are different methods to reduce placebo response in cns ...€¦ · affect placebo response when...

The SAFER Interview• TheSAFERInterviewisanoperationalizedversionoftheSAFER(Specificity, Assessability, Face Validity, Ecological Validity, and Rule of the 3 Ps)principlesdevelopedbycliniciansatMassachusettsGeneralHospital(MGH)4.

• Inarecentanalysisoffivetrialsintreatmentresistantdepression(TRD)allsubjectshadpassedscreeningproceduresatthesiteandwereconsideredtobeeligibleforthetrial.

• Inadditiontothe45-minuteSAFERInterviewperformedremotelybyMGHclinicianswhocalledsubjects(N=1935)directly,astructuredseverityinterviewwasperformed;1562(81percent)ofthesubjectsweredeemedeligibleforcontinuedscreening.(Figure 5)

• Ofthe373(19percent)subjectsdeemedineligible:97(5percent)didnotmeetseveritycriteria;39(2percent)didnotmeetonlySAFERcriteria;151(8percent)didnotmeetATRQcriteriafortreatmentresistance;and,86(4percent)didnotmeetcriteriaonmorethanonecomponentoftheSAFERInterview.

• AcrossallofthesefiveTRDstudies,placeboresponserateswerewithinarangeof18to28percent,belowthe30to37percentaverageinstudiesoftreatmentsapprovedforTRD(olanzapine-fluoxetinecombination,quetiapineandaripiprazole).

Ziprasadone Monotherapy Study: Papakostas et al, J Clin Psychiatry. 2012 Dec; 73(12): 1541-7; TRD-1 and TRD-2: Papakostas et al, Am J Psychiatry. 2012 Dec 1; 169(12): 1267-74; ADAPT-A: Fava et al, Psychother Psychosom. 2012; 81(2): 87-97; Alkermes ALK 5461: Poster, ISCTM, 2013 Oct Note: ALK response rates relate to MADRS (secondary) outcomes.

19%

81%Zip Mono TRD-1 ADAPT-A TRD-2 ALK Average

12

68

54 52

81

53

100

80

60

40

20

0

% Reduction

Randomize

PlaceboActive

Treatment (Dose X)

Active Treatment

(Dose X)

Active Treatment

(Dose X)Placebo

ResponderResponderNon

ResponderNon

Responder

PlaceboActive

Treatment (Dose X)

Randomize

Randomize

TreatmentPhase 1

Results

TreatmentPhase 2

Rater Cohort Characteristics 1.00 80% 100

Experienced, calibrated 0.933 78% 108

0.80 71% 125

Inexperienced, calibrated 0.773 70% 130 0.60 59% 167


Reliability (ICC) Study Power1 Sample Size per Arm2

Q1 Q2 Q3 Q4

HAM-A (N=100) (Study 1)* 0.90 0.96 0.95 0.97

HAM-A (N=68) (Study 2)* 0.98 0.97 0.97 0.96

PANSS (N=131) (Study 3)* 0.90 0.90 0.96 0.88

PANSS (N=67) (Study 4)* 0.98 0.97 0.98 0.98

*MedAvante study data on file and made available to sponsor during study.


19%


12

68

54 52

81

53

100

80

60

40

20

0

% Reduction

Randomize

PlaceboActive

Treatment (Dose X)

Active Treatment

(Dose X)

Active Treatment

(Dose X)Placebo


ResponderNon

Responder

PlaceboActive

Treatment (Dose X)

Randomize

Randomize

TreatmentPhase 1

Results

TreatmentPhase 2



0.80 71% 125




Q1 Q2 Q3 Q4

HAM-A (N=100) (Study 1)* 0.90 0.96 0.95 0.97

HAM-A (N=68) (Study 2)* 0.98 0.97 0.97 0.96

PANSS (N=131) (Study 3)* 0.90 0.90 0.96 0.88

PANSS (N=67) (Study 4)* 0.98 0.97 0.98 0.98



19%


12

68

54 52

81

53

100

80

60

40

20

0

% Reduction

Randomize

PlaceboActive

Treatment (Dose X)

