Implementation of MAMS 1 Mahesh Parmar, Feb-2010 Practical Implementation of Multi-Arm Multi-Stage Trials Mahesh Parmar Director MRC Clinical Trials Unit

Implementation of MAMS 2 Mahesh Parmar, Feb-2010 Structure of presentation 1.MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.Conclusions

Implementation of MAMS 3 Mahesh Parmar, Feb MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.Conclusions

Implementation of MAMS 4 Mahesh Parmar, Feb-2010 Why Multi-Arm, Multi-Stage trials? More often than not ‘new’ is not better Academia  624 NCI sponsored phase III trials (Arch. Int Med, 2008)  ~30% of trials ‘statistically significant’  ~40% of trials ‘new’ therapy preferred Industry  Agents successful at phase I: only 10-20% receive a marketing authorisation  Success rate of phase III trials ~30-40%

Implementation of MAMS 5 Mahesh Parmar, Feb-2010 Why Multi-Arm, Multi-Stage trials? Typical (academic) Phase III trial  Hundreds or thousands of patients  5 to 10 years from idea to result  Hundreds of research staff  Cost millions in development  Years of investment from the key players High chance of finding new treatment is not superior

Implementation of MAMS 6 Mahesh Parmar, Feb-2010 Principles of a New Strategy Need better mechanism  than single arm phase II trial to decide Test as many new promising treatments as possible  in the same timescale  maximise potential for a “positive trial” Potential to discontinue unpromising arms  quickly and reliably Start to randomise as quickly as possible Multi-Arm, Multi-Stage (MAMS) trials

Implementation of MAMS 7 Mahesh Parmar, Feb-2010 Activity (phase II stages) Asks the question  Are there reasons why we should not continue testing this treatment? Testing for a lack-of-activity  Emphasis not testing for activity but for lack-of- activity or lack-of-sufficient-benefit

Implementation of MAMS 8 Mahesh Parmar, Feb-2010 Activity phase II stages Assume definitive outcome measure is most clinically relevant  eg overall survival in cancer trials In Activity Stages focus on an earlier outcome measure  eg “failure-free survival” (FFS) in cancer trials

Implementation of MAMS 9 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient

Implementation of MAMS 10 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Stage 1 accrual Final analyses Stage 3 accrual Stage 2 accrual Intermediate analyses 2 Intermediate analyses 1

Implementation of MAMS 11 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Final analyses Stage 3 accrual Stage 2 accrual Intermediate analyses 2 Intermediate analyses 1

Implementation of MAMS 12 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Intermediate analyses 1 Final analyses Stage 3 accrual Stage 2 accrual Intermediate analyses 2

Implementation of MAMS 13 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Intermediate analyses 1 Yes Stage 2 accrual Final analyses Stage 3 accrual Intermediate analyses 2

Implementation of MAMS 14 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Intermediate analyses 1 Yes Stage 2 accrual Final analyses Stage 3 accrual Intermediate analyses 2

Implementation of MAMS 15 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Intermediate analyses 1 Yes No Stage 2 accrual Final analyses Stage 3 accrual Intermediate analyses 2

Implementation of MAMS 16 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Yes No Stage 2 accrual Final analyses Stage 3 accrual Intermediate analyses 2

Implementation of MAMS 17 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Yes No Yes Stage 2 accrual Final analyses Stage 3 accrual Intermediate analyses 2

Implementation of MAMS 18 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Yes No Yes Stage 2 accrual Control Sufficient activity Intermediate analyses 2 Final analyses Stage 3 accrual

Implementation of MAMS 19 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Yes No Yes Stage 2 accrual Control Sufficient activity Intermediate analyses 2 Yes Stage 3 accrual Final analyses

Implementation of MAMS 20 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Final analyses Yes No Yes Stage 2 accrual Control Sufficient activity (N/A) Intermediate analyses 2 Yes No Stage 3 accrual

