December 4-8, 2017 at Atlantic City, New Jersey
The 73rd Deming Conference on Applied Statistics
Sponsored by Metropolitan Section, ASQ and Biopharmaceutical Section, ASA
   
Abstract
       

Back to Program

12/6/2016 (Tuesday) 8:30 AM - 11:45 AM

Session E: Biomarker evaluation and subgroup discovery in clinical trials

Dr. Ilya Lipkovich, QuintilesIMS
Dr. Alex Dmitrienko, Mediana Inc.

Moderator: Kalyan Ghosh
 
This half-day course focuses on a broad class of statistical problems arising in the development of personalized medicine/tailored therapeutics. Vast literature has been generated in medical and statistical journals over the last 15 years concerning identification of biomarker-based patient subgroups and assessment of their validity/credibility. Principled analytical approaches have been replacing ad-hoc approaches and this half-day course will present a survey of recently developed methods aimed at identifying predictive biomarkers and subgroups with desirable characteristics (e.g., enhanced efficacy). These methods can be utilized in Phase II clinical trials to select promising biomarkers for later stages of drug development (e.g., for Phase III enrichment trials) or for rescuing failed Phase III trials.

Commonly used approaches to biomarker evaluation and subgroup identification in the context of personalized medicine will be introduced. This introductory module will be followed by a detailed discussion of the SIDES method (Subgroup Identification based on Differential Effect Search) introduced in Lipkovich et al. (2011). SIDES is based on recursive partitioning and can be used in prospective and retrospective analysis. Key elements of SIDES will be discussed, including generation of multiple promising subgroups based on different splitting criteria, complexity control to reduce the size of the search space and resampling-based approach to controlling the Type I error rate. In addition, efficient two-stage SIDEScreen procedures (Lipkovich and Dmitrienko, 2014a) that compute variable importance scores to pre-screen candidate biomarkers will be introduced.

Multiple case studies will be used to illustrate the principles and statistical methods introduced in this course, including design and analysis of Phase III trials with target subgroups and biomarker discovery in Phase III development programs (Lipkovich and Dmitrienko, 2014b; Dmitrienko et al., 2015). Software tools for implementing the biomarker and subgroup analysis methods in clinical trials will be presented, including the R package developed by the authors and Windows application.


Ilya Lipkovich, PhD, Senior Director, Quintiles, has 17 years of statistical consulting experience working in various areas including econometrics, manufacturing and quality control, and pharmaceutical industry. His research interests include clustering, predictive modeling and subgroup identification in clinical data using recursive partitioning, missing data and multiple imputation, and causal inference for observational data including marginal structural models and propensity-based estimation. He is a co-developer of novel subgroup identification methods (SIDES and SIDEScreen) and chairs the QSPI Subgroup Analysis Working Group sponsored by the Society of Clinical Trials.

Alex Dmitrienko, PhD, is Founder and President of Mediana Inc. Dr. Dmitrienko has been involved in pharmaceutical statistics for 20 years and, prior to founding Mediana, worked at Quintiles (Vice President, Innovation Unit) and at Lilly (Research Advisor, Advanced Analytics). He has been actively involved in biostatistical research and has published over 80 papers on key topics in clinical trial statistics, including multiple comparisons, subgroup analysis, adaptive designs and analysis of safety data. He has authored/edited two SAS Press books (Analysis of Clinical Trials Using SAS, Pharmaceutical Statistics Using SAS) and a Chapman and Hall/CRC Press book (Multiple Testing Problems in Pharmaceutical Statistics). Dr. Dmitrienko has served as an Associate Editor for The American Statistician, Biometrics and Statistics in Medicine, and is a Fellow of the American Statistical Association.


Back to Program