How to read a peptide research study: a non-scientist's fieldguide

Most peptide articles — including this one’s neighbours on Wolverine — assume the reader can already read a peptide research study. The skill is not innate. Almost every mistake the consumer-facing peptide market makes about the evidence base traces back to one of five specific places.

Here are the five questions to ask of any cited paper before letting it carry weight in a prescribing or buying decision. Each one is illustrated with a real published study from the peptides currently under FDA review — because the worked examples are how the questions get easier to ask the next time.

Question 1 — What molecule did they actually dose?

The most common error in peptide marketing is citing a real trial for the wrong peptide. The drugs in the GHRH-receptor-agonist class are the worst offenders: sermorelin, tesamorelin, CJC-1295, and the older [Nle27]GHRH(1-29) analog are all closely related but structurally distinct, and a trial in one is often presented as evidence for another.

Baker and colleagues (2012, Arch Neurol) is the worked example. 152 adults aged 55–87, twenty weeks of nightly subcutaneous treatment, favourable cognition effect at p=0.03. The intervention was tesamorelin. The trial is routinely cited in sermorelin marketing copy as the evidence base for that compound’s cognitive effects. It is not. Sermorelin and tesamorelin share a receptor and a class but they are not the same molecule, the half-lives are different by an order of magnitude, and only one of the two has a current FDA drug approval. Read the methods section; check which compound the participants actually received.

Question 2 — What was the route of administration?

The bioavailability of a peptide depends on how it was delivered. Most peptides are degraded by digestive enzymes; oral bioavailability for unprotected sequences is typically under one percent. So a paper showing a positive result in rats fed a peptide by gavage (a controlled-volume dose delivered directly into the stomach) is not automatically evidence that a capsule of the same compound will work in humans. We covered the route-and-evidence problem in detail in Peptide bioavailability.

Wu and colleagues (2020, Mater Sci Eng C) is a sharper version of the same problem. Thymosin β4 was loaded onto electrospun nanofiber scaffolds and showed improved migration, proliferation, and tenogenic differentiation in human stem cells across 28 days. The work is strictly in vitro — cells in a dish, on a synthetic scaffold. No animal received an injection. No human received an injection. The paper is repeatedly miscited as evidence that TB-500 (synthetic Tβ4) injected into a human tendon will repair the tendon. The route — no injection at all — is the entire reason the citation does not support that claim.

Question 3 — What was the population, and does it match the use case?

A trial in HIV-infected adults with abdominal lipodystrophy is not automatically a trial in healthy adult men trying to lose belly fat. A trial in adults with mild cognitive impairment is not automatically a trial in healthy older adults trying to sharpen attention. A trial in 70-year-olds is not a trial in 35-year-olds.

The relevant question for any cited study is: were the people in this trial similar enough to me, or to the person I am prescribing for, that the effect size should be expected to transfer? Sometimes the answer is yes. Often the answer is no, and the trial is still being used as if the answer is yes. The fix is to read the population description (usually paragraph one of the Methods section) and decide whether the off-label use case sits inside that population or outside it.

Question 4 — What was the primary endpoint, and did it reach significance?

Trials register a primary endpoint in advance — the one outcome that defines whether the trial succeeded or failed. Secondary endpoints are exploratory; they are not the standard of evidence.

Beck and colleagues (2014, Int J Colorectal Dis) is the standard worked example. 117 adults undergoing bowel resection were randomised between ipamorelin and placebo for postoperative ileus. Median time to first tolerated meal was 25.3 hours on ipamorelin vs 32.6 hours on placebo. That 7.3-hour gap looks suggestive. The p-value, 0.15, says it did not reach statistical significance. The primary efficacy endpoint was null. Helsinn discontinued the development programme after the readout.

The Beck trial is sometimes cited in ipamorelin marketing as evidence the molecule helps gut recovery. It is the opposite. The trial that ran did not show that. Confusion happens when someone reads phrases like comparable trend or directional improvement and reads them as positive — they are the language used when the primary endpoint failed but the data is presented as gently as the authors can manage.

The Miller GHK-Cu trial is the variant of the same trap. Miller and colleagues (2006, Arch Facial Plast Surg) randomised 13 patients undergoing CO2 laser resurfacing to a GHK-Cu skincare regimen or the same regimen without the peptide. Blinded evaluators found no significant differences in erythema, wrinkles, or overall skin quality at 12 weeks. Only patient-reported satisfaction reached significance (P=0.04). That is the only positive endpoint in the trial, and it is also the easiest endpoint for almost any plausible intervention to move regardless of mechanism. The single published randomised human trial of GHK-Cu skincare is null on every objective endpoint; it is sometimes presented as a positive trial by quoting the satisfaction P-value.

Question 5 — Has anyone else replicated it?

A finding from a single research group, even a real and well-conducted one, is not the same as an established result. Mechanism stories that survive only inside the originating lab tend to look different once an outside team tries to reproduce them.

BPC-157’s mechanism literature is dominated by Predrag Sikirić’s group at the University of Zagreb across three decades; epitalon’s by Vladimir Khavinson’s group in St Petersburg across roughly the same span; GHK-Cu’s by Loren Pickart and collaborators since the 1970s. Each of those bodies of work contains real findings. None of them has been independently replicated in the volume of trials that would normally make a result a settled fact. Single-group dominance is a feature of the peptide evidence base; the question for a careful reader is whether the result has been confirmed outside the originating lab.

Bonus — what null actually means, and what to do with the n

The Lee and Burgess 2025 BPC-157 IV-safety pilot is a useful example of how a no adverse events finding can be misleading. The trial reported no adverse effects and no biomarker changes. The n was two. The duration was three days. The journal was Alternative Therapies in Health and Medicine. None of the surface positives — directional safety signal, IV route, well-tolerated — should outweigh those design specifics. Two people for three days is not safety evidence at scale, with or without a positive direction.

The Mendias and Awan 2026 Sports Medicine review puts the broader picture in print: this is a peptide landscape with thin controlled human safety data, scarce published pharmacokinetics, and a placebo effect amplified by social-media reach. The skill of reading the studies cleanly is what separates an honest assessment of any individual peptide from the volume reproduced in the marketing.

What a credible peptide study would actually look like

A credible peptide study, against the five questions above, is one where you can answer cleanly: the exact molecule dosed (not a related analog), the route administered (matching the intended consumer use), the population studied (similar to who the prescription is for), the primary endpoint as registered (and the direction of the result on that endpoint, not on secondaries), and at least one independent replication outside the originating research group.

Most peptide studies in current consumer marketing do not clear all five. The ones that do tend to be the trials of approved drugs — insulin, GLP-1 analogs, recombinant growth hormone, tesamorelin in HIV lipodystrophy — and a small handful of registered controlled trials in the molecules currently under FDA Pharmacy Compounding Advisory Committee review. We covered how PCAC reviews actually work in What is the PCAC.

The reason this matters: the prescribing question for any peptide is whether the available evidence supports a decision in this patient, this indication, this dose, this route. Reading the underlying studies cleanly is what separates a prescription from a guess. Wolverine Health is being built around the version of the prescribing conversation where a supervising physician has read the studies as carefully as the article above asks the reader to. Join the waitlist for a note the day the prescribing files we cover actually have the kind of trials this checklist would clear.