Reading the research, and why the headlines are usually wrong

Nutrition research is genuinely difficult, and the difficulties explain most of the apparent contradictions.

Why nutrition research is hard

You cannot randomise diets easily. Randomised controlled trials require people to eat an assigned diet, blinded if possible, for long enough to affect outcomes.

Blinding is generally impossible. Adherence over years is poor. Feeding studies where all food is provided are expensive and can only run for weeks.

Which means most long-term nutrition evidence is observational.

Observational studies have confounding. People who eat more vegetables also tend to smoke less, exercise more, drink less and have higher incomes.

Statistical adjustment helps and cannot fully remove it. Residual confounding is the standard limitation of the field.

Dietary assessment is inaccurate. Most studies rely on people reporting what they ate, via questionnaires or recall.

People misremember, underestimate, and report what they think they should have eaten. The error is substantial and it is not random.

Outcomes take decades. Cardiovascular disease and cancer develop over many years, which means either very long studies or reliance on surrogate markers that may not predict outcomes.

Everything is correlated. Eating more of one thing means eating less of another. The question is never simply whether a food is good, but what it replaces.

The questions to ask

Is this a trial or an observational study? Both are informative and they support different strengths of conclusion.

Observational studies are good at generating hypotheses and at studying long-term outcomes. They cannot establish causation on their own.

How many people, and for how long?

What was measured? An outcome people care about — heart attacks, deaths, diagnoses — or a surrogate marker such as a blood measurement?

Surrogate markers frequently fail to predict outcomes. This has caught out entire fields.

How large is the effect? Relative risk is what headlines report and it is misleading without the absolute risk.

A fifty percent increase in a risk that affects one person in ten thousand is very different from the same increase in a risk affecting one in ten.

Look for absolute numbers.

What was the comparison? Replacing saturated fat with polyunsaturated fat gives a different answer from replacing it with refined carbohydrate.

The substitution matters as much as the food.

Who funded it? Industry funding is common in nutrition research and is associated with results favourable to the funder in analyses of the literature. It does not invalidate a study and it is worth knowing.

Was it in humans? A great deal of nutrition coverage derives from cell studies and animal studies, which do not establish clinical effects.

The specific traps in plant-based nutrition coverage

Healthy user bias. People who choose plant-based diets differ systematically from those who do not — in smoking, exercise, alcohol, income and health awareness.

Studies finding better outcomes in vegetarians must contend with this, and the better ones attempt to adjust for it.

Not all plant-based diets are the same. Research distinguishing healthful from less healthful plant-based diets — whole foods versus refined grains, sugar and processed products — finds substantially different outcomes.

A diet of chips and white bread is plant-based.

Advocacy on both sides. This is a topic where people have strong commitments, and coverage on both sides selects evidence accordingly.

Be equally sceptical of sources arguing that plant-based diets are uniquely healthy and of those arguing they are dangerous.

Single nutrient thinking. The history of nutrition science includes repeated cycles of identifying a nutrient, isolating it, supplementing it, and finding that the supplement does not reproduce the effect of the food.

Where the evidence is reasonably strong

Worth stating, since the scepticism above could suggest nothing is known.

Fibre intake and health outcomes. Fruit and vegetable intake and cardiovascular outcomes. Legume intake. Wholegrains. Replacing saturated with unsaturated fat. The harms of excessive alcohol, of smoking and of high intakes of processed meat.

These are supported by multiple lines of evidence — observational, mechanistic and where possible trials — and they are broadly consistent across expert bodies internationally.

That convergence is the strongest signal available in this field.

The practical heuristic

When a headline appears:

Wait. A single study rarely changes what anyone should do.

Look at what the major dietetic and health bodies say, since they synthesise the whole literature rather than one paper.

Ask whether the effect is large enough to matter.

And notice that the advice that has survived decades of research is unglamorous and stable: eat plenty of vegetables, fruit, legumes and wholegrains; limit ultra-processed food, alcohol and excess energy; and do not smoke.

Almost everything else is at the margin.

General information rather than individual advice. Consult a registered dietitian for guidance specific to your situation.