Third-Party vs In-House Testing: What a Peptide COA Actually Proves
Every research-peptide supplier in the United States now advertises testing. Most advertise “third-party tested” specifically, on the reasonable theory that a buyer trusts an outside laboratory more than the vendor’s own bench. The theory is reasonable. The conclusion usually drawn from it — that a third-party certificate settles the question of what is in the vial — is not.
The strongest available evidence on this point does not come from the research-supply market at all. It comes from the world’s national measurement institutes, and it is uncomfortable reading for anyone who treats a purity percentage as a fact rather than as a measurement with a method attached.
Eight of the best laboratories on earth, one vial, a 220 mg/g spread
Between 2016 and 2020 the Consultative Committee for Amount of Substance ran a formal key comparison on peptide purity, coordinated by the Bureau International des Poids et Mesures and the National Institute of Metrology in Beijing. The study material was synthetic oxytocin. Eight national metrology and designated institutes took part: BIPM, NIM (China), NRC (Canada), NMIJ (Japan), LGC (United Kingdom), INMETRO (Brazil), TÜBİTAK UME (Turkey) and NMIM (Malaysia).
These are not vendor laboratories. They are the institutions that define what a kilogram and a mole mean in their respective countries, working on a material distributed from a single homogeneous batch, with unlimited method freedom and no commercial incentive of any kind.
The agreed reference value for the oxytocin content came out at 787.2 mg/g, with an expanded uncertainty of ±12.9 mg/g. The individual laboratory results ranged from 766.3 mg/g at the low end to 987 mg/g at the high end — a spread of roughly 220 mg/g on the same material. Every result except the highest agreed with the reference value within uncertainty; the outlier came from a determination that had not been corrected for peptide-related impurities.
Read that last sentence twice, because it is the whole argument. Expressed as a percentage, the rejected result was 98.7% — a number that would look entirely at home on a supplier’s product page. The results that survived scrutiny clustered near 78.7%. The difference between them was not the quality of the laboratory. It was whether the measurement corrected for the things in the vial that are not the peptide: water, counterion, residual solvent, and the structurally related peptide species that co-purify with the target. The same comparison put those peptide-related impurities at 31.6 mg/g, with an expanded uncertainty of ±1.4 mg/g.
An independent laboratory that runs an uncorrected method produces an independent number that is wrong. Independence and correctness are separate properties, and only one of them is visible on a certificate.
The method contributes more variance than the letterhead
The metrology result is not an isolated curiosity. In 2019 the United States Pharmacopeia’s Biologics Department published a multi-laboratory collaborative study in the Journal of Pharmaceutical and Biomedical Analysis (Li and colleagues, volume 166, pages 105–112; PMID 30640042), comparing three ways of assigning content to a peptide reference standard: an HPLC assay against an external standard, quantitative nuclear magnetic resonance, and amino acid analysis. The author list runs across USP, NRC Canada, NIST, Sweden’s Medical Products Agency and the UK’s NIBSC — nineteen authors, drawn from institutions with no commercial stake in the answer.
Their finding was that the HPLC assay showed the lowest inter-laboratory variability of the three, and that quantitative NMR merited further exploration as a primary method on grounds of simpler operation and shorter analysis time. The useful part for a research buyer is not which method won. It is that three defensible methods, run by competent laboratories on one material, did not return one number. Method selection is a variable of the same order as laboratory selection, and a certificate that names the laboratory but describes the method as “HPLC” has disclosed the less important half.
What in-house testing is genuinely good at
The case against in-house testing is usually made as a conflict-of-interest argument, and the conflict is real: a supplier grading its own material has an obvious incentive in one direction. But the argument is often stretched into a claim that in-house data is worthless, which is false and worth correcting.
In-house analysis is fast, it is cheap enough to run on every batch rather than on a sample of batches, and it is run by people who know the synthesis that produced the material and therefore know which impurities to look for. A manufacturer that lyophilizes its own material knows whether a given lot saw a long drying cycle and can target oxidation products accordingly. An outside laboratory receiving an unlabeled vial cannot. In pharmaceutical manufacturing, in-process control is in-house by definition, and nobody regards that as a scandal.
The problem with in-house data is not accuracy. It is verifiability. A buyer has no way to distinguish a careful in-house determination from a careless one, or from a number that was never measured at all, because the only evidence for it is the assertion of the party selling the material.
What third-party testing is genuinely good at — and its four weak points
Third-party testing solves the verifiability problem, partially. It introduces a party whose reputation is damaged by publishing a wrong number, and it produces a document with a report identifier that a buyer can quote back. That is a real improvement, and it is why the practice is worth paying for.
It does not solve four things, and a research buyer evaluating a supplier should treat these as the actual checklist:
Who chose the sample. The vendor selects which vial goes to the laboratory. Unless the sampling is documented — how many vials, drawn from where in the lot, by whom — the certificate describes one vial, not a batch. This is a different question from purity and it is the one most certificates leave unanswered.
