Estimating SARS-CoV-2 variant strain infectiousness in Japan as of March 28, 2021

infectiousness, reproduction number, SARS-CoV-2, variant strain, 13 VOC202012/01 14 15 16 17 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2

18

A variant strain of SARS-CoV-2, VOC202012/01, emerged in the UK 19 in September, 2020. Its infectiousness was estimated as higher than that of the original 20 strain. 21 Object: We estimated the infectiousness of the variant strain of SARS-CoV-2 in 22 comparison to that of the original strain under conditions prevailing in Japan. 23

We estimated infectiousness through a simple susceptible–infected–recovered 24 (SIR) model by strain. The study period was March 1–28, 2020. The information used 25 for the study was available as of April 3, 2020. 26

The estimated reproduction number of the SARS-CoV-2 variant strain was 27 1.799; its 95% CI was [1.642, 1.938]. The onset date of the first case in Japan was 28 estimated as December 4 ([November 16, December 14]), 2020. However, 29 infectiousness of the original strain was estimated as 1.123 ([1.093, 1.166]). 30

and Conclusion: We demonstrated that infectiousness increased by 0.684 or 31 60%, increasing from the original to the variant. That finding might be comparable to 32 that of a study conducted in the UK. However, the difference must be monitored 33 continuously and carefully. 34 35 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 3

36 37 A variant strain of S ARS- COV - 2, VOC202012/01, emer ge d in the UK in 38 September 2020. Its infectiousness and pathogenicity were estimated as 75% higher 39 than those of the original strain, which was previously circulating before emerging 40 variant strain[1]. 41 In Japan, the first case was detected on January 6 as an imported case from the UK. 42 Subsequently until the end of March, 2021, community outbreak of the variant strain 43 occurred in Japan. Since the beginning of March, 2021, the outbreak was remarkable in 44 Japan. Actually, the Ministry of Health, Labour and Welfare reported prevalence of the 45 variant strain in the community outbreak as 7, 10, 16, and 20% for each week of March 46 in Japan [2]. 47 However, the estimated infectiousness of the variant strain was not evaluated 48 outside Europe. Even for the outbreak caused by the original strain, outbreak situations 49 differed among countries. For example, prevalence of SARS-COV-2, with mainly the 50 original strain, in the USA were 9361 and 6558 cases in the UK per 100 thousand 51 residents, but in Japan it was limited to 372 as of the end of March, 2021 [3]. 52 Infectiousness of the variant strain might differ country-by-country, as that of the 53 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 4 original strain has. Therefore, we evaluated the infectiousness of the variant strain 54 compared to that of the original strain under the same conditions using a simple 55 deterministic SIR model in Japan. 56 57

