Harvard  Catalyst  Adap.ve  Clinical  Trials  

Design  and  Implementa.on  

Zoran  Antonijevic    

•  Overview  of  Novel  Adap.ve  Designs  •  Implementa.on  •  DMC  issues  

Cytel  2013   Harvard  Catalyst  31May2013   2  

Presentation Contents

Cytel  2013   Harvard  Catalyst  31May2013   3  

   

Overview  of    Novel  Adap.ve  Designs  

Definition of an Adaptive Design

•  A  study  design  is  called  "Adap.ve"  if  it  allows  modifica.on  of  an  essen.al  design  feature  (e.g.,  sample  size,  randomiza.on  ra.o,  number  of  treatment  arms),  based  on  accruing  data  from  within  that  clinical  trial  –  Should  be  carried  out  without  compromising  the  integrity  of  the  trial  

–  For  confirmatory  trials  full  details  of  adapta.ons  need  to  be  pre-­‐specified  

Cytel  2013   4  Harvard  Catalyst  31May2013  

•  Also  known  as  “Less-­‐Well  Understood”  Adap.ve  Designs  (DraX  FDA  AD  Guidance):  

•  Phase  2  –  Adap.ve  Dose  Selec.on  Studies  

•  Phase  3  –  Unblinded  Sample  Size  Re-­‐Assessment  –  Confirmatory  Trials  with  Dose  Selec.on  (Seamless  Phase  2b/3)  

–  Confirmatory  Trials  with  Sub-­‐Popula.on  selec.on  (Adap.ve  Enrichment  Design)  

 Cytel  2013   Harvard  Catalyst  31May2013   5  

Novel Adaptive Designs

When  applicable:  entering  Phase  3  there  is  s.ll  residual  uncertainty    regarding  the  treatment  effect  

GSD  vs.  Unblined  SSR:  1.  Both  address  the  same  

problem  2.  With  GSD  one  commits  to  a  

larger  study  ini.ally,  can  stop  early    if  efficacy  becer  than  assumed,  or  for  fu.lity  

3.  With  USSR  one  ini.ally  assumes  a  more  op.mis.c  effect,  increases  the  sample  size  at  the  interim  as  needed.    

Unblinded Sample Size Re-estimation

2  Harvard  Catalyst  31May2013  

When  applicable:  entering  Phase  3    there  isn’t  enough  evidence  that    Biomarker  subpopula.on  would  perform  (substan.ally)  becer  than  complementary  (or  full)  popula.on  

Randomiza.on  stra.fied  by  the  Biomarker  status.  At  the  interim  the  enrolment  in  the  complementary  sub    may  be  stopped  as  guided  by  the  data.    

IA50%

IA70%

Stop for E!cacy

Stop for Futility

Continueas Planned

IncreaseSample Size

EnrichPopulation

Confirmatory Trial with Sub-population Selection (Adaptive Enrichment)

2  Harvard  Catalyst  31May2013  

When  applicable:  aXer  comple.on  of  Phase2    there  is  s.ll  residual  uncertainty    regarding  the  best  dose    Formerly  known  as  “Seamless  Phase  2b/3,  and  has  caused  some  controversy.  Now  actually  recommended  in  the  guidance.    

Preferably  ini.ated  as  a  follow-­‐up  to  Phase  2b,  although  many  sponsors  s.ll    use  it  as  a  replacement  for  (or  skipping)    Phase  2b.  

 

2  

Confirmatory Trial with Dose Selection

Harvard  Catalyst  31May2013  

Cytel  2013   Harvard  Catalyst  31May2013   9  

     

Implementa.on  

•  Required  exper.se  –  Sta.s.cal  methodology  –  Independent  data  monitoring  commicee  (IDMC)  involving  both  clinicians  and  

sta.s.cians  –  Experienced  opera.onal  team  to  manage  the  logis.cal  complexi.es  and  firewalls  to  

limit  poten.al  opera.onal  biases  •  Adequate  planning  .me  –  Pre-­‐study  simula.ons  –  Early  communica.on  with  the  regulators  for  confirmatory  trials  (e.g.  SPA,  Parallel  

Scien.fic  Advice  -­‐  FDA/EMA)  –  Document  prepara.on  (DMC  Charter  and  SAP  submission)  and  opera.onal  setup  •  Choosing  the  right  technology  –  IVRS  and/or  IWRS:  Drug  logis.c  management  and  randomiza.on  –  EDC:  Real-­‐.me  access  to  high-­‐quality  data  –  Endpoint  Adjudica.on:  Centralized  processing  and  independent  review  –  Integrated  Workflow  Management  System:  Strict  security  levels  (firewalls)  to  ensure  

data  and  trial  integrity  in  compliance  with  regulatory  guidance  Cytel  2013   Harvard  Catalyst  31May2013   10  

Key Considerations when Planning Adaptive Studies

•  Key  issues:  – Who  knows  what  and  when?  –  Access  control  and  verifica.on  –  Firewalls    –  Storage  of  DMC  materials  –  Communica.on  of  Results  and  Recommenda.ons  

