Claims on this page checked 2026-08-21

Certificate of analysis meaning: what does a COA prove?

A certificate of analysis is a report of tests on a batch. For a peptide vial, it can show which tests a sample passed. The batch link tells you which goods the report covers. A test of one vial does not prove the contents of a different, untested vial.

A COA can be useful. Read the test list as well as the result. Purity, endotoxin and sterility are distinct checks in the EMA peptide guideline, EMA/CHMP/CVMP/QWP/367182/2025. A claim of purity does not show that all those tests were done. This source was checked on 21 August 2026.

For tests of vial contents, see what is actually in the vial. For findings on germs and endotoxin, see sterility and infection. For symptoms and phone numbers, see when to get help.

What does a certificate of analysis mean?

A certificate of analysis is a test report, not the product itself. It is not a drug regulator's full set of tests and limits. A useful report names the batch. It also names the test method, result, limit and lab. Those fields let you check what a purity claim means.

A single large purity number gives less detail. It does not state which other tests were run. The EMA source below shows why the test list matters.

Which tests does the EMA peptide guideline list?

The EMA guideline lists the checks for the active substance in a synthetic peptide drug. Its list is not exhaustive. The source is an agency guideline, not a funded trial.

The guideline took effect on 1 June 2026 (EMA/CHMP/CVMP/QWP/367182/2025; checked 21 August 2026).

Its specification list has twelve entries (EMA/CHMP/CVMP/QWP/367182/2025; checked 21 August 2026).

Source for each entry: EMA/CHMP/CVMP/QWP/367182/2025, checked 21 August 2026.

The EMA says large-molecule impurities must be specified so their identities are known. A vague entry for that group is not enough. The same guideline gives the European Pharmacopoeia thresholds below.

These limits are not a test of total purity in a finished vial. They give a scale for the named impurities. They do not make unlike test methods the same.

What happened when 98 certificates were checked independently?

Verbeke and colleagues tested synthetic quorum-sensing peptides made for research. The peer-reviewed paper appeared in 2015 (PMID 29403929; checked 21 August 2026).

IWT-Vlaanderen helped fund the work. The Ghent University Special Research Fund also helped fund it. The source gives the grants in the table below.

The source is Verbeke et al., Journal of Pharmaceutical Analysis, DOI 10.1016/j.jpha.2014.12.002.

These were research peptides, not finished vials for injection. The suppliers and the research lab used different chromatography methods. The paper directly compares COAs with lab checks. It does not give a match rate for COAs on grey-market injectable vials.

What did the online vial study measure?

The test-purchase team ordered semaglutide from six illegal online pharmacies (PMID 39509151; checked 21 August 2026). Three orders arrived (PMID 39509151; checked 21 August 2026). The other three were non-delivery scams (PMID 39509151; checked 21 August 2026).

The Hungarian Scientific Research Fund supported the work through grant 143684 (PMID 39509151; checked 21 August 2026).

The full report appeared in Journal of Medical Internet Research in 2024, DOI 10.2196/65440. The same purchase study also appeared in JAMA Network Open in 2024, PMID 39093567, DOI 10.1001/jamanetworkopen.2024.28280. Both sources were checked on 21 August 2026. These papers report one purchase study, not two independent datasets.

The chromatograms showed only semaglutide in the delivered vials (PMID 39509151). Low purity did not mean that the team found insulin or a different peptide. Endotoxin was found in all three vials (PMID 39509151; checked 21 August 2026). In one vial it reached 8.95 EU/mg, a level the authors called elevated (PMID 39509151; checked 21 August 2026).

The team found no viable microorganisms in the freeze-dried samples when they tested them (PMID 39509151; checked 21 August 2026). The separate results show why purity, endotoxin and sterility need their own checks. This was the only peer-reviewed semaglutide test-purchase study found for this page as of 21 August 2026.

What does the larger submitted-sample dataset add?

Mendias and Awan studied reports sent to Finnrick Analytics. Consumers and peptide-making firms sent the samples. The preprint reports no outside funding. Its authors report no conflicts of interest.

The exact title is Evaluation of Research Grade Peptides Marketed Directly to Consumers Reveals Extensive Variability in Purity and Measured Abundance. The source is DOI 10.20944/preprints202604.1748.v1, checked 5 September 2026.

The middle of the tested group was close to the label on both amount and purity. The model failure rates below measure a different question. A sample had to meet both the amount and purity limits to pass a model.

The failure rates came from limits applied to measured amount and purity. They are not rates of samples with the wrong peptide. The same source reports both the medians and the failures.

Did purity predict endotoxin?

The regression used 36 samples with quantifiable endotoxin (DOI 10.20944/preprints202604.1748.v1; checked 2026-09-05). The authors reported no correlation between purity and endotoxin in that group. The model returned R2 < 0.01 (DOI 10.20944/preprints202604.1748.v1; checked 5 September 2026). The reported P value was 0.75 (DOI 10.20944/preprints202604.1748.v1; checked 5 September 2026).

A purity result did not predict the endotoxin result in samples with both tests. This does not give an endotoxin rate for the whole market.

What limits does the submitted-sample dataset have?

The preprint has not been peer-reviewed. The sample was not random. The authors say firms that trust their product quality may be more likely to send samples. Poor batches may be less likely to appear in the data.

The tested panel covered purity, amount and endotoxin. It did not cover sterility. It did not test residual solvents, particles, stability or structural integrity. These limits come from the same preprint, checked 5 September 2026.

The authors call the failure rates a lower bound on the true quality deficit. That is their stated limit, not a market-wide rate proved by this sample. The sample's self-selection also stays with the findings.

What did older forensic testing find?

Janvier and colleagues studied falsified peptide drugs from the Belgian market. Their full paper names no funder. The authors state that they had no conflict of interest.

The source is Janvier et al., Talanta 2018;188:795–807, PMID 30029448, DOI 10.1016/j.talanta.2018.06.023.

All arsenic found was in the more toxic inorganic form. Each preparation had at least one trace residual solvent. All calculated maximum solvent levels were below the ICH limits. These findings are from the same paper, checked 26 August 2026.

The study has a small sample. It predates the current counterfeit GLP-1 reports. It is not a test of every research-peptide vial.

What remains unclear?

These studies do not show that a particular COA is false. Nor do they show that every COA leaves out a test. They show what was checked in the samples each team studied.

There is a direct COA-versus-QC study of 98 research peptides (PMID 29403929; checked 21 August 2026).

A targeted Europe PMC search on 21 August 2026 found no title or abstract about an equivalent semaglutide comparison. A second search found no title or abstract combining certificate of analysis, vial and independent testing. These were bounded searches, not a systematic review. They do not rule out a study outside those search results.

A separate editorial reported on 24 patches sold as natural GLP-1 (PMID 41480958; checked 21 August 2026). It also covered one gel (PMID 41480958; checked 21 August 2026). The authors found no posted COA for those products. That is a finding about absent certificates, not a lab check of their vial contents.

The source is White, Tai and Nuzi, Annals of Pharmacotherapy, DOI 10.1177/10600280251407145. Europe PMC labels it an editorial. Only the abstract was open. That abstract gave no funding details. These source limits were checked on 21 August 2026.

A COA can help you see what a lab tested. Its reach depends on the tests, methods and batch link it states. For the full set of vial and test questions, return to the evidence hub.