modeling of analytical (u)hplc: an important element in ... · modelling of analytical (u)hplc: an...
TRANSCRIPT
14/11/2013
1
Modelling of analytical (U)HPLC: An important element in the QbD toolbox
Pharma Lab Congress 2013, Düsseldorf, Germany, November 13-14th 2013Petersson, Patrik , Novo Nordisk A/S , +45 3079 2146, [email protected]
Presentation title Date 1
• QbD in general• Analytical QbD at Novo Nordisk• Modelling of analytical (U)HPLC
Outline
2Presentation title Date
Time, min
454035302520151050
Col
umn
Tem
pera
ture
, °C
30
35
40
45
50
55
60
65
0.66
0.63
0.61
0.58
0.55
0.52
0.50
0.47
0.44
0.41
0.39
0.36
0.33
0.30
0.28
0.25
0.22
0.19
0.17
0.14
0.11
0.08
0.06
0.03
0.00
14/11/2013
2
“A systematic approach to development that begins with predefined
objectives and emphasizes product and process understanding and
process control, based on sound science and quality risk management”
Quality by Design (QbD)
Presentation title Date 3
from ICH Q8 pharm. dev.
Quality by Design
Presentation title Date 4
~1980/90s 2004-2012
__________________________________Pharmaceutical Quality by Design: Product and Process Development, Understanding, and ControlLawrence X. Yu, Pharmaceutical Research, Vol. 25, No. 4, April 2008
14/11/2013
3
• The pharma industry presented their vision for analytical QbD 2009*• Expectation of regulatory relief
_____________________*EFPIA/PhRMA white paper to FDA and EMA 2009
Position paper, “Implications and Opportunities of Applying QbD Principles to Analytical Measurements”, Schweitzer et al. Pharm Tech 34 (2010) 52-59
Quality by Design for analytical methods
5Presentation title Date
• Currently no guidance• EMA/FDA - “There is currently no international
consensus … until this is achieved, any application that includes such proposals will be evaluated on a case-by-case basis.”*
• No regulatory relief at this stage • A way to increase method understanding and
thereby improve performance and robustness
Quality by Design for analytical methods
6Presentation title Date
_____________________________________*EMA-FDA pilot program for parallel assessment of Quality-by-Design applications: lessons learnt and Q&A resulting from the first parallel assessment EMA/430501/2013
14/11/2013
4
Analytical Quality by Design at NN
Presentation title Date 7
Analytical Target Profile
Risk assessment
Method understanding
Control strategy
Knowledge management
Time, min
2624222018161412108642
Col
um
n Te
mpe
ratu
re,
°C
36
38
40
42
44
46
48
50
52
541.57
1.50
1.44
1.37
1.31
1.24
1.18
1.11
1.05
0.98
0.92
0.85
0.79
0.72
0.65
0.59
0.52
0.46
0.39
0.33
0.26
0.20
0.13
0.07
0.00
• “User requirement specification”• Giving clear directions for analytical development• Ensures agreement between stakeholders• Internal targets – no regulatory relevance
• Method specific targets• Performance aspects typically included in validations (accuracy …)• User aspects (handling, safety …)• Customer aspects (re-runs, cycle time …)• Sample aspects (matrix, ranges … )• …
Analytical Target Profile
Presentation title Date 8
14/11/2013
5
Presentation title Date 9
Analytical Target Profile
• Evaluation against • Validation data• Customer/user experiences
Analytical Target Profile
Presentation title Date 10
Methodvalidationphase 1
Methodvalidationphase 3
14/11/2013
6
• Special focus on intermediate precisionand specifications/OOS• Validation
Analytical Target Profile
Presentation title Date 11
Methodvalidationphase 1
Controlsamples
Controlsamples
Controlsamples
• Control sample included in routine analysis sample sets
Methodvalidationphase 3
Stab. Dataphase 3
Stab. dataphase 1
Stab. Dataphase 2
• Stability study residual variation
Risk assessment
• Method specific risk assessment and knowledge sharing• Brainstorm to share knowledge, spot problems/possibilities and anchor actions• At transfer to/from discovery, external vendors or QC• Different focus (development, re-runs, robustness … )
14/11/2013
7
Method understanding (by modelling)
• Some techniques generic and not optimized or typically involve one factor at the time optimization (e.g. spectroscopic)
• Most techniques are well suited for optimization and/orrobustness testing based on statistical models (DoE)
• Chromatography also suitable for models based on theory (later … )
FVVk
FtV
m
Ftt DM
GM
GR
1)13.2log( 0
• Control charts
A B C D E
• Instructions for data evaluation
Control strategy
• To ensure that the analysis is fit purpose each time• Involves
• System suitability tests• Selection of parameters• Definition of data based acceptance criteria• Adjustments … later
• Preventive maintenace parametersP
14/11/2013
8
Method operational design ranges (MODR)
• MODR = design space for analytical method• MODR with acceptable results typically very small for (U)HPLC
Method operational design ranges (MODR)
• MODR very costly to define• “demonstration of statistical confidence throughout the MODR.
