FDA Nitrosamine Impurity Requirements in NDA, ANDA, and Supplement Reviews

FDA nitrosamine requirements in NDA, ANDA, and supplement reviews: risk assessments, acceptable intake limits, and control strategies

FDA now treats nitrosamine risk as a standard part of the CMC and nonclinical review of new drug applications, generics, and postapproval supplements. Across recent Drugs@FDA review packages, the pattern is consistent. Applicants are expected to submit a nitrosamine risk assessment that covers both small-molecule nitrosamines and nitrosamine drug substance-related impurities (NDSRIs). Where a risk exists, they must justify an acceptable intake (AI), usually through FDA's carcinogenic potency categorization approach (CPCA). They must then show, with batch data, that a control strategy keeps the impurity at or below that AI through shelf life. Missing or inadequate nitrosamine data has been treated as a preapproval deficiency and has contributed to at least one complete response.

Many impurity identities and numerical limits in FDA's public reviews are redacted under (b)(4). Where a value is not given below, the review withheld it. FDA did not necessarily leave the value unset.

The regulatory framework reviewers apply

Guidance documents

Reviewers measure submissions against two final guidances. The first is Control of Nitrosamine Impurities in Human Drugs (September 5, 2024), which revises the February 2021 guidance of the same title. The second is Recommended Acceptable Intake Limits for Nitrosamine Drug Substance-Related Impurities (NDSRIs) (August 7, 2023) ^387388389.

Three-step process

The guidance sets out three steps ³⁹⁰:

1. Assess nitrosamine risk in APIs and drug products, including NDSRIs, and keep the assessment available for FDA on request.
2. If a risk is identified, run confirmatory testing with sensitive, appropriately validated methods.
3. Report changes made to prevent or reduce nitrosamines through the applicable DMF amendment or application-change pathway.

FDA recommends revisiting earlier assessments that left out NDSRIs and reassessing risk periodically ³⁹⁰. Confirmatory testing should generally cover at least three representative batches unless a data-driven risk assessment justifies omitting it ³⁹¹.

Acceptable intake limits

An AI approximates one additional cancer case per 100,000 people, assuming daily exposure over 70 years ³⁹². Among small-molecule nitrosamines, FDA gives an AI of 96 ng/day for NDMA and 26.5 ng/day for NDEA ⁶⁶⁸⁶⁶⁹. If no AI can be determined by the approaches in the guidance, FDA recommends 26.5 ng/day as the fallback ⁶⁷¹.

For NDSRIs, the CPCA assigns a potency category based on activating and deactivating structural features, including how favorable α-hydroxylation is ³⁹⁵³⁹⁶:

Categories 4 and 5 rest on the threshold of toxicological concern. Category 5 covers structures not predicted to undergo α-hydroxylation ³⁹⁷. Some structures are excluded from the CPCA, including nitroso compounds with a carbon double-bonded to a heteroatom and compounds whose N-nitroso group sits in an aromatic ring ³⁹⁸. If FDA has already recommended a compound-specific or read-across AI, that value takes precedence over the CPCA-derived limit ³⁹⁵.

An applicant whose NDSRI exceeds its CPCA-associated AI should pursue mitigation. To propose a higher AI, the applicant needs a scientifically justified rationale supported by compound-specific safety data or a justified read-across to a suitable surrogate, and FDA may ask for more data ⁶⁹⁸. For enhanced Ames testing, FDA recommends the full OECD 471 strain set, a 30-minute preincubation, and both rat and hamster S9 at 30% ⁶⁹⁹.

Multiple nitrosamines

When more than one nitrosamine is present, the total should generally not exceed the AI of the most potent one. FDA also describes a flexible approach in which each nitrosamine's percentage of its own AI is summed and the total is kept at or below 100% ⁶⁷²⁶⁷³.

Specification thresholds and reporting category

The guidance ties the reporting route to where confirmatory results fall ^405406407408:

Deadlines

For marketed products, FDA asked manufacturers to reassess NDSRI risk by November 1, 2023. It asked them to complete confirmatory testing, submit the required changes, and meet recommended NDSRI AIs by August 1, 2025 ³⁹⁹⁴⁰⁰. Applicants with pending applications should assess risk promptly and tell FDA if testing shows levels above the AI ⁴⁰¹.

Mitigation measures named in the guidance

The guidance names several ways to reduce formation ^573574575576:

• Qualify excipient suppliers to account for nitrite variation, or reformulate with lower-nitrite excipients.
• Add antioxidants such as ascorbic acid, sodium ascorbate, alpha-tocopherol, or propyl gallate.
• Move the formulation microenvironment to neutral or basic pH, for example with sodium carbonate.
• Control precursor amines in the drug substance.

Root causes FDA identifies include nitrite salts reacting with amines under acidic conditions, nitrous acid used to quench azide, nitrite impurities in excipients, and leaching from container-closure systems ^674675677.

Risk assessment expectations in NDA reviews

Advice before submission

At pre-NDA meetings, FDA routinely tells sponsors that the NDA must include a nitrosamine risk assessment.

