key: cord-0994130-rzaydn7o authors: Berumen, Jaime; Schmulson, Max; Guerrero, Guadalupe; Barrera, Elizabeth; Larriva-Sahd, Jorge; Olaiz, Gustavo; Garcia-Leyva, Rebeca; Wong-Chew, Rosa M; Betancourt-Cravioto, Miguel; Gallardo, Hector; Fajardo-Dolci, German; Tapia-Conyer, Roberto title: Trends of SARS-Cov-2 infection in 67 countries: Role of climate zone, temperature, humidity and curve behavior of cumulative frequency on duplication time date: 2020-04-23 journal: nan DOI: 10.1101/2020.04.18.20070920 sha: 31a8c61158e3d1c3c2dc4c0f1a16ca29deee4383 doc_id: 994130 cord_uid: rzaydn7o Summary Objective. To analyze the role of temperature, humidity, date of first case diagnosed (DFC) and the behavior of the growth-curve of cumulative frequency (CF) [number of days to rise (DCS) and reach the first 100 cases (D100), and the difference between them (ΔDD)] with the doubling time (Td) of Covid-19 cases in 67 countries grouped by climate zone. Design. Retrospective incident case study. Setting. WHO based register of cumulative incidence of Covid-19 cases. Participants. 1,706,914 subjects diagnosed between 12-29-2019 and 4-15-2020. Exposures. SARS-Cov-2 virus, ambient humidity, temperature and climate areas (temperate, tropical/subtropical). Main outcome measures. Comparison of DCS, D100, ΔDD, DFC, humidity, temperature, Td for the first (Td10) and second (Td20) ten days of the CF growth-curve between countries according to climate zone, and identification of factors involved in Td, as well as predictors of CF using lineal regression models. Results. Td10 and Td20 were ≥3 days longer in tropical/subtropical vs. temperate areas (2.8[plusmn]1.2 vs. 5.7[plusmn]3.4; p=1.41E-05 and 4.6[plusmn]1.8 vs. 8.6[plusmn]4.2; p=9.7E-05, respectively). The factors involved in Td10 (DFC and ΔDD) were different than those in Td20 (Td10 and climate areas). After D100, the fastest growth-curves during the first 10 days, were associated with Td10<2 and Td10<3 in temperate and tropical/subtropical countries, respectively. The fold change Td20/Td10 >2 was associated with earlier flattening of the growth-curve. In multivariate models, Td10, DFC and ambient temperature were negatively related with CF and explained 44.7% (r2 = 0.447) of CF variability at day 20 of the growth-curve, while Td20 and DFC were negatively related with CF and explained 63.8% (r2 = 0.638) of CF variability towards day 30 of the growth-curve. Conclusions. The larger Td in tropical/subtropical countries is positively related to DFC and temperature. Td and environmental factors explain 64% of CF variability in the best of cases. Therefore, other factors, such as pandemic containment measures, would explain the remaining variability. 0.10% in the Netherlands despite a much higher population density of 508 inhabitants per Km 2 or 0.11% in Germany with a 240 inhabitant density per Km 2 . 6 In contrast, in the Americas, up to the same date in April, the United States reported the highest number of cases with a total of 327,920. However, this only comprised 0.01% of its total population, and Canada had a 0.04% infectious rate. The data from these North American countries, were low, but still 10 times more frequent than in the rest of the countries of Central and South America, with rates ranging from 0.04% in Panama, 0.001% in Mexico, 0.002% in Colombia and 0.0003% in Guatemala. 6 The above described infectious rates, decreasing from Europe to North America and South America, is in agreement with the time frame of the spreading path that this pandemic has followed. 5 However, the much lower infection rates in South America, compared to the USA and Canada as well as that in Europe, may also reflect an underestimation on the illness, although other variables including climatologic differences deserve to be studied. First, the cold winter season and severe drought were both present in China during the first SARS 2002 and the current COVID-19 epidemics. 4 Second, the combination of cold and dry environment seem to be more adequate than cold weather by itself for viral transmission. 7 Even more, cold temperature and lower humidity reduce both ciliary movement and the mucous secretion as defensive mechanism to remove particles such as infectious agents, and alters the integrity of the nose-mucosal epithelium. Overall, cold and dry environment may enhance the susceptibility to COVID-19 infections. 4, 8 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint In addition, cold season and low temperatures cause stress to the human organism impairing the immune system, increasing norepinephrine and cortisol levels, producing lymphocytosis and decreasing the lymphoproliferative responses, and TH1 cytokine levels and salivary IgA hemostasis. 8 These mechanisms may be predisposing factors to acquire viral infections in winter. 8 Although in Europe and North America, including the USA and Canada, cold weather remains even as spring season has begun, in the northern countries of Latin America such as Mexico, the predominant climate is warm humid weather. And in more southern countries such as Argentina, it is autumn with template weather. Thus, it is plausible that the higher temperatures and humidity in these countries, explain the much lower infectious rates of COVID-19 when compared to the Northern hemisphere. 