Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming...
Transcript of Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming...
![Page 1: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/1.jpg)
Proof of Concept Concepts
Proof of Concept Concepts
Donald A. [email protected]
Donald A. [email protected]
BERRY
STATISTICAL INNOVATION
CONSULTANTS
![Page 2: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/2.jpg)
22
“Proof of concept”“Proof of concept”“Proof”: Impossible“Concept”: Not completely known“of”Best can do:
ProbabilitiesUtilities
“Proof”: Impossible“Concept”: Not completely known“of”Best can do:
ProbabilitiesUtilities
Modeling plays critical role
Modeling plays critical role
My perspective Bayesian
My perspective Bayesian
![Page 3: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/3.jpg)
33
Some IssuesSome Issues
Sequential learningPredictive distributionsBorrowing strength
Historical dataOther drugs in classAcross patient/disease groups
Sequential learningPredictive distributionsBorrowing strength
Historical dataOther drugs in classAcross patient/disease groups
DesignDesign
Design & Analysis
Design & Analysis
![Page 4: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/4.jpg)
44
Important PointsImportant PointsEndpoints*
EarlyLateModeling both
Historical informationBorrowing in classPreclinical ⇔ Clinical modelingSequential monitoring*Prespecify futility/promotion criteria*Utility, marketing, costs*Dynamic programming
Endpoints*EarlyLateModeling both
Historical informationBorrowing in classPreclinical ⇔ Clinical modelingSequential monitoring*Prespecify futility/promotion criteria*Utility, marketing, costs*Dynamic programming
![Page 5: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/5.jpg)
55
EXAMPLESEXAMPLESPredictive probabilitiesAdaptive randomization
Drug trial in AMLPhase I/II cancer trialDrug screening processDose findingPoC study in pain
Modeling early & late endpoints
Predictive probabilitiesAdaptive randomization
Drug trial in AMLPhase I/II cancer trialDrug screening processDose findingPoC study in pain
Modeling early & late endpoints
![Page 6: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/6.jpg)
66
PREDICTIVE PROBABILITYPREDICTIVE PROBABILITYGiven results so far in the trial, what is probability of
Statistical significance at trial’s end?Drug will make money?Futility?Etc?
Critical role inTrial designMonitoring trials
Given results so far in the trial, what is probability of
Statistical significance at trial’s end?Drug will make money?Futility?Etc?
Critical role inTrial designMonitoring trials
![Page 7: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/7.jpg)
77
Example* Using Pred ProbsExample* Using Pred ProbsDisease: HER2+ neoadj breast cancerEndpoint: tumor responseBalanced randomized: CT±HerceptinSample size planned: n = 164DMC; Interim results after n = 34:
Not Herc: 4/16 = 25% vs Herc: 12/18 = 67%Consistent with other studies
Predictive probability of stat sig: 95%Update: Adj trials w/12,000 patients!
