Determining Recommended Acceptable Intake Limits for N-nitrosamine Impurities in Pharmaceuticals: Development and Application of the Carcinogenic Potency Categorization Approach
Background
N-nitrosamine impurities are a class of potentially cancer-causing chemicals that can form during manufacture or storage of a drug. Uncertainty about the presence and acceptable amounts of nitrosamines in drug products has raised regulatory challenges and has led to some drug applicants and manufacturers conducting additional studies or, in some cases, discontinuing drug products from the market. Identification of nitrosamines in drug products has also led to disruptions in supply and access, sometimes resulting in drug shortages.
Determining an acceptable intake (AI) limit (the acceptable amount of an impurity in a drug that is assessed as posing no discernible increased risk in cancer due to daily exposure to the impurity) can often be difficult due to the limited availability of safety data for the impurity. This is particularly challenging for nitrosamine drug substance-related impurities (NDSRIs), a subset of nitrosamine impurities that have more complex structures that include a nitrosamine group, although they share similarity with the drug itself. NDSRIs are generally unique to each drug and, therefore, there are usually no existing safety data available to establish AI limits.
When safety data are not available for a nitrosamine impurity, information from nitrosamine comparators (structurally similar compounds) has been used in some instances to identify AI limits. However, there are many cases where an appropriate comparator is not available. Previously in these cases a conservative default AI limit was applied, but the default AI limit posed practical challenges to both industry and regulators, significantly impacting drug supply chains.
To that end, FDA and international regulators developed a methodology, the Carcinogenic Potency Categorization Approach (CPCA), that uses the chemical structure of a nitrosamine impurity to recommend AI limits by assignment to 1 of 5 predicted potency categories reflecting carcinogenic risk. The CPCA is based on knowledge that the α-hydroxylation mechanism of metabolic activation is responsible for the potent carcinogenic response observed for many nitrosamines. Therefore, the number and distribution of α-hydrogens combined with other activating and deactivating features (molecular substructures that are associated with an increase or decrease, respectively, in carcinogenic potency) in the nitrosamine are considered.
Materials and Methods
A group of 81 nitrosamines was assembled to serve as a training set to identify five potency category (PC) thresholds with four unique corresponding AI limits. The training set (and the CPCA) was limited to N-nitroso-compounds with a carbon atom directly bonded to both sides of the N-nitroso group, and where the carbon atom is not directly double-bonded to a heteroatom (i.e., N-nitrosamides, N-nitrosoureas, N-nitrosoguanidines and other related structures are excluded). Additionally, nitrosamines where the N-nitroso group is bonded to a nitrogen within an aromatic ring (e.g., nitrosated indole) were excluded. These excluded classes undergo metabolic activation by a different biological pathway than typical nitrosamines.
Activating and deactivating structural features of nitrosamines were then compiled from scientific publications, as well as through manual review of nitrosamine structures to identify additional patterns that affect carcinogenic potency. These α-hydrogen features and other structural features were assigned relative weights based on visual inspection of matching training set structures and their associated carcinogenic potency data, chemical reactivity, metabolism, and other evidence from the published literature. Scores (weights) of individual structural features (ranging from -1 to 3) reflect the magnitude of the effect, where the higher the score, the stronger the deactivating effect of the feature.
A potency scoring scheme for the whole molecule was developed based on the sum of the individual scores of matching features: the higher the total potency score, the lower the predicted carcinogenic potency.
Potency Score = α-Hydrogen Score + Deactivating Feature Scores + Activating Feature Scores
The total potency score could then be used to assign a nitrosamine to a PC with a corresponding AI limit based on the established thresholds.
The count and distribution of α-hydrogens on each side of the N-nitroso group were encoded as structural features in the format “n,n” where the lowest number of hydrogen atoms is listed first.
The other structural features considered in the calculation of the score were:
- tertiary α-carbon atom (deactivating)
- carboxylic acid group (deactivating)
- N-nitroso group in a 4–7 membered ring (deactivating)
- chains of ≥5 consecutive non-hydrogen atoms (deactivating)
- electron-withdrawing group (EWG) bonded to the α-carbon (deactivating)
- β-hydroxyl group (deactivating)
- benzylic group bonded to the N-nitroso group (activating)
- methyl group bonded to the β-carbon (activating)
Results
Structural features included in the CPCA are those that have been identified as directly affecting the α-hydroxylation pathway and, consequently, carcinogenic risk. Therefore, the combined effect of these features is used to generate a prediction of carcinogenic potency of a nitrosamine based on its chemical structure without the need for compound-specific safety data or a comparator (Table 1).
Table 1. Model structural features, their relative carcinogenic potency or activating/deactivating effect, and their CPCA individual feature score or PC (for features that result in automatic assignment to PC 5 by the CPCA flow chart).
