EMA Nitrosamine Article 5(3)

In June 2018, NDMA (N-nitrosodimethylamine, a probable human carcinogen) was detected in valsartan API manufactured by a Chinese supplier. Investigation traced the impurity to the synthetic route — an interaction between solvents, catalysts and sodium nitrite under acidic conditions. The recall spread across the sartan class, and over 2018-2019 NDMA and other nitrosamines were detected in ranitidine, metformin, and several other widely-used products.

EMA's CHMP triggered an Article 5(3) referral in September 2019. Article 5(3) of Regulation (EC) 726/2004 lets EMA issue Union-wide scientific opinions on matters of public-health importance. The 2019 opinion required every MAH of chemical drugs in the EU to conduct a three-step programme; FDA issued aligned guidance shortly after (September 2020, revised September 2024).

Steps are sequential per product but products are managed in parallel. Many MAHs are still working through Step 2/3 for NDSRI-prone portfolios in 2025-2026.

Nitrosamine Drug Substance Related Impurities (NDSRIs) are nitrosamines formed in the finished product from a secondary-amine API reacting with trace nitrite in excipients. They were not the original valsartan concern (those were API-process nitrosamines), but they have become the dominant residual risk because they are product-specific and there is no analogue compound to anchor an Acceptable Intake.

EMA and FDA have published category-based AIs (the 'CPCA' approach in EMA's Q&A, the 'Carcinogenic Potency Categorisation Approach' in FDA's revised guidance) so MAHs can calculate an AI for an NDSRI without compound-specific carcinogenicity data. Where the resulting AI is too restrictive to meet, MAHs can pursue compound-specific carcinogenicity testing or reformulation.

The AI is the maximum daily lifetime exposure to a nitrosamine that is considered to carry an acceptable cancer risk (typically a 1-in-100,000 lifetime risk per ICH M7). EMA's Q&A document is the living source of nitrosamine-specific AIs. It is revised continuously as new toxicology data and new NDSRIs emerge — MAHs cannot rely on a printed snapshot.

• NDMA: 96 ng/day (FDA), 128 ng/day (EMA)
• NDEA: 26.5 ng/day (FDA), 26.5 ng/day (EMA)
• NMBA: 96 ng/day
• NDIPA: 26.5 ng/day
• NEIPA: 26.5 ng/day
• NDSRIs: per the CPCA category (18 / 100 / 400 / 1500 / 1500 ng/day)

A defensible nitrosamine control strategy combines source-reduction (eliminate the formation pathway where possible), analytical control (LC-MS/MS / GC-MS/MS as a release test where formation cannot be eliminated), and lifecycle review (the risk evaluation is a living document, re-reviewed when raw materials, suppliers, processes or packaging change).

1. Source reduction: replace nitrite-generating raw materials; change synthetic route; control acidic conditions; remove susceptible solvents.
2. Excipient screening: nitrite in excipients drives most NDSRI formation. Specify nitrite limits with the excipient supplier.
3. Analytical: validated, low-LOQ LC-MS/MS or GC-MS/MS; release specifications expressed as percent-of-AI for clarity.
4. Stability: NDSRI formation can be time-dependent; stability data must include the nitrosamine endpoint.
5. Lifecycle: any change to API supplier, excipient supplier, process or container-closure triggers re-evaluation.

When Step 3 requires a change to controls, it is filed as a variation (Type IA, IB or II depending on the change). EMA and the national competent authorities have streamlined the variation handling for nitrosamine changes; some changes can be handled by notification rather than full assessment. The MAH must communicate proactively if a confirmed result exceeds the AI — both to the authority and (if a recall is necessary) to the supply chain.

• Step 1 done once and never revisited when raw materials or suppliers change.
• Confirmatory testing not validated to low enough LOQ for the AI.
• AI snapshots in SOPs that are not updated when EMA Q&A revises.
• NDSRI risk underestimated because the API is not obviously a secondary amine in protonated form.
• Stability nitrosamine data missing — release passes but end-of-shelf-life would not.
• Excipient nitrite limits not specified in the supplier agreement.