Reading a certificate of analysis, line by line

What a certificate of analysis is

A certificate of analysis is a laboratory’s record of what it did to a sample and what it found. That is the whole of it. It is a report about a specific quantity of material that a specific client sent to a specific laboratory on a specific date, and everything that makes it useful or useless follows from how tightly those three specifics are pinned down.

In licensed manufacture a certificate sits inside a system — the sample is drawn under procedure from a batch that is identified, sealed, tracked and released by a named person who carries legal responsibility. Outside that system the same document format persists, with the system removed. This is why members find these documents so disorientating: they look exactly like the regulated version, and the parts that made the regulated version meaningful are the parts that are silently absent.

What follows is how our quality circle reads one. It needs no chemistry. Most of the faults people find are in the fields a bored person skips.

The header: who tested what, and for whom

Start at the top and read it as a form. Which laboratory produced this, and is the laboratory’s own identity stated in a way that could be checked — name, location, a report reference? Is there an accreditation claim, and if so, to what standard and for which methods? Accreditation is granted within a defined scope; a laboratory accredited for one technique is not thereby accredited for everything on its price list.

Then the client field, which people skip and should not. The certificate was commissioned by somebody. If the client is the supplier, that is ordinary and worth knowing. If the client is a third party you have never heard of, then this document was produced for somebody else about somebody else’s sample, and its appearance on a product page is a marketing decision. If the client field has been removed, ask why.

Then the two dates: sample received, and report issued. A gap of months, or a report date that precedes the stated manufacturing date, is not always sinister but always needs explaining.

Sample identity: the line doing most of the work

Everything hinges on the sample identifier. You are trying to establish a chain from the vial in your hand back to the material that was tested, and the identifier is the only link in that chain that appears on the page.

So: is there a lot or batch number, and does it match the lot you have been quoted? Does the identifier appear anywhere else — on the vial label, on the invoice, in the supplier’s written answer to your batch question? A certificate whose identifier appears nowhere except on the certificate has established nothing. Members are often told, when they push, that the tested sample came from a reference batch rather than from the shipping lot. That is a legitimate practice and a completely different claim, and it should be stated on the document rather than extracted by interrogation.

Also read what the sample was described as when it arrived at the laboratory. Analytical laboratories generally test what they are sent and report it as declared by the client. The document usually says so, in small type, and that sentence is the laboratory correctly declining to vouch for provenance.

The method block

The method section is where a real report distinguishes itself from a designed one. You are not assessing whether the method was well chosen; you are checking that it is described at all, in enough detail that another laboratory could repeat it.

For purity by chromatography, expect the technique, the column, the mobile phases, the gradient, the flow rate, the detection wavelength, and the injection volume. For identity by mass spectrometry, expect the instrument mode, the observed mass, the theoretical mass, and the tolerance. For anything reported as content, expect the technique named explicitly — amino acid analysis and nitrogen determination are the usual ones, and a content figure with no stated method is not a content figure.

A certificate carrying a single purity percentage with no method block is a claim on headed paper. Members sometimes defend these on the grounds that the number is probably fine. It might be. You simply have no way to tell, which is the situation the document was supposed to remove.

The results block, and the chromatogram

Read the numbers as a set rather than picking the headline. If a purity percentage is given, is the basis stated — area per cent at the stated wavelength, normalised? Is there an impurity table listing peaks by retention time and area, or only a single figure? A report that lists what else was seen is telling you more, not less, and members occasionally read that honesty as a defect, which is backwards.

If a chromatogram is included, you can check things without interpreting chemistry. Does the trace have axis labels and a scale? Does the main peak’s retention time match the number written in the table? Is there a solvent front and a settled baseline, or has the image been cropped so tightly that everything either side of the main peak is out of frame? Cropping is the most common cosmetic intervention we see.

For mass spectrometry, compare the observed mass with the theoretical mass yourself. They should agree within the tolerance stated. If no theoretical mass is given, the comparison you are being invited to make has been quietly removed.

What is not on the page

Then do the most important pass, which is a pass for absence. Nearly every certificate in this market covers identity and purity and nothing else, and the omissions are not disclosed because nobody claimed them in the first place.

Not present: sterility. Not present: bacterial endotoxin. Not present: residual solvents. Not present: elemental impurities. Not present: water content. Not present: net peptide content, unless explicitly stated with a method. Not present: fill weight per vial. Not present: any stability data, so nothing on the page tells you how long the material remains what it was.

What purity actually means goes through why each of those needs its own assay. The practical instruction here is simply to say the absent list out loud when you finish reading, because a well-produced document creates a strong impression of coverage, and the impression is the thing that has to be actively resisted.

When it does not check out

Sometimes a check fails: the identifier does not match, the dates are impossible, the client is somebody else, the crop is obvious. Members ask what to do next, and the answer is duller than they hope.

Ask the supplier once, in writing, specifically. Keep the reply. If the reply is a non-answer, you have your result — not proof of fraud, but proof that lot-level accountability is not available to you, which is the thing you were trying to establish. Some laboratories will confirm whether a report reference exists, and members have occasionally had that confirmation refused on client-confidentiality grounds, which is a legitimate answer and not an evasion.

What we ask members not to do is post the document publicly with an accusation attached. Our quality-and-testing circle reads them together instead, without naming a conclusion the document cannot support, and the recorded picture per supplier lives on the reference pages at vendors. Being wrong in public about a company is both unkind and useless.