The document that does the persuading in this market is the certificate of analysis. It carries one number, usually a purity figure in the high nineties, and it’s treated as the end of the argument. The published analytical literature says that number answers a different question than the one a buyer actually has.
Three national laboratories have independently found injectable material whose chromatographic purity would have read clean and whose molecule was not the one named on the label. A federal regulator has twice read real grey-market certificates in public and written down what those documents contained and what they left out. Meanwhile some of the most-quoted figures in this space, including the reconstituted storage windows everyone repeats, trace back to seller web copy rather than to any study.
This is a review of what’s been measured, by whom, in what material. Where nothing has been measured, that’s said plainly, because several of the confident claims circulating here sit on top of an empty evidence base.
What analysis of seized material has found
The largest dataset comes from border seizures and doping-control laboratories, not from consumer testing.
| Study | Material | Finding | |—|—|—| | Weber, Krug, Kamber, Thevis, Subst Use Misuse 2017, PMID 28156209 | 1,190 products seized at the Swiss border, 146 of them labelled as peptide hormones or growth factors | Fewer than 20% contained the claimed substance in the claimed amount | | Høj, Rasmussen, Dalsgaard, Linnet, Drug Test Anal 2021, PMID 33686802 | 36 samples submitted to Danish forensic chemistry for peptide or protein analysis, 2019 caseload | 9 of 36 (25%) contained no peptides or proteins at all | | Krug et al., Eur J Clin Pharmacol 2014, PMID 25168622 | 337 black-market products, German customs and the Cologne doping laboratory, 2010 to 2013 | 67 active ingredients found, 49 of them WADA-prohibited. 12.8% of findings were peptide hormones or growth factors | | Hullstein et al., Drug Test Anal 2015, PMID 26607218 | Confiscated products, Norwegian doping laboratory, 2011 to 2014 | 28% contained peptide or protein-based doping substances | | Fabresse et al., Forensic Sci Int 2021, PMID 33838562 | 110 products seized from French bodybuilders, 2016 to 2019 | Of 75 seized pharmaceuticals, 33% substandard on dose, 32% counterfeit on qualitative formulation, 19% genuine | | Barrios et al., Drug Test Anal 2025, PMID 40551438 | 324 samples across 14 laboratories in 13 countries, 5 years, covering growth hormone secretagogues and SARMs | 24% were presented to the buyer as dietary supplements. Quantitative data was limited, and some products were overdosed rather than under |
A 2024 report from the Cologne laboratory covering 41 products found that 53% of ingredients were not declared or were falsely declared, and describes one product labelled as semaglutide that contained insulin. Read the denominator carefully: the 53% is of 72 findings across 21 drugs, not of the 41 products, and peptide hormones were only about 7% of the doping-relevant compounds identified, with anabolic agents making up 89%. It is a conference proceedings chapter (Recent Advances in Doping Analysis vol 33, DOI 10.25847/mdi.2025.033) from the same laboratory behind several of the peer-reviewed papers cited here.
For context rather than as peptide data: a meta-analysis of 19 studies and 5,413 anabolic steroid samples across Europe and the Americas found 36% counterfeit (95% CI 29 to 43) and a further 37% substandard (95% CI 17 to 63), with high heterogeneity (Magnolini et al., BMC Public Health 2022, PMID 35842594). That’s a different product class and shouldn’t be used as a peptide failure rate.
The one study that bought from sellers and tested what arrived
Ashraf et al., J Med Internet Res 2024, PMID 39509151, is the only test-purchase study in this space with full analytical follow-through. The group screened 1,080 search-result links, identified 59 unique illegal online pharmacy sites, and measured more than 4.7 million visits to the top 30 affiliated domains between July and September 2023. Six test purchases were placed. Three vials were delivered. The three prefilled-pen orders were non-delivery scams, meaning half the purchases produced no product at all.
