johnson mark ispe pv samp acc crit ppq (johnson 12 sep · pdf fileispe process validation...

12
ISPE Process Validation Conference 12 – 14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria for Process Performance Qualification Helen Strickland (GSK): Principal Author Mark Johnson (AbbVie): Presenter 12 September 2017 Connecting Pharmaceutical Knowledge ispe.org Outline Sampling Concepts Populations & Samples Simple Random Sampling Stratified Random Sampling Process Validation Lifecycle Flowchart Sampling Criteria Sampling Scheme and Control Strategy AQL & LQ Sampling Plans Sampling & OC Curves Example Summary 2

Upload: vohanh

Post on 01-Feb-2018

227 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

1

ISPE PV

Sampling and Acceptance Criteria for Process Performance Qualification

Helen Strickland (GSK): Principal AuthorMark Johnson (AbbVie): Presenter12 September 2017

Connecting Pharmaceutical Knowledge ispe.org

Outline• Sampling Concepts

• Populations & Samples• Simple Random Sampling• Stratified Random Sampling

• Process Validation Lifecycle

• Flowchart• Sampling Criteria• Sampling Scheme and Control Strategy

• AQL & LQ Sampling Plans

• Sampling & OC Curves

• Example

• Summary

2

Page 2: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

2

Connecting Pharmaceutical Knowledge ispe.org

Sampling Concepts: Populations & Samples

Population—totality of items under consideration

• Single lot from a process• Series of lots from a process• Series of lots from future output of a process

Sample—miniature representation of population

• Representative sampling: randomly sample such that the observed values have the same distributions in the sample as the population

• Random sampling: each possible combination of ‘n’ sampling units has particular probability of being selected

• Sampling scheme should ensure correct selection probabilities—that is all possible samples have an equal opportunity of being included in the sample selected for testing.

• A sampling scheme with incorrect selection techniques lacks reliability of obtaining a representative sample.

3

Connecting Pharmaceutical Knowledge ispe.org

• A random selection is collected from a population

• Each unit in population has same probability of selection

Sampling Concepts: Simple Random Sampling

Sample

X X X X X X X X XX X X X XX

X X X XX X X X

X X X X X X X X X X X X

X X X X X

X X X X X X X XX X X X

Population

5

Page 3: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

3

Connecting Pharmaceutical Knowledge ispe.org

Sampling Concepts: Simple Random Sampling

It is possible – but unlikely – that a random sample could be pulled entirely from one spot

Sample

X X X X X X X X XX X X X XX

X X X XX X X X

X X X X X X X X X X X X

X X X X X

X X X X X X X XX X X X

Population

6

Connecting Pharmaceutical Knowledge ispe.org

Sampling Concepts: Stratified Random Sampling

• What does stratified sampling do?

– Assures you get units from each subgroup

• When would this be important?

– To check identifiable (or potential) subgroups within the batch / process: top/middle/bottom of container; every hour; every fill nozzle, etc.

– To prove we do NOT have uniformity problems, by sampling from higher-risk physical locations or time points (worst case)

• How is this done?

– Divide the population into “strata” or subpopulations

– Take samples randomly from each stratum in the population

6

Page 4: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

4

Connecting Pharmaceutical Knowledge ispe.org

Sampling Concepts: Stratified Random Sampling

Random samples are taken from each stratum / subgroup

In this example, the first and last units are also selected

Scientific reasons, not purely statistical

SamplePopulation with strata

X X XY Y Y Y Y YZ

X XX X X X X X X XY Y Y Y Y

w W W W W WW W W W W W

Y Y zZZZZZZZZZZZ

8

Connecting Pharmaceutical Knowledge ispe.org

Process Validation Lifecycle: FlowchartPD

Preliminary estimates of nonconforming rates of Critical Quality Attributes (CQA) (Lagging Indicator)

Technical Linkages of Critical Process Parameters (CPP) & Critical Material Attributes (CMA) to CQAs (Leading Indicators)

Some Acceptance Sampling-Pilot

PPQ

Isolated

Heightened Sampling: CQAs’ conformance to specification…

Heightened Monitoring: CPPs & CMAs in state of control…

Focus >> Intra-batch

CPV 3A

Isolated

Heightened Sampling: CQAs, but reduction justifiable…

Heightened Monitoring: CPPs & CMAs in state of control…

Focus >> Inter-Batch

CPV 3B

Continuing Series of Lots – process quality consistent, maintaining control state…

