© Guidant 2004 Medical Device versus Drug Similarities and Differences Jeng Mah, David Breiter Guidant Corporation FDA Industry Workshop September 16,

© Guidant 2004

Medical Device versus Drug

Similarities and Differences

Jeng Mah, David Breiter

Guidant Corporation

FDA Industry WorkshopSeptember 16, 2005

2 © Guidant 2004

Workshop 2005

Device vs. Drug - physical

Device Drug

3 © Guidant 2004

Workshop 2005

Device vs. Drug - functional

Device Actions

• Mechanical

• Physical

• Dynamic

• Adaptive

• User dependent

Device Effects

• Local

• Direct/immediate

• Measurable

• Reversible

Drug Actions

• Chemical

• Physiological

• Fixed

• Not adaptive

• Simple

Drug Effects

• Systemic

• Indirect/deferred

• Difficult to measure

• irreversible

4 © Guidant 2004

Workshop 2005

Device vs. Drug - statistical

Device Studies

• May perform multiple adaptive functions (bundled features)

• Extensive & informative bench and acute tests

• Non-blinded, non-placebo pivotal studies

• Automation & decision making

• Subject specific optimal programming

• User interface/skill affects/determines out-comes

• Real time data collection generates lots of data

Drug Studies

• One drug one desired effect (usually), or deal with drug interactions

• Phase I, II studies serve different purposes

• Active control, blinded pivotal studies

• Subjects receive identi-cal, fixed treatments

• Data collection focuses on final, pragmatic outcomes

5 © Guidant 2004

Workshop 2005

Device vs. Drug - statistical

Device Studies

• Difficult to administer

• FDA requirement: single pivotal trial

• Reliability and quality control issues

Drug Studies

• Easy to administer

• FDA requirement: two pivotal trials

• Reliability and quality control issues are less prominent

6 © Guidant 2004

Workshop 2005

Totest an ‘intelligent’ device with a new feature: the new feature include one programmable parameter; physician needs to find the optimal setting of each patient based on the responses of an acute test. This parameter can be modified anytime.

Questions: how do we model a true ‘device effect’ adjusted for the physician, parameter, and patient? What is the ‘treatment’ we are interested in? Feature? Parameter? Programmability? Utility? What is the experimental unit? When do we need to do new studies?

An Example