Implementation of MAMS 21 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Final analyses Yes No Yes Stage 2 accrual Control Sufficient activity (N/A) Insufficient activity Intermediate analyses 2 Yes No Stage 3 accrual

Implementation of MAMS 22 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Intermediate analyses 1 Final analyses Yes No Yes Stage 2 accrual Control Sufficient activity (N/A) Insufficient activity Sufficient activity Intermediate analyses 2 Yes No Yes Stage 3 accrual

Implementation of MAMS 23 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Control Primary analysis Intermediate analyses 1 Final analyses Yes No Yes Stage 2 accrual Control Sufficient activity (N/A) Insufficient activity Sufficient activity Intermediate analyses 2 Yes No Yes Stage 3 accrual Primary analysis

Implementation of MAMS 24 Mahesh Parmar, Feb-2010 Multi-Arm, Multi-Stage Trials RANDOMISERANDOMISE Test 1 Test 2 Test 3 Test 4 Con- trol Eligible patient Yes Stage 1 accrual Control Sufficient activity Insufficient activity Sufficient activity Control Primary analysis Secondary analysis Primary analysis Intermediate analyses 1 Final analyses Yes No Yes Stage 2 accrual Control Sufficient activity (N/A) Insufficient activity Sufficient activity Intermediate analyses 2 Yes No Yes Stage 3 accrual

Implementation of MAMS 25 Mahesh Parmar, Feb-2010 Advantages of MAMS Fewer patients Less overall time  Randomised from the start  Concurrent (not sequentially)  No delay between phase II and phase III  Fewer applications for finance and approvals one grant application one protocol one CTA submission (per country) one ethics submission (per country) one R&D approval (per site)  Saves many years!

Implementation of MAMS 26 Mahesh Parmar, Feb-2010 Advantages of MAMS Fewer patients Less overall time Increased flexibility Reduced costs

Implementation of MAMS 27 Mahesh Parmar, Feb MAMS application in STAMPEDE 1.MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.Conclusions

Implementation of MAMS 28 Mahesh Parmar, Feb-2010 Most common male cancer  UK diagnosis: 34,000 in 2005  UK deaths: 10,000 in 2006  Global deaths: 250,000 in 2000 Rising rates of diagnosis  Aging population, awareness, PSA screening Modest treatment effect = big impact Impact

Implementation of MAMS 29 Mahesh Parmar, Feb-2010 Urgent need to improve outcomes for men starting hormone therapy with prostate cancer  Median survival ~4 years  Median failure-free survival ~2 years No new therapies improving survival for this group of men for many years Needs in prostate cancer

Implementation of MAMS 30 Mahesh Parmar, Feb-2010 Design rationale Many interesting agents  Different classes and modes of action No obvious reason to choose one  Many used in later stages of disease  Don’t want to choose arbitrarily Quicker and efficient to use MAMS design

Implementation of MAMS 31 Mahesh Parmar, Feb-2010 Trial Design

Implementation of MAMS 32 Mahesh Parmar, Feb-2010 Trial Design Stages StageOutcome Measures PrimarySecondary PilotSafetyFeasibility Activity I-IIIFailure-free survivalOverall survival (phase II)(PSA-driven)Toxicity Skeletal-related events Efficacy IVOverall survivalFailure-free survival (phase III)Toxicity Skeletal-related events Quality of life

Implementation of MAMS 33 Mahesh Parmar, Feb-2010 Design Assumptions: for all stages Pairwise comparison of each research arm against control Hazard ratios for design 10% improvement in FFS: 50 to 60% at 2 yr Overall survival Failure-free survival Null (H 0 )1.00 Alternative (H A )0.75

Implementation of MAMS 34 Mahesh Parmar, Feb-2010 Significance level and power Stage Primary Outcome Significance Level Power IFFS0.5095% IIFFS0.2595% IIIFFS0.1095% IVOS % Overall %

