New paper - Use of molar, instead of weight-based, safety limits

I’ve said a few times on here and in talks “watch this space” for more evidence for the use of molar limits for nitrosamines - the paper is now out!

This should be particularly relevant for scaling read-across analogues, often far larger molecules than the compounds they are read-across from…


The authors analysis of the chemical and manufacturing aspects, emphasizing the difficulties in meeting non-MW-corrected AI criteria compared to the practicality of MW-corrected criteria compliance, is thought-provoking. This is an important topic that warrants more attention and discussion in our field.


Challenging even the CPCA limits - I like it!! :smiley:

I’ve always argued that molecular weight needs to be taken into consideration, for exactly the reasons here, great to see it published - congratulations to all.


I calculated 163pmol/day of Category 1 nitrosamines in EMA and FDA AI list from their Mw. The correction makes the AI higher from 1.5 to 5 times. It’s practical!!

In EMA AI list;
N-nitroso orphenadrine: 46.4ng/day
N-Nitroso Betahistine: 26.9ng/day
N-nitroso-desmethyl-chloropyramine: 49.7ng/day
N-nitroso-desmethyl-tripelennamine: 44.1ng/day

In FDA AI list (20 compounds from 47 compounds);
N-nitroso-nortriptyline: 47.7ng/day
N-nitroso-desmethyl-imipramine: 48.1ng/day
N-nitroso-desmethyl-almotriptan: 57.1ng/day
N-nitroso-desmethyl-amitriptyline: 47.7ng/day
N-nitroso-atomoxetine: 46.3ng/day
N-Nitroso Desmethyl Bedaquiline: 93.0ng/day
N-nitroso-desmethyl-brompheniramine: 54.5ng/day
N-Nitroso Desmethyl Cabergoline: 76.1ng/day
N-nitroso-desmethyl-carbinoxamine: 49.8ng/day
N-nitroso-desmethyl-chlophedianol: 49.7ng/day
N-Nitroso N-Desmethyl-Chlorpheniramine: 47.2ng/day
N-nitroso-desmethyl-chlorpromazine: 54.4ng/day
N-Nitroso Desmethylcitalopram: 55.3ng/day
N-nitroso-desmethyl-clomipramine: 53.8ng/day
N-nitroso-desmethyl-cyclobenzaprine: 47.3ng/day
N-nitroso-desipramine: 48.1ng/day
N-nitroso-desmethyl-desvenlafaxine: 45.4ng/day
N-nitroso-desmethyl-dexbrompheniramine: 54.5ng/day
N-nitroso-desmethyl-dexchlorpheniramine: 47.2ng/day
N-nitroso-desmethyl-diltiazem: 70.0ng/day


It is an improvement value-wise, but probably often not the preferred option compared to other AI derivatisation options. Readacross would also allow MW-corrections, so the magnitude of difference between initial AI and MW corrected AI does not differ in the end, meaning that the MW correction unfortunately doesn’t lower the relative gap between readacross and CPCA e.g.:
N-nitroso-fluoxetine: 100 ng/day → 297 ng/day (via MW correction and ICH M7 most robust study (Rivenson 1988), most sensitive tumor site/tissue approach) = factor 3 correction
N-nitroso-fluoxetine: CPCA cat. 1 18 ng/day → 55 ng/day = factor 3 correction (when making abstraction of 26.5 ng/day CPCA cat. 1 in US)

I’m also wondering whether the CPCA MW-rules would be expected to be capped (just like Q&A 22 capping on LTL). Could you, due to very high MW (exceptional though), jump categories and have a MW-corrected CPCA cat. 1 above 100 ng/day? What if MW corrections can be developed for other categories, if and how to cap?

Probably for CPCA category 2 we can also already think about a MW correction mechanism and category improvement (but this should probably go hand in hand with optimization of (de-)activating features and industry verifications on the CPCA design).

CPCA category 1:

  • 18 ng/day for MW <120 Da

  • MW >120 Da: Fine 2023 rule: 163 pmol/day x MW nitrosamine (ng/pmol) = AI (ng/day) (with or without a capping at 100 ng/day?)

CPCA category 2:

  • 150 ng/day based on NDMA (145 ng/day) and NNK (182 ng/day) readacross.

  • MW > 200 Da: 880 pmol/day x MW nitrosamine (ng/pmol) = AI (ng/day) (with or without a capping at 400 ng/day?) (Capping here should mean extra capping on Q&A might be a bridge too far?)

