key: cord-0781697-8ndl8zjz
authors: Hoogeveen, M. J.; Van Gorp, E. C.; Hoogeveen, E. K.
title: Pollen Explains Flu-Like and COVID-19 Seasonality
date: 2020-06-07
journal: nan
DOI: 10.1101/2020.06.05.20123133
sha: dca43f0f52ceccf80bb4720fcffcbc8b73899b9d
doc_id: 781697
cord_uid: 8ndl8zjz

Current models for flu-like epidemics insufficiently explain multi-cycle seasonality. Meteorological factors alone do not predict seasonality, given substantial climate differences between countries that are subject to flu-like epidemics or COVID-19. Pollen is documented to be antiviral and allergenic, play a role in immuno-activation, and seems to create a bio-aerosol lowering the reproduction number of flu-like viruses. Therefore, we hypothesize that pollen may explain the seasonality of flu-like epidemics including COVID-19. We tested the Pollen-Flu Seasonality Theory for 2016-2020 flu-like seasons, including COVID-19, in The Netherlands with its 17 million inhabitants. We combined changes in flu-like incidence per 100K/Dutch citizens (code: ILI) with weekly pollen counts and meteorological data for the same period. Finally, a discrete, predictive model is tested using pollen and meteorological threshold values displaying inhibitory effects on flu-like incidence. We found a highly significant inverse association of r(224)= -.38 between pollen and changes in flu-like incidence corrected for incubation period, confirming our expectations for the 2019/2020 COVID-19 season. We found that our predictive model has the highest inverse correlation with changes in flu-like incidence of r(222) = -.48 (p < .001) when pollen thresholds of 610 total pollen grains/m3 per week, 120 allergenic pollen grains/m3 per week, and a solar radiation threshold of 510 J/cm2 are passed. The passing of at least the pollen thresholds, preludes the beginning and end of flu-like seasons. Solar radiation is a supportive factor, temperature makes no difference, and relative humidity associates even with flu-like incidence increases. We conclude that pollen is a predictor for the inverse seasonality of flu-like epidemics including COVID-19, and solar radiation is a co-inhibitor. The observed seasonality of COVID-19 during Spring, suggests that COVID-19 may revive in The Netherlands after week 33, the start being preceded by the relative absence of pollen, and follows standard pollen-flu seasonality patterns.

Current models for flu-like epidemics insufficiently explain multi-cycle seasonality. Meteorological Recently, we identified pollen bio-aerosol as a discrete seasonal factor in inhibiting flu-like epidemics 45 for the period 2016 to 2019 in The Netherlands (Hoogeveen, 2020) . In this epidemiological study, we 46 found strong inverse associations between allergenic pollen counts and hay fever on the one hand, and 47 flu-like incidence on the other hand. The study was based on the persistent observation that pollen and 48 flu season predictably alternate each other in moderate climate zones, and the absence of sufficient 49 meteorological explanations (Tamerius et al., 2011) . We further observed that the passing of pollen 50 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 June 7, 2020. To study the relation between pollen and flu-like incidence in The Netherlands, we used the public 87 data sets of Elkerliek Hospital (Elkerliek.nl) about the weekly allergenic, non-allergenic and total 88 pollen counts in The Netherlands in grains/m 3 . Total pollen count is the combination of allergenic and 89 non-allergenic pollen counts. Further, we use the data from the Dutch State Institute for Public Health 90 (RIVM.nl) gathered by Nivel (Nivel.nl) about weekly flu-like incidence (WHO code "ILI") reports at 91 the primary medical care, per 100,000 citizens in The Netherlands with a population of currently 17.4 92 million. ILI (influenza like illnesses) is defined as a combination of a measured fever of ≥ 38 °C, and 93 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint cough, with an onset within the last 10 days. The flu-like incidence metric is based on a representative 94 group of 40 primary care units, and calculated using the number of influenza-like reports per primary 95 care unit divided by the number of patients registered at that unit, averaged for all primary care units, 96 and next extrapolated to the complete population. Compared to our previous study (Hoogeveen, 2020) , 97

we have expanded the datasets with 42 weeks so that it now runs from week 1 of 2016 till week 18 of 98 2020 (n = 226 data points) to include the recent COVID-19 pandemic in the tail-end of flu season. 99

