Do compounded GLP-1 drugs have the same human evidence?

Search for compounded semaglutide or compounded tirzepatide and the page fills with sellers. Several state plainly that a compounded product works the same as the approved one, because it contains the same active ingredient. Others insist just as confidently that compounded preparations have never been studied in anyone.

Neither claim holds, and they fail for different reasons.

Compounded preparations have been studied, though the published record is far smaller than either side implies. None of it shows what the marketing figures claim, and almost every study in that record was produced by someone with money riding on the answer.

PubMed carries 106 records on compounded GLP-1 products. Exactly two administered such a product to a person and reported what happened.

What was actually given to people

The larger of the two is a retrospective cohort published in Diabetes, Obesity & Metabolism in March 2025. Ninety-four adults at a commercial weight-management studio received a compounded product weekly between June 2023 and January 2024. After three months, mean weight loss was 4.11 kg (SD 2.77), or 4.57 % of body weight (SD 2.96). The underlying data are not public, though the paper offers them on request.

Wherever the study is cited, its own limits go unmentioned. It had no control group, so nothing separates the drug’s effect from the programme built around it (the coaching, the weighing, the motivation of people who had already paid to be there). The product itself was a semaglutide and cyanocobalamin combination, a formulation with no approved-medicine equivalent, and the paper’s title never names the added vitamin. The interest declaration is one full sentence, not a note tucked into the acknowledgements, and it covers every author: all authors received salary support from the wellness chain whose customers were studied, among them co-authors at Stanford, Texas A&M and Mass General Brigham.

The body-composition numbers show why that omission matters. Participants lost fat (2.67 kg) and trunk fat (1.10 kg), but they also lost lean mass (1.43 kg) and skeletal muscle (0.88 kg), losses that account for roughly a third of the total weight lost. The paper reports both figures, then adds that as a proportion of total weight, lean and skeletal muscle increased, and concludes in its discussion that “body composition improved.”

Both statements are arithmetically true. If fat falls faster than lean tissue, lean mass rises as a share of body weight even as it falls in absolute terms. The two point in opposite directions, and it is the second that travels. The chain’s own summary of the study reports the weight figures accurately (9.06 lb is 4.11 kg, 5.89 lb of fat is 2.67 kg), then states an “approximately 1.2 % and 0.5 % increase in lean and skeletal muscle” without mentioning the absolute losses.

A reader of that page comes away believing they gained muscle, when the study itself says they lost 0.88 kg of it. That is claim transfer at a finer grain than borrowing a number from a different study: it presents the flattering half of the same study’s own data and leaves the rest out.

The second study is considerably smaller.

Compounded Tirzepatide Therapy for Weight Loss: A Health Economics & Outcomes Research Analysis, published in the International Journal of Pharmaceutical Compounding in January 2025, studied one participant: the author, a health-outcomes investigator between 55 and 65 with a BMI of 27 and no type 2 diabetes, who took a compounded product for roughly four weeks and wrote up the result. There was no control group and no blinding, and the sole listed affiliation was a one-person consultancy, in a journal published for the compounding industry. The paper states outright that it is the first article focused on compounded tirzepatide, all previous tirzepatide literature having addressed the branded drug.

That remains true. The only publication in the indexed literature centred on compounded tirzepatide is a four-week self-experiment with one subject, and it circulates anyway as scientific backing for a product class dispensed to hundreds of thousands of people.

The numbers in the advertising belong to different products

This is where the claim and the evidence quietly part ways.

A clinic page promises 16 to 22.5 % weight loss over 72 weeks. Another cites 10 to 15 % over 68 weeks. Both figures are real, measured in the phase 3 programmes of the approved products: randomised trials with control arms, thousands of participants and prespecified endpoints. Neither number was ever observed in anyone taking a compounded preparation, and no compounded cohort has run for 68 weeks.

Set beside those is the figure from the only published cohort of compounded users: 4.57 % over three months, uncontrolled, at a single site, in 94 people.

These are not competing estimates of the same quantity: the trial figures answer “what did this approved drug do against placebo over 68 weeks”, while the cohort figure answers “what happened to 94 customers of one studio over 12 weeks”. A page that prints the first number under the heading of the second has moved past overstating a result. It has attributed someone else’s result to a product that has never been measured that way.

There is no standard to compound against

Ask whether these products match, and a prior question surfaces: what would they be matching to? Per the American Diabetes Association’s February 2025 statement, the answer is nothing.

Compounding a copy of an approved medicine would normally follow a US Pharmacopeia monograph: a published specification for identity, strength, purity and performance, with the test methods to verify each. USP generally starts writing one a few years before a patent expires. For semaglutide and tirzepatide, no monograph exists.

