A preprint study that has not yet been peer reviewed used a large language model to scan anonymized electronic health records for mentions of the peptide BPC-157, and its authors say documented use has climbed sharply. The research was conducted by nference, an AI startup that mines clinical records for insights, according to Venky Soundararajan, the study's author and the company's chief scientific officer. The work offers one of the few attempts to measure real-world use of a gray-market substance that has no insurance code and no uniform way for providers to record it.
The preprint counted roughly 54 confirmed patients using BPC-157 in the first quarter of 2022. That figure rose to 1,018 in the first quarter of 2026. Soundararajan said use appeared relatively limited during the pandemic before rising significantly over the last three years. He acknowledged the numbers are likely far higher than what the records captured.
BPC-157, short for body protection compound, was first identified in gastric juice in 1993 and is believed to support injury recovery by creating new blood vessels, encouraging cell growth, and repairing tendons, muscles, and ligaments. It is not approved by any regulatory body, has been studied mainly in animals, and has limited human clinical evidence. A 2025 systematic review found only 35 preclinical studies and one clinical study in humans.
The same review reported more than 50 million views for BPC-157 videos on YouTube and TikTok in 2024. Google Trends shows a 1,600 percent increase in traditional US web searches for the peptide over five years, and over 5,000 percent for YouTube searches. Soundararajan said the typical patient discussing BPC-157 with a doctor is getting younger, averaging in the upper 40s, and is most likely Caucasian. About two months pass between first use and disclosure to a physician, and most patients took the peptide orally rather than by injection.
Roughly a third of patients used BPC-157 without other therapies or drugs, while about half stacked it with substances including testosterone, non-steroidal anti-inflammatory drugs, and TB-500, the other half of the so-called Wolverine stack. Family practice and internal medicine doctors received the most reports, followed by gastroenterologists and orthopedic surgeons. Soundararajan flagged gaps in the data: 96 percent of patients who disclosed use were Caucasian, compared with 2 percent African American, far from the source data's 78 percent and 17 percent.
Only about 2 to 3 percent of cases reported adverse reactions, and Soundararajan said that data is anecdotal because severe cases requiring hospitalization may go unrecorded. Most positive effects were self-reported, physicians rarely documented their own observations, and 66 percent of cases showed no documented response on whether BPC-157 worked. Patients obtained the peptide from gray-market sources, so some may have taken something else entirely, and any recorded benefits could reflect the placebo effect.
Soundararajan said researchers are handicapped because they must rely on clinical notes alone, and he argued health record manufacturers should make it easier for doctors to record where patients obtained unapproved products. He also said prospective studies and gold-standard clinical trials are needed.
More technology news from TechManNews.
