When System Suitability Fails, the Batch Is Not Guilty

System suitability is the least dramatic check in the laboratory and the most consequential. It runs before the samples, on a standard solution, and it asks one question: is this chromatographic system capable of producing a defensible number today?

When the answer is no, nothing about the batch is known, because the batch has not been measured. What happens in the following hour is where laboratories reveal what their records are for.

What the four parameters guard

Parameter	What it protects	Typical failure cause
Resolution	The impurity is distinguishable from the main peak	Column ageing, wrong mobile phase pH
Tailing factor	Integration of the main peak is unambiguous	Dead volume, column contamination
Plate count	The separation is performing as designed	Column collapse, flow inaccuracy
Injection precision	The quantitative basis is stable	Autosampler carryover, bubble in the needle

Each parameter guards a different part of the number a certificate will eventually carry. Resolution protects the impurity figure, tailing protects the assay, plate count protects the method's identity, and injection precision protects everything that depends on a ratio. A run that skips the check is reporting into a system none of whose guards have been tested.

The order of blame

A failed suitability call starts a search, and the search has a natural order from cheap to expensive. Instrument function first: is the flow what it says, is the oven at temperature, does the detector lamp have hours left. Column second: age, back pressure, and whether a reference standard from a different method behaves normally. Mobile phase third: preparation date, pH, degassing, and whether somebody substituted a grade of solvent. Standard preparation fourth, because a mis-weighed standard moves every number at once. Technique fifth. The method last, and in practice the method is almost never the cause.

The order matters because each step is cheaper than the next, and because the record of the search is what makes the eventual result defensible. A record that shows the failure and the passing run, with no account of what was checked in between, describes luck rather than diagnosis.

The sequence is not negotiable

The rules are simple and they are binary. Suitability runs before samples. If it fails, no samples are injected until it passes. If it passes and then fails partway through the sequence, the results since the last passing check are in question.

The rule that gets broken is the second one, usually under time pressure, in the form of "inject the samples now, fix the suitability in parallel". The numbers produced that way may well be correct. They are also unattributable, because the system they came from had not been shown to work, and no amount of post-hoc suitability repairs the sequence. This is the same logic that governs an out-of-specification investigation, and it fails in the same way when it is bent.

Repeat-until-pass

A criterion that can be retried until it passes is not a criterion. Every attempt belongs in the record, including the failed ones, and the passing attempt counts only because the sequence leading to it is visible. A chromatography system that makes failures easy to delete is the single most common data integrity finding in the industry, and the audit trail is the mechanism that prevents it.

The test a buyer can apply without visiting the laboratory: ask whether failed suitability runs are retained in the system. A laboratory that says yes has a system. One that says the run was repeated has a habit.

What the customer should expect to see

Four items, when the question arises. The failed check, with its parameters. The corrective action, in the order of blame above. The passing check that followed. And confirmation that no sample results were reported from the failed interval.

The fourth item is the one that changes a recovered run into a trustworthy one. Without it, the customer is being asked to trust a sequence they cannot see, and the trust is doing work that documentation should be doing.

Suitability and the method's range

One subtlety connects suitability to validation. The criteria are chosen during method development or validation, and they are only meaningful across the range the method will be used in. A resolution criterion of 1.5 demonstrated at the main peak says little about the resolution between two impurities at 0.05 per cent, which is where the decision will actually be made.

A method whose suitability criteria only test the comfortable part of the chromatogram has been validated in the wrong place, and the validation summary is where that choice becomes visible.

Where this leaves a supply-chain partner

We treat suitability as a condition of reporting rather than an obstacle to it: the run waits, the search is recorded in the order of blame, and no result leaves the laboratory from a failed interval. Where a customer's method sets suitability criteria, we run theirs rather than ours, because the criteria are part of the method's definition.

One question maps the discipline in a sentence: when suitability failed last month, how many attempts are in the record? A number is a system. A single number is a system with an edit button.

What the parameters are set against

Acceptance criteria for suitability are chosen twice in a method's life: once during development, when they describe what the system achieved, and once at validation, when they become what the system must achieve. The second choice is the one that matters commercially, because it converts an observation into a gate.