Active Treatment

(Dose X)

Active Treatment

(Dose X)Placebo


ResponderNon

Responder

PlaceboActive

Treatment (Dose X)

Randomize

Randomize

TreatmentPhase 1

Results

TreatmentPhase 2



0.80 71% 125




Q1 Q2 Q3 Q4

HAM-A (N=100) (Study 1)* 0.90 0.96 0.95 0.97

HAM-A (N=68) (Study 2)* 0.98 0.97 0.97 0.96

PANSS (N=131) (Study 3)* 0.90 0.90 0.96 0.88

PANSS (N=67) (Study 4)* 0.98 0.97 0.98 0.98


Number of Central Raters Estimated number of site raters

Num

ber o

f Rat

ers

180

160

140

120

100

80

60

40

20

0

STUDY A

STUDY B

STUDY C

STUDY D

STUDY E

STUDY F

STUDY G

STUDY I

STUDY J

STUDY K

STUDY L

STUDY M

STUDY H

STUDY N

STUDY O

STUDY P

STUDY Q

STUDY R

STUDY S

STUDY T

STUDY U

STUDY V

STUDY W

STUDY X

600

400

200

0

Freq

uenc

y

Freq

uenc

y

Freq

uenc

y

Freq

uenc

y

BPRS Total Score (n=382)PANSS Total Score (n=563)HAM-A Total Score (n=462)MADRS Total Score (n=4140)

80

60

40

20

0

100

80

60

40

20

0

50

40

30

20

10

00 4 8 12 16 20 24 28 32 36 40 44 48 6 10 14 18 22 26 30 34 38 42 50 60 70 80 90 100 110 120 130 140 24 28 32 36 40 44 48 52 56 60 64 68

Number of Central Raters Estimated number of site raters

Num

ber o

f Rat

ers

180

160

140

120

100

80

60

40

20

0

STUDY A

STUDY B

STUDY C

STUDY D

STUDY E

STUDY F

STUDY G

STUDY I

STUDY J

STUDY K

STUDY L

STUDY M

STUDY H

STUDY N

STUDY O

STUDY P

STUDY Q

STUDY R

STUDY S

STUDY T

STUDY U

STUDY V

STUDY W

STUDY X

600

400

200

0

Freq

uenc

y

Freq

uenc

y

Freq

uenc

y

Freq

uenc

y

BPRS Total Score (n=382)PANSS Total Score (n=563)HAM-A Total Score (n=462)MADRS Total Score (n=4140)

80

60

40

20

0

100

80

60

40

20

0

50

40

30

20

10

00 4 8 12 16 20 24 28 32 36 40 44 48 6 10 14 18 22 26 30 34 38 42 50 60 70 80 90 100 110 120 130 140 24 28 32 36 40 44 48 52 56 60 64 68


19%


12

68

54 52

81

53

100

80

60

40

20

0

% Reduction

Randomize

PlaceboActive

Treatment (Dose X)

Active Treatment

(Dose X)

Active Treatment

(Dose X)Placebo


ResponderNon

Responder

PlaceboActive

Treatment (Dose X)

Randomize

Randomize

TreatmentPhase 1

Results

TreatmentPhase 2



0.80 71% 125




Q1 Q2 Q3 Q4

HAM-A (N=100) (Study 1)* 0.90 0.96 0.95 0.97

HAM-A (N=68) (Study 2)* 0.98 0.97 0.97 0.96

PANSS (N=131) (Study 3)* 0.90 0.90 0.96 0.88

PANSS (N=67) (Study 4)* 0.98 0.97 0.98 0.98


Fava,M1,2, Williams JBW3,4, Freeman, M1,2 , Detke MJ3,5,6

1Massachusetts General Hospital Dept. of Psychiatry, 2Clinical Trials Network and Institute 3MedAvante, Inc., 4Columbia University Depts. of Psychiatry and Neurology, 5Detke Biopharma Consulting LLC, 6Indiana University School of Medicine

Are Different Methods to Reduce Placebo Response in CNS Trials Mutually Exclusive or Have Additive Effects?

