1. Introduction Background to the work of the BUWG Garth Boehm
BUWG Draft Recommendations
Tom Garcia
Data Mining: Challenging the Theory

2. Why Test Blend Uniformity?
21CFR
(a) To assure batch uniformity and integrity of drug products, written procedures shall be established and followed that describe the in-process controls, tests, or examinations to be conducted on appropriate samples of in-process materials of each batch. ……..
(3) Adequacy of mixing to assure uniformity and homogeneity; …...

3. Why Test Blend Uniformity?
OGD’s Draft Guidance
All Solid Dosage forms <50% active or <50 mg require routine BUA
Use 6 to 10 samples, unit weights
Must meet mean 90.0% to 110.0% label claim, RSD NMT 5.0%

4. Product Quality Research Institute
PQRI ( is a collaborative effort between FDA, Industry, and Academia.
PQRI’s mission is to provide a scientific basis for developing Regulatory Policy.
One of PQRI’s initiatives is to set up ‘expert’ Working Groups to analyze particular areas and make recommendations on future Regulatory Policy.

5. Blend Uniformity Working Group
The Blend Uniformity Working Group was established in late 1999
The group is chaired by Tom Garcia and has members from academia, FDA (CDER and DMPQ), and industry (innovator and generic).
The group is charged with making scientifically based recommendations on suitable procedures for assuring batch homogeneity.

6. BUWG Actions Conducted Industry Practices Survey
Published Uniformity Troubleshooting Guide
Held Public Workshop on BU testing issues
Held several Working Group meetings
Written Draft Proposal for use of Stratified Testing of Dosage Units
Sought data to challenge proposed method

7. Industry Practices Survey
Survey was blinded to assure confidentiality
Sent to all solid dose sponsors with at least one approved NDA or ANDA that could be located
Designed to elicit information on general practices regarding BU sampling and testing

8. Industry Practices Survey
28 of 134 replied (20%), mostly large manufacturers
Survey asked questions on Demographics, Sampling, Routine Testing, Validation Testing, Cause of Failure, Cost, & New Technology
Full Survey and Results can be found at and a summary in August 2001 Pharmaceutical Technology

9. Industry Practices Survey
The picture that emerged was of a conservative industry that:
Samples with conventional sampling thieves taking unit dose sample sizes
Tests samples with conventional ‘wet’ analytical methods
Uses established acceptance criteria

10. Industry Practices Survey
About 2/3 for routine batches and 1/2 for validation batches are prepared to defeat failing BU results with enhanced testing
Have trouble with 10% to 20% of products
Think failures are due to sampling or analytical error
Have not adopted any ‘new technology’

11. Troubleshooting Guide
Early in the BUWG discussions it became apparent that no concise guide was available for diagnosing blend or dosage unit uniformity problems
Jim Prescott and Tom Garcia took on the task of writing the guide and designing a companion chart
Published in March 2001 Pharmaceutical Technology

12. Public Workshop
Based around the theme “Is BU Testing a Value Added Test?”
Held September 2000, about 200 people attended
Several formal presentations on aspects of blending, blend sampling, acceptance criteria, new technology, BUWG progress
Summary published in September 2001 Pharm Tech

13. Public Workshop Presentations based around the following:
Blending of solids is a poorly understood process
Very difficult to sample powder bed with conventional sampling thieves
Sampling errors are common & occur both ways
Post-blending segregation can be a serious problem

14. Public Workshop
Major part involved break-out sessions to elicit feedback from attendees.
Is Blend Uniformity Testing on every batch a value-added test?
How do you validate a process when you have a sampling problem?
What new technologies are available to assess blend uniformity?

15. Public Workshop Conclusions
Blend Uniformity Testing on every batch is not a value-added test
Appropriate and meaningful BU testing should be conducted during development and validation
Higher costs are acceptable if they yield meaningful data

16. Desired Attributes of a BUWG Recommendation
BUWG Draft Proposal
“The Use of Stratified Testing of Blend and Dosage Units to demonstrate Adequacy of Mix for Powder Blends”
1. The test should be simple to perform, maximizing the use of data
2. Acceptance criteria should be easy to evaluate and interpret
3. Acceptance criteria should demonstrate when lack of homogeneity is suspected

17. PQRI BUWG Recommendation for the Use of Stratified Sampling to Demonstrate Blend & Dosage Unit Content Uniformity

18. PQRI BUWG Recommendation
Utilizes stratified sampling
Collectively considers the uniformity of the powder blends and dosage units.
Acknowledges that the best way to assess blend uniformity may be indirectly by measuring the uniformity of the dosage units.