Disease: HER2+ neoadj breast cancerEndpoint: tumor responseBalanced randomized: CT±HerceptinSample size planned: n = 164DMC; Interim results after n = 34:
Not Herc: 4/16 = 25% vs Herc: 12/18 = 67%Consistent with other studies
Predictive probability of stat sig: 95%Update: Adj trials w/12,000 patients!*Buzdar, et al. ASCO 2004, JCO 2005*Buzdar, et al. ASCO 2004, JCO 2005
![Page 8: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/8.jpg)
88
Troxacitabine (T) in acute myeloid leukemia (AML) combined withcytarabine (A) or idarubicin (I)Adaptive randomization to:
IA vs TA vs TIMax n = 75End point: Time to CR (< 50 days)
Troxacitabine (T) in acute myeloid leukemia (AML) combined withcytarabine (A) or idarubicin (I)Adaptive randomization to:
IA vs TA vs TIMax n = 75End point: Time to CR (< 50 days)
ADAPTIVE RANDOMIZATIONGiles, et al JCO (2003)
ADAPTIVE RANDOMIZATIONGiles, et al JCO (2003)
![Page 9: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/9.jpg)
99
Adaptive RandomizationAdaptive Randomization
Assign 1/3 to IA (standard) throughout (unless only 2 arms)Adaptive to TA and TI based on current results: P(pE > pIA|data)
Results →
Assign 1/3 to IA (standard) throughout (unless only 2 arms)Adaptive to TA and TI based on current results: P(pE > pIA|data)
Results →
![Page 10: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/10.jpg)
1010
Patient Prob IA Prob TA Prob TI Arm CR<501 0.33 0.33 0.33 TI not2 0.33 0.34 0.32 IA CR3 0.33 0.35 0.32 TI not4 0.33 0.37 0.30 IA not5 0.33 0.38 0.28 IA not6 0.33 0.39 0.28 IA CR7 0.33 0.39 0.27 IA not8 0.33 0.44 0.23 TI not9 0.33 0.47 0.20 TI not
10 0.33 0.43 0.24 TA CR11 0.33 0.50 0.17 TA not12 0.33 0.50 0.17 TA not13 0.33 0.47 0.20 TA not14 0.33 0.57 0.10 TI not15 0.33 0.57 0.10 TA CR16 0.33 0.56 0.11 IA not17 0.33 0.56 0.11 TA CR
![Page 11: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/11.jpg)
1111
Patient Prob IA Prob TA Prob TI Arm CR<5018 0.33 0.55 0.11 TA not19 0.33 0.54 0.13 TA not20 0.33 0.53 0.14 IA CR21 0.33 0.49 0.18 IA CR22 0.33 0.46 0.21 IA CR23 0.33 0.58 0.09 IA CR24 0.33 0.59 0.07 IA CR25 0.87 0.13 0 IA not26 0.87 0.13 0 TA not27 0.96 0.04 0 TA not28 0.96 0.04 0 IA CR29 0.96 0.04 0 IA not30 0.96 0.04 0 IA CR31 0.96 0.04 0 IA not32 0.96 0.04 0 TA not33 0.96 0.04 0 IA not34 0.96 0.04 0 IA CR
Compare n = 75
DropTI
![Page 12: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/12.jpg)
1212
Summary of resultsSummary of results
CR < 50 days:IA: 10/18 = 56%TA: 3/11 = 27%TI: 0/5 = 0%
Criticisms . . .
CR < 50 days:IA: 10/18 = 56%TA: 3/11 = 27%TI: 0/5 = 0%
Criticisms . . .
![Page 13: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/13.jpg)
1313
Example: Developing phase I/II cancer trialExample: Developing phase I/II cancer trialTwo drugs, A & D
DosingSequential or concurrent?
Phase I and MTD? PI’s plans …FactorialAdmissible dose combosAdaptive randomization
Two drugs, A & DDosingSequential or concurrent?
Phase I and MTD? PI’s plans …FactorialAdmissible dose combosAdaptive randomization
![Page 14: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/14.jpg)
1414
Dose/schedule possibilitiesDose/schedule possibilitiesD
ose
of A
Dose of DD
ose
of A
Dose of D
Concurrent A —> D
Admissible dosesAdmissible doses
![Page 15: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/15.jpg)
1515
At any given timeAt any given time
Expand admissible doses if toxicity allowsRandomize to admissible doses, adapting to tumor response(So might expand dose-range but still focus on lower doses)
Expand admissible doses if toxicity allowsRandomize to admissible doses, adapting to tumor response(So might expand dose-range but still focus on lower doses)
![Page 16: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/16.jpg)
1616
PoC trials for many drugs!PoC trials for many drugs!
Tumor responseGoals:
Treat effectivelyLearn quickly
Bury losersPromote potential winners
Tumor responseGoals:
Treat effectivelyLearn quickly
Bury losersPromote potential winners
![Page 17: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/17.jpg)
1717
Standard designsStandard designs
One drug (or dose) at a time; no drug/dose comparisonsTypical comparison by null hypothesis: RR = 20%Progress hopelessly slow!