Issues for further reflection include the assessment of validationrequirements as identified in ICH Q2(R1) throughout the MODR and confirmation of system suitability across all areas of the MODR.“*
• Is it worth the effort?!
______________________________________________*EMA-FDA pilot program for parallel assessment of Quality-by-Design applications: lessons learnt and Q&A resulting from the first parallel assessment EMA/430501/2013
14/11/2013
9
• We do not define any MODR for chromatographic methods• Instead SST to justify changes in %ACN and temperature to
compensate batch to batch and instrument to instrument differences• Described in EP, USP and JP ±5°C
Method operational design ranges (MODR)
48°C
45°C
42°C
• Graphical guidance e.g.
Knowledge management
• ”Follow the molecule”• Hands on experience and personal contacts
• Concise method specific development report• A table with conclusions and references
• Purpose - easy documentation/retrieval of method developmentand validation information
• Continuously updated
14/11/2013
10
Analytical Quality by Design at NN
Presentation title Date 19
Analytical Target Profile
Risk assessment
Method understanding
Control strategy
Knowledge management
Time, min
2624222018161412108642
Col
um
n Te
mpe
ratu
re,
°C
36
38
40
42
44
46
48
50
52
541.57
1.50
1.44
1.37
1.31
1.24
1.18
1.11
1.05
0.98
0.92
0.85
0.79
0.72
0.65
0.59
0.52
0.46
0.39
0.33
0.26
0.20
0.13
0.07
0.00
Method understanding by modelling
Generic screen ofcolumns and
mobile phases
DoE: Optimisation of mobile
phase composition
Semi-theoretical models:Optimisation of
gradient and temperatureDoE: Robustness testing
14/11/2013
11
Semi-theoretical models
Rt
dtk
uL0
0 1
1
Linear gradient analytical solution
Non-linear gradient Numerical solutionNon-linear least squares fitting
FVVk
FtV
b
Ftt DM
GM
GR
1)13.2log( 0
time, tve
loci
ty,
distance, L
1)(ln
0
0 bkR e
u
Lt
Solvent strength models
• The retention models that can be found in most commercial software have been developed for RPC and small molecules
• Novo Nordisk is a companydeveloping products based on proteins
• In order to model proteins other chromatographic techniques and retention models are needed
• Collaboration between and
14/11/2013
12
Solvent strength models
• RPC Reversed phase chromatography
• IEC Ion exchange chromatography
• HILIC hydrophilic interaction chromatography
• HIC hydrophobic interaction chromatography
• NPC Normal phase chromatography
_____________________L.R. Snyder, J.W. Dolan, High-Performance Gradient Elution: The Practical Application of the Linear-Solvent-Strength Model, Wiley, Hoboken, NJ, 2007.
)ln(ln bak
bakln
)ln(ln bak
)ln(ln bak
bakln
Solvent strength models
44.01.1ln MWabak
• Proteins respond much more strongly to changes in %ACN
_____________________L.R. Snyder, J.W. Dolan, High-Performance Gradient Elution: The Practical Application of the Linear-Solvent-Strength Model, Wiley, Hoboken, NJ, 2007.