9 Also, viral infections have cyclic patterns such as the annual flu or the human respiratory syncytial virus epidemic in which hosts are more susceptible during the winter months. 10 However, not all viruses have the same seasonality, and some spread during the summer season (e.g. Enteroviruses), other have a spring/fall cycle (e.g. Rhinovirus), or all year round (e.g. Adenovirus). 11 Also, transmission can occur in different weather patterns. 12, 13 which is especially true for tropical regions in which there is no clear seasonal pattern. 14 When a new virus such as the SARS CoV2 arise with no immunity for humans, and with a great capacity to disseminate, the seasonality might or might not be a determinant factor for its spread. 15 Therefore we aimed at analyzing the role of climate areas, outside temperature, humidity, the number of days it took for the spreading curve to raise or elevate (DCS) and to reach the first 100 cases (D100) and the difference of the latter ones (ΔDD) 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint with the duplication speed of infected patients, in 67 countries around six continents. We hypothesized that the elevation of the curve, the time to reach the initial 100 cases, as well as the time for duplicating the number of cases, should be faster in number of days in countries within lower temperate areas when compared to those in tropical or subtropical ones. The study populations are the daily confirmed newly diagnosed cases of COVID-19 officially reported by the WHO from 67 countries, 18 located in temperate or cold areas, and 49 in tropical or subtropical regions from December 29, 2019 to April 15, 2020. The population data was collected from the reports released on the official websites of the World Health Organization about Covid-2019. 16 The cumulative frequency (CF) and the date when the first case was diagnosed in each country were obtained. Thus, no ethical review was required. The average temperature and relative humidity were collected from Time and Date The CF of Covid-19 cases of each country was plotted in Excel and the exponential equation was obtained. The days of the curve rise (DCS) and of reaching the first 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint 100 cases (D100), as well as the difference between these two days (ΔDD) were graphically identified with the WHO data. The duplication time (Td) of the number of positives was obtained from the slope (λ) of the exponential graph (N=Noe λt ) as follows: Td=ln(2)/λ. 18 It was calculated for the first (Td10), second (Td20) and third (Td30) 10-day periods of the CF curve, as well as for the entire 30-day period (TdT), starting from day D100. Descriptive analysis was performed. Numerical variables were described with medians and interquartile range (IQR) or means and standard deviations. The variables were compared between the groups (temperate vs. tropical/subtropical zones). The significance of differences between the groups was assessed with the Mann-Whitney U-test or the t-test. The Pearson´s correlation was performed for some numerical variables. The association of significant variables with Td was explored using univariate (ULR) and multivariate (MLR) linear regression models. Finally, we built models to predict the CF of Covid-19 cases expected to happen on day 20 and 30 of the pandemic growth curve with the Td and the rest of variables using ULR and MLR models. To enter the date of the first Covid-19 case diagnosed and the climate zone in the linear regression models the nominal data were transformed into numeric values, as follows: 1) the dates of the first Covid-19 cases diagnosed in the 67 countries were sorted in ascending order, and numerical values in ascending order, starting with the number 1, were assigned to each date, 2) for the climate zones, the values of 1 and 2 were assigned to temperate and tropical/subtropical zones, respectively. 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint The association was expressed as the β coefficient and 95% confidence interval (CI), and the contribution to the variability of Td or CF was expressed as adjusted r 2 . Confounders were identified using a theoretical strategy based on a backstep, stepwise MLR model and the change-in-estimate criterion. Variables with p < 0.20 in the univariate analysis were considered for entry in the multivariate model. Confounders were defined as those variables for which the percentage difference between the values of the regression β and the adjusted and non-adjusted variables in the stepwise multivariate model were higher than 10% (p>0.1). Therefore, the total variability, the contribution of each factor, and interaction between the factors on Td or CF was calculated using this MLR models. The factors and interactions were included in the model in one block. A post hoc power analysis was performed for each linear regression model using the software G * Power 3.1.9.2, considering the sample size, the β and an α = 0.05. In addition, for MLR models, the value of the total r 2 obtained at the end of the model was introduced for power calculation. All statistical tests were two-sided, and differences were considered significant when p < 0.05. The statistical analyses were conducted using SPSS version 20 software (SPSS Inc., Chicago, IL, USA). Analysis