There is a very specific way a claim gets overstated, and this molecule is a clean example of it. A peptide shows a strong effect in mice. A human trial is run. The human trial is reported as a success. If you read only the first paragraph, or only the press release, or only the sentence with the word “significantly” in it, you will come away with the wrong idea about what was actually shown.
The two species did not agree. Here is exactly where they parted.
The mouse arm, which worked
The 2025 paper in Cardiovascular Research ran thymosin beta 4 through two mouse models and a human trial in one design.
In C57BL/6J mice subjected to ischaemia and reperfusion, seven days of recombinant human TB4 prevented cardiac dysfunction and fibrosis 28 days after the surgery, and significantly reduced plasma NT-proBNP at both one day and 28 days after the injury. In a second model with permanent ligation, TB4 improved cardiac function and reduced infarct size at eight weeks after the infarction.
The molecular side is unusually well specified for a compound like this. RNA-sequencing of the injured heart tissue showed TB4 modulating the ErbB signalling pathway. In cell models of hypoxia and reoxygenation, TB4 activated the ErbB2/Raf1 pathway, attenuated cardiomyocyte apoptosis and suppressed the pro-apoptotic protein Bad.
And then the part that should be in every write-up: the cardioprotective effects in the mice were abolished by ErbB2 inhibition. Block the pathway, and the effect disappears. That is a mechanism claim with a falsification test attached, and it is much stronger than a compound that only moves a marker.
The human arm, which did not
Same paper, same compound. A randomised, placebo-controlled, double-blind trial in 96 people with ST-elevation myocardial infarction after primary angioplasty. The first TB4 was given within eight hours of the procedure.
The result, as printed:
- in the group that received the first TB4 within 8 hours after the procedure (n = 43), the infarcted areas were significantly reduced at the 90-day follow-up compared with placebo;
- across the whole trial (n = 96), the overall difference in infarcted areas was not significant.
Both sentences are in the same abstract. They are not in tension with each other, and neither is a typo. What they describe is a trial whose primary comparison did not reach significance, plus a subgroup comparison that did.
The authors do not dress it up. The conclusion states that the clinical relevance is demonstrated through a randomised controlled trial, and in the same breath that further rigorous randomised studies are needed to assess the significance of early TB4 use after myocardial infarction reperfusion.
How a subgroup result becomes a headline
The subgroup is 43 of 96 people. A subgroup analysis on a trial that missed its primary endpoint is, structurally, a hypothesis generator. It is where you go to design the next trial, not where you stop.
There are well-known reasons subgroup findings run hot: the more subgroups you cut a trial into, the more likely one of them crosses a threshold by chance alone. The paper does not report a correction for the number of subgroups it examined, and it does not claim to.
The honest reading of both sentences together is narrower than either one alone: a plausible early-treatment window worth testing properly, in a trial powered to find it. That is a real result. It is not a demonstrated treatment, and the paper does not present it as one.
Where the rest of the human file actually sits
The cardiac trial is the interesting arm, but it is not the whole human record, and reading only it gives a distorted picture in the other direction.
The ophthalmic record is the one that has been carried furthest. A 2018 review in Expert Opinion of Biological Therapeutics by Sosne, who describes the route from the National Institutes of Health in 1998, tracks TB4 as a candidate for dry eye and neurotrophic keratopathy, with Phase 3 trials described as ongoing at the time of writing. The same peptide, in a different tissue, has a substantially further clinical file than the cardiac one.
Dermatology sits in between. A 2016 review in Vitamins and Hormones notes that TB4 increases the rate of dermal healing across several preclinical animal models, including diabetic and aged animals, and is active for burns, and that it also accelerated repair in Phase 2 trials in people with pressure ulcers, stasis ulcers and epidermolysis bullosa wounds — describing it as safe and well tolerated.
So the shape of the human file is: promising and incomplete in cardiology, further along in the eye, further still in wounds. That is a very different picture from “one positive mouse study”, and also a very different one from “a proven cardiac treatment”.
The mechanism argument is the part worth keeping
If you read the paper once, read the ErbB2 line again. The finding is that TB4’s effect on cardiac function and adverse remodelling in the mice was abolished by ErbB2 inhibition. That is the kind of detail that is easy to leave out of a summary and expensive to lose, because it is the difference between a compound that does something to a readout and a compound that acts through a pathway somebody has identified.
What it does not do is bridge to a person. A blocked pathway in a mouse explains the mouse. The human arm of the same paper is the part that failed to replicate, and the mechanism does not rescue it.
What the store page on this molecule is for
The commercial pages for peptides carry cardiac claims with varying degrees of care, and the honest thing to check on any of them is whether the claim points at the animal work or at the human trial. The cardiac research page on primalform sets out the same body of work, and the point of a page like that is the same as the point of this article: it should be possible to tell from the document which species each number came from.
What this does not show
It does not show that TB4 does not work on the heart. It shows that one 96-person trial did not reach significance on its primary comparison. That is a much weaker statement, and “did not work” is not it.
It does not show that the early-treatment window is real. The 43-person subgroup is a signal worth a trial. It is not a finding.
It does not show that the mouse data were wrong. The mouse arm was internally consistent, mechanistically specified and falsifiable — and the same blockade that removed the benefit is what makes it falsifiable rather than merely plausible.
It does not say anything about the eye or the skin. Different tissue, different trials, different state of knowledge. Lumping them together is how a thin cardiac file gets presented as a broad one.
If I had to put it in one sentence
The strongest claim for thymosin beta 4 in the heart rests on a mouse study with a named, blockable mechanism, while the one randomised human trial of 96 people missed its primary endpoint and reports its significant result in a 43-person subgroup — which is a reason to run the next trial, not a reason to skip it.