The three delivered vials were tested for sterility, endotoxin, and content by LC-MS. All three were free of viable microorganisms at the time of testing under Ph. Eur. 2.6.1. Content ran over label in every case, by 28.56%, 33.58% and 38.69%. The chromatograms showed a single peak per sample with no peptide-like impurities identified. Visual and packaging compliance failed on 13 to 14 of 22 criteria.
That study is also the source of a figure that gets badly misused. Its reported “purity” values of 7.7%, 8.97% and 14.37%, set against a vendor claim of “at least 99%”, are not chromatographic purity. The full text defines them as measured peptide mass divided by total lyophilised powder mass in the vial. In any lyophilised injectable that fraction is legitimately low, because the cake is mostly excipient, buffer salt, counterion and residual water. The vendor’s 99% is an area-percent claim. Comparing the two is a category error, and the study’s own chromatography found no peptide-related impurities. What the data supports is overage against label and an uncharacterised non-peptide mass.
Two further caveats on the same paper. Endotoxin was reported as under 2.1645, under 2.8658, and 8.9511 EU/mg, so two of the three values sit below quantitation rather than being quantified detections, and the authors applied no pharmacopoeial limit and made no pass or fail determination. On disclosure, two authors are employees of a data-science startup funded by government public-health contracts, one of them a co-founder and co-owner, and the same author edits a journal in the publishing family that published the paper.
Purity and identity are different measurements
An HPLC-UV purity figure answers a narrow question. Of the UV-absorbing material that eluted from that column under those gradient conditions, what fraction of total integrated peak area sat under the main peak.
Each qualifier carries weight. Detection at 214 to 220 nm or 280 nm under-represents or misses non-chromophoric material such as salts, counterion and many synthesis scavengers. Anything retained on the column or eluting in the void isn’t in the denominator. Closely related impurities can co-elute with the main peak, which FDA describes as difficult to identify and quantify without sophisticated analytical methods. And the figure is relative, so it says nothing about how much peptide is in the vial.
Breindahl et al., Drug Test Anal 2015, PMID 24771717, is the clean demonstration of that last point. Vials were bought from three online shops, all claiming 10 mg. Measured content ranged from 4.32 to 8.84 mg, which is 43% to 88% of claim. Not one vial met label. Unknown impurities ran 4.1% to 5.9% in vials from two shops and below the limit of quantitation in the third.
The identity failures are the more instructive set, because none of them would move an area-percent number.
Gajda et al., Drug Test Anal 2019, PMID 30136411, analysed preparations seized by Danish customs and identified analogues of GHRP-2, GHRP-6, ipamorelin and modified GRF (1-29). The paper states that “in all cases, the detected modification involved the addition of an extra glycine amino acid at the N-terminus” and concludes that targeted detection methods need updating. The abstract doesn’t give a sample size, so no n should be quoted for it.
Krug et al., Growth Horm IGF Res 2018, PMID 29864719, working across the German Sport University Cologne, the Norwegian doping laboratory, the German federal revenue administration and a Leibniz analytical institute, found the same defect independently: Gly-GHRP-6, Gly-GHRP-2 and Gly-ipamorelin, with the identity of two confirmed by custom synthesis rather than spectral inference alone. The same paper reports a modified growth hormone of 192 amino acids instead of 191, carrying an extra N-terminal alanine, at a monoisotopic mass of 22,195 Da against 22,124 Da for authentic 22 kDa hGH.
Popławska and Błażewicz at the National Medicines Institute in Warsaw, Drug Test Anal 2019, PMID 30051972, sequenced a novel heptapeptide of MW 874.02 Da from a seized injection vial and identified it as a glycine analogue of GHRP-2, noting it “may indicate a new approach to circumvent detection of doping practices.”
Three national laboratories, four different secretagogues, the same single-residue extension. An added glycine is 57.02 Da. On a chromatogram it’s still one sharp peak.
