How to Read a Certificate of Analysis

A COA is a document. It is not a guarantee, a license, a grade, or a seal. It is a record of what one laboratory says it observed when it ran specified tests on a specified sample on a specified date.

By Makay Groves · Paralegal and law student · Written for people evaluating laboratory research material. Nothing here concerns what any compound does, or any use in humans.

Everything useful about a COA follows from taking that opening sentence literally — and everything misleading about a COA follows from readers who don't.

1. What a COA is, and who signed it

The single most important field on a COA is the one most buyers skip: who generated the data.

There are two fundamentally different documents that both get called "COA":

Manufacturer- or vendor-issued

The party selling the material tested it, or received test data from its own supplier, and reproduced the results on its own letterhead. This is self-reported data. That characterization is not an accusation — it is a description of the chain of custody. The party with a commercial interest in the result controlled the sample, the instrument, the analyst, and the document. There is no independent step anywhere in that chain.

Independent third-party report

A contract laboratory with no financial stake in the outcome received a sample, ran it under its own methods, and issued a report under its own name and report number.

Regulated pharmaceutical manufacturing treats this distinction as load-bearing. Under U.S. drug GMP regulations, a manufacturer may accept a supplier's certificate of analysis in place of full testing only if it performs at least one specific identity test itself and establishes the reliability of the supplier's analyses through appropriate validation at appropriate intervals (21 CFR 211.84(d)(2)). The regulatory system does not accept a supplier COA as sufficient on its own. Neither should a buyer.

2. Identity testing — what mass spectrometry establishes

Identity testing answers one question: is this the molecule the label says it is?

The standard tools are mass spectrometry — electrospray ionization (ESI-MS, often coupled to liquid chromatography as LC-MS) or matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF). What a COA usually reports is a theoretical (calculated) molecular weight and an observed molecular weight, with agreement between them treated as confirmation.

What that actually establishes

A close match between observed and theoretical mass establishes that the sample contains material of the expected molecular mass. That is real information. It rules out a great many gross substitutions and it rules out material that is simply something else entirely.

What it does not establish

Molecular weight is a scalar. Many distinct molecules share one.

• Isobaric residues. Leucine and isoleucine have identical elemental composition and therefore identical mass. Mass alone cannot distinguish them.
• Stereochemistry. A D-amino acid substituted for its L-counterpart is the same mass. Racemization is invisible to a mass measurement.
• Isomerization. Aspartic acid and iso-aspartic acid are isobaric, and the mass-spectrometry literature notes they are indistinguishable even in MS/MS under common fragmentation methods.
• Sequence order. Any rearrangement of the same residues yields the same total mass. Intact mass says nothing about sequence.
• Everything that isn't the analyte. Salts, water, solvents, and other poorly ionizing material do not appear as the reported mass.

Mass spectrometry is also not a reliable quantitative purity method on its own, because different species ionize with different efficiencies and can suppress one another. A COA that reports MS and calls the result "purity" is using the word loosely.

If a COA reports only a number — "MW confirmed" — without the spectrum, the observed mass, the charge states, and the mass error, it is asking to be taken on faith.

3. Purity — reading an HPLC number correctly

Peptide purity is conventionally determined by reversed-phase high-performance liquid chromatography (RP-HPLC) with UV detection, typically in the 210–220 nm range. That range is chosen because the peptide bond itself absorbs there.

Purity is then reported as area percent: the integrated area of the main peak divided by the total integrated area of all peaks in the chromatogram.

Read that definition again, because the entire section follows from it.

Area percent of what the detector sees, at one wavelength

An HPLC purity figure is a ratio of UV absorbance signal at a chosen wavelength. It is not a mass fraction. It cannot be a mass fraction. Anything present in the vial that does not absorb UV at that wavelength contributes nothing to either the numerator or the denominator — it is not counted as an impurity because it is not counted at all.

Manufacturer technical literature states this plainly: purity is quantified as the area of the main peak relative to the total area of all peaks, and water and residual salts are not detected UV-spectrophotometrically.