of the initial phase of the growth curve. Figure 1 shows how the CF grew from the day the first case of Covid-19 was diagnosed in 67 countries. The day the curve started to ascend (DCS) and the slope of the curves were observed. There is a wide variation between countries, but similar 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. Man-Whitney U-test; Figure 2 panel A). On the other hand, the slope of the curves was much higher in temperate than in tropical/subtropical countries ( Figure 1 ). An indirect way to evaluate the slope, is to measure the number of days it takes for the curve to reach 100 cases, once the ascend begins. Thus, we identified the number of days it took for the curves to reach 100 cases (D100) from the diagnosis of the first case and subtracted the DCS to obtain the difference in days between the two points (ΔDD). While the D100 is much higher [median (IQR) 32 vs. 20 (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) ; p=0.108, Man-Whitney U-test; Figure 2 panel B], the ΔDD is much lower in temperate countries [median (IQR) 3 (2-5) vs 7 (6-9); p=0.002, Man-Whitney U-test; (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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint correlation; Figure 3 panel B); that is, the later the first case was diagnosed, the faster the DCS and D100 days were reached; and vice versa, in countries where the first Covid-19 case was diagnosed very early in the year, for example in January, the DCS and D100 were longer (Figure 3 The cumulative frequency, converted to log10, was plotted against the number of days of evolution of the epidemic in each country from D100 ( Figure 4 ). In addition, doubling times of 1 to 4 days were calculated and included in the graph to locate the growth curve of each country between these intervals (black dotted lines in Figure 4 ). Changes in the trends of the accumulated frequencies over time are represented. For example, Figure 4 shows how the epidemic quickly grew in China, Korea and Iran during the first 10 days closer to Td= 1 day (dotted line, Figure 4 ), then the curves start to lie down after the 10th and 20th days of the evolution of the epidemic. Something similar was seen in the European countries. To compare the growth curves in more detail, the doubling time was calculated for the first (Td10), second 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint (Td20), and third (Td30) 10 days, and for the entire 30-day period (TdT) in each country. The TdT was lower in temperate areas when compared with tropical/subtropical areas [4.2 ± 1.5 vs. 6.5 ± 3.1; p= 3.08E-04]; this difference became greater when comparing the Td10 (2.8 ± 1.2 vs. 5.7 ± 3.4; p= 1.41E-05) and the Td20 (4.6 ± 1.8 vs. 8.6 ± 4.2; p= 9.7E-05) (all with t-test). This clearly indicates that Td is lower in temperate countries. In addition, Figure 5 panels A and B depict the individual Td10, Td20 and Td30 from several temperate and tropical or subtropical countries. Countries that had a very rapid increase in the growth curve such as Italy, Spain, China, Korea and Iran (see Singapore ( Figure 5 panel B), which indicates that the rate for doubling the number of infected people was lower in most countries during days 11 to 20 of the curve. In the case of the USA, it is possible that the infection duplication rate grew during that time period. In fact, the Td20 falls below 2, and as seen in Figure 4 , the slope of the curve increases from day 7 and the curve ascends almost as a straight line between Td= 2 and Td= 3. Something similar happened with Singapore ( Figure 4) , however, the curve remained below the Td= 4 line, which indicates that the Td > 4 days. In temperate countries, where the value of Td20 is two times that of Td10 [Fold Change (FCTd20/Td10)], such as Korea (FCTd20/Td10 = 4.5), Sweden (FCTd20/Td10= 2.5), Norway (FCTd20/Td10= 2.6), or that have a close value as in the case of the Czech Republic (FCTd20/Td10= 1.9; Figure 5 ), coincides with the early flattening of the curves, growing towards the area of Td= 4 or to a larger Td ( Figure 4 ). It seems clear that the increase of the Td20, more than Td30, is essential for the early flattening of the 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint curve. Examples of this include Austria (FCTd20/Td10= 1.6) and Switzerland (FCTd20/Td10= 1.5) that have a high Td30 but did not have a FCd20/Td10 > 2, and the curves flattened later. This is even more difficult for Spain, France and Italy that started with a very low Td10 and FCTd20/Td10= 1.7. In tropical and subtropical areas, in addition to Iran that had a Td10= 1.8, countries that had a Td10 ≤ 3, such as Brazil, Chile, Ecuador and South Africa, had fastgrowing curves during the first 10 days. However, most of them, except for Chile, had a FCTd20/Td10 ≥ 2. In conclusion, it seems clear that the Td10 value is closely related to the initial momentum of the curve growth, with a cut-off point of Td10= 2 and Td10= 3, for temperate and tropical/subtropical countries, respectively, above which the growth of the curve is more manageable. The other important time-point is Td20, and when its value is ≥ 2 times that of Td10, there is a trend for flattening the curve earlier than when it is lower than 2. The relation of each variable (DCS, D100, ΔDD, first