Other documented wrong-molecule findings fill out the picture. Esposito, Deventer and Van Eenoo, Drug Test Anal 2014, PMID 25283153, found a hGHRH analogue with incomplete C-terminal amidation in a confiscated product, a difference of 0.98 Da, and wrote that “the presence of partially amidated molecule reveals the poor pharmaceutical quality of the preparation.” Kohler et al., Growth Horm IGF Res 2010, PMID 20675162, found a His6-tagged Long-R3-IGF-I in an injection vial and concluded “the product may rather be a by-product from biochemical studies than synthesized for injection purposes.” Walpurgis et al., Drug Test Anal 2014, PMID 25219942, characterised an 18 kDa fusion protein appearing in several unlabelled products from independent sources and found, after thrombin cleavage and LC-MS/MS, that “the unknown protein is only the product of an empty expression vector without the DNA insert of interest.” The same group had earlier identified fibroblast growth factor 1 in an unlabelled product taken by German customs, Drug Test Anal 2011, PMID 21998075.
Mass spectrometry is what caught all of these, and it has limits of its own. Erckes et al. at the ETH Zurich Laboratory of Pharmaceutical Analytics, RSC Med Chem 2026, PMID 41541711, state that “conventional chromatographic methods and standard mass spectrometric analyses often fail to distinguish structurally similar peptides with nearly identical physicochemical properties and masses,” and show that CID and ETD fragmentation can separate deamidation and isoaspartate products without chromatographic separation. The same work reports that amidated peptides remained stable in neutral aqueous-organic mixtures or at lower temperatures, which is the other half of the picture. A review from a large pharmaceutical manufacturer’s laboratories, Lian et al., J Am Soc Mass Spectrom 2021, PMID 34110145, reaches the same place from the manufacturing side, with structural isomers and epimers as the hard case. All four authors of that review are employees of the manufacturer.
Why identity is the measurement that matters has now been tested rather than asserted. Mattei et al., Front Immunol 2025, PMID 41445733, applied three in-silico tools, in-vitro class II HLA-DR binding and a human PBMC T-cell proliferation assay to teriparatide and its product-related impurities, and found that “the orthogonal approaches identified multiple impurities as more immunogenic than” the parent peptide. Read that paper with its disclosure in view: five authors work for a private immunogenicity-screening company whose tools are among the methods evaluated, and four are FDA staff.
What a certificate of analysis actually certifies
The strongest evidence here isn’t from a journal. It’s from FDA’s Pharmacy Compounding Advisory Committee briefing documents, which are a rare case of a regulator auditing certificates from this supply chain and publishing what it found.
On ipamorelin (FDA briefing document, PCAC meeting 29 October 2024, fda.gov/media/182088), FDA wrote that “in the CoA provided in the nomination package, there is only a purity test limit of ≥95% with the testing result of 99.72%, and there is no impurity attribute control to demonstrate the impurity profiles.” A wider literature search turned up a single impurity limit of under 2% in one other certificate, with no information on what those impurities were. For the acetate salt, FDA recorded that “there is no testing result for the control on impurities, aggregates, and bioburden/endotoxin levels.” No certificate at all was supplied for the free base. FDA concluded both forms were “not well-characterized from the physical and chemical characterization perspective,” noted there’s no USP drug substance monograph for either after searching USP-NF, the European Pharmacopoeia and others, and stated it couldn’t rule out immunogenicity from the uncharacterised impurities and aggregates, particularly for injection.
The same document records a certificate that contradicted itself. Two nominating pharmacies submitted documents where “the Certificate of Analysis (CoA) submitted with each nomination package refers to one BDS by name in the title and a different BDS by the molecular weight/formula.” Both were titled for the acetate salt and both listed the CAS number, molecular formula and molecular weight of the free base. That happened in a package prepared for a federal advisory committee, under maximum scrutiny.
