Reading a peptide RCT — a practical guide

A typical reader sees "randomized controlled trial" and assumes high-quality evidence. Most published peptide studies are not what they appear. Trial design tells most of the story before you read a single result.

Here is the seven-question filter we apply when grading the evidence on this site. You can apply the same filter to any paper you encounter.

1. Is the population the population that matters?

A trial of BPC-157 in healthy 22-year-old male medical students tells you very little about whether it works in 55-year-old patients with degenerative tendinopathy. Inclusion and exclusion criteria narrow generalizability sharply. We discount studies where the trial population is far from the population the peptide would be used in clinically.

2. What's the comparator?

Comparing peptide vs no-treatment is much weaker than peptide vs placebo, which is much weaker than peptide vs an active comparator that already works. The strongest evidence in the peptide world — semaglutide vs placebo in SUSTAIN-6, tirzepatide vs semaglutide in SURPASS-2 — is strong precisely because the comparator was meaningful and the trial was powered to detect a real difference.

3. Was it actually blinded?

"Double-blind" on the cover sometimes means "we tried." Look for a description of how blinding was maintained, whether participants could plausibly tell which arm they were in (peptides with notable side effects can break blinding), and whether outcome assessors were independent of the patients.

4. Is the primary endpoint pre-specified — and clinically meaningful?

A peptide that improves a biomarker without improving any patient-experienced outcome is not a peptide that has been shown to work. Pre-specification matters because post-hoc selection of significant endpoints from a basket of measured ones is one of the most common ways trials look more impressive than they are.

5. Is the effect size meaningful?

Statistical significance with n=2,000 is not the same as a clinically meaningful effect. A 0.1% reduction in HbA1c that hits p<0.05 is not a useful finding even though it's "statistically significant." We grade against effect size, not p-value.

6. Was the trial powered for the question being asked?

A 22-patient open-label pilot can establish "this seems plausible." It cannot establish "this works." Peptide literature is full of small open-label trials that produce a positive signal that vanishes in a properly powered Phase 2 or Phase 3. The grade reflects what's been shown — not what was hinted at in underpowered work.

7. Who ran it, and what's the conflict of interest disclosure?

We don't reflexively discount industry-funded trials — most pivotal evidence in modern medicine is industry-funded by necessity. But we do read the conflict of interest disclosures, look at trial registration history (was the protocol changed? which endpoints?), and weigh independent replication. A single-lab result without replication weighs less, regardless of who funded it.

Putting it together

When you see a new peptide claim circulating online, walk through these seven questions before forming a view. If any of the first six fail badly, the result is hypothesis-generating at best. If the seventh raises a flag, expect to want independent replication before treating the result as established.

This is exactly the filter that informs every grade on Peptigrade. We owe you the work of applying it; you owe yourself the discipline of asking for the work to be shown.