58 59 We defined the variant strains as mutated at N501Y and/or E484K together because 60 the numbers of patients were not reported by subtype among variant strains. Therefore, 61 we cannot specifically examine differences among variant strains. 62 For a simple deterministic SIR model [4–6] applied to the epidemic curve for 63 Japan’s 120 million populations, we used the numbers of symptomatic patients reported 64 by strain from MHLW for March 1–28, 2021, which were published [7] on April 4, 65 2021. We excluded patients who were presumed to be infected abroad because those 66 patients presumably do not represent cases of community-acquired infection in Japan. 67 We assumed that the two strains had similar properties, except for infectiousness. 68 Because the SIR model is a system reflecting daily differences, we converted weekly 69 data simply to daily data divided by 7. 70 The incubation period is assumed to conform to the empirical distribution in Japan. 71 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 5 Letting the probability of onset on day k after exposure in the empirical distribution be 72 p(k), then p(k)/Π m=0 k-1(1-p(m)) of infected patients showed no symptom until k-1 days 73 from exposure showed onset on day k. That is, 1- p(k)/Π m=0 k-1(1-p(m)) of them showed 74 onset on k+1 day after exposure, or later. 75 For symptomatic patients with unknown onset date, we estimated the onset date as 76 follows: Letting f(k) represent the probability of a k day delay in the empirical 77 distribution of reporting delay from onset and letting Nt denote the number of patients 78 for whom onset dates were not available for publication at date t, then the number of 79 patients for whom the onset date was known is t−1. The number of patients for whom 80 onset dates were not available was estimated as f(1)Nt. Similarly, the number of patients 81 with onset date t−2 and for whom onset dates were not available were estimated as 82 f(2)Nt. Therefore, the total number of patients for whom the onset date was not available, 83 given an onset date of s, was estimated as Σ k=1f(k)Ns+k for the long duration extending 84 from s. 85 Moreover, the reporting delay from the Ministry of Health, Labour and Welfare 86 (MHLW) might be considerable. In other words, if s+k is larger than that in the current 87 period t, then s+k represents the future for period t. For that reason, Ns+k is not 88 observable. Such a reporting delay engenders underestimation bias of the number of 89 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 6 patients. For that reason, it must be adjusted as Σ k=1 t-sf(k)Ns+k /Σ k=1 t-sf(k). Similarly, 90 patients for whom the onset dates were available are expected to be affected by the 91 reporting delay. Therefore, Ms|t /Σ k=1 t-sf(k), where Ms|t represents the reported number of 92 patients for whom onset dates were within period s, extended until the current period t. 93 We used these arranged data to produce the epidemic curve for SIR modelling. 94 The distribution of infectiousness in symptomatic cases was assumed to be 30% on 95 the onset day, 20% on the following day, and 10% for the subsequent 5 days [8]. If 96 designating the distribution of infectiousness as g(k), where k represents days from 97 onset and the number of symptomatic patients on day t and onset was k days ago as 98 I(t,s), then the number of newly infected patients on day t was R(t) S(t)Σ k=0 g(k) I(t,k) 99 /Y(t), where S(t) denotes the number of susceptible population and Y(t) represents the 100 total population on day t. 101 We run the simulation from the initial case under the presumed reproduction 102 number. Simulation results of the epidemic curve, which was defined as the number of 103 patients by onset date, was converted to the predicted reporting number of 104 laboratory-confirmed patients by the empirical distribution of reporting delay from 105 onset. To that was added a constant proportion of asymptomatic patients, 20%, 106 assuming no infectiousness among asymptomatic patients [8]. 107 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 7 If the number of the estimated reported laboratory-confirmed patients on a day in 108 the simulation was higher than the actual data on March 1, then we compared the model 109 prediction and observation for four weeks to seek reproduction number by types to 110 achieve the minimum absolute difference among them through grid search on [1,2] by 111 0.0001. The bootstrapping procedure we used was fully replicated bootstrapping for a 112 fixed number of initial cases. There were N patients in the data, with numbering of the 113 patients from the initial case to the last case. Initially, no patient was on the 114 bootstrapped epidemic curve. If a random variable drawn from a uniform distribution of 115 (0,1) was t included in the internal [i/(N-1), (i+1)/(N-1)], then we added one to the onset 116 date of the i+1th patient to the bootstrapped epidemic curve. We replicated this 117 procedure N-1 times. Thereby, we obtain the bootstrapped epidemic curve with N-1 118 patients. Finally, we added the initial patient, for whom the onset date was January 14, 119 to the bootstrapped epidemic curve. Consequently, we obtained a bootstrapped 120 epidemic curve with N patients. 121 The study period was March 1–28, 2021 for estimating the infectiousness of the 122 two strains. The information used for the study was available on April 4, 2021. The 123 empirical distribution of the incubation period and reporting delay were based on the 124 outbreak in Japan from January 14, 2020 to the end of February, 2021. 125 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 8 126

