HPLC qualification and system suitability: what each one actually proves

The most common mistake in chromatography qualification is a logical one rather than a technical one: treating a passing system suitability test as proof that the instrument is qualified, or treating a completed performance qualification as proof that today's run is valid. They are tests of different things, and a system can pass one while failing the other without any contradiction.

A system suitability test asks whether this method, on this system, on this day, with this column and this mobile phase and this standard, is performing well enough for the result to be reported. It is a method-level control, defined in the method or the monograph. Instrument qualification asks whether the pump delivers the flow it claims, whether the injector injects the volume it claims and whether the detector reads the wavelength it claims — independent of any method at all.

What has to be measured, module by module

An HPLC is not one instrument; it is five or six instruments plumbed together, and a modular qualification measures each one against something traceable. The acceptance limits are yours — they come from the manufacturer's specification, from what your methods actually need, or from both — but the list of things worth measuring is fairly settled:

The detector row is worth dwelling on. A UV or diode-array detector is a spectrophotometer, and it is subject to the same wavelength and absorbance questions as a bench instrument — which is why a wavelength check belongs in an HPLC qualification and not only in the UV-Vis procedure.

USP <1058> and scaling the work

USP <1058> sets out analytical instrument qualification as the base of a data quality framework and describes a lifecycle of design, installation, operational and performance qualification. Its central device is a three-group classification — Group A, Group B and Group C — which scales how much qualification an item needs to how much the result depends on the instrument's own internal behaviour.

The point people miss is that the grouping follows intended use, not the model number. The same instrument can be classified differently in two laboratories, and a magnetic stirrer and a chromatography system are not expected to be treated alike. That is also the answer to the complaint that qualification is disproportionate: if it is disproportionate, the grouping was done without thinking about use.

USP <1058> does not tell you what your acceptance limits are. It tells you to have them, to have justified them, and to hold the records. The numbers come from the manufacturer's specification and from what your methods require, and the second of those is the one that gets left out.

What <621> and 2.2.46 actually ask for

USP <621> gives general system suitability provisions and the rules for adjusting a compendial method without revalidating it; Ph. Eur. 2.2.46 is the European counterpart, revised to carry the harmonised text agreed by the Pharmacopoeial Discussion Group in September 2021. In both, criteria stated in an individual monograph take priority over the general chapter — a point worth checking before copying a general figure into a method.

One concrete requirement is worth knowing because it is routinely got wrong: USP <621> specifies that data from five replicate injections are used to calculate the relative standard deviation where the requirement is 2.0 % or less, and six replicate injections where the requirement is more than 2.0 %. The number of injections is part of the test, not a matter of local habit. The permitted-adjustment provisions are similarly specific, and are narrower for gradient methods than for isocratic ones, because a gradient is more sensitive to a change in conditions.

Dwell volume: the failure mode that neither activity covers

A gradient method developed on one instrument and transferred to another frequently fails on arrival, with shifted retention times, lost resolution between an early pair of peaks, or a peak that has moved into the injection disturbance. Both instruments are qualified. Both pass their own system suitability on other methods. Nothing is broken.

The cause is usually the gradient delay volume — the volume between the point where the gradient is formed and the head of the column, which includes the mixer, the connecting tubing and the injector path. It differs between instrument designs, between low-pressure and high-pressure mixing, and between two instruments of the same model fitted with different mixers. The gradient arrives at the column later on the system with the larger delay volume, so every peak eluting during the early part of the gradient moves.

• Measure the delay volume rather than inferring it — a step gradient with a tracer and a known flow gives it directly.
• Record it as an instrument attribute in the qualification record, so a transfer starts from numbers instead of a guess.
• Where a method has to run on dissimilar systems, an initial isocratic hold long enough to absorb the difference is the usual fix, and it is a method change that has to be made through the permitted-adjustment provisions or through revalidation.
• Extra-column volume affects isocratic methods too, through peak dispersion rather than retention shift — a narrow early peak loses plates in long or wide tubing.

Records, and the part auditors look at

Qualification records are expected to show what was measured, with what, against which limit, and by whom — the same evidence chain as any calibration. Three specifics come up repeatedly: the reference instruments used for flow, temperature and wavelength need their own current calibration certificates, traceable and with stated uncertainty; the chromatographic data system is a computerised system in its own right, with audit trail, access control and the usual data integrity expectations; and a qualification performed by a service engineer is still your record, so the acceptance limits in it have to be ones you have reviewed rather than ones that arrived on the vendor's form.

ICH Q2(R2) and ICH Q14, both adopted at Step 4 on 1 November 2023, sit above all of this: they govern how the analytical procedure itself is developed and validated. A qualified instrument running an unvalidated procedure produces traceable numbers about nothing in particular.

ValiTrac AI is not an accredited calibration laboratory and confers no compliance. The acceptance criteria for any of the measurements above belong to your own specification, your own methods and the pharmacopoeial chapters you work to.