Topic · Toxicology
“QA wants an ICH M7 classification this week, with a written rationale.”
ICH M7 assessment of impurities
What ICH M7 (R2) requires for mutagenic impurities and how CovaSyn delivers the groundwork: both assessment methods in one run, the applicability domain checked per prediction, a table ready for review by your experts.
| Peak | t (min) | Area % |
|---|---|---|
| V1 | 2.1 | 0.09 |
| V2 | 3.4 | 0.24 |
| API | 6.2 | 99.15 |
| V3 | 7.8 | 0.38 |
| V4 | 9.1 | 0.14 |
Regulation
Obligations and deadlines
What the guideline sets out, as stated in our Insights articles and the ICH M7 guide. The final classification remains the job of your experts.
| Framework | What applies | Threshold or deadline |
|---|---|---|
| ICH M7 (R2), assessment | For bacterial mutagenicity, two complementary (Q)SAR methods: one expert rule-based and one statistical. If both are negative, class 5 is possible without further testing. | Two methods, both documented |
| ICH M7 (R2), class 1 | Known mutagenic carcinogen: control at or below the compound-specific acceptable intake. | Compound-specific, e.g. NDMA with a published limit of 96 ng/day |
| ICH M7 (R2), class 2 | Known mutagen with unknown carcinogenic potential: control at or below the acceptable limit. | TTC-based, except for the cohort of concern |
| ICH M7 (R2), class 3 | Alerting structure unrelated to the drug substance, no mutagenicity data: control at the TTC limit or a bacterial mutagenicity test to reclassify to 5 or 2. | TTC or Ames test |
| ICH M7 (R2), classes 4 and 5 | Alert shared with the drug substance or a tested non-mutagenic relative (4), or no alert or an alert with data showing no mutagenicity (5): treat as non-mutagenic. | ICH Q3A/Q3B applies |
| ICH M7 (R2), TTC | Threshold of toxicological concern for controlling mutagenic impurities at lifetime exposure. | 1.5 µg/day |
| ICH M7 (R2), cohort of concern | N-nitroso compounds, aflatoxin-like and alkyl-azoxy structures: the general TTC does not apply. A compound-specific acceptable intake from carcinogenicity data, read-across or a framework such as CPCA is required. | Not 1.5 µg/day |
| ICH M7 (R2), prediction without data | A prediction alone gives class 3 or 5 at most. Classes 1, 2 and 4 require data or comparison with the drug substance. |
Sources: our articles on ICH M7 classification and nitrosamine risk assessment and the ICH M7 guide, linked below.
CovaSyn
How CovaSyn does it
CovaSyn makes the groundwork reproducible and traceable. The classification, control strategy and filing are owned by your experts in toxicology, QA and regulatory affairs.
Both methods in one run
Rule-based structural alerts with mechanism and literature reference, and a statistical Ames probability with confidence. Both results come back separately, not as a merged verdict.
Many impurities against the API
Up to 50 impurities in one run, with class per impurity, class distribution, worst case and similarity to the drug substance. That separates drug-related from process impurities.
Applicability domain per prediction
If a structure lies outside the model space, the result says so. The expert rule then carries the rationale, not the statistical score.
Audit trail from the first call
Timestamp, tool, version, result status and key ID per call, outputs with a SHA-256 checksum, export as CSV or JSON.
GAMP 5 category 4
A configured product, no custom code at your site. Validation takes place in your organisation.
Your experts decide
Class 3 still needs review by qualified toxicologists. Classes 1 and 2 are reported, not blocked automatically.
What the computation does not replace
- The tool does not replace qualified toxicology experts. The final classification and filing strategy remain with QA and regulatory affairs.
- A statistical score outside the applicability domain is not evidence but a directional signal.
- Listing nitrosatable amines in the drug substance, the route and the formulation is chemistry, not software.
Insights
Read on and try it yourself
ICH M7 Classification: Class 1 to 5, by Example
ICH M7 classification explained class by class, with the dual (Q)SAR assessment and five real impurities scored live by CovaTox, including NDMA.
Read articleNitrosamine Risk Assessment: An ICH M7 Workflow
A practical nitrosamine risk assessment workflow for NDSRIs: ICH M7 dual assessment, real CovaTox classifications, and where the QSAR goes out of domain.
Read articleICH M7 automated: mutagenicity triage in seconds on a Losartan worked example
ICH M7 requires two (Q)SAR methods per impurity. On a Losartan worked example: 10 impurities classified in seconds, 5 probable mutagens identified, including the well-known azido impurity (AZBT). How CovaTox covers both methods in one call, and where the limits are.
Read articleApplicability Domain in QSAR: When a Model Says I Don't Know
What the applicability domain in QSAR is, how Tanimoto, feature range and leverage methods define it, and how CovaTox downgrades out-of-domain predictions.
Read articlePDE vs TTC in Cleaning Validation: The Genotoxic Case
PDE vs TTC cleaning validation: when a health-based PDE applies, when ICH M7 forces the 1.5 µg/day TTC, and how to check the residue is even cleanable.
Read article
Questions
Common questions
Does the 1.5 µg/day TTC apply to nitrosamines?
No. N-nitroso compounds belong to the ICH M7 cohort of concern. A compound-specific acceptable intake is required, derived from carcinogenicity data, read-across or a framework such as CPCA.
Why does ICH M7 require two methods?
The rule set captures mechanisms that a model trained on Ames data cannot express traceably. The statistical model catches structures no rule covers. Both must be allowed to disagree.
Next step
Assess your own impurities.
Create an account, get your API key, use the tools in Claude, ChatGPT or Cursor.