Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Best Practices for OINDP Pharmaceutical Development Programs Leachables and Extractables
VIII. Quality Control and Specification Setting
PQRI Leachables & Extractables Working Group
PQRI Training CourseApril 12-13, 2007
Chicago, IL
April 2007 PQRI Training Course 2
Definition ReviewDefinition Review
►► A A LeachablesLeachables StudyStudy is a laboratory investigation into the is a laboratory investigation into the qualitative and quantitative nature of a particular OINDP qualitative and quantitative nature of a particular OINDP leachablesleachables profile(sprofile(s) over the proposed shelf) over the proposed shelf--life of the life of the product. Supports:product. Supports:§§ Developing an Developing an extractables/leachablesextractables/leachables correlationcorrelation§§ Establishment of drug product Establishment of drug product leachablesleachables acceptance criteria.acceptance criteria.
►► Routine Routine ExtractablesExtractables Testing Testing is the testing by which OINDP is the testing by which OINDP container closure system critical components are container closure system critical components are qualitatively and quantitatively profiled for qualitatively and quantitatively profiled for extractablesextractables, , for:for:§§ Establishing Establishing extractablesextractables acceptance criteriaacceptance criteria§§ Release by established acceptance criteria.Release by established acceptance criteria.
April 2007 PQRI Training Course 3
Control of Control of LeachablesLeachables Through Through Control of Control of ExtractablesExtractables
►► Specifications and acceptance criteria are required Specifications and acceptance criteria are required for for leachablesleachables profiles in OINDP.profiles in OINDP.
►► Implementation of routine Implementation of routine leachablesleachables testing and testing and specifications/acceptance criteria is a policy specifications/acceptance criteria is a policy matter.matter.
►► If If extractables/leachablesextractables/leachables correlations can be correlations can be established, then established, then leachablesleachablesspecifications/acceptance criteria may be specifications/acceptance criteria may be established as “established as “if tested will comply”if tested will comply”..
►► Therefore, in the ideal situation Therefore, in the ideal situation leachablesleachables can be can be controlled through routine testing of controlled through routine testing of extractablesextractables..
April 2007 PQRI Training Course 4
Routine Routine ExtractablesExtractables TestingTestingPerformed on all critical components of OINDP container Performed on all critical components of OINDP container
closure systems with following general goals:closure systems with following general goals:
►► To establish To establish extractablesextractables acceptance criteria for OINDP acceptance criteria for OINDP critical container closure system components.critical container closure system components.
►► To help ensure that the To help ensure that the leachablesleachables profile in the drug profile in the drug product is maintained within appropriate limits.product is maintained within appropriate limits.
►► To release OINDP container closure system critical To release OINDP container closure system critical components according to established acceptance criteria, components according to established acceptance criteria, which are designed to:which are designed to:§§ Confirm the identities and levels of known Confirm the identities and levels of known extractablesextractables;;§§ Detect “unspecified” Detect “unspecified” extractablesextractables which could be present as the which could be present as the
result of component ingredient changes, manufacturing changes, result of component ingredient changes, manufacturing changes, external contamination, or other causes.external contamination, or other causes.
