Evidence-Based Nutrition
Why Almost Every Nutrition Headline You've Read Is Technically True and Practically Useless
The study that proved coffee causes cancer and the study that proved it prevents cancer were both rigorous, peer-reviewed, and funded by reputable institutions — and that tells you something important about how nutrition science actually works.
The Idea
Nutrition research has a structural problem that most health coverage never mentions: almost all of it is built on observational data. Researchers ask people what they ate over the past year, assign them a risk score, and then track what diseases they develop. The method is called a food frequency questionnaire, and it is, charitably, an educated guess. People misremember meals. They underreport indulgences. They round portions. And then these self-reports get fed into statistical models that produce relative risk figures — '32% increased risk of X' — that get laundered through press releases into headlines that feel like facts. The deeper issue is confounding. People who eat lots of processed meat also tend to sleep less, exercise less, and earn less. People who eat lots of kale tend to do the opposite. Unpicking the effect of any single food from the tangle of lifestyle, class, stress, and genetics is extraordinarily difficult — often impossible without a randomised controlled trial, which you cannot ethically run on humans for decades at a time. So the field operates in a permanent state of approximation. This is not a scandal. It is the honest limitation of a genuinely hard science. But it does mean the right posture toward nutrition headlines is not cynicism — it is calibrated scepticism. The broad patterns that survive across multiple methodologies and populations — eat mostly whole foods, limit ultra-processed food, don't overeat — are robust. The specifics — this superfood, that macronutrient ratio — almost never are.
In the World
In 2015, epidemiologist John Ioannidis — already famous for a paper arguing that most published research findings are false — turned his attention to nutrition science specifically. He and his colleagues analysed a random sample of ingredients from a standard cookbook and found that 80% of them had been the subject of at least one study claiming they either caused or prevented cancer. Butter. Eggs. Milk. Parsley. The data was not fabricated; the studies were real. The problem was that each one was small, relied on self-reported dietary recall, and had never been replicated at scale. Ioannidis's broader point was uncomfortable for a field that had grown enormously influential: the tools nutrition epidemiology uses are poorly matched to the questions it is trying to answer. A single observational study showing that people who eat more walnuts have better heart health cannot distinguish between the walnuts, the fact that walnut-eaters tend to be wealthier and more health-conscious, and a dozen other variables moving in parallel. What followed his critique was not a collapse of nutrition science but a useful reckoning. Researchers began investing more in rigorous controlled trials — like the PREDIMED study on the Mediterranean diet — and in better biomarker data rather than self-report. The science is improving. But the gap between what researchers actually know and what gets published as a confident dietary recommendation remains wide enough to drive a food truck through.
Why It Matters
Understanding the architecture of nutrition evidence changes how you read health information — and how much mental energy you spend on it. If you have ever felt whiplashed by contradictory advice about eggs, fat, gluten, or red wine, it is not because scientists are incompetent or corrupt. It is because the question 'is this specific food good or bad for you' is genuinely difficult to answer cleanly, and the media ecosystem rewards certainty over nuance. The practical upshot is liberating rather than paralysing. It means you can stop treating every new nutritional study as a mandate to reorganise your diet. The signal-to-noise ratio in individual studies is low. The signal in broad dietary patterns, sustained over years, is much higher. Eating mostly whole, minimally processed food, in amounts that keep your energy stable, is about as well-supported as nutrition science gets — and it does not require you to track a single superfood or ban a single ingredient. The lesson is not 'nothing matters.' It is 'the things that matter most are also the least exciting to write headlines about.'
A Question to Ponder
If the nutrition advice you feel most certain about turned out to be based on the weakest evidence, would you find that freeing — or would it feel like losing something you needed to believe?
Get a new one of these every morning.
Start learning with Thinkable