powerpoint presentation tablet issue multivariate matrix from the database then proceed to a closer...

29

Upload: dongoc

Post on 15-May-2018

216 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 2: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 3: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 4: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 5: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 6: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 7: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Multi-disciplinary Team

Page 8: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 9: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 10: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Product Master

Database

Coating

(~10 factors)

Compression

(~20 factors)

Compaction

(~15 factors)

Raw material

(~40 factors)

Xs – Input process parameters, material attributes; the “causes” Ys – Output parameters, product attributes; the “effects”

Aim to look for “cause – effect” relationships

Page 11: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

API

API API

API

Multivariate analysis as first-step screening of potential cause-effect relationships related to capping/broken tablet issue

Multivariate matrix from the database

Then proceed to a closer look of the “interesting” correlations one by one:

Page 12: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

12

Looked at individual R2 from each correlation and come up with a cumulative overall parameter ranking of higher to lower risk of impacting FCT AQL results…

Screening DoE with top-ranking parameters Parameters not only determined by data analysis

but also by SME suggestion eg. Compaction force was not varied in historical

database but was considered important factor for DoE based on process understanding

Parameter

Sum

Normalized

R2 Rank Stage

API PSD <150 um 0.64 1 Mix

API Carr 0.48 2 Mix

COM AVG - Calculated (TD-BD)/BD 0.43 3 Com

COM AVG - 80 Mesh 0.40 4 Com

Precomp 0.39 5 Core

Binder LOD 0.37 6 Mix

Roller Speed (force conc) 0.36 7 Com

Mset 0.33 8 Tab

Page 13: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Screening DoE (top 8 parameters. Used Binder from manufacturing process 1 for high/low trials)

Screening

DoE Test

Variable

API % PSD

< 150 mm

Gran screen

size (mm)

Compaction

force (kN)

Press

Speed

(rpm)

Compaction

gap (mm) Binder LOD Pre-comp (kN) Mset (kN)

10

20

30

40

50

60

70

80

90

100

TA

BIn

dex

-H

ard

ness/

(FR

I A%

* Wt R

SD

)A

ctu

al

10 20 30 40 50 60 70 80 90

TAB Index - Hardness/(FRIA%*WtRSD)

Predicted P=0.0029 RSq=0.8 RMSE=11.791

Actual by Predicted Plot

Response TAB Index - Hardness/(FRIA%*WtRSD)

Optimization DoE (Used Binder from manufacturing process 3-supplier driven change)

Opt DoE Test

Variable

Gran screen

size (mm)

Compaction force x gap

Roller

Speed (rpm)

Pre-comp

(kN) Mset (kN)

Based on Tablet hardness, friability, Wt%RSD and AQL from Screening DoE

Multivariate Regression modeling of DoE results further determined the top four effects

TA

V2-RM-Binder LOD

Page 14: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Optimized Recommendations on Screen Size, Compaction Force, Gap Width, Roller Speed and Precompression Force

Confirmed by Confirmatory trials 1 and 2 need Compaction/Compression

range determination studies

Compaction Study (Used binder from manufacturing process 3 for Sieve Spec determination)

Compression Specification Range Determination Study (Used binder from manufacturing process 3 for Compression Spec Determination)

Full-Scale manufacturing

Based on Tablet hardness, friability, Wt%RSD and AQL from optimization DoE

Page 15: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 16: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 17: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Continuous process improvement

Full scale demonstration batches (one lower strength and one higher strength) were manufactured with zero capping/broken tablet observed.

Page 18: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 19: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Pre Change Ppk: 0.48 (~536640 PPM) Post Change Ppk: 1.89 (~22 PPM)

Pre Change Ppk: 0.79 (~195800 PPM) Post Change Ppk: optimal (not calculated since no defect observed)

Pre and Post process improvement Evaluation

Higher Strength

Lower Strength

After 1 year of post-process improvement, There are close to 200 batches of lower strength and 100 batches of higher strength manufactured and released. No appearance related product defects were observed. Product Robustness is significantly improved.

Page 20: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 21: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Design Considerations

Formulation: Simple ingredients designed to provide consistent dissolution. Process: Moisture sensitive ingredients and dry granulation was chosen.

Page 22: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Non-Critical Non-Critical Non-Critical Critical Equipment

Critical Equipment

36”pan→48”pan→66”pan Y kg →1.7 Y kg → 6.7 Y kg

Non-Critical

X kg → 7X kg

Scale up to 66” pan is a risk and requires core tablets with good mechanical strength

Page 23: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 24: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 25: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Factors Responses Desirability

Press Speed (rpm) Average Hardness (n=10) Maximize in the specified range

Pre-compression Force (kN) Friability Minimize

Main Compression Force Extended Friability Minimize

Page 26: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12
Page 27: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Strength Content Uniformity

Dissolution (Market 1) 80% in

X min

Dissolution (Other markets) 80% in Y min

1 (56 batches) 2.55 3.40 3.26

2 (142 batches) 3.29 4.63 4.11

3 (242 batches) 3.74 4.83 4.71

4 (157 batches) 4.16 4.80 5.13

Page 28: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Page 29: PowerPoint Presentation tablet issue Multivariate matrix from the database Then proceed to a closer look of the “interesting” correlations one by one: 12

Thank You