Why do nutrition studies keep contradicting each other?

Partly because what people eat is hard to measure, and partly because the study designs that can look at diet over decades are not the ones that can prove cause. Nutritional epidemiology examines dietary and nutritional factors in relation to disease at a population level. Some argue that its findings are unreliable because diet is an exposure susceptible to considerable measurement error.

Many big studies measure diet with a food frequency questionnaire, which asks how often a person ate foods over the past month, three months or year. Because it looks back so far, it depends on memory and on people judging correctly whether they eat something once a month or once a week. Walter Willett, who developed the Harvard version, estimated a correlation of 0.60 to 0.70 between such questionnaires and diet recalls. Obese people consistently under-report their food consumption compared with people of normal weight, as shown in a calorimeter room and by direct observation.

Then there are the people. Those who volunteer for studies tend to be healthier because they care about their health and are predisposed to follow medical advice, a bias that can damage the validity of studies testing interventions. Observational studies simply watch what people do, while experimental ones assign the exposure, which can give stronger evidence but is sometimes unethical.

On the other side of the ledger, nutritional epidemiology has linked folate deficiency to a higher risk of neural tube defects, vitamin C deficiency to scurvy, and trans fat to a higher risk of cardiovascular disease.