Reduced sampling, but revert back to heightened sampling if acceptance criteria not met, to determine if product quality has deteriorated to unacceptable level

8

Page 5: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

5

Connecting Pharmaceutical Knowledge ispe.org

Process Validation Lifecycle: Sampling Criteria

Routine Commercial ManufacturePD PPQ

DOE requires heightened sampling to build relationships between CPP & CQA

Verification/confirmatorysampling should illustrate control strategy sufficiently supports process performance & CQA estimates

Acceptance sampling confirms our product conforms to specification & meets our process assumptions

Initial Cml Mfg

9

Connecting Pharmaceutical Knowledge ispe.org

Traditional Control Strategy >> Population: Process >> Subgroup: Sample Each Batch

(retrospective assessment)

After the event sampling for product control

Statistical process control techniques applied to lagging indicators of quality (i.e., CQAs) –

Acceptance sampling (a.k.a. End product testing) used to assess conformance to specification

CQAs are output quality characteristics

Adjustments made to process based on sampling output—lagging indicators

Detection of product quality drift

Process Validation Lifecycle: Sampling Scheme & Control Strategy

Enhanced Control Strategy >> Population: Process >> Subgroup: Sample Each Batch

(proactive assessment)

Statistical process control techniques applied to leading indicators of quality (CMAs and CPPs) -

Acceptance sampling used as adjunct to verify conformance to specification

Output quality characteristics of a previous unit operation may be input characteristics of subsequent unit operation

Adjustments made to process based on relationship of CMA s and /or CPP s to CQAs

Detect/prevent process drift before product quality drift

10

Page 6: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

6

Connecting Pharmaceutical Knowledge ispe.org

AQL and LQ Sampling Plans, Sampling & OC CurvesActual Batch or Lot Quality

Producer’s Risk Quality (PRQ)

Consumer’s Risk Quality (CRQ)

Out

com

e

Acceptance Sampling

Plan’s Acceptance Criteria Are

Met

Producer’s Confidence

Proportion of times a batch or lot will meet

the acceptance sampling plan’s

acceptance criteria when the batch’s

quality level is equal to PRQ.

Consumer’s RiskProportion of times a batch or lot will meet

the acceptance sampling plan’s

acceptance criteria when the batch’s quality

level is equal to CRQ (β=0.1).

Acceptance Sampling

Plan’s Acceptance Criteria Are

Not Met

Producer’s RiskProportion of times a batch or lot will not

meet the acceptance sampling plan’s

acceptance criteria when the batch’s

quality level is equal to PRQ (α=0.05).

Consumer’s Confidence

Proportion of times a batch or lot will not

meet the acceptance sampling plan’s

acceptance criteria when the batch’s quality level is equal to CRQ.

0 2 4 6 8 10

0.0

0.2

0.4

0.6

0.8

1.0

HIGH probability means lots will typically be found acceptable.

LOW probability means lots will typically be found unacceptable.

Metrics on X axis may be:%RSD, Distance of the Mean from Target

Probability of Passing Criteria

AQL=0.95

LQ=0.10

11

Acceptance Quality Limit (AQL) – Quality level a sampling plan accepts with high probability (e.g., 95%; α=0.05)

Limiting Quality (LQ) – Quality level a sampling plan rejects with high probability (e.g., 90%; β=0.1) (Also RQL or LTPD)

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity Syringe after Filling with Drug Product

General Information Regarding Product Quality Characteristic

Product Quality Characteristic

Functional Appearance of Syringe Needle.

Nonconformity Classification PDA-43: Major A, Deformed Needle Point

Nonconformity that leads to serious impairments, therefore, considered a CQA.

CQA is treated as coming from a binomial distribution as items are countable and quality characteristic is classifiable or measurable as being conforming or nonconforming.

Process Design: Process Control Strategy for CQA: Process controls address variability to ensure product quality: examination of materials and equipment. Full range of input variability not typically known during PD—Use process knowledge and understanding to predict contributions of input variations.

CPPs and CMAs related to CQA

Supplier audit indicates syringe manufacturing process operates in a state of control with a process average where the syringe defect rate is less than 0.01% (100 per 1 million).

Technical linkages between CPPs and CQA (e.g. functional appearance of syringe) were established during PD.

SPC methods are utilized for the CPPs during the syringe filling operation for each batch.

SQC methods are utilized for CQA, acceptance sampling plan is used demonstrate conformance to specification.