April 2007 PQRI Training Course 5
Recommendations for Routine Recommendations for Routine ExtractablesExtractables TestingTesting
►► Analytical methods for Routine Analytical methods for Routine ExtractablesExtractablesTesting should be based on the analytical Testing should be based on the analytical technique(s)/method(stechnique(s)/method(s) used in the Controlled ) used in the Controlled Extraction Studies. Consider the following:Extraction Studies. Consider the following:
§§ Simplicity relative to R&D methodsSimplicity relative to R&D methods§§ Ruggedness and robustnessRuggedness and robustness§§ TransferabilityTransferability§§ Cost effectivenessCost effectiveness
April 2007 PQRI Training Course 6
Transition from Transition from ExtractablesExtractables Studies Studies Method to Routine QC Method to Routine QC ExtrablesExtrables TestingTesting
5.00 10.00 15.00 20.00 25.00 30.00 35.000
200000
400000
600000
800000
1000000
1200000
1400000
1600000
1800000
2000000
2200000
2400000
2600000
2800000
3000000
3200000
3400000
Time-->
Abundance
TIC: 07300307.D
min5 10 15 20 25 30 35
pA
0
50
100
150
200
250
300
350
400
GC/MS GC/MS ExtractablesExtractables Profile of an ElastomerProfile of an Elastomer
min0 2 4 6 8 10 12 14 16 18
mAU
0
25
50
75
100
125
150
175
DAD1 A, Sig=200,4 Ref=550,100 (I:\HPCHEM\1\DATA\022569\022569\JAN31008.D)
2-propanol (Reflux)
Routine Routine ExtractablesExtractables Method Method –– GC/FIDGC/FID
Routine Routine ExtractablesExtractables Method Method –– HPLC/UVHPLC/UV
Development MethodDevelopment Method
Quality Control Quality Control Method(sMethod(s))
April 2007 PQRI Training Course 7
Routine Routine ExtractablesExtractables TestingTesting-- Method Method Development and ValidationDevelopment and Validation--ReferencesReferences
1.1. ICH ICH HarominzedHarominzed Tripartite Guideline, “Text on Validation of Tripartite Guideline, “Text on Validation of Analytical Procedures Q2A”, International Conference on Analytical Procedures Q2A”, International Conference on Harmonization of Technical Requirements for Registration of Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use.Pharmaceuticals for Human Use.
2.2. ICH ICH HarominzedHarominzed Tripartite Guideline, “Validation of Analytical Tripartite Guideline, “Validation of Analytical Procedures: Methodology Q2B”, International Conference on Procedures: Methodology Q2B”, International Conference on Harmonization of Technical Requirements for Registration of Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use.Pharmaceuticals for Human Use.
3.3. “Reviewer Guidance “Reviewer Guidance –– Validation of Chromatographic Methods”, Validation of Chromatographic Methods”, Center for Drug Evaluation and Research (CDER), United States Center for Drug Evaluation and Research (CDER), United States Food and Drug Administration, November, 1994.Food and Drug Administration, November, 1994.
4.4. “Guidance for Industry “Guidance for Industry –– Analytical Procedures and Methods Analytical Procedures and Methods Validation Validation –– Chemistry, Manufacturing, and Controls Chemistry, Manufacturing, and Controls Documentation”, Documentation”, Draft GuidanceDraft Guidance, Center for Drug Evaluation and , Center for Drug Evaluation and Research (CDER), United States Food and Drug Administration, Research (CDER), United States Food and Drug Administration, August, 2000.August, 2000.
5.5. Michael E. Swartz and Ira S. Michael E. Swartz and Ira S. KrullKrull, , Analytical Method Development Analytical Method Development and Validationand Validation, Marcel , Marcel DekkerDekker, Inc., New York, 1997., Inc., New York, 1997.
April 2007 PQRI Training Course 8
Routine Routine ExtractablesExtractables TestingTesting-- Method Development Method Development and Validationand Validation
►► Extraction procedures for critical components Extraction procedures for critical components should be based on the optimized procedures from should be based on the optimized procedures from the quantitative Controlled Extraction Studiesthe quantitative Controlled Extraction Studies§§ Demonstrate asymptotic levels of Demonstrate asymptotic levels of extractablesextractables..
►► The linear dynamic range of the analytical method The linear dynamic range of the analytical method should be established based on levels of should be established based on levels of extractablesextractables anticipated from quantitative anticipated from quantitative Controlled Extraction StudiesControlled Extraction Studies
►► The LimitThe Limit--ofof--QuantitationQuantitation of the method should be of the method should be established with consideration of the appropriate established with consideration of the appropriate AET.AET.
April 2007 PQRI Training Course 9
Routine Routine ExtractablesExtractables TestingTesting-- Method Method Development and Validation (cont.)Development and Validation (cont.)
►► Method validated according to the ICH validation Method validated according to the ICH validation characteristics of a quantitative impurity test, characteristics of a quantitative impurity test, §§ Include: Accuracy, Precision (Repeatability, Intermediate Include: Accuracy, Precision (Repeatability, Intermediate
Precision), Specificity, LimitPrecision), Specificity, Limit--ofof--QuantitationQuantitation (LOQ), Linearity, and (LOQ), Linearity, and Range. Range.