case date, average relative humidity, average temperature, climate zone) with Td10 and Td20 was investigated in ULR models. Variables that had a significance of p< 0.2 were selected to be introduced in multivariate models, one for the Td10 and one for the Td20, and variables were selected in the final models when they remained in the MLR analysis with a p< 0.1. In addition, Td10 was included in the Td20 model ( Table 1) . Five of these variables were individually associated with the Td10, four showed a positive association (ΔDD, date of first case diagnosed, average temperature and climate 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 preprint this version posted April 23, 2020. in fact, they also showed the highest values of r 2 (0.237 and 0.182, respectively), which multiplied by 100, indicated the percentage of Td10 variability explained by that variable. However, in the MLR analysis, the only variables that remained in the model were the date of the first case diagnosed (β= 0.077 95%CI: 0.037 to 0.116, p= 2.6E-04) and the ΔDD (β=0.242 95%CI: 0.097 to 0.388, p= 1.5E-03). That is, for each day that the ΔDD increases, the Td10 increases by 0.242 days and for each day that passes from the diagnosis of the first case, the Td10 increases 0.077 days. These two variables explain 29.7% of the variability of the Td10 (r 2 = 0.297) ( Table 1 ). It is not rare that temperature stands out of the model, as there is a very high correlation between temperature and ascending dates in the calendar from January to March. In fact, when the date of the first case is not introduced into the MLR model, the temperature remains significant in the model (data not shown). However, the exclusion of the temperature in the model and the permanence of the date of diagnosis of the first case may have an additional explanation. In the ULR models for Td20, all variables are significant, except relative humidity and ΔDD. But only four variables remained in the MLR model, including Td10 (β=0.999 CI95%: 0.468 to 1.53, p= 4.91E-04) and the climate zone (β=5.7 CI95%: 2.396 to 9.012, p= 1.22E-03), which appeared to be the most important factors influencing the value of Td20 (Table 1 ). For each day that the Td10 increases, the Td20 increases by 0.999 days, conversely, for each day the Td10 decreases, the 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint Td20 will decrease by 0.999 days. In tropical/subtropical countries, Td20 increases 5.7 days above the Td20 of temperate countries. For Td20, the temperature no longer seems to be an important factor, while important components of the initial curve behavior remained, such as DCS and D100 (Table 1 ). All these factors explain 42.6% (r 2 = 0.426) the variability of Td20. When the model is stratified by zone, the Td10 variable remains in both area models as the most important variable (data not shown). We explored whether the Td10, Td20, and TdT, with the other variables, could predict the Covid-19 CF that would be reached 20 and 30 days after the day D100 in the growth curve using ULR and MLR models. Variables that had a p=0.2 in the ULR models were selected to be introduced into the MLR model. A clear relationship of Td10 with CF at day 20 is observed in the ULR model (r 2 = 0.189), as discussed above in relation to Figures 4 and 6 , however, the relationship of TdT to the CF is much higher (r 2 = 0.268), therefore this variable of the doubling time was the one introduced in the MLR model (Table 2) . Also, ΔDD, date of first case, the average relative humidity, the average temperature and climate zone passed the cut-off value in the ULR models (Table 2) (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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint such as epidemic containment measures, would explain more than 55% of the remaining variability. For the prediction of the CF at the 30th day of growth curve the model has a better performance ( Table 2) (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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint Discussion. In this study we disclosed that the behavior of the growth curve during the period between the diagnosis of the initial and the first 100 cases, especially the day the curve begin to rise (DCS), the day the first 100 cases (D100) were reached and the difference between these two time points (ΔDD), were related to temperature, the date of the first case and the doubling time (Td). In addition, these values were substantially different between countries located in temperate and tropical/subtropical areas, especially the Td during the first ten days (Td10), after D100, was on average 3 days longer in tropical/subtropical than in temperate countries. We also identified that the factors involved in Td the first ten (Td10) days (date of first case diagnosed and ΔDD) are different than those involved the second ten (Td20) days (mainly Td10 and climate zone) of the growth curve. The fastest growth curves during the first 10 days, after D100 day, were associated with Td10 <2 and Td10< 3 in temperate and tropical/subtropical countries, respectively. And the fold change Td20/Td10 > 2 was associated with earlier flattening of the growth curve. In the MLR predictive models, the Td10, the date of the first Covid-19 case diagnosis and the ambient temperature, were negatively related to the cumulated frequency for the 20 th day of the growth curve, while only Td20 and date of the first Covid-19 case were negatively related to the CF reached on day 30 of the growth curve, although for the latter, only the data of temperate countries was used. 