One drug (or dose) at a time; no drug/dose comparisonsTypical comparison by null hypothesis: RR = 20%Progress hopelessly slow!
![Page 18: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/18.jpg)
1818
Standard 2-stage designStandard 2-stage design
First stage 20 patients: Stop if ≤ 4 or ≥ 9 responsesElse second set of 20
First stage 20 patients: Stop if ≤ 4 or ≥ 9 responsesElse second set of 20
![Page 19: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/19.jpg)
1919
An adaptive allocationAn adaptive allocation
When assigning next patient, find r = P(rate ≥ 20%|data) for each drugAssign drugs in proportion to rAdd drugs as become availableDrop drugs that have small rDrugs with large r → phase 3
When assigning next patient, find r = P(rate ≥ 20%|data) for each drugAssign drugs in proportion to rAdd drugs as become availableDrop drugs that have small rDrugs with large r → phase 3
![Page 20: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/20.jpg)
2020
Suppose 10 drugs, 200 patientsSuppose 10 drugs, 200 patients
9 drugs have mix of RRs 20% & 40%, 1 has 60%(“nugget”)
9 drugs have mix of RRs 20% & 40%, 1 has 60%(“nugget”)
<70%<70%
>99%>99%
Identify nugget …With probability: In average n:
Identify nugget …With probability: In average n:
110110
5050
Adaptive also better at finding “40%”, soonerAdaptive also better at finding “40%”, sooner
Stan
dard
Ada
ptiv
e
Ada
ptiv
e
Stan
dard
![Page 21: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/21.jpg)
2121
Suppose 100 drugs, 2000 patientsSuppose 100 drugs, 2000 patients
99 drugs have mix of RRs 20% & 40%, 1 has 60%(“nugget”)
99 drugs have mix of RRs 20% & 40%, 1 has 60%(“nugget”)
Adaptive also better at finding “40%”, soonerAdaptive also better at finding “40%”, sooner
<70%<70%
>99%>99%
Identify nugget …With probability: In average n:
Identify nugget …With probability: In average n:
11001100
500500
Stan
dard
Ada
ptiv
e
Ada
ptiv
e
Stan
dard
![Page 22: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/22.jpg)
2222
ConsequencesConsequences
Treat pts in trial effectivelyLearn quicklyAttractive to patients, in and out of the trialBetter drugs identified sooner; move through faster
Treat pts in trial effectivelyLearn quicklyAttractive to patients, in and out of the trialBetter drugs identified sooner; move through faster
![Page 23: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/23.jpg)
SEAMLESS PHASES II/III*SEAMLESS PHASES II/III*Early endpoint (tumor response, biomarker) may predict survival?May depend on treatmentShould model the possibilitiesPrimary endpoint: survivalBut observe relationships
Early endpoint (tumor response, biomarker) may predict survival?May depend on treatmentShould model the possibilitiesPrimary endpoint: survivalBut observe relationships
*Inoue, et al (2002 Biometrics)*Inoue, et al (2002 Biometrics)
![Page 24: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/24.jpg)
Goodresp
Goodresp
No respNo resp
SurvivaladvantageSurvival
advantageNo survivaladvantage
No survivaladvantage
Phase IIPhase II Phase IIIPhase III
Conventional drug developmentConventional drug development
6 mos6 mos 9-12 mos9-12 mos > 2 yrs> 2 yrs
StopStop
Seamless phase II/IIISeamless phase II/III< 2 yrs (usually)< 2 yrs (usually)
NotNot
MarketMarket
![Page 25: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/25.jpg)
Seamless phasesSeamless phasesPhase II: 1 or 2 centers; 10 pts/mo, randomize E vs C If pred probs “look good,” expand to Phase III: Many centers; 50 pts/mo (Initial centers continue accrual)Max n = 900
[Single trial: survival data combined in final analysis]
Phase II: 1 or 2 centers; 10 pts/mo, randomize E vs C If pred probs “look good,” expand to Phase III: Many centers; 50 pts/mo (Initial centers continue accrual)Max n = 900
