CUI 2006 MUFPADA

The Use of a Clinical Utility Index (CUI) in Decision Making to Select an Insomnia Compound: 

The Use of a Clinical Utility Index (CUI) in Decision Making to Select an Insomnia Compound Daniele Ouellet, Ph.D. Clinical Pharmacology MUFPADA May 3, 2006

Clinical Utility Index Approach: 

Clinical Utility Index Approach Rowland and Tozer, 1989, Chapter 5 Not a new concept, however, only few examples

Another Example…: 

Another Example… Graham G, Gupta S, and Aarons L J Pharmacokin Pharmacodyn 2002

Background: 

Background 2 compounds with same mechanism of action for treatment of insomnia Strategy Advance leader as long as it remains market viable and/or back-up clearly differentiable Status after FIH Compound 2 less potent but potentially with a better profile than Compound 1 Phase 2a study planned for Compound 2 Design nearly identical

Background - Continued: 

Background - Continued Question to Clinical Team: How to make decision? Decide a priori Communicate to team / governing bodies Clinical Utility Index (CUI) Integrated measure of benefit/risk over dose range Proposed as a tool to compare compounds Translate clinical judgment into systematic measure for decision making

POC Study: 

POC Study Similar Study Design Primary Chronic Insomnia Patients (N=43) 4- or 5-way cross-over (1-week washout) 2 doses of Compound 1, PBO, AC 3 doses of Compound 2, PBO, AC Differences in Expected Outcome Dose-Response Analysis to Account for Differences in Dose/Baseline Different Efficacy on Different Endpoints Use CUI to Balance Risk/Benefit

Differences in insomnia severity between studies: 

Differences in insomnia severity between studies BSL  Difference BSL  Difference Dose Response Analysis

Development of CUI: 

Development of CUI How to select endpoints Efficacy Residual Effects Consider number of attributes / correlation Dose response analysis Account for baseline differences Account for different doses How important is each attribute Weighting of attributes How to get consensus Adding apples to oranges Normalize by meaningful difference

Many Clinically Relevant Outcomes…: 

Many Clinically Relevant Outcomes… Getting to Sleep Latency to persistent sleep (LPS)* Subjective latency to sleep Ease of getting to sleep (Leeds Questionnaire) Sleep Time Total sleep time (TST)* Subjective sleep time Sleep efficiency* Residual Effects (morning after dosing) Digit Symbol Substitution Test (DSST) Leeds Questionnaire (ease of awakening and early morning behavior) Sleep Consolidation # Awakenings* Wake after sleep onset (time)* Subjective # of awakenings # Arousals* Sleep Architecture % Time in sleep stages 1, 3-4 * % Time in sleep stages 2, REM* Latency to REM* Sleep Maintenance Sleep efficiency by period* Quality of Sleep Leeds Questionnaire Subjective rating

Market Research vs. CUI: 

Market Research vs. CUI Results of Market Research Available 581 Physicians Rank relative importance of different attributes by hybrid conjoint analysis (1 to 20) Attributes Tested Different Measures of Efficacy (sleep onset, sleep quality, sleep architecture) Different Measures of Safety (hangover effects, memory impairment, abuse potential) Other such as drug-drug interaction, etc. Map Attribute to Clinical Measure in POC study Help Build Consensus across Team Members

Calculation of CUI: 

CUI = Calculation of CUI - Determine shape of CUI Wgt = Relative importance as % determined by market research CLIN = clinically important difference k = number of attributes POS and NEG = change from pcb in positive and negative attributes

CUI - Methods: 

CUI - Methods Dose Response Analysis 7 clinical endpoints Bootstrap (N=500) 500 sets of parameters (e.g., Emax, ED50, PCB, SIG) Calculate CUI vs. Dose Normalize for Study Baseline (PCB) Apply weight and clinically meaningful difference used to construct CUI Calculate Peak CUI for each Bootstrap Calculate Difference Between Compounds for Peak CUI for each Bootstrap

Dose Response  CUI: 

Dose Response  CUI  Drive CUI Down  Drive CUI Up

Slide14: 

CUI and Confidence Interval Minimum Clinically Relevant Benefit Start of Diminishing Returns Recommended Doses 10-90th percentiles of bootstrap dose response

Sensitivity Analysis: 

Sensitivity Analysis Test Alternate Profiles of Efficacy/Residual Effects and Compare CUI Curves Alternate Profiles Tested:  Less Residual Effects  Greater Efficacy on onset (LPS)  Differences on Sleep Architecture  Less Efficacy on maintenance (WASO)  Combination of All of the Above

Sensitivity Analysis: 

Sensitivity Analysis Difference of CUI of 0.1  Clinically Relevant

Slide17: 

Median CUI and Uncertainty Dotted Line is Alternate Profile; Dot-Dash is another compound (worse profile); Symbols are Observed Data; Solid Line is Predicted Profile

Slide18: 

Results: Compound 1 vs. 2 Median (Solid Line) and 80% CI (Shaded Area); Symbols are Observed Data CUI adjusted for placebo/baseline from Study 2 Peak CUI of 0.345 (0.25-0.43) Peak CUI of 0.436 (0.35-0.52) CUI (400 mg) of 0.324 (0.28-0.37)

Slide19: 

Difference in Peak CUI (N=500 Bootstraps) Median (80% CI) 0.09 (-0.03, 0.21)

Recommendation: 

Recommendation Peak CUI for Compound 2 observed at doses > 600 mg Although Compound 2 has higher peak CUI value than Compound 1, the improvement is not observed at a dose that is viable. Recommendation was made to discontinue Compound 2 as a back up compound for insomnia

Use of CUI: 

Use of CUI Quantitative Approach to integrate efficacy/safety Go/No Go Criteria (Phase 2a) Select Optimal Dose (End of Phase 2b) Characterize therapeutic Index (width) Compare Population (elderly vs. non-elderly) Adjust Endpoints along the way

Slide22: 

Acknowledgements Neha Patel John Werth Doug Feltner Bruce McCarthy Randy Stone David Mitchell Richard L. Lalonde Departments of Clinical Pharmacology, Clinical R&D, and PGP

Slide23: 

Questions? Comments?