STAMPEDE Joint TSC/IDMC Meeting35 Matt Sydes, 20-July-2009 Cutpoints in STAMPEDE 95% power

Implementation of MAMS 36 Mahesh Parmar, Feb Issues in implementation 1.MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.Conclusions

Implementation of MAMS 37 Mahesh Parmar, Feb-2010 Groups to convince – 1 Medical community  Mostly seen by urologists  Trial treatments need to be given by oncologists Would it appear complex in clinics?  Or in MDT meetings? Previous multi-arm trials  Excellent recruitment to:  FOCUS – colorectal cancer – 5 arms  ICON5 – ovarian cancer – 5 arms

Implementation of MAMS 38 Mahesh Parmar, Feb-2010 Groups to convince – 2 Men with prostate cancer  Involved patient groups  Two patients on Trial Management Group  Very positive opinions  Patient involvement has been excellent for trial Two-part PIS  General information (few pages) – prior to randomisation  Arm-specific information All before randomisation or Only relevant one after randomisation (few pages) Patient choice – as informed as patient would like to be

Implementation of MAMS 39 Mahesh Parmar, Feb-2010 Groups to convince – 3 Funding bodies Regulatory and ethical committees Hospital governance Potential for conservative reviewers  No precedent for such approaches Approved – after a bit of ‘discussion’

Implementation of MAMS 40 Mahesh Parmar, Feb-2010 Groups to convince – 4 Industry partners  3 industry partners in STAMPEDE  All keen on design because…  Not primarily comparing their drugs head to head  Efficient design  Early “get-out” if agent not so beneficial More companies = more negotiations  More contracts  More time  …

Implementation of MAMS 41 Mahesh Parmar, Feb-2010 Recruitment How many patients are required? Total N dependent on:  Observed accrual rates  Observed event rates Do we have the predicted mix of patients?  # arms recruiting in each Activity & Efficacy Stage Likely 2300 to 3600 patients  Over 5.5 to 7.5 years Faster recruitment:  Requires more patients  Takes less time

Implementation of MAMS 42 Mahesh Parmar, Feb-2010 STAMPEDE Accrual

Implementation of MAMS 43 Mahesh Parmar, Feb-2010 Pilot Phase Assessing safety & feasibility  Particularly for the combination arms Target: 210 patients in 18 months  Limited centres Completed: Oct-2005 to Spring-2007  On schedule IDMC recommended continuing all arms  None stopped because of clear safety signals

Implementation of MAMS 44 Mahesh Parmar, Feb-2010 Stage 1 Completed ahead of schedule IDMC recommended continuation of all arms

Implementation of MAMS 45 Mahesh Parmar, Feb-2010 Hurdles 1.Funding 2.Finalising choice of drug 3.Discussions with pharma 4.Funding (revisited) 5.EU Clinical Trials Directive 6.Vioxx 7.Increased number of stages 8.Continuity of staff 9.Pilot phase

Implementation of MAMS 46 Mahesh Parmar, Feb-2010 Funding – main grant Outline Application to MRC – May 2001  Accept for Full Application  Novel design & novel drugs vs conservatism  Advised to submit to CTAAC instead  New at time – for cancer trials, incl. MRC funds Outline Application to CTAAC – November 2001  Accept for Full Application Full Application to CTAAC – November 2002  Accepted for partial funding – Feb 2003  Agreements/funding needed from industry partners

Implementation of MAMS 47 Mahesh Parmar, Feb-2010 Funding – industry partners Novartis – zoledronic acid  Research grant  Free drug  Drug distributed to sites Sanofi-Aventis – docetaxel  Research grant  Discounted drug (buy 1, get 2 free)  Sites to buy drug & claim back Pfizer – celecoxib  Free drug  Research grant, including distribution costs

Implementation of MAMS 48 Mahesh Parmar, Feb-2010 Negotiations Discussions required in many trials More difficult if many companies?  Different pace  Different contracts  Different comments on protocol Flagged plans for future discussions  International expansion planned  All agreements made with UK affiliates

Implementation of MAMS 49 Mahesh Parmar, Feb-2010 Changes in CTU Personnel Continuity 3 Project Leads 5 Trial Managers 7 Data Managers 3 Patients/Consumer representatives 3 Statisticians (including the original 2!)

