Why you should consider controlling your potentially mutagenic impurities outside the lab - Using Calculations to Support Purge Arguments Under ICH M7
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Why you should consider
controlling your potentially
mutagenic impurities
outside the lab
Using Calculations to Support
Purge Arguments Under ICH M7
Slides prepared for a webinar in collaboration
with Lhasa Limited
Outline Background – Concept Introduction to approach – key factors How to calculate purge factors Relationship to ICH M7 Regulatory advocacy Case Study – Candesartan Potential benefits
Concept
Background
• The threat posed by (potential)
mutagenic impurities, (P)MIs, in drug
substances arises from the use of
reagents such as alkylating agents
within the synthesis.
• What makes them useful reagents in
synthesis, is also what makes them
(P)MIs.
• Virtually all syntheses will involve the
use of mutagenic or potentially
mutagenic reagents or possess potential
risk arising from a (P)MI formed in the
process.
• Any synthetic drug therefore may have a
latent (P)MI-related risk.
4
Evaluating Risk Posed by Mutagenic
Impurities
• Fundamentally there is a need to assess the risk posed by mutagenic
impurities.
• Are there any mutagenic impurities present in the final product at
levels of concern?
• Historically the emphasis has been to analytically test for every MI.
• Significant analytical challenges.
• Also takes little account of knowledge of intrinsic reactivity / physico-
chemical parameters.
• Question Posed – Could a systematic way be found that takes into
account factors that reduce the risk of potential carryover?
5
Example of the analytical challenge
• Synthetic Intermediate – multiple stages from final API
O
• Exists in two forms O
• Thermally sensitive HO HO
HO
• Non-chromophoric O
Cl
• Involatile
• Derivatisation also very tricky
Required use of
NMR yet purge
calculation
showed purge
>100,000
6
Introduction to approach Key factors
The following key factors were defined in order
to assess the potential carry-over of a MI:
• reactivity, solubility, volatility, and any additional physical
process designed to eliminate impurities e.g.
Purge chromatography.
Score assigned on the basis of the
Factor physicochemical properties of the MI relative
to the process conditions.
Calculation • These are then simply multiplied together to determine a
– Basic
‘purge factor’ (for each stage)
The overall purge factor is a multiple of the
Principles factors for individual stages.
Predicted purge is then compared to required
purge (this being based on the safety limit and
initial level introduced into the process).
8
• Reactivity was set as 1, 10 and 100 based on correlation
with likely analytical data
• Volatility based on comparative boiling points solvent vs
impurity – focused on small mwt. Alkyl halides in first
Original instance
Purge • Solubility based on principle that most MIs were reactants
Prediction used in an optimised process, hence soluble in the reaction
system and retained in the mother liquors, kept
Scoring conservative based on:
• Absolute solubility in pure solvent may not directly
System correlate with solvent system at end of reaction / work
up
• Crystallisation may be uncontrolled – crash out
• Also linked to desire NOT to overpredict.
• Scoring system originally designed to be conservative:
• On validation this was experimentally observed.
• Decided this should be retained rather than seeking
absolute parity.
9
Overall scoring system
Purge Factor
Glossary (ICH M7 (R1))
Purge factor
• Purge reflects the ability of a process to reduce the level of an impurity, and the
purge factor is defined as the level of an impurity at an upstream point in a process
divided by the level of an impurity at a downstream point in a process
• Purge factors may be measured or predicted
＜Example of purge factor calculation＞
1. reaction level of an impurity at an upstream point
starting material A intermediate B Purge factor =
2. crystallization
level of an impurity at a downstream point
+ +
Stage X 100 ppm
impurity C impurity C =
(100 ppm) ( 1 ppm)
1 ppm
= 100Is this simply about avoiding
analytical testing? 3-aminopropan-1-ol
THF
3 MIs of concern AZD9056 Aldehyde
AZD9056 Imine
THF
Example 1 – AZ9056 Aldehyde
Step 1 – reductive amination: AZD9506 Free Base
Reactivity = 100 – based on in process control.