©2014MedAvanteInc.

RESEARCH INSTITUTE

INTRODUCTION

Overthepasttwodecades,numerousmethodologieshavebeendevelopedwiththespecificgoalofreducingtheplaceboeffectinCNStrials.Thenumberoffailedtrialsinpsychiatryhasincreasedsubstantiallyovertime.Thismaybeduetobaselinescoreinflationandtheinclusionofinappropriatesubjects,variabilityinbaselineandpost-baselineclinicalratings,andexpectationorrelationshipbiasesassociatedwiththeperceptionsofsubjects,cliniciansorboththatcanaffectplaceboresponsewhenclinicalratersfunctioninmultiplerolesatstudysites.Thispresentationwilldiscusshowacombinationofthreestrategies,CentralRatings,useoftheSAFERInterview,anduseofSequentialParallelComparisonDesign(SPCD),mayoffera“triplesafety”nettoensuremaximumreductionoftheriskofafailedtrial.

THE THREE STRATEGIES

COMBINING THE THREE STRATEGIES

Get the free mobile app athttp:/ /gettag.mobi

Eachoftheseriskreductionstrategiescanimproveresults,butcombiningallthreemethodstoworkinsequenceinatrialcanincreasestudypowerandmosteffectivelyminimizethemethodologyfactorsthatcancauseatrialtofailtodetectasignal.Analyseshaveshownthatthecostofreducingtheriskoftrialfailureincreasestheexpectednetpresentvalueoftheoverallprograminvestment,manytimesover5.• Calibratedindependentcentralratersavoidthebiasesandreducetheratingvariabilitythatadverselyaffectssubjectselection,accuratebaselineandpost-baselineratings.

• TheSAFERInterview,administeredbyindependentcentralraters,canhelpensureselectionofappropriatesubjects.

• SPCDmethodology,inconjunctionwiththeaccuracyandprecisionachievablewithremotecentralratings,candecreaseplaceboresponseandallowforasmallersamplesize.

Incombination,thesethreeuniquebutcomplementarystrategiesoffera“triplesafetynet”thatmaydramaticallyreducetheriskofafailedtrialandofferthebestchanceforatrialtosucceed.

References1Mueller&Szegedi.(2002).EffectsofInterraterReliabilityofPsychopathologicAssessmentonPowerandSampleSizeCalculationsinClinicalTrials.JournalofClinicalPsychopharmacology.22(3),318-325.Calculatedwheren=100subjectsperarmandeffectsize=.402PerkinsDOetal.(2000).Penny-wiseandpoundfoolish:Theimpactofmeasurementerroronsamplesizerequirementsinclinicaltrials.BiologicalPsychiatry.47,762-766.Calculatedwheren=100subjectsperarm3KobakKAetal.(2009).SourcesofUnreliabilityinDepressionRatings..JournalofClinicalPsychopharmacology.29(1),82-85.4TargumSD,PollackMH,FavaM.(2008).Redefiningaffectivedisorders:Relevancefordrugdevelopment.CNSNeuroscience&Therapeutics.14,2-9;doi:10.1111/j.1527-3458.2008.00038.x.5GreenblattWetal.(2012).TheValueofRiskReductioninCNSDrugDevelopment:UseofExpectedNetPresentValue(eNPV)asaModel.PresentedatAmericanCollegeofNeuropsychopharmacology51stAnnualMeeting,HollywoodFL.

Remote Independent Central Ratings• Theuseofwell-trainedandblinded,calibratedexpertcentralratersminimizesperceptionbiasesandreducesvariability,whichenhancesseparationbetweenplaceboandactivetreatment.

