Twenty Five Years Designing Clinical Trials . and Still Learning - PowerPoint PPT Presentation

Twenty Five Years Designing Clinical Trials …. and Still Learning Roy N. Tamura Health Informatics Institute University of South Florida

Career Timeline Type 1 Diabetes, Rare Diseases USF - Health Informatics Strattera Cymbalta Zyprexa Eli Lilly Clinical Eli Lilly Non-Clinical Finished PhD ! Started at NCSU 1978 1983 1992 2012

Clinical Trial Design Advancements – Past 25 Years 1. Group Sequential Designs: alpha spending functions, efficacy monitoring futility analyses 2. Adaptive Designs: sample size, randomization, stratification 3. Enriched Designs: biomarker based, response based 4. Multi-stage Designs: SMART (sequential multiple assignment randomized trials) and many other useful proposals.

NIH Rare Disease Network Project Isolated Skin Vasculitis: No current effectiveness information on any drug Trial: Compare dapsone, colchicine, azathioprine over six month period. Goal: Find the best drug and compare that drug with the next best drug. Slide 4 of 11

ARAMIS - Small n Sequential Multiple Assignment Randomized Trial (snSMART) Stage 1 Stage 2 responders M6 continue Azathioprine Azathioprine N = 30 non-responders Dapsone randomized 1:1 Colchicine Enrollment Dapsone responders M6 continue Dapsone Randomization N = 30 n=90 (1:1:1) non-responders Azathioprine randomized 1:1 Colchicine responders M6 continue Colchicine Colchicine N = 30 non-responders Azathioprine randomized 1:1 Dapsone

Weighted Z Statistic Let d 1 and d 2 be the difference in response rates between Drugs A and B in Stages 1 and 2 respectively. Let Var 1 and Var 2 be the usual pooled variances in Stage 1 and Stage 2. Z w = (wd 1 + (1-w)d 2 ) / ( w 2 Var 1 + (1−w ) 2 Var 2 ) w is the weight for each Stage. See Tamura, et al. 2016, Contemporary Clinical Trials on how to chose w). Slide 6 of 11

PCORI Improving Methods Contract U Michigan Biostatistics (Kelley Kidwell – PI) and U South Florida Rare Disease Network (me) Consolidate information from both stages of an snSMART to make overall drug recommendation Investigate incorporation of adaptive design elements into snSMART Develop user friendly app for power, sample size

Appreciation for NCSU – Statistics Department Role models for both research and application. Emphasis on experimental design / collaboration with researchers. Support, advice, and encouragement throughout my career. Best wishes to the Department for their Next 75 Years!