19. Scope of Recommendation
Applies to:
Process validation and marketed batches for solid oral drug products.
Products where the active ingredients are introduced into the blend.
Does not apply to:
Drug products where the determination of dosage-form uniformity by weight variation is allowed.

20. Stratified Sampling
“The process of selecting units deliberately from various locations within a lot or batch or from various phases or periods of a process to obtain a sample.” [Glossary and Tables for Statistical Quality Control , ASQC Quality Press, copyright 1983.]
Stratified sampling of the blend and dosage units specifically targets locations either in the blender or throughout the compression/filling operation which have a higher risk of producing failing content uniformity results.

21. Stratified Sampling of Dosage Units
Advantages
More accurate & relevant
Eliminates blend sampling errors
Detects segregation following blending
Eliminates issues related to blend sampling of toxic or potent drugs (operator safety)
Disadvantages
“Too late”
Batch compressed/filled
Not consistent with “quality by design”
Parametric release
Note: Control vs. Test
BUA is utilized as a test

22. Process Development
Stratified sampling plan is not a substitution for poor process development
Sampling technique should be defined during process development
Determine appropriate sampling device
Identify an acceptable sampling plan (for both blend and dosage units)
Recommendation allows blend sample sizes > 1-3X, if they can be scientifically justified

23. Validation Approach

24. Process Validation Blend: 10 locations 3 samples per location
Assay 1 sample per location
Acceptance Criteria:
RSD  5.0%
All individuals within +/- 10% of mean
Fail
Pass
Assay 2nd and 3rd blend samples
from each location
Proceed to Stage 1
Dosage Unit Testing
Mixing problem
identified
Yes No
Investigation points to sampling
bias or some other attributable cause
Proceed to Stage 2
Dosage Unit Testing
Blend is not uniform.
Go back to development

25. Process Validation During compression/filling, sample from at least
20 locations, taking at least
7 dosage units per location
Assay at least 3 dosage
units per location
Acceptance Criteria: RSD of all individuals  6.0%
Each location mean within % target potency
All individual within % target potency
Pass
Process
Validated
Fail
Assay at least 4 additional dosage units per location
Pass
Acceptance Criteria: RSD of all individuals  6.0%
Each location mean within % target potency
All individual within % target potency
Fail
Blend is not uniform or post-blending
practices cause segregation

26. Justification of Blend Sample Sizes and Acceptance Criteria
Number of Sampling Locations
At least 10 locations should be used for tumbling mixers to adequately map blender
At least 20 locations should be used for convection mixers, which are more likely to have dead spots
Replicates Per Location
At least 3 samples/location required to perform component variance analysis to detect the presence of sampling error

27. Justification of Dosage Unit Sample Sizes and Acceptance Criteria
Number of dosage unit samples and sample size through the use of OC curves, considering:
Weight variability
Assay variability
Between location error
Within location error
USP Content Uniformity Test used as a reference for comparison

30. Justification of Dosage Unit Acceptance Criteria
RSD  6.0%
Consistent with Stage 1 USP Test
All location means % target potency
Detects drifting/segregation or non-uniform spots in the blend
All individuals within % target potency
Will pick up outliers, such as subpotent or superpotent (agglomeration) dosage units

31. Justification of Dosage Unit Acceptance Criteria
Two stage test is consistent with USP Content Uniformity Test
Stage 1 and Stage 2 criteria are the same
Stage 2 test offers a second opportunity to comply with acceptance criteria, for those batches which barely fail Stage 1 testing

32. Routine Manufacture

33. Merging the cGMP Requirement with Compendial Release Testing
Dosage units to be tested are in-process samples
Perform two calculations on a single set of data
cGMP Compliance - Normalize for weight
Compendial Testing - No weight correction
Acceptance criteria the same as that described in the USP Content Uniformity Test
If the in-process sample is not the finished dosage form, you must demonstrate during validation that the in-process results provide the same or better control as the content uniformity data generated during compendial release testing of the corresponding finished dosage units.