The BPC-157 briefing document (PCAC meeting 23 to 24 July 2026, fda.gov/media/193343) reads the same way. One nominator’s certificate reported purity of 97.8% against a limit of ≥95%, plus maximum individual and total impurity results of 1.1% and 2.2%. FDA’s comment: “the impurity profiles are unclear because the impurities were not identified or specified.” The second nominator’s certificate carried a purity result of 98.34% and no impurity control at all. FDA also noted no information on residual solvent testing and no way to tell whether bioburden or bacterial endotoxin testing was in place for a substance proposed for an injectable.
FDA lists what a certificate for an injectable would need to cover and what these didn’t: sterility, bacterial endotoxins, foreign particulates, bioburden, peptide-related impurities, aggregates, particle size and residual solvents. FDA also explains, in both documents, why the impurities are the hard part. Peptide synthesis produces impurities from “incomplete coupling reactions, truncations, or side reactions,” and those impurities are “typically similar in structure to the target peptide.” Structurally similar peptides are exactly the species most likely to co-elute with the main peak.
What drug manufacturers are allowed to do with a supplier’s certificate
US drug GMP answers this directly. 21 CFR 211.84(d)(1) requires that “at least one test shall be conducted to verify the identity of each component of a drug product.” Subsection (d)(2) permits a manufacturer to accept a report of analysis from a component supplier only “provided that at least one specific identity test is conducted on such component by the manufacturer, and provided that the manufacturer establishes the reliability of the supplier’s analyses through appropriate validation of the supplier’s test results at appropriate intervals.”
A licensed manufacturer is forbidden from accepting a supplier certificate as proof of identity. It has to run its own identity test on every lot and periodically re-validate the supplier’s numbers. Anyone holding only a PDF occupies a position the regulations specifically don’t allow a regulated firm to occupy.
What third-party testing can and can’t settle
Two things here are documented and the rest is mechanism.
Documented: the failure modes above are observed, not hypothesised. Purity-only reporting, unidentified impurities, missing endotoxin and bioburden and aggregate and residual-solvent data, a certificate whose title and chemistry describe different substances, and no certificate at all for one of two nominated forms.
Not documented, and it should be labelled as inference: a certificate attests to a sample that reached a laboratory, and nothing in the transaction binds that sample to the vial in front of a buyer. A lot number that doesn’t appear on the vial, or a certificate dated well before the shipment, breaks the link entirely. ISO 17025 accreditation is scope-specific, attesting that a laboratory is competent in particular methods within its accredited scope, and says nothing about the product, the sampling, or whether the sample was representative. A laboratory reports what it’s asked to test, so “third-party tested” against a purity-only scope is a true statement that answers none of the safety questions.
One more absence is worth stating. No peer-reviewed or regulatory work could be located that audits grey-market certificates against independent testing of the same shipped lot. The instructional content on how to read a peptide certificate appears to be produced entirely by sellers or seller-adjacent sites. That’s a finding about the information supply, not evidence about product quality.
Batch to batch is unmeasured, not measured and bad
No study was found that tested multiple lots of the same labelled product from the same source. Anyone claiming batch consistency, and anyone claiming batch chaos, is going past the data.
What can be shown is spread between sources for the same nominal product. Melanotan II across three shops ran 4.32 to 8.84 mg against a 10 mg claim, with quantified impurities in two of three shops and the third below the limit of quantification (PMID 24771717). Semaglutide across three sellers ran 28% to 39% over label, consistently over but nowhere near target (PMID 39509151).
There’s also a structural reason to doubt stability of identity over time from one source. The Polish group framed its novel glycine analogue as a possible route to circumvent detection. A supply chain with an incentive to change the molecule isn’t one where batch identity can be assumed constant.
The storage numbers circulating here came from product pages
FDA’s briefing documents contain the stability figures that get quoted everywhere: lyophilised material stable at room temperature for 3 weeks, reconstituted solution stable 2 to 3 weeks at 4 °C and 3 to 4 months at -20 °C. The footnotes show FDA’s sources for those numbers are commercial peptide vendor product web pages, cited because, as the agency states, characterization data weren’t found in the publicly available scientific literature.
