When 23andMe was in its most ridiculous phase of existence, the co-founder was telling FDA to talk to the hand while she assumed campaign donations to members of Congress would exempt them from being told to stop lying about their tests allowing consumers to “take steps toward mitigating serious diseases” like breast cancer. It didn't work, and after giving them years to obey the law FDA finally sent a warning letter and then Jeffrey Shuren, M.D., J.D., Director of FDA’s Center for Devices and Radiological Health gave their customers hilarious subtext about how useless the tests really were:
FDA is not standing in the way of 23andMe selling tests intended to help consumers trace their ancestry, identify relatives and tell them why they like or don’t like the taste of cilantro.
They just could no longer "suggest" their tests were a “first step in prevention” of disease. The company was then forced to admit its tests were 'just for entertainment' the way supplements sold in the GNC also have no known scientific value and have to put that disclaimer on their bottles.

That word "suggest" is used for disinformation and misinformation quite a lot. To suggest some Miracle Vegetable can make you live longer, or some Scary Chemical at homeopathic levels will murder you, all you have to do is find data with someone who had either and lived or died less than average.

That's not science. Unfortunately, it is often food and chemical epidemiology. It's easy to pretend that people did not trust guidelines on COVID-19 because of politics or stupidity but it's a weak argument(1) and suggesting an genetic switch controlling distrust of government mandates would be an even crazier case of putting statistical lipstick on a correlational pig.(2)

Though more credible epidemiologists wanted them gone decades ago, the brainless zombie of Food Frequency Questionnaires is back to torment actual science again. This time claiming they provide genetic evidence that heart health influences coffee consumption. Because on surveys, people with high blood pressure, angina, etc. report drinking less coffee.

Do the authors not realize how crazy it is to therefore claim our genes tell us to drink less coffee if we're more prone to high blood pressure? 

Colorado became popular with tuberculosis patients in the 1800s because of its dry climate. Over 30 percent of Colorado residents had the disease and Colorado led the nation in deaths from "consumption." If you were doing bad epidemiology then you could correlate living in Colorado to getting tuberculosis. And dying from it. 

That is the problem with taking people who develop heart issues and who follow doctor recommendations and creating a causation arrow back to the food product the doctor told them to reduce. It would be like looking at a university classroom and seeing more people wearing vision correction than in fighter pilots and concluding that universities create bad eyes.

What's bizarre is that they make these claims but claim their study wasn't designed for that at all. It was just to see if people who drank too much coffee get jittery, which pretty much everyone who has ever had too much caffeine in any form can confirm it true; stimulants work.

No fooling? So some people say they don't drink coffee in the afternoons because it prevents sleep and aren't making it up?

Even their conclusion is half fine: "people tend to naturally regulate their coffee consumption based on blood pressure levels and heart rate."

Okay, people who say they get jittery if they have more than two cups of coffee are right in not having more than two, but 'naturally regulate' and 'blood pressure' are giant red flags that science is about to leave the discussion.

"Causal relationships were determined via Mendelian randomization" shows it left and never came back.

Because no, they weren't determined. Putting some genetic science into epidemiology definitely helped save the field, and it can help estimate causation if the authors are agnostic about confounders and genetic differences and test the assumptions Mendelian randomization uses, but it only works if the genetic variants can't affect the outcome in other ways.

This does not pass that test. It is instead the kind of paper that 20 years ago caused epidemiologists to suggest taking their field out behind the barn and shooting it.(3)

They used estimates from meta-analyses for heart rate associations and consider them settled science. They stumbled into survivor bias while using a proxy to claim a risk factor for a risk factor has a genetic basis.

It gets worse. The authors double down on voodoo and claims our genes actively regulate how much coffee we want to drink. Our genes are protecting us from consuming more than our genes tell us we should. And then "people subconsciously self-regulate safe levels of caffeine based on how high their blood pressure is, and this is likely a result of a protective genetic a mechanism," claims University of South Australia epidemiologist Prof. Elina Hyppönen

Come on. Your subconscious brain tells you what kind of coffee to drink and how much? Freud died over 80 years ago and his 'theory' was dead before that, but it keeps showing up when people inclined to believe that correlation is causation need something science-y to feel good about. Yet even he would think this is a stretch.

The high number of survey results (390,435 - UK Biobank) they looked at sounds authoritative but is actually not. If you take enough diseases or benefits and try to match them with enough foods, using 400,000 people you are 100 percent certain to be able to claim correlation from lots of things to lots of other things and claim statistical significance.

Source: https://xkcd.com/882/

Abuse of statistical significance has gotten so bad in food and psychology that the American Statistical Association issued a warning about how bad most papers claiming statistical significance probably were.

And yet food epidemiologists do it all of the time. No wonder the public is confused. Along with causing heart disease and forcing our genes to take on sentience, does coffee also cure or cause cancer? Yes.

With only 61 foods and 10 diseases you are going to get a statistically significant link to lots of them. Using even a small data pool, you can show coin flips are not random, they are biased against heads. Or tails.

It's bad epidemiology, but it's what gave us garbage claims like gluten sensitivity and the Grain Brain and Wheat Belly  diet books that followed. It's what gave us IARC claims that bacon was as "hazardous" to humans as plutonium. The US National Institute of Environmental Health Sciences is trotting out links to trace chemicals and some disease every week with no more evidence than 'we looked at data of people who had disease X and say they cleaned their kitchens so we claimed Comet probably caused the disease.'

Do "genes" decide if you like coffee at all? Well, biology matters. I don't like many vegetables, they make me gag despite growing up in a family that ate them all. I didn't learn to dislike salad, salad makes me feel ill just thinking about it. I don't like alcohol. That is biological no different than some Koreans have enough of an allergy they get a red face when they drink it.

But this - "Whether we drink a lot of coffee, a little, or avoid caffeine altogether, this study shows that genetics are guiding our decisions to protect our cardio health" - is scientific gibberish.


(1) Left-wing people in the US are overwhelmingly more anti-vaccine than right-wing so it can't be Republicans are the problem.

 (2) Instead of being science, this is the same thing that has harmed the reputation of serious epidemiology - what we need to get people to act smartly when it comes to smoking and corornavirus - for decades. It is more like voodoo, a claim that you have a supernatural switch that limits how much coffee you want to drink, and your brain knows it and your subconscious makes that happen. No plausible biological mechanism is even proposed because to get written up by corporate journalists, you don't need any science, just a provocative claim with confounders dismissed.

(3) They were concerned because the big stuff had been solved and all that was left was FFQs and increasingly bizarre links to new diet books.