Publications/Public health
Public health
September 7, 2026
Clinical-note surveillance confirmed BPC-157 use in 1,039 patients and found a 33-fold rise in newly documented quarterly users from 2020 to 2026; among patients with a known consumer source, compounding pharmacies and gray-market peptide vendors predominated. Outcome documentation was incomplete and confounded, so this retrospective preprint cannot establish safety or efficacy.
Authors
AJ Venkatakrishnan, Adhikaar Marwaha, Christopher J. Gregg, Gowtham Varma, Matthew Hurchik, Karthik Murugadoss, Santhosh Shivabasappa, Venky Soundararajan
nference; Metabolism Agentic Intelligence Atlas (MAIA); nference Labs

Overview
BPC-157 is an unapproved synthetic peptide marketed for musculoskeletal healing, gastrointestinal complaints, and performance despite very limited human evidence. Because use often occurs outside conventional prescribing pathways, structured medication records may not capture exposure. Investigators used clinical-note curation to describe how documented BPC-157 use changed over time, where patients reported obtaining it, why they used it, and what outcomes appeared in routine care documentation.
Study design & methods
This retrospective observational study searched de-identified clinical notes from a federated U.S. EHR network containing 15.2 million patients with at least one note from 2020 through the second quarter of 2026. An LLM reviewed the complete note history of 1,536 patients with a BPC-157 mention and classified 1,039 as confirmed users. Additional schema-constrained passes extracted supply route, indication, co-therapies, administration route, discontinuation, symptomatic or functional response, and adverse-event documentation. A physician reviewed stratified samples at one participating site; validation accuracy varied by task, including 87.2% for exposure ascertainment, 93.8% for supply route, 88.2% for response direction, and 93.9% for drug-attributed adverse-event extraction. Analyses were descriptive, used Wilson 95% confidence intervals for proportions, controlled prespecified comparison families with the Benjamini-Hochberg procedure, and suppressed cells with fewer than 11 patients.
Use rose 33-fold as sourcing and outcomes remained sparsely documented
Quarterly newly confirmed users increased from 4 in the first quarter of 2020 to 134 in the first quarter of 2026, an approximately 33-fold rise. Among 205 users with a documented consumer source, 74 obtained BPC-157 from compounding pharmacies and 71 from peptide or gray-market vendors. Co-use of at least one other therapeutic agent was documented for 525 of 1,039 users, most commonly testosterone, NSAIDs, TB-500, and corticosteroids. A directional symptomatic response was available for 354 users: 279 reported improvement, 58 no change, and 17 worsening, while 685 users had no ascertainable response direction. Documented adverse-event signals were uncommon but likely under-reported, including neuropsychiatric, dermatologic or injection-site, and gastrointestinal events in approximately 1.3% to 1.4% of confirmed users.
Interpretation & limitations
The study documents rapidly increasing BPC-157 disclosure in participating health systems and shows how clinical-note curation can support surveillance of unapproved products that lack structured prescription records. It does not establish efficacy, safety, product identity, dose, adherence, or causality. Exposure and outcomes were patient-reported, follow-up was as observed, there was no active comparator, the cohort represents people who disclosed use to clinicians, most users lacked complete source, duration, and response information, and concomitant treatments, placebo effects, selection, reporting, and confirmation bias could explain perceived outcomes. Validation was single-site and not blinded, several fine-grained extraction tasks had less than 80% agreement, and the federated network is not representative of the U.S. population. Prospective pharmacokinetic, tolerability, and controlled efficacy studies are needed. Evidence confidence: Preliminary. Rationale: the large underlying EHR network and physician-validated extraction support the descriptive utilization signal, but incomplete documentation, self-selection, no comparator, unmeasured co-treatment, and preprint status sharply limit outcome interpretation. Confidence variables: design-retrospective descriptive study; exposure-note-confirmed but self-reported; outcomes-incomplete and nonstandardized; validation-single-site and task-dependent.
BPC-157 Gray-market therapeutics Clinical-note surveillance
Retrospective observational EHR study LLM-assisted clinical-note curation
Correspondence to:
Venky Soundararajan (venky@nference.net)
On this page
Overview
Methods
Results
Interpretation