CPCA category 3:

  • To be evaluated when transparency on design of CPCA and 84 molecules and their MW?

  • Piperidine and pyrrolidine corrections to be optimized?

  • MW corrections?

  • AS IS: N-nitroso group in a 5- or 6-membered ring: CPCA score + 2

  • TO BE: N -nitroso group in a 5- or 6-membered ring not identifying as a piperidine or pyrrolidine: CPCA score +2; N -nitroso group in a piperidine or pyrrolidine ring: CPCA score +3

CPCA category 4-5

  • 1500 ng/day

  • To be evaluated when transparency on design of CPCA and 84 molecules

  • MW corrections incompatible with TTC concept (category 4)

  • Finetuning of (de-)activating factors?


Is this Nitrosamines Limits 3.0 ? @trust_user_c @trust_user_b @trust_user_a


Nitrosamines limits will have more versions as with other GTIs. The more the data and confidence and the more the refinement!


Thank you for sharing! Regards.

Dear Yosukemino
The limit for NDSRI of Chlorpromazine in FDA list 24 ng/mL , but you have mentioned as 54.4 ng/mL.Would ypu please reference of the same. It would be of great help.

Thank you for asking, @RCREDDYP. The molecular weight of N-Nitroso Desmethyl Chlorpromazine is 333.83 g/mol.

163 pmol/day = 163 x 333.83/ 1000 = 54.4 ng/day

The limit of N-Nitroso Desmethyl Chlorpromazine is 26.5 ng/day(Cat.1). The corrected limit is two times higher than the FDA limit.

Appreciated your clarification. But , the same is not reflecting in FDA website.

@RCREDDYP The discussion here is about new approaches proposed to assess and establish limits, no regulatory agency has expressed in their guidance document the use or acceptability of these approaches. i.e. You will not find it referenced in the FDA or EMA

1 Like

We chemists have been trying to demonstrate that the AI of nitrosamines or NDSRIs is too strict. And CPCA is still considered conservative. The AI of some nitrosamines is unrealistic for measurement. The new approach is one of the next steps. We are going toward the good ways, but it takes more time…

1 Like

Nonetheless, compounds in Category 1, still with the conversion factor are way conservative in some cases. More like the 18 ng/day or 26.5 ng/day is still a bit incompatible with the approaches industry, etc are proposing.

Plus something I read on Linkedln, but I am not the one to comment if could be a constrain with the methodology is: “If we plan to use NGRA, then cellular perturbation can be taken into consideration and the level further goes down or modifies.”

At the end is a summary of different proposals that will make this topic more realistically manageable.

1 Like

NGRA can only be used a weight of evidence call for now on nitrosamines.

The challenges with current NGRA application is- we may also end up with more conservative limits which are actually not required because physiological response manifestation is different from cellular perturbation.

I think we will definitely see some strategy coming up for class-I NDSRIs because practicality remains tough job in controlling them.


No guidance, but in the European context, it has come up a few times on the discussion table in meetings with Industry and the Agency since 2022. On the request to comment on the Fine 2023 paper, the NS OEG confirmed in December 2023 that international discussions continue and have stumbled on the molar thinking not been implemented on the experimental side as well:

MW adjustment is an issue that has cropped up ever since we started using some of the smaller nitrosamines as surrogates with the readacross approach and you know we have been considering it with other international regulators. But the principal argument against using these adjustments is that regulatory guidelines on toxicology studies and impurities, have always been considered based on weights and not other quantities and if we now start applying this MW adjustment on nitrosamines: just for the comparative assessment we would also have to think about perhaps extending it to Ames testing using molarity and in vivo studies as well (Ames: adjusting the reference doses to be applied to higher levels tested in case of higher molecular weight compound that is studied), and then higher levels of NDSRIs could need to be tested in Ames or in vivo mutagenicity testing, but we do have the CPCA: the readacross approach would probably rarely be used because we only have about 45 nitrosamine surrogates where we do have robust carcinogenicity data and it’s very rare that one of those could be used as a good comparator for an NDSRI.

