🇺🇸 FDA - Updated information on recommended AI for NDSRIs (23/Feb/2024) - Tables 2 & 3

FDA updated information on recommended AI for NDSRIs (23/Feb/2024), as follows:

Table 2: FDA Recommended AI Limits for Certain NDSRIs - Based on Compound-Specific Data or Read-Across Analysis from a Surrogate

N-nitroso-bumetanide → Added Testing Method for the Determination of N-nitroso-bumetanide

N-nitroso-propranolol → Added Testing Method for the Determination of N-nitroso-propranolol

N-nitroso-varenicline → Added Testing Method for the Determination of N-nitroso-varenicline

Table 3: Recommended Interim AI Limits for Certain NDSRIs for Approved or Currently Marketed Products

N-nitroso-duloxetine → Added to Table 3


An important update from FDA on NDSRIs was posted last 23 Feb 24.

Table 3. Recommended Interim AI limits for certain NDSRIs for approved or currently marketed products

NDSRI Name Source Recommended Interim AI Limit (ng/day) Recommended Interim Control Limit (ppm) Estimated Duration***
N-nitroso-duloxetine Duloxetine 600 ng/day 5 ppm 10/1/2024

*Note that an AI limit can be converted into a parts per million (ppm) control limit. The conversion varies by product and is calculated based on a drug’s maximum daily dose (MDD) as reflected in the drug labeling (ppm = AI (nanograms (ng))/MDD (milligrams)).
**A decision to release lots remains solely the firm’s responsibility. Firms are responsible for ensuring that their drugs are manufactured in compliance with all applicable requirements, including CGMP, and FDA expects them to vigilantly monitor and promptly report to FDA any adverse drug experiences or other findings that may affect product quality or safety.
***“Estimated Duration” is the date by which FDA intends to reassess the recommended interim control limit. As indicated in the NDSRI guidance, while FDA recommends conclusion of NDSRI confirmatory testing of drug products and submission of required changes in drug applications by August 1, 2025 for approved or currently marketed products, FDA may recommend that firms complete an expedited risk assessment, confirmatory testing, or other regulatory action based on information available to the Agency. With respect to FDA-recommended interim control limits, FDA generally does not intend to object to the distribution by manufacturers and applicants of drug products that contain NDSRI levels at or below the recommended interim control limit, until the date identified in the “Estimated Duration” column. However, if FDA becomes aware that the shortage risk has been alleviated, FDA may provide additional notification or may reconsider its recommended interim control limit.

Recommended Testing Methods for Confirmatory Testing of Certain NDSRIs

FDA intends to provide information on FDA-generated testing methods to provide options for regulators and industry to detect NDSRI impurities in specific drug substances and drug products. The methods should be validated by the user if the resulting data are used to support a required quality assessment of the drug substance or drug product, or if the results are used in a regulatory submission.

LC-ESI-HRMS Method for the Determination of N-Nitroso-Bumetanide

LC-ESI-HRMS Method for the Determination of N-Nitroso-Propranolol

LC-ESI-HRMS Method for the Determination of N-Nitroso-Varenicline

@trust_user_a @trust_user_b @trust_user_c @trust_user_d
Thanks @E_da_Silva for the post.


Hi @Naiffer_Host thank you for sharing this news, I hope that soon the EMA can also reconsider the AI for the same NDSRI.


Where is this 600 ng/day interim limit for nitroso-duloxetine coming from? Based on what is this limit set?

1 Like

Thank you for sharing the information, @E_da_Silva and @Naiffer_Host. As N-nitroso-duloxetine was category 1 in CPCA (26.5ng/day), it moved to Table 2 with AI of 100ng/day which was determined from NNK by read-across. The Interim AI of 600ng/day is now available to avoid drug shortage until the estimated duration (10/1/2024). 100ng/day(0.83ppm?) seemed challenging for current products.

We should pay attention to the further actions by the FDA.


Here the discussion we had last year about Duloxetin…

1 Like

Raphy, it doesn’t seem to be the Interim-limit approach proposed by EMA on
Question 22. What is the approach to control the presence of Nnitrosamine exceeding the AI during CAPA implementation?


This shows how random the agencies have become. Nobody knows where the 600 ng/day is coming from. Even a surrogate study of N-nitroso duloxetine does not justify a limit of more than 142 ng/day. I was initially excited, thinking some company did a carc. study and found the TD50 to be 600 ng/day but the table says “interim”. Again, if they are giving the NOC of duloxetine this limit, why not give this to fluoxetine and other similar compounds. While I am happy for my cllients, this really does not make any sense.


As we know FDA don’t accept the use of LTL in any form. Even if you were to say that they are applying a x6.7 LTL adjustment for an interim period, giving a 670 ng/day limit (close enough to 600 ng/day), what is the rationale of setting the duration until Oct 2024?
It could very well be that the interim limit is set by FDA (in accordance with their nitrosamine guideline (see excerpt below from the guideline) due to potential disruption in the drug supply by multiple drug manufacturers.

“When contacted about a potential disruption in the drug supply, FDA intends to evaluate each circumstance on a case-by-case basis. FDA may work directly with a specific manufacturer or applicant of the marketed drug and intends to consider whether it is appropriate or not appropriate to recommend an interim AI limit for a temporary period. If FDA recommends an interim AI limit, it generally does not intend to object, for example based on applicable underlying CGMP violations, to distribution of such drug product batches that contain NDSRI levels at or below the recommended interim AI limit during the specified period under certain circumstances on a case-by-case basis.
In certain cases where FDA does not intend to object to the distribution of drug products from multiple drug manufacturers that contain NDSRI levels at or below the recommended interim AI limit, FDA intends to post such recommended interim AI limit on the FDA website in connection with this guidance.”