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint Considering the fact that the Td was calculated based on the CF reported daily by the WHO, there is great uncertainty about the variability of this information with the reality in each country depending on the number of tests carried out in each of them. The number of tests have been changing and we do not have a reliable information about the time the mass use of SARS-Cov-19 screenings began in each country, nor how many tests were performed by days 10, 20 and 30 days of the growth curve, after the day that case 100 (D100) was reached. Moreover, the promptness and type of containment measures adopted by each country could have also affected the Tds. However, we can discuss which data from our analysis could have little or a lot of deviation related to the number of tests and containment measures, and therefore the importance of the findings. Notably, the findings from the analysis of the first stage of the growth curve, from the day on which the first case was diagnosed until the first 100 cases of Covid-19 were diagnosed, as well as the date on which the first diagnosis was made and probably the value of Td10, would have a modest influence from the number of tests or containment measures that were carried out in the different countries, since most of them were similar. The date of the first case diagnosed, the average differences in the day on which the growth curve raised (DCS) and from there to the time of the first 100 cases (ΔDD) were significantly different among temperate and tropical/subtropical countries. Although the date of the first case diagnosed is directly related to temperature, this variable remains independent and heavier in the MLR models associated to Td10 and CF. The above suggest that the date of the first case is not only related to ambient temperature but perhaps also related to the way the infection was disseminated. The fact that the epidemic reached the 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 preprint this version posted April 23, 2020. when the infection spread from China, was 58.5 ± 15.2 days. The question is why in these countries the cumulated positive cases curve rose rapidly but with a smaller slope than in countries of temperate zones, and why in the later ones, once the rise of the curve began, the slope was steeper. In discussing the last issue first, an important possibility could be that in European and other temperate countries, asymptomatic individuals were the first to accumulate and these subjects spread the infection. 19 This seems very feasible as up to 80% of infected individuals are asymptomatic. 20, 21 22, 23 In addition, common use of mass transportation and small housing spaces in European cities or those such as New York with a high population density, could have facilitated infection dissemination. Further, the spread could have then affected the older population which in temperate countries make up a much higher percentage of the population pyramid than in countries of tropical/subtropical areas. For example, in countries like Spain and Italy the 65 years and older age-group, constitutes 19% and 23% of their populations, respectively. Conversely, the same age-group represents only 7% of the Mexican and Ecuadorean populations. The older population may have greater susceptibility to infection in cold climates, and it has been suggested that social and biological factors 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 preprint this version posted April 23, 2020. Health of this country (45 years old), 28 than that in China (56 years old) 29 and Italy (64 years old). 30 In addition, the expectation in these tropical/subtropical countries, is that there is a protective effect of ambient temperature, at least on the airway function. 8 Temperature variation as a function of climatic status has a profound influence in virtually all stages of the host virus interaction; 31-33 therefore, the relationship of ambient temperature and doubling time depicted here becomes highly relevant. There is a direct effect of temperature on the enzyme-mediated reactions resulting in cell homeostasis and immune responses. 31, 34 More explicitly, the frequency of 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint upper respiratory tract viral infections relates inversely to every degree Celsius that the temperature drops. 31, 33, 34 Although available information strongly suggests a climatic influence on Covid-19 infection and spreading, 18, 32 the novelty of this illness, warrants further assessment of this issue. While the information gathered herein provides further support to the notion that an overall increased temperature and humidity limits the spreading and occurrence of Covid-19, systematic studies are required to ruled out possible ethnical and/or genetical influences 35 upon the spreading of and susceptibility to the disease. In this regard, as mentioned earlier, the human ACE2 receptor has now been recognized as the receptor for the SARS-CoV-2 S protein, and variations in this gene may confer susceptibility or some type of resistance to the viral infection. 