Further, we have added the non-allergenic pollen counts to the allergenic ones as it does not matter for 100 the bio-aerosol filter function whether pollen is allergenic or not. 101

Further, we added the meteorological datasets form the Royal Dutch Meteorological Institute 102 (KNMI.nl), including relative humidity/day, average temperature/day and global solar radiation in 103 J/cm 2 per day as an indicator of UV radiation, from its centrally located De Bilt weather station, and 104 calculated the weekly averages for the same period. 105

To test allergenic versus non-allergenic pollen assumptions, against hay fever and pre-covid-19 flu-106 like incidence, we make use of the hay fever index. The hay fever index is defined as turnover for hay 107 fever medication as reported by all Dutch pharmacies to the Dutch Central Bureau of Statistics 108 (CBS.nl) based on respective ATC codes (R01A/R01AC). We use a dataset from week 1 of 2016 till 109

week 10 of 2019 (n=166 data points), as no further data has been made available. 110

There is no missing data in the data sets, except that there are only three weekly pollen counts missing 111 because of a malfunctioning metering station. We used a 4 weeks surrounding average to estimate the 112 three missing data points to avoid breaking lines in visuals. However, given the large number of 226 113 data points, and by using this trend-based average, we assume that these missing data have no material 114 impact on the results. 115

Regarding the incidence of flu-like symptoms, we calculated the weekly change compared to the 116 previous period to get an indication of the flu-like epidemic life cycle progression, whereby a decline 117 is interpreted as Ro<1 and an increase as Ro>1 (Ro is the reproduction number of flu-like viruses). 118

Further, to cover, in one time-series metric, for changes in flu-like incidence as well as an incubation 119 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint period of up to two weeks, we calculated a three weeks moving average (3WMA) of changes in flu-120 like incidence, of which two weeks are forward looking. 121

We tested the following statistical null hypotheses: 122 H10: there are no inverse correlations for non-allergenic and total pollen counts with flu-like incidence 123 (corrected for incubation period). 124 H20: there are no inverse correlations between pollen and changes in flu-like incidence (corrected for 125 incubation time 3WMA). 126

Checking assumptions: 127 H30: non-allergenic pollen has no effect on hay fever. 128 H40: meteorological variablessolar radiation, temperature and relative humidityhave no effect on 129 pollen and flu-like incidence change (3WMA). 130 H50: there is no predictive significance of a discrete model's compound value, based on pollen and 131 solar radiation thresholds, on changes in flu-like incidence (3WMA). 132

Note that except H3, all hypotheses are related to potential causality: the temporal sequentiality of the 133 respective independent variables, and flu-like consults corrected for incubation period. 134 135

Variables are presented with their mean and standard deviation (SD). 137

We calculated correlation coefficients to do a first test of hypotheses, as data sets show normal 138 distributions, to assess the strength and direction of relationships. Next, linear regression (F-test) on 139 identified inhibitors and interactions is used to determine the linear equation using estimates and 140 intercept values, and produce the test values such as probability, significance level, F-value, and the 141 Multiple R squared correlation to understand the predictive power of the respective inhibitor. Standard 142 deviations and errors, and degrees of freedom (DF) are used as input for calculating the 95% 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 preprint this version posted June 7, 2020. (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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint In our previous study, we found a significant inverse association between allergenic pollen counts and 167 changes in flu-like incidence, but no significant inverse correlations between allergenic pollen counts 168 and flu-like incidence due to the short and extreme bursts of these allergenic pollen (Hoogeveen, 169 2020 ). However, the addition of non-allergenic pollen reveals highly significant negative correlations 170 (r(224)= -.19, p < .01) with flu-like incidence, which become stronger when correcting for two weeks 171 incubation time (r(222)= -.31, p < .00001). We can thus reject the null-hypotheses (H10) that non-172 allergenic pollen, has no negative association with flu-like incidence, including the first cycle of the 173 COVID-19 pandemic, and thus we cannot rule out the non-allergenic pathway. That the associations 174 become stronger when taking into account incubation time, implies temporality, also in the case of 175 total pollen count (see Table 1 ). Furthermore, we can also observe from Figure 2 that flu-like consults 176 start to decline after the first pollen bursts. And that flu-like consults start to increase sharply after 177 pollen counts become very low or close to zero. This adds to the sense of causality as well. Further, we 178

can notice that the first COVID-19 cycle behaves according to pollen-flu seasonality. 179