MW b b(MW)/b(100)100 8 1
1000 23 310000 63 8
100000 174 21
14/11/2013
13
Temperature models
Presentation title Date 25
2.8 2.9 3 3.1 3.2
x 10-3
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
1/T [1/K]
ln(k
)
40°C
Ibuprofen ~0.2 kDa
80°C
Toluene ~0.1 kDa
3 3.1 3.2 3.3 3.4
x 10-3
2
2.05
2.1
2.15
1/T [1/K]
ln(k
)
20°C60°C
NN >100 kDa
NN >100 kDa
Tbak /ln
_____________________Small molecules - M. Euerby et al. , Hichrom LtdProteins – P. Petersson et al., Novo Nordisk A/S
3 3.1 3.2 3.3 3.4
x 10-3
3.2
3.25
3.3
3.35
3.4
1/T [1/K]
ln(k
)
20°C
60°C
NN ~5 kDa
NN ~5 kDa
2//ln TcTbak
Temperature models
Presentation title Date 26
3 3.1 3.2 3.3 3.4
x 10-3
2
2.05
2.1
2.15
1/T [1/K]
ln(k
)
Temperature
20°C60°C
14/11/2013
14
• From 2, 6 or 9 experiments we need:• Retention time• Peak width (main + 1 to 2 imps.)• Peak area• Peak asymmetry for main peak• Dead volume• Dwell volume• Flow
Input
Presentation title Date 27
Column
A
B
Pump
Detector
VD “dwell volume”
VM dead volume
tG (min)
T(
C)
A 1st example: Protein >100 kDa
• Demonstration • [LINK]• Improved resolution• Improved robustness• Isolation/preparative
14/11/2013
15
A 2nd example: Peptide map ~0.5-3.5 kDa
• In general:• 60 to 80% reduction in
cycle time• Maintained or better
resolution• Improved robustness • Identification of critical
peak pairs
A 3rd example: Excipient <0.5 kDa
• 2 methods 1
14/11/2013
16
Accuracy
• tR <2% and systematic
• w <20%
• w 1.2 Rs 0.83
-50
0
50
100
150
200
250
300
350
400
450
500
1.9 2.4 2.9 3.4 3.9 4.4
t [min]
S [
-]
SA
SB
SC
Stot
-50
0
50
100
150
200
250
300
350
400
1.9 2.4 2.9 3.4 3.9 4.4
t [min]
S [
-]
SA
SB
SC
Stot
Peak tracking
• Peak areas• Retention change linear
or somewhat curved• Guess and confirm• MS!!!
14/11/2013
17
Peak tracking overwhelming?
• Model the first and the last peak• Model the API and critical peak(s)
which are easy to follow• Ignore the rest
Peak tracking overwhelming?
• Predict e.g. 6 conditions which gives good retention and resolution for the peaks which have been modelled
Rs-map based on modelled peaks
14/11/2013
18
Peak tracking overwhelming?
• Test these conditions experimentally• Pick the condition that gave the best results
What the Rs-map would look like if all peaks had been modelled
Conclusions
• Analytical QbD• NN interpretation of analytical QbD• A way to obtain better methods
• Understanding• Performance• Robustness
14/11/2013
19
Conclusion
• Semi-theoretical modelling• Works very well also for large proteins after some adaption• Not all softwares provide the necessary models• Application areas:
• Optimisation• Robustness evaluation (DoE still needed)• Fraction collection• Troubleshooting
• Anders Dybdal Nielsen• Anette Skammelsen Schmidt• Babak Jamali• Birgit Gaarde-Nielsen• Bonnie Schmidt• Helle Holton• Jytte Pedersen
• Karin Juul Jensen• Mads Wichmann Matthiessen• Marika Ejby Reinau• Steffen Frandsen• Thorbjørn Strøm-Hansen• Tine Sloth Høg
Acknowledgements