how to evaluate a food claim: the larderlab methodology

The questions people actually bring us.

Q01

How do you tell a real food claim from internet panic?

Run it through seven checks: separate the bundled layers of the claim, identify whether the evidence is mechanism, association, or causation, rank the study quality (meta-analysis and RCT beat cohort beat case report beat a single mechanism study), check whether the scary dose resembles a realistic human exposure, ask who benefits from the claim, find the one quantified number that settles the practical decision, and state both consensus and dissent. Panic claims almost always fail on layers, dose, or evidence tier. The seed-oil-vs-butter-vs-olive-oil and arsenic-in-rice deep-dives show the method applied end to end.

Q02

Why does 'studies show' tell you almost nothing?

Because the study type carries the weight, and 'studies show' hides it. A meta-analysis of randomized controlled trials is strong evidence for causation; a single observational cohort can only show association (confounders, healthy-user effects); an in-vitro or rodent study shows a mechanism that may or may not translate to humans at realistic doses. Most alarming food claims cite the weakest tier, often a mechanism study, while implying the strongest. We name the study, the year, the design, and the sample size for exactly this reason, and so should any source you trust.

Q03

Why is dose the step most people skip?

Because the dose makes the poison, and headlines almost never state it. A compound that causes harm at 100x typical exposure is not a meaningful risk at 1x. This is the entire resolution of several recurring scares: inorganic arsenic in rice matters at high daily consumption and for infants, not at one serving a day (see arsenic-in-rice); MSG's 'symptoms' do not reproduce in blinded trials at dietary doses (see is-msg-bad). Always convert the scary claim into a realistic daily exposure before reacting to it.

Q04

How should a claim handle uncertainty?

By stating it. When the literature is genuinely mixed, the honest format is: here is the consensus position, here is the strongest dissenting evidence and who holds it, and here is our read with the reason. Pretending certainty in either direction (a food is 'toxic' or 'totally safe') is the tell of a low-quality source. Our is-canola-oil-bad deep-dive is built this way: the consensus, the strongest steel-manned objection, and a quantified bottom line, rather than a verdict dressed up as settled science.

Q05

What number actually settles a food decision?

The one that maps directly to the choice you have to make: micrograms of a contaminant per serving against the tolerable intake, the glycemic load of a portion rather than the abstract glycemic index (see glycemic-index), the cost per gram of an effective nutrient, or the percent risk change in absolute rather than relative terms. A claim that cannot be reduced to a decision-relevant number usually should not change your behavior. Every deep-dive on this hub ends on that single quantified bottom line.

Q06

Who funded it, and why does that matter?

Funding and incentive shape which questions get asked and how results get framed, on both sides. Industry-funded studies can understate harm; supplement and 'natural food' sellers can overstate the harm of a competitor ingredient to sell an alternative. The fix is not to dismiss funded research but to weight it against independent replication and to notice who profits from either the alarm or the reassurance. A claim that only the seller of the solution is making deserves extra scrutiny, which is the same logic behind this site's no-paid-placement rule.