Implementation of MAMS 50 Mahesh Parmar, Feb-2010 Future hurdles 1.Recruitment rates 2.Contracts revisited 3.Moving through MAMS stages 4.Dropping arms 5.Completing recruitment 6.Adding arms

Implementation of MAMS 51 Mahesh Parmar, Feb Issues in intermediate analyses 1.MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.Conclusions

Implementation of MAMS 52 Mahesh Parmar, Feb-2010 Moving through stages IDMC review interim data  Safety and activity data  Recommendations to TSC and TMG Education & training  For all committees  Trust in relationships  Hypothetical examples

Implementation of MAMS 53 Mahesh Parmar, Feb-2010 Moving through stages Face-to-face meeting between TSC and IDMC Discussion of different scenarios Agree a lengthy communication plan – including timelines  Between IDMC and TSC  To Centres  To Participants  To Regulators  Changes in randomisation programme etc…

Implementation of MAMS 54 Mahesh Parmar, Feb-2010 Dropping arms or agents If combination arm stopped for lack of sufficient effect  Should “single” agent arm stop too? If single agent arm stopped for lack of sufficient effect  Should combination arm stop too? Training and discussion

Implementation of MAMS 55 Mahesh Parmar, Feb-2010 Adding new arms Could an extra arm be added?  Promising agents could be added later!  “Rolling trial” or “roundabout design”?  Use same approach as for other arms, but delayed Implementing in STAMPEDE  Considering abiraterone  Application to CTAAC (CRUK) approved  Company considering proposal Design should be appealing to funders and industry  Pre-existing network of recruiting sites  Saves ~3 years

Implementation of MAMS 56 Mahesh Parmar, Feb Conclusions 1.MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.Conclusions

Implementation of MAMS 57 Mahesh Parmar, Feb-2010 Conclusions Multi-arm, multi-stage trials are  Being done in a number of other cancer types  Feasible  Efficient  Supported by patients, clinicians, funders, companies STAMPEDE  Is recruiting well  Is running well

MAMS Trials for GSK58 Mahesh Parmar, Nov References – MAMS trials Royston P, Parmar MKB, Qian W Novel Designs for Multi-Arm Clinical Trials with Survival Outcomes, with an Application in Ovarian Cancer. Statistics in Medicine 2003;22: 2239 – Barthel FM-S, Royston P, Parmar MKB A menu-driven facility for sample size calculation in multi- arm, multi-stage randomised controlled trials with a survival-time outcome. The Stata Journal 2007 (submitted) Parmar MKB Speeding up the Evaluation of New Agents in Cancer. J.Natl.Cancer Inst. 100 (17): , 2008.

MAMS Trials for GSK59 Mahesh Parmar, Nov References - STAMPEDE Sydes MR, MKB Parmar, ND James et al Issues in applying multi-arm multi-stage (MAMS) methodology to a clinical trial in prostate cancer: the MRC STAMPEDE trial. Trials 10 (39), James ND, Sydes MR, Clarke NW et al STAMPEDE: Systemic Therapy for Advancing or Metastatic Prostate Cancer -- A Multi-Arm Multi-Stage Randomised Controlled Trial. Clinical Oncology 20 (8): , James ND, Sydes MR, Clarke NW et al Systemic therapy for advancing or metastatic prostate cancer (STAMPEDE): a multi-arm, multistage randomized controlled trial. BJU.Int 103 (4): , 2009.