§§ System Suitability parameters should be established System Suitability parameters should be established §§ Robustness should be evaluatedRobustness should be evaluated§§ Note that in certain cases it may be appropriate to validate rouNote that in certain cases it may be appropriate to validate routine tine
extractablesextractables methods as “Limit Tests”, in which case only methods as “Limit Tests”, in which case only Specificity and LimitSpecificity and Limit--ofof--Detection (LOD) need be considered.Detection (LOD) need be considered.
►► Accuracy can be determined through the analysis of spiked Accuracy can be determined through the analysis of spiked samples.samples.§§ Spiking matrix could be an extract taken through the extraction Spiking matrix could be an extract taken through the extraction
procedure minus the component sample.procedure minus the component sample.§§ Spiking levels should be chosen so as to be representative of Spiking levels should be chosen so as to be representative of
anticipated anticipated extractablesextractables levels based on results from quantitative levels based on results from quantitative Controlled Extraction Studies.Controlled Extraction Studies.
April 2007 PQRI Training Course 10
Specifications and Acceptance Specifications and Acceptance Criteria for Criteria for LeachablesLeachables
►► LeachablesLeachables specifications should include a fully specifications should include a fully validated analytical test method. validated analytical test method.
►► Acceptance criteria for Acceptance criteria for leachablesleachables should apply should apply over the proposed shelfover the proposed shelf--life of the drug product, life of the drug product, and should include:and should include:§§ Quantitative limits for known drug product Quantitative limits for known drug product leachablesleachables
monitored during product registration stability studies.monitored during product registration stability studies.§§ A quantitative limit for “new” or “unspecified” A quantitative limit for “new” or “unspecified” leachablesleachables
not detected or monitored during product registration not detected or monitored during product registration stability studies.stability studies.
April 2007 PQRI Training Course 11
Specifications and Acceptance Criteria for Specifications and Acceptance Criteria for LeachablesLeachables
►► Quantitative acceptance criteria should be based Quantitative acceptance criteria should be based on safety considerations as outlined in the ‘L&E on safety considerations as outlined in the ‘L&E Best Practices‘Best Practices‘§§ Actual Actual leachablesleachables levels, and trends in levels, and trends in leachablesleachables levels, levels,
observed over time and across various storage observed over time and across various storage conditions and drug product orientations during product conditions and drug product orientations during product registration stability studies should be considered.registration stability studies should be considered.
►► Ability to consistently meet should be established Ability to consistently meet should be established with appropriate statistical analysis.with appropriate statistical analysis.
April 2007 PQRI Training Course 12
Specifications and Acceptance Criteria for Specifications and Acceptance Criteria for LeachablesLeachables
►► Comprehensive correlation should obviate the need for Comprehensive correlation should obviate the need for routine implementation of drug product routine implementation of drug product leachablesleachablesspecifications and acceptance criteria, assuming:specifications and acceptance criteria, assuming:§§ Adequate information from critical component suppliersAdequate information from critical component suppliers§§ Understanding and control of critical component Understanding and control of critical component
fabricationfabrication§§ Controlled Extraction Studies on critical components.Controlled Extraction Studies on critical components.§§ Validated Validated leachablesleachables methods and a methods and a LeachablesLeachables Study.Study.§§ Validated Routine Validated Routine ExtractablesExtractables Testing methods and Testing methods and
database of critical component database of critical component extractablesextractables profiles.profiles.§§ Appropriate specifications and acceptance criteria for Appropriate specifications and acceptance criteria for
extractablesextractables
April 2007 PQRI Training Course 13
Specifications and Acceptance Criteria for Specifications and Acceptance Criteria for ExtractablesExtractables
►► Routine Routine ExtractablesExtractables Testing should be performed Testing should be performed on OINDP critical components prior to drug on OINDP critical components prior to drug product manufacture. product manufacture.
►► Critical components should be released to drug Critical components should be released to drug product manufacture based on defined product manufacture based on defined specifications and acceptance criteria established specifications and acceptance criteria established through:through:§§ Understanding of critical component Understanding of critical component composition(scomposition(s), ),
ingredients, and compounding/fabrication processes.ingredients, and compounding/fabrication processes.§§ Comprehensive Controlled Extraction Studies.Comprehensive Controlled Extraction Studies.§§ A significant database of A significant database of extractablesextractables profiles obtained profiles obtained
with validated Routine with validated Routine ExtractablesExtractables Testing methods.Testing methods.§§ A complete A complete leachables/extractablesleachables/extractables correlation.correlation.