12

Page 7: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

7

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity Syringe after Filling with Drug Product

PPQ Data Collection

Leading Process Based Indicators (CPPs and CMAs)

Enhanced sampling and testing performed on CMAs

CPP data such as force required to fill syringe electronically collected.

Lagging Product Based Indicators (CQAs)

Stratified Random Sampling

10 stratum with 15,000 syringes filled per stratum 3 Batches for PPQ Protocol 30 stratum

PPQ Data Evaluation

CPPs

CMAs

Retrospective control charting performed on electronic capture of CPP data to assess within batch state of control

Shewhart Xbar-S Chart Boxplot/Scatter Plots Analysis of Variance Methods

13

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity Syringe after Filling with Drug Product

Isolated Lot Acceptance Sampling Plan Indexed on LQ of 0.5% (ISO 2859-2 (1985))

Inspect 800 syringes (80 per stratum), Accept ≤ 1, Reject ≥ 2 Deformed Syringe Needle Points

CR ≤ 9.1% with CRQ ≥ 0.50 % non-conforming & PR ≤ 5.0% when PRQ ≤ 0.044% non-conforming

(n=800, a=1 plan: CR=5% at CRQ=0.59%)

Batch must contain less than 0.044% deformed syringes (non-conforming syringes produced randomly and at a constant rate) to have at least a 95% probability of passing n=800, a=1 plan.

Process must produce less than 0.0188% deformed syringes to have at least a 95% probability of 5 consecutive batches passing n=800, a=1 plan.

14

Page 8: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

8

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity Syringe after Filling with Drug Product

Customized Acceptance Sampling Plan, Accept Number = 0

Inspect 600 syringes (60 per stratum), Accept = 0, Reject ≥ 1 Deformed Syringe Needle Points

CR ≤ 5.0% with CR Quality ≥ 0.498 % non-conforming & PR ≤ 5.0% when PR Quality ≤ 0.0086% non-conforming

Batch must contain less than 0.0086% deformed syringes (non-conforming syringes produced randomly and at a constant rate) to have at least a 95% probability of passing n=600, a=0 plan.

Process must produce less than 0.0017% deformed syringes to have at least a 95% probability of 5 consecutive batches passing n=600, a=0 plan.

15

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity Syringe after Filling with Drug Product

Customized Acceptance Sampling Plan, Accept Number ≤ 1

Inspect 950 syringes (95 per stratum), Accept ≤ 1, Reject ≥ 2 Deformed Syringe Needle Points

CR ≤ 5.0% with CRQ ≥ 0.498 % non-conforming & PR ≤ 5.0% when PRQ ≤ 0.037% non-conforming

Batch must contain less than 0.037% deformed syringes (non-conforming syringes produced randomly and at a constant rate) to have at least a 95% probability of passing n=950, a=1 plan.

Process must produce less than 0.0158% deformed syringes to have at least a 95% probability of 5 consecutive batches passing n=950, a=1 plan.

16

Page 9: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

9

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity, Syringe after Filling with Drug Product

PPQ: confirm process design and demonstrate commercial manufacturing process performs as expected (FDA, 2011)

Two hypothetical scenarios can illustrate how the PPQ results based on option 3 may be summarized and used to determine the data collection and evaluation strategies for CPV3A and CPV3B.

1) All five batches met the functional appearance acceptance criteria stated in the PPQ protocol.

a) Scenario A: Zero defective syringes were observed in each of the five PPQ batchesb) Scenario B: One defective syringe was observed in each of the five PPQ batches.

2) Statistical evaluation of the leading indicator data (CPPs and CMAs) related to the functional appearance of the syringe needle indicates that the process met the conditions established in the PPQ protocol, and support the conclusion that the process was operated in a state of control during the manufacturing of the five PPQ batches.

17

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity, Syringe after Filling with Drug Product

PPQ: confirm process design and demonstrate commercial manufacturing process performs as expected (FDA, 2011)

3) Conclusion of a constant and common defect rate for each batch was supported by the results of the retrospective evaluation of the CPPs and CMAs. The acceptance sampling results obtained on the 5 PPQ batches were pooled to determine preliminary estimates of the process defect rate under each scenario.

a) Under Scenario A, the 95%, 50% and 5% one-sided upper confidence bounds on the true percent nonconforming was computed where no defective syringe needles were observed in a total of 4,750 syringes inspected.