These findings indicate that it makes sense to include the so-called non-allergenic pollen in our model 180 and pollen tests as well. (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 June 7, 2020. When testing the impact on the changes in medical flu-like consults (mean: -0.25, SD: 15.4), the 185 extended dataset till 2020, including COVID-19, shows a stronger and highly significant inverse 186 correlation with allergenic pollen (r(223) = -.23, p = .000413) compared to our previous study. For 187 total pollen, including non-allergenic pollen, the association is even somewhat stronger (r(223) = -.27, 188 p = .000059). Therefore, we can falsify the null-hypothesis (H20) that there are no inverse correlations 189 between the weekly pollen countstotal or only allergenic -and changes in flu-like incidence, Linear regression analysis shows that there is a highly significant inhibitory effect of pollen on flu-like 204 incidence change (3WMA) of F(1, 222) = 37.1, p < .001 (see Table 2 line 1), as a further basis for 205 using total pollen count as a predictor. 206 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 June 7, 2020. 213 Table 3 : Summary of univariable regression analyses of total pollen count on hay fever (p < .001) and 214 hay fever on flu-like incidence/100K citizens per week (p < .001), whereby pollen leads to an increase 215 in hay fever, which in turn is associated with a decrease in flu-like incidence. 216

Previously, we established a highly significant inverse correlation of r(164)= -.34 (p = .00001) 217 between the hay fever index (mean: 101 SD: 115.7) and flu-like medical incidence (Hoogeveen, 218 2020). Univariable regression analyses show that there is a highly significant positive effect of all 219 pollen on hay fever incidence, which in turn has a highly significant inhibitory effect on flu-like 220 incidence (see Table 3 ). In line with the association between pollen and flu-like incidence, the inverse 221 association between total pollen count and hay fever is stronger (r(164)= -.51, p < .00001) than those 222 for allergenic and non-allergenic pollen individually. This confirms that we can best use total pollen 223 count as predictor. Counter to expectation, non-allergenic pollen has a highly significant effect on hay 224 fever as well (r(164)= .48, p < .00001). We can thus reject the null-hypothesis that non-allergenic 225 pollen (H30) has no effect on hay fever. This might imply that pollen classified as non-allergenic might 226 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint still be responsible for certain allergic effects, and not just allergenic pollen. Therefore, trying to use 227 non-allergenic pollen to discriminate effects outside the allergenic path regarding the immune system 228 might be challenging. Given the scatter diagram in Figure 3 , the nature of the relation between hay fever and flu-like 234 consults, might however not be strictly linear, but logarithmic. Such a logarithmic relation might 235 support the idea that switching thresholds are important to understand the relation between the 236 independent variables and flu-like incidence. 237 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint warmer and dry weather. We can reject the null-hypothesis that meteorological variables (H40) have 243 no effect on pollen, for solar radiation (mean: 1047, SD: 709) and relative humidity (mean: 79, SD: (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 June 7, 2020. Only for solar radiation there is a highly significant inverse association with changes in flu-like 257 incidence (3WMA) (r(224) = -.25, p = .000156), so it is not unlikely that apart from its indirect effects, 258

it is playing a role in destroying the aerosol flu-like viruses before they can infect someone. 259

Thus, from the meteorological variables, for solar radiation and relative humidity the null-hypothesis 260 (H40) can also be rejected for that they have no effect on the flu-like epidemic lifecycle. But, of these 261 two only solar radiation is a flu-like inhibitor in line with its positive effect on pollen count. 262

