Eleven studies of a peptide-based hemostatic gel applied prophylactically after endoscopic resection did not reduce delayed bleeding, in a meta-analysis that pooled 928 patients and looked at the comparative data from five trials head-on. The risk ratio compared with standard care was 1.34 (95 percent confidence interval 0.83 to 2.18), meaning patients who got the gel had roughly the same bleeding rate as patients who got conventional clip and thermal hemostasis, with the upper end of the confidence interval admitting some possibility of a small harm.
The paper, published online May 19 in Surgical Endoscopy ↗, is the cleanest pooled readout to date on whether self-assembling peptide gel earns its place in the post-resection toolkit. The answer it gives is no, not as currently used.
What the meta-analysis looked at
Endoscopic submucosal dissection (ESD) and endoscopic mucosal resection (EMR) are the two procedures used to remove suspect lesions from the lining of the stomach, esophagus, and colon through a flexible camera tube, without opening a patient up. Both leave a fresh wound on the inside of the gut. Delayed bleeding, where a clot dislodges hours or days after the procedure, is the main adverse event clinicians worry about. The standard prevention is mechanical: clip the visible vessels, cauterize the wound bed, watch the patient.
Self-assembling peptide (SAP) gel was supposed to add a passive layer to that. The most-used product is RADA16-I, sold commercially as PuraStat by 3-D Matrix. The peptide is a sixteen-residue sequence (alternating arginine, alanine, and aspartate) that exists as a soluble solution in syringe form and self-assembles into a nanofiber hydrogel scaffold the moment it contacts physiologic salt at the wound surface. The scaffold is meant to coat the resection bed, provide a passive barrier against rebleed, and dissolve over days. It has CE marking in Europe, regulatory clearance in Japan and Australia, and is used off-label in several US centers. The pitch is "spray on, walk away."
Observational data has been encouraging. Randomized data has been mixed. The authors searched PubMed, Embase, and the Cochrane Library through October 2025 and assembled 11 cohorts (prospective and retrospective) reporting delayed bleeding after ESD or EMR with SAP, with or without a comparator arm. Five of those cohorts had a direct comparison group on standard care. They are the ones the headline number comes from.
The numbers do not separate
In the five comparative studies, SAP plus standard therapy did not lower bleeding versus standard therapy alone (RR 1.34, 0.83 to 2.18, no heterogeneity). A sensitivity analysis restricted to the randomized controlled trials gave RR 1.53 (0.85 to 2.76, also no heterogeneity), drifting numerically in the same wrong direction. The subgroup limited to ESD-only studies came in at RR 1.16 (0.65 to 2.08). None of those intervals crossed clean below 1.0. None of them gave a winning point estimate, either.
The single-arm side of the meta-analysis pooled all 11 cohorts to get an absolute SAP bleeding rate of 6.7 percent (3.9 to 10.2 percent). That is in line with historical baseline rates after ESD and EMR in the literature. Among the 108 anticoagulated patients pulled out as a separate subgroup, the bleeding rate was 11.5 percent (6.3 to 18.0 percent), again consistent with the higher historical baseline that subgroup carries. The single-arm pool had real heterogeneity (I-squared 71.5 percent), which the authors flag as the main limit on how confidently any pooled rate can stand on its own. The comparative pools were homogeneous (I-squared zero), which makes the null result harder to argue with.
The conclusion the authors land on is the cautious version: "Further studies are required to define the patients most likely to benefit from SAP." The data they show in the body of the paper does not promise that any such subgroup exists. The comparative analyses were null in the overall pool, in the RCT-restricted pool, and in the ESD-only pool, all with no detectable between-study variability.
What this says about the peptide-as-material claim
Self-assembling peptide hydrogels are one of the cleanest demonstrations the peptide field has that a small designed sequence can act as a structural material instead of as a signaling agent. The RADA16 family has built a small literature on tissue engineering, regenerative scaffolds, drug delivery, and surgical adjuncts. The PuraStat product line has been the field's most prominent clinical translation of that work. The data this section has covered on peptide hemostats and surgical adjuncts has been mostly observational, mostly enthusiastic, mostly underpowered.
A 928-patient comparative meta-analysis is the bar that an established procedure has to clear once enough randomized data accumulates. SAP did not clear it for delayed post-endoscopy bleeding prophylaxis. That does not invalidate the peptide-as-material idea. It does say that the specific clinical use the field has marketed the platform hardest into (endoscopic hemostasis) is now the use case where the comparative evidence is least supportive.
The platform reading on peptidemodel is that "peptide as scaffold material" and "peptide as therapeutic ligand" remain two genuinely different research programs, and the readouts that matter for each are different. A null meta-analysis on a hemostatic gel does not transfer to a null read on receptor-targeted peptide therapeutics. It is a discipline-specific result that the field of self-assembling peptide hydrogels has to absorb on its own terms, and the next round of trials in this product class has to be designed to win against a much harder baseline than the observational literature implied.