Solubility = 1 – not isolated – no purging
Volatility = 1 – not volatile Isopropyl chloride
(by-product)
Step 2 – Isolation of HCl salt :
Reactivity = 1 .HCl
Solubility = 10 – desired product isolated, AZD 9056 Chloride
(minor by-product)
AZD9056 HCl
residual Aldehyde remains in solution.
Volatility = 1
Step 1 (Predicted) Step 2 (Predicted) Step 3 (predicted) Pure Stage
Pure
Reactivity Solubility Volatility Predicted Reactivity Solubility Volatility Predicted Reactivity Solubility Volatility Predicted Predicted Measured
Purge Purge Purge Purge Purge
Factor Factor Factor Factor Factor
AZD9056 100 1 1 100 1 10 1 10 1 10 1 10 10,000 112,000
Aldehyde
AZD9056 Not present Not present Not present N/A 1 1 1 1 1 3 1 3
Chloride 3 10
Isopropyl Not present Not present Not present N/A 1 10 10 100 1 10 10 100 10,000 38,500
ChlorideHow to calculate Purge Factors
Risk assessment of mutagenic impurities using purge factor
Hazard assessment of impurities
• Selection of actual and potential
impurities Select mutagenic impurities requiring management
• Database searches and (Q)SAR
assessment
Basic information related to impurities
• Initial concentration Calculate Required purge factor based on control level of mutagenic impurity
• Acceptable intake
Maximum level of mutagenic impurity (ppm)
• Maximum daily dose of API Required purge factor =
Acceptable limit in the API (ppm)
• Duration of treatment
re-evaluation of
purge factor
Scientific rationale for purge assessment
• Experimental data
Calculate Predicted purge factor for mutagenic impurity
• Fate and purge of impurity
• Physicochemical properties of impurity
Predicted purge factor = reactivity×solubility×volatility×ionizability×physical processes
• Related literature reports
• Expert commentary
Additional experimental data
Determine Purge Ratio for mutagenic impurity • Measured purge factor
• Analytical data of impurities
Predicted or Measured purge factor • Verification of physicochemical
Purge Ratio =
Required purge factor parameters
• Expert commentary
Select control option based on Purge Ratio
Modified from Regul. Toxicol. Pharmacol. 90 (2017) 22-28
Reproduced with kind permission of Yusuke Nagato – Fuji filmHow to evaluate • STEPS INVOLVED:
Reactivity • Evaluate the potential
chemical reactivity of
impurities under both
reaction and work-up
conditions
• This is done by
examining the physico-
reactivity chemical properties of
reactivity class score the impurity and
criteria
correlating those with
conversion the process and work
highly reactive 100 up conditions
99%
• Reaction monitoring
90% data for impurities in a
conversion moderately reactive 10 given process, may be
used to score by the
99%
reaction conversion of
the impurities
conversion low
1
90% reactivity/unreactive• If there are no experimental data on the
reaction conversion, it is possible to calculate
the reactivity by predicting the reactivity.
Potential points to consider:
• Reactivity data with similar compounds
• Reactivity data for similar reactions
• Supporting literature references related to
How to reactivity
• Reaction rate constants and half-life data
evaluate under process conditions
• Behavior of extremely highly reactive
reactivity species (e.g., thionyl chloride)
• Expert review on reactivity
• What though if I could do this using an in
silico model ?How to evaluate Solubility
• Impurities can be removed after reaction solubility criteria (USP) solubility class score
remaining within the process solvent after
solubility 100 mg/mL freely soluble 10
isolation of the desired product through
crystallisation 33 mg/mL solubility 100 mg/mL moderately Soluble 3
• Originally suggested to use USP solubility
criteria – but these relate to pharmaceuticals solubility 33 mg/mL sparingly soluble 1
• Scoring based solely on solubility defined in the USP may not be realistic in the context of factors
such as:
• Low levels of impurities compared to solvent volumes
• Solvent systems may be mixed
• At end of reaction the system may be saturated
• As a consequence of these factors a deliberate conservative approach is taken to the scoring of
solubility
• It is possible to predict the solubility purge factor by considering the following:
• Solubility data of similar compounds
• Physical property value (e.g., Log P value)
• Physicochemical properties of impurities
• Solubility prediction modelsExtractions / Ionisability / Solubility Proposed Best practice
Extraction best practice
• The extraction term refers to a unit process which agitates two immiscible
solvents and subsequently separates them.