A simple regression test confirms that there is a highly significant inhibitory effect for solar radiation 263 on flu-like incidence change (3WMA) (F(1, 222) = 14.43, p < .001 (see Table 2 line 2). As the 264 correlation is weak (Multiple R-squared = .06), we can see solar radiation only as a supportive factor 265 to pollen. 266

Taking into account all these findings, we developed a discrete, compound model in which we take the 267 changes in flu-like incidence (3WMA), a threshold value for solar radiation (kr), and both pollen 268 threshold values for allergenic (kap) and all pollen (kp). We found that the compound model (mean: 269 1.4, SD: 1.1) has the highest inverse correlation (r(222) = -.48, p < .001) for the following threshold 270 values: kr: 510 J/cm 2 , kap: 120 allergenic pollen grains/m 3 per week, and kp: 610 total pollen grains/m 3 271 per week. In line with the previous outcomes, inclusion of relative humidity, non-allergenic pollen or 272 temperature did not improve the correlation strength of this model. As they also did not show 273 significant interaction effects with pollen, also such interactions are not meaningful to consider for the 274

In each of the observed years the now (re)defined pollen thresholds are being past in week 10 (± 5 276 weeks) depending on meteorological conditions controlling the pollen calendar, which coincides also 277 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint with reaching flu-like peaks, and again in week 33 (± 2 weeks) marking the start of the new flu-like 278

There is a highly significant inhibitory effect of our compound thresholds based predictor value 280 (outcomes are in the range [0, 3]) on flu-like incidence change (3WMA) of F(1, 222) = 65.59, p < .001 281 and a Multiple R-squared correlation of 0.2281 (see Table 2 line 3). This confirms the usefulness of a 282 discrete, pollen and solar radiation thresholds based model as a predictor of switches in flu-like 283 seasonality, whereby the effect of pollen is stronger than that of solar radiation. As a consequence we 284

can reject the null-hypothesis (H50) that this compound pollen/solar radiation value has no predictive 285 significance for flu-like seasonality. 286 287 Discussion 288

We found highly significant inverse relations between pollen and (changes in) flu-like incidence: a 289 higher number of pollen associates with a decline in flu-like incidence. This association becoming 290 stronger when including the 2019/2020 period. Therefore, we conclude that also COVID-19 is subject 291

to pollen-flu seasonality as all previous pandemics since the end of the 19 th century. 292

When taking into account all pollen, both allergenic and non-allergenic, the inverse association 293 between the increment of pollen and decline of flu-like incidence becomes stronger. This finding is in 294 line with our virus filtering aerosol hypothesis in which aerosol pollen de-activate aerosol viruses 295 before they can infect a new host. However, as so-called non-allergenic pollen also contributes to the 296 incidence of hay fever in our study, another explanation might be that non-allergenic pollen have 297 immuno-activating effects as well. Therefore, we included the total number of pollen as a predictor of 298 flu-like seasonality, not just the number of allergenic pollen. The fact that the inhibitive effects of 299 pollen gets stronger when taking into account an incubation period of up to two weeks strengthens the 300 idea of causality. 301

The highly significant inverse association between hay fever and flu-like incidence can be interpreted 302 in a number of ways, which are not mutually exclusive. A) Allergic rhinitis symptoms might make it 303 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint more difficult for flu-like viruses to find their way to the lung cells that are vulnerable to it, such as 304 ACE-2 receptor positive cells in case of SARS-CoV-2 (Wan et al., 2020) . B) The application of hay 305 fever medicationthe basis of the index -likely reduces over-reporting (false positives) of flu-like 306 incidence because they suppress symptoms that are typically confused with flu. C) The anti-histamine 307 hay fever medication possibly suppresses flu-like symptoms as well. In this respect interesting to note 308 that COVID-19 patients typically suffer from an over-expression of the immune system. 309