1) 95% UCB on the true proportion nonconforming of the sampled process: 0.063%2) 50% UCB on the true proportion nonconforming of the sampled process: 0.015%3) 5% UCB on the true proportion nonconforming of the sampled process: 0.001%

18

Page 10: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

10

Connecting Pharmaceutical Knowledge ispe.org

Example: Major A Nonconformity Syringe after Filling with Drug Product

PPQ: confirm process design and demonstrate commercial manufacturing process performs as expected (FDA, 2011)

Under Scenario B, the 95%, 50% and 5% one-sided upper confidence bounds on the true percent nonconforming was computed where a total of 5 defective syringe needles (1 per batch) were observed in a total of 4,750 syringes inspected where one defective syringe was observed in each of the five PPQ batches, (n=4,750 and a=5)

95% UCB on the true proportion nonconforming of the sampled process: 0.22% 50% UCB on the true proportion nonconforming of the sampled process: 0.12% 5% UCB on the true proportion nonconforming of the sampled process: 0.055%

19

Connecting Pharmaceutical Knowledge ispe.org

Summary

20

• Samples from the process population may be a single lot, series of lots or projected number of future lots

• An effective sampling strategy is required, to assure the sample is representative of the various population distributions and that all possible units have an equal likelihood of inclusion in the test sample

• Sampling plan and inspection strategies vary across the process validation lifecycle from product development through continued process verification activities

• Development of an effective sampling scheme & control strategy in the process validation lifecycle creates better understanding of relationships between CMAs and CPPs and their impact on CQAs that are usable for establishing PPQ sampling and acceptance criteria requirements

Page 11: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

11

Connecting Pharmaceutical Knowledge ispe.org

Backup Slides

21

Connecting Pharmaceutical Knowledge ispe.org

Sampling Concepts: Types of Sampling

Simple Random Sampling – Selecting samples so that each unit has an equal chance of being selected

Stratified Random Sampling – Selecting samples deliberately from each time period or location in a batch

Nested Sampling – Selecting units from locations within a batch and obtaining multiple samples from each location

Systematic sampling – Selecting units periodically over time

Convenience Sampling – Selecting samples based on their ease of access or taken as convenience, opportunity, or expertise permits

Tail-Gate Sampling – Selecting pre-delivery samples for the customer's incoming quality inspection

22

Page 12: Johnson Mark ISPE PV Samp Acc Crit PPQ (Johnson 12 Sep · PDF fileISPE Process Validation Conference 12 –14 September 2017 Bethesda, MD 1 ISPE PV Sampling and Acceptance Criteria

ISPE Process Validation Conference12 – 14 September 2017

Bethesda, MD

12

Connecting Pharmaceutical Knowledge ispe.org

1. Well-defined product requirements

2. Valid and capable measurement methods

3. Understand relationship CMAs, CPPs to CQAs

4. Monitoring strategy

a) how to constitute a sample/subgroup

b) how many observations,

c) frequency

5. Select control chart type

6. Determine out-of-control action plan

(From “10 Requirements for Effective Process Control: A Case Study”, Thomas Little, Quality Progress, February 2001, pp 46-52)

Process Validation Lifecycle: Process Control Implementation Strategy

• Leading indicators—process based

• Lagging indicators—product based

• Single streams versus multiple streams

• Collapsing multiple input and output streams into one measurement requires experimental sampling scheme to estimate between and within variation to establish sampling plan

• Process behavior over time for individual batch?

• Process behavior over time between batches?

• Instantaneous distributions, resulting distributions

• Did we meet all PV requirements sufficiently to reduce sampling?

• Often there are qualitative differences between process validation conditions & routine operational conditions, how will this affect ongoing process quality?

23

Connecting Pharmaceutical Knowledge ispe.org

AQL and LQ Sampling Plans for Attributes: ISO 2859

Sampling procedures for inspection by attributes — Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection (1999)

AQLs for sampling = 0.01%, 0.015%, 0.025%, 0.040%, 0.065%, 0.10%, 0.15%, 0.25%, 0.40%, 0.65%, 1.0%, 1.50%, 2.0%, 2.5%, 4.0%, 6.5%, 10.0%

Sampling procedures for inspection by attributes – Part 2: Sampling plans indexed by limiting quality (LQ) for isolated lot inspection (1985)

LQs for sampling = 0.5%, 0.8%, 1.25%, 2.0%, 3.15%, 5.0%, 8.0%, 12.5%, 20%, 32%

24