Indeed, in the light of LTL principles for a temporary limit, MW corrected readacross to NNK or the AI based on PhysChem readacross or QM modelling and uncertainties, 600 ng/day t-AI might not sound too conservative. However, this indeed mostly disregards the in vivo mutagenicity data in BigBlue rats for N-nitroso-duloxetine:
N-nitroso-duloxetine:
• 6.8 mg/kg/day/50000 50 kg1000000 ng/mg = 6800 ng/day (Jolly 2024, in vivo TGR, liver BMDL, with justification to derive AI based on a mutagenicity endpoint)
• 6.8 mg/kg/day/0.21 mg/kg/day x 145 ng/day = 4695 ng/day (relative to NDMA)
• 6.8 mg/kg/day/0.1 mg/kg/day x 26.5 ng/day = 1802 ng/day (relative to NDEA)
• Overall Jolly 2024 conclusion based on in vivo data: >5000 ng/day
Strangely, this data has not been publicly endorsed yet by regulators. The studies seem to have been properly designed, including tissue selection and 3 compounds of the class have been tested showing a coherent picture, while a broader WoE on the compounds was presented. Ames test data and in vitro mammalian cell mutation assay data from other sources also point to a likely alpha-hydroxylation mechanism.
One of the general topics here is the in vivo mutagenicity vs. in vivo carcinogenicity correlation (which is the basis for calculations 2 and 3). The fact that this correlation has been built for a set of small nitrosamines now strangely leads to challenges on if this correlation would be true for NDSRIs as well. In absence of carcinogenicity data on a lot of NDSRIs this cannot be easily shown.
For small nitrosamines non-mutagenic carcinogenicity or mutagenicity non-carcinogenicity might indeed be rarer than for NDSRIs, which often have increased structural complexity and thus more theoretic options for alternative mechanisms. Possibly the quality of the correlation is therefore a little bit lower if it would be built for NDSRIs, here and there an NDSRI might be off the curve. But one should question if this can be a relevant blind spot, as there are only two out of curve situations possible (while actually the exceptions would influence the quality of the correlation and not there being a correlation):
- Either you overestimate the carcinogenicity relative to the mutagenicity of the NDSRI, meaning that AI setting based on a mutagenicity endpoint is sufficiently conservative (no impact in case of slightly lower quality of the correlation).
- Either you underestimate the carcinogenicity relative to the mutagenicity of the NDSRI, but that would be an indication of an non-nitrosotypical carcinogenicity occurring at least as part of the mechanism, meaning that the lack of (full) correlation muta-carci has nothing to do with the CoC principle on nitrosamines. Based on typical cancer mechanisms and the structure of the NDSRIs, I don’t think it would be often plausible to have a theory that such unidentified alternative cancer mechanisms would likely not be compatible with application of a 1500 ng/day TTC, which is fairly low. (And there is also the fact that mammalian cell line experiments point indirectly to there being a nitroso-typical mechanism as well.)
So I don’t understand what prevents to apply the Powley 2024 idea that 1500 ng/day is precautious for NDSRIs with such lower potency from in vivo mutagenicity studies, as it seems unlikely the aimed LCR cannot be maintained that way (especially when there is also a LTL element due to temporary application while science advances). Cutting the 6800 ng/day in 10 to 600 ng/day seems a little bit too conservative (especially for temporary use), good if you can reach it, but if it is leading to recalls and possible shortages I don’t know if it is pragmatic enough. (Remediation might also not be too easy given that the 600 ng/day introduced in February 2024 was prolonged from ending October 2024 to ending December 2024, linked to a shortage justification).
For this class of the oxetines, in vitro experiments have given some suspicion on cytotoxicity effects of the NDSRIs that are probably being investigated further, but I still believe this rationale is possible.
I hope the 31 December 2024 end of the t-AI also coincides with regulatory advancements on AI setting from in vivo mutagenicity data and that it is not prolonged to give time for in vivo carcinogenicity studies (though the cytotoxicity observations are probably explored).