1 Incorporating Virtual Patients into Clinical Studies Adam Himes, Tarek Haddad, Medtronic Laura Thompson 1 , Telba Irony 2 , Rajesh Nair 1 1 CDRH / FDA, 2 CBER / FDA on behalf of MDIC working group colleagues: Dawn Bardot, MDIC Anita Bestelmeyer, BD Dan Cooke, Boston Scientific Mark Horner, ANSYS Russ Klehn, St. Jude Medical Tina Morrison, OSEL / FDA Kyle Myers, OSEL / FDA Val Parvu, BD MDIC.org
15
Embed
Incorporating Virtual Patients into Clinical Studies Himes.pdf · Virtual Patient R Code Access and Use: Public release of the Virtual Patient R Code (on CRAN) with demonstrations
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Incorporating Virtual Patients into Clinical Studies Adam Himes, Tarek Haddad, Medtronic Laura Thompson 1 , Telba Irony2 , Rajesh Nair1 1CDRH / FDA, 2CBER / FDA
on behalf of MDIC working group colleagues: Dawn Bardot, MDIC Anita Bestelmeyer, BD Dan Cooke, Boston Scientific Mark Horner, ANSYS Russ Klehn, St. Jude Medical Tina Morrison, OSEL / FDA Kyle Myers, OSEL / FDA Val Parvu, BD
“If it can be shown that these virtual patients are similar, in a precisely defined way, to real patients, future trials may be able to rely partially on virtual-patient information, thus lessening the burden of enrolling additional real patients.”
• Many applicable models, implantable defibrillator lead fracture is a good example − Straightforward − Relevant − Public domain examples
Haddad, et.al., Reliability Engineering and System Safety, 123 (2014): 145-157. Swerdlow, et.al., JACC, 67 (2016): 1358-1368.
8
Virtual Patient Example: ICD Lead Fracture
• Simulate many combinations of virtual patients & clinical trial
• Propagate variability and uncertainty to predict survival with confidence bounds
Field data
Projection with 95% Confidence
Interval
in-vivo bending
patient activity
fatigue strength
INPUT OUTPUT
Haddad, et.al., Reliability Engineering and System Safety, 123 (2014): 145-157
9
Bayesian Statistical Methods
• How much influence is given to the prior data?
• What if the clinical study data disagrees?
Challenges: Solution: • Method developed by MDIC
working group
• Influence of prior data determined by agreement with study data
• Maintain statistical power with fewer patients
influ
ence
disagree agree disagree agree
(ideal state)
prior data
study data
discount function
Provide a way to incorporate prior data into analysis of a clinical study
10
Incorporate Virtual Patients: Step By Step
1. Compare virtual patient and current data
2. Compute strength of historical data using discount function
3. Combine virtual patient and current data
4. Statistical analysis using combined data
virtual patient data
current data
𝑝𝑝
𝑝𝑝
𝛼𝛼0
𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = 𝑛𝑛𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑡𝑡 + 𝛼𝛼0𝑛𝑛𝑉𝑉𝑉𝑉
combined data
This part is new
Haddad, et.al, J. Biopharm Stat (2017) DOI: 10.1080/10543406.2017.1300907
11
Implementation: Mock Submission • Collaboration between MDIC and FDA CDRH Division of Cardiovascular Devices • Demonstrate the engineering and statistical framework for virtual patients • MDIC sponsor team includes industry and FDA • FDA review team, just like for a real device
MDIC Virtual Patient Statistical Framework: 2017 Communication Activities
PAST PRESENTATIONS: • Cardiovascular Research Technologies (CRT17): Leveraging Existing Information for Future
Studies: The Case for Bayesian Methods • AdvaMed Innovations Summit: Innovation in Clinical Evidence Generation, Synthesis and
Appraisal to Advance Regulatory Science for the Total Product Life Cycle • 10th Annual FDA/AdvaMed Medical Devices & Diagnostics Statistical Issues Conference:
Bayesian and Adaptive Designs UPCOMING PRESENTSTIONS • Joint Statistical Meeting (JSM): Improving the Efficiency of Medical Device Clinical Trials by
Combining Simulations and Experiments. Baltimore, 8/01/17, 2:00 PM - 3:50 PM ONLINE MDICx SERIES EVENTS: mdic.org/MDICx
• Leveraging Existing Information for Future Studies: The Case for Bayesian Methods: Encore
presentation from CRT17 with Case Study and Bayesian/Adaptive Regulatory negotiation updates. • Virtual Patient R Code Access and Use: Public release of the Virtual Patient R Code (on CRAN)
with demonstrations applying the model to various study types
15
Summary
• Virtual patients can improve the clinical decision process while exposing fewer patients to clinical trials
• Discount function controls the influence of virtual patients
• The statistical methods are ready – we just need the right applications!
• Without collaboration between FDA and industry, we would not be here today!