→ Regulators perspective that if you use molecular weight corrections surrogate to NDSRI in CPCA and readacross, you also have to think about implementing the molecular weight correction in designing in vivo and vitro studies protocols, meaning a re-evaluation of the OECD 471 guidance on Ames testing? For example the Ames OECD 471 maximum plate concentration of 5000 µg/plate should then be defined as X µmole/plate (and thus be translated to µg/plate by multiplying with the molecular weight of the test compound). This means larger doses for NDSRIs with higher MW are to be applied (increasing testing cost, whereas background lawn clearance (toxicity) might anyway be seen for higher plate concentrations), whereas this also means an excellent purity of the NDSRIs will be needed to avoid misinterpretations, e.g. cf. Lijinsky recommendation:
Although it was possible - and was observed in some cases - that much larger quantities of nitrosamine than 1 mg could give rise to a number of revertants slightly above the minimum for significance, this cannot be considered acceptable procedure. The difficulty of excluding the presence of an undetectable trace of strongly mutagenic impurity in several milligrams of nitrosamine is so severe, that the significance of such marginal responses is moot. Therefore, the highest dose we used was 1 mg Redirecting.

The change in thinking in terms of molarity for mutagenicity evaluation from the perspective that 1 mole of a mutagenic function like nitrosamine (1 mole of mononitrosated compound) can lead to one mole of DNA alkylation should be done holistically, not only in readacross, but also in the toxicology study guidance: we fully think in terms of moles in mutagenicity or we don’t?

But then again, the nitrosamine saga isn’t over yet, shouldn’t we be more pragmatic?
As Industry commented to the EMA:
Over the years we’ve used milligram per kilogram body weight thinking in toxicology, but the toxicological data that was available was for a vast range of molecular weight of compounds. (So by studying classes of certain compounds, there was nothing wrong with thinking in mg/kg body weight because we had data for a certain class of compounds with sufficient variability in molecular weights and/or representative molecular weight for what we actually wanted to do readacross on). Nitrosamines as we know are different: we have only carcinogenicity data for very small nitrosamines (low MW, often dialkylnitrosamines), but we are studying higher MW nitrosamines (high MW) and are being confronted with the need to extrapolate low MW data to high MW applications, which we might not per se have seen when studying other mutagenicity cases (so the comparability can away not be fully restored). We just don’t have the data, but from understanding the mechanism that it is really stoichiometrically dependent mutagenesis that causes the carcinogenicity: the molecular weight argument is strong.

On a positive note: no MW correction capping discussions were confirmed by regulators, so I still believe that the caps I started to propose above can be arguable and visualise the thinking that MW corrections are needed to close important MW gaps for higher MW NDSRIs and can be verified with examples (e.g. NNK to NDMA readacross) to show the risks that the MW-corrected SAR-based AI is higher than the real AI. When the readacross is well justified and the nitrosamine is reasonably high in MW, an application of a capped MW correction as a next step should be possible?


Thanks @ccdw for expressing your thoughts and sharing the discussed points on MW correction from NS OEG.

I have following views as a regulatory toxicologist:

  1. I agree molarity consideration is more apt for mechanistic understanding rather than mass when we discuss mutagenicity- But the fact is, this is not applicable for only mutagenicity- if we plan to apply- then it is applicable to all contexts of physiological action of drugs and their respective impurities.

  2. The debate around molarity versus mass is not new. It was there and it will be there. Most drugs to patients are sensed as mass and there are few exceptions. Interestingly endogenous substances are mostly dealt in molar.

  3. Definitely NDSRIs should never be treated as small alkyl/aryl nitrosamines. To address this gap we all make several efforts, few among these are molarity and molecular weights corrections. But the fact to be taken into consideration is generally how we communicate dose strength units when it is given to patient and how we develop drugs does really matter.

  4. As per my understanding still representation in “mass - (mg/kg or mg/day)” holds more pragmatical representation because that’s how we see the patient gets exposed to drugs or other components in daily life and evolution of drug development practices.

  5. In regard to context that was mentioned- application of molarity to toxicological studies particularly Ames assay and to comment for most of the toxicity studies- I see a great practical challenge in application, that would compromise the purpose of the study.

  6. Coming to mutations- it is always a double edged sword because there are arguments one mutation is enough to cause de-railment for cell and another argument of minimum 6-8 somatic mutations required to manifest an onco environment. However, evidence demonstrates most of genotoxic mechanisms obey dose -response, threshold based effects and several other physiological factors.

Last not but least- I completely welcome this debate because, irrespective of above challenges, I foresee these debates might trigger regulatory agencies at one point to think for application of some common factors to bridge the stochiometric differences between small nitrosamines and complex NDSRIs.