April 2007 PQRI Training Course 14
Specifications and Acceptance Criteria for Specifications and Acceptance Criteria for ExtractablesExtractables
►►Acceptance criteria for OINDP critical Acceptance criteria for OINDP critical component component extractablesextractables can include the can include the following:following:§§ Confirmation of Confirmation of extractablesextractables identified in identified in
Controlled Extraction Studies.Controlled Extraction Studies.§§ Quantitative limits for Quantitative limits for extractablesextractables identified in identified in
Controlled Extraction Studies.Controlled Extraction Studies.§§ A quantitative limit for ”new” or “ unspecified” A quantitative limit for ”new” or “ unspecified”
extractablesextractables not detected during Controlled not detected during Controlled Extraction Studies.Extraction Studies.
April 2007 PQRI Training Course 15
Establishing Specifications:Establishing Specifications:Widgets vs. PillsWidgets vs. Pills
►► Rest of World Rest of World (Planes, (Planes,
Trains & Automobiles…)Trains & Automobiles…)
§§ Known requirements Known requirements that must be met to that must be met to insure product insure product performanceperformance§§ Establish that process is Establish that process is
capable of meeting capable of meeting requirementsrequirements
►► PharmaceuticalsPharmaceuticals
§§ Vaguely known Vaguely known requirements (requirements (vsvsproduct performance)product performance)§§ Establish requirements Establish requirements
from vaguely known from vaguely known process capabilitiesprocess capabilities
April 2007 PQRI Training Course 16
Statistical Tools Related to Statistical Tools Related to SpecificationsSpecifications
►► Process Capability & Performance AnalysisProcess Capability & Performance Analysis§§ Statistical evaluation of process variability with respect Statistical evaluation of process variability with respect
to limitsto limits§§ Typically includes both process and measurement Typically includes both process and measurement
variabilityvariability
►► Operating Characteristic CurvesOperating Characteristic Curves§§ Statistical evaluation of decision making process related Statistical evaluation of decision making process related
to an individual testto an individual test§§ Considers influence of different test structures: numbers Considers influence of different test structures: numbers
of samples, average vs. individuals, tiered testing…of samples, average vs. individuals, tiered testing…
April 2007 PQRI Training Course 17
Types of Quality InspectionsTypes of Quality Inspections►►Inspection by AttributesInspection by Attributes§§ Defect testing (pass/fail by unit)Defect testing (pass/fail by unit)Visual inspection of containers for foreign material or defectsVisual inspection of containers for foreign material or defectsSpray test of Spray test of MDIsMDIs
►►Inspection by Variables Inspection by Variables §§ Estimation of Batch Parameters (central Estimation of Batch Parameters (central
tendency, variability)tendency, variability)HPLC Assay of tablets for active ingredientHPLC Assay of tablets for active ingredientDelivered Dose Uniformity of an MDIDelivered Dose Uniformity of an MDIContent Uniformity of a tabletContent Uniformity of a tabletLeachableLeachable/extractable testing/extractable testing
April 2007 PQRI Training Course 18
What we would like to have to What we would like to have to establish/verify acceptance criteria:establish/verify acceptance criteria:
April 2007 PQRI Training Course 19
What we typically have to What we typically have to establish acceptance criteria:establish acceptance criteria:
Impurity X
0.24
0.07
0.15
ND
April 2007 PQRI Training Course 20
Performance of Limits Established with Performance of Limits Established with Small DatasetsSmall Datasets
Fraction of Results Complying with Limits1.00.80.60.40.20.0
100
80
60
40
20
0
Variable
n=7n=8n=9n=10n=20
n=3n=4n=5n=6
Robustness of Establishing Acceptance Criteria with Small Datasets(limits established via +/- 3 standard deviations)
Ris
k of
obt
aini
ng li
mits
wor
se
than
the
asso
ciat
ed c
ompl
ianc
e ra
te
April 2007 PQRI Training Course 21
Comparison of Different Comparison of Different Approaches to Setting LimitsApproaches to Setting Limits
Fraction of Results Complying with Limits1.00.90.80.70.60.50.40.30.20.1
100
80
60
40
20
0
Variablep(accept) 3 stdp(accept) min/maxp(accept) 95/95p(accept) 99/99
Comparison of Different Approaches to Establishing Limits
Ris
k of
obt
aini
ng li
mits
wor
se
than
the
ass
ocia
ted
com
plia
nce
rate
n=7
April 2007 PQRI Training Course 22
Process Capability & PerformanceProcess Capability & Performance
►►Process CapabilityProcess Capability: Statistical comparison of : Statistical comparison of inherent process variability (inherent process variability (common causecommon causevariability only) to some limits. Generally, variability only) to some limits. Generally, represents the best possible performance. represents the best possible performance.
