratory viruses, but measles is a well-known example of a respiratory infection that can be spread before symptom onset, 27 and viral shedding during the incubation period has also been reported for influenza. 28 Viral shedding has been reported during the incubation period for COVID-19, 29 and there were also asymptomatic cases for COVID-19 across all age groups, in varying proportions depending on the intensity of surveillance and testing. 30, 31 In conducting investigations of clusters of cases, it is often found that secondary cases have on average milder illnesses than primary cases; for example, this was noted for Middle East Respiratory Syndrome coronavirus infections. 32 Here, we found that secondary cases were more likely to have milder disease (Table 1) , which would be consistent with the existence of some milder COVID-19

infections, which were not laboratory confirmed. The cluster investigations that we report here are not likely to represent the full spectrum of mild infections, and better information would be provided by prospective studies of close contacts with repeated collection of respiratory swabs and sera.

Our estimates of the growth rate of infections in Wuhan in early January 2020 of around 0.14 per day (95% CI: 0.11, 0.17) are very consistent with previous reports. 4, 6 Our estimate of the effective reproductive number in Wuhan is now slightly lower than the previous estimate of R 0 4 and some other estimates of R 0 5-7 but that is because of the shorter serial interval. We included a limited number of cases in other cities in China that were detected in persons that had not been to Wuhan ( Figure 1B ), but the sample size was insufficient at the time of analysis to determine whether there has been sustained transmission in any other cities.

The incubation period averaged 4.6 days and up to 11.1 days in 95% of infections (Figure 2A) , which is important for specifying quarantine periods and also for understanding transmission dynamics. The city of Wuhan was put under lockdown on 23 January 2020, and other nearby cities were also locked down on the following days.

This stopped exportation of infections to other Chinese cities, and the effect of this intervention has become more apparent in early February, given the delays that occur between infection, illness onset, admission to hospital and then laboratory testing.

We restricted the study participants to COVID- 19 analysis were also subject to ascertainment of symptom onset date, which was self-reported by the patient and could be less reliable especially at the early stage of symptom presentation.

In conclusion, our analysis showed evidence indicative of pre-symptomatic transmission of COVID-19. Quarantine of exposed persons having close contact with infected cases before symptom onset could reduce the risk of further transmission.

We gratefully thank Benjamin J. Cowling, Peng Wu, Jessica Y. Wong, Yiu Chung Lau, Tim K. Tsang from School of Public Health, University of Hong Kong for their guidance and support on study design, data analysis and interpretation. We appreciate the staff members of the county-, city-and province-level Centers for Disease Control and Prevention of China for their fieldwork on case investigation, data collection and validation. 

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

A novel coronavirus from patients with pneumonia in China

A novel coronavirus outbreak of global health concern

Early transmission dynamics in Wuhan, China, of novel coronavirus-infected pneumonia

Pattern of early human-to-human transmission of Wuhan

Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling study

Preliminary estimation of the basic reproduction number of novel coronavirus (2019-nCoV in China, from 2019 to 2020: a data-driven analysis in the early phase of the outbreak

Journey of a Thai taxi driver and novel coronavirus

First case of 2019 novel coronavirus in the United States

Potential presymptomatic transmission of SARS-CoV-2

COVID-19 transmission within a family cluster by presymptomatic carriers in China

Presumed asymptomatic carrier transmission of COVID-19

Serial interval of COVID-19 among publicly reported confirmed cases

New Coronavirus Pneumonia Prevention and Control Program (Second Edition) 2020. in Chinese

Nowcasting the number of new symptomatic cases during infectious disease outbreaks using constrained P-spline smoothing

Alternative methods of estimating an incubation distribution: examples from severe acute respiratory syndrome

How generation intervals shape the relationship between growth rates and reproductive numbers

Serial interval of novel coronavirus (COVID-19) infections

Temporal dynamics in viral shedding and transmissibility of COVID-19

Estimating the generation interval for coronavirus disease (COVID-19) based on symptom onset data

Quantifying SARS-CoV-2 transmission suggests epidemic control with digital contact tracing

Presymptomatic transmission of SARS-CoV-2 -Singapore

Delivery of infection from asymptomatic carriers of COVID-19 in a familial cluster

Asymptomatic and presymptomatic SARS-CoV-2 Infections in residents of a long-term care skilled nursing facility

A COVID-19 transmission within a family cluster by presymptomatic infectors in China

Transmission of COVID-19 in the terminal stages of the incubation period: a familial cluster

Evidence base of incubation periods, periods of infectiousness and exclusion policies for the control of communicable diseases in schools and preschools

The dynamic relationship between clinical symptomatology and viral shedding in naturally acquired seasonal and pandemic influenza virus infections

Viral load of SARS-CoV-2 in clinical samples

Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship

A well infant with coronavirus disease 2019 (COVID-19) with high viral load

Middle East respiratory syndrome coronavirus: quantification of the extent of the epidemic, surveillance biases, and transmissibility

Evidence for pre-symptomatic transmission of coronavirus disease 2019 (COVID-19) in China