What is the TTC threshold?

The threshold of toxicological concern is 1.5 micrograms per day under ICH M7 for a mutagenic impurity at lifetime exposure. It applies as long as no compound-specific limit exists. For shorter treatment durations ICH M7 allows higher values under the less-than-lifetime staging. The cohort of concern (aflatoxin-like, N-nitroso and alkyl-azoxy compounds) is excluded: the TTC does not apply to them.

What exactly does class 3 mean?

Class 3 means alerting structure, unrelated to the structure of the drug substance, and no mutagenicity data available. ICH M7 offers two equivalent routes. Either you control the impurity permanently at the acceptable intake, usually the TTC threshold, in which case no assay is required. Or you run a bacterial mutagenicity assay: a negative result reclassifies the impurity to class 5 and it is treated as an ordinary impurity, a positive result places it in class 2 and it stays limited at TTC level. The assay is therefore not an obligation but the way out of the limit. Class 3 is explicitly not an all-clear.

Does the prediction replace the Ames test?

No. ICH M7 permits two complementary prediction methodologies in place of a test when both are negative and the rationale is documented. As soon as one of the two fires, the test or the limit applies. So the pre-screen decides which cases need the test at all.

Where do the structural alerts come from?

From a rule-based set of SMARTS patterns, each carrying a mechanism and a literature reference. For the nitrosamine alert the tool cites the IARC Monographs (carcinogenicity classification), for the aromatic acetamide alert Benigni and Bossa 2006, for the aminophenol alert Bolton et al. 2000. Every alert is therefore checkable, not merely asserted.

What happens to my structure?

The submitted structure is processed solely to handle the request, is not used for training, and is kept tenant-separated. Processing takes place in the EU, and deletion on request applies. This is anchored contractually in the data processing agreement, not only in this paragraph.

Can I screen an entire candidate list at once?

Yes. For series screening there is a batch variant that returns the same classification for many structures in one call. The difference from the single check is the call form, not the method.

Is this enough for a submission?

The output is a documented prediction with method, mechanism and source, so it is a defensible first pass. What becomes of it in a submission is your assessment. We do not deliver a validated system, we deliver the documentation you validate with.

ICH M7 pre-screening of nitrosamine impurities: what the class actually tells you

Author: CovaSyn team, experts in chemistry and analytics · Updated 2026-08-03

The short answer

An ICH M7 pre-screen answers exactly one question: does this impurity have to be treated as a mutagen or not? The four computed examples below show the result pattern. For the three nitrosamines NDMA, NDEA and NMBA, structural alert and statistical model agree, and the tool reports them as mutagenic (tool class 2). The regulatory picture goes further: as known mutagenic carcinogens they belong to ICH M7 class 1 and to the cohort of concern. The TTC of 1.5 micrograms per day explicitly does not apply to them; compound-specific limits do, for example 96 nanograms per day for NDMA.

The fourth case, paracetamol, is the instructive one. The tool finds two structural alerts, the statistical model disagrees at 0.17, and the classification from structure alone lands on class 3: alerting structure with no mutagenicity data. This is exactly where a prediction reaches its limit. Paracetamol has extensive negative Ames data, and with those data it belongs to ICH M7 class 5. For an unknown impurity with the same pattern, class 3 forces a decision: a permanent limit at TTC level, or a bacterial mutagenicity assay that moves it to class 5 or class 2 depending on the outcome.

The computed result

Compound	SMILES	Class per tool	Structural alerts	Statistical model	Confidence
NDMA (N-nitrosodimethylamine)	CN(C)N=O	2	nitrosamine	0.6905	0.628
NDEA (N-nitrosodiethylamine)	CCN(CC)N=O	2	nitrosamine	0.7215	0.656
NMBA (N-nitroso-N-methyl-4-aminobutyric acid)	CN(CCCC(=O)O)N=O	2	nitrosamine	0.8901	0.809
Paracetamol (control, not a nitrosamine)	CC(=O)Nc1ccc(O)cc1	3	aromatic_amine_acetamide, aminophenol	0.1677	0.648

Computed 2026-08-03 with covatox_ich_m7 against mcp.covasyn.com. The values are verbatim tool output, not editorially rounded.

Why is the structural alert alone not enough?

ICH M7 explicitly requires two complementary prediction methodologies, one expert rule-based and one statistical. The reason is visible in the paracetamol result: the aromatic acetamide alert fires on any acetylated aromatic amine, regardless of whether that specific compound is Ames-positive. The statistical model contributes the counter-evidence. A tool that delivers only one of the two either produces false alarms or misses cases.