MAMS Trials for GSK60 Mahesh Parmar, Nov Software and wokshop Free software available  Design MAMS trials  Available from MRC CTU  Implemented in Stata CTU Workshop in February 2011  Design and analysis of MAMs trials  Advanced Issues in Design and Analysis of Trials with Time to Event Outcomes

MAMS Trials for GSK61 Mahesh Parmar, Nov. 2010

Implementation of MAMS 62 Mahesh Parmar, Feb-2010 Other MAMS trials ICON 6  3-arm, 4-stage trial in 2nd line ovarian cancer using progression-free survival as the intermediate outcome measure AML 16  5-arm, 2-stage trial in acute myeloid leukaemia

Implementation of MAMS 63 Mahesh Parmar, Feb-2010 Conclusions MAMS trials are  Feasible  Efficient  Supported by patients, clinicians, companies

Implementation of MAMS 64 Mahesh Parmar, Feb-2010 Contacts Matthew Sydes E T +44 (0)

Implementation of MAMS 65 Mahesh Parmar, Feb-2010 Extra slides

Implementation of MAMS 66 Mahesh Parmar, Feb-2010 Trial Initiation 6 trial arms Pilot Phase Recruit 210 patients Recruitment continues to control arm + other research arms in next stage Due to safety issues or a lack of feasibility, accrual may stop to one or more research arms IDMC review KeyIDMC = Independent Data Monitoring Committee FFS = Failure Free survival (Total ~210 pts) Trial Stage - Pilot

Implementation of MAMS 67 Mahesh Parmar, Feb-2010 Trial Stage - Efficacy Stage I Efficacy Stage I Recruit until 113 control arm FFS events (Total ~1200 pts) IDMC review Accrual may be stopped in any research arm for safety or lack of efficacy HR=1.00 KeyIDMC = Independent Data Monitoring Committee FFS = Failure Free survival Recruitment continues to control arm + other research arms in next stage

Implementation of MAMS 68 Mahesh Parmar, Feb-2010 Trial Stage - Efficacy Stage II Efficacy Stage II Recruit until 216 control arm FFS events (Total ~1800 pts) IDMC review KeyIDMC = Independent Data Monitoring Committee FFS = Failure Free survival Accrual may be stopped in any research arm for safety or lack of efficacy HR=0.92 Recruitment continues to control arm + other research arms in next stage

Implementation of MAMS 69 Mahesh Parmar, Feb-2010 Trial Stage - Efficacy Stage III KeyIDMC = Independent Data Monitoring Committee FFS = Failure Free survival Efficacy Stage III Recruit until 334 control arm FFS events (Total ~2300 pts) IDMC review Accrual may be stopped in any research arm for safety or lack of efficacyHR=0.89 Recruitment continues to control arm + other research arms in next stage

Implementation of MAMS 70 Mahesh Parmar, Feb-2010 Trial Stage – Efficacy Stage IV Efficacy Stage IV Recruit until 405 control arm deaths (Total ~3200 pts) Main Analyses (1) Overall survival in arms recruiting in Efficacy Stage IV (2) Secondary outcome measures in arms recruiting in Efficacy Stage IV (3) All outcome measures in all 6 arms involved in trial (regardless of when recruitment stopped)

Implementation of MAMS 71 Mahesh Parmar, Feb-2010 Statistical Issues Pair-wise comparison of each research vs control Randomisation ratio (A:B:C:D:E:F) = 2:1:1:1:1:1  Two control arm patients for every research arm patients  Efficient for power Randomisation centrally  Computer based algorithm  Minimisation with an additional random element  7 stratification factors for balancing groups

