19 August 20081 Case Studies in Quality by Design with Design of Experiments From Pharmaceutical...

Preview:

Citation preview

19 August 2008 1

Case Studies in Quality by Design with Design of Experiments From Pharmaceutical Technology

Lynn Torbeck19 August 2008

19 August 2008 2

Overview

A little, very little, history3 types of controlled experimentsKey literature and datesToday’s driving force behind QbD“Show me an example in my area of interest”Case Studies from Pharm Tech

19 August 2008 3

19 August 2008 4

A Short Bit of History

Sir Ronald A. FisherBorn 1890, EnglandDied 1962, AustraliaGraduated college in 1913, math, genetics1919 joined Rothamsted Experimental Station in Harpenden, EnglandThe right person in the right place.

19 August 2008 5

Three Controlled Experiments

John S. Mill, System of Logic, 18431. Success / Failure One run, no factors varied, one outcome, yes/no Easy to design, easy to analyze Lack of comparison, inefficient

2. OFAT, One-Factor-at-a-Time We all learned this in school Several runs, one factor varied, two outcomes Easy to Design, has comparison of outcomes Can’t find interactions and is inefficient

19 August 2008 6

Fisher’s Experiments

Multiple runs, multiple factors variedMultiple outcomesWill find interactionsIs much more efficientComparison of outcomes

19 August 2008 7

Key Literature

1926, “The Arrangements of Field Experiments.” Journal of the Ministry of Agriculture of Great Britain. Fisher.1935, The Design of Experiments, Oliver & Boyd, London. Fisher.1951, “On the Experimental Attainment of Optimum Conditions,” Box and Wilson. The original source for QbD !

19 August 2008 8

Today’s Driving Force

FDA / PAT guidanceICH Q8 – Quality by DesignICH Q8 _ Annex with DOE exampleThe freedom of Design SpaceAbility to change within Design SpaceEconomics and cost savingsProduct / Process Knowledge

19 August 2008 9

State of the Topic

While there is more to Quality by Design than DOE, it seems to be the part that most people have the most trouble with.Chemometrics is many times more complicated than DOE but yet it seems to be more readily accepted than DOE.

19 August 2008 10

Show Me an Example

Many people have taken a DOE class at some time, but still have difficulty in getting started.The most common request is for examples in specific areas.Examples here show that it is not all that difficult to get started.QbD was being done before ICH Q8

19 August 2008 11

Six Steps to Designing

1. Do your homework2. Define the measured responses

(CQA)3. Brainstorm factors (CPP)4. Select 2-7 factors to be treatments5. Select levels or values for

treatments6. Select a design

19 August 2008 12

A Short List of DesignsNumber of Runs

4 8 8 9 12 16

Number 2 22 3*3, 32

of 3 23-1III 23

Factors 4 24-1IV PB9 24

5 25-2III PB8 PB9 25-1

V

6 26-3III PB8 PB9 26-2

IV

7 27-4III PB8 PB9 27-3

IV

8 PB9 28-4IV

9 PB12

10 PB12

11 PB12

19 August 2008 13

Pharm Tech Yearbook, 1999

“Functionality Testing of a Co-processed Diluent Containing Lactose and Microcrystalline Cellulose”Gohel, M., et allPre-formulation development of excipients

19 August 2008 14

Objective

“The objective of the present study was to prepare the directly compressible adjuvant by using a simpler process that could be adopted by any pharmaceutical company.Product is a tablet

19 August 2008 15

Treatments

A: Ratio of lactose to MCC 75:25, 85:15

Binding Agent Dextrin, HPMC

% binding agent 1.0%, 1.5%

19 August 2008 16

Held Constant

Stirring speed at 35 rpmStirring time at 90 minutes

19 August 2008 17

Agglomerate Responses

Bulk Density, Tapped DensityAngle of Repose, Flow RateHausner ratioCarr’s IndexFriability IndexMoisture uptake

19 August 2008 18

Statistical Design

Three treatmentsEach at two levelsEight sets of conditions or runs A 23 full factorial design

19 August 2008 19

Results

This is a complicated set of data with many two factor interactions, but it can be understood by looking at a geometric presentation of the factors and the responses for flow rate.Ratio is on the horizontal, A, axisAgent is on the vertical, B, axisPercent is on the third, C, axis

19 August 2008 20

AgentB

HPMC

16.00 16.00

19.00 14.00

RatioDextrin 14.80 18.00 A

75/25% 85/25%

1.0%

15.00 14.60

C Percent1.5%

19 August 2008 21

Observations for Flow Rate

1. Within these bounds, flow is 14.0 to 19.0 g/s

2. Slowest is 85/15, HPMC, 1.5%.3. Fastest is 75/25, HPMC, 1.5%4. Fast is 85/15, Dextrin, 1.0%

19 August 2008 22

Pharm Tech, November 1999

This is a related example.“An Investigation of the Direct-Compression Characteristics of Co-processed Lactose-Microcrystalline Cellulose Using Statistical Design.”Gohel, M., and Jogami, P.

