What system suitability numbers should I insist on seeing before I believe a purity or content figure?

I have started asking labs for "the system suitability data" and getting back either silence, a shrug, or a screenshot of six retention times. I clearly do not know what I am asking for well enough to know when I have received it.

What is the actual list of parameters, what are the conventional acceptance criteria, and — most importantly — what does each one protect against? I would rather understand why a tailing factor limit exists than memorise that it should be under 2.0, because then I can tell when a borderline value matters and when it does not.

Also, one specific thing I have been told twice and want to check: I have heard that theoretical plate counts are not really meaningful for gradient peptide methods, which surprised me since plates are on every list of suitability parameters I have seen. Is that right?

Here is the list with criteria and, for each, what it is actually protecting. And yes, the plate-count point is right, for a specific and interesting reason.

The plate-count point

You have been told correctly. The plate-count formulae — N = 16 (tR / W)^2 using tangent widths, or N = 5.54 (tR / W_half)^2 using width at half height — are derived from plate theory for isocratic elution, where a band spreads freely throughout its passage down the column and peak width grows with retention time.

Under a gradient that assumption fails, because the gradient compresses the band. A peptide sits near the head of the column while the organic fraction is too low to move it, then elutes in a narrow window once the eluent reaches its critical strength. Gradient peaks are therefore far narrower than isocratic peaks at the same retention time, and feeding their widths into the isocratic formula gives physically meaningless plate counts — tens or hundreds of thousands on a column that could not deliver a quarter of that isocratically.

Worked, to show the magnitude. A peptide eluting at 18.42 minutes with a width at half height of 0.29 minutes:

N = 5.54 x (18.42 / 0.29)^2 = 5.54 x (63.52)^2 = 5.54 x 4035 = 22,350

That is a plausible-looking number for a 150 mm column packed with 2.6 micron core-shell particles, and it is a coincidence. Shorten the gradient and the same column will report 60,000; lengthen it and it will report 8,000. The number tracks the gradient, not the column.

So a plate-count criterion in a gradient peptide method is a system consistency check, not a measure of column quality: useful for confirming today's column behaves like last week's, useless across methods. The parameters carrying real information under gradient conditions are resolution and tailing, which is why peptide monographs specify those and treat plates as optional.

Resolution, worked, because it is the one that moves your purity number

Half-height resolution: R = 1.18 (t2 - t1) / (W_half,1 + W_half,2). For a main peak at 18.42 minutes and an impurity shoulder at 18.97, with half-height widths of 0.29 and 0.31:

R = 1.18 x 0.55 / (0.29 + 0.31) = 0.649 / 0.60 = 1.08

At R = 1.08 the peaks share a substantial valley, so the split between them depends on whether the analyst drops a perpendicular, skims valley-to-valley or uses a tangent skim. The smaller peak is the impurity, so the reported purity moves with that choice. That is precisely why the criterion is 1.5 and not 1.0, and why a run reporting R = 1.0 has already conceded that its purity figure is an integration decision.

What to ask for, concretely

One message: please send the system suitability summary for the run containing my sample — replicate standard injection area %RSD, resolution between the main peak and the nearest impurity, tailing factor, the blank, and the bracketing standard recovery. That is a specific, ordinary request and any laboratory running a validated method has it in the run record.

Interpreting the response is most of the value:

• Sends the numbers. The method is validated and the run was controlled. Read the values.
• Sends the chromatogram with the parameters annotated. Better still. PeptideMeter publishes method conditions alongside results, which lets you evaluate the separation rather than just the conclusion.
• Sends six retention times. They have the data acquisition but not a suitability protocol. Retention repeatability is real information and the absence of an area %RSD means nobody established the precision of the quantitation.
• Does not know what you mean. There is no validated method behind the number, and its uncertainty is unknown rather than large. That is a different and worse situation than a wide error bar.

edited 3 May 2026 by Dr_Ilse_Vandenberg — removed a claim I could not source

Two additions on the %RSD criterion, since it is the one people fixate on and it is easy to over-read.

First, %RSD measures precision only, and only of the injection-and-detection step. Six replicate injections of one standard solution from one vial tells you the autosampler is reproducible and the detector is stable. It tells you nothing about whether the solution was prepared correctly, whether the standard's assigned value is right, or whether the sample was recovered quantitatively from its vial. An 0.3% RSD sits perfectly happily on top of a 15% calibration error. Precision is cheap and accuracy is expensive, and suitability data speaks mostly to the cheap one.

The measurement that speaks to accuracy is a spike recovery: add a known mass of standard to a sample matrix, re-assay, and confirm you get back what you put in, conventionally 98 to 102%. That is a validation experiment rather than a per-run suitability check, so it lives in a method validation report, not a run record. Asking whether the method was validated for accuracy, and at what recovery, is the higher-value question and almost nobody asks it.

Second, a suspiciously good %RSD is a flag. Peak area RSD below about 0.1% on six real injections of a gradient peptide method is better than the pumps and the injector are usually capable of. It sometimes indicates that the "replicates" were re-integrations of one injection, or that the reported figure is a retention-time RSD mislabelled. Ordinary excellent performance is 0.2 to 0.6%.

Conversely, borderline values are worth reading in context rather than as pass or fail. An area RSD of 2.1% against a 2.0% criterion on a related-substances method is a technical failure and a trivial practical problem, since a 2% relative error on a 0.4% impurity is 0.008% absolute. The same 2.1% on an assay where the answer is the number you care about is a real limitation on the result. Criteria are tighter for assays than for impurity profiling for exactly that reason, and knowing which method you are looking at determines how much a marginal value should worry you.

Worth naming the one suitability element that is specific to peptides and does not appear on generic small-molecule lists: autosampler stability of the prepared solution.

Peptides in dilute aqueous solution in an autosampler tray do three things over a run that a small molecule does not:

• Adsorb to the vial and the needle. At the low microgram-per-millilitre concentrations used for impurity work, loss to glass and polypropylene surfaces is measurable over hours and is concentration-dependent, so it hits your dilute solutions harder than your concentrated ones and can bend a calibration curve.
• Degrade. Deamidation, oxidation and aspartimide chemistry all run at ambient temperature, slowly. A sample sitting at 25 degrees for fourteen hours while a long run works through the queue is not the same sample that was prepared.
• Aggregate. Acylated analogues with a fatty side chain are surface active and can form aggregates or adsorb at the air-liquid interface, which produces a slow drift in apparent concentration.

The controls are a chilled autosampler compartment, typically 4 to 8 degrees, a demonstrated solution stability window from method validation, and the bracketing standard at the end of the run — which is the check that catches all three at once if the drift is monotonic. A closing standard recovering at 96% against the opening standard is not a shrug; it is telling you the run drifted 4% and your sample was quantified somewhere in the middle of that.

So when you ask for suitability data, the bracketing standard recovery is the single most informative line for a peptide method, ahead of tailing factor and well ahead of plates. It is also the one most often omitted, because reporting it requires having run it.