This webinar examined developments in early-phase oncology dose selection and trial design.

The US-FDA’s Project Optimus has reshaped the early phase development pathway for Oncology therapies, regardless of where the studies are conducted. The Maximum Tolerated Dose is no longer the sole basis for choosing a dose, and a randomized dose optimization phase is now required. This dose optimization phase in turn, then changes the optimal design of the escalation phase – choosing between 3+3, BOIN, CRM or mTPI is no longer critical. In contrast, details that apply similarly to all designs do become critical. This includes the use of accelerated titration, choice of target toxicity, cap on patients at each dose level, whether to re-test a dose from which you previously de-escalated, and judicious use of backfill.

During randomized dose optimization, solutions that keep this phase flexible, such as multiple interim analyses, parallel escalation or further expansion in the event of exceptional efficacy enable this phase to be used as the primary tool for dose selection.

Counter-intuitively, rather than increasing the size and complexity of oncology programs, the randomized dose-optimization phase creates opportunities for programs to have lower sample sizes, and run quicker than traditional single-dose programs, as well as opening new pathways for an early claim of efficacy. For example, randomized studies provide a valid basis for the use of Progression Free Survival and Overall Survival as endpoints in addition to Overall Response Rate. To leverage the potential of these designs, studies need to be analyzed with appropriate statistical models.

Confidence distributions provide a frequentist analogue of Bayesian posterior distributions, but without the requirement to specify a prior distribution. In randomised clinical trials, confidence distributions are particularly useful for quantifying the strength of evidence for a treatment effect, using confidence statements such as Conf(Superiority)=93.5% and Conf(Non-inferiority)=98.2%. In an industry context this can be particularly useful for guiding decisions about whether to continue product development. In this talk, Ian reviewed the application of confidence distributions to clinical trials using various case studies. Confidence distributions are useful at all stages of a clinical trial, from design to monitoring to analysis, and are also useful in adaptive designs. In some contexts, confidence distributions may provide the advantages of a Bayesian analysis but with less complexity and sensitivity to assumptions. They have recently found their way into major medical journals and are a promising new tool for clinical biostatisticians.

Recording of a 1-hour online panel discussion held on 28 April 2026 of “Careers as Biostatisticians in the Pharmaceutical Industry. The speakers walked through the different paths they have taken on their journeys as biostatisticians.

Speakers: Alan Brnabic (Lilly), Patrick Lim (Resolutum Global), William Reece (Fortrea), Richard Hutchinson (Parexel) and Motoko Yoshihara (NSLHD, NSW Health)

Moderator: Belinda Butcher (WriteSource Medical)

Recognising a need to build better connections between early-career and more established statisticians in Australia, Karen Lamb created a mentoring program through the Biostatistics & Bioinformatics Section of the Statistical Society of Australia (SSA) in 2020. The program involved 10 mentor-mentee pairs from around Australia and ran for 12 months. The program has since expanded broadly across the SSA and is run by the Mentoring Program Committee, co-chaired by Karen and Lizzie Korevaar. The broader program is now in its 4th year, pairing between 20 and 30 mentees and mentors each year. In this talk, Lizzie will provide an overview of the program and its benefits to participants.

Recognising a need to build better connections between early-career and more established statisticians in Australia, Karen Lamb created a mentoring program through the Biostatistics & Bioinformatics Section of the Statistical Society of Australia (SSA) in 2020. The program involved 10 mentor-mentee pairs from around Australia and ran for 12 months. The program has since expanded broadly across the SSA and is run by the Mentoring Program Committee, co-chaired by Karen and Lizzie Korevaar. The broader program is now in its 4th year, pairing between 20 and 30 mentees and mentors each year. In this talk, Lizzie will provide an overview of the program and its benefits to participants.

Pharmacokinetics (PK) plus pharmacodynamics (PD) equal to pharmacology, which plays an essential role in drug development. PK is the study of how our body treats the drug and PD is the study of how the drug treats our body. PK exposure is a key factor connecting dose to PD, efficacy and safety of the studied drug in question. PK/PD modelling and simulation has become a powerful approach to support drug development and regulatory review, which is often referred as Model-informed drug development (MIDD). This presentation will provide an introduction about the concepts of PK, PD, and seven case studies about how PK/PD modelling and simulation can support drug development.

The mission of R&D-oriented pharmaceutical companies is to contribute to the improvement of health and welfare of people around the world through continuous R&D and stable supply of innovative pharmaceutical products.

One way to efficiently meet society’s demands is to shorten development time. Adaptive design, Bayesian statistics, and the use of external controls are some examples. Some trials that are considered complex innovative design are presented as examples of urgent clinical trials. In trials such as these, it is sometimes difficult to analytically assure control of Type I errors and can only be assessed by simulation. It is desirable to examine the operating characteristics of the estimated amount of bias before starting the test. The following is an organized discussion of precautions to be taken when simulation is used to demonstrate operating characteristics. The Potvin’s method is known as a method of sample size re-estimation in BE testing, but it does not always control for type 1 errors. There is some controversy about the estimand of BE tests, so we will introduce that point as well.

The Japan Pharmaceutical Manufacturers Association and how-to step-up ICH topics will be introduced.

You will be guided through a case study (CheckMate 037) and interactively deepen the knowledge to gain hands-on experience in developing estimands. You will be introduced to the background and disease context of the case study, the concept of an estimand, and important clinical events (increased dropout rate and crossover therapy) that occurred in the trial and affected interpretation of the results. Using non-technical language and clear graphical presentation of the concepts, you will experience how the estimand framework provides a common language to describe the diversity of patient journeys and why it is important to address the right question in clinical trials. Target audience (no prior knowledge in oncology or of estimands required): Clinicians, Investigators, Regulatory Experts, Medical Writers, Ethics Committees, Statisticians.

The targeted learning outcomes are as follows: – Recognize the benefits of following the estimand framework (ICH E9 (R1) addendum) in the context of a clinical trial, in order to: – have a common language to describe the diversity of patient journeys – address the right question in clinical trials – Be able to construct an estimand, including identification of relevant intercurrent events and application of relevant strategies to address them – Gain insights from a cross-industry international working group on estimands in oncology Presented by members of the Estimands in Oncology special interest group www.oncoestimand.org