19 August 2008 23

Pharm Tech, June, 1993

A bottle packaging example.“The Effect of Rayon Coiler on the Dissolution of Hard-Shell Gelatin Capsules.Hartauer, K.; Bucko, J.; Cooke, G; Mayer, R.; Schwier, J. and Sullivan, G.

19 August 2008 24

BioPharm, October 1997

“Demonstrating Process Robustness for Chromatographic Purification of a Recombinant Protein.”Kelly, B.; Jennings, P.; Wright, R. and Briasco, C.

19 August 2008 25

Objective

“Control is achieved by setting operating ranges for manipulated process variables. Those ranges should ensure that a process does not fail within the multidimensional operating space defined by those limits.”That is, the Design Space !

19 August 2008 26

Treatments

1. Load Mass 2.4 – 15.52. Load Conductivity 2.5 – 4.23. % Cleavage 63 – 754. Wash pH 9.4 – 9.65. Wash volume 9.7 – 11.66. Elution pH 9.4 – 9.67. Elution conductivity 8.6 – 14.4

19 August 2008 27

Responses

1. Recovery %2. Purity %3. rhIL-11 mass4. Product pool volume5. Elution pool concentration

19 August 2008 28

Statistical Design

Wash pH / Wash volume confoundedElution pH / Elution conductivity confounded

1. Five factors each at two levels2. 16 runs will still find the two factor

interactions3. Design is a 25-1 fractional factorial

19 August 2008 29

A:Elution pH B:Conductivity C:Cleavage D:Load Mass E:Wash pH Recovery % Purity %-1 -1 -1 -1 1 112.6 96.31 -1 -1 -1 -1 90.7 96.9-1 1 -1 -1 -1 104.9 97.11 1 -1 -1 1 72.8 97.3-1 -1 1 -1 -1 99.6 96.11 -1 1 -1 1 84.2 97.1-1 1 1 -1 1 98.4 97.31 1 1 -1 -1 104.2 97.8-1 -1 -1 1 -1 104.3 91.81 -1 -1 1 1 79.0 94.9-1 1 -1 1 1 94.8 96.61 1 -1 1 -1 93.7 96.0-1 -1 1 1 1 95.5 94.41 -1 1 1 -1 88.5 93.9-1 1 1 1 -1 78.7 96.51 1 1 1 1 58.7 98.4

19 August 2008 30

Design Space

IndependentFactorSpace

?DependentResponse

Space

19 August 2008 31

Conceptual Design Space

Uncertain space

Region of operability

Operation Space Opt

Region of Interest

19 August 2008 32

Statistical Design Space

“The mathematically and statistically defined combination of Factor Space and Response Space that results in a system, product or process that consistently meets its quality characteristics, SSQuIP, with a high degree of assurance.” LDT

19 August 2008 33

Analysis

Analysis is done by fitting a mathematical model to the factors (CPP) and the responses (CQA) that includes the factor main effects and the significant two factor interactionsThe model is then used to find contour plots for recovery and purity.

19 August 2008 34

Design-Expert® Software

Recovery112.6

58.7

X1 = A: Elution pHX2 = B: Conductivity

Actual FactorsC: Cleavage = -1.00D: Load Mass = 0.00E: Wash pH = 0.00

-1.00 -0.50 0.00 0.50 1.00

-1.00

-0.50

0.00

0.50

1.00Recovery

A: Elution pH

B: C

ondu

ctiv

ity86.779291.133395.487599.8417

104.196

19 August 2008 35

Design-Expert® Software

Purity98.4

91.8

X1 = A: Elution pHX2 = B: Conductivity

Actual FactorsC: Cleavage = -1.00D: Load Mass = 0.00E: Wash pH = 0.00

-1.00 -0.50 0.00 0.50 1.00

-1.00

-0.50

0.00

0.50

1.00Purity

A: Elution pH

B: C

ondu

ctiv

ity

94.8167

95.2708

95.725

96.1792

96.6333

19 August 2008 36

Design-Expert® Software

Overlay Plot

RecoveryPurity

X1 = A: Elution pHX2 = B: Conductivity

Actual FactorsC: Cleavage = -1.00D: Load Mass = 0.00E: Wash pH = 0.00

-1.00 -0.50 0.00 0.50 1.00

-1.00

-0.50

0.00

0.50

1.00Overlay Plot

A: Elution pH

B:

Co

nd

uct

ivity

Recovery: 90Recovery: 100

19 August 2008 37

Pharm Tech, February 1999

“Blow-Fill-Seal Technology: Part II, Design Optimization of a Particulate Control System.”Price, J.

19 August 2008 38

Objectives

1. Optimize the particulate control system

2. Find cause and effect relationships

3. Alter the system settings to improve performance

4. Find interactions between factors

19 August 2008 39

Treatments

1. HEPA flow rate % 20 50 802. Damper % open 30 55 803. Chimney air ft/min 300 550 8004. HEPA height in 0 0.3755. Isolation plate Slotted –

Hole6. Knife cut Double Single

19 August 2008 40

Response

Particulate level.Three measurements at each of the 24 conditions

19 August 2008 41

Statistical Design

Six factors Three at two levels Three at three levels

16 combinations8 center pointsDesign is a 26-2 fractional factorialDesign is resolution IV

19 August 2008 42

Analysis

Analysis of Variance, ANOVA, was used.15 effects were included5 were statistically significant Damper HEPA height Knife cur Isolation plate HEPA flow * HEPA height OR {damper*knife cut}

19 August 2008 43

Conclusions

“The study met the design objective of minimizing the particulate levels while the particulate control system operated in the dynamic state. … a more thorough understanding of the cause and effect relationships between the critical input factors and the particulate levels was obtained using the DOE.”

