How to read a supplement study

Learn · Reading the evidence · Explainer, Typed Supps

One study can say almost anything. What matters is what kind of study, in whom, how big the effect — and whether anyone could repeat it.

Stage by stage

The ladder (Mechanism → meta-analysis): Cells and mice generate ideas; associations suggest; randomised trials test; pooled trials conclude. Marketing mostly lives on the bottom rungs.
The tricks (Tells to watch for): Surrogate markers, twelve-person samples, wrong populations, vanished null results, and “clinically proven” — a phrase with no legal meaning.
The size test (Significant ≠ noticeable): Find the absolute change, then benchmark it against sleep, protein and training. Most products lose to the free stuff.
Our grades (Per claim, not per product): Strong, moderate, limited, insufficient — the human evidence for that specific claim. Insufficient means unproven, not disproven.

What the evidence supports

For context: randomised, blinded trials control for placebo and expectation — strong evidence. The reason RCTs anchor every serious grading system, including this one.
For context: animal and in-vitro findings translate to human outcomes — limited evidence. The default rung of supplement marketing. Some survive translation; most quietly don’t.
For context: industry-funded trials report more favourable outcomes — moderate evidence. A consistent pattern across fields — operating through design and publication choices more than fabrication.
For context: small, flexible studies produce a high rate of false positives — strong evidence. The core of Ioannidis’s argument, and the replication record since.

Timing

A habit, not a dose: before buying, find the claim, find the study type, find the effect size in humans like you — then price it against an hour of sleep.

Time to effect

1 study: A reason to watch. Rarely a reason to buy.
5–10 yrs: How long a literature usually takes to settle.

Common myth

“It’s backed by a clinical study.” In reality: So is nearly everything — the question is which study. In cells? In mice? Twelve untrained students for a fortnight, measuring a blood marker, funded by the seller, never replicated? Each of those gets compressed into “clinically proven”. The truth lives in the details, which is why this site reads them for you.