A) α-Hydrogen features in the format “n,n” where the lowest number of hydrogens is listed first
| α-Hydrogen feature | Example structure | Relative carcinogenic potency (α-Hydrogen Score or PC) |
|---|---|---|
| 0,0 | 
| Low (PC 5) |
| 0,1 | 
| Low (PC 5) |
| 0,2 | 
| Medium&% (3) |
| 0,3 | 
| Medium& (2) |
| 1,1 | 
| Low (PC 5) |
| 1,2 | 
| Medium (3) |
| 1,3 | 
| Medium (3) |
| 2,2 | 
| High (1) |
| 2,3 | 
| High (1) |
| 3,3 | 
| High (None - NDMA) |
B) Deactivating features
| Deactivating Feature | Example Structure | Deactivating Effect (Feature Score or PC) |
|---|---|---|
| Tertiary α-carbon | 
| Very Strong (PC 5) |
| Carboxylic acid group anywhere on molecule | 
| Strong (+3) |
| N-nitroso group in a pyrrolidine ring | 
| Strong (+3) |
| N-nitroso group in a 6-membered ring containing at least one sulfur atom | 
| Strong (+3) |
| N-nitroso group in a 5- or 6-membered ring* | 
| Medium (+2) |
| N-nitroso group in a morpholine ring | 
| Weak (+1) |
| N-nitroso group in a 7-membered ring | 
| Weak (+1) |
| Chains of ≥5 consecutive non-hydrogen atoms (cyclic or acyclic) on both sides of acyclic N-nitroso group. Not more than 4 atoms in each chain may be in the same ring | 
| Weak (+1) |
| Electron-withdrawing group** bonded to α-carbon on only one side of N-nitroso group (cyclic or acyclic) | 
| Weak (+1) |
| Electron-withdrawing groups** bonded to α-carbons on both sides of N-nitroso group (cyclic or acyclic) | 
| Medium (+2) |
| Hydroxyl group bonded to β-carbon*** on only one side of N-nitroso group (cyclic or acyclic) | 
| Weak (+1) |
| Hydroxyl group bonded to β-carbons*** on both sides of N-nitroso group (cyclic or acyclic) | 
| Medium (+2) |
C) Activating features
| Activating Feature | Example Structure | Activating Effect (Feature Score) |
|---|---|---|
| Aryl group bonded to α-carbon (i.e., benzylic or pseudo-benzylic substituent on N-nitroso group) | 
| Weak (-1) |
| Methyl group bonded to β-carbon (cyclic or acyclic) | 
| Weak (-1) |
&Activity is very low if “0” hydrogen count corresponds to a tertiary α-carbon.
%If methylene α-carbon is part of a terminal ethyl group, a score of 2 should be applied.
*Excludes examples where N-nitroso group is in a pyrrolidine ring, a 6-membered ring containing at least one sulfur atom, or a morpholine ring (all counted separately).
**Excludes carboxylic acid and aryl (counted separately), and ketone (conflicting data). Additional electron withdrawing group examples are limited to those described by Cross and Ponting (2021), where they are referred to as “β-carbon electron withdrawing groups.”
***β-Carbon must be in an sp3 hybridization state for this feature to apply.
To operationalize the CPCA, a flow chart (Figure 1) was developed to assess nitrosamine structures to determine if a potency score needs to be calculated and to assign a nitrosamine to a PC. The flow chart first checks for the presence of features that are predicted to result in a lack of activation or lack of reaction with DNA through the α-hydroxylation pathway (meaning low cancer-causing potential). The 0,0 α-hydrogen feature lacks α-hydrogens for metabolic activation, the 0,1 or 1,1 α-hydrogen features strongly disfavor metabolic activation, and the tertiary α-carbon feature results in detoxification by water. The presence of any one of these features automatically results in assignment of the nitrosamine to the least carcinogenic PC (5). For nitrosamines without these features, the flow chart directs the user to the “Calculate Potency Score” step, which then translates to a PC with associated recommended AI limits ranging from 18 ng/day or 26.5 ng/day2 (most potent, lowest AI limit) up to 1500 ng/day (least potent, highest AI limit).
The ability of the CPCA to accurately predict the carcinogenicity of nitrosamines was tested and shown to be able to appropriately identify nitrosamines of the highest risk.
Figure A) CPCA flowchart to predict the carcinogenic potency category of a nitrosamine structure and its associated recommended AI limit. *AI limit of 18 ng/day or 26.5 ng/day depending on regulatory region.
Discussion
FDA co-led the development of the CPCA, which was developed in 2023 in collaboration with the Nitrosamine International Technical Working Group (NITWG), a consortium of international drug regulatory authorities. Now published in the nitrosamine guidances of several NITWG member organizations including FDA, the CPCA considers the relative activating and deactivating effects of multiple features in nitrosamines and uses this information collectively to generate a prediction of carcinogenic potency and recommend an AI limit. The CPCA is easy to use and can be rapidly applied without the need for specialized software. The CPCA draws on a larger body of carcinogenicity data than determining an AI limit based on a single comparator and it allows an AI to be determined when no safety data for the impurity are available. Since it uses only a chemical structure, it can be applied to both known and hypothetical nitrosamine impurities. Importantly, it offers applicants and manufacturers a path forward that does not require generating compound-specific data using toxicology studies, which was a significant limitation in the past.
The application of the CPCA to determine recommended AI limits for nitrosamine impurities has expedited the regulatory review of drug safety assessments submitted by drug companies as an AI limit can be identified more efficiently and with greater transparency and predictability.
Further development of the CPCA is expected as new experimental studies help identify additional features and properties that may be appropriate to include. CDER continues to work with international regulators in the development and use of this modeling approach to identify recommended AI limits for nitrosamines and to more effectively address the potential impact of these impurities on the drug supply.
Other resources
- Recommended Acceptable Intake Limits for Nitrosamine Drug Substance-Related Impurities
- Control of Nitrosamine Impurities in Human Drugs
1Kruhlak, NL, Schmidt, M, Froetschl, R, Graber, S, Haas, B, Horne, I, Horne, S, King, ST, Koval, IA, Kumaran, G, Langenkamp, A, McGovern, TJ, Peryea, T, Sanh, A, Siqueira Ferreira, A, van Aerts, L, Vespa, A, & Whomsley, R, 2024, Determining recommended acceptable intake limits for N-nitrosamine impurities in pharmaceuticals: Development and application of the Carcinogenic Potency Categorization Approach (CPCA). Regul Toxicol Pharmacol, 150:105640. https://doi.org/10.1016/j.yrtph.2024.105640
2For products intended for marketing in the United States, FDA recommends an AI limit of 26.5 ng/day for Potency Category 1, even if a different limit is recommended in other regulatory regions.