MAMS Trials for GSK72 Matt Sydes, Oct-2009 Phase II studies of single agent Phase II studies of combinations Phase III trials of combinations R1 C R2 C R3 C A – Traditional approach B – MAMS design Phase II studies of single agent R1 R2 R3 CR1R2R3 MAMS: Phase II/III trial of combinations Time Notes C = control arm; R1 = research arm R1; R2 = Research arm R2; R3 = research arm R3 Assumes that single agent studies would be carried out before combination studies Assumes that phase II studies require smaller numbers of patients and so smaller number of centres. Therefore, phase II studies of different agents may be carried out concurrently Assumes that phase III trials require larger numbers of patients and a network of centres that can only run one trial at a time: therefore, phase III trials of different agents must be carried out sequentially MAMS design rolls phase II assessment of combinations into the same trial as the phase III assessment of effectiveness

Implementation of MAMS 73 Mahesh Parmar, Feb-2010 Impact of accrual rates Accrual/ Year Total ptsDuration (yrs) Difference in pts Difference in length months months Assuming: median FFS=24m, median OS=48m, 6 arms recruiting at each stage

Implementation of MAMS 74 Mahesh Parmar, Feb-2010 Assumptions in STAMPEDE Broad eligibility spectrum  Patients starting hormone therapy Newly metastatic New & locally advanced Previously treated & relapsing  Assume more patients are M0 than M1  Failure-free survival (FFS): median = 2 years  Overall survival (OS): median = 4 years

Implementation of MAMS 75 Mahesh Parmar, Feb-2010 Traditional Approach Phase II Phase III ABC Multi-arm, Multi-stage ABC Phase II = 50 pts, Phase III = 600pts 4*50 + 4*300 = 1400 Traditional vs MAMS = = = = =1950 Total

STAMPEDE Joint TSC/IDMC Meeting76 Matt Sydes, 20-July-2009 ContinueStop Estimated Treatment Effect (HR) (Favours experimental)(Favours control) 1-sided 75% CI around 0.92 End of Stage II if HR>0.92 A+B B A x x x Example Stop A Stop A+B Stop B Action?

STAMPEDE Joint TSC/IDMC Meeting77 Matt Sydes, 20-July-2009 ContinueStop Estimated Treatment Effect (HR) (Favours experimental)(Favours control) 1-sided 75% CI around 0.92 End of Stage II if HR>0.92 A+B B A x x x Example Action? Continue A Continue A+B Continue B

STAMPEDE Joint TSC/IDMC Meeting78 Matt Sydes, 20-July-2009 ContinueStop Estimated Treatment Effect (HR) (Favours experimental)(Favours control) 1-sided 75% CI around 0.92 End of Stage II if HR>0.92 A+B B A x x x Example Requires exploration and intra-arm comparisons Stop A Stop A+B Continue B Action?

STAMPEDE Joint TSC/IDMC Meeting79 Matt Sydes, 20-July-2009 ContinueStop Estimated Treatment Effect (HR) (Favours experimental)(Favours control) 1-sided 75% CI around 0.92 End of Stage II if HR>0.92 A+B B A x x x Example Continue A Continue A+B Continue B Action?

STAMPEDE Joint TSC/IDMC Meeting80 Matt Sydes, 20-July-2009 ContinueStop Estimated Treatment Effect (HR) (Favours experimental)(Favours control) 1-sided 75% CI around 0.92 End of Stage II if HR>0.92 A+B B A x x x Example Action? Continue A Continue A+B Continue B

STAMPEDE Joint TSC/IDMC Meeting81 Matt Sydes, 20-July-2009 Activity phase II stages At any time should be more intermediate events than definitive events –eg more FFS events than deaths (OS) Therefore, more power for intermediate outcome measure at any time Design assumes: –To see an effect on OS you have to see an effect on FFS –Just because you see an effect on FFS does not mean that you will see an effect on OS

STAMPEDE Joint TSC/IDMC Meeting82 Matt Sydes, 20-July-2009 Advantages of MAMS Fewer patients Less overall time

STAMPEDE Joint TSC/IDMC Meeting83 Matt Sydes, 20-July-2009 Traditional Approach Phase II Phase III ABC Multi-Arm, Multi-Stage ABC Phase II = 50 pts, Phase III = 600pts 4*50 + 4*300 = 1400 Traditional vs MAMS = = = = =1950 Total