36 Allelic variants may differ according to countries or populations and explain, at least partially, differences in infection rates. On the other hand, the differences in the rise of the curve, the ΔDD and the Td10 were not uniform among the countries within temperate or tropical/subtropical areas. For example, in European countries such as Switzerland, Austria, Holland, Norway and the Czech Republic, these parameters were similar to those observed in tropical/subtropical areas and not to those in the rest of Europe. This finding coincides with the diagnosis of Covid-19 later than in the rest of Europe (Figure 3 panel A). In fact, these countries and the tropical/subtropical ones had more time to implement containment measures, which could have also contributed to the increase of Td. In the case of Td20, there is a possibility that this calculation is biased or at least severely related to the number of tests and containment measures that each country has undertaken and may have contributed, at least in part, to the differences 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint between temperate and tropical/subtropical countries. If this would have been the case, we would have observed a greater difference between the comparisons of partial doubling times between the climate zone as the curve progresses, that is, that there would have been a greater difference (Δ) in ΔTd30 >ΔTd20 > ΔTd10. We were not able to compare the Td30 as the curve in most of the countries in tropical/subtropical areas have not reached that period yet, but we found that ΔTd20>ΔTd10. Although, in the lineal regression models, Td10 participated in the value of Td20, only explaining about 17% of the variability of that value. Therefore, the expected difference in the Td20s between countries within temperate and tropical/subtropical zones should not be greater than 1.17 times Td10, and in opposition to that, the difference of the Td20s was 2.6 times that of Td10. This suggests that the differences in Td20 between countries in the different climate areas, may be related to factors such as the number of tests performed and the implemented measures to mitigate the epidemic. However, the relationship or change folds between Td20 and Td10 must remain valid, on the assumption, at least in principle, that the policies on the number of tests and implemented containment measures, did not change too much in each country. In addition, no statistically significant difference was found between FCTd20/Td10 in countries within temperate or tropical/subtropical regions. This implies that the value of FCTd20/Td10 ≥ 2 would be a good parameter for both regions to positively assess the evolution of the epidemic during the 21-30 day-periods of the curves. This study has several strengths including the large sample of incident cases of Covid-19 collected from the WHO database, to analyze the behavior of the epidemic 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint curves for more than 90 days. Another strength is the fact that epidemic curves and environmental variables of 67 countries distributed in different climatic zones and six continents, were studied. These two issues allowed the comparison of several variables in various segments of the epidemic growth curve and to establish the differences between the two climatic zones. Notwithstanding, the study has some limitations such as the inability to incorporate in the analysis the number of tests carried out or the containment measures implemented in each country, during the different periods of the epidemic curves. Although it is possible that these issues most likely do not affect or have very little impact on the analysis during the first 10 days of the epidemic curve, it is likely that they substantially affect the analysis in the second ten and third ten days of the epidemic curves. Despite this, the fold change ratio between Td20 and Td10 was no different between the two compared climate zones. Having a different pattern of infection spread between temperate and tropical/subtropical countries could only slow the speed in which the virus is being transmitted, and although this is good news for health services utilization, it does not necessarily imply that the proportion of the population infected will be smaller in the tropical/subtropical countries. 9 The latter will depend on the timing, related to the cumulative frequency growth curve, in which the containment measures are established, and their magnitude to reduce the spread of the epidemic in each country. 