During this pandemic year, much has been written about whether the SARS-CoV-2 coronavirus, which causes the COVID-19 disease, is seasonal like other respiratory viruses. That is, if its ability to infect will depend on the climate and will be greater each year in autumn or winter. But the answer is made to beg.
Last week, I read in big headlines that it had finally been confirmed that the coronavirus can be considered seasonal. “Everything is in the ‘cold’: find why the coronavirus could become a seasonal enemy,” sentenced a headline. “As with the flu, the cases will be concentrated mainly in winter and will almost disappear in summer,” announced many media.
You can make a great headline like this about coronavirus – or any other topic – based on a single result published by science people in science journals. But be careful, because at least you have to contrast and validate the results of the study in question. And make sure they can be reproduced, of course.
We do ourselves a disservice if we bias the interpretation of a scientific result towards what, for one reason or another, interests us. How could this have been the case?
What we know about the seasonality of the coronavirus
As far as is known, the transmission of the SARS-CoV-2 coronavirus is not only influenced by environmental conditions, but also by various other factors, including sociological, microbiological and physiological ones.
However, there are studies that completely ignore these factors. And it is not the only methodological error that is repeated. As the atmospheric scientist David Pino explained a few months ago in The Conversation, the studies often do not consider other variables that may be influencing the statistical analysis of the data, nor the time lag of the analyzed data, or even that the temperature and humidity are dependent variables (which influences the statistical results). Moreover, as a general rule there is a tendency to confuse correlation with causality.
I’m not quite sure if we science people actually mistake correlation for causality or if we just prefer to ignore that fact. For example, I think everyone understands that the correlation between temperature and the incidence of COVID-19 does not imply that temperature is the cause of a higher or lower incidence of the disease.
An eye-catching headline based on a weak correlation
A few days ago, an American university launched a press release with the headline “ Global analysis suggests COVID-19 is seasonal ” (“A global analysis suggests that COVID-19 is seasonal”). Interested, I read the content of the news in detail, which included statements by the author responsible for the article stating: “One conclusion is that the disease may be seasonal, like the flu. This is very relevant to what we should expect from now on after the vaccine controls these first waves of COVID-19. ” from the moment the vaccine controls these first waves of COVID-19. ”)
Although this “may be seasonal” is not a categorical assertion, the following affirmation “This is very relevant” may make us think that the analyzes carried out have yielded results that lead us to take it for granted. In search of arguments that would allow me, when the time came, to speak with knowledge of the facts, I went to the original article (“Temperature and Latitude Correlate with SARS-CoV-2 Epidemiological Variables but not with Genomic Change Worldwide”), and read carefully its content.
I take this opportunity to say that the article has been published in the scientific journal Evolutionay Bioinformatics, with a good reputation and with peer review, and with whose editor-in-chief, Dennis Wall, I have a great friendship.
In my opinion, the article is very well written and argued. The analyzes carried out, for the statistical methodology they have used, are adequate. The results are correctly expressed and the conclusions of the summary are consistent with the results obtained. In them, the authors are cautious and use words like “trends… suggest a seasonal effect”. And they also conclude that “more studies will be necessary to determine if the correlations are more likely due to …”. Nothing definitive.
I do not explain myself and it is difficult for me to understand how, in the public eye (because in the scientific article he expresses everything adequately), the senior author of the work “sells” the results almost as a categorical statement. Moreover, and put to being finicky, the correlations obtained at work, although statistically significant, are, for the most part, weak (with an absolute mean value of 0.24, in an absolute range between 0 and 1). And if we look at the most interesting, such as the relationship between temperature and incidence, or between temperature and mortality, the numbers tell us that it can hardly be explained beyond 7% of its possible relationship.
Other studies with the same purpose use a methodology that is more suitable for this type of analysis, such as Bayesian mathematical models, and not a simple correlation, no matter how statistically significant it may be. This is the case of the article “Seasonality and uncertainty in global COVID-19 growth rates”, published in the PNAS magazine, to indicate an example of rigor even in its own article title.
Bertrand Russel already said it:
“The concept of probability is the most important in modern science, especially since no one has the slightest idea what it means.”
Francisco José Esteban Ruiz, Associate Professor of Cell Biology, University of Jaén
This article was originally published on The Conversation. Read the original.
