Degradation Prediction Goals - Welcome to Lhasa Limited · Degradation Prediction Goals QbD...
Transcript of Degradation Prediction Goals - Welcome to Lhasa Limited · Degradation Prediction Goals QbD...
Degradation Prediction Goals
QbD Approach
Degradation Knowledge Space
In Silico Prediction Tools
In Cerebro Chemistry of Drug Degradation
Prediction vs Actual Results
How well are we progressing?
Experimental Conditions
Is wet chemistry still required?
Degradation Workflow
3
Degradation prediction enables understanding
labile functionalities critical in designing less
reactive, more stable analogs.
With efforts to reduce time and cost to market, the
potential for stability issues increases dramatically.
Degradation studies conducted by a chemistry
guided predictive stability approach enables
analysts to deliver stability indicating methodology
more efficiently.
Stress testing of the drug substance can
help identify the likely degradation
products, which can in turn help establish the degradation pathways
and the intrinsic stability of the molecule
and validate the stability indicating
power of the analytical procedures used. The nature of the stress testing will
depend on the individual drug
substance and the type of drug product
involved.
QbD - Controlling the Quality: defining design space that the product has demonstrated in development to consistently meet required specifications
Design Space(Acceptable Operating Space) Knowledge Space
(Failure to Operate)Control Space
(Normal Operating Space)
Design Space
Control Space
BC
Parent
"Actual" Degradation Productsin final packaging / storage conditions
H
AB
CD
E
F
GH
Parent
I
"Potential" Degradation Products(Stress Testing Results)
Knowledge Space: Full Suite of
Conditions (All likely modes of
degradation)
B
CD
EParent
"Actual" Degradation Products(Accel. / Long-Term RT Stability)
H
Exploratory Screen Development
(EDS)
Screening Designed Synthesis
(SDS)
Lead Development
(LD)
CAN Seeking (CS)
Phase 1
Strong chemistry understanding applied upfront creates a solid
knowledge space in our QbD model
1- PREDICT DEGRADANTSPredict most likely degradants.
2- DESIGN PROTOCOLDevelop based on the chemistry
of the API/drug product formulation.
3- PERFORM EXPERIMENTSSample at appropriate points using
‘reasonable’ stress conditions.
4- CHALLENGE METHODOLOGYScreen degradation samples
using suitable methodology (HPLC).
5- EVALUATE PURITY/POTENCYObtain purity/potency data including mass
balancewhere appropriate.
6- SELECT KPSS/TRACK PEAKSDetermine the primary degradants.
Track KPSS across orthogonal methods.
7- IDENTIFY DEGRADANTSUtilize LC-MS, LC-NMR,.
8- DOCUMENTPrepare reports and share degradation
structures and mechanisms.
Predict Degradants based on
In-cerebro-chemistry knowledge
In silico approaches and databases :
CAMEO, DELPHI, Pharma D3 and
Zeneth
Baertschi, Alsante, Santafianos Chapter 3 “The
Chemistry of Drug Degradation” Pharmaceutical Stress
Testing, Second Edition, S. Baertschi, K. Alsante, R. Reed, Editors, Informa, July 2011
CAMEO: Computer Assisted Mechanistic Evaluation of Organic Reactions
By Prof. W.L. Jorgensen to predict the products of organic reactions based on the concepts of organic chemistry.
Analyzes molecule and the reactants and applies organic chemistry to predict reactions
Converted to a plug-in program available with CamSoft ChemOffice
No ability to teach new chemistry
DELPHI: Degradant Expert Leading to PHarmaceutical InsightReference: Pole et al. Molecular Pharmaceutics, Vol. 4, No. 4, 539-549, 2007.
Challenging to teach new chemistry, involves outdated code
CAMEO understands that the nitrogen is nucleophilic and looks for something to react with. In this case we have told CAMEO that hydrogen peroxide is in the reaction. CAMEO can be told what is in the reaction and can deal with new reagents and substrates.