[Single trial: survival data combined in final analysis]
![Page 26: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/26.jpg)
Early stoppingEarly stoppingUse pred probs of stat sigFrequent analyses (total of 18) using pred probs to:
Switch to Phase IIIStop accrual for
FutilityEfficacy
Submit NDA
Use pred probs of stat sigFrequent analyses (total of 18) using pred probs to:
Switch to Phase IIIStop accrual for
FutilityEfficacy
Submit NDA
![Page 27: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/27.jpg)
Conventional Phase III designs: Conv4 & Conv18, max N = 900(O’Brien-Fleming, same power as adaptive design)
Conventional Phase III designs: Conv4 & Conv18, max N = 900(O’Brien-Fleming, same power as adaptive design)
ComparisonsComparisons
![Page 28: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/28.jpg)
Expected N under H0Expected N under H0
0
200
400
600
800
1000
431
855 884
Bayes Conv4 Conv18
![Page 29: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/29.jpg)
Expected N under H0*Expected N under H0*
0
200
400
600
800
1000
424
854 884
Bayes Conv4 Conv18
![Page 30: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/30.jpg)
Expected N under H0**Expected N under H0**
0
200
400
600
800
1000
470
867 884
Bayes Conv4 Conv18
![Page 31: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/31.jpg)
Expected N under H1Expected N under H1
0
200
400
600
800
1000
649
887 888
Bayes Conv4 Conv18
![Page 32: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/32.jpg)
Expected N under H1*Expected N under H1*
0
200
400
600
800
1000
533
881833
Bayes Conv4 Conv18
![Page 33: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/33.jpg)
Expected N under H1**Expected N under H1**
0
200
400
600
800
1000
588
881 881
Bayes Conv4 Conv18
![Page 34: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/34.jpg)
BenefitsBenefitsDuration of drug development is greatly shortened under adaptive design:
Fewer patients in trialNo hiatus for setting up phase IIIAll patients used for
Phase III endpointRelation between response & survival
Duration of drug development is greatly shortened under adaptive design:
Fewer patients in trialNo hiatus for setting up phase IIIAll patients used for
Phase III endpointRelation between response & survival
![Page 35: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/35.jpg)
Possibility of large NPossibility of large NN seldom near 900When it is, it’s necessary!This possibility gives Bayesian design its edge[Other reason for edge is modeling response/survival]
N seldom near 900When it is, it’s necessary!This possibility gives Bayesian design its edge[Other reason for edge is modeling response/survival]
![Page 36: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/36.jpg)
3636
ConsequencesConsequences
Treat pts in trial effectivelyLearn quicklyAttractive to patients, in and out of the trialBetter drugs identified sooner; move through faster
Treat pts in trial effectivelyLearn quicklyAttractive to patients, in and out of the trialBetter drugs identified sooner; move through faster
![Page 37: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/37.jpg)
3737
Adaptive designs for proof of concept and dose-finding trialsAdaptive designs for proof of
concept and dose-finding trials
Mike K Smith1, Mark F. Morris1, Ieuan Jones1, Andy P. Grieve1,
Keith Tan2
1 Biostatistics and Reporting, PGRD Sandwich; 2 Clinical Sciences, PGRD Sandwich
Mike K Smith1, Mark F. Morris1, Ieuan Jones1, Andy P. Grieve1,
Keith Tan2
1 Biostatistics and Reporting, PGRD Sandwich; 2 Clinical Sciences, PGRD Sandwich
Mike SmithMike Smith
![Page 38: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/38.jpg)
3838
Mike K. Smith at U.S. JSMMike K. Smith at U.S. JSM
Proof of concept:If drug works well: want dose-response, efficacy & safetyIf doesn’t work: KILL IT!