STAMPEDE Joint TSC/IDMC Meeting84 Matt Sydes, 20-July-2009 Traditional Approach Phase II Phase III ABC Multi-Arm, Multi-Stage ABC Phase II = 50 pts, Phase III = 600pts 4*50 + 4*300 = 1400 Traditional vs MAMS = = = = =1950 Total

STAMPEDE Joint TSC/IDMC Meeting85 Matt Sydes, 20-July-2009 Cutpoints in STAMPEDE 95% power

STAMPEDE Joint TSC/IDMC Meeting86 Matt Sydes, 20-July-2009 Cutpoints in STAMPEDE 95% power

STAMPEDE Joint TSC/IDMC Meeting87 Matt Sydes, 20-July General issues in implementation 1.MAMS Designs 2.MAMS application in STAMPEDE 3.Issues in implementation 4.General issues in implementation 5.Issues in intermediate analyses 6.Conclusions

STAMPEDE Joint TSC/IDMC Meeting88 Matt Sydes, 20-July-2009 No intermediate outcome? Design depends on the use of an intermediate outcome What happens if no such outcome exists? Use the primary outcome, earlier in time –Lack-of-benefit analysis useful anyway –Applying this to a number of ongoing trials

STAMPEDE Joint TSC/IDMC Meeting89 Matt Sydes, 20-July-2009 How many arms? Could be many arms –From 2 to 10 or more Consider –Accrual rates –Rationale for inclusion –Adjust randomisation ratio

STAMPEDE Joint TSC/IDMC Meeting90 Matt Sydes, 20-July-2009 Arms to compare Different drugs –Classes of agent –Combinations of agents Same drug –Dose levels –Duration –Initiation time –Method of administration Non-drug therapy

STAMPEDE Joint TSC/IDMC Meeting91 Matt Sydes, 20-July-2009 Adding new arms Could an extra arm be added? –Promising agents could be added later! –“Rolling trial” or “roundabout design”? –Use same rules as for other arms, but delayed Exploring in STAMPEDE Design should be appealing to industry –Pre-existing network of recruiting sites –GSK?

STAMPEDE Joint TSC/IDMC Meeting92 Matt Sydes, 20-July-2009 Stopping for efficacy? Stopping rules specified for lack-of-benefit No formal stopping rules for efficacy –Early data looking for sufficient evidence of activity to support continued investment –Could draw in usual conservative early stopping guidelines

STAMPEDE Joint TSC/IDMC Meeting93 Matt Sydes, 20-July-2009 STAMPEDE TMG Current TMG members Nick JamesKaren Sanders Noel ClarkeRachel Morgan David DearnaleyRichard Gracie Max ParmarJim Stansfeld Matt SydesJohn Dwyer Malcolm Mason John Anderson Rick Popert

STAMPEDE Joint TSC/IDMC Meeting94 Matt Sydes, 20-July-2009 How long to get here? Oct 2000 = first discussion Nov 2000 = first formal meeting Oct 2005 = first patient randomised Why so long? –Some issues of design ie MAMS-specific issues –Some issues of collaboration –Some issues beyond control

STAMPEDE Joint TSC/IDMC Meeting95 Matt Sydes, 20-July-2009 Hurdles 1.Funding

STAMPEDE Joint TSC/IDMC Meeting96 Matt Sydes, 20-July-2009 Hurdles 1.Funding 2.Finalising choice of drug

STAMPEDE Joint TSC/IDMC Meeting97 Matt Sydes, 20-July-2009 Changes in drugs 1 st mtg = Nov st appl n = May 2001 Hormone therapy Mitoxantrone (MTZ)MTZ+pred DocetaxelDocetaxel + pred Prednisolone- StilboestrolStilboestrol + aspirin EstramustineEstramustine + aspirin ?Bisphosphonates (factorial)?Bisphosphonates ?Herceptin- ?ECF- ?Strontium- ?Epirubicin-

