Invited Session 3: Dose Optimization

3:40-5:00 PM, November 22, 2026; Room 10A-D

Moderator
  • Tony JiangAstellas
Speakers
  • Yeonhee ParkSungkyunkwan University
  • Masahiro KojimaChuo University
  • Kanji KomatsuAstellas Pharma Inc.
  • Marc BuyseIDDI and One2treat; Hasselt University, Belgium

Tony Jiang

Moderator: Head of Statistical Science, Integrated Oncology, Astellas Pharma
Biography

Tony Jiang

Dr. Tony Jiang is Head of Statistical Science, Integrated Oncology at Astellas Pharma. In this role, he provides statistical leadership across Astellas’ oncology and hematology pipeline and leads statistical teams supporting clinical development programs. Dr. Jiang earned his PhD in Statistics from the University of Connecticut and brings extensive expertise in oncology drug development, clinical trial methodology, and Bayesian statistics.


A Bayesian Basket Design with Adaptive Patient Enrichment

Speaker: Yeonhee Park, Sungkyunkwan University, South Korea
Biography

Yeonhee Park

Dr. Yeonhee Park is an Associate Professor in the Department of Statistics at Sungkyunkwan University, South Korea, where she has been a faculty member since March 2025. Previously, she was an Assistant Professor in the Department of Biostatistics and Medical Informatics at the University of Wisconsin–Madison, where she also served as Director of the Biostatistics Shared Resource. Her research focuses on clinical trial design, precision medicine, dimension reduction, multivariate analysis, and Bayesian methods.

Abstract

Basket trials provide an efficient framework for evaluating treatments across multiple indications and dose levels, yet patient heterogeneity and limited sample sizes within indications pose challenges for dose optimization. We propose a Bayesian enrichment design that integrates adaptive patient selection with a hierarchical Bayesian model to optimize doses across multiple dose levels in basket trials. The hierarchical structure enables adaptive information borrowing across indications while preserving indication-specific decision-making, and the enrichment strategy selectively recruits treatment-sensitive patients to improve efficiency. Simulation studies demonstrate that the proposed approach outperforms fully pooled and independent designs by achieving more accurate dose selection and higher decision reliability in heterogeneous settings. These results highlight the advantages of combining enrichment strategies with hierarchical Bayesian modeling for robust and efficient dose optimization in basket trials.


A Staggered Seamless Dose-Optimization Design for Co-Developing Monotherapy and Combination Therapy

Speaker: Masahiro Kojima, Chuo University, Japan
Biography

Masahiro Kojima

Masahiro Kojima is an Associate Professor in the Department of Data Science for Business Innovation at Chuo University, Japan, and a Visiting Associate Professor at the Institute of Statistical Mathematics. He joined Kyowa Kirin in 2014 and received his Ph.D. in Statistical Science from the Graduate University for Advanced Studies (SOKENDAI) in 2022. He joined Chuo University in 2025 and now focuses primarily on his academic role, while continuing to work at Kyowa Kirin on a reduced-hours basis. His research focuses on statistical methods for clinical trials, particularly adaptive designs and dose optimization in early-phase oncology. Drawing on his experience in both industry and academia, he emphasizes developing intuitive, practical methods that can be readily implemented in clinical development.

Abstract

Evaluating an anticancer agent both alone and in combination requires efficient dose selection based on efficacy and tolerability. This talk presents a staggered, seamless phase I/II design integrating dose optimization for monotherapy and combination therapy. Monotherapy evaluation begins first, and combination subtrials are initiated before monotherapy dose exploration is completed once acceptable safety and a prespecified early activity signal are observed. Dose allocation follows a model-assisted algorithm, with backfilling at promising and tolerable regimens. Two candidate regimens are selected based on an efficacy–toxicity utility. Their phase I data are then incorporated into Bayesian futility and efficacy monitoring in phase II to support final selection of the optimal biological dose. Monte Carlo simulations evaluate the operating characteristics of the proposed design, including trial duration, dose-selection performance, and safety.