►►Process PerformanceProcess Performance: Statistical comparison : Statistical comparison of the total observed variability to some of the total observed variability to some limits. May include limits. May include special causespecial cause variability.variability.
April 2007 PQRI Training Course 23
Process CapabilityProcess Capability
►► Several different ‘Capability Indices’ existSeveral different ‘Capability Indices’ exist►► Designed to show whether Designed to show whether process+measurementprocess+measurement are are
capable of meeting limitscapable of meeting limitsCp=(USLCp=(USL--LSL)/6LSL)/6σσww
CpkCpk=Min{[(USL=Min{[(USL--Avg)/3σAvg)/3σww],[(Avg],[(Avg--LSL)/ 3σLSL)/ 3σww]]}}
►► Minimum Minimum CpkCpk of 1.33 expected for new processof 1.33 expected for new process►► Cp ~ Cp ~ CpkCpk when process is ‘centered’when process is ‘centered’►► Above is for twoAbove is for two--sided limit, for a onesided limit, for a one--sided limit Cp is sided limit Cp is
meaningless and meaningless and CpkCpk considers only the range to the considers only the range to the specified limitspecified limit
April 2007 PQRI Training Course 24
Process PerformanceProcess Performance
►► Several different ‘Performance Indices’ existSeveral different ‘Performance Indices’ exist►► Designed to show Designed to show process+measurementprocess+measurement performance performance
relative to limitsrelative to limitsPp=(USLPp=(USL--LSL)/6LSL)/6σσ
PpkPpk=Min{[(USL=Min{[(USL--Avg)/3σ],[(AvgAvg)/3σ],[(Avg--LSL)/ 3σ]LSL)/ 3σ]}}
►► Minimum Minimum PpkPpk of 1.33 expected for new processof 1.33 expected for new process►► PpkPpk ~ ~ CpkCpk when no ‘special cause’ source of errorwhen no ‘special cause’ source of error►► Above is for twoAbove is for two--sided limit, for a onesided limit, for a one--sided limit Pp is sided limit Pp is
meaningless and meaningless and PpkPpk considers only the range to the considers only the range to the specified limitspecified limit
April 2007 PQRI Training Course 25
Example Example Cp,Cpk,Pp,PpkCp,Cpk,Pp,Ppk
►► Consider a set of extractable data collected as follows:Consider a set of extractable data collected as follows:§§ 5 gaskets where sampled from each of 10 lots of gaskets5 gaskets where sampled from each of 10 lots of gaskets§§ The level of extractable “A” was determined for all 50 The level of extractable “A” was determined for all 50
samplessamples§§ Based on Based on toxtox information and a information and a leachablesleachables correlation an correlation an
upper limit of 120 upper limit of 120 ppmppm is being consideredis being considered§§ Normally an extractable limit is a oneNormally an extractable limit is a one--sided limit, but for sake sided limit, but for sake
of this example suppose we are also interested in a lower of this example suppose we are also interested in a lower limit i.e. 80 limit i.e. 80 ppmppm (this allows calculation of Cp & Pp which are (this allows calculation of Cp & Pp which are meaningless for a onemeaningless for a one--sided limit)sided limit)
►► Can the gasket manufacturing process support this limit?Can the gasket manufacturing process support this limit?►► What else can we conclude about the gasket process?What else can we conclude about the gasket process?