Things a 214 nm RP-HPLC purity figure is structurally blind to:

• Water
• Inorganic salts
• Counterions (trifluoroacetate, acetate)
• Residual organic solvents that do not absorb at that wavelength
• Bacterial endotoxin
• Particulates
• Anything that co-elutes underneath the main peak

That last item deserves emphasis. Chromatographic purity assessment can produce false negatives when an impurity co-elutes with the main compound and has a similar UV spectrum; the chromatography literature treats this as a recognized limitation, not an edge case. For peptides specifically, deletion sequences — the target minus one residue — can be chemically similar enough to elute at or near the same retention time.

Area percent also assumes that the analyte and its impurities have comparable UV response at the detection wavelength. They frequently do not. An impurity with weaker absorbance is systematically under-counted; one with stronger absorbance is over-counted.

Why "99% purity" is a narrower claim than it reads

"99% purity by HPLC at 214 nm" translates to: of the UV-absorbing material that eluted from this column and was integrated by this software, 99% of the absorbance area was in the main peak.

It does not mean the vial is 99% peptide by mass. It does not mean 1% impurity. It does not speak to what fraction of the powder is water or salt. A lot can honestly report very high HPLC purity while a substantial share of the vial's mass is not peptide at all.

4. What is usually missing, and why it matters

Most research-grade peptide COAs report two things — an MS identity check and an HPLC purity number — and stop. Here is what is typically absent.

Net peptide content. The fraction of total sample mass that is actually peptide, as distinct from purity. Contract manufacturer documentation describes net peptide content as commonly falling in the range of roughly 60–90% of total weight, with the balance made up of water, absorbed solvents, counterions, and salts. It is determined by amino acid analysis, elemental analysis, or UV spectrophotometry — not by HPLC. Purity and net peptide content are not equivalent, and one cannot be inferred from the other.

Counterion content. Peptides purified by RP-HPLC using trifluoroacetic acid are typically isolated as the TFA salt. The counterion is part of the mass on the balance. Ion chromatography quantifies it. Almost no research-grade COA reports it.

Water content by Karl Fischer titration. Lyophilized peptide salts are hygroscopic. Water is invisible to UV detection and can be a meaningful share of vial mass. Its absence from the COA means the mass fraction of the vial is unquantified.

Residual solvents. Synthesis and purification use organic solvents. The pharmaceutical framework is ICH Q3C — currently at revision R9, approved in 2024 — which classifies solvents by toxicity, and USP General Chapter <467>, which specifies procedures. A generic "<0.5%" with no named solvents is not a residual solvent result.

Elemental impurities. USP General Chapters <232> (limits) and <233> (procedures), aligned with ICH Q3D, cover heavy metals and other elemental contaminants. Catalysts, reagents, and equipment are plausible sources. Rarely tested for research material.

Bacterial endotoxin. Measured by Limulus amebocyte lysate assay or recombinant Factor C alternatives, under USP <85>. Endotoxin is a bacterial cell-wall product; it is not detectable by HPLC, MS, or visual inspection.

Sterility. A distinct test from endotoxin — a sample can be sterile and still contain endotoxin from bacteria that died earlier. Research-grade material is generally not sterility tested.

None of these are exotic. They are routine in regulated manufacturing. Their absence from a research COA is normal, and it is precisely why "research grade" and "pharmaceutical grade" are not synonyms.

5. Red flags in the document itself

Before evaluating the data, evaluate the document as a document.

• No lot or batch number. A COA without a lot number is untraceable to any specific material. It is not a certificate of analysis; it is a brochure.
• Lot number on the COA does not match the vial. The most common failure mode and the easiest to check.
• No test date. Undated results cannot be placed in time relative to synthesis, storage, or shipping.
• No named laboratory, analyst, or signatory. Accountability requires a name. Anonymous data has no author to question.
• No instrument or method identified. A credible report names the technique, the column, the mobile phase, the detection wavelength, and the instrument. "HPLC: 99.2%" with no method is an assertion, not a measurement.
• No chromatogram or spectrum — only a summary number. The raw trace is the evidence; the number is the conclusion. Omitting the trace also conceals baseline quality, integration choices, peak shape, and whether anything was cropped out of the run.
• Image-only PDFs. A flattened scan with no text layer and no metadata is trivially editable and effectively unverifiable. Document properties — creation date, producer application, author — sometimes tell you more than the body text.
• Identical documents across lots. If two lots produce chromatograms with the same retention times to three decimals and the same purity figure, either the document was reused or the numbers are not measurements.
• Implausibly clean results. Zero mass error, purity above 99.9% on every lot, perfectly round numbers across an entire catalog. Real analytical data has variance.
• A laboratory that cannot be found. No website, no address, no phone, no accreditation listing — the report has no verifiable author.