Cytel  2013   11  

Data monitoring of Adaptive Clinical Trials

Harvard  Catalyst  31May2013  

Cytel  2013   12  

Improving the Process

ISC  

DMC  

Sponsor  

EDC   IRT   CTMS  

Harvard  Catalyst  31May2013  

Cytel  2013   Harvard  Catalyst  31May2013   13  

     

DMC  Issues  

•  The  FDA  Adap.ve  Design  Guidance:  “Because  the  DMC  is  unblinded  to  interim  study  results  it  can  help  implement  the  adapta2on  decision  according  to  the  prospec2ve  adapta2on  algorithm,  but  it  should  not  be  in  a  posi.on  to  otherwise  change  the  study  design  except  for  serious  safety-­‐related  concerns  that  are  the  usual  responsibility  of  a  DMC”  

Cytel  2013   14  

DMC Role

Harvard  Catalyst  31May2013  

•  Similari.es  with  GSD:  -­‐  The  adapta2on  algorithm  can  be  pre-­‐specified  in  a  rela2vely  straigh>orward  manner      

-­‐  Both  based  on  the  unblinded  results  of  the  treatment  effect  -­‐  The  pre-­‐specified  sta.s.cal  algorithm  can  be  compromised  by  safety  findings  for  both  of  these  designs  

•  Differences:  –  GSD;  concerns  re  overflow  –  SSR;    design  carefully  not  to  reveal  treatment  effect  

Cytel  2013   15  

Unblinded Sample Size Reassessment

Harvard  Catalyst  31May2013  

•  Decisions  involve  parameters  beyond  just  ethical  considera.ons.  •  Include  decisions  that  tradi.onally  have  been  sponsor  

responsibili.es.    •  May  have  major  commercial  implica.ons,  for  example  dose  

selec.on  or  sub-­‐popula.on  selec.on.    •  Complex  algorithms  •  DMC  may  desire  to  consult  with  the  steering  commicee  that  

includes  a  sponsor  representa.ve  not  involved  in  trial  management.        

Cytel  2013   16  

Complex Adaptive Designs at Confirmatory Stage

Harvard  Catalyst  31May2013  

•  Not  intended  to  demonstrate  confirmatory  evidence;  less  rigor  in  restric.ng  knowledge  of  interim  results    

•  Lesser  regulatory  implica.ons,  though  avoiding  opera.onal  bias  should  be  recommended  in  any  trial  

•  May  require  frequent  updates  based  on  interim  data  •  Adapta.ons  are  oXen  driven  by  a  pre-­‐specified  algorithm  

automated  through  an  integrated  response  system    •  Therefore,  it  may  be  preferable  that  recommenda.ons  are  

endorsed  by  an  Internal  Review  Commicee  

Cytel  2013   17  

Adaptive Designs in Exploratory Studies

Harvard  Catalyst  31May2013  

•  US  Food  and  Drug  Administra.on.    Guidance  for  industry:  adap.ve  design  clinical  trials  for  drugs  and  biologics  (draX  guidance).  February  2010.  Available  online  at  hcp://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInforma.on/Guidances/UCM201790.pdf  

 

Cytel  2013   Harvard  Catalyst  31May2013   18  

References

Cytel  2013   Harvard  Catalyst  31May2013   19  

     

Back-­‐up  Slides  

Value  of  Design  •  Easily  under-­‐appreciated,  trial  designs  have  a  major  impact  

on  the  value  of  a  pharmaceu.cal  product  

Adap.ve  Phase  2  Designs  •  The  FDA  AD  Guidance  recommends  use  of  AD  in  Phase  2  •  AD  is  the  most  efficient  approach  for  this  stage  of  

development  

Development  Program  Op.miza.on  •  Phase  2  most  cri.cal  for  op2mal  drug  development    

Development  Scenario  Simula.on  •  Comparison  of  rela.ve  efficiency  of  various  development  

op2ons  can  be  accomplished  only  with  simula2ons    

  Harvard  Catalyst  31May2013   20  

Phase 2b Dose Finding Studies

Cytel  2013  

Scenarios  –  (7)  different  sta.s.cal  approaches  –  (2)  Phase  2  strategies  –  (3)  Phase  3  strategies  –  For  a  total  of  42  scenarios.  Same  assump.ons  for  safety  and  efficacy  used  for  all  scenarios  

–  Cost  and  development  .me  calculated  using  real  data  from  exis.ng  databases  

Results  –  The  expected  NPV  for  different  scenarios  ranged  from  $0.5B  to  $1.2B  

–  Adap.ve  Designs  far  outperformed  tradi.onal  approaches  

Case Study

Harvard  Catalyst  31May2013   21  Cytel  2013  

Expected NPV

22  Average NPV (millions)

ANOVA

Dopt

GADA

MCPMod

MTT

BMA

LOCFIT

600 800 1000

N = 150

ANOVA

Dopt

GADA

MCPMod

MTT

BMA

LOCFIT

N = 250

1 dose 2 doses, fast 2 doses, normal

Harvard  Catalyst  31May2013  Cytel  2013  

Method