The confidence column in the result above is therefore not decoration. For NMBA it sits at 0.81 because both methods point the same way. For paracetamol it sits at 0.65 despite two alerts, because the statistical model pulls against them. That number is the most honest figure on this page.

What follows for nitrosamines in practice?

Under ICH M7, class 1 means known mutagenic carcinogen and class 2 means known mutagen with unknown carcinogenic potential. The TTC of 1.5 micrograms per day applies to class 2. N-nitroso compounds, however, belong to the cohort of concern, for which ICH M7 explicitly excludes the TTC. For NDMA, NDEA and NMBA, EMA and FDA publish compound-specific acceptable intakes in the nanogram-per-day range. The pre-screen does not replace checking the current regulatory list; it tells you that you need to.

The mechanism the tool returns is not an extra. CYP2E1 activation to the diazonium ion followed by DNA alkylation is the reason the whole nitrosamine class is treated this way. Anyone who has to defend the classification to a regulator needs that sentence, not just the class number.

How fast is this in daily work?

The four runs above took under a minute in total. The point is not the speed as such, it is that a pre-screen moves to the start of the decision rather than the end. A candidate list of twenty possible degradation products can be reviewed before the first synthesis, not once the analytical result is in.

Frequently asked questions

What is the TTC threshold?: The threshold of toxicological concern is 1.5 micrograms per day under ICH M7 for a mutagenic impurity at lifetime exposure. It applies as long as no compound-specific limit exists. For shorter treatment durations ICH M7 allows higher values under the less-than-lifetime staging. The cohort of concern (aflatoxin-like, N-nitroso and alkyl-azoxy compounds) is excluded: the TTC does not apply to them.
What exactly does class 3 mean?: Class 3 means alerting structure, unrelated to the structure of the drug substance, and no mutagenicity data available. ICH M7 offers two equivalent routes. Either you control the impurity permanently at the acceptable intake, usually the TTC threshold, in which case no assay is required. Or you run a bacterial mutagenicity assay: a negative result reclassifies the impurity to class 5 and it is treated as an ordinary impurity, a positive result places it in class 2 and it stays limited at TTC level. The assay is therefore not an obligation but the way out of the limit. Class 3 is explicitly not an all-clear.
Does the prediction replace the Ames test?: No. ICH M7 permits two complementary prediction methodologies in place of a test when both are negative and the rationale is documented. As soon as one of the two fires, the test or the limit applies. So the pre-screen decides which cases need the test at all.
Where do the structural alerts come from?: From a rule-based set of SMARTS patterns, each carrying a mechanism and a literature reference. For the nitrosamine alert the tool cites the IARC Monographs (carcinogenicity classification), for the aromatic acetamide alert Benigni and Bossa 2006, for the aminophenol alert Bolton et al. 2000. Every alert is therefore checkable, not merely asserted.
What happens to my structure?: The submitted structure is processed solely to handle the request, is not used for training, and is kept tenant-separated. Processing takes place in the EU, and deletion on request applies. This is anchored contractually in the data processing agreement, not only in this paragraph.
Can I screen an entire candidate list at once?: Yes. For series screening there is a batch variant that returns the same classification for many structures in one call. The difference from the single check is the call form, not the method.
Is this enough for a submission?: The output is a documented prediction with method, mechanism and source, so it is a defensible first pass. What becomes of it in a submission is your assessment. We do not deliver a validated system, we deliver the documentation you validate with.

Sources

ICH M7(R2): Assessment and control of DNA reactive mutagenic impurities in pharmaceuticals to limit potential carcinogenic risk. Source of the class 1 to 5 scheme and the TTC threshold.
IARC Monographs: Reference the tool cites for the nitrosamine alert and the stated mechanism; IARC classifies carcinogenicity.
Bolton JL et al. (2000): Role of quinones in toxicology, Chemical Research in Toxicology 13, 135 to 160. Source of the aminophenol alert (oxidation to the quinone imine).
Benigni R, Bossa C (2006): Rule set of structural alerts for mutagenicity and carcinogenicity; cited by the tool as the source of the acetamide alert.

The same calculation on your structures

Send us your structures and receive the same evaluation as a report ready for approval. Or run it yourself, directly in your usual working environment.

Intro call Try the tools

How access works