sample size re-estimation: controlling the type-1 error regulatory-industry presentation - y...

Post on 08-Jul-2020

2 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

YannisJemiai,Ph.D.26September2017

ASABiopharmaceuticalSectionRegulatory-IndustryStatisticsWorkshop

SampleSizeRe-Estimation:ControllingtheType-1Error

2

unblinded samplesizere-estimationisanessentialdesigntool

YJemiai– 26Sep2017

Addressesuncertaintyintrialdesignassumptions

Oneofthemostpopularadaptations,especiallywhenusingaPromisingZoneapproach

21st CenturyCuresAct,PDUFAVI,encouragetheuseofadaptivedesigns

RegulatoryguidancedocumentsexistfromEMA(2007),FDACDER/CBER(2010),andCDRH(2016)

Increasinglymanyexamplesofregulatoryacceptance

Regulatory-IndustryStatistics

3

SowhataresomeoftheissuesconcerninguSSR designs?

YJemiai– 26Sep2017

Cantype-1errorbecontrolled?

Cansoundadaptivedecisionrulesbedeveloped?

Howdowegetapointestimateandconfidenceintervalsforthetreatmenteffect?

Howdoweavoidoperationalbiasduringtrialconduct?

Wefocushereontype-1errorcontrol

Regulatory-IndustryStatistics

4

Whydoestype-1errorgetinflated?

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

Consideratwo-stagedesignwithoutsamplesizeincrease

SupposenowthatweincreasethesamplesizeinstageIIfromn(2) ton*(2),butwedonotchangethecriticalvalue

Thiswillleadtotype-1errorinflation

5

Howcanwecontroltype-1errorthen?

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

6

Chronologyofdevelopment(partiallist)

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

7

1.Useaweightedstatisticwithpre-specifiedweights

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

8

Alsocalledthep-valuecombinationapproach

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

9

2.UsetheConventionalWaldStatistic

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

10

ExtendedCDLMethod

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

11

Whydoesitwork?

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

12

… andwhataretheconcerns?

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

13

3.PreserveConditionaltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

14

Preservingtheoveralltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

15

Conditionaltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

16

Conditionaltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

17

Conditionaltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

18

Conditionaltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

19

Conditionaltype-1errorrate

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

20

Pointstoconsider

YJemiai– 26Sep2017

Handlingsurvivalendpoints

Usableinformationatinterimanalysis

Non-inferiority&equivalencesettings

Independentincrements

Smallsamples

Regulatory-IndustryStatistics

21

Recap:challengesinunblinded SSRtrials

YJemiai– 26Sep2017

Type-1errorcontrolisnotanobstacle.Methodsexisttoensurestrongcontrol

Inferenceremainsachallenge,butmakingsomeprogress

Decision-makingalgorithmcanbeoptimizedusingsimulationsandlatestresearch

Operationalbiascanbeaddressed/minimizedbyusingiDMCs,puttinginplaceproperprocesses,andmakinguseoftechnology

Regulatory-IndustryStatistics

22

“Byfailingtoprepare,youarepreparingtofail.”

- BenjaminFranklin

22YJemiai– 26Sep2017 Regulatory-IndustryStatistics

23

Thankyou

YJemiai– 26Sep2017 Regulatory-IndustryStatistics

top related