Presentation is loading. Please wait.

Presentation is loading. Please wait.

Valerie Durkalski Medical University of South Carolina

Similar presentations


Presentation on theme: "Valerie Durkalski Medical University of South Carolina"— Presentation transcript:

1 Valerie Durkalski Medical University of South Carolina
The Pros/Cons of Implementing Response Adaptive Randomization in Your Clinical Trial Valerie Durkalski Medical University of South Carolina

2 Swiss Army knife Clinical Trial Toolbox
Importance of conducting simulation studies Just as important….. Resources for these simulation studies(when needed) Basic vs complex simulations Expectations and education of our collaborators Industry or Academic settings Affordable toolboxes Planning grants Methods grants

3 Incorporating RAR into your clinical trial
Planning Aspects Why RAR? (check all that apply) Innovative checkbox Ethical advantages for the participants Efficiency (in terms of power) Because it is in our toolbox

4 Personal Life Lesson #1 The EPISOD Trial (NCT 00688662) – funded 2005
RAR: innovative checkbox….and assumed patient ethics Binary outcome – complex definition of success in a sham surgical trial (12month follow up; N=214; ~7sites) Disability score <6 (month 9 and 12 visits) No narcotic use in past 6 months for abdominal pain No reinterventions Burn in period (n= 25% of enrollments) Enrollment delays, narcotic information difficult to obtain and required monitoring and bad news travels faster than good news. Discussions with DSMB; ended up never implementing RAR. Motivated me to take a harder look at why RAR

5 Personal Life Lesson #2 SHINE Trial (NCT 01369096, funded 2011)
Binary outcome (90-day mRS; N=1400; ~60 sites) Single blinded with blinded outcome assessment Group sequential design with frequent interim looks, blinded N-reestimation, RAR (Bruno, Durkalski, et al 2013) RAR: innovative checkbox / simulations studies RAR with covariate balancing (Zhao and Durkalski, 2015)

6 Personal Experience #3 NINDS-funded Clinical Trial Network (2006)
Neurological Emergencies Treatment Trial Network (NETT) Primarily confirmatory trial setting Blinded N re-estimation, RAR with covariate balancing, bayesian approach with frequent looks, seamless two-stage, multi-arm (no control) NIH-FDA Grant - Advancing Regulatory Science through Novel Research and Science Based Technologies (2010) Studied the design phase of the adaptive process Four confirmatory trial settings - focus groups, simulation studies, grant submissions Shadow designs

7 Personal Experience #4 PhD dissertation topic: Yunyun Jiang
The adaptive allocation methods vary across studies, the performance of different approaches remain unclear with respect to the overall trial operating characteristics and treatment effect estimation. Compare various RAR formulas with respect to overall operating characteristics and treatment effect estimation. What is the expected final allocation and its variability? What is the impact on the type I error and power? Total #failures as well as prob(more in inferior arm)? When and how often should the allocation ratio be updated? Every one subject or every M subjects?

8 Current Literature Potential benefits Debates
Reduce patient risk of receiving inferior treatment. Improve the efficiency of identifying treatment effect. Debates Trade-off between statistical power and patients’ ethics. (Yin et al 2012) May assign more patients to the inferior treatment. (Thall et al. 2015) Updating frequency impact of type I error – every subject or every group/block? (Thall et al. 2015; Conner et al 2013) Bias in treatment effect estimation. (Thall et al 2015, Bowden et al 2015)

9 Two vs Multi-Arm Two-arm trials:
Performance variability. (Thall and Wathen 2007 et al; Lee et al 2012) Clear trade-off between statistical power and patient ethics. (Yin et al 2012) Can end up with an increased total number of failures compared to equal allocation (Zhao and Durkalski, 2015) The actual benefit of RAR is questionable Multi-arm trials Greater potential benefits for RAR in both trial efficiency and patient ethics. (Berry et al. 2011; Connor et al 2013) Optimal allocation varies based on trial setting. treatment effect/frequency update/approach No simple generalizable solution available. (Joen and Hu et al 2009) Extensive simulation study for each setting to understand the potential peformance

10 RAR Implementation aspects

11 Incorporating RAR into your clinical trial
Implementation Aspects “Further factors beyond methods will impact the choice for or against a certain design…” Drug re-supply and tracking…..impact on budget Updating frequency and data validation ITT or As treated Handling of missing data Complexities in implementation equally important May be able to do it but why?

12 Things to Consider Drug supply/resupply
Ability to resupply without breaking the treatment blind When planning, consider keeping some study team members blind to this process Estimate cost of resupply and ‘wasting of drug supply’ Fully integrated clinical trials management system If more frequent update of allocation, consider impact of data validation. Related to enrollment rate and outcome collection schedule.

13 Things to Consider ITT vs PP Safety implications
Expected cross over/non-adherence rate Consider impact in your simulation study Safety implications Handling of missing data Similar to primary analysis plan? Details of RAR plan (burn in, updating frequency) Consider not sharing this with the entire study team particularly if single blinded study design

14 Experience #5 ESETT (NCT01960075)
Bayesian adaptive CE trial conducted within the NETT (Connor, Elm et al 2013) Adaptive randomization with potential early stopping Three active arms (no control) Extensive simulation studies

15 Summary We are learning a great deal about RAR…but still aspects to uncover!

16 Thank you NETT Network (NINDS U01 050941)
ADAPT-IT Project (FDA/NIH/NCATS U ) SHINE Trial (NINDS U ) Yunyun Jiang and Wenle Zhao


Download ppt "Valerie Durkalski Medical University of South Carolina"

Similar presentations


Ads by Google