• Journavx (NDA 219209): FDA said a risk assessment for drug-substance manufacture was required. It also said the drug product should be evaluated for potential nitrosamines and any found controlled at acceptable levels ⁴⁶⁸⁴⁶⁹.
• Zurnai, NDA 218590 (nalmefene autoinjector); phenylephrine HCl in 0.9% NaCl (NDA 216830); Cypsedo (NDA 220482): reviews record the same request. For Cypsedo, FDA explicitly asked the sponsor to consider the NDSRI guidance ^326327338.
• Orladeyo (NDA 219776): FDA called the AI a review issue that required consultation with its Nitrosamine Working Group, and warned it might request more data after submission ³³¹.
• Widaplik (NDA 219423): FDA said it would evaluate the risk assessment together with the proposed drug-product specification after submission ⁴³⁶.

Low-risk assessments FDA accepted without added testing

A documented low-risk conclusion has repeatedly been enough to approve without routine nitrosamine testing.

• Veozah (fezolinetant, NDA 216578): the assessment covered drug substance, excipients, process, and facilities, found no risk of nitrosamines or nitrosating agents, and was accepted ³³⁷.
• Myqorzo (aficamten, NDA 219083): risk of nitrosamine formation was low and there was no significant NDSRI risk, so no additional controls or tests were included ³⁰¹.
• Vykat XR (diazoxide choline, NDA 216665): formation risk was very low, so routine release testing was not warranted ³⁰².
• Rocuronium bromide injection (NDA 217472): risk was low and routine testing for commercial batch release was not required ³⁰³.
• Eliquis Sprinkle (NDA 220073): FDA found a low risk of forming nitrosamine impurities ⁵⁵³⁵⁵⁴.
• Utebzi (NDA 215960): the applicant reported no risk of NDSRI formation in drug substance or drug product ³³².
• Nereus (tradipitant, NDA 220152): the review found no safety concern from individual or total nitrosamine levels ³³⁶.
• VTAMA (tapinarof, NDA 215272): no separate drug-product assessment was submitted. The reviewer did not raise this as an issue, citing the drug-substance assessment and the small amounts of low-amine, low-nitrite excipients in the formulation ³⁰⁴.

Inadequate assessments treated as deficiencies

neffy (epinephrine nasal spray, NDA 214697). FDA identified a potential NDSRI, N-nitroso-epinephrine. The original submission had no adequate NDSRI risk assessment or confirmatory data. FDA classed this as a preapproval deficiency rather than something that could be handled through a postmarketing commitment ^321322323324.

FDA's request had several parts ²⁹³²⁹⁴:

• If the impurity was detected, validated testing of at least three primary batches against the recommended AI at release and through shelf life.
• A control strategy that reliably keeps levels at the AI.
• If levels exceeded the AI, a root-cause investigation and process or formulation changes, or a scientifically justified alternative AI.

The resubmission resolved the issue, and the quality team recommended approval ³²⁵.

Cardamyst (etripamil, NDA 218571). The nitrosamine at issue was N-nitroso-etripamil. FDA did not accept the applicant's justification ^559560563564:

• The read-across analysis supporting the proposed AI was not acceptable.
• A recommended mammalian-cell in vitro mutation test had not been submitted.
• The in vitro metabolism assay was inadequate.
• Batch results showing acceptable levels through shelf life were missing.

These quality and AI-justification deficiencies contributed to a complete response on March 27, 2025. After resubmission, the Office of Pharmaceutical Quality (OPQ) and pharmacology/toxicology recommended approval ⁵⁶⁵⁵⁵⁸.

How acceptable intake limits were set in recent reviews

The numerical ng/day values are redacted in most of these reviews. The pattern still shows that FDA independently verifies CPCA assignments rather than accepting them as submitted. It also shows that read-across claims need supporting mutagenicity and metabolism data.

Control strategies FDA required or accepted

Specifications at release and through shelf life. In the first Brynovin review (sitagliptin oral solution, NDA 219122), FDA found the proposed stability limit inadequate for three reasons ²⁹⁵⁶⁰²:

• The limit exceeded the concentration corresponding to the lifetime AI at the 100 mg/day maximum dose.
• Aged batches exceeded the required limit.
• Release and stability data were insufficient.

FDA asked for release, stability, and end-of-shelf-life in-use stability data from three commercial-scale batches. A revised shelf-life limit then met the AI, supported by 12 months of storage at 2 to 8°C. The later complete-response recommendation for Brynovin concerned a CGMP facility deficiency, not nitrosamines ^296257603605.

Specification testing added during review, with a postmarketing commitment. For Zaynich (cefepime/zidebactam, NDA 220787), a redacted nitrosamine was a theoretical risk in cefepime manufacture. It was not detected in seven drug-substance batches or 19 batches of the approved Cefepime for Injection ⁵²¹. Nitrosamine testing was still added to the drug-product specification ⁵²²⁵²³. Because testing was moving to a new site, FDA obtained a postmarketing commitment to revalidate the method there with an internal reference standard and to submit a method-equivalency report ²⁹⁷.