There is therefore no external standard defining what a correct compounded semaglutide contains, and no required test to check it against. The ADA notes what fills that gap instead: some compounders add vitamin B12 or B6, and some substitute salt forms (semaglutide sodium or semaglutide acetate) for the base used in the approved products. A salt form is not chemically identical to the base.

Unlike an approved generic, a compounded product is never subjected to bioequivalence testing, and that single fact answers the article’s question at the regulatory level. Nobody is required to establish equivalence for these products, and as a result nobody has.

What the samples actually contained

The equivalence argument assumes that an identical active ingredient guarantees an identical outcome. Two laboratory studies published in 2026 show what that assumption skips.

The first, in Expert Opinion on Drug Safety in May 2026, tested ten samples of compounded tirzepatide with B12, obtained from compounding pharmacies, medspas and telehealth providers across seven US states. Potency ranged widely against the tirzepatide reference, with some samples containing as little as 43 % of the labelled amount, and every B12-containing sample carried a previously unknown impurity, a tirzepatide–B12 adduct: the two molecules chemically bound, present at up to 10 % of total polypeptide content. NMR work showed the binding changes the structure of the tirzepatide molecule itself and not merely what surrounds it.

Whether that matters clinically is unknown, and the authors say so, though they do raise a precedent: a deliberately engineered B12–GLP-1 conjugate, characterised and tested in animals, that turned out to have meaningfully different distribution and effects from its parent drug. That conjugate went through characterisation and animal testing before anyone used it; this one forms by accident in a vial and reaches a patient without anyone having looked at it first.

Ten samples is a small study. It establishes that the adduct occurs and is widespread among the products tested, without measuring how often or testing any people at all.

The second study, in Pharmaceutical Research in July 2026, is the larger one. Its authors bought semaglutide and liraglutide products from 36 commercial suppliers across North and South America, Europe and Asia (13 injectable compounded semaglutide products among them) and ran immunogenicity assays against dendritic cells from 15 healthy donors, plus mass spectrometry, photostability and fibrillation testing.

Every compounded and follow-on sample showed an impurity profile distinct from the originator, including amino-acid deletions and additions and impurities the authors could not identify. Almost all were made by chemically linking amino acids in sequence. The approved products are grown in cells instead, and that manufacturing difference is a plausible route to exactly the divergence the authors found. Under light exposure, compounded semaglutide diverged significantly from the originator in strength, total impurities and high-molecular-weight protein.

The paper’s own conclusion is the careful one: the effect of these impurity profiles on safety and efficacy has not been tested clinically. What the authors establish is potential immunogenicity, not demonstrated harm.

Who paid for the evidence

The record turns strange here, and no page in the search results mentions it.

The impurity paper is funded by Eli Lilly and Company, and all five authors are Lilly employees. Lilly manufactures the approved tirzepatide products that compounded tirzepatide substitutes for.

The immunogenicity paper comes from Novo Nordisk. Every author is an employee and shareholder, or was at the time of the work, and Novo Nordisk also supplied the originator products used as the comparison. Novo Nordisk manufactures the approved semaglutide and liraglutide products.

So the two studies establishing that compounded products differ in composition were run by the two companies whose approved products compete with them. The chain selling compounded semaglutide co-authored the one study reporting that it worked. The only tirzepatide paper in the record is a self-experiment in the compounding industry’s own journal.

None of this makes the findings false. Both manufacturer studies disclose their funding, methods and instruments. A mass spectrum is checkable, and a competitor is well placed to run one. A conflict of interest calls for reading the methods carefully rather than dismissing the findings.

The pattern itself is the finding. On whether these products are equivalent, almost nobody without a stake has looked. The closest thing to a disinterested source is the ADA statement, which declares it was funded from the association’s general revenue with no industry support of any kind. The FAERS analysis below comes from Geisinger and Binghamton University, with no product on either side, though its lead author sits on the board of the small medication-safety non-profit that contributed $850 to the work, and a co-author was on a research team supported by Pfizer and Lilly years earlier. Those are disclosures, and they do not disqualify the work. They are also the most independent funding this literature has.

Nobody is paying for neutral testing of what is in these syringes, and that absence is far from neutral ground between the two marketing claims: it is the reason the question stays open.

What the adverse-event reports show, and what they cannot

The pharmacovigilance study, in Expert Opinion on Drug Safety in March 2026, searched the FDA’s adverse event reporting system from 2018 to 2024: 81,078 reports for GLP-1 receptor agonists, of which 707 involved compounded products. With adjustment by logistic regression, reporting odds ratios were elevated for preparation errors (48.92; 95 % CI 12.63–189.6), contamination (19.00; 4.24–85.03), compounding and manufacturing issues (8.51; 5.17–14.0) and hospitalisation (2.35; 1.94–2.83).