The only meteorological variable that has a co-inhibitive effect on flu-like life cycles, solar radiation, 310 has a stimulating effect on aerosol pollen formation, and is responsible for melatonin-induced 311 immuno-activation. Relative humidity reduces pollen aerosol formation, and correlates positively with 312 flu-like incidence. In our study we showed that temperature, except for influencing the onset of pollen 313 season, has no predictive value for the flu-like lifecycle, and therefore its inverse correlation with flu-314 like epidemics is probably spurious: the common causal factor seems to be solar radiation. properties, and an interaction that prevents viruses to effectively reach new potential hosts. 329 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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint Meteorological variables such as increased solar radiation and temperature are not only triggering 330 pollen ripening, but also affect the bio-aerosol formation: dry and warm conditions stimulate pollen to 331 be airborne, and makes the bio-aerosol more hostile to flu-like viruses as it occasionally might reach 332 sterilizing temperatures. Rain, to the opposite, makes pollen less airborne, cools the bio-aerosol down, 333 and very high humidity levels (RH 98%) are even detrimental for pollen (Guarnieri, 2006) . The RH 334 98% effect on pollen, could thus provide an alternative explanation of why flu-like incidence in 335 tropical countries is higher during rainy season, and reduced during the rest of the year. 336

A third explanation might be related to circadian and seasonal rhythms being coded into the 337 functioning of our immune system, "switching off" to a lower gear during the night, and 338

Autumn/Winter, and "switching on" during day light and Spring/Summer. Not only solar (UV) 339 radiation and temperature, but also pollen, could function as a gene-coded trigger associated to the un-340 depression of the immune system, which would make evolutionary sense as pollen is the distinct 341 harbinger of the seasons of opportunity. Hemisphere with its larger populations are higher to be the initial breeding ground for a new flu-like 347 pandemic than the Southern Hemisphere. (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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint

Pollen likely seasonal factor in inhibiting flu-like epidemics. A Dutch study into the 405 inverse relation between pollen counts, hay fever and flu-like incidence

Allergic rhinitis: not purely a histamine-related disease

Mediators and 410 cytokines in allergic and viral-triggered rhinitis

Projecting the transmission 413 dynamics of SARS-CoV-2 through the post-pandemic period

The reproductive number of COVID-19 is higher 415 compared to SARS coronavirus

Seasonality of Respiratory Viral Infections. Annual 418 Review of Virology

Why Lungs Keep Time: Circadian Rhythms and Lung Immunity

Nectar and Pollen Phytochemicals Stimulate Honey Bee (Hymenoptera: Apidae) Immunity to

Viral Infection

The vaccination coverage required to establish herd immunity against influenza 427 viruses

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

Dynamical prediction of flu seasonality driven by ambient temperature: 430 influenza vs. common cold

Current Understanding of the Pathophysiology of Allergic Rhinitis

The Influence of

Simulated Sunlight on the Inactivation of Influenza Virus in Aerosols

Absolute Humidity and Pandemic Versus Epidemic 437

The Role of

Global 445 influenza seasonality: reconciling patterns across temperate and tropical regions. Environmental health 446 perspectives

Potential utility of melatonin in deadly infectious diseases 448 related to the overreaction of innate immune response and destructive inflammation: focus on COVID-449 19

Atopic diseases and related risk 452 factors among Dutch adolescents

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

Antiviral effect 455 of brassinosteroids against herpes virus and arenaviruses

Receptor Recognition by the Novel Coronavirus from Wuhan: an

Analysis Based on Decade-Long Structural Studies of SARS Coronavirus

No Association of

COVID-19 transmission with temperature or UV radiation in Chinese cities

including their anti-viral properties. Another question is whether immuno-activation by pollen is 356 indeed a causal factor in reducing the spread of flu-like viruses and whether that depends on anti-viral 357 immune responses or indiscriminate ones. Understanding this might lead to new therapeutic 358

On the basis of current data, we can conclude that also the covid-19 pandemic is seasonal and as a 360 consequence multicycle, and will thus likely return from week 33 on, like all other flu-like viruses, 361 when pollen season is over in the Northern Hemisphere. It is important to be prepared for it as long as (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 June 7, 2020. (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 June 7, 2020. . https://doi.org/10.1101/2020.06.05.20123133 doi: medRxiv preprint