What the laboratory’s accreditation covers. ISO/IEC 17025:2017, General requirements for the competence of testing and calibration laboratories, is scoped: an accredited laboratory is accredited for named methods, not in general. “ISO 17025 accredited” on a supplier page tells a buyer nothing unless the scope includes the method that produced the number on the certificate.
Which corrections were applied. As CCQM-K115.b demonstrated on the world’s most capable laboratories, this variable can move the answer by twenty percentage points. Area-percent purity from a chromatogram and mass-fraction content are different quantities, and a certificate that reports only the first has not told a buyer how much peptide is in the vial.
What the method cannot see. Which brings us to the failure mode that neither model catches.
The undeclared-constituent problem
In 2019, researchers including staff of the United States Pharmacopeia published a short study in the Journal of Pharmaceutical and Biomedical Analysis (Choules and colleagues; PMID 31671336) describing quantitative proton NMR analysis of custom synthetic peptides obtained commercially. Two of the materials contained undeclared mannitol — at roughly 20% and 43% by weight.
Mannitol is UV-transparent and poorly retained on a reversed-phase column. A chromatographic purity assay with ultraviolet detection cannot see it. Both materials could therefore carry a truthful high-purity certificate from any laboratory, in-house or independent, that ran the standard method — because area-percent purity is a statement about the peaks a detector produced, not about the contents of the vial.
This is the argument for orthogonal methods rather than for outsourcing. A third-party laboratory running the same blind method reaches the same blind conclusion.
Why this is worth a research group’s attention
Freedman, Cockburn and Simcoe estimated in PLoS Biology in 2015 (13(6):e1002165; PMID 26057340) that irreproducible preclinical work exceeds 50% prevalence and accounts for roughly US$28 billion per year of United States preclinical spending. The following year Nature published Baker’s survey of 1,500 researchers (533:452–454; PMID 27225100), in which a majority reported failing to reproduce another group’s result. Reagent identity and characterization are recurrently named in this literature as a contributing factor, and a peptide whose actual content is 78% rather than the 98% assumed at the bench is a reagent-characterization error that propagates silently into every concentration in the study.
What our own certificates say, and what they do not
Applying the checklist above to ourselves, in public, is the only version of this article worth publishing.
Our BPC-157 10MG batch certificate carries report number DBAV-BPC-10-051226, issued by Testides Analytical and dated 23 May 2026, reporting 99.41% by HPLC-UV against a stated 10 mg fill. It names the laboratory, the report identifier, the date and the method, and it is linked from the product page as a downloadable PDF rather than summarized in marketing copy. Those are the properties that make a number checkable, and they are the reason we publish the document rather than the percentage.
Three limitations belong in the same paragraph. First, the certificate reports HPLC-UV; there is no mass-spectrometry section on it, so it establishes chromatographic purity and content mass, not independent identity confirmation. Second, we selected the laboratory and supplied the sample, which is true of every third-party certificate in this market and is not fixed by the certificate being third-party. Third, a UV-based purity assay carries the blind spot Choules and colleagues documented, and no reading of our certificate should be taken as excluding a UV-transparent constituent.
We would rather state those three things than let a 99.41% figure imply more than it supports. A supplier that will not tell you what its certificate cannot see is telling you something about its certificate.
The practical position
“Third-party tested” is a necessary claim and an insufficient one. A certificate earns trust through four disclosed properties — an identified laboratory with relevant accreditation scope, a named method with its corrections stated, a documented sampling basis, and a report identifier and date that permit the document to be checked — and the identity of the party running the assay is only the first of them. Eight national metrology institutes measuring one vial of oxytocin produced a 220 mg/g spread, and the outlier was the one that looked most like a marketing number.
All materials described here are supplied for research use only. They are laboratory chemicals intended for in vitro and preclinical research conducted by qualified personnel. They are not for human use, not for veterinary use, and not for diagnostic or therapeutic application. Nothing on this page is guidance for administration to people.
References
- CCQM-K115.b, Key Comparison Study on Peptide Purity — Synthetic Oxytocin, final report, Bureau International des Poids et Mesures and National Institute of Metrology (Beijing), 2020. Eight participating national metrology and designated institutes; reference value 787.2 ± 12.9 mg/g; peptide-related impurities 31.6 ± 1.4 mg/g.
- Li C, Bhavaraju S, Melanson J, et al. Multi-laboratory comparison of three peptide quantification approaches using oxytocin. Journal of Pharmaceutical and Biomedical Analysis, 2019; 166:105–112. PMID 30640042.
- Choules MP, et al. NMR detection of an undeclared constituent in custom synthetic peptides. Journal of Pharmaceutical and Biomedical Analysis, 2020. PMID 31671336.
- Freedman LP, Cockburn IM, Simcoe TS. The economics of reproducibility in preclinical research. PLoS Biology, 2015; 13(6):e1002165. PMID 26057340.
- Baker M. 1,500 scientists lift the lid on reproducibility. Nature, 2016; 533:452–454. PMID 27225100.
- ISO/IEC 17025:2017, General requirements for the competence of testing and calibration laboratories.