Is Coffee Good for Us? Maybe Machine Learning Can Help Figure It Out.

Should you drink coffee If yes, how much? These seem to be questions that a society capable of producing vaccines for a new respiratory virus should have no problem answering in a year. And yet the scientific literature on coffee shows a frustration that readers, not to mention a lot of researchers, have with nutritional studies: the findings are always changing, and they often contradict each other.

Such disagreement does not matter much if we are talking about foods or beverages that are not consumed very widely. But in 1991, when the World Health Organization classified coffee as a possible carcinogen, its implications were enormous: more than half of the American population drinks coffee daily. A possible link between drink and bladder and pancreatic cancer was revealed by observational studies. But it turns out that such studies – in which researchers ask a large number of people to get information about things like their dietary intake and daily habits, and then seek engagement with particular health outcomes – validation Smokers were not more likely to drink coffee. It was smoking that increased their risk of cancer; Once that association (along with others) was understood, coffee was removed from the list of carcinogens in 2016. The following year, the available evidence published in The British Medical Journal was reviewed, A link between coffee and some low risk for cancer, Also for heart disease and death from any cause.

Now a new analysis of existing data published in the American Heart Association journal Circulation: Heart Failure suggests that Two to three (or more) cups of coffee per day can reduce the risk of heart failure. Of course, common caveats apply: it is union, not work-cause. It could be that people with heart disease avoid coffee, presumably thinking it would be bad for them. So … good for you or not good for you, which is it? And if we can never tell, what do these studies mean?

Critics have argued that this is not, in fact – that nutrition research should shift its focus from observational studies to randomized control trials. By randomly giving coffee to one group and stopping from another, such tests can be tried to separate the cause and effect. Yet when it comes to understanding how any aspect of our diet affects our health, both approaches have important limitations. Our diet works on us throughout our lives; It is not possible to keep people in a lab, monitoring their coffee intake until they develop heart failure. But it is extremely difficult for people to report correctly what they eat and drink at home. Ideally, to get to the bottom of the coffee question, you will know what type of coffee beans are used and how it is roasted, ground and brewed – all of which affect its biochemistry Do – as well as the exact amount, its temperature, and the amount of any type of sweetener or dairy. You will then consider all other variables that affect the coffee drinker’s metabolism and overall health: genome, microbiome, lifestyle (eg sleep) and socioeconomic status (is there domestic stress? Poor local air quality?).

Randomized control trials can still give useful insights into how coffee affects biological processes in a short period of time. This can help explain, and thus, validate some long-term associations. But before testing on a given nutrient, there must be some reason for scientists to think that it can have a meaningful effect on many people; They are already required to give plausible evidence that testing of the compound on human subjects will not cause them permanent harm.

Circulation studies employed observational data, but its initial aim was not to assess the association between coffee and heart failure. This is what lead author David Cao, a cardiologist at the University of Colorado School of Medicine, told me: “The overall question was, what are the factors in daily life that affect cardiovascular health that we don’t know about Are potentially to be replaced at lower risk. “Because one in five Americans will develop heart failure, even small changes in their behavior can have a large cumulative effect.

Traditionally, researchers start with a hypothesis – coffee, for example, lowers the risk of heart disease. They then compare the subjects’ coffee intake with their cardiac history. A drawback to this process is that researchers have all sorts of predetermined assumptions, which can affect what variables they affect and in the data to fit their theory by researchers included or excluded in the analysis. Are influenced by manipulation. “You can achieve any discovery in science using your own biases, and you’ll get its publication,” says Steven Hemsfield, professor of metabolism and body composition at the Pennington Biomedical Research Center at the University of Louisiana. To illustrate this point, a 2013 comprehensive review in The American Journal of Clinical Nutrition discovered 50 common kitchen ingredients in the scientific literature; 36 was personally linked to one Increased or decreased the risk of cancer including celery and peas.

Cao, however, did not start with a hypothesis. Instead, he used a powerful and increasingly popular data-analysis technique, known as machine learning, to characterize the thousands of patients collected in the famous Framingham Heart Study and the development of heart failure in those patients. To see the relationship between. “The algorithm will introduce the variables that contribute the most to the variance in the data,” says Diana Thomas, a professor of mathematics at West Point, or the range of cardiac outcomes. “And that’s the purpose.”

Machine learning’s ability to process large amounts of data may change nutrition researchers ‘ability to study their subjects’ behavior more accurately and in real-time, wrote Amanda West, Medical Director of the Cardiac Transplant Program at Tufts Medical Center is. An editorial published with circulation studies. For example, it can be trained to scan pictures of subjects’ food and interpret their macronutrient levels. It can also analyze geolocation devices, activity sensors and social media data.

But machine learning is only as good as the data being analyzed. “Without careful control,” says Michael Kosorok, a professor of biotechnology at the University of North Carolina at Chapel Hill, “it gives us the ability to make more and more mistakes. “If, for example, it applies to data sets that are not diverse or random enough, then the patterns it sees won’t catch when the algorithm then uses them to make real-world predictions. This has been a serious problem with facial recognition software: trained primarily on white male subjects, algorithms have been much less accurate in identifying women and people of color. Algorithms to handle uncertainty in data are also Must be programmed – when one person reports a “cup of coffee” is six ounces and another has eight ounces.

An analysis, like Cao’s, begins with no presumption assumptions about the data that no one thought could reveal the connection. But those findings should be rigorously tested to see if they can be replicated in other contexts. After the link between coffee intake and reduced risk of heart failure was seen in the Framingham data, Cao results in two other respected data sets using an algorithm to correctly estimate the relationship between coffee intake and heart failure. Confirmed. Kosoroka describes the approach as “thoughtful” and says that it “seems like very good evidence.”

Nevertheless, it is not certain. Rather, it is part of a growing body of evidence that, at this time, may say very little about how much people should drink about coffee. “It might be good for you,” says Darush Mozazarian, dean of the Friedman School of Nutrition Science and Policy at Tufts University. “I think we can say with good certainty that it is not bad for you.” (Additives are another story.) More research will be required if it is more specific. Last year, Mozaffarian and others called for the National Institutes of Health to establish an institute for nutritional science that could coordinate those efforts and, importantly, help people interpret the results. “We need a well-funded, well-organized, coordinated effort to learn about nutrition,” he said. “No study is done for truth.”


Kim Tingley is a contributing writer for the magazine.

Source link

Leave a Comment