N
NAr
O
N
N Ar
HO OH2
Compound A
11
Hydrogen bond dissociation enthalpy
a model for
oxidative
susceptibility?
How hard
(energetically) is it to
pull the hydrogen
off?
BDE: A Model for Predictive Stability?
Computed numbers are all over the place
However, comparison is relative. Differences wash out
(as long as we use a consistent computational model!)
Interested in relative (intramolecular) trends
Conclusion: Calculations were time consuming/
resource intensive and not sustainable model
Cl
Cl
NH
1
2
3
45
6
7
8
2'
5'
6'
9
Species Hformation1
kcal/mole
BDEkcal/mole
Hformation2
kcal/moleBDEkcal/mole
Hformation3
kcal/moleBDEkcal/mol
e
Hformation4
kcal/moleBDEkcal/mole
sertraline
29.3 22.0 19.9 10.0
-H1 47.8 70.6 41.5 71.6 40.1 72.3 32.1 74.2
-H2 63.0 85.8 56.6 86.7 50.8 83.0 42.3 84.4
-H4 55.7 78.5 50.3 80.4 47.1 79.3 39.2 81.3
-H6 87.4 110.2 81.6 111.7 73.4 105.6 65.3 107.4
-H9 53.3 76.1 48.1 78.2 43.3 75.5 35.8 77.9
-H2' 89.3 112.1 83.4 113.5 76.5 108.7 68.2 110.3
c
Uses high quality knowledge base and
reasoning engine. Detailed trees showing
chemical degradation pathways have
been implemented. Including:
Level of likelihood
• Chemical formula
• Exact mass
• Degradation pathway description
• Literature references
Novel "hybrid" publishing/database paradigm enabling a proactive approach to drug stability by establishing trends with functional groups to allow enhanced prediction of degradation results
Work with 2 verisons: internal – proprietary; external – published data mining
http://d3.cambridgesoft.com/
15
D RH
O2 h
H+ OH-
16
Degradants:
Acid/Base:
Mechanism, Conditions
Oxidation:
Mechanism, Conditions
Thermal/Humidity:
Mechanism, Conditions
17
Change in MW
Functional Group Search
With a benzylic functionality, oxidation will be emphasized in
our experimental protocol
1000 Degradants
Substructure searchable
Predicts degradants from
API
API/Excipient Impurity
API/ low MW Excipient
Captures change in MW
Records actual results
Notebook references
LC method described
Sample prep. described
Example chromatogram
Solid state stress testing
Solution stress testing
PGI degradant alerts
Flag Potential Genotoxic Impurity (PGI) structure alerts
Goal of understanding/detecting PGIs to assist analysts
Proactively identify PGIs
Developing stage sensitive strategies for controlling PGIs
Search
Pharma D3
Run Predictive Software Zeneth
Design/Execute Experimental
Protocol
Identify Actual Degradants
Compare Actual Vs Predicted Degradants
Teach Zeneth/
Archive Phama D3Not predicted in Zeneth
KB 2009-2011
Study Radical Oxidation 30 mol%ACVA, 60°C
New Zeneth KB
Rule: 2012.1.0
Enrichmnet/isolation and
characterization by MS and NMRObserved but not predicted
N
N NH
R'N
NH2
R NH
O
NH
R'
O
ACVA/O2 25%ACTUAL RESULT
<0.1% dissolving solvent H2O-15%MeCN,60C, 24h<0.1% 1N-HCl in 85%H2O-15%MeCN
5.0% 0.8N-KOH in 85%H2O-15%MeCN89.0% H2O2 oxidation31.0% Oxidation ACVA, 60C, 24h
XX% 0.8N KOH
XX% 0.8N KOH
XX% 0.8N KOHXX% 1.0N HCl
ACTUAL RESULTS
Summary of most common degradation reactions
Link to MW Pharma D3 mining
Degradation vs. synthetic chemistry
Synthetic chemistry focuses on bond making and
stoichiometric reactions
Degradation chemistry pays attention to slow, low
yielding (0.1%) and bond breaking reactions
Dehydration
Hydration
Oxidation
Oxidation
KetoneOxidation
Epimerization, Rearrangement
Hydroxyl to Ketone
Olefin
DecarboxylationMethyl Ether/
Ester Hydrolysis
Ethyl Ether Hydrolysis
Four Late stage compounds evaluated and disguised as Compounds A, B, C and D for
presentation (Tofacitinib, Crizotinib, Axitinib,
Bostutinib)
Comparison of DELPHI, Zeneth 3 (knowledge base 1-3) and Zeneth 4 beta version
Predictions performed using auto Zeneth conditions
at pH 1 and 13 with an equivocal threshold
Some differences observed in pH 1 and 13 predictions
DEG.