April 2007 PQRI Training Course 26
April 2007 PQRI Training Course 27
Calculating Cp, Calculating Cp, CpkCpk, Pp, , Pp, PpkPpk
Using actual numbers from previous figure:
Note: Std Dev (s) can be estimated from range (R) by s=R/d2d2 is a tabulated by n; for n=5, d2=2.33
April 2007 PQRI Training Course 28
Calculating Cp, Calculating Cp, CpkCpk, Pp, , Pp, PpkPpk
Cp=(120Cp=(120--80)/(6)(5.25)= 1.2780)/(6)(5.25)= 1.27
CpkCpk=min[(120=min[(120--99.1)/(3)(5.25)],99.1)/(3)(5.25)],[(99.1[(99.1--80)/(3)(5.25) ]= 1.2180)/(3)(5.25) ]= 1.21
Pp=(120Pp=(120--80)/(6)(8.72)= 0.7680)/(6)(8.72)= 0.76
PpkPpk=min[(120=min[(120--99.1)/(3)(8.72)], 99.1)/(3)(8.72)], [(99.1[(99.1--80)/(3)(8.72)]= 0.7380)/(3)(8.72)]= 0.73
April 2007 PQRI Training Course 29
Process Capability/Performance Process Capability/Performance using MINITABusing MINITAB®®
Sam
ple
Mea
n
10987654321
110
100
90
__X=99.09
UCL=106.18
LCL=92.01
Sam
ple
Ran
ge
10987654321
20
10
0
_R=12.28
UCL=25.97
LCL=0
Sample
Valu
es
108642
120
100
80
1201101009080
12010080
Within
O v erall
Specs
WithinStDev 5.28118C p 1.26C pk 1.2C Cpk 1.26
O v erallStDev 8.76404Pp 0.76Ppk 0.73C pm *
11
11
Xbar Chart
R Chart
Last 10 Subgroups
Capability Histogram
Normal Prob PlotA D: 0.413, P: 0.326
Capability Plot
April 2007 PQRI Training Course 30
Quality Decisions: Possible Outcomes and
Consequences
True Situation
Decision
Batch is of Acceptable Quality
Batch is not of Acceptable Quality
Accept Batch
Correct Decision
Type II error (β) (‘consumer’s risk’)
Reject Batch
Type I error (α) (‘producer’s risk’)
Correct Decision
Designing & Evaluating Test Structures: Designing & Evaluating Test Structures: Operating Characteristic CurvesOperating Characteristic Curves
April 2007 PQRI Training Course 31
Quality Standards vs. Quality Standards vs. Acceptance CriteriaAcceptance Criteria
Quality Standard:Quality Standard:►► All units must have an All units must have an
assay greater than 95%assay greater than 95%
Test Acceptance Criteria:Test Acceptance Criteria:►► Assay of 2 of 2 Samples Assay of 2 of 2 Samples
must be between 98must be between 98--102%102%
Quality standard should Quality standard should drive acceptance criteria drive acceptance criteria and test structureand test structure
OCCsOCCs used in this contextused in this context
April 2007 PQRI Training Course 32
Operating Characteristic CurvesOperating Characteristic Curves►► Used to characterize the statistical qualities of the Used to characterize the statistical qualities of the
decision making process associated with a decision making process associated with a particular test’s structure/formparticular test’s structure/form§§ Test structure/form includes: numbers of samples, Test structure/form includes: numbers of samples,
limits, tiers, decision process flow, quaniti(es) limits, tiers, decision process flow, quaniti(es) compared to limit compared to limit
►► Comment on the ability of the test structure to Comment on the ability of the test structure to discriminate between acceptable and discriminate between acceptable and unacceptable ‘batches’unacceptable ‘batches’
►► Allows estimation of type I & II error ratesAllows estimation of type I & II error rates§§ risk of failing an acceptable batch risk of failing an acceptable batch §§ risk of passing an unacceptable batchrisk of passing an unacceptable batch
April 2007 PQRI Training Course 33
Operating Characteristic CurvesOperating Characteristic Curves
►►Plot of the probability of acceptance (or Plot of the probability of acceptance (or rejection) vs. the quality variablerejection) vs. the quality variable§§ P(acceptP(accept) vs. true batch mean) vs. true batch mean§§ P(acceptP(accept) vs. true batch standard deviation) vs. true batch standard deviation§§ P(acceptP(accept) vs. true % defects) vs. true % defects
