Impurity profile comparison: what a reference standard proves, and what it does not

A profile is a description, not a number

Both impurity guidelines define the term the same way, at the level of the substance and of the product: a description of the identified and unidentified impurities present in the material. That definition carries three obligations which a total-impurity figure cannot satisfy. Every peak that matters needs a designation, a structure or a stated qualitative label. Every peak needs a level, with the basis of the calculation named. Every identified impurity needs a classification, because a process impurity and a degradation product are controlled in different places.

Comparisons fail at the first obligation. One laboratory reports an unidentified peak at relative retention 0.87, another at 0.89, and the two are treated as the same species because the numbers are close. They may be one species on two column lots, or two species sitting near each other. A profile is comparable only when designation, level and basis travel together.

The threshold ladder fixes what a comparison may ignore

The ladder is dose-dependent, so the scope of a comparison is fixed by the maximum daily dose of the product, not by the habit of the analyst. For a drug substance: reporting 0.05 per cent up to 2 g a day and 0.03 per cent above it, identification 0.10 per cent or 1.0 mg a day whichever is lower, qualification 0.15 per cent or 1.0 mg a day. For a drug product the bands are set by dose, because a degradation product is judged against daily intake rather than against the mass of the unit:

Two consequences follow. Each threshold exists in two currencies, a percentage and a daily intake, and the lower applies: 0.2 per cent at a 100 mg dose is 200 micrograms a day, and at 1.5 g it is 3 mg, so the 2 mg a day cap takes over. And below 1 per cent, results are reported to the number of decimal places of the threshold, so 0.06 is not rounded to 0.1.

The acceptance criterion in a specification is a separate question. Criteria are extrapolated from batch and stability data, and the specifications guideline is explicit that criteria tightly encompassing the batch data available at filing are inappropriate. A comparison borrowed from another strength, or from a monograph for a different salt, answers a question nobody asked.

Area per cent is not mass per cent

An ultraviolet detector responds to mass through a molar absorptivity that differs from compound to compound, so equal areas do not mean equal amounts. Both guidelines allow the drug substance itself as the standard for estimating impurity levels, on one condition: where the response factors are not close, a correction factor is applied, or the impurity is knowingly overestimated, and the assumption of equivalent response is stated in the application.

The correction is a division. A relative response factor of 0.5 means the impurity gives half the response per unit mass, so its true level is twice its area per cent. On a 100 mg tablet, with an identification threshold of 0.2 per cent:

Two of the four decisions invert. Peak B is 0.24 per cent by mass, above both the identification and the qualification threshold, and needs a structure and a safety argument; read straight off the chromatogram it would have been reported and left unidentified. Peak C is 0.15 per cent by mass and needs reporting only, where the uncorrected area would have triggered an identification study that answers nothing. One injection of the standard at a known concentration corrects all four, and is also how the factor is measured.

What the correction does not fix: a factor taken at one wavelength does not transfer to another detection mode, a factor measured against a standard of the wrong salt form carries that salt's mass into the arithmetic, and a factor taken at one level assumes a linearity an overloaded injection has already broken.

What a co-injection actually proves

Retention time is a qualitative property with a quantitative reputation. Two compounds can share it, and one can move off it as a column ages. The evidence for identity has a recognisable order. A retention match excludes a grossly different species and little else. Spiking the sample with the standard excludes one peak hiding two, since the spike either stays single or splits. A photodiode array spectrum excludes a different chromophore but not an isomer. Mass confirmation excludes a different formula, not a positional variant. Only NMR on an isolated portion settles constitution. The hidden condition on the first line is the standard itself: it has a form and an isomeric purity, and a match against the wrong form is not a match. Where the impurity exists as a salt and the standard is the free base, retention shifts and the comparison is recorded as a missing peak. The conventions are in salt, hydrate and free base conversion.

Without a standard: relative retention and its drift

Most peaks in a comparison have no commercial standard, and the guidelines allow an unidentified impurity to be defined and named by its relative retention. The number is not a constant. It moves with column lot, mobile phase pH, gradient delay volume, temperature and sample solvent strength, and it moves most on early peaks.

A retention window must therefore be established from the laboratory's own injections across at least two column lots, and re-established at method transfer rather than carried over, as covered in method transfer acceptance criteria. Quantitation without a standard proceeds on an assumed factor of one, and the assumption is written down. An assumed factor is not neutral: for a chromophore-poor impurity it can understate the level several fold, which is the direction that hides a finding.

Close the mass balance before arguing about a peak

Assay, organic impurities, water, residual solvents and inorganic residue are parts of one budget. Where they do not add to the whole, something is present and unseen, and each peak percentage is computed against a total that is itself incomplete. The usual explanations: a water or counter-ion term left out, a volatile degradant lost during drying, a species with no chromophore, material retained on the column, loss during sample preparation. Where the assigned content comes from quantitative NMR rather than a subtraction, qNMR against mass balance explains why the two can disagree without either being wrong.

Two preparation effects mimic chemistry: a solvent stronger than the mobile phase distorts early peaks, and an injection past the linear range inflates the baseline.

A structural alert outranks the ladder

The ladder answers "at what level must this be described". It does not answer "is this safe at that level". For a potentially mutagenic impurity the operative number is a threshold of toxicological concern of 1.5 micrograms per day, corresponding to a theoretical excess lifetime cancer risk of one in a hundred thousand, and a small cohort of high-potency structures is treated below it.

Set the two frameworks side by side. A peak at 0.1 per cent of a 100 mg dose is 100 micrograms a day, about sixty-seven times that threshold; the 0.24 per cent peak above is 240 micrograms a day, one hundred and sixty times. Being small does not clear an impurity; having a structure that has been assessed does. Where a new peak appears and its structure is unknown, classification comes before any discussion of limits, as set out in ICH M7 in practice.

Decisions by comparison outcome

What to ask for when the standard is for profile work

• The identity and the form, stated separately, since a free base standard cannot confirm a salt impurity.
• The method behind the assigned content, and whether it separates the impurity from its own isomers.
• A relative retention or response factor measured on a stated system, with the wavelength named.
• The certificate convention for the assigned value: as-is, dried or anhydrous basis, with the water content stated.
• The storage and validity statement, since a standard that sat warm for a year no longer represents the peak it was bought to identify.

Each of those items changes what a comparison can conclude. A standard without a stated basis turns a 0.24 per cent finding into a range; a standard of the wrong form turns a present peak into a missing one. The documentary side is in how to read a certificate of analysis, and the distinction between an impurity standard and a characterised material in what an impurity reference standard is.

How a comparison usually goes wrong

• Area per cent compared across methods with different wavelengths, and a difference in purity reported.
• A retention match treated as identification, with the spiking experiment skipped.
• The comparison run at one dose strength when the thresholds are banded by dose.

A profile comparison is an argument about three things at once: what is there, how much of it there is, and whether anyone has decided it is acceptable. A standard settles the first, and only when its form, its basis and its evidence of identity are known. The second needs a response factor and a calibrated injection. The third needs a structure and a toxicological judgement that no chromatogram contains.