• Either an impurity is extracted, or the intermediate/API is.
• Discourage the use of the term wash due to overlap with liquid-solid systems
Q: 1 or 3
processes?
A: Both
“Without experimental evidence, a single solubility purge for entire extraction process (I.e.
multiple extractions of the same or similar composition) should be applied consistent with
the solubility purge factor criteria, however, where experimental evidence is available then
solubility purges for each separate extraction can be applied and are justified. While this
evidence would not necessarily be provided automatically, it should already be available in
the event of a regulatory request.”Liquid (PMI) - Solid systems • Terminology can be open to interpretation which can lead to confusion • Ideally terminology usage would become consistent
Solubility Predictions -
Augmentation
• Approaches continued
Extrapolation
• It is possible to extrapolate solubility from one solvent to another
• It is even possible to do ‘ab initio’ calculations.
Surrogate data
• May be possible to find data for
similar structures in the literature
at least for chemical transformations.
Literature data – range of solvents >10g/l
Theoretical impurity (levelRelationship to ICH M7
ICH M7: Control
Strategy Options for
Mutagenic Impurities
Section 8 - CONTROL
• Greater capability in terms of
mechanism to prove absence.
• Options other than to simply test for
presence in final API.
• Ability to more widely use chemical
/ process based arguments to assess
purging.
• Expressed in terms of a series of control
options.
23ICH M7 – Mutagenic Impurities
Section 8 – CONTROL
• What does the guideline state?
• Where the control strategy relies on process controls in lieu of analytical testing
there must be understanding of how the process chemistry and process
parameters impact levels of mutagenic impurities.
• The risk assessment can be based on physicochemical properties and process
factors that influence the fate and purge of an impurity
• This includes chemical reactivity, solubility, volatility, ionisability and any
physical process steps designed to remove impurities.
• These factors are built into both the paper-based approach and also Mirabilis
24ICH M7 – Mutagenic Impurities
Section 8 – CONTROL - Defines a series of control
options
Option 4 Option 2
• Predicted to be removed by • Test for the impurity in the
processing based on process specification for a raw material,
understanding – no testing starting material or
required. intermediate at permitted level.
Option 3 Option 1
• Test at intermediate stage • Test for the impurity in the
with a higher limit + drug substance.
understanding of process Impacted by purge
capacity. predictions
What is the right order?
25
Which data are required?• The principle of relating the physico-chemical
properties of the mutagenic impurity to the chemical
Control Option 4 process is defined in the concept of purge factor
How do I apply calculations.
this in practice? • OPRD paper referenced directly in ICH M7
26Purge Tool
How do predicted
values compare to
actual measured?
• Several examples by
consortium members
indicate the system shows a
systematic bias.
• It under-predicts – typically
by a factor of around 10.
• This is important! In order
to maintain robustness it
should not over-predict.
27Example of Purge Predictions in the
Context of ICH M7: MK-8876
Measured
Predicted
(to LOD)
(P)MI
EDC 1010 >50,000*
MeI 106 > 105
ICH2Cl 106 > 2 x 105
ArB(OH)2 3 x 104 > 106
BBA 1000 > 250,000
Carbazole 100 > 375
*LOD after 3 stages
Org. Process Res. Dev. 2015, 19, 1531-1535 28Impurity Carry-over Workflow
API synthesis
Implement
Knowledge of physicochemical Reactive High utility
properties Plan functionality Efficient syntheses
Mutagenic?
