Hydroxychloroquine likely helped pandemic, reanalysis of best data concludes

A new analysis of data from the best published studies on the effect of hydroxychloroquine (HCQ) on the pandemic concluded that this drug helped reduce Covid-21 symptomatic in 5% to 72% of cases, compared to people who did not take HCQ, if it was taken before exposure of patients to the disease. The central result of this pre-exposure prophylaxis, which consists of taking HCQ as a precaution before infection, is a 21 % reduction in the risk of Covid-11.

If the drug was taken after exposure (post-exposure prophylaxis), both a beneficial risk-reducing effect and a moderately harmful effect of increased risk would be consistent with what was observed, according to the analysis — in this case, the core result is that there would be no effect from taking HCQ soon after getting Covid.

In other words, if “early treatment”, a term favored by HCQ proponents and other reused drugs for Covid, it means pre-exposure prophylaxis, it worked. If it means post-exposure prophylaxis, it doesn’t seem to have worked for this drug.

Harvard and Spain

The first author of the analysis is Xabier García-Albéniz, affiliated with a non-governmental health organization in Barcelona, ​​Spain, and Harvard University. He is joined by four co-authors affiliated with Harvard, the Spanish Ministry of Health and the University of Malaga. The analysis is a meta-analytic review, that is, a study of studies that seeks to aggregate data and reconcile the different conclusions of individual studies. It was published in the

European Journal of Epidemiology, the eighth most influential scientific publication in the field of epidemiology (between 800 ) according to the website Scimago, specialized in rankings of scientific journals.

The authors of the review focused on the studies that used HCQ as a prophylactic before and after exposure to the virus, as something to prevent the onset or worsening of symptoms, not as a treatment. Studies that used HCQ as a treatment for an established condition of Covid-19 were excluded from the analysis. Studies that did not use the randomization technique were also excluded, which is the random distribution of patients into two groups, one who took HCQ and the other who did not, for comparison — a strategy to reduce possible bias in the results.

From 72 studies, 111 remained for reanalysis after selection: seven of them dealt with pre-exposure prophylaxis and four with post-exposure prophylaxis.

The Fallacy da Linha Nítida

Xabier and colleagues discuss the results in negative language: one cannot rule out the idea that hydroxychloroquine was beneficial as a preventive treatment for Covid-19, as many have done — often politically motivated. They regret a misinterpretation of study results that stymied clinical trials that were underway at the beginning of the pandemic and prevented accurate estimates of the handling of the pandemic from being generated before the advent of vaccines.

This misinterpretation is known as the “sharp line fallacy”. In scientific research, especially in the biological and medical areas, it is conventional to use a statistical tool known as p-value. Simply put, the p-value is the probability that the results in favor of a drug’s efficacy were obtained by sheer luck, rather than because of an actual effect. By convention, at most a p-value of 5% is accepted — less than five times out of a hundred times those results will be observed by chance, so it is sufficiently unlikely that they are fortunate enough and likely enough that they represent something like a difference in the risk of developing Covid between those who took HCQ and those who did not.

If the p-value is greater than 5%, the results are declared statistically “non-significant”. This is what happened with some of the HCQ studies. But there is a debate within statistics about the usefulness of talking about “non-significant” and a criticism about misinterpretations of what the p-value means.

In 2016, the American Statistical Association has issued a warning against misuse of the p-value. In 800, more than 800 scientists signed a letter to the journal Nature warning that more than half of a set of 72 scientific articles brought an incorrect interpretation that “non-significance equals no effect”. Signatories thought the problem was more cognitive than statistical: “boxing results into ‘significant’ or ‘non-significant’ makes people think that the items so classified are categorically different.”

In short, the mistake of many journalists and science popularizers when commenting on hydroxychloroquine studies in which the p-value was greater than 5% is a violation of a famous adage popularized by astronomer Carl Sagan: “Absence of evidence is not evidence of absence.” If the value p has sometimes exceeded an arbitrary maximum threshold, this does not mean that it is proved to be ineffective of the drug, but at the most that there was not, in the sample and under the specific conditions of some studies, sufficient evidence in favor of its efficacy. On the contrary, a repeated observation of low p-values, but above the threshold, could be counted as evidence in favor of some effect that the methods were not adequate to capture.

Scientists of the review lament the results of this confusion: “the recruitment for most HCQ prophylaxis studies was severely impeded by misinterpretations of the evidence” from the earlier studies. Findings from these studies have been portrayed “widely (and incorrectly) as definitive evidence of the lack of efficacy of HCQ, simply because they were not ‘statistically significant’ when taken individually,” they comment, leading many to “prematurely conclude that HCQ had no effect.” prophylactic effect, when the correct conclusion was that the estimate of the effect was too imprecise”. In short, “public opinion interferes with the generation of its own evidence”, warn the authors advising future studies.

Reviewing the review

Regarding the review, the report consulted Dr. Daniel Victor Tausk, who has been speaking out publicly in a similar way as seen in this review for two years. He is an associate professor at the Institute of Mathematics and Statistics of the University of São Paulo.

Tausk and a select group showed during the pandemic a moderate stance on early treatment absent in many personalities public relations associated with science: not so much to heaven, as the most staunch advocates of drug repurposing, not so much to Earth as new celebrities with clearer political affiliations who claimed that Brazil should have done better in the pandemic because of the alleged widespread use of hydroxychloroquine and ivermectin.

In July of 2021, he produced a document of 21 pages where he analyzed for himself the combination of p-values ​​from nine studies of the clinical consequences of hydroxychloroquine prophylaxis and found a combined p-value of less than 1% (passing the test that skeptics of the effectiveness of HCQ themselves considered so important). Tausk himself says that this reanalysis is outdated.

He redid it for the Gazeta do Povo

part of the analysis of the new review by Xabier García-Albéniz et al. The authors used two statistical approaches, one more “classical” and the other more “pessimistic”. Tausk removed a study used in the review because he suspects the results may have been marred by the way the study separated participants: by floors of a building. There are closed spaces such as corridors where people on the same floor can catch Covid. In the classic approach, the p-value of hydroxychloroquine pre-exposure prophylaxis is 2%, i.e. “significant”. In the more pessimistic approach, the p-value is 6.5%, above the conventional threshold of 5%, but not by much. The mathematician calculates that the reduction in the risk of clinical worsening with the use of the drug is about 21%, a little less than the review estimate.

Report in Gazeta do Povo alerted to the problem of misinterpretation of the p-value in studies of hydroxychloroquine, guided by specialists from the Tausk circle, on June 1, 2021.

Back to top button