34. Definition of “Readily Complies” and Impact on Degree of Testing Required
“Readily Comply” is demonstrated if for each ANDA exhibit and/or validation batch:
RSD  4.0%, all mean results within 90.0 – 110.0%, all individual results between 75.0 – 125.0%
Stage 1 testing allowed (10 dosage units)
Testing for products that do not “readily comply”
Stage 2 testing (30 dosage units) for at least 5 batches
If after testing 5 consecutive batches, the criteria for the mean is met and the RSD routinely is  5.0%, then Stage 1 testing is allowed

35. Routine Manufacture For ANDA exhibit and/or validation batches:
RSD  4.0%, all mean results
90-110%, all values between %
Yes [“Readily Complies”]
No [Does not “Readily Comply”]
Stage 1: Test 1 sample/location
mean %, RSD  5.0%
Stage 2: Test 3 samples/location
mean %, RSD  6.0% No
Yes
Yes No
Adequacy of mix demonstrated;
perform 2nd calculation to satisfy
compendial release requirements
Adequacy of mix demonstrated;
perform 2nd calculation to satisfy
compendial release requirements
Adequacy of mix
not demonstrated
Future
lots
After passing 5
Consecutive Batches

36. Justification of cGMP Compliance Sample Sizes and Acceptance Criteria
Sample Size: At least 10 locations, 3 dosage units per location
Consistent with the USP Content Uniformity Test
cGMP Acceptance Criteria: RSD  5.0% and mean of all samples between % target potency
Consistent with FDA Validation Acceptance Criteria for demonstrating adequacy of mix for powder blends

37. Alternative Approaches
BUWG recommendation is one approach for evaluation of adequacy of mix
The cGMP requirements are open to other approaches
On-line monitoring techniques such as NIR
PDA 25 approach
Traditional methods (direct sampling/analysis of blend sample)

38. Results of PQRI Datamining Effort

39. Objectives of Datamining Effort
Test the hypothesis “blend uniformity is not value added testing”
Test the assumption that means are normally distributed
Validate the use of computer simulated data
Subject batches to PQRI, OGD, FDA Validation, PDA 25, USP, and modified USP (ICH) acceptance criteria

40. Summary of Data Analyzed
Desired Categories of Data
Active ingredient < 5% and between 15-25%
Product made using direct compression and granulation processes (either wet or dry)
Data for tablets and capsules
Commercial batches both small ( kg) and large (>400 kg)
8 companies submitted 149 batches

41. Characteristics of Submitted Data
Dosage Form
Tablets: 149
Capsules: 0
Manufacturing Process
Direct Comp: 12
Wet Granulation: 67
Dry Granulation: 70

42. Test for Normality of Means
Tested both location and within location means for normality using the Wilk-Shapiro test for normality
Location: ~ 11% of batches had at least 1 value that was statistically different
Most were at beginning/end of run
Within Location: ~15% of batches had at least 1 value that was statistically different
Conclusion: Computer simulations to estimate criteria rejection rates yield slightly smaller values (conservative) than reject rates based on actual data

43. Comparison of Blend and Dosage Unit Content Uniformity Data
Primary means to test they hypothesis “blend uniformity testing is not value added”
Plots prepared comparing dosage unit RSD as a function of blend RSD
Break the curve down into 3 zones:
Blend RSD <3%
Blend RSD 3-5%
Blend RSD > 5%

44. Comparison of Blend and Dosage Form RSDs

45. Blend RSD < 3%

46. Blend RSD 3-5%

47. Blend RSD >5%

48. Correlation Between Blend and Dosage Unit RSD
Blend RSD < 3%: Blend data is predictive of final dosage form uniformity
Dosage form RSD often higher (weight variability, segregation?)
Blend RSD 3-5%: Diminished correlation between blend data & dosage form uniformity
Blend RSD >5%: Blend data is not predictive of content uniformity of the final dosage form

49. Comparison of Acceptance Criteria

50. Datamining Results: “Readily” vs. “Marginally” Comply
83/88 (94%) passed PQRI Validation acceptance criteria
Of the batches that met PQRI Validation acceptance criteria
Readily Comply: 79/83
Marginally Comply: 4/83

51. Acknowledgements Jerry Planchard (Aventis) Garth Boehm (Purepac)
Joep Timmermans (Merck)
Jerry Mergen (McNeil)
Fernando Muzzio (Rutgers)
Jean-Marie Geoffroy (Abbott)
Jim Prescott (Jenike & Johanson)
Pedro Jimenez (Lilly)
John Dietrick (FDA)
Jon Clark (FDA)
Neeru Takiar (FDA)
Muralidhara Gavini (FDA)
Laura Foust (Lilly)