Nutrition
What the evidence actually supports
Description
In 1948, researchers in a mill town outside Boston sat down with roughly 5,000 residents and started asking them, at intervals, what they ate. The Framingham Heart Study is still running today, three generations deep, and it gave us much of what we think we know about diet and the heart. It also handed nutrition science its founding tool and its founding weakness: the food questionnaire. People were asked to recall, from memory, how often they ate eggs, butter, red meat. From those answers, an entire field of dietary advice was built.
The trouble is that this method — following large groups of people and correlating what they say they eat with how they later fare — cannot easily tell cause from coincidence. A person who eats more vegetables also tends to smoke less, exercise more, earn more, and see a doctor sooner. When that person stays healthier, the salad gets the credit, but the confounders were all riding along. Over seventy-five years, this has produced a discipline that generates confident headlines and reverses them a decade later, leaving most of us unsure whether eggs are friend or foe this year.
So it is worth separating two things that usually travel together: the noise, which is enormous, and the signal, which is smaller and steadier than the news cycle suggests. A handful of dietary findings have survived every methodological beating we can give them. Most of the rest are provisional at best. The gap between those two categories is where nutrition advice goes to die — and where the interesting story actually lives.
The question we’re asking : Once we strip out the confounders and the reversals, what does the evidence on food actually support?What we’ll see : How a fragile method built a field, which of its findings genuinely hold, and why the same data keeps producing opposite headlines.
Table of contents
01 Chapter 1 — The trouble with the questionnaire
Most of what circulates as nutritional fact comes from observational studies, and observational studies rest on a shaky foundation: asking people what they ate. The food-frequency questionnaire, the workhorse of the field, requires someone to remember whether they had fish once a week or twice, how many servings of grain, roughly how much oil. Human memory is bad at this, and human honesty is worse. People underreport what embarrasses them and overreport what they think they should eat. One well-known analysis of the data feeding U.S. dietary guidelines concluded that the reported calorie intakes were, for large fractions of the population, physiologically implausible — too low to sustain life if taken literally.
On top of the measurement problem sits the confounding problem, which is deeper. Diet does not vary at random across a population. People who eat more whole grains tend to be wealthier, more educated, less likely to smoke, more likely to exercise. So when a study finds that whole-grain eaters live longer, the grain is entangled with a dozen other things that also extend life. Statisticians try to adjust for these, but you can only adjust for what you measured, and the lived texture of a healthy life resists being fully captured in a survey.
02 Chapter 2 — The handful of findings that hold
Given all that, it is striking that anything survives — but a few things do, and they survive precisely because they show up across every kind of evidence: observational studies, randomized trials, biochemistry, and populations that differ wildly from one another. The first is trans fats. Industrially produced partially hydrogenated oils raise the harmful fraction of blood cholesterol and lower the protective fraction, and the link to heart disease held up so robustly that governments moved to ban them outright, a rare case of nutrition science ending an argument rather than starting one.
The second robust finding concerns severe deficiency. Scurvy, rickets, goitre, pellagra, anaemia — the diseases of missing vitamin C, vitamin D, iodine, niacin, iron — are settled science, because here the experiment is brutal and clear: remove the nutrient, the disease appears; restore it, the disease resolves. This is the ground on which nutrition earned its early authority, and it remains unshaken. The complication is that this logic tempted the field into assuming more is always better, which is where supplements went wrong.
03 Chapter 3 — Why the headlines keep swinging
If a stable core exists, why does the coverage feel like whiplash? Part of the answer is structural. A single observational study, even a weak one, produces a publishable correlation, and a correlation makes a headline. Coffee causes cancer; coffee prevents cancer; both can be true findings from different studies of the same beverage, because each is picking up a different tangle of confounders. The press covers the study, not the slow accumulation of evidence that would let you weigh it, and each new paper arrives as if it overturned the last.
There is also the problem of the effect size. Most real dietary effects, outside of deficiency and trans fats, are small. A food might shift your risk of some outcome by a few percentage points over decades — genuinely real, but so faint it can be swamped by the confounders and measurement error already baked into the method. When the signal is that quiet, the noise easily flips its apparent direction from one study to the next. So the same nutrient oscillates between hero and villain, not because the science is broken, but because it is trying to detect something near the edge of what its instruments can see.
04 Chapter 4 — A science that measures a moving target
Step back, and nutrition looks like a case study in what happens when a discipline is handed a question everyone cares about, that unfolds over a lifetime, and that it cannot cleanly experiment on. Physics can isolate a variable in a chamber. Nutrition studies human beings who eat thousands of meals a year, in social settings, over decades, changing their habits as fashions and incomes and moods shift. The thing being measured will not hold still, and the measurer cannot ethically lock people in a metabolic ward for thirty years to watch. The field's uncertainty is not a sign of incompetence; it is honest reporting from an almost impossible vantage point.
This reframes what we should expect from it. We keep asking nutrition to behave like pharmacology — this dose of this compound produces this outcome — when it is closer to ecology or economics, sciences of complex systems where interventions ripple unpredictably and clean answers are the exception. The people who navigate this best tend to hold the robust core firmly and the periphery loosely, treating the daily churn of single-food studies as weather rather than climate.
05 Conclusion
The Framingham questionnaire is still being filled out, and its descendants now come with wearable trackers and genetic panels and phone apps that photograph your plate. The tools have grown sophisticated, but the underlying bind has not moved: we are still trying to learn how food shapes a human life by watching people live, imperfectly recalled and impossibly entangled. What has survived that scrutiny is real, and smaller than the shelf of diet books suggests — trans fats harm, deficiencies kill, plant-heavy patterns help, and excess weight costs.
Related books
