\n
\n Pantelis Vlachos
\n (Cytel Inc.) \;
MCP-Mod in East®\;: \;Early development d ose-finding design and analysis
\nSelection of a dose (or doses) to carry into a confirmatory phase III study is among th e most difficult decisions in drug development. A prerequisite for informe d decision making and dose selection at the end of phase II is a solid cha racterization of the dose-response relationship(s).The MCP-Mod method comb ines principles of multiple comparisons with modelling techniques to provi de an efficient alternative to traditional dose-finding studies which are either designed and analyzed based on multiple comparisons of active doses vs placebo within an ANOVA framework\, of assume a functional relationshi p between response and dose according to a certain parametric model. We il lustrate MCP-Mod design and analysis capabilities with East®\;. \; \;
\n\;
\nBio: \ ;Pantelis is Director/Strategic Consultant for Cytel\, Inc. based in Geneva. He joined the company in January 2013. Before that\, he was a Principal Biostatistician at Merck Serono as well as a Professor of Statis tics at Carnegie Mellon University \; for 12 years. His research inter ests lie in the area of adaptive designs\, mainly from a Bayesian perspect ive\, as well as hierarchical model testing and checking although his secr et passion is Text Mining. He has served as Managing Editor of the journal &ldquo\;Bayesian Analysis&rdquo\; as well as \;editorial boards of s everal other journals and online statistical data and software archives.\n
\n
\n Neal Thomas \;
\n (Pfizer Inc.)
Understanding MCP-Mod dose finding as a method based o n linear regression
\nMCP-MOD \; is a testi ng and model selection approach utilizing contrast-based test statistics a nd p-values adjusted for multiple comparisons. The MCP-Mod procedure can b e alternatively represented as a method based on simple linear regression\ , where 'simple' refers to the inclusion of an intercept and a single pred ictor variable\, which is a transformation of dose. It is shown that the c ontrasts are equal to least squares linear regression slope estimates. The test for each contrast is the usual t-statistic for a null slope paramete r\, except that a variance estimate with fewer degrees of freedom is used in the standard error. Selecting \; the model corresponding to the mos t significant contrast p-value is equivalent to selecting the predictor va riable yielding the smallest residual sum of squares. Many of the properti es of MCP-Mod procedure can be understood and quantified using results fro m least squares linear model theory.
\nBio: \;Neal received a PhD in Statistics from the University of Chicag o. \; He is the \; leader of the Statistical Research and Innovati on center at \; Pfizer working on clinical and non-clinical applicatio ns in several therapeutic areas. Previous work experience includes sample surveys\, educational statistics (ETS)\, and health policy applications.&n bsp\; Statistical research interests include design of observational studi es\, dose response\, missing data methods\, matrix sampling\, psychometric models\, and Bayesian statistics.
\n
\n Saswati Saha
\n
(Inserm\, Aix-Marseille University)
Model based dose-finding methods in Phase II cl inical trials
\nThe primary objective of this p resentation is to discuss dose-finding methods in Phase II clinical trials that can simultaneously establish the dose-response relationship and iden tify the right dose. MCP‐Mod is one of the pioneer approaches developed wi thin the last 10 years. Though MCP-Mod is identified as an efficient stati stical methodology for model-based design and analysis of Phase II dose fi nding studies under model uncertainty\, a major disadvantage of MCP-Mod is that the parameter values of the candidate models need to be pre-specifie d a priori for the PoC testing step. This may lead to loss in power and un reliable model selection. Off late several new variations and alternatives to MCP-Mod are explored where the parameter values need not be pre-specif ied in the PoC testing step and can be estimated after the model selection step. We will briefly introduce four such state-of-art dose-finding metho ds\, show how to implement the methods in R software and present a numeric al comparison between the different new methods and the MCP-Mod approach.< /p>\n
Bio: \; Saswati completed her Ph.D as a part of ID EAS network on December 2018 from the Competence Center for Clinical Trial s (KKSB) at University of Bremen under the supervision of Professor Werner Brannath. Her primary areas of research during her PhD were dose response modelling\, multiple testing\, drug combination studies\, dose finding an d confidence interval estimation for target doses in drug development.
\nSaswati studied at the Indian Statistical Institut e\, where she completed her Bachelor&rsquo\;s degree (2011) and Master&rsq uo\;s degree (2013) in Statistics. After her masters she worked on credit risk modelling in two renowned financial institutions\, Ernst &\; Young and Genpact\, for two years and dealt with time series modelling for stre ss testing and logistic regression modelling for building scorecards.
\ n