The same windows appear verbatim for two chemically unrelated compounds, a C-terminally amidated pentapeptide containing unnatural residues and a 15-residue free-acid gastric peptide. Identical numbers for unrelated molecules is what boilerplate looks like. A search for peer-reviewed solution-stability data on GHRP-2, GHRP-6, ipamorelin or melanotan II returned nothing indexed.
The degradation chemistry that has been measured is less forgiving than the marketing copy. Malgave et al., Eur J Pharm Biopharm 2025, PMID 40490042, stressed semaglutide at 25, 40 and 60 °C for 28 days and at 80 °C for 7 days across pH, buffer species and molarity, and identified 13 distinct degradation impurities, 6 of which formed under every condition tested. pH was the dominant factor, and four impurities were absent across all buffered conditions but appeared when water alone was the solvent. The ETH Zurich group (PMID 41541711) confirmed isoaspartate formation under mildly basic conditions such as PBS, and found that acidic conditions, particularly with the trifluoroacetic acid used as standard in solid-phase synthesis and HPLC purification, caused substantial deamidation by hydrolysis. C-terminal amides were markedly more susceptible, and the authors note these events “can occur under routine laboratory conditions.” Ipamorelin, GHRP-6 and melanotan II are C-terminally amidated, and the one confirmed amidation failure found in a real seized product sits at exactly that site (PMID 25283153).
On preservatives the literature is genuinely split and shouldn’t be flattened. Roy et al., J Pharm Sci 2005, PMID 15614819, studied a lyophilised protein reconstituted with 0.9% w/v benzyl alcohol in water and found aggregation that correlated with structural perturbation in the dried solid, with similar effects documented for recombinant interferon-gamma (PMID 9210188) and, at pH 7.0 but not pH 3.5, for G-CSF (PMID 16729274). The same Roy paper reports that during storage of the reconstituted material at room temperature, benzyl alcohol did not accelerate aggregation. Against that, Li et al., Mol Pharm 2022, PMID 35917158, studied an acylated 31-residue peptide by solution NMR and reported that “the addition of benzyl alcohol does not induce aggregation… and has no chemical shift perturbation,” while 1% m-cresol produced insoluble aggregates amounting to 25% w/w of the peptide after 24 hours at room temperature. All ten authors of that paper work for a large pharmaceutical manufacturer. Preservative-induced aggregation is real, peptide-specific and not predictable from first principles, which is what formulation development exists to settle.
FDA states the framing point directly in both briefing documents: the stability, pharmacological activity and immunogenic properties of peptides are highly sensitive to the manufacturing process and the quality attributes of the finished product.
Regulatory status, and what “research use only” does not do
Since 23 March 2020, under the final rule defining the term biological product (85 FR 10057, 21 February 2020), a polymer of 40 or fewer amino acids is a peptide and is regulated as a drug under FD&C Act section 505 rather than as a biologic. Semaglutide at 31 residues, BPC-157 at 15, ipamorelin at 5 and melanotan II at 7 all sit on the drug side of that line.
Under section 503A a compounding pharmacy may use a bulk drug substance only if it has a USP or NF monograph, is a component of an FDA-approved drug, or appears on the 503A Bulks List, and the substance must come with a valid certificate of analysis. FDA’s interim policy sorts nominated substances into three categories, where category 2 means adequately documented but carrying identified safety risks. Peptides currently listed in category 2 include GHRP-2, GHRP-6, ipamorelin acetate, ibutamoren mesylate and kisspeptin-10, with FDA’s cited concerns running to immunogenicity and, for two of them, reported serious adverse events including death.
