How to read a peptide study

Most peptide claims online trace back to a study. The question is what kind of study, and whether it supports the claim built on top of it. You do not need a medical degree to check. You need five questions, asked in order. Below we work through each one with real studies from our database, including the ones behind semaglutide, BPC-157, and TB-500.

1. Who or what was studied?

Start with the subjects, because they cap how far the conclusion can travel.

• Cells in a dish tell you about mechanism. They say nothing about whether a peptide reaches the target tissue in a living body.
• Animals tell you whether an effect survives a whole organism. In a 2011 rat study, BPC-157 sped healing of a cut Achilles tendon and increased tendon cell outgrowth. [4] That is a real finding in rats. It is not evidence for your shoulder.
• People are the only subjects that answer the question most readers are asking. For BPC-157, a 2019 review found the human evidence limited to unpublished pilot work and said human trials are needed before clinical claims can be made. [5]

Also check which people. STEP 1 enrolled adults with obesity, or overweight plus a weight-related condition, who did not have diabetes. [1] A result in that group does not automatically apply to lean athletes or to people with type 2 diabetes.

2. How was it designed?

Design determines whether a study can show cause and effect.

• Randomized, double-blind, placebo-controlled is the strongest common design. Participants are assigned by chance, and neither they nor the investigators know who got the drug. STEP 1 used this design in 1,961 adults for 68 weeks. [1]
• Active comparator trials pit two treatments against each other. SURPASS-2 compared tirzepatide with semaglutide in type 2 diabetes and was open label, meaning participants knew which drug they received. [2] Our semaglutide vs tirzepatide comparison is built on trials like this.
• Open-label or uncontrolled studies can suggest an effect but cannot rule out placebo response or bias. The clinical trials of semax in stroke were open comparative studies published in Russian-language journals, without independent replication. [7]
• Observational and genetic association studies find links, not causes. A 2021 study linked a mitochondrial DNA variant that changes the MOTS-c sequence to type 2 diabetes in men. [8] It did not give anyone MOTS-c, so it cannot show that taking the peptide does anything.

3. What exactly was measured?

Every trial names a primary endpoint in advance: the one question it is built to answer. Everything else is secondary.

The phase 2 trial of thymosin beta-4 eye drops for dry eye is a useful example. In 72 adults, neither primary endpoint (eye discomfort and corneal staining) was significantly different from placebo, although several secondary measures improved. [6] A fair summary is "missed its primary goals, with some encouraging secondary signals." It is sometimes cited as proof that TB-500 works in humans. It is a different product, given in the eye, for a different condition, and it did not meet its main goals.

Watch for surrogate endpoints too. A change in a blood marker, a gene expression level, or telomere length is not the same as feeling better, healing faster, or living longer.

4. How big was the effect, and how sure are we?

A statistically significant result can still be small. Look for the actual numbers.

In STEP 1, average weight change was minus 14.9% with semaglutide versus minus 2.4% with placebo, and 86.4% versus 31.5% of participants lost at least 5% of body weight. [1] Those are large, concrete differences.

For outcomes that take years, trials report hazard ratios. In SELECT, semaglutide cut major cardiovascular events with a hazard ratio of 0.80 and a 95% confidence interval of 0.72 to 0.90. [3] The confidence interval matters: because the whole range sits below 1.0, the benefit is unlikely to be chance. An interval that crosses 1.0 means the study cannot rule out no effect.

Also look at size. A trial of 17,604 people, as in SELECT, can detect modest effects and uncommon harms. [3] A study of 15 animals cannot.

5. Who ran it, and has anyone repeated it?

Replication by independent groups is what turns a finding into knowledge. Two warning signs come up often in peptide research:

• One lab. Much of the BPC-157 literature comes from a small number of laboratories, mostly in Zagreb. [5]
• Reviews by the discoverer. Much of what is summarized about GHK-Cu in skin comes from review articles co-authored by the scientist who discovered the peptide, summarizing small cosmetic studies. [9] That is not misconduct, but it is not independent confirmation either.

A narrative review is also not a new study. It summarizes others, and its conclusions are only as strong as the studies it picks.

Putting it together

Here is the checklist in one place:

1. Cells, animals, or people? Which people?
2. Randomized and blinded, or open and uncontrolled?
3. Was the primary endpoint met? Is it a real outcome or a surrogate?
4. How large was the effect, and does the confidence interval exclude no effect?
5. Independent replication, or one group?

We apply the same questions to every peptide and publish the answer as an evidence grade. You can see the grades across a whole topic on condition pages such as obesity, where approved drugs with large randomized trials sit next to peptides supported only by animal work.