•RemoteindependentcentralratersscreeningsubjectsviaevaluationofsymptomseverityandtheSAFERInterviewhelpensuretheunbiasedselectionofappropriatesubjectswhileavoidingbaselinescoreinflation.(Figure 1)

Figure 1.Sample of Baseline Score Distributions in Completed Studies MDD GAD SCHIZOPHRENIA PSYCHOSIS

•Centralratingsensureaccuratepost-baselineratingsandincreasestudypowerviahighICCSfurtherenhancingtheimpactoftheSPCDidentificationofplacebonon-respondersforrandomizationinStage2.

•Centralratersreducevariabilitybylimitingthenumberofratersneededforanyonetrial.(Figure 2)

•Higherinterraterreliabilitymeanshigherstudy

power.Experiencedandcalibratedratersensurethehighestlevelofreliability.(Figure 3)

•Ongoingcalibrationofcentralraterspreventsraterdriftthroughoutthecourseofthetrial.(Figure 4)

Figure 3.Impact of Experience vs. Calibration on Interrater Reliability

Figure 4.Impact of Continuous Calibration on Interrater Reliability

Figure 2.Centralized vs. Decentralized Study: Number of Raters

Figure 5.SAFER Interview Pass/Fail Rate

Sequential Parallel Comparison Design (SPCD)• GoalsoftheSequentialParallelComparisonDesign(SPCD)aretoreducetheimpactofhighplaceboresponseanddecreasestudysamplesize.

• TheSPCDmodel(Figure6)incorporatestworelativelyshortphasesoftreatmentofequalduration:

Phase 1:allsubjectsrandomizedtoplacebooractivetreatment;moresubjectsthanusualassignedtoplaceboarmtoenrichphase2;

Phase 2:subjectsonactivetreatmentandplacebonon-respondersinPhase1arerandomizedtoplacebooractivetreatment.

Figure 6.Sequential Parallel Comparison Design

• InPhase2theblindismaintainedforallsubjectsasthoseonactivedruginPhase1remainonactivetreatmentandplacebonon-respondersarerandomizedtoplacebooractivetreatment.

• StandardanalysesofSPCDstudiescombinedatainPhase1fromallsubjectsandinPhase2onlyfromplacebonon-respondersduringPhase1.

• Poolingthedatawithapre-specifiedweightedaveragesignificantlyreducestheoverallplaceboresponse,asitistheaverageoftheplaceboresponseinPhase1(standardresponse)andinPhase2(lowerresponseamongsubjectswhofailedtorespondtoplaceboprospectively).

• InfivecompletedSPCDtrials,thismethodhasreducedtheoverallplaceboresponserateonaverageby53percentcomparedtothePhase1placeboresponse.(Figure 7)

Figure 7.Percent Reduction in Placebo Response Rates in Stage 2 SPCD Trials (Chronological Order)

Passes

Fails

ZiprasadoneMonotherapyStudy:Papakostasetal,JClinPsychiatry.2012Dec;73(12):1541-7;TRD-1andTRD-2:Papakostasetal,AmJPsychiatry.2012Dec1;169(12):1267-74;ADAPT-A:Favaetal,PsychotherPsychosom.2012;81(2):87-97;AlkermesALK5461:Poster,ISCTM,2013OctNote:ALKresponseratesrelatetoMADRS(secondary)outcomes.


19%


12

68

54 52

81

53

100

80

60

40

20

0

% Reduction

Randomize

PlaceboActive

Treatment (Dose X)

Active Treatment

(Dose X)

Active Treatment

(Dose X)Placebo


ResponderNon

Responder

PlaceboActive

Treatment (Dose X)

Randomize

Randomize

TreatmentPhase 1

Results

TreatmentPhase 2



0.80 71% 125




Q1 Q2 Q3 Q4

HAM-A (N=100) (Study 1)* 0.90 0.96 0.95 0.97

HAM-A (N=68) (Study 2)* 0.98 0.97 0.97 0.96

PANSS (N=131) (Study 3)* 0.90 0.90 0.96 0.88

PANSS (N=67) (Study 4)* 0.98 0.97 0.98 0.98


are different methods to reduce placebo response in cns ...€¦ · affect placebo response when...

Documents