127 128 In the four weeks I March 2021, 7264, 7940, 8960 and 12045 laboratory- confirmed 129 cases including variant strain cases as well as the original strain cases were reported [7]. 130 Multiplying proportion of the variant strain described above to these number, the 131 weekly number of reported laboratory-confirmed patients by strain in March, 2021 were 132 shown in Figure 1. One seventh of these data were used for the following analysis as 133 daily data. 134 Figure 2 presents an empirical distribution of the duration of onset to reporting in 135 Japan. The maximum delay was 31 days. Figure 3 depicts the empirical distribution of 136 incubation periods among 125 cases for which the exposed date and onset date were 137 published by MHLW in Japan. The mode was 6 days. The average was 6.6 days. 138 The estimated reproduction number of the variant strain of SARS-CoV-2 was 139 1.799; its 95% CI was [1.642, 1.938]. The onset date of the first case in Japan was 140 estimated as December 4 ([November 16, December 14]), 2020. However, the 141 infectiousness of the original strain was estimated as 1.123 ([1.093, 1.166]). Figures 4 142 and 5 show the fitted line and its 95% CI of the number of laboratory-confirmed cases 143 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 9 by the model for the two strains. Although stepwise data were difficult to mimic by the 144 model, bands of CI fairly covered the observation data. 145 146

147 148 Similar differences of infectiousness were found for the variant strain and the 149 original strain, as described for an earlier study [1]. The infectiousness increased by 150 0.684, indicating a 60% increase from the original to the variant. The rate of increase 151 was slightly lower than that presented in an earlier report [1]. It might reflect the 152 outbreak situation including vaccine coverage. 153 The vaccine coverage in March, 2021 in Japan was less than 1%. Therefore, it 154 probably did not affect the outbreak at all. However, with increasing vaccine coverage 155 in Japan, differences among strains are not expected to be affected if the vaccine 156 effectiveness is equal for all strains. Alternatively, differences among strains will 157 become increasingly evident if vaccine effectiveness against variant strains is less than 158 the effectiveness against the original strain. 159 We estimated the onset date of the initial case of the community outbreak by a 160 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 10 variant strain as December 4 ([November 16, December 14]), 2020. Actually, the first 161 case of the variant strain was detected on January 6, 2021 as an imported case from UK. 162 Therefore, we predicted the initial case of the variant strain as earlier than one month 163 from the officially reported imported case. As one might expect, the earlier stage of the 164 outbreak in the variant strain was not recognized in December, 2020 and January, 2021 165 because viral genomic sequencing, which can detect a variant strain, was not applied to 166 the specimen from the community outbreak in Japan. Therefore, we cannot deny the 167 possibility that a community outbreak by the variant strain occurred earlier than 168 December, 2020, in Japan. The earliest case with the variant strain in the UK was 169 reportedly identified on September 20, 2020 [1]. Therefore, no community outbreak was 170 likely to have started in Japan two months later than in the UK. 171 The present study has some limitations. First, because of data limitations, some 172 differences among variant strains were not considered. However, examination of those 173 differences might be important for predicting which subtype of variant strain is 174 dominant in the outbreak. 175 Secondly, the obtained result reflects information up to March, 2021. Changes in 176 climate or mobility might have affected the results. In addition, as described above, 177 increasing vaccine coverage might eventually affect results. Therefore, it is necessary to 178 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 11 monitor differences of infectiousness among strains carefully and continuously. 179 180

181 182 We demonstrated that infectiousness of the variant was greater by 0.684, or 60%, 183 compared to the original strain of COVID-19. The findings might be comparable to 184 those of a study conducted in the UK [1]. Nevertheless, careful and continuous 185 monitoring of differences among strains is necessary. 186 The present study is based on the authors’ opinions. Its results do not reflect any 187 stance or policy of their professionally affiliated bodies. 188 189 Acknowledgments 190 191 We acknowledge the great efforts of all staff at public health centers, medical 192 institutions, and other facilities who are fighting the spread and destruction associated 193 with COVID-19. 194 195 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 12