Dose Optimization in Oncology: Contributions and Limitations of Pharmacometrics

Speaker: Kanji Komatsu, Astellas Pharma Inc.
Biography

Kanji Komatsu is a Pharmacometrics Methodology Expert at Astellas Pharma Inc. He has 25 years of experience in the pharmaceutical industry, working in the areas of pharmacokinetics, clinical pharmacology, and pharmacometrics. His work focuses on methodological research and the application of quantitative approaches in clinical drug development, particularly in pharmacometrics and its interface with biostatistics. He has experience in population pharmacokinetic/pharmacodynamic modeling, exposure-response analysis, and quantitative approaches to dose optimization. His recent interests include the integration of statistical methodology into pharmacometric analyses and interdisciplinary collaboration between pharmacometricians and statisticians.

Abstract

Dose optimization in oncology requires integration of evidence on efficacy, safety, and tolerability to support the selection of a dose and schedule. Pharmacometrics (PMx) provides a quantitative framework for connecting dose, drug exposure, and response, but translating these relationships into dosing decisions requires careful consideration of both pharmacological and statistical questions.

This presentation will introduce the contributions and limitations of PMx in oncology dose optimization for an audience with diverse disciplinary backgrounds. General principles and publicly available examples will illustrate how quantitative models can integrate information and inform the evaluation of candidate dosing regimens.

Recently, an attention will be given to the interpretation of exposure–response relationships. In oncology, dose modifications may depend on evolving treatment response or toxicity, creating feedback between treatment, exposure, and subsequent outcomes. This motivates consideration of causal inference when distinguishing observed associations from the effects of changing a dosing regimen. The discussion will highlight the importance of aligning the question of interest with data, assumptions, and analytical approach.

The presentation will also consider how connections with statistical thinking and trial design can support the continued development of PMx. Collaborative research will be introduced as one example. The overall aim is to provide a shared perspective on integrating pharmacological understanding, statistical reasoning, and clinical evidence to support informed dose optimization decisions.


Use of the Net Treatment Benefit for Dose Optimization in Oncology

Speaker: Marc Buyse, IDDI and One2treat, Louvain-la-Neuve, Belgium; I-BioStat, Hasselt University, Belgium
Biography

Marc Buyse

Marc Buyse holds degrees in engineering, management, statistics and a ScD in biostatistics from the Harvard School of Public Health. He worked at the European Organization for Research and Treatment of Cancer (EORTC) in Brussels and at the Dana Farber Cancer Institute in Boston, MA, prior to founding IDDI, CluePoints and One2Treat, companies offering services and software for clinical research. He is Associate Professor of biostatistics at Hasselt University in Belgium. He has co-authored several books and ~ 500 peer-reviewed papers on statistical methods for surrogate endpoint validation, detection of errors in clinical data, analysis of multiple prioritized outcomes for patient-centric medicine. He proposed the Net Treatment Benefit as a measure of treatment effect in randomized clinical trials (see Marc E Buyse - Publications List).

Abstract

FDA’s Project Optimus emphasizes the importance of dose optimization in oncology. The Net Treatment Benefit (NTB) is a quantitative framework to assess multiple prioritized outcomes. NTB is calculated as the difference between the proportions of generalized pairwise comparisons favoring one dose regimen over the other across multiple prioritized outcomes. The advantage of generalized pairwise comparisons is that any number of outcomes of any type can be combined (times to events or rates of events, ordinal or continuous outcomes), with thresholds of clinical relevance where appropriate. In dose optimization trials, such flexibility is important as relevant outcomes may include efficacy, toxicity, patient-reported outcomes, PK/PD data and biomarkers.

NTB analyses will be illustrated using data from two randomized phase 2 dose-selection trials of experimental therapies for solid tumors. We will show that NTB offers a quantitative, transparent and readily interpretable framework for integrating efficacy, safety and tolerability in dose selection. This approach is aligned with the principles of FDA’s Project Optimus and may be prospectively applied to support benefit–risk assessment and dose optimization for novel oncology therapeutic agents.