Statistical Approaches to Addressing the Requirements of the New FDA Process Validation Guidance for Small Molecules

1
Statistical Approaches to Addressing the
Requirements of the New FDA Process Validation
Guidance for Small Molecules

• Jason Marlin, MS/T Statistics, Eli Lilly Co.

2
Key Takeaways from The FDA Guidance for Industry,
Process Validation General Principles and
Practices, January 2011

• the level of sampling and testing in validation
  must be sufficient to confirm uniform product
  quality throughout the batch during processing
• the level of sampling and testing in validation
  must provide statistical assurance that the
  process is reproducible and consistently delivers
  quality products
• Considerations
• 1A statisticians review is required for approval
  of a Process Validation protocol to ensure the
  testing achieves the above requirements
• Statistical assurance statements
• 1The FDA Guidance (footnote reference 1)
  recommends this in Stage 3 We recommend that a
  statistician or person with adequate training in
  statistical process control techniques develop
  the data collection plan and statistical methods
  and procedures used in measuring and evaluating
  process stability and process capability.

3
Reality

• Limited statistical resources
• Many PV protocols
• World-wide sites
• Need for general PV requirements.

4
What we did, in general

• Readily Pass and Marginally Pass terminology are
  employed
• The terms Readily Pass and Marginally Pass were
  introduced in the FDA draft Guidance for Industry
  Powder Blends and Finished Dosage Units --
  Stratified In-Process Dosage Unit Sampling and
  Assessment, published October 2003.

5
Attributes addressed by this work

• UDU
• Potency
• Purity
• Water Content
• Release Profile
• Appearance

6
UDU

• Unit Dose Uniformity is confirmed by two types of
  data in a PV.
• Blend Uniformity
• Potency samples are obtained to demonstrate
  uniformity of the blend for tablet and
  powder-filled capsules.
• Drug Product Dosage Form Uniformity
• Dosage unit samples are collected using an
  appropriate stratified sampling method.

7
Blend Uniformity Weight-Corrected Potencies

• Used to demonstrate blend uniformity
• Follow PQRI guidance (readily marginally pass)

8
UDU Sample Collection and Testing

• n 60 Dosage Units
• 3 dosage units from each of 20 stratified
  locations (pull at least 7 per location)
• Unit dose samples are tested for potency with
  corresponding individual weights recorded.
• Weight corrected potency recorded
• Non-weight corrected potency recorded

9
UDU Criteria

• Readily Pass - meet Bergums criteria
• Will be ASTM standard soon (E2709)
• Test provides criteria for the Mean and Maximum
  Allowable Standard Deviation of the 60 Dosage
  Units (reported in of Label Claim)
• If criteria met, we have 95 confidence that the
  UDU lt905gt criteria would be met at least 95 of
  the time for the batch in question
• Marginally pass
• 100 - Mean 2.4 s lt 15.0 , RSD
  (wt-corrected)6.0.

10
Output from Bergums CuDAL Simulation
11
Potency

• Release Limits accounting for change on stability
  and assay variability must be established
• To Readily Pass potency
• Must meet the Marginally Pass criteria (see
  below),
• Using potency data from Non-Weight Corrected
  Potency UDU results, assess within-batch and
  between-batch capability.
• Combine all the mean non weight-corrected potency
  results from each location within and across
  validation batches (NLT 20 location means per
  batch for CU products).
• For 3-batch validations of standard CU products,
  this should be 60 or more location means.
• Compute the Cpk for the location means, compared
  to the shelf-life registered criteria and the Cpk
  of the location means must be NLT 1.0
• To Marginally Pass potency
• the composite test result must meet the
  corresponding Release Limits

12
Purity and Water

• Release Limits accounting for change on stability
  and assay variability must be established
• To Pass (no distinction between Readily and
  Marginally Pass)
• test results must meet the corresponding Release
  Limits

13
Dissolution

• Three Dissolution runs (B/M/E) of at least n6
  Dosage Units per location are performed for each
  validation batch.
• Readily pass
• Pass the marginally pass criteria
• Pass Bergums 95/95 criteria for dissolution mean
  and SD
• Marginally pass
• All dissolution runs at each location pass
  Dissolution Specification by Stage 2

14
Appearance

• Select a random sample from the batch or
  subsection, per the corresponding plan medium
  risk (n800) or high risk (n1250). The
  acceptance criteria (for both medium and high
  risk plans) are provided in the following table
• We specify and control consumer risk , LQL
• Individual sample plans are selected from
  ANSI/ASQ Z1.4-2003, indexed by the AQL, with LQL
  implied.

Defect classification Accept/reject
Critical 0/1
Major 7/8
Minor 21/22
15
Additional Challenges

• Statistical reviews of all PV documents?
• Stage 2 to Stage 3 Transitionwhat constitutes
  sufficient post- validation monitoring?

16
Conclusions

• Increased scrutiny of validation protocols and
  of validation data
• Still some uncertainty on post-Stage II
  monitoring
• Three validation batches are the standard
• Within batch variation is well understood
• Between batch variation requires time element

17

18
PQRI Guidance

• Marginally Pass Criteria
• All results within 75.0-125.0 of target mg/mg
• All location averages are within 90.0-110.0 of
  target mg/mg
• RSD of weight-corrected potencies not more than
  6.0
• Readily Pass Criteria
• All results within 75.0-125.0 of target mg/mg
• All location averages are within 90.0-110.0 of
  target mg/mg
• RSD of weight-corrected potencies not more than
  4.0