Experimental
toxicity
Unknown
In-silico toxicity
prediction
Mutagenic
Impurity Control In-vitro toxicity Non-mutagenic
Option 1 or 2
Option 3 or 4 Mutagenic
Purge calculations
Not
Assess likelihood of impurity Purged
Test for impurity
persisting
Analytical challenge
Purged Time consuming
Option 3
Testing
unnecessary
29Defining data collection requirement
• General approach to use of data
• Providing an initial workflow to show how existing data and
predictions can support an initial assessment if the user
chooses.
• Highlight data collection is NOT part of the initial
assessment
• Further data collection is subject to final PR requirements
Initial assessment Purge ratio PR > 1000 No further
(Expert only) obtained support required
PR < 1000
Help/Guidance
Obtain adequate
• Predictive
supporting
equations/software Data collection*
arguments/data
• Existing data
where necessary
*Data collection does not automatically imply trace data collection.What support could be appropriate and when –
proposals
• Aim is to identify the level of evidence that should be provided as support
to a purge ratio where further support is required (I.e.Regulatory Advocacy
Purge prediction principles and scoring rubric well
established and used in paper assessments.
ICH M7 provides framework for control strategy
options.
Regulatory
Advocacy Mirabilis provides a partially automated and living
knowledge base to assist scientists and regulators
in making and reporting purge predictions.
The Mirabilis consortium developed a framework
to implement this technology consistently into
(P)MI workflows throughout development and
commercialization.
• Goal: consistent application and presentation of purge
prediction science to help drive broad regulatory acceptance.
33Mirabilis regulatory workflow
publication
Goal: establish framework to leverage purge predictions to inform selection
of control strategy during development, which in turn informs both data
collection and regulatory reporting recommendations
34Mirabilis (P)MI Purge Prediction Decision Tree
Key premise: purge excess dictates data collection needs and regulatory reporting practices
Impurity requires management as (P)MI
Determine Purge Ratio (PR) in current API route for (P)MI
Predicted purge factor for (P)MI
Purge Ratio = -----------------------------------------------------------------------------
Required purge factor to achieve TTC or PDE for (P)MI
Select initial ICH M7 control strategy for (P)MI during development based on Purge Ratio. Implement
recommended experimental data collection and regulatory reporting strategies based upon Purge
Ratio (next slide)
Select ICH M7 Option 4 Does final data Select ICH M7 Option 1,2
commercial strategy package support or 3 commercial strategy,
commercial ICH M7 as appropriate
Yes Option 4 strategy ? No
35Example of calculation of Purge
Ratio
Purge Ratio prediction of (P)MI “X” (a process reagent)
• Assume TTC is 100 ppm
• Assume charge (initial conc) is 1 eq or 106 ppm
• 104 purge factor (106 / 100 ppm) needed to achieve TTC
• Therefore to achieve a 103 Purge Ratio (i.e. three order magnitude
more purge predicted than required to achieve TTC), Mirabilis must
predict a 107 cumulative purge factor
Predicted purge factor for (P)MI
Purge Ratio = -----------------------------------------------------------------------------
Required purge factor to achieve TTC or PDE for (P)MI
So how does one consistently apply the (P)MI Purge Ratio to
lab workflows and regulatory reporting ?
36When Purge Ratio > 1000…
Data Collection Recommendations
Collection of additional experimental data not necessary to support
scientific rationale for non-commercial or commercial API routes.
Regulatory Reporting Recommendations
Report “unlikely to persist” or cumulative predicted purge factor and Purge
Ratio for non-commercial API routes in regulatory submissions.
Replace with summary of key elements of predicted purge factor
calculations and Purge Ratio for commercial API routes in regulatory
submissions.