COND.
Oxidation
R (ACVA)
60C, 24h
H2O2
48 h,
25C
H2O2
FeSO4
48
h/25C
1N-HCl
RT, 48h
0.8N-
KOH
25C, 48h
High
Reactivity
>10%
>90%
X
30%
X
60%
X
Moderate
Reactivity
1-10%
1%
X
5%
X
Low
Reactivity
<1%
DEG.
COND.
Oxidation
R (ACVA)
60C, 24h
H2O2
6 days,
25C
0.1N-HCl
RT, 2.5h
0.2N-KOH
RT, 0.4h
High
Reactivity
>10%
10%
X
27%
X
Moderate
Reactivity
1-10%
7%
X
Low
Reactivity
<1%
<1%
X
X = assesment of degradation by degradation type using in-silico; in-cerebro prediction
Degradant
Structure
ID Observations
API&DP Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6
2012.1.3
Predicted Degs-
Observed Degs-
Total predicted Degs pH 1-
Total predicted Degs pH 13-
3*
6
39
34
2
6
31
170
3
6
158
197
5
6
188
321
A
1
API Solution
Acid/Base
Solid State Thermal
DP-
70C/75%RH/3wks
API-not observed
pH 1(yes)
pH13 (yes)STEP 1
Vey Likely
pH 1(yes)
pH13 (yes)STEP 1
Very Likely
pH 1(yes)
pH13 (yes)STEP 1
Very Likely
pH 1(yes)
pH13 (yes)STEP 1
Very Likely
pH 1(yes)
pH13 (yes)
STEP 1
Very Likely
A
2
API Solution
Not observed
Solid State Thermal
DP-
30ºC/65%RH/2y
rs
API-not observed
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(yes)
pH13 (yes)
A
3
API Solution
1) ACVA Oxidation
2) H2O2 Oxidation
Solid State Thermal
DP-
30ºC/65%RH/2y
rs
DP70ºC/75%RH/3w
s
API-not observed
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(yes)
pH13 (yes)
STEP 1
Likely
Amide hydrolysis
Ring cleavage
Pyrrolo Pyrimidine oxidation
Degradant
Structure
ID Observations
API&DP Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6
2012.1.3
A4
API Solution
Not observed
Solid State Thermal
DP-70ºC/75%RH/3wks
DP-30ºC/65%RH/2 yrs
API-not observed
pH 1(yes)
pH13 (no)STEP 3Likely
pH 1(no)
pH13 (no)pH 1(yes)
pH13 (yes)STEP 2
equivocal
pH 1(yes)
pH13 (yes)STEP 2
equivocal
pH 1(yes)
pH13 (yes)STEP 2
equivocal
A5
API Solution
ACVA Oxidation
Solid State Thermal
DP-70ºC/75%RH/3wks
DP-30ºC/65%RH/2 yrs
API-not observed
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)pH 1(yes)
pH13 (yes)STEP 2likely
A6
API Solution
1) ACVA Oxidation
2) Fenton
3) Metals
Solid State Thermal
DP-70ºC/75%RH/3wks
DP-30ºC/65%RH/2 yrs
API-not observed
pH 1(yes)
pH13 (yes)STEP 2Likely
pH 1(yes)
pH13 (yes)STEP 2Likely
pH 1(yes)
pH13 (yes)STEP 2
equivocal
pH 1(yes)
pH13 (yes)STEP 2
equivocal
pH 1(yes)
pH13 (yes)STEP 2likely
Pyrrolo Pyrimidine oxidation
Amine dealkylation
Ring cleavage
Degradant
Structure
ID Observations
API&DP Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6
2012.1.3
Predicted Degs-Observed Degs-Total predicted Degs pH 1-Total predicted Degs pH 13-
1*
3
92
41
1
3
150
167
1
3
175
170
1*
3
34
60
B
3
API Solution
Not observed
Solid State Thermal
DP-70ºC/75%RH/4wks
API-not observed
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
B
1
API Solution
ACVA Oxidation
Solid State Thermal
API -not observed
DP(tablet) -not observed
DP(oral soln)-50ºC/4wks
