Why nitrosamines remain a critical issue The presence of nitrosamines in medicines has been a major concern since the first recalls in 2018. These compounds are classified as probable human carcinogens, and even trace amounts can raise questions about patient safety. Regulators across the world, including the European Medicines Agency and the Food and Drug Administration, have issued strict guidance to control the presence of nitrosamine impurities in both active substances and finished products (European Medicines Agency, 2025; U.S. Food and Drug Administration, 2023). The issue is not limited to one therapeutic class. It spans small molecules, complex generics, and even some biological products. The challenge lies in the many possible routes for nitrosamine formation during synthesis, formulation, packaging, and storage. These risks demand a structured, science-based approach that covers the entire product lifecycle — and in our experience, the firms that handle it well treat it as a continuous control problem, not a one-off testing campaign. What does a nitrosamine control programme actually require? A defensible control programme rests on three pillars: a documented risk assessment that maps every plausible formation pathway, confirmatory analytical testing at sensitivity well below the acceptable intake (AI) limit, and a control strategy that prevents formation rather than catching it at release. The acceptable intake for N-nitrosodimethylamine (NDMA) sits at 96 ng/day for chronic exposure (published-survey: European Medicines Agency, 2025) — a threshold that forces analytical methods into the trace range and shapes every upstream decision. Both EMA and FDA require marketing authorisation holders to perform risk assessments, confirmatory testing, and implement corrective actions. These steps are not one-off exercises. They form part of a continuous monitoring process because new information, new suppliers, or changes in manufacturing can alter the risk profile. For pharma teams already running a GxP programme, the nitrosamine workstream sits inside the broader compliance envelope we explore in GxP regulations explained for AI software in pharma. The data-integrity and audit-trail expectations are the same; only the analytical and chemistry inputs differ. How nitrosamines enter the picture Understanding the chemistry is the entry point. Nitrosamine formation typically involves secondary or tertiary amines reacting with nitrosating agents such as nitrite under acidic conditions. This can happen in the API synthesis route, during recovery of solvents, or through contaminated raw materials. Excipients can also contribute. Some contain trace nitrite, which can react with amines in the formulation. Packaging is another source. Certain lidding foils, inks, and adhesives may release nitrite or nitrogen oxides into the headspace of a sealed pack. Even storage conditions matter — heat and humidity can accelerate reactions that create nitrosamine impurities over time. This breadth is why a single test result, however clean, is not evidence of a controlled process. The pathways are too varied and too sensitive to lot-level variability for a snapshot to be reassuring on its own. A structured approach to risk assessment The first step is a thorough risk evaluation. Map every plausible pathway for nitrosamine formation across the manufacturing process and supply chain. Consider starting materials, reagents, catalysts, and recycled solvents. Review excipient specifications for nitrite content. Audit packaging components for potential migration. Once the map is complete, prioritise scenarios by likelihood and patient exposure. High-risk cases move to confirmatory testing. This testing must use sensitive and selective methods. For volatile nitrosamines like NDMA, GC–MS with headspace sampling is common. For non-volatile species, LC–HRMS or LC–MS/MS is preferred. Detection limits should sit well below the AI to provide confidence in negative results. Quick reference: analytical method by nitrosamine class Nitrosamine class Preferred technique Typical sensitivity target Volatile (NDMA, NDEA) GC–MS with headspace sampling ≤ 30% of AI limit Non-volatile / drug-substance-specific (NDSRIs) LC–HRMS or LC–MS/MS ≤ 30% of AI limit Trace nitrite in excipients Ion chromatography or LC–MS ppm-level reliable Packaging headspace migration GC–MS with thermal desorption Sub-ppb in headspace The table is a starting point, not a substitute for method validation against the specific matrix. NDSRIs (nitrosamine drug-substance-related impurities) in particular often need bespoke method development because each API generates a structurally distinct nitrosamine. Controlling nitrosamine levels in practice When testing confirms the presence of nitrosamines, the priority is to remove the root cause. This could mean switching to low-nitrite excipients, tightening pH control, or replacing a reagent. In some cases, adding scavengers or antioxidants can help. Packaging changes may also be necessary, such as moving to foils with proven low migration. Process adjustments should be documented and justified with data. Each change must show a measurable reduction in nitrosamine levels. End-product testing alone is not enough. Regulators expect a control strategy that prevents formation rather than relying on detection after the fact — the same logic that drives continuous validation thinking elsewhere in pharma compliance. Lifecycle monitoring and trending Risk does not end after implementation. Companies must trend results over time to confirm that controls remain effective. This includes routine testing of high-risk products, periodic checks on excipient lots, and monitoring of packaging suppliers. Trending helps detect early signals of drift and supports decisions on shelf-life or storage conditions. The EMA’s Q&A guidance stresses that marketing authorisation holders remain responsible for ongoing vigilance. The FDA echoes this in its updates, reminding firms that presence of nitrosamine impurities can occur even after years on the market if processes or materials change (published-survey: European Medicines Agency, 2025; U.S. Food and Drug Administration, 2023). The implication is operational: trend dashboards, supplier change-control hooks, and a defined re-assessment cadence belong in the system from day one. Documentation that stands up to inspection Regulators expect