I agree, N- nitroso fluoxetine may also have same interim AI, eventhough they share few structural differences.

I heard from one agency N-nitroso fluoxetine when tested experimentally found to be strong mutagen. I am not sure of the veracity of this information.

1 Like

May be FDA and few manufacturers found the middle ground of 5 ppm which can avoid market shortage.


I want to remind you of the cases where the FDA proposed interim Als for varenicline and sitagliptin.


Agency scientists evaluated the risk of exposure to N-nitroso-varenicline at interim acceptable intake levels up to 185 ng per day and determined that it presents minimal additional cancer risk when compared to a lifetime of exposure to N-nitroso-varenicline at the 37 ng per day level.


Agency scientists evaluated the risk of exposure to NTTP at interim acceptable intake levels up to 246.7 ng per day and determined that it presents minimal additional cancer risk when compared to a lifetime of exposure to NTTP at the 37 ng per day level.

Although the justification for “minimal risk” is not explained, the interim AI range is between 5x and 6.67x in actual values.

1 Like

My hunch is that N-nitrosoduloxetine will act similarly as N-nitrosofluoxetine in Ames and other tests. This is a bit scary.

It would be useful to understand the reasoning better, but probably just pragmatism backed up by science as @Pradpharma @conudel suggest already?

The 5 ppm temporary limit we also saw for MNP in rifampin (compared to 14 ppm or 20 ppm for CPNP in rifapentine).

Possibly these “FDA does not object against” limits are indeed based on pragmatic proposals from applicants and round-up values linked to as is doable levels?
Reason for the other cases not being integrated now in Table 3 as well is not fully clear to me, 1-cyclopentyl-4-nitrosopiperazine hasn’t been integrated in any of the tables yet (also not the CPCA one), but in general the FDA tables overly represent nitroso-API and nitroso-desmethylAPI.

Of note in the MNP case a factor 31.25 and 140 applied between the temporary and final AI reported at the time, order of magnitude wise 26.5 (in case of CPCA) to 600 is comparable (factor 22.6).

Indeed @Yosukemino
For temporary limits sometimes the 6.7 popped up in FDA temporary cases, like sitagliptin.
FDA has reported: “In place of the LTL approach, the FDA implements a flexible approach that allows values higher than the AI as interim limits yet maintaining a 1:100,000 cancer risk. ” The recent guidance updates from FDA are still not focusing on LTL too explicitly. But going from 37 to 246.7 ng/day just screams the 6.7 LTL factor doesn’t it?

The reporting of 600 ng/day leads indeed to questions because if it is a rounding up thing one could think of calculations and wonder what elements FDA does not object against.

145 ng/day (NDMA) x 326.41/74 = 640 ng/day (ca. 600 ng/day)?
(although NDMA to NDSRI MW correction is less conservative than NNK to NDSRI MW correction)
100 ng/day (NNK) x 6.7 = 670 ng/day (ca. 600 ng/day)?
96 ng/day (NDMA) x 6.7 = 643 ng/day (ca. 600 ng/day)?

None of these are probably linked to the exact reasoning, but it is noteworthy 600 ng/day is higher than readacross based on ICH M7 + MW correction:

  • TD50 from the most sensitive species and site for tumors from the most robust study: 145 ng/day for NDMA readacross, 182 ng/day for NNK readacross (“150 ng/day” sometimes for combining NNK and NDMA as category).
  • N-nitroso-duloxetine (MW 326.41 g/mol): MW factor 1.58 to NNK (and 4.40 to NDMA)

182 ng/day x 1.58 = 288 ng/day

So getting to 600 ng/day could possibly be based on ALARP/Risk/benefit in combination with factor 2 or 6 not significantly impacting the LCR (as well explained by Dr. Johnson in the MfE paper), especially when temporarily applied, but that’s then going against the earlier confirmation the LCR would be maintained (unless you don’t interpret it too exactly)? Unless there is in vivo or in vitro data further supporting the WoE an alternative calculation can be done? (e.g. removal of part of the substitution upon metabolisation as an alternative way of MW correction?)
As FDA said before “yet maintaining a 1:100,000 cancer risk” for temporary strategies it should be something LTL- or MW-based?

A little bit more out of the box:
For LTL factors, in principle one can also calculate with the exact days of application, possibly FDA or applicant assumes worst case applicability somewhere from 2022-2023 (retroactively) to August 2025 or something similar:
365 x 70/x days = y = 22.6
y = 600/26.5 = 22.6
x = 365 x 70 / 22.6 = 3.097 years (37.7 months)
This is the only calculation so far exactly yielding 600 ng/day I believe (26.5 ng/day + LTL correction for 3 years without going to the categorical factor of 6.7). So staying on the straight line of Figure 1 of ICH M7 instead of the dotted extrapolation to more conservative categorical values (motivated because true AI is likely > 26.5 ng/day? e.g. 22.6/6.7 = 3.4 => this is close to 100/26.5 or 96/26.5 = 3.6).
(FDA assumes prolonging until August 2025 could be needed, based on October 2024 assessment, so I’m assuming this is already in the value. Considering applicants probably did report in 2022-2023 already, 3 year seems logic, calculating exactly 3 years is again 70/3 x 26.5 = 618 ng/day.)


Seems justified!