The criteria should be set where the method makes decisions. A resolution criterion measured between the main peak and the nearest impurity protects the impurity result. One measured between two distant peaks protects nothing, because those peaks were never in danger of merging. Reading the criteria against the chromatogram is the fastest way to judge whether a method's gate is guarding the entrance or a side door.

The compendial general chapters from the USP and the EDQM describe the customary parameters and their definitions, and the ICH Q2 framework places the whole check within validation. None of them sets the numbers; the numbers belong to the method, which is why a suitability criterion inherited from a different method is a borrowed gate.

Preventive maintenance sits upstream of all of it

Most suitability failures trace to a component whose condition was known. A column past its injection count, a lamp with hours left, a pump seal that was due. The laboratories that rarely fail suitability are not luckier; they replace the consumables on a schedule that starts before the failure instead of after it.

Two records make that visible without any inspection: the column log, with injections against a defined retirement point, and the lamp hours where the detector reports them. Both cost minutes per run to maintain, and both convert a suitability failure from a mystery into an appointment.

The training variable sits beside them. A new analyst produces different preparation variability for the first weeks, and a method whose intermediate precision was validated across analysts absorbs that; one validated on a single analyst does not. That is the validation question resurfacing in the daily run, and it is one of the reasons intermediate precision is the figure worth reading.

What the record owes a future reader

A failed suitability event is complete when a reader who was not present can reconstruct it: what failed, by how much, what was checked in the order of blame, what was changed, and what passed afterwards. Five facts, and the event becomes part of the method's history rather than a gap in it.

The same five facts are what a buyer is entitled to when a certificate is questioned. The question to a supplier is not whether suitability ever fails, because it fails everywhere. It is what the record of the last failure looks like, and whether the sequence that produced the passing result is visible in it.

The multi-method laboratory

A laboratory running forty methods across five instruments has a suitability problem that a single-product site does not: the criteria belong to methods, and the instruments are shared. A column that satisfies one method may be marginal for another, and the schedule decides which method gets the good one.

The controlled way to handle it is to treat columns as method-specific assets with their own logs and retirement points, rather than as shared consumables. The uncontrolled way is to let the analyst choose, which produces a pattern where the difficult method is run on whatever is available and its suitability failures are attributed to the sample preparation.

There is a scheduling dimension as well. A system suitability run costs twenty to forty minutes, and a laboratory under pressure will batch the suitability runs at the start of a shift and the samples later. That ordering breaks the rule that suitability precedes samples on the same system, and it breaks it invisibly, because both runs pass.

The rule's purpose is proximity, not ceremony: the check describes the system as it is when the samples run. A check four hours and thirty injections earlier describes a system that no longer exists.

What good looks like from outside

Three observable habits separate a disciplined laboratory. Columns have logs with injection counts. Suitability runs sit immediately before their sample sets in the sequence. And failed runs remain in the system with their outcome visible.

All three are visible in a single screen-share session, and none of them requires the customer to understand chromatography. That is what makes them useful questions at qualification: they test the system's habits rather than the analyst's knowledge.

Frequently asked questions

What is system suitability actually testing?

The chromatographic system, on the day, under the conditions in use. The parameters, usually resolution, tailing factor, plate count and injection precision, are measured with a standard solution before samples are injected. If they fail, the system is not ready to produce defensible numbers, whatever the sample contains.

Does a suitability failure mean the batch is bad?

No. The suitability solution contains no sample. A failure before the first injection is a statement about the instrument, the column or the preparation of the standard. The batch has not yet been involved.

Can samples be injected while suitability is being fixed?

No. Injecting samples before suitability passes produces numbers with no validated context, and repeating suitability until it passes does not repair that. The sequence is a condition of the method, not a recommendation.

What is the order of blame when suitability fails?

Instrument function, column condition, mobile phase preparation, standard preparation, technique, and only then the method itself. Each step is cheaper to check than the next, and the method is usually innocent.

Why is repeat-until-pass a data integrity issue?

Because the acceptance criterion loses meaning if it can be retried until met. A criterion that passes on the third attempt describes a system that was not suitable twice. The record has to show every attempt, and the last one only counts if the sequence leading to it is documented.

What should a buyer see if a result was produced after a failure?

The failure, the corrective action, the passing suitability run, and the evidence that no sample results were reported from the failed interval. Four items, and they are the difference between a recovered run and a hidden one.