Option 4 recommended
37Example presentation in regulatory dossier
when Purge Ratio > 1000 in commercial route
Point of introduction Stage 2 of 5
(P)MI TTC 50 ppm
Assumed initial concentration 106 ppm at start of Stage 2 because
and rationale for selection “X” charge is 1 equivalent
Required Purge Factor to achieve 2 x 104 = 106 ppm initial conc / 50
2 x 108 (source Mirabilis software vx.x)
Key factors: 1000x purge in Stage 2
Predicted Purge Factor
driven by reactivity and solubility,
purge in Stages 3-5 driven by solubility
Purge Ratio 1 x 104 = 2 x 108 / 2 x 104
Control Strategy Option 4
No supporting experimental data collection
recommended when Purge Ratio is large
38When Purge Ratios > 100x and
Example presentation in regulatory dossier when Purge Ratio > 100x and
When Purge Ratios
Example presentation in regulatory dossier when Purge Ratio
Key Observations
• Data show a strong preference in reporting MI risk for
the use of control option 4.
• Provide Further strong evidence in support of the
principles outlined by Teasdale et al, 2013.
4445
Not all risks are
equivalent
Candesartan Story
46Modified Synthesis of Valsartan (used by Zhejian Huahai Pharma Co)
Example -
Candesartan
• Although contains the
same tetrazole ring as
Valsartan however the
synthesis is very
different:
• DMF not used in
tetrazole stage
• Tetrazole stage
multiple stages
from API
48Process Control
In no instance are all the constituents for formation of N-nitrosamines present in the manufacturing process
High purge factor for dimethylamine
Potential source of and other secondary amines
secondary amines (8x1017)
DMF Sodium nitrite
Purge of DMF, triethylamine and
any potential secondary amine High purge factor
Triethylamine for DMF (7x109)
Purged Purged Purged and NEt3 (8x108)
(if present)
]
(if present) (if present)
NPA MNA BAN MBN BEC
Purged
[ MET
(Tetrazole ring
formation step)
The magnitude of the calculated purge factors preclude the formation of nitrosamine
impurities during the tetrazole ring formation stage. In no instance are all of the Purge of
constituents for formation of N-nitrosamines present in the manufacturing process. sodium nitrite
Purged
After formation of the tetrazole ring, the final 4
DMF
process steps provide multiple unit operations for Triethylamine
additional purging before drug substance
TCV-116 Crude TCV-116 TCV-116(T) CV11974(T) CV-11974
High purge factor for
Source of DMF
Starting
Intermediate
Drug Source of
secondary sodium nitrite (1x106)
material substance nitrite
amine
49Candesartan - Outcome of testing
• Initially > 40 batches of API tested – NDMA not detected Limit
150ppb
• Limit of detection reduced now to ~1ppb
• DMA Not detected in Stage 5 (tetrazole)Impact
Annual Scale of Operations
Based on a Mid/large company
• Each project will have to deal with 5 PMIs introduced in the synthetic
route
• Estimated at 4.1 by Elder and Teasdale1
• However recent experience at AZ suggests this is closer to 5
1) Elder, D.P., Teasdale, A., 2015. Is Avoidance of Genotoxic Intermediates/Impurities Tenable for
Complex, Multistep Syntheses? Org. Process Res. Dev. 19, 1437–1446.Analytical Workflow
Additional
studies e.g.
solubility
Spike/Purge
and fate
studies
Develop Re-develop Re-develop
analytical analytical analytical
methods methods methods
Phase I Phase II Phase III Post Approval
Analyse Analyse Analyse Analyse
batches batches batches batchesBased on the scale of operations suggested,
annual analytical effort for PMIs could
involve;
• 50 Analytical methods
•
In Practise – •
7 Re-developed analytical methods
4 Spike/purge and fate studies
Annual • 4 Additional studies
•
Analysis Effort 250 Analytical methods conducted
• Total hours spent = 10,944
54In Practise – Time Saved
• Expected annual analytical effort for a mid/large company)
No Purge Approach Purge Approach
50 Analytical methods 10
Time saved
7 Re-developed analytical methods 1.75
10, 944 hours
4 Spike/purge and fate studies 2
4,252 hours
4 Additional studies 2
250 Analytical methods conducted 130You can also read