9, [37] [38] [39] On the other hand, the value of Td10 allows the evaluation of the growth of infections during the second 10 days of the epidemic curve, while the ratio 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 preprint this version posted April 23, 2020. . Td20/Td10 allows the study to evaluate the growth of the curve during the third 10 days and with it, may help in evaluating how effective are the implemented containment measures. This study showed that the behavior of the growth curve (DCS, D100 and ΔDD) in the first stage of the epidemic is related to the date that the first Covid-19 case was diagnosed, the ambient temperature and doubling time of infection cases, which are different between countries located in temperate and tropical/subtropical areas. The Td10 is on average 3 days longer in tropical/subtropical countries than in those located in temperate zones and can predict the growth of the curve for the following 10 days of the evolution of the epidemic after D100. These differences appear to be related to ambient temperature and the date of the first case that was identified and how the infection spread in both climatic zones. In addition, the Td10 and Td20 values helped predict the cumulative frequency of Covid-19 at the 20 th and 30 th days of the epidemic after D100 day, while the Td20/Td10 ratio helps to evaluate the growth curve behavior during the next third 10 days of the epidemic evolution. • Data from China and other Asian countries during the first month of the Covid-19 pandemic suggest that ambient temperature and humidity have a direct relationship to the doubling time (Td) for the number of cases. 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04. 18.20070920 doi: medRxiv preprint • There is a relationship not only between the ambient temperature, but also with the date when the first case was diagnosed and the behavior of the growth curve of the cumulated frequency, with the doubling time during the different 10-day time periods of the pandemic growth in 67 countries around the world. (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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint of these point-differences (ΔDD). The data included the analysis of 67 countries. Medians were compared between the groups by the Man-Whitney U-test. b. Td10 and Td20 = doubling time the fisrt 10 and 20 days of the curve, respectively, DCS= day start the curve climb, D100= day when the 100th was diagnosed, ΔDD=D100-DCS, climatic zones = 1 for temperate zones and 2 for tropic/subtropic zones in the models. a. The variables that pass in the univariate models at p<0.2 were introduced in the multivariate models using the backward method. Lineal regression models a 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 preprint this version posted April 23, 2020. b. Td10 and Td20 = doubling time the fisrt 10 and 20 days of the curve, respectively, DCS= day start the curve climb, D100= day when the 100th was diagnosed, ΔDD=D100-DCS, climatic zones = 1 for temperate zones and 2 for tropic/subtropic zones in the models. *The parameters of the multivariate model: b=23,067, r=0.67, Durbin-Watson=1.65, ANOVA p<0.0001 Prediction of total cases of Covid-19 at day 30 of the curve** 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 preprint this version posted April 23, 2020. . https://doi.org/10.1101/2020.04.18.20070920 doi: medRxiv preprint Clinical Characteristics of Coronavirus Disease 2019 in China A Novel Coronavirus from Patients with Pneumonia in China Coronavirus disease 2019 (COVID19) Situation report 76 Potential Factors Influencing Repeated SARS Outbreaks in China Coronavirus world map which countries have the most cases and deaths. Coronavirus updates The Effects of Temperature and Relative Humidity on the Viability of the SARS Coronavirus Low ambient humidity impairs barrier function and innate resistance against influenza infection Weathering the pandemic: How the Caribbean Basin can use viral and environmental patterns to predict, prepare and respond to COVID-19 Influenza seasonality: underlying causes and modeling theories Seasonality of Respiratory Viral Infections Ecological and immunological determinants of influenza evolution Seasonal variation in host susceptibility and cycles of certain infectious diseases Latitudinal variations in seasonal activity of influenza and respiratory syncytial virus (RSV): a global comparative review The National Academies Press Role of temperature and humidity in the modulation of the doubling time of COVID-19 cases Presymptomatic Transmission of SARS-CoV-2 -Singapore Covid-19: identifying and isolating asymptomatic people helped eliminate virus in Italian village Q&A: Similarities and differences -COVID-19 and influenza Il 70% dei donatori di sangue è positivo". La Stampa, 2020. 23. Prevention CfDCa. Interim Clinical Guidance for Management of Patients with Confirmed Coronavirus Disease (COVID-19) Ageing and respiratory infections: the airway of ageing Immunosenescence of ageing Lung infections and aging Innate immunosenescence: effect of aging on cells and receptors of the innate immune system in humans Reporte diario COVID-19 Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study Coronavirus Disease 2019 (COVID-19) in Italy Climate change, thermal stress and mortality changes Temperature, humidity, and latitude analysis to predict potential spread and seasonality for COVID-19 An explanation for the seasonality of acute upper respiratory tract viral infections Season and mortality Critical Care Utilization for the COVID-19 Early Experience and Forecast During an Emergency Response Structural variations in human ACE2 may influence its binding with SARS-CoV-2 spike protein SARS-CoV-2 and Europe: timing of containment measures for outbreak control Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder 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.