►►Constructed using the appropriate Constructed using the appropriate cumulative density probability distributioncumulative density probability distribution
April 2007 PQRI Training Course 34
Ideal OC CurvesIdeal OC Curves
True Mean (of Batch)
Pro
bab
ility
of
Acc
ept
anc
e
1501251007550
1.0
0.8
0.6
0.4
0.2
0.0
Ideal Operating Characteristic CurveTwo-Sided Limit (85-115)
0.0
0.2
0.4
0.6
0.8
1.0
Pro
bab
ility of Re
jection
True Mean (of Batch)
Pro
bab
ility
of
Acc
ept
anc
e
11010090807060
1.0
0.8
0.6
0.4
0.2
0.0
Ideal Operating Characteristic CurveOne-Sided Limit (>85)
Proba
bility o
f Reje
ction
0.0
0.2
0.4
0.6
0.8
1.0
True Std. Dev. (of Batch)
Pro
bab
ility
of
Acc
epta
nce
543210
1.0
0.8
0.6
0.4
0.2
0.0
Ideal Operating Characteristic Curve
Proba
bility of R
ejectio
n
Variability Limit (std. dev. <2.5)
0.0
0.2
0.4
0.6
0.8
1.0
April 2007 PQRI Training Course 35
Typical OC CurvesTypical OC Curves
True Mean (of Batch)
Prob
abili
ty o
f A
ccep
tanc
e
1.0
0.8
0.6
0.4
0.2
0.0
Real Operating Characteristic Curve
0.0
0.2
0.4
0.6
0.8
1.0
Probability of Rejection
True Std. Dev. (of Batch)
Prob
abili
ty o
f A
ccep
tanc
e
1.0
0.8
0.6
0.4
0.2
0.0
Real Operating Characteristic Curve
Probability of Rejection
Variability Limit
0.0
0.2
0.4
0.6
0.8
1.0
True Mean (of Batch)
Prob
abili
ty o
f A
ccep
tanc
e
1.0
0.8
0.6
0.4
0.2
0.0
Real Operating Characteristic Curve
Probability of Rejection
0.0
0.2
0.4
0.6
0.8
1.0
April 2007 PQRI Training Course 36
0
0.2
0.4
0.6
0.8
1 0
0.2
0.4
0.6
0.8
170 80 90 100 110 120 130
True Mean
p(ac
cept
) P(reject)
Limiting Quality
Risk of acceptingunacceptable lot
Risk of rejectingacceptable lot
Risks Associated with Testingin Relation to Operating Characteristic Curve
April 2007 PQRI Training Course 37
Process for Constructing OC Curves: Process for Constructing OC Curves: p(acceptp(accept) vs. Mean) vs. Mean
►► Need model probability distribution for individual Need model probability distribution for individual measurementsmeasurements
►► Need estimate of standard deviationNeed estimate of standard deviation§§ Curve is for an assumed standard deviation of the individual Curve is for an assumed standard deviation of the individual
measurementsmeasurements
►► Calculate probability to accept for a given value of the Calculate probability to accept for a given value of the mean from the appropriate cumulative density probability mean from the appropriate cumulative density probability distribution based on the test constructdistribution based on the test construct
►► Alternatively can estimate through numeric approachAlternatively can estimate through numeric approach►► Repeat over range of means of interestRepeat over range of means of interest
April 2007 PQRI Training Course 38
Example Calculations OC Curve:Example Calculations OC Curve:Sample mean compared to a twoSample mean compared to a two--sided limitsided limit
►► Normal Distribution: Normal Distribution:
►► Need areas under Need areas under distributiondistribution
►► No analytical solutionNo analytical solution►► Numeric approaches Numeric approaches
used leading to used leading to tabulations of tabulations of cdfcdf: : cumulative density cumulative density functionfunction
22 2/)(
21)( σµ
πσ−−= yeyf
April 2007 PQRI Training Course 39
Example Calculations OC Curve:Example Calculations OC Curve:Sample mean compared to a twoSample mean compared to a two--sided limitsided limit
►►For each point on the OC Curve, need to For each point on the OC Curve, need to calculate area under distribution (calculate area under distribution (μμ,,σσ) ) between limits, i.e. between limits, i.e. prob(acceptprob(accept) for a given ) for a given value of the mean (value of the mean (μμ))
