Why Nutrition Research Is So Difficult to Get Right

Both approaches to the methodological problems in nutrition research work. What differs is what they cost you, and the cost is what this sets out.

The experiment nobody can run

The ideal study would assign thousands of people to different diets at birth and follow them for their entire lives. That is impossible practically and unacceptable ethically, so researchers work with designs that approximate it badly.

Feeding studies achieve real control but only for weeks or months and usually in small numbers of volunteers. Long-term evidence therefore comes from observation, where people choose their own diets and researchers try to account for the consequences. Every well-known difficulty in this field follows from that single unavoidable compromise.

Measuring what people actually eat

Most large studies rely on people recalling their own intake through questionnaires covering typical consumption over months. Recall is unreliable, portion estimation is worse, and people under-report foods they feel they should be eating less of. The measurement error is not random, since it varies systematically with weight, sex and how healthy someone believes their diet to be.

On the label, better methods exist, including biological markers of intake, and they are expensive and available for only some nutrients. A study cannot be more accurate than its exposure measurement, and this is the weakest link in most nutrition research.

The healthy-user problem

People who follow dietary advice also tend to exercise more, smoke less, drink less and attend medical appointments more reliably. They are also, on average, wealthier and better educated, both of which independently affect health outcomes considerably.

Statistical adjustment attempts to remove these influences but can only adjust for factors that were measured. Residual confounding from unmeasured or imperfectly measured factors is the standard explanation for many modest associations. This is why an observed association between a food and an outcome is a starting point rather than a conclusion.

Everything is a substitution

Eating less of one thing necessarily means eating more of something else, so no dietary change happens in isolation. The health effect of reducing a food depends entirely on what replaced it, which studies often fail to specify.

Replacing one fat with another produces different results from replacing it with refined carbohydrate. Headlines describing a single food as good or bad ignore the substitution that any real change would involve.

Asking replaced with what converts an unanswerable question into one that research can actually address.

Why the headlines keep flipping

Single studies are reported individually, while the field advances through the accumulation of many studies over years. Each new study is newsworthy precisely when it contradicts what came before, which guarantees an appearance of chaos. Small effects in noisy data will scatter in both directions, and each scatter point becomes a headline.

Where the claim is carefully worded, press releases from institutions compete for attention and frequently overstate what a single study established. Following the reviews rather than the studies produces a far more stable and far less exciting picture.

What has nonetheless held up

Some findings have survived decades of scrutiny across different methods, populations and research groups. The benefits of not smoking, of physical activity, of adequate fibre and of limiting alcohol fall into that category. So does the broad pattern favouring vegetables, legumes, whole grains and limited highly processed food.

Against the trial data, convergence across independent lines of evidence is what makes these conclusions durable while individual headlines are not. A field being difficult does not mean it knows nothing, and treating it that way is its own kind of error.

Side by side

The takeaway

The field is difficult for structural reasons, not lazy ones. Follow the reviews and the guidelines rather than the individual headline.

The body already has organs for this, and none of them are sold in a box.