[tsimionescu]

Research methodologies in nutrition are almost universally underpowered for the results they claim. You'll very rarely find double-blind clinical trials with large numbers of participants, and since we know little of the possible confounding factors between individuals when it comes to nutrition, that's what you need.

Instead, most knowledge in nutrition comes either from extremely small clinical studies (often less than 50 people), often with co-morbidities such as obesity or diabetes or endocrine diseases; or from professional atheltics. Or, from huge national-level statistics that rely on interviews for self-reported diet information, and are thus forced to mix diet, diet perception, excersise levels, lifestyle (stress, sleep) and environmental factors in one big mess, out of which come recommendations like the mediteranean diet, instead of siesta or 5-7 weeks of vacation per year or year-round sun.

[nradov]

Nutritional studies also frequently suffer from the healthy subject effect. Dietary is highly correlated with other behaviors in ways that are difficult to control for.