Precursor control. For escitalopram oxalate capsules (NDA 219130), the applicant limited a precursor impurity to keep the NDSRI below a set percentage of its AI. Confirmatory testing of three registration batches found no detectable NDSRI. FDA judged the assessment and confirmatory data adequate to support the impurity's absence ^300568569570.

Specification-based control. For Lopressor oral solution (NDA 219373), the review found that the proposed drug-product specifications adequately controlled impurities, including nitrosamines ⁶²¹⁶²².

Manufacturing and storage changes. For Kisqali, the applicant proposed several measures ²⁹⁸²⁹⁹:

• An HPLC-MS nitrosamine test and NMT ppm limit in the drug-substance and drug-product specifications.
• Refrigerated storage and a shorter shelf life.
• A two-month room-temperature period after dispensing.

The 10%-of-AI decision rule appears directly in review language. The Wellcovorin supplement review restates that routine testing may be omitted below 10% of AI, and that release and stability specifications are needed when levels fall between 10% and 100% of AI ³⁰⁵.

ANDA reviews

The clearest ANDA example is Laurus's sacubitril/valsartan (ANDA 213676). FDA's deficiency cited concern about nitrosamines, particularly NDMA, in valsartan and other ARB products ^273274688689. FDA asked for:

• A validated method to detect and quantify nitrosamines in the drug product.
• Full method-validation and batch data.
• If an in-house method was used, a one-time comparison against FDA's published headspace method using spiked samples.

FDA initially treated the missing drug-product method as a major deficiency ⁶⁸⁹.

Laurus responded with three lines of evidence ^271275276:

• A formation-risk assessment covering the drug substance, excipients, solvents, packaging, and cGMP cross-contamination controls.
• Testing of every incoming batch of the sacubitril-valsartan complex for NDMA and NDEA. Both were not detected at an LOQ of 0.01 ppm.
• Upstream valsartan controls for NDMA, NDIPA, NEIPA, NMBA, and NDBA.

FDA concluded that the listed nitrosamines could not occur in the finished product and found finished-product testing unnecessary ⁶⁹⁰.

Recent ANDA quality summaries now carry an "NDSRI-impacted ANDA?" field. Sacubitril/valsartan ANDA 213748, risperidone ANDA 214068, liraglutide ANDA 215503, and prednisolone acetate ANDA 216935 were each marked not impacted ^657692693694.

Supplement reviews

Kisqali (NDA 209092/S-018, with the copack NDA 209935/S-027). This is the most consequential supplement case in the record ^333334335357:

• A CPCA Category 3 NDSRI was found in both drug substance and tablets, at levels above the limit applicable to early-stage breast cancer patients, for whom the ICH S9 advanced-cancer flexibilities do not apply.
• An enhanced Ames test was negative.
• A follow-up in vivo MutaMouse study found significant increases in mutant frequency in liver and duodenum.
• FDA placed early breast cancer trials on partial clinical hold in March 2024, and required investigator notification, patient reconsent, and an updated Investigator's Brochure.

In its NATALEE assessment, FDA found no apparent increase in second primary malignancies at that time. It cautioned that cancers from nitrosamine exposure may take years or decades to appear ³⁵³³⁵⁸.

Paxlovid (NDA 217188/S-08). An NDSRI was found in manufactured ritonavir batches. FDA recommended an AI and required that changes to prevent or reduce the impurity be reported under 21 CFR 314.70 ³⁴⁹.

Wellcovorin (NDA 018342/S-015). FDA identified two Category 4 NDSRIs. Because the product had been discontinued and none existed to test, FDA said the risk could not be assessed. Any future marketed product would need a supplement addressing current impurity-control recommendations, including nitrosamines ²⁶⁹³⁵⁹.

Blujepa (gepotidacin, NDA 218230/S-001). A nitrosamine was among eight drug-product impurities that needed qualification. The NDSRI assessment from the original NDA was updated for the new clinical dose, supported by (Q)SAR analysis ³²⁹³⁵⁶.

Tremfya (BLA 761061/S-021). A container-closure extractables assessment found no nitrosamines above the reporting limit in stopper material. This shows nitrosamine screening extending to packaging components for biologics ³⁶⁰.

Practical takeaways for applicants

• Submit a complete NDSRI risk assessment with the original application. FDA has treated its absence as a preapproval deficiency that cannot be deferred to a postmarketing commitment ³²¹³²².
• Expect FDA to check CPCA assignments independently ^250258259. Read-across or alternative AIs need supporting mutagenicity (including mammalian-cell assays) and metabolism data ⁵⁶³⁶⁹⁸.
• Confirmatory data should come from at least three representative or commercial-scale batches and cover release, shelf life, and in-use conditions where relevant ^391293605.
• Where levels approach the AI, likely remedies include process changes, precursor controls, specification limits, and storage or shelf-life restrictions ^440569298.
• A well-documented low-risk assessment can support approval without routine nitrosamine testing ^301302303.