ICH M7ICH M7(R2) DNA-Reactive (Mutagenic) Impurities

ICH Q3A (impurities in new drug substances) and Q3B (in new drug products) set thresholds for identification and qualification of impurities based on patient daily exposure. M7 is a parallel framework specifically for impurities that are DNA-reactive — that is, capable of forming DNA adducts and therefore potentially mutagenic and carcinogenic. For DNA-reactive impurities, the standard Q3A/Q3B thresholds (typically 0.10-0.15 % of the API) are far too permissive because mutagenic impurities can pose cancer risk at much lower levels. M7 introduces a separate, much tighter TTC-based framework with a default of 1.5 µg/day for a lifetime dose and tighter compound-specific AIs for high-potency cohort-of-concern impurities.

Classification uses (Q)SAR predictions from at least two complementary methodologies (an expert-rule-based system such as Derek and a statistical system such as Sarah, Leadscope, etc.). When predictions are negative or expert-reviewed-negative, the impurity moves to Class 4 or 5. When predictions are positive, Ames testing can be performed to confirm/refute the prediction; a negative Ames result allows Class 5 classification.

For Class 3 impurities (no compound-specific data), the default limit is the Threshold of Toxicological Concern (TTC) of 1.5 µg/day for lifetime exposure — derived from a probabilistic analysis of the cancer-slope distribution across known carcinogens, calibrated to a 1-in-100,000 lifetime cancer risk. For Class 1 and Class 2 with available data, compound-specific Acceptable Intakes (AIs) are derived using the TD50 method (TD50 ÷ 50,000 = AI for lifetime exposure).

Less-than-lifetime (LTL) exposure adjustments allow proportionally higher limits when treatment duration is bounded: ≤1 month at 120 µg/day, ≤1 year at 20 µg/day, >1 year (lifetime) at 1.5 µg/day. The LTL framework is widely used for clinical-trial materials and acute-care medications.

M7 designates a 'cohort of concern' — compound classes whose toxicity is significantly higher than the TTC default, requiring compound-specific AIs rather than the 1.5 µg/day TTC. The cohort includes: aflatoxin-like compounds, N-nitroso compounds, alkyl-azoxy compounds, and certain high-potency genotoxins like benzo[a]pyrene. For these classes, the AI is typically 1-2 orders of magnitude tighter than the TTC. N-nitrosamines drove the entire post-2018 programme; FDA's published nitrosamine AIs (NDMA 96 ng/day, NDEA 26.5 ng/day, etc.) sit in this cohort-of-concern regime.

1. Option 1 — control the impurity in the drug substance at or below 30 % of the AI (default).
2. Option 2 — control the impurity at higher levels in an in-process intermediate provided the downstream chemistry demonstrably purges it to below the AI by drug-substance release.
3. Option 3 — control the impurity in an upstream starting material or reagent provided downstream purging is established.
4. Option 4 — predict the level by understanding the fate-and-purge of the impurity (no routine testing needed) when the calculated worst-case downstream level is below 30 % of the AI.

Option 4 is the most efficient when the chemistry supports it — but requires a robust justification that any change in upstream conditions won't invalidate the calculation. Options 2-4 are particularly relevant for impurities that form transiently in synthesis but are purged by crystallisation, distillation or downstream wash steps.

1. Risk assessment using a single (Q)SAR engine instead of two complementary systems.
2. Negative (Q)SAR prediction accepted without expert review — particularly for borderline structures.
3. Option 4 'predicted to purge' claim with no quantitative purge data — regulator wants a defensible spike-and-purge model.
4. Cohort-of-concern impurity (e.g. a nitrosamine, an alkyl azoxy) limited to the 1.5 µg/day TTC rather than to the much tighter compound-specific AI.
5. LTL adjustment used for a chronic-treatment indication.
6. Risk assessment not refreshed after a synthetic-route change, supplier change or starting-material change.
7. Test method not validated to the very low ppb / ppt levels required by cohort-of-concern AIs.

• Impurity register per drug substance / drug product holds each potential mutagenic impurity, its M7 class (1-5), the (Q)SAR predictions, the AI/TTC derivation, the chosen control option (1-4) and the supporting purge or test data — all as controlled records under change control.
• Change-control workflow auto-routes synthetic-route changes, supplier changes, starting-material changes and formulation changes to the M7 risk-assessment owner; the file is re-opened, not silently invalidated.
• LIMS captures M7 impurity test results as critical release attributes; the AI and the percent-of-AI are visible on the CoA and the BMR, not only on a separate impurity-control spreadsheet.
• Cohort-of-concern handling (nitrosamines particularly) integrates with the published per-compound AIs; updates to AIs from FDA/EMA flow into the impurity register via a regulated-update workflow.
• Recall-readiness: an exceedance triggers genealogy lookup of all impacted lots and downstream consignees in minutes — minimising the 'how many days of supply has this been on the shelf' question that dominates M7 exceedance investigations.