A reporting odds ratio is a disproportionality measure. It is not a risk estimate and it does not establish cause. The database is voluntary and unverified, anyone may file, the denominator is unknown, and compounded GLP-1 products drew sustained press attention across this exact window, a factor that by itself lifts reporting. The authors say the study was not designed to show cause and effect.

The picture complicates in two directions at once, and neither can be left out.

The same analysis found compounded products had lower reporting odds for administration errors (0.29) and dosing errors (0.24). Set against that is FDA’s July 2024 alert describing dosing errors with compounded semaglutide at 5 to 20 times the intended dose, some requiring hospitalisation, driven by multi-dose vials, concentrations that varied between compounders, and confusion over whether a label meant millilitres, milligrams or units. A voluntary-report database and a safety alert built from case reports measure different things. One does not cancel the other.

The reporting channels themselves are not comparable. Manufacturers of approved drugs must report every adverse event they receive. State-licensed compounding pharmacies do not report to FDA at all, and outsourcing facilities report only events that are both serious and unexpected. The ADA adds that reporting often falls to the patient instead of the clinician. The 707 compounded reports come from a channel built to capture less.

For scale from FDA’s own dashboards: compounded drugs accounted for 45 % of US drug product recalls between 2012 and 2021 while making up under 3 % of prescriptions, and more than 70,000 vials of compounded GLP-1 products have been recalled since 2023, mostly for lack of sterility.

The legal footing changed, and the record is widely misreported

Large-scale compounding of these drugs was lawful because they were in shortage. That condition has ended, and the sequence is reported inconsistently almost everywhere.

FDA declared the tirzepatide shortage resolved on 2 October 2024. That decision was remanded on 22 October 2024 for re-evaluation as part of litigation, and replaced by a new decision on 19 December 2024. Anyone citing October 2024 is citing a determination that was withdrawn. The semaglutide shortage was declared resolved on 21 February 2025. Enforcement discretion for 503A pharmacies and 503B outsourcing facilities ran out on staggered dates through May 2025, and in 2025 a district court declined to block either determination.

What followed was a shift in product instead of a withdrawal. By autumn 2025, industry data cited in the impurity paper put 80 % of compounded semaglutide and tirzepatide prescriptions as combinations with added ingredients such as B vitamins, a formulation that is not a copy of an approved product and therefore not restricted in the same way.

FDA addressed exactly this on 1 April 2026, in guidance that consumer coverage largely left out. The agency names the semaglutide plus vitamin B12 combination specifically, stating it may be considered essentially a copy of a commercially available product when the route of administration matches and both components fall within 10 % of the approved strengths. A four-prescription threshold exists. FDA does not intend to act against a compounder filling four or fewer such prescriptions in a calendar month. Neither semaglutide nor tirzepatide appears on the 503B bulks list or the shortage list, either, which is the footing outsourcing-facility compounding would otherwise need.

What equivalence would actually require

None of that makes compounded products fake or the pharmacies selling them dishonest. It leaves one specific claim unsupported, and that claim is the one being sold.

Establishing that two products perform the same is a defined exercise. It requires comparing them in the same people or in matched groups, measuring how much active drug reaches the bloodstream over time, and showing the results fall within accepted bounds. For an injected peptide it also requires characterising impurities and testing immunogenicity, because it is the manufacturing route, not the molecular formula, that determines what is in the syringe.

No published study does this for a compounded GLP-1 product, and no manufacturer claims it does. The gap is not contested. It is simply not mentioned where the products are sold.

Original sources

• Chun et al., 2025, Diabetes, Obesity & Metabolism
• PubMed 39921911 — compounded tirzepatide HEOR analysis
• Jordan et al., 2026, Expert Opinion on Drug Safety — novel impurity in compounded tirzepatide/B12 products
• Kopp et al., 2026, Pharmaceutical Research — impurities and potential immunogenicity
• McCall et al., 2026, Expert Opinion on Drug Safety — FAERS analysis of compounded GLP-1 receptor agonists
• American Diabetes Association statement, Diabetes Care, 2025
• FDA — policies for compounders as national GLP-1 supply stabilizes

Independent editorial summary. The study authors, their employers, the journals, FDA and the American Diabetes Association are not affiliated with LifeScore. LifeScore has no commercial interest in any product named here. This article does not provide medical, dosing, sourcing or purchase advice.