Post-herpetic Neuralgia (PHN): Pain after shingles lasting >3 mos
Proof of concept:If drug works well: want dose-response, efficacy & safetyIf doesn’t work: KILL IT!
Post-herpetic Neuralgia (PHN): Pain after shingles lasting >3 mos
Mike SmithMike Smith
![Page 39: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/39.jpg)
3939
Pain endpointsPain endpoints
4-week to 12-week endpoints“Average pain score” over last 7 days’ pain scoresChange from baselinePerhaps not monotonic due to dropouts
4-week to 12-week endpoints“Average pain score” over last 7 days’ pain scoresChange from baselinePerhaps not monotonic due to dropouts
Mike SmithMike Smith
![Page 40: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/40.jpg)
4040
Case study: PD-217,014Case study: PD-217,014Experiment treatment for PHNPreclinical tests suggest 2 outcomes:
NO DIFFERENCE from conventional Rx based on gold standard preclinical testHUGE effect based on novel preclinical test
Normal Dynamic Linear Model (NDLM) as used in stroke dose-response study*
Experiment treatment for PHNPreclinical tests suggest 2 outcomes:
NO DIFFERENCE from conventional Rx based on gold standard preclinical testHUGE effect based on novel preclinical test
Normal Dynamic Linear Model (NDLM) as used in stroke dose-response study*
Mike SmithMike Smith
* ref: Berry DA, Müller P, Grieve AP, Smith MK, Parke T, Blazek R, Mitchard N and Krams M. Adaptive Bayesian designs for dose-ranging trials. In Case Studies in Bayesian Statistics V, (Ed: Carlin B, Carriquiry A, Gatsonis C,Gelman A, Kass RE, Verdinelli I, West M.) Springer Verlag, (2002) pp99-181.
![Page 41: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/41.jpg)
4141
DesignDesignObjective: dose that’s 1.5 pts better than placeboDoses, sample size:
Max N = 2807 active doses + placebo + active comparatorN=35/dose; oral dosage form; parallel groupEqual allocation to all treatmentsInterim analyses to drop ineffective doses or stop study
Continue to max N if any dose ≥ 1.5 pts better than placebo
Objective: dose that’s 1.5 pts better than placeboDoses, sample size:
Max N = 2807 active doses + placebo + active comparatorN=35/dose; oral dosage form; parallel groupEqual allocation to all treatmentsInterim analyses to drop ineffective doses or stop study
Continue to max N if any dose ≥ 1.5 pts better than placebo
Mike SmithMike Smith
![Page 42: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/42.jpg)
4242
Operating characteristics: power (average N)
Operating characteristics: power (average N)
0.83 (280)0.42 (280)0 (280)0 interims
0.83 (249)0.41 (209)0 (97)3 interims
0.81 (256)0.42 (225)0 (118)2 interims
0.84 (265)0.41 (247)0 (156)1 interim
Clinically important
improvement
Modest improvement
No dose response
Simulation Scenario
Mike SmithMike Smith
![Page 43: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/43.jpg)
4343
Case study: Pfizer A1461006Case study: Pfizer A1461006
Europe, Canada, AustraliaStarted Feb 2004First interim Aug 2004N=80 subjects evaluable, 110 randomised, 133 recruited
Europe, Canada, AustraliaStarted Feb 2004First interim Aug 2004N=80 subjects evaluable, 110 randomised, 133 recruited
Mike SmithMike Smith
![Page 44: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/44.jpg)
4444
Case study: results @ N=80Case study: results @ N=80
Number of evaluable subjects in analysis: N = 80
1.001.00-1.530.64-0.3600mg
1.000.93-1.400.59-0.3450mg
1.000.75-1.520.57-0.4300mg
1.000.92-1.330.57-0.2200mg
1.000.99-1.230.55-0.1150mg
1.001.16-0.990.530.0100mg
1.000.90-0.860.42-0.050mg
Posteriorprob offutilityUpperLowerSE
NDLMEstimated
Dose Effect(Active-Pbo)Dose
80% CredibleInterval
Mike SmithMike Smith
![Page 45: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/45.jpg)
4545
Case study: results @ N=80Case study: results @ N=80
Dose (mg)
Cha
nge
from
bas
elin
e
-10
-50
5
Placebo 50mg 100mg 150mg 200mg 300mg 450mg 600mg
NDLM estimate80% CI limits
NDLM dose response curve with observed data. Change from baseline in mean pain score
This is great; saved millions $
Mike SmithMike Smith
![Page 46: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/46.jpg)
4646
Case study: results @ N=80Case study: results @ N=80
Dose (mg)
Cha
nge
from
bas
elin
e
-10
-50
5
Placebo 50mg 100mg 150mg 200mg 300mg 450mg 600mg
NDLM estimate80% CI limits
NDLM dose response curve with observed data. Change from baseline in mean pain score
But suppose these observations were missing?