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
Amine dealkylation
Oxidation
Degradant
Structure
ID Observations
API&DP Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
B2
API Solution
photo
Solid State Thermal
API-not observed
DP(tablet)-not
observed
DP(oral soln)-
observed
pH 1(yes)
pH13 (no)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
NA Excipient API Interaction Predicted by Zeneth
Ether hydrolysis
Dimer
Degradant
Structure
ID Observations
API&DP Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6
2012.1.3
Predicted Degs-Observed Degs-Total predicted Degs pH 1-Total predicted Degs pH 13-
1
3
158
44
1
3
167
95
1
3
83
95
2
3
90
115
3
3
51
80
C1
API Solution
1) ACVA Oxidation
2) H2O2 Oxidation
3) NMP Oxidation
Solid State Thermal
DP-
70ºC/75%RH/6wks
API-not observed
pH 1(yes)
pH13 (yes)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
pH 1(yes)
pH13 (yes)Very Likely
Step 1
C2
API Solution
Photo
Solid State
API photo
DP photo
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)pH 1(yes)
pH13 (yes)Very Likely
Step 1
Sulfoxide
dimer
Degradant
Structure
ID Observations
API&DP
Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6
2012.1.3
C3
API Solution
Photo
Solid State
API photo
DP photo
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(yes)
pH13 (yes)
Likely
Step 1
pH 1(yes)
pH13 (yes)
Likely
Step 1
NA
*
API Solution
Photo
Solid State
API photo
DP photo
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
pH 1(no)
pH13 (no)
Cis isolmerization
dimer
Degradant
Structure
ID Observations
API&DP
Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6 2012.1.3
Predicted Degs
Observed Degs
Total predicted Degs pH 1
Total predicted Degs pH 13
2
5
226
208
2*
5
192
229
5*
5
311
311
5*
5
270
294
D1
API Solution
H2O2
oxidation
Solid State
H2O2 +API
H2O2+DP
yes pH 1(yes)
pH13 (yes)
Likely
Step1
pH 1(no)
pH13 (yes)
Very Likely
Step 1
pH 1(no)
pH13 (yes)
Very Likely
Step 1
pH 1(no)
pH13 (yes)
Very Likely
Step 1
D2
API Solution
H2O2
oxidation
Solid State
H2O2 +API
Light +API
H2O2+DP
Light + API
yes pH 1(no)
pH13 (no)
-
-
pH 1(no)
pH13 (no)
-
-
pH 1(yes)
pH13 (yes)
Very Likely
Step 1
pH 1(yes)
pH13 (yes)
Very Likely
Step 1
D3
API Solution
H2O2
oxidation
Solid State
H2O2 +API
H2O2+DP
yes pH 1(no)
pH13 (no)
-
-
pH 1(no)
pH13 (no)
-
-
pH 1(no)
pH13 (yes)
Equivocal
Step 2
pH 1(no)
pH13 (yes)
Equivocal
Step 2
Oxidation
hydroxyl
Oxidation + hydroxyl
Degradant
Structure
ID Observations
API&DP Stressed
Conditions
Prediction
KB1
2009.1.0
Prediction
KB2
2009.2.0
Prediction
KB3
2011.1.0
Prediction
KB4
2011.2.0
Prediction
KB6
2012.1.3
D4
Solid State
Photo
API
pH 1(yes)
pH13 (yes)
Likely
Step2
pH 1(yes)
pH13 (yes)
Likely
Step2
pH 1(yes)
pH13 (yes)
Likely
Step2
pH 1(yes)
pH13 (yes)
Likely
Step2
pH 1(yes)
pH13 (yes)
Likely
Step2
D5
Solid State
photo
API
pH 1(no)
pH13 (no)
-
-
pH 1(no)
pH13 (no)
-
-
pH 1(yes)