6. How to verify a COA independently

Verification is a sequence, and most people stop at step zero.

1. Match the lot. Compare the lot or batch number on the COA against the number physically printed on the vial and its packaging. If they don't match, stop.
2. Confirm the laboratory exists. A real contract analytical laboratory has a physical address, a working phone number, a scope of services, and staff. Confirm the lab actually performs peptide analysis — plenty of real labs do environmental or food testing and have never run a peptide.
3. Check accreditation, and check its scope. The relevant standard is ISO/IEC 17025:2017. Accreditation bodies maintain public directories. Critically, accreditation is granted for a defined scope — specific methods on specific matrices. A lab accredited for something unrelated is not accredited for the test on your COA. Look up the scope document, not just the logo.
4. Contact the lab with the report number. This is the step that separates a real document from a fabricated one.
5. Request the raw chromatogram and full spectrum from the seller, ideally in a form that includes acquisition metadata rather than a cropped image.

What a legitimate laboratory will and will not tell you

Expect a narrow answer, and do not read the narrowness as evasion.

A legitimate lab will typically confirm whether a given report number was issued by that laboratory, whether the document format is theirs, and whether they perform the method named. That alone catches fabricated reports and invented lab names.

A legitimate lab will not hand over a client's analytical results to a stranger. ISO/IEC 17025 imposes confidentiality obligations regarding customer information; results belong to the client who commissioned the testing. The lab's customer is the seller, not you.

So the realistic outcome of a verification call is: yes, that report number is ours, or no, we have no record of that. Both are decisive. If the lab has no record, you have your answer. If the lab does not exist, you had your answer before you called.

7. What a clean COA does not tell you

Assume everything above checks out — real lab, real accreditation, real chromatogram, matching lot, confirmed report number. Here is the honest boundary of what you now know.

You know something about one sample. A COA describes the aliquot that was tested. Lyophilization is not perfectly homogeneous, and vials are filled sequentially. The tested sample is evidence about the lot; it is not a measurement of every vial in it.

You know something about one moment. The test date is a timestamp. Peptides are chemically dynamic — oxidation, hydrolysis, deamidation, and aggregation are time- and condition-dependent. A COA dated eighteen months ago describes material as it was eighteen months ago.

You know nothing about the intervening custody. Storage temperature, freeze-thaw cycles, humidity exposure, courier delays, time on a loading dock — none of it appears on any COA, and all of it happens after the test.

You know nothing about anything the tests didn't cover. A COA is silent on every parameter it does not list. Silence is not a passing result. It is silence.

Sources

• 21 CFR 211.84(d)(2) — supplier COA reliance and identity testing
• Bachem — Quality Control of Amino Acids & Peptides
• AmbioPharm — What is net peptide content?
• Chromatography Online — Peak purity: concepts and limitations
• ICH Q3C(R9) — Residual Solvents (2024)
• ISO/IEC 17025:2017
• USP <232> — Elemental Impurities: Limits

What I could not verify

Typical water content ranges for lyophilized peptides could not be corroborated against a manufacturer or compendial source, so no number is given. I found no published enforcement action or investigation documenting fabricated research-peptide COAs; the red-flag section above is therefore framed as document-integrity reasoning — what a document must contain to be verifiable — and not as a description of a documented fraud pattern. Whether an individual contract lab will confirm a report number to a non-client is a matter of that lab's policy, not a requirement of any standard. No pass/fail purity threshold, mass-error tolerance, or endotoxin limit is stated here, because compendial limits are product- and route-specific and do not transfer to uncharacterized research material.