![Page 47: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/47.jpg)
4747
Case study: results @ N=80 40Case study: results @ N=80 40
Dose (mg)
Cha
nge
from
bas
elin
e
-10
-50
5
Placebo 50mg 100mg 150mg 200mg 300mg 450mg 600mg
NDLM estimate80% CI limits
NDLM dose response curve with observed data. Change from baseline in mean pain score
Little loss(could fill out if needed)
![Page 48: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/48.jpg)
4848Dose (mg)
Cha
nge
from
bas
elin
e
-10
-50
5
Placebo 50mg 100mg 150mg 200mg 300mg 450mg 600mg
NDLM estimate80% CI limits
NDLM dose response curve with observed data. Change from baseline in mean pain score
Or this? (Find predictive probability of eventually positive)
Case study: results @ N=80 40Case study: results @ N=80 40
![Page 49: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/49.jpg)
4949Dose (mg)
Cha
nge
from
bas
elin
e
-10
-50
5
Placebo 50mg 100mg 150mg 200mg 300mg 450mg 600mg
NDLM estimate80% CI limits
NDLM dose response curve with observed data. Change from baseline in mean pain score
Best yet! Continual monitoring; sequential dose assignment
Case study: results @ N=80Case study: results @ N=80 4040
![Page 50: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/50.jpg)
5050
Case study conclusionsCase study conclusions
133 subjects recruited (280 max)Little dose effect; little drug effectPositive control group gave expected effectPfizer stopped developmentCost savings > $2M
133 subjects recruited (280 max)Little dose effect; little drug effectPositive control group gave expected effectPfizer stopped developmentCost savings > $2M
Mike SmithMike Smith
![Page 51: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/51.jpg)
5151
CA125 data & predictive distributions of survival for two of many patients* ——>
CA125 data & predictive distributions of survival for two of many patients* ——>
*Modeling due to Scott Berry<[email protected]>*Modeling due to Scott Berry<[email protected]>
![Page 52: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/52.jpg)
Days
Patient #1
Treatment
![Page 53: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/53.jpg)
Patient #1
![Page 54: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/54.jpg)
Days
Patient #2
![Page 55: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/55.jpg)
Patient #2
![Page 56: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/56.jpg)
5656
MethodsMethods
Analytical
Multiple imputation
Analytical
Multiple imputation
![Page 57: Proof of Concept Concepts - Graz University of …zUtility, marketing, costs* zDynamic programming Endpoints* Early Late Modeling both Historical information Borrowing in class Preclinical](https://reader034.fdocuments.in/reader034/viewer/2022050115/5f4c616ee334e23252040682/html5/thumbnails/57.jpg)
5757
EXAMPLESEXAMPLESPredictive probabilitiesAdaptive randomization
Drug trial in AMLPhase I/II cancer trialDrug screening processDose findingPoC study in pain
Modeling early & late endpoints
Predictive probabilitiesAdaptive randomization
Drug trial in AMLPhase I/II cancer trialDrug screening processDose findingPoC study in pain
Modeling early & late endpoints