"Most studies are wrong," warns a researcher

Two researchers selected the top 50 most used ingredients in a cookbook and looked at how many had been badociated with cancer risk or benefit in various studies published in the literature. scientific journals.

The answer: 40 out of 50, a list including salt, flour, parsley and even sugar. "Is everything we eat related to cancer? They then asked, ironically, in their 2013 article.

Samples too small

Their question touches on a known but persistent problem in the world of research: too many studies use samples too small to reach generalizable conclusions

But the pressure on researchers, the competition between journals and the insatiable appetite of the media for studies announcing revolutions or major discoveries, makes these articles continue

"The majority of articles published, even in serious journals, are bad," says one of the authors, John Ioannidis, professor of medicine at Stanford, de facto specializing in the study studies

"Scientists have only superficial training in statistics"

This slayer of bad scientific research demonstrated in 2005 in an art resounding "Why most published studies are wrong". Since then, he says, only some progress has been made.

Some journals require that authors provide their raw data and publish their protocol in advance. This transparency prevents researchers from shredding their methods and data in order to find a result, whatever it may be. They allow others to verify or "replicate" the study.

Because when they are redone, the experiments rarely achieve the same results. Only a third of 100 studies published in the three most prestigious psychology journals were reproduced by researchers, in an badysis published in 2015.

Medicine, epidemiology, clinical trials of drugs and … nutrition studies do not do much better, John Ioannidis insists, especially during replications.

"In the biomedical sciences and elsewhere, scientists have only a superficial training in statistics and methodology," adds John Ioannidis . Too many studies focus on only a few individuals, preventing generalization to an entire population, because the selected participants are unlikely to be representative.

The field of diets

"The diet is one the most appalling areas, "continues Professor Ioannidis, not only because of conflicts of interest with the agri-food industry. Researchers often go in search of correlations in huge databases, without any hypothesis.

In addition, "measuring a diet is extremely difficult," he says. How to quantify exactly what people eat?

Even when the method is good, with a randomized study, where participants are randomly selected, the execution sometimes leaves something to be desired.

A Famous 2013 Benefit Study The Mediterranean diet for heart disease had to be withdrawn in June by the most prestigious medical journal, the New England Journal of Medicine, because not all participants were randomly recruited; the results have been revised downwards

What questions should be asked?

So what to remember in the stream of published studies every day? John Ioannidis recommends asking the following questions: Is such a study isolated or is it reinforcing existing work? Is the sample small or large? Is this a randomized experiment? Who financed it? Are researchers transparent?

These precautions are fundamental in medicine, where bad studies contribute to the adoption of treatments that are at best ineffective, and at worst harmful.

In their book "Ending Medical Reversal", Vinayak Prasad and Adam Cifu offer terrifying examples of practices adopted on the basis of studies that have been invalidated years later, such as stenting (mini prostheses) in a brain artery to reduce the risk of stroke. It was not until after ten years that a rigorous study showed that the practice … actually increased the risk of stroke.

The solution involves the collective tightening of the criteria of all research, not only journals: universities, public funding agencies, laboratories … But these institutions are all subject to competition.

"The system does not encourage people to go in the right direction," says AFP Ivan Oransky, co-founder of the Retraction Watch website, which covers withdrawals of scientific articles. "We want to develop a culture where transparency is rewarded."

The problem also comes from the media, which in his opinion should better explain to their readers the uncertainties inherent in scientific research, and resist sensationalism. "The problem is the endless succession of studies on coffee, chocolate and red wine," he complains. "We must stop."