19 August 2008 44

Pharm Tech, Analytical Validation, 1999

Robustness Testing of an HPLC Method Using Experimental Design.”Peters, P. and Paino, T.

19 August 2008 45

Objective

“This article describes an experimental design that challenged an analytical method that assays two components in a solid dosage drug product.”Confirm the robustness of an HPLC method.

19 August 2008 46

Treatments

HPLC system A, BHPLC column Y, XWavelength A 270, 290 B 215, 235

Flow rate 0.7, 1.3

19 August 2008 47

Treatments

Injection volume 10, 30Column tempAmbient, 30Mobile phase TFA 85, 75 MeCN 15, 25

19 August 2008 48

Responses

1. Resolution of component A and B2. Theoretical plates for A and B3. Tailing factor for A and B4. %RSD of the peaks for A and B

19 August 2008 49

Statistical Design

7 factors each at two levelsWavelength A and B are confoundedMobile phase TFA and MeCN are confounded8 runs done in triplicate for 24 totalDesign is a 27-4 fractional factorialDesign is resolution III.

19 August 2008 50

Analysis and Results

Visual inspection of an overlay of the 8 chromatograms shows that the method is robust within the tolerance limits of the parameters tested. They have acceptable resolution and peak shape.

19 August 2008 51

Compare Chromatograms

19 August 2008 52

Pharm Tech, May 1998

“A Systematic Formulation Optimization Process for a Generic Pharmaceutical Tablet.”Hwang, R.; Gemoules, M; Ramlose, D. and Thomasson, C.

19 August 2008 53

Objective

“ … optimizing an immediate release tablet formulation for a generic pharmaceutical product.”Develop a generic tablet with a disintegration time of 6-12 minutes, 5 minute dissolution of 40-60% and 45 minute dissolution of greater than 90%.

19 August 2008 54

Treatments

API particle size small largeAPI % 5% 10%Lactose MCC ratio 1:3 3:1MCC particle size small largeMCC density low high

19 August 2008 55

Treatments

Disintegrant cornstarch, glycolateDisintegrant % 1% 5%Talc 0 5%Mag Sterate 0.5% 1%

19 August 2008 56

Responses

Blend homogeneityCompression force %RSDEjection forceTablet weight %RSDTablet hardness

19 August 2008 57

Responses

Tablet friabilityTablet disintegration timeTablet dissolution at 5 minutesTablet dissolution at 45 minutes

19 August 2008 58

Statistical Design

9 factors each at two levels16 runsDesign is a 29-5 fractional factorialResolution III

19 August 2008 59

The best formulation:

API 7.14%Fast-Flo lactose 60.74%Avicel PH-302 30.37%Talc 1%Mag Stearate 0.75%

19 August 2008 60

Conclusion

“The formulation was successfully scaled up to a 120 kg batch size and the manufacturability and product quality were confirmed.”“This study has demonstrated the efficiency and effectiveness of using a systematic formulation optimization process … “

19 August 2008 61

Pharm Tech, March 1994

“Evaluation of a Cartridge and a Bag Filer System in Fluid-Bed Drying.Bolyard, K. and McCurdy, V.

19 August 2008 62

Pharm Tech Europe, April 2000

“Response Surface Methodology Applied to Fluid Bed Granulation.”Wehrle, P. et all

19 August 2008 63

Pharm TechMarch 1992 and May 1992

“A Compaction Study of Directly Compressible Vitamin Preparations for the Development of a Chewable Tablet, Parts I and II.Konkel, P. and Mielck, B.

19 August 2008 64

Pharm Tech, March 1994

“Computer Assisted Experimental Design in Pharmaceutical Formulation.”Dobberstein, R. et all.

19 August 2008 65

Pharm Tech, April 1998

“A Unique Application of Extrusion for the Preparation of Water Soluble Tablets.”Murphy, M. and Hollenbeck, R.

19 August 2008 66

Pharm Tech, June 2000

“Artificial Neural Network and Simplex Optimization for Mixing of Aqueous Coated Beads to Obtain Controlled Release Formulations.”Vaithiyalingam, S. et all.

19 August 2008 67

Summary

Looked at 13 Case studiesShown 3 types of analysis Shown several areas of applicationIllustrated how to get startedShown that Q8 QbD has a precedentDOE has been used for a long time

19 August 2008 68

Acknowledgements

The University of Adelaide Library is the owner of the image of Sir R. A. Fisher.Pharmaceutical Technology holds the copyright for the journal articles used in this presentation.Opinions in this presentation are that of Lynn Torbeck alone.

Recommended