196 1. Leung K, Shum MHH, Leung GM, Lam TTY , Wu JT. Early transmissibility 197 assessment of the N501Y mutant strains of SARS-CoV-2 in the United Kingdom, 198 October to November 2020. Euro Surveill 2021;26:2002106. doi: 199 10.2807/1560-7917.ES.2020.26.1.2002106. 200 2. Ministry of Health, Labour and Welfare, Implementation rate and positive rate of 201 screening test for the variant strain (mechanical estimation) preliminary figures. 202 https://www.mhlw.go.jp/content/000766381.pdf (in Japanese) [accessed on April 9, 203 2021] 204 3. World Health Organization. Coronavirus disease (COVID-19) Weekly 205 Epidemiological Update and Weekly Operational Update 206 https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-report207 s [accessed on April4, 2021] 208 4. Kurita J, Sugawara T, Ohkusa Y . Estimated effectiveness of school closure and 209 voluntary event cancellation as COVID-19 countermeasures in Japan. J Infect 210 Chemother 2021, 27:62-4. doi: 10.1016/j.jiac.2020.08.012. Epub 2020 Aug 19. 211 5. Sugishita Y, Kurita J, Sugawara T, Ohkusa Y. Effects of voluntary event 212 cancellation and school closure as countermeasures against COVID-19 outbreak in 213 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 13 Japan. Plos one 2020. https://doi.org/10.1371/journal.pone.0239455 214 6. Ohkusa Y, Sugawara T, Taniguchi K, Okabe N. Real-time estimation and 215 prediction for pandemic A/H1N1(2009) in Japan. J Infect Chemother. 2011; 17: 216 468-472. 217 7. Japan Ministry of Health, Labour and Welfare. Press releases. Available from: 218 https://www.mhlw.go.jp/stf/newpage_10723.html (in Japanese) [accessed on April 219 4, 2021] 220 8. Cao S, Gan Y, Wang C, Bachmann M, Wei S, Gong J, Huang Y, Wang T, Li L, Lu 221 K, Jiang H, Gong Y, Xu H, Shen X, Tian Q, Lv C, Song F, Yin X, Lu Z. 222 Post-lockdown SARS-CoV-2 nucleic acid screening in nearly ten million residents 223 of Wuhan, China. Nat Commun 2020;11:5917. doi: 10.1038/s41467-020-19802-w. 224 225 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 1 Fig. 1. Number of laboratory confirmed cases infected by original and variant strain of 1 SARS-CoV-2 in March, 2021 reported by MHL W, Japan. 2 (Cases) 3 4 ( D a y s i n M a r c h , 2 0 2 1 ) 5 Note: Orange bars represent numbers of laboratory-confirmed cases infected by the 6 original strain weekly. Blue bars represent patients infected by the variant strain. 7 Laboratory-confirmed cases included asymptomatic and presymptomatic cases. Dates 8 are the reported date: not the onset or infected date. 9 0 2000 4000 6000 8000 10000 12000 1-7 8-14 15-21 21-28 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 1 Fig. 2. Empirical distribution of durations from onset to report by MHLW, Japan. (%) 1 2 Note: Bars represent the probabilities of duration from onset to report based on data of 3 657 patients in Japan for whom onset dates were available. The horizontal axis shows 4 the duration from exposure to onset in days. The vertical axis shows the probability of 5 the incubation period. Data were obtained from MHLW, Japan. 6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 1 Fig. 3. Empirical distribution of the incubation period published by MHLW, Japan. (%) 1 (Cases) 2 3 ( D a y s ) 4 Notes: Bars show the distribution of incubation periods for 125 cases for which the 5 exposure date and onset date were published by MHLW, Japan. The horizontal axis was 6 measured in duration from exposure to onset in days. The vertical axis was the 7 probability of the incubation period. Patients for whom incubation was longer than 14 8 days are included in the bar shown for day 14. 9 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 1 Fig. 4. Observed data of the original strain and prediction with 95% CI 1 (Number of cases) 2 3 (Days in March, 2021) 4 Note; Bars represent the reported number of the test confirmed cases. The bold line 5 represents its prediction by the model. Thin lines represent its 95% CI. 6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint 1 Fig. 5. Observed data of the variant strain and its prediction with 95% CI. 1 (Number of cases) 2 3 (Days in March, 2021) 4 Note: Bars represent the reported numbers of the test confirmed cases. Bold line 5 represents its prediction by the model. Thin lines represent 95% CI. 6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted April 17, 2021. ; https://doi.org/10.1101/2021.04.11.21255283doi: medRxiv preprint