STAMPEDE Joint TSC/IDMC Meeting98 Matt Sydes, 20-July-2009 Changes in drugs 1 st appl n = May nd appl n = ??? Hormone therapy MTZ+predMTZ+pred OR Docetaxel + pred Stilboestrol + aspirinStilboestrol + aspirin OR Estramustine + aspirin ?Bisphosphonates ?Iressa ?Atrasentan ?Celecoxib

STAMPEDE Joint TSC/IDMC Meeting99 Matt Sydes, 20-July-2009 Changes in drugs 2 nd appl n = ???Final appl n = Nov 2002 Hormone therapy MTZ+pred OR- Docetaxel + pred Stilboestrol + aspirin OR- Estramustine + aspirin ?BisphosphonatesZoledronic acid ?Iressa- ?Atrasentan- ?CelecoxibCelecoxib Docetaxel + zoledronic acid Celecoxib + zoledronic acid

STAMPEDE Joint TSC/IDMC Meeting100 Matt Sydes, 20-July-2009 Change in population Originally new M1 only Expand to new M1 + new M0 Expand to new M1 + new M0 + relapsing

STAMPEDE Joint TSC/IDMC Meeting101 Matt Sydes, 20-July-2009 Hurdles 1.Funding 2.Finalising choice of drug 3.Discussions with pharma

STAMPEDE Joint TSC/IDMC Meeting102 Matt Sydes, 20-July-2009 Hurdles 1.Funding 2.Finalising choice of drug 3.Discussions with pharma 4.Funding (revisited) 5.EU Clinical Trials Directive

STAMPEDE Joint TSC/IDMC Meeting103 Matt Sydes, 20-July-2009 EU Clinical Trials Directive May 2004 Many uncertainties over sponsorship Application for regulatory approval –Put in before new system to get CTA-rollover –Easier in old system? –Less easy not for old system trials

STAMPEDE Joint TSC/IDMC Meeting104 Matt Sydes, 20-July-2009 Hurdles 1.Funding 2.Finalising choice of drug 3.Discussions with pharma 4.Funding (revisited) 5.EU Clinical Trials Directive 6.Increased number of MAMS stages

STAMPEDE Joint TSC/IDMC Meeting105 Matt Sydes, 20-July-2009 Increased stages Planned as two-stage trial (plus Pilot) IDMC to meet annually Therefore, plan more interim hurdles Increased from 2 to 4 stages –Required new program –-stage2- to -stagen- –Revise protocol Revised again in Summer 2007 –-stagen- to -nstage- –Small decrease in number of required events

STAMPEDE Joint TSC/IDMC Meeting106 Matt Sydes, 20-July-2009 Changes in trial personnel NameInvolvement Nick JamesTMG (CI) Noel ClarkeTMG David DearnaleyTMG Max ParmarTMG Matt SydesTMG Sarah MeredithMoved on Sharon NaylorMoved on David KirkRetired (TSC) Clare MoynihanMoved on

STAMPEDE Joint TSC/IDMC Meeting107 Matt Sydes, 20-July Recruitment rates Accrual/ Year Total ptsDuration (yrs) Difference in pts Difference in length months months Assuming: median FFS=24m, median OS=48m, 6 arms recruiting at each stage

STAMPEDE Joint TSC/IDMC Meeting108 Matt Sydes, 20-July-2009 Key points – 1 For many diseases we have many potential new treatments Majority are likely to prove no more effective than control We need to change the question we ask –How do we improve outcomes as rapidly as possible for this disease?

STAMPEDE Joint TSC/IDMC Meeting109 Matt Sydes, 20-July-2009 Key points – 2 Intermediate outcome can be very helpful –Need not be true surrogate –Often assume larger effect size on intermediate outcome MAMS trials speed evaluation of new treatments by testing many treatments at the same time and using lack of benefit analyses