clear, concise evidence. Three files do most of the work in an inspection: A live risk assessment with dates, rationales, and decisions on what was tested versus excluded. Analytical reports with method details, validation data, and raw results. A control strategy that links each mitigation to its measurable effect on nitrosamine levels. When acceptable intake limits cannot be met immediately, companies may apply for temporary measures under less-than-lifetime exposure (LTL) principles. These cases require strong justification and a clear timeline for corrective action. We see them treated as bridging arrangements, not destinations — inspectors expect a credible path back to the chronic AI. Common pitfalls to avoid One frequent error is assuming that absence of nitrite equals zero risk. Trace amounts can vary between lots and still drive nitrosamine formation under the right conditions. Another mistake is overlooking packaging — migration from foils or inks has caused several confirmed cases. A third is treating the initial assessment as a one-time task. In reality, this is a lifecycle obligation that demands periodic review, supplier-change triggers, and a clear owner. A fourth, subtler pitfall: confusing GxP-critical analytical systems with auxiliary lab tooling. The LC–HRMS instrument generating release data is squarely in scope; a spreadsheet used by chemists to plan an experimental design may not be. Drawing that line proportionately — neither over-validating nor under-validating — is the same boundary problem we examine for AI software in pharmaceutical inspections and compliance essentials. Why this matters for global compliance Both EMA and FDA align on the fundamentals: science-based risk assessment, sensitive testing, and proactive control. Other agencies, supported by the World Health Organization, follow similar principles to protect patients worldwide. For companies, this means a harmonised approach can satisfy multiple markets and reduce duplication. For patients, it means safer medicines and fewer recalls. The serialisation and traceability layer matters here too. When a nitrosamine signal emerges post-market, the speed of containment depends on barcode-level traceability — a topic we cover in barcodes in pharma from DSCSA to FMD in practice. How TechnoLynx can help We work with pharmaceutical companies to design and implement robust nitrosamine control programmes. We start with a detailed risk map tailored to your processes and materials. Our team develops and validates advanced analytical methods for trace detection, including LC–HRMS and GC–MS workflows. We also build trending dashboards that link results to suppliers, batches, and packaging lots — using the same data-engineering stack (Python, PyTorch where ML is appropriate, containerised pipelines on Docker and Kubernetes) we deploy across our pharma engagements. Every solution comes with audit-ready documentation and a lifecycle monitoring plan, consistent with EMA, FDA, and other global requirements. FAQ What does GxP compliance specifically require when the software is AI/ML rather than deterministic code? GxP compliance still demands the same data integrity, audit-trail, and change-control discipline, but the validation evidence shifts. For ML components, the validation package must cover training-data provenance, model-version control, drift monitoring, and a documented retraining trigger — alongside the conventional installation, operational, and performance qualification. Which GxP rules apply to AI training data, models, and inference outputs? Training data inherits the data-integrity expectations of GMP records (ALCOA+). Models are treated as configured software under GAMP 5 categorisation, typically Category 4 or 5 depending on customisation. Inference outputs that influence a GxP decision — release, batch disposition, deviation classification — are GxP records and must be retained, traceable, and reproducible. How is a GxP-validated AI system kept compliant as the model retrains or drifts? Through a defined change-control workflow that treats retraining as a controlled change. That means pre-specified performance thresholds, a documented validation re-run for material changes, and continuous monitoring of drift indicators against the qualified baseline. Silent retraining is not compliant. Where is the boundary between GxP and non-GxP usage of AI inside a pharma manufacturing workflow? The boundary is set by impact on product quality, patient safety, or data integrity of a GMP record. An AI tool that ranks experimental designs for R&D scientists is generally non-GxP; the same model architecture deployed to classify deviations or approve batch release crosses the line into GxP scope. Which GxP roles own AI-specific risks and how is that documented? System owner, QA, and validation lead retain their conventional responsibilities, augmented by a data-science or ML-engineering function that owns model lifecycle artefacts. Roles are documented in the validation plan, the data-governance charter, and the SOPs covering retraining and drift response. How do ISPE’s GAMP AI guidance and the ISPE AI maturity model fit into an existing GxP programme? They extend rather than replace the existing programme. GAMP 5 remains the categorisation backbone; the AI guidance adds specifics on training data, model governance, and continuous validation. The maturity model is a self-assessment lens, useful for prioritising which capabilities to build first rather than as a compliance gate in itself. References European Medicines Agency (2025) Nitrosamine impurities: guidance for marketing authorisation holders. Available at: https://www.ema.europa.eu/en/human-regulatory-overview/post-authorisation/pharmacovigilance-post-authorisation/referral-procedures-human-medicines/nitrosamine-impurities/nitrosamine-impurities-guidance-marketing-authorisation-holders (Accessed: 26 September 2025). European Medicines Agency (2025) Q&A on the CHMP Article 5(3) opinion on nitrosamine impurities. Available at: https://www.ema.europa.eu (Accessed: 26 September 2025). U.S. Food and Drug Administration (2023) Control of nitrosamine impurities in human drugs. Available at: https://www.fda.gov (Accessed: 26 September 2025). World Health Organization (2023) Medication safety and nitrosamine risk management. Available at: https://www.who.int (Accessed: 26 September 2025). Image credits: Freepik.
