Statistical Innovation in Pharma
Title: A two-stage dose optimization framework for oncology drug development via an integrated benefit-risk assessment.
Friday, Aug 21st, 2026
Speaker: Shubhadeep Chakraborty , Senior Manager Biostatistics
Bristol Myers Squibb
Shubhadeep is currently working as a Senior Manager within Oncology Biostatistics at Bristol Myers Squibb (BMS). He obtained his PhD in Statistics from Texas A&M University in 2020 and worked as a postdoctoral scholar at University of Washington, Seattle prior to joining BMS. There he contributed to innovating statistical methodology to support drug development as part of the Statistical Methodology team, before transitioning to Oncology.
Abstract: The traditional oncology dose selection strategy of targeting the maximum tolerated dose (MTD) has proven inadequate for targeted and immunotherapies, as doses lower than MTD may provide similar efficacy benefit with reduced toxicity. This has necessitated a shift towards targeting the optimal biological dose (OBD) instead, where the drug achieves maximum efficacy with minimum toxicity. We propose a two-stage dose optimization framework that can accommodate multiple efficacy, safety, and tolerability endpoints of interest, and a hierarchy of the relative clinical importances of the risk factors. In Stage 1, we eliminate inadmissible doses based on pre-defined probability cut-offs for each of the endpoints of interest. The dose admissibility criteria ensure that the risks are controlled below a minimum acceptable threshold for all the endpoints. In Stage 2, we introduce a benefit-risk score based on the estimated probabilities (making it independent of original scale) from Stage 1 for each of the admissible doses, and select the dose with the highest benefit-risk score as the OBD. Extensive simulations demonstrate the superiority of the proposed method over existing methods in identifying the OBD. We use BMS clinical trial data to illustrate the application of the method.
Title: SIMONE - An Integrated Bayesian Simulation Framework for Clinical Trial Designs with AI-Orchestrated Workflow
Friday, Aug 21st, 2026
Speaker: Bingying Dai , Principal Biostatistician
Regeneron
Dr. Bingying Dai is a Principal Biostatistician in the Quantitative Innovation and Statistical Strategy (QISS) organization at Regeneron. She received her PhD in Statistics from Colorado State University in 2025. Her work focuses on innovative study designs and framework development for clinical trials, with a particular interest in leveraging AI and machine learning to enhance efficiency in statistical practice and trial design.
Abstract: The FDA published its draft guidance for industry on applying Bayesian designs in clinical trials this year. Although the field is still digesting the document, it is clear that Bayesian designs can be critically valuable in specific scenarios. The need for practical, accessible simulation tools therefore has become increasingly pressing. We present SIMONE, a YAML-driven simulation framework that supports Bayesian group sequential and borrowing designs under a single platform allowing for continuous and binary outcomes, recurrent event and survival analysis. The framework features a reproducible pipeline accessible through both an interactive R shiny app and a conversational AI assistant that orchestrates the entire workflow from design specification through simulation execution to report generation. SIMONE makes rigorous Bayesian trial planning accessible across quantitative disciplines in drug development.
Title: The Evolving Statistician: Bridging Statistical Modeling and AI/ML to Transform Clinical Developmen
Friday, Aug 21st, 2026
Speaker: Zhaoling Meng , Associate Vice President
Sanofi
Dr. Zhaoling Meng is Associate Vice President, Global Head of Clinical Statistical Modeling in Evidence Generation and Decision Science, sanofi R&D. Zhaoling has 25+ years of experience in Pharma. She built and leads a global team to support and promote modeling & simulation (M&S) and quantitative decision making across drug development stages and various disease areas. Her research interests include development scenario and design decision via competitive landscape modeling, real world data use in informing clinical studies, clinical prediction modeling to trial optimization, and AI/ML method in clinical applications. Prior to sanofi, Dr. Meng also worked at Bill and Melinda Gates Medical Research Institute, Merck, Pfizer and GSK.
Abstract: As clinical development becomes increasingly complex, the role of the statistician is shifting from traditional hypothesis testing toward data-integrated predictive modeling. While modeling and simulation are already vital for accelerating drug development, the integration of AI/ML offers a robust framework for evidence-based decision-making. By synthesizing diverse data streams—including biological insights, disease physiology, and real-world evidence—statisticians can move beyond rigid assumptions to generate dynamic insights. This session examines the transformative impact of these methodologies through case studies on trial planning, enrollment optimization, and quantitative decision frameworks. Ultimately, we demonstrate how embracing statistical modeling and AI/ML enables a more efficient, adaptive, and successful path to bringing new therapies to patients.
Title: A seamless two-stage phase I/II trial design with backfill and joint monitoring for dose optimization
Friday, Aug 21st, 2026
Speaker:Kentaro Takeda , Senior Director
Astellas
Dr. Kentaro Takeda is a Senior Director, Head of innovative statistics at Astellas Pharma Global Development, Inc. His research interest covers Bayesian clinical trial designs, including dose-finding trials, basket trials, using real-world data, and oncology statistics. Dr. Takeda has published more than 60 peer-reviewed papers in the research area. Dr. Takeda is leading an innovative statistics group in Astellas and has implemented novel statistical approaches, including his proposed methods, in new drug developments. Dr. Takeda is also an associate editor of several academic journals
Abstract: The FDA's Project Optimus initiative emphasizes dose optimization through randomized cohorts and comprehensive evaluation across dose levels. Additionally, early-phase oncology trials must efficiently evaluate antitumor activity while maintaining patient safety, requiring robust statistical frameworks for futility monitoring. We propose a seamless two-stage Phase I/II trial design integrating dose optimization with efficacy evaluation. Stage 1 employs a dose-finding approach with patient backfilling, utilizing Bayesian optimal boundaries for efficacy and toxicity to select two promising doses for further evaluation. The backfill strategy enables sequential enrollment while previous patients continue their evaluation periods, thereby accelerating the trial. Stage 2 simultaneously identifies the optimal dose and evaluates treatment effectiveness through joint monitoring of efficacy and toxicity outcomes. Stage 2 incorporates Bayesian optimal boundaries for both futility and efficacy stopping, enabling early decision-making while explicitly controlling Type I error rates. Simulation studies across realistic scenarios demonstrate superior operating characteristics of the proposed design compared to existing designs, making this approach particularly valuable for modern oncology drug development where efficiency, accuracy, and patient safety are paramount.