ICH M7 in practice: classifying mutagenic impurities

What the guideline asks for, in three steps

A hazard assessment assigns every actual and potential impurity to one of five classes. A risk characterisation turns the class into a number. A control strategy decides where that number is enforced.

The first step runs on two kinds of evidence: literature and database data where they exist, and structure-activity relationships aimed at bacterial mutagenicity where they do not.

The computational requirement catches people out: it is not one model but two complementary (Q)SAR methodologies, one expert rule based and one statistical, both expected to follow the OECD validation principles for (Q)SAR models. The value of the pairing is the disagreement. Two models built on the same principles fail in the same places; a rule-based and a statistical system fail differently, so a negative from both is a stronger statement than a negative from either. That is also where the guideline permits you to stop: if neither flags a structural alert, the impurity is treated as non-mutagenic and no further testing is recommended.

The five classes

Classes 4 and 5 exist to stop the same chemistry being tested twice. Commercially that cuts both ways: a Class 4 argument rests on a compound already tested rather than a new standard, while the cheapest route out of Class 3 is usually an assay on the impurity — which needs it as a characterised material.

The default limit, and what it does not cover

For Classes 2 and 3 the default acceptable intake is 1.5 micrograms per person per day: a threshold of toxicological concern corresponding to a theoretical excess cancer risk below 1 in 100,000 across a lifetime of daily exposure. It is deliberately conservative, and it is a default rather than a universal answer. Three things sit outside it:

• Class 1 impurities, calculated from carcinogenic potency by a compound-specific assessment that can land above or below the default.
• The cohort of concern — aflatoxin-like, N-nitroso and alkyl-azoxy structures — which can be far more potent, and whose intakes are set case by case at levels expected to be significantly lower.
• Shorter treatments, which get higher limits, and the change is substantial.

Why a one-month programme gets a much higher limit

Risk is assumed to track cumulative dose, so a low dose over a lifetime and the same total dose compressed into a few months carry comparable risk. The default of 1.5 micrograms per day across 25,550 days of lifetime exposure gives a cumulative budget of about 38.3 mg, which is then spread over the days the patient is actually treated.

Two details do real work. For intermittent dosing the bands apply to dosing days, not the calendar interval: a product dosed weekly for two years has 104 dosing days and falls in the 20 microgram band per dose. And a compound-specific limit can be scaled up for shorter treatment but is capped at 0.5 percent of the maximum daily dose, whichever is lower — so 15 micrograms per day can rise to 1,200 for a course under a month, but a drug dosed at 100 mg per day caps that at 500 micrograms.

Clinical development uses the same logic with tighter targets: 1 in a million in early development, where benefit has not yet been established, and 1 in 100,000 later. For phase 1 trials of up to 14 days the scope narrows further — only Class 1, Class 2 and cohort-of-concern impurities must be controlled; a Class 3 alert alone does not trigger action.

So the same impurity can be a limit test in one programme and a quantitative assay in another: the molecule did not change, the treatment duration did.

When more than one impurity is in scope

Two Class 2 or Class 3 impurities are controlled individually. From three upwards — counting only those specified on the drug substance specification — the total is capped as well.

Class 1 impurities are not counted into that total, since they already carry compound-specific limits of their own. Degradants formed in the drug product are controlled individually, with no total applied.

For a laboratory this is the difference between one standard and a panel: a total limit is only demonstrable if every analyte inside it can be measured.

The four control options

Where the limit is enforced is a separate decision from what the limit is, and the four recognised approaches consume standards in very different quantities.

The direction of travel is deliberate: the further upstream the control sits, the more process understanding it demands and the less analytical work it leaves. Option 4 suits impurities that are inherently unstable, or that enter early and are effectively cleared by later steps — not one formed in the last synthetic step, where an Option 1 control is the expectation unless specifically justified. Below the acceptable limit, there is no requirement to drive the impurity as low as reasonably practicable.

Where the standard actually gets used

A purge factor is the level of an impurity at an upstream point in the process divided by its level downstream. It may be measured or predicted, and a predicted factor is what carries an Option 3 or Option 4 argument.

That is where a standard stops being a formality. A spiking study adds a known amount of the impurity and follows it through the process — and a known amount means weighing a characterised material, so the assigned content, not the area percent on a chromatogram, is the number that enters the calculation. Spiking at percent levels and tracking across several downstream steps can yield a purge factor in the hundreds: exactly the evidence an Option 3 control rests on. If the standard is not quantified on a mass basis, the purge factor inherits the error.

What to have on the shelf

1. A standard for every Class 1 and Class 2 impurity in scope, with an assigned content and its basis stated.
2. The material for each Class 3 impurity, whether you control it or resolve it by assay.
3. Labelled analogues where the method is quantitative at trace level; they are usually the longest lead, and isotope dilution is what controls recovery and ionisation losses.
4. For a Class 4 argument, the related compound already tested: the classification rests on existing data, not a new measurement.
5. Enough quantity for spiking work, not just a system suitability injection — purge studies consume material across several process steps, often at more than one concentration.
6. An uncertainty statement for anything feeding a purge factor or a limit calculation.

Traps worth naming

• Treating a computational negative as permanent. A (Q)SAR conclusion follows the structure; change the route, starting material or solvent and the impurity profile changes with it, so the assessment has to be revisited.
• Applying the 1.5 microgram default to a Class 1 impurity. Where carcinogenicity data exist, the compound-specific intake governs, and it can be lower.
• Reading the duration bands off the calendar. Intermittent regimens are counted in dosing days, which can move an impurity up a band or more.
• Forgetting the 0.5 percent ceiling when scaling a compound-specific limit, or buying one standard for a total limit covering three or more analytes.
• Scheduling the spiking study before the standard arrives. The lead time on a non-catalogue impurity sets the date, not the lab slot.
• Feeding an area-percent value into a mass calculation. See what an impurity reference standard is for why the basis of the purity figure decides what it can be used for.

Where to start

What turns a mutagenic-impurity enquiry into an accurate quotation is short: the structure or SMILES of the impurity, the parent drug substance, the class if assigned, the quantity, and whether the intended use is identification, a limit test, or a quantitative assay with an uncertainty statement. The last is a different product from the first two, and it is the one that purge and spiking work depends on. Impurities in no catalogue are the normal case here rather than an obstacle: send the structure, or the parent drug substance and the site of the alert, to technical@rssechem.com. The impurity standards index is organised by parent drug substance, and the certificate of analysis page shows the fields we report.