A widely repeated claim says FDA removed a dozen peptides from category 2. FDA’s page, last updated 22 April 2026, lists those substances under a separate heading introduced with the words that they “were withdrawn by the nominators.” The nominators withdrew. FDA didn’t clear anything. Those substances still have no monograph, aren’t components of approved drugs and aren’t on the Bulks List, so compounding them under 503A remains outside the statutory conditions. Withdrawal moved them out of a warning category, not into a permitted one.
The committee record is worth reading for how the voting members weigh documentation. At the 29 October 2024 meeting (minutes at fda.gov/media/185412), ipamorelin free base was rejected for the 503A list 0 yes to 12 no with 1 abstention, ipamorelin acetate by the same margin, kisspeptin-10 by 0 to 11, and ibutamoren mesylate by 1 to 13. The recorded reasoning for ipamorelin was “a lack of information supporting safety and efficacy shown in the available data,” and on a separate substance members cited “the lack of a USP monograph or a certificate of analysis” as grounds to vote against. The same meeting included an FDA presentation titled Immunogenicity Risk of Compounded Peptides. Seven more peptides went before the committee on 23 and 24 July 2026 with FDA proposing not to add any of them. Secondary reports describe the committee going against that proposal on several, but no FDA minutes, transcript or vote summary were posted at time of writing, so that outcome shouldn’t be stated as fact.
One more correction. The impurity thresholds, where anything above 0.5% is not acceptable and the 0.10 to 0.5% band requires identification, characterisation and justification, quoted in press coverage and in at least one vendor-commissioned release aren’t general peptide rules. They come from a May 2021 guidance covering five named peptides in the narrow case of a generic synthetic peptide referencing an rDNA-origin listed drug, and FDA has withdrawn that guidance on the basis that it no longer reflects current scientific thinking. There’s no approved specification for ipamorelin, BPC-157, TB-500, melanotan II or retatrutide. Nothing exists to be in spec against.
As for the disclaimer on the vial, FDA’s position is settled and recent. Warning letter 735063, dated 24 August 2026, states: “Despite statements on your product labeling marketing your products ‘for research use only’ and ‘not for human or veterinary use,’ evidence obtained from your website establishes that your products are intended to be drugs for human use.” Under 21 CFR 201.128 intended use is established by objective evidence of the seller’s intent, and in these letters FDA drew that evidence from website claims of physiological benefit and from the practice of offering reconstitution solution alongside the injectable products. The letters cite unapproved new drug violations under sections 301(d) and 505(a). An earlier letter from February 2025 added a misbranding count under section 502(f)(1). Roughly fourteen such letters have issued since February 2025, five of them on a single day. The separate FDA guidance on research-use-only labelling governs in vitro diagnostic products and gives no shelter here.
What this leaves a reader able to check
The documented record supports a short list of questions, and FDA supplied the list.
Identity confirmed by a method orthogonal to HPLC, against a reference standard, on the lot in hand. That’s the measurement that would have caught a 57.02 Da glycine extension, a 0.98 Da amidation failure, a 192-residue growth hormone or an empty expression-vector product, and it’s the one measurement GMP refuses to let a manufacturer take on trust from a supplier.
Impurities named, not counted. A figure of 2.2% total means nothing without knowing what the 2.2% is, which is FDA’s own objection to two real certificates.
The attributes FDA identifies as critical for an injectable and found missing from every certificate it read: sterility, bacterial endotoxins, bioburden, aggregates, particulates, residual solvents. Add water content and counterion content, both of which determine how much peptide a stated milligram figure actually represents.
A traceable line from the document to the vial. Lot number matching, dated in the right order, sampled from the material shipped.
The failure mode in this market isn’t the one people picture. The best test-purchase data found sterile vials, clean single-peak chromatograms and content over label rather than under. What the literature documents instead is that no specification exists to test against, no lot is traceable, no identity method is routinely applied, and the single document provided answers a question nobody needed answered.
This article is a review of published research. It is not medical advice, and nothing here is a recommendation to use any compound. The substances discussed are research-use-only materials, they are not approved for human use, and nothing above describes how to use them.