pH13 (no)
Equivocal
Step2
pH 1(yes)
pH13 (no)
Equivocal
Step2
pH 1(yes)
pH13 (no)
Equivocal
Step2
Ether hydrolysis
Ketone + hydrolysis
Benchmarked against 4 recently filed Pfizer compounds
Zeneth’s accuracy has improved from an average
of 39% to 79% on benchmarked data
Zeneth Version Z3 Z3 Z3 Z4 Z4 Z5
Knowledge Base KB1 KB2 KB3 KB4 KB5 KB6
Drug Observed Predicted % Predicted Predicted % Predicted Predicted % Predicted Predicted % Predicted Predicted % Predicted Predicted % Predicted
A 6 3 50.0 2 33.3 3 50.0 3 50.0 5 83.3 5 83.3
B 3 1 33.3 1 33.3 1 33.3 1 33.3 1 33.3 1 33.3
C 3 1 33.3 1 33.3 1 33.3 2 66.7 2 66.7 3 100.0
D 5 2 40.0 2 40.0 5 100.0 5 100.0 5 100.0 5 100.0
Average
Predicted 39.2 35.0 54.2 62.5 70.8 79.2
CORE INGREDIENTS EXCIPIENT IMPURITY
Varenicline -
Mannitol D-Mannose
Microcrystalline Cellulose (Avicel) Formic acid, formaldehyde, D-glucose, acetic acid
Dibasic Calcium Phosphate (A-Tab, Di-Cal)
Calcium phosphate
FILM COAT COMPONENTS
PEG 3350 Aldehydes, peroxides, organic acid, formic
acid, formaldehyde, acetic acid
Cellulose Acetate Acetic acid, D-glucose, formaldehyde, formic acid, acetic acid
API degradant prediction process becomes too
complicated when incorporating excipients and
their impurities in one processing step as illustrated
below.
Easier to process API against individual reactant
Examples of valenicline excipient impurity degradants in tablets
Reaction of formaldehyde with amines (Eschweiler-Clarke)
formic acid/formaldehyde
can act as a reducing agent.
Eschweiler-Clarke Methylation of Primary or Secondary Amine
Reaction of formic acid with amines-amide formation
Default Zeneth data bases below can be used to create custom
data bases for excipients and their contaminants
Chemistry guided approach using predictive tools assists in
targeting most likely reactive functional groups and experimental
conditions to focus on (by using Zeneth, Pharma D3, etc.)
Knowledge Space
Design Space
Control Space
Fine-TunedKnowledge Space
based on chemistryguided tools
Identifying Experimental Conditions:
Based on prediction knowledge, degradation conditions are
selected/optimized
1- PREDICT DEGRADANTSPredict most likely degradants.
2- DESIGN PROTOCOLDevelop based on the chemistry
of the API/drug product formulation.
3- PERFORM EXPERIMENTSSample at appropriate points using
‘reasonable’ stress conditions.
4- CHALLENGE METHODOLOGYScreen degradation samples
using suitable methodology (HPLC).
5- EVALUATE PURITY/POTENCYObtain purity/potency data including
mass balance where appropriate.
6- SELECT KPSS/TRACK PEAKSDetermine the primary degradants.
Track KPSS across orthogonal methods.
7- IDENTIFY DEGRADANTSUtilize LC-MS, LC-NMR,.
8- DOCUMENTPrepare reports and share degradation
structures and mechanisms.
Knowledge Space
Dinos Santafianos
Martin Ott and LHASA Zeneth Consortium
Steering Committee