How labs get peptide results wrong

If a report is genuine, can it still be wrong?

Yes. Lessons 14 and 15 deal with documents that were edited or invented. This lesson covers a real report, from a real laboratory, carrying a number that does not describe the vial. It names no particular lab; it lists where analysis is known to be fragile.

The reference point is ICH Q2(R2), which sets out what a method must be shown to do before its results are trusted: specificity, range, accuracy and precision among them, with robustness evaluated during development (ICH Q2(R2)). Most headings below are one of those properties failing in practice.

What goes wrong when a method is not validated for the peptide?

An HPLC method is a recipe: column, solvents, gradient, temperature, detection wavelength. A recipe tuned for one peptide may not separate the impurities of another. ICH Q2(R2) says specificity for a separation method can be shown "by the resolution of the two components which elute closest to each other", and that where impurities are not available it can be shown by comparing results with "an orthogonal procedure". A lab that runs every peptide through one generic gradient, without checking that the critical pair separates, has not shown specificity for that peptide.

Validation also covers how the method behaves day to day. Q2(R2) defines intermediate precision as variation within one lab across "different days, different environmental conditions, different analysts and different equipment", and robustness as the method's "capacity to meet the expected performance criteria during normal use". A system suitability test, run before samples, is meant to catch a column or instrument that has drifted. A report that names no method, no column and no system suitability result gives you no way to tell whether any of this was done.

How can an impurity hide under the main peak?

Lesson 3 showed purity as the main peak's share of all detected peak area. That calculation assumes the main peak is one compound. If an impurity elutes at the same time, it is counted as product and purity looks higher than it is. This is co-elution.

Peptide impurities are often close relatives of the peptide itself: a sequence missing one residue, an oxidised methionine, a residue in the wrong mirror-image form. These can behave almost identically on a column. The defences are a method shown to resolve the closest pair, a second method run under different conditions, or a mass spectrometer reading the peak, since an impurity with a different mass shows up even when it co-elutes. The LC-MS method page covers that last approach. A purity figure from a single UV trace with none of these is only as good as the separation behind it; the chromatogram tool shows how a merged shoulder moves the percentage.

How do integration choices change the purity number?

Software finds peaks, but people set the rules: where the baseline runs, how small a peak must be before it is ignored, where to drop a line between two touching peaks. Lesson 6 introduced these choices. They are legitimate, and they move the answer.

Neither row is fabricated. Regulators deal with this by requiring a record. The US FDA's data integrity guidance says the audit trail for an HPLC run "should include the user name, date/time of the run, the integration parameters used, and details of a reprocessing, if any", with a justification for any reprocessing (FDA, Data Integrity and Compliance With Drug CGMP). A research-market report rarely shows integration settings, which is why a printed chromatogram is valuable: you can at least see whether there was a shoulder or a small peak that the number seems to ignore.

What can go wrong with a reference standard?

Amount results, covered in lesson 4, depend on comparing the sample with a reference standard of known content. If the standard is wrong, every amount measured against it is wrong by the same proportion, and nothing in the sample's own chromatogram will reveal it.

• The standard's content is assumed, not assigned. A vial of reference material labelled 10 mg holds 10 mg of powder, not 10 mg of peptide. USP scientists describe a two-step process that first assigns a value to bulk material by mass balance, then uses that bulk to assign a value to each vialed standard (McCarthy et al., Pharmaceutical Research, 2023).
• The standard has taken up water. The EMA warns that peptides "are often very hygroscopic powders" and that moisture content of a reference standard should be confirmed before use where appropriate (EMA, Guideline on the Development and Manufacture of Synthetic Peptides). A standard that gained water makes samples read high. Lesson 10 covers why.
• The standard is a different salt form. A TFA salt and an acetate salt of the same peptide have different peptide shares per milligram, as lesson 5 showed.
• The standard drifts between batches. The EMA asks manufacturers to describe how they will "prevent drift in peptide content" when qualifying future reference standard batches.

A USP multi-laboratory study compared HPLC assay, quantitative NMR and amino acid analysis for oxytocin. The HPLC assay "using the same peptide bulk material as the standard showed the lowest inter-lab variability", and that variability was calculated without the uncertainty of the standard's own purity assignment (Li et al., J Pharm Biomed Anal, 2019, via NIST). Labs agree best when they share a standard, so agreement between labs cannot reveal an error in the standard itself.

How common are plain transcription errors?

Common enough that regulators write rules about them. The FDA guidance says it is "not acceptable to record data on pieces of paper that will be discarded after the data are transcribed", and says data should be attributable, legible, contemporaneously recorded, original or a true copy, and accurate (ALCOA).

• A sample code mistyped, so one vial's result is attached to another vial.
• A report template reused with the previous sample's batch code or date left in place.
• Units confused, such as mg and micrograms, or a decimal point moved.
• Expected and observed mass swapped, or the wrong expected mass typed for the compound.
• A purity figure rounded up in a summary but not in the attached data.

Most of these leave a trace if you read carefully: a date that falls before the sample was received, an expected mass that does not match the compound's published value, an amount that is impossible for the vial size. For semaglutide, for example, the compound page lists the expected masses so you can check that the right one was used.

What about repeating a test until it passes?

The FDA calls "testing different samples until the desired passing result is obtained" testing into compliance, and says the practice is not consistent with CGMP. In a research market the pressure comes from the client: someone who receives a poor result may ask for a retest. A retest after a documented fault is sometimes correct, but a third attempt with the first two never mentioned is not the same evidence as a first result. Lesson 12 explains why coding samples makes this harder to do.

How can I tell a solid result from a fragile one?

The lesson on what HPLC measures covers what the purity figure measures. The COA checker turns the table above into questions to work through with a report in hand.

Check your understanding