How to read a peptide study in five minutes

A peptide paper will usually tell you how much it actually shows, if you read it in the right order. Seven questions, all answerable from the abstract and the methods, separate a solid finding from an interesting hint and from a press release in disguise. Here they are, with real peptide papers as examples.

Every article on HPR carries one of five evidence levels, from pooled reviews of human trials down to first-hand reports. The questions below are how we decide which label a study earns. You can use them on anything you’re sent.

1. Who, or what, was studied?

Start here, because it overrides everything else. Cells in a dish, rats, and people are three different kinds of evidence, and a striking result in the first two often doesn’t carry over.

This isn’t a peptide-specific problem. When researchers followed up highly cited animal studies from leading journals, only about a third had been replicated in human randomised trials [1]. For many peptides discussed online, rat work is most of the literature. That makes the study preclinical, which tells you what is worth testing in people next.

Check the abstract’s methods line. If it says “rats”, “mice”, “in vitro” or “cell line”, read it as a preclinical result and look for a human study before drawing conclusions about people.

2. How many people?

Look for “n =”. Small numbers aren’t worthless, but they can’t detect uncommon side effects and they exaggerate chance findings.

Two real examples. An intravenous safety study of BPC-157 enrolled two people, both of whom had received the peptide before [2]. A knee-pain study of the same peptide reported on 16 patients [3]. The first tells you that two people tolerated two infusions over three days, which is precisely what a safety pilot is built to show. The second gives you a signal worth following up. Neither tells you how the peptide performs in a general population.

3. Compared with what?

A result without a comparison group can’t separate the treatment from time, natural recovery, or the placebo effect. Outcomes people rate themselves, such as pain, mood, sleep and energy, are the most open to expectation.

Three features make a comparison fair [4]:

• Randomisation. A random process, not the researcher, decides who gets what. This balances the groups on things nobody measured.
• Blinding. Participants and ideally assessors don’t know who got the real thing, so expectations can’t shape the result.
• A proper control. Placebo, or the current standard treatment.

The BPC-157 knee study had none of these. It was a retrospective chart review: staff phoned patients, most of them six months to a year after their injections, and asked them to rate their pain, with no validated scoring tools [3]. Fourteen of 16 reported relief. That is a signal worth testing properly, and the reason to run the trial this design cannot replace.

4. What was the main outcome, and did it hit it?

A well-run trial names one primary outcome before it starts. That is the question it was built to answer. Secondary outcomes are extras, and with enough of them something will reach significance by chance.

Thymosin beta-4 eye drops give a clean example. A small phase 2 trial in nine people with severe dry eye reported significant improvements in discomfort and corneal staining [5]. A larger phase 2 trial, with 72 participants, missed both of its primary endpoints; it reported improvements in several secondary measures instead [6]. Its abstract still concluded that the study “confirms the efficacy” of the drops [6]. Read the results line before the conclusion line, and check which outcome was primary.

5. How big was the effect?

Statistical significance tells you a difference probably isn’t chance. It doesn’t tell you whether the difference matters.

Look for the size of the gap between groups, in real units. In PIONEER 1, a well-run 26-week trial in 703 adults with type 2 diabetes, oral semaglutide 14 mg lowered HbA1c by 1.1 percentage points more than placebo, and body weight by 2.3 kg more [7]. Those are numbers you can weigh. Be wary of results given only as percentages of a percentage, or only as a p-value.

6. How long, and who dropped out?

Short studies can’t show whether a benefit lasts or what happens with long-term use. Six weeks tells you about six weeks.

Also check how many people left and why. If far more dropped out of the treatment group than the placebo group, side effects may be the reason, and the people left may be the ones it suited. PIONEER 1 reported discontinuation rates of 2.3–7.4% on semaglutide and 2.2% on placebo, which is the kind of disclosure you want to see [7].

7. Who ran it, and who paid?

Industry funding doesn’t make a result wrong. Large drug trials are often company-funded, and PIONEER 1 is typical: three of its ten named authors worked for the manufacturer [7]. What funding changes is how much independent confirmation you should want before relying on a result.

The broader warning comes from a much-cited 2005 essay. It argued that a published finding is less likely to be true when studies are small, effects are small, designs and outcomes are flexible, and there is financial or other interest in the answer [8]. Many peptide studies tick several of those boxes at once.

Matching a study to our evidence levels

The five-minute version

1. Cells, animals or people?
2. How many people?
3. Randomised, blinded, and compared with what?
4. Did it hit its primary outcome?
5. How big was the difference, in real units?
6. How long, and how many dropped out?
7. Who paid, and has anyone repeated it?

If a claim about a peptide survives all seven, it is worth taking seriously. Plenty clear the first two and stall later, and knowing exactly where a claim stalls is what tells you how much weight to give it.

What this means

Our position is simple: a peptide study is not a verdict, it is a step, and the honest way to read one is to ask how far along that step it has travelled. Most of the peptides people ask us about sit somewhere in the middle — a small human study, a chart review, a handful of animal papers — and the middle ground is where an early answer normally lives. Reporting it clearly is more useful than pretending it does not exist.

That is why we label every article instead of sorting peptides into “real” and “fake”. A pilot in sixteen patients and a 703-person randomised trial are both human evidence, but they carry very different weight, and flattening them into a single yes-or-no is where most online arguments about peptides go wrong. The point of reading carefully is not to declare a winner; it is to see the distance between what has been shown and what has been claimed.

We think you are better served by a plain statement of how strong the evidence is than by a confident conclusion either way. If a claim rests on cells, we say so. If it rests on one small trial that missed its primary outcome, we say that too — and we still report the signal, because a signal that fails a strict test is often exactly what the next, better-designed study is built to chase.

You are free to disagree with where we draw those lines. What we will not do is hide them. If you can finish a paper and name which of the seven questions it fails, and why that matters to you, the labelling has done its job.

Educational content only — not medical advice. Many peptides discussed on HPR are not approved for human use. Talk to a qualified clinician before making any decision about your health.