►►Consider two sided limits of 95Consider two sided limits of 95--105 and 105 and σσ=2=2§§ σσ: standard deviation of sample means: standard deviation of sample means§§ This This σσ related to related to σσ of individual measurements of individual measurements
by a factor of 1/√nby a factor of 1/√n
April 2007 PQRI Training Course 40
Example Calculations OC Curve:Example Calculations OC Curve:Sample mean compared to a twoSample mean compared to a two--sided limitsided limit
►► For example, to calculate For example, to calculate p(acceptp(accept) at ) at μμ=99 first =99 first convert limits to standardized units (z)convert limits to standardized units (z)
►► (95(95--99)/2 = 99)/2 = --2; (1052; (105--99)/2 = 399)/2 = 3►► From tabulation of CDF or stat program:From tabulation of CDF or stat program:§§ Area below z=Area below z=--2 is 0.022752 is 0.02275§§ Area above z=3 is 0.00135Area above z=3 is 0.00135§§ Area between z=Area between z=--2 to z=3 is 2 to z=3 is
11--0.022750.02275--0.00135=0.975900.00135=0.97590
April 2007 PQRI Training Course 41
Example Calculations OC Curve:Example Calculations OC Curve:n of n compared to a twon of n compared to a two--sided limitsided limit
►►Follow above procedure to calculate the Follow above procedure to calculate the prob(acceptprob(accept) a single value.) a single value.
►►The probability to accept n of n values is:The probability to accept n of n values is:P(acceptP(accept11))
nn
►►If the previous example instead required 3 If the previous example instead required 3 of 3 results to be within 95of 3 results to be within 95--105, then105, then
P(acceptP(accept)=(0.97590))=(0.97590)33=0.92943=0.92943
April 2007 PQRI Training Course 42
Influence of Different Test Designs Influence of Different Test Designs on the OC Curveon the OC Curve
►►Tests Designed to Control MeanTests Designed to Control Mean§§ Vary n, set requirement on sample meanVary n, set requirement on sample mean§§ Vary n, set requirement on individual valuesVary n, set requirement on individual values§§ Influence of acceptance criteriaInfluence of acceptance criteria
April 2007 PQRI Training Course 43
Control on Batch MeanControl on Batch MeanImprovement in OC Curve as Sample Size IncreasesImprovement in OC Curve as Sample Size Increases
Acceptance Criteria: Sample Mean > 100Acceptance Criteria: Sample Mean > 100
80 90 100 110 120
0.0
0.5
1.0
true mean
p(ac
cept
) n=1
n=9
n=3
n=5n=7
1.0
0.0
p(reject)
April 2007 PQRI Training Course 44
Control on Batch MeanControl on Batch MeanEffect on OC Curve as Sample Size IncreasesEffect on OC Curve as Sample Size Increases
for n of n Requirementfor n of n RequirementAcceptance Criteria: n of n > 100Acceptance Criteria: n of n > 100
80 90 100 110 120
0.0
0.5
1.0
true mean
p(ac
cept
) p(reject)
n=1
n=3
n=5
n=7
n=9
0.0
1.0
April 2007 PQRI Training Course 45
Relationship of OC Curve to Relationship of OC Curve to Specification Limits (one sided)Specification Limits (one sided)
April 2007 PQRI Training Course 46
Relationship of OC Curve to Relationship of OC Curve to Specification Limits (two sided)Specification Limits (two sided)
0
0.2
0.4
0.6
0.8
1 0
0.2
0.4
0.6
0.8
160 70 80 90 100 110 120 130 140
true mean
P(ac
cept
)
80-12085-11590-110 95-105
P(reject)
April 2007 PQRI Training Course 47
Conclusions & Final ThoughtsConclusions & Final Thoughts
►►Appropriate L&E testing schemes should Appropriate L&E testing schemes should reflect:reflect:§§ InIn--depth understanding of component depth understanding of component
composition and the L&E characteristics of the composition and the L&E characteristics of the product/componentproduct/component§§ Thoughtful selection of critical testsThoughtful selection of critical tests§§ Robust validated methodsRobust validated methods§§ Statistical design and evaluation of tests and Statistical design and evaluation of tests and
acceptance criteriaacceptance criteria