Limit for API having two NDSRI

@ Yosukemino:

If drug substance contains two NDSRI then
Can individual limit be assigned to each NDSRI?
What would be the overall limit for nitrosamine impurities.

Thank you for asking, @sarushail. First, I want you to read our previous discussion.

After discussion, FDA published new guidance for NDSRIs. It includes the following parts.

If more than one nitrosamine impurity (NDSRI(s) and/or small molecule nitrosamine(s)) is identified and the total level of nitrosamines exceeds the recommended AI limit for the most potent nitrosamine in the drug product based on the maximum daily dose, the manufacturer or applicant should contact the Agency. FDA also recognizes that if the recommended AI limits for the individual nitrosamines vary greatly, basing the total nitrosamine limit on the most potent individual nitrosamine limit might not be practical. In such cases, FDA encourages manufacturers to contact the Agency if proposing a nitrosamine limit other than the recommended AI of the most potent nitrosamine impurity detected.


@ Yosukmino: Thank you for your prompt revert.

I am further elaborating my query as- A drug substance has two NDSRIs. Out of that AI for one NDSRI is 400 ng /day and AI for other NDSRI is 100 ng / day.
In this case, what would be AI for total nitrosamine?

Depends on what of the options you use for limit calculation and also depends on the detections to see what is most optimal.

Note that initially you are always starting from the individual limits, only when both NDSRI exceed 10% of their individual limit you would have to go into group limit setting in my opinion.
You can of course proactively start from methods reaching LOQ at least 10 ng/day for both nitrosamines if practical.

Some examples:


For European countries, see:

EMA/409815/2020 Rev.19 Questions and answers for marketing authorisation holders/applicants on the CHMP Opinion for the Article 5(3) of Regulation (EC) No 726/2004 referral on nitrosamine impurities in human medicinal products
pages 16-18

The examples are related to NDMA and NDEA, but the three options may be theoretically applied also for one or more NDSRI.

1 Like

Thanks Wybon for your prompt revert.

Thanks Paliog for explanation

As per EMA:

Hi Waybon,
Thanks for explaination.
You mentioned Option-O (recalculation of cancer risk), can you please further elaborate this?

1 Like

This is just another way to keep openness for the CPCA-AI not being (close to) the real AI (or any other critique on the conservativeness or limitations of the 1:100000 LCR approach. The LCR concept (see also ICH M7 (R2)) and critical considerations on its are also embedded the nice contribution of Dr. George Johnson in the MfE report: Review of NDSRIs in Pharmaceutical Drugs -Pub).

Option 0 is added because I provided CPCA-based examples to match the original context of the question also having a CPCA link.

In case of presence of multiple nitrosamines, if the total nitrosamine risk remains below a theoretical lifetime excess risk of <1:100000, that is an acceptable total risk level, as recognised for example by the EMA guidance other responders have cited. (Again, for a critical review: Dr. George Johnson report).

However, one does not have to automatically assume that a CPCA-based categorical AI as point of departure is an AI that is linked to the 1:100000 LCR in case of one nitrosamine detected (which is the basis for the other calculation options provided in my example). The real AI could be lower or higher than the CPCA categorical AI, meaning that the related cancer risk is higher respectively lower, and that possible alternative limits via the flexible or fixed approach are still evaluatable.

It was thus more a critical preliminary note on how to link CPCA with a LCR, especially considering the design of categorical CPCA AIs is not always fully transparent. (One can assume there is somewhere a 1:100000 LCR consideration in the model, if not model-wise (e.g. 400 ng/day?) then via the 84 molecule AIs for model validation).

An AI corresponding to a 1:100000 LCR originates from dividing the TD50 by 50000
(This is a linear assumption!)
AI (ng/day) = TD50 (µg/kg/day) x 50 kg/50.000
So CPCA 100 ng/day would assume a TD50 of 100000 µg/kg/day.
If you have a robust indication of the real TD50, you can re-evaluate that against what corresponds to the TD50 for 1:100000 LCR, a significant factorial difference would mean an impact on the targeted LCR and thus a need for revaluation of the individual AI for departure.
(A similar consideration can be made when talking in terms of LCR on the individual guided AIs of course.)

1 Like