Gray-Market Retatrutide
Real-World EHR Data

An August 18, 2026 preprint used de-identified records from a federated U.S. electronic-health-record network covering about 29 million patients. Investigators identified 652 people whose notes documented retatrutide exposure. Among the 531 with a documented supply route, 71.2% reportedly obtained a product outside clinical-trial participation.

In a matched analysis, the blinded trial-participant cohort lost 15.5% of body weight at 6–12 months. The direct-to-consumer or compounded cohort lost 7.2%—less than half the trial-cohort result and statistically indistinguishable from matched tirzepatide users. Both retatrutide cohorts showed an early increase in heart rate.

These findings do not prove what any outside-trial product contained. “Confirmed exposure” means the clinical notes supported that a patient reported taking retatrutide; no direct chemical analysis verified the identity, dose, purity, sterility, or formulation of those products.

Retatrutide remains investigational and has no approved product or National Drug Code. The researchers therefore searched clinical text rather than pharmacy claims. A large language model reviewed full notes to distinguish documented use from discussion, family-member use, or unclear mentions, then extracted supply route and reported start date.

A clinician manually reviewed 320 sampled extractions at one participating site. Prevalence-weighted accuracy was 99.8% for exposure classification and 90.3% for supply-route assignment. That supports the note-classification method, but it does not validate the underlying truthfulness or completeness of patient disclosures.

The analytic cohorts included 89 blinded trial participants, 243 direct-to-consumer or compounded users, 890 semaglutide initiators, and 890 tirzepatide initiators. Matching covered age, sex, race, baseline BMI, diabetes status, prior incretin exposure, and time since the most recent prior exposure.

Among 378 direct-to-consumer users with an outside-trial route, 57.4% reportedly obtained the product through online or telehealth vendors, 29.4% through compounding pharmacies, and 8.7% through wellness, medical-spa, or weight clinics. Smaller categories fell below the study’s reporting threshold.

Documented initiations grew approximately 1.8-fold per quarter from October 2023 through March 2026. The notes also described co-formulations with cagrilintide, high-dose tirzepatide, BPC-157, AOD-9604, NAD+, or IGF analogues, along with nonstandard labels such as “GLP-3” and “triple-G.”

At 6–12 months, mean weight loss was 15.5% in the blinded trial-participant cohort, 7.2% in the direct-to-consumer or compounded cohort, 7.7% among matched tirzepatide users, and 4.6% among matched semaglutide users.

The outside-trial result was significantly lower than the trial-cohort result (P=0.004) and statistically indistinguishable from tirzepatide (P=0.79). The preprint compared the 15.5% trial-cohort estimate with a 16.9% weighted mean across retatrutide and placebo arms in TRIUMPH-1 through TRIUMPH-4.

That comparison requires care. The EHR trial cohort remained blinded, so it could include placebo recipients. Follow-up and persistence also differed: treatment duration documented in notes had a median of 83 days in the trial cohort and 56 days in the direct-to-consumer cohort. Shorter exposure, discontinuation, adherence, and missing measurements may all contribute to the gap.

At three months, heart rate increased by 4.3 beats per minute in the trial cohort (P=0.028) and 2.5 beats per minute in the direct-to-consumer cohort (P=0.011). The matched semaglutide and tirzepatide groups did not show a significant change. By six months, the mean changes in both retatrutide cohorts had returned toward baseline.

An exploratory analysis suggested a 0.28-bpm increase for each percentage point of weight lost among pooled retatrutide users. The trend did not reach conventional statistical significance (P=0.099), included only 79 retatrutide users with paired measurements, and was not monotonic across weight-loss bands.

After Benjamini–Hochberg correction, the pooled retatrutide cohorts had higher note-derived incidence in 12 of 16 symptom categories versus both semaglutide and tirzepatide. Compared with semaglutide, cardiovascular symptoms had a rate ratio of 1.56 (95% CI 1.26–1.92; q<0.001), neuropsychiatric symptoms 1.95 (1.61–2.36; q<0.001), and hypersensitivity 1.82 (1.32–2.47; q=0.001).

Tachycardia specifically had a rate ratio of 1.90 versus semaglutide and 2.88 versus tirzepatide. A negative-control group of 108 benign or structural findings was null against semaglutide, which argues against a purely generic documentation effect.

Association is not causation. Retatrutide users had denser clinical documentation, symptom detection relied partly on note text, and the analysis did not adjust for documentation volume. Product uncertainty, co-formulations, selection effects, underlying illness, and residual confounding could all influence the observed rates.

Three-point MACE and expanded MACE estimates were based on low event counts and approximately 290 retatrutide person-years. Confidence intervals were wide and crossed 1.0. No statistically significant difference emerged versus either matched comparator.

This is an underpowered null result, not evidence that risk is absent. The measurable heart-rate effect and symptom signals require longer follow-up and randomized cardiovascular-outcomes data before their clinical-event meaning can be established.

The strongest conclusion is not that every outside-trial product was inactive. The early heart-rate rise and the observed weight loss suggest that at least some users encountered pharmacologically active material. What the study cannot show is whether that material was authentic retatrutide, correctly dosed, uncontaminated, or consistent across suppliers.

The outside-trial cohort achieved roughly half the trial-cohort weight loss while retaining a recognizable physiologic signal and carrying more documented symptoms than matched users of approved incretin therapies. That combination makes product identity, concentration, co-formulation, persistence, and clinical supervision central unanswered questions.

For compound background, see the Retatrutide research guide. For evaluating laboratory evidence attached to any research product, review how to read a peptide COA and how Peptide Protocols COA grades work.