Nitrosamine impurities in medicines (presented at the ARCS Annual Conference 2024)

Dr. Helen Hughes from TGA presented at the ARCS Annual Conference.

I am interested in this slide. Nitrosamine A is likely to have the same molecular weight as N-Nitroso-Fluoxetine. I could not find a different surrogate with a TD50 of 0.537mg/kg/day.


very interesting case Yosuke and characteristic of the confusing situation which remains even after the adotping of CPCA approach almost globally.
Anyway, i think that Sponcor C request could not be denied and in the next step TGA should updated the limit adopting the AI of 5000ng/day.
Thanks for sharing
best regards

1 Like

But what limit did the TGA accept?? I couldn’t see it in the presentation.



May I ask why it should not be denied? As of now and is clearly indicated in the ICH M7, mutagenicity cant define limits for carcinogenicity end-points. Industry is generating data to see if there is a correlation between carc va muta but that may take still some time until is fully diggested the topic by regulators (and if the correlation actually exist).


Hi Diego,
well your point is acknowledge as we usually stated to the assessors :blush:
Of course you are right about the relation between mutagenicity and carcinogenicity.
My state was based in EMA Q and A guideline (EMA/409815/2020 Rev.20) where it is note that (page 16): ‘‘A negative result in a relevant well-conducted in vivo mutagenicity study can allow control of the N-nitrosamine as a non-mutagenic impurity (NMI), i.e. according to ICH Q3A(R2) and ICH Q3B(R2) limits’’
Do you want to share with us your choise between those 5 approaches?
thanks for your question
kind regards


Thanks for your reply, as my personal opinion I would actually say a Weigh of Evidence approach with Sponsor B and Sponsor C proposal, but to move the AI limit to Cat 3 for example. I like Sponsor C approach but 50 times fold increase I do not see realistically possible right now. Considering how conservative regulators may be.

1 Like

In the FDA/HESI meeting in 2023 a lot of attention was given to WoE approaches and the importance of mechanistic understanding of the metabolism (also for designing in vitro or in vivo mutagenicity studies). We are still seeing too little in vitro metabolisation assays or QM mechanistic studies in new literature or appreciation of such understanding in AI setting, would be great if the fluoxetine case would get published in literature (so that B and C get connected somehow). Yet AI setting often supposes nitrosamine selectivity or little risk for false positives (out of precaution, but if the data is there…).

N-nitroso-fluoxetine (MW relative to NNK MW is 1.633) is a great example here that in simplified test batteries you always have to try to understand what you are seeing (and in general interpretation of in vitro data including aromatic structures is often more difficult):

  • Enhanced Ames test: positive (–> 30% hamster S9 and WP2 uvrA pKM 101) (FDA)
  • In vitro mammalian cell gene mutation assay: positive (FDA)
  • In vitro micronucleus test without S9 (with human lymphoblastoid TK6 cells): positive (FDA)
  • In vitro micronucleus test with S9 (with human lymphoblastoid TK6 cells): positive (FDA)
  • Human CYP activation: CYP2C19, 2B6 (FDA)
  • In vitro Comet assay in primary rat hepatocytes (to try to do potency ranking with BMD approach): higher DNA damage N-nitroso-fluoxetine compared to NDMA, NDELA; N-nitroso-fluoxetine is cytotoxic (17 µM (24 h)) (Mutamind)

But then metabolic study with quantum modeling: probably identifying ether related metabolites / hydroxy-substituted or related small nitrosamines as suitable surrogate for SAR / dominant denitrosation as toxification : this might not be as potent….
And then the in vivo test (validating the quantum mechanical data?!)… it is 30 to 50 times less mutagenic than NDMA.
(The SOT 2024 poster of FDA also in general did suggest in vivo mutagenicity data as follow-up, so I don’t see why 5000 ng/day wouldn’t be justifiable if a properly conducted study is available. When doing the BMD that would be done with regards for carcinogenicity/mutagenicity correlation (and then again all the other available data (except readacross) also is focused on mutagenicity).

In vivo priority over in vitro, but in vivo should stimulate mechanistic or toxification/detoxification balance interest?

Sponsor B was definitely on to something.


Dear all,

Thank you for sharing your insights. I asked whether TGA accepted those approaches through LinkedIn. The answer was that it is a hypothetical example demonstrating that there are many ways to derive an acceptable intake (AI) for nitrosamine impurity. Different approaches (such as EAT or in vivo test) in TGA guidance may be appropriate. Anyway, no further information was available. We have to wait for a time before some new mitigations will be published.


I think it’s quite appropriate to call here the publication from Schlingemann et al. @schlinjo1975 where all these points are discussed in detail.


It has now been clarified that this is based on N-nitroso N-methyl dodecylamine as a surrogate.

The quantumchemical work was linked to calculating the C-N stretching frequencies of the diazonium metabolite of the NDSRI and surrogates and correlating with published TD50 data, allowing to orient N-nitroso-fluoxetine in the spectrum of the correlation, suggesting a TD50 of 1.14 mg/kg bw.

Also Ames test data is presented, next to the in vivo mutagenicity data supporting 5000 ng/day.

Sponsor B and sponsor C are the same company it seems (Eli Lilly).

Conclusions for AI N-nitroso-fluoxetine:

  • 537 ng/day based on Phys Chem readacross
  • 1140 ng/day based on QM modeling (taken bij TGA under the above point)
  • More than 5000 ng/day based on TGR study

Estimation of Acceptable Daily Intake Values based on Modeling and In Vivo Mutagenicity of NDSRIs of Fluoxetine, Duloxetine and Atomoxetine - ScienceDirect


so, apparently it was not so ‘‘hypothetical’’ example as the TGA answered to Yosuke.
Thank you very much ccdw

Hi Wybon,

Thank you for the clarification. The paper you shared is interesting. The comparison of Molar correction AI, read-across, quantum mechanical modeling, and BMD approach is worth reading. The potency of N-nitroso-fluoxetine does not look so high compared to NNK.

the paper is actually fantastic…and i think that the correlation of the stretching frequencies of the C-N bond of the diazonium intermidiates to TD50 is brilliant and very right to the point…thanks again Wybon

1 Like