Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

 
 
Session Overview
Session
Innovation in oncology dose escalation trials and beyond
Time:
Wednesday, 27/Aug/2025:
2:00pm - 3:30pm

Location: ETH E27

D-BSSE, ETH, 84 seats

Show help for 'Increase or decrease the abstract text size'
Presentations
40-onco-dose-escalation: 1

Declaring Doses as Safe in Ongoing Oncology Dose-Escalation Trials: Are They Truly Safe? A Critical Assessment of Safety Criteria

Stefan Englert1, Thomas J. Prior2, Anirban Mitra3, Busola Sanusi2, Liangcai Zhang2, Lilla Di Scala4

1Janssen-Cilag GmbH, a Johnson & Johnson company, Germany; 2Janssen Research & Development, LLC, a Johnson & Johnson company, USA; 3Johnson & Johnson Limited, India; 4Actelion Pharmaceuticals Ltd, a Johnson & Johnson Company, Switzerland.

Background / Introduction

Ongoing dose-escalation trials present unique challenges in assessing safety, all with the goal to establish the maximum tolerated dose (MTD) and/or the recommend phase II dose (RP2D). Due to the extended duration of trials, it is common to declare doses as safe even if the entire adaptive and iterative dose escalation process has not been completed. This practice is often applied to aid trial management decisions, such as backfilling previous cohorts, allowing intra-patient dose escalation, or initiating concurrently combination therapies, to enhance efficiency and accelerate the pace of drug development.

There exists a misperception within the research community that once an escalation assessment team has cleared a cohort for escalation, the associated dose level is inherently safe. Zhao et al. (2024) referred to this as being cleared for safety. Given that dose escalation algorithms permit repeated de-escalations, previously cleared doses may not be truly safe, necessitating a quantification of this risk.

Methods

We investigated criteria for declaring doses as safe, focusing on the BOIN design. Through rigorous simulation studies, we identified criteria that must be met to confidently declare dose levels as safe, minimizing the risk of treating additional patients at doses that exceed the MTD.

Results

Our findings suggest that the criteria proposed by Zhao et al. are overly permissive, with up to 12.4% of doses classified as safe being above the MTD as established at the trial’s conclusion. By mandating clearance of two consecutive dose levels, this percentage decreases to less than 1%. A more practical approach requires a minimum of three subjects at the next dose level, which keeps the misclassification rate below 5% while allowing, on average, 18% more doses to be deemed safe.

Conclusions

Criteria for declaring doses as safe are often neither specified nor are the implications of different safety rules examined. This may put patients at risk of receiving doses above the MTD determined only at the conclusion of the trial.

Our comparative analysis of safety criteria shows that evaluating a minimum of three subjects at the dose level above the one to be declared safe is essential for accurate safety assessments and protecting patients from toxic therapies. More aggressive dose-escalation paradigms would risk exposing patients to doses that are not ultimately declared safe.

References

Zhao Y et al., "Backfilling Patients in Phase I Dose-Escalation Trials Using Bayesian Optimal Interval Design (BOIN)," Clinical Cancer Research, 30(4), 673-79, 2024.



40-onco-dose-escalation: 2

Guiding Phase I Dose Escalation for Modern Oncology Therapies: Tackling (Informative) Dropout with Bayesian Multi-Cycle Time-to-Event Models

Lukas Andreas Widmer, Sebastian Weber

Novartis Pharma AG, Basel, Switzerland

The design of phase I trials in oncology predominantly utilizes a dose-escalation approach, monitoring dose-limiting toxicity (DLT) during the first treatment cycle to systematically determine the maximum tolerated dose (MTD). While appropriate for cytotoxic treatments, with the advent of targeted therapies, immunotherapies, and chimeric antigen receptor T-cells, this traditional model is increasingly inadequate for modern oncology treatments.

For contemporary non-cytotoxic therapies, efficacy and safety mechanisms are no longer inherently linked, rendering the "target toxicity" concept to establish a therapeutic dose as obsolete. Consequently, the MTD may not represent the optimal dose for further development, as the primary objective is to maximize efficacy while ensuring sufficient tolerability within a robust safety framework.

Furthermore, long-term administration requires sustained tolerability and efficient patient enrollment by leveraging partially-observed data. Long-term tolerability assessments and treatments for aggressively progressive cancers necessitate addressing patient dropout scenarios, which often lead to informative censoring. Aggressive cancers can cause dropout due to lack of efficacy at lower doses, while higher doses might result in dropout due to tolerability issues such as sustained lower-grade adverse events. Traditional dose-toxicity models typically overlook these dropout scenarios.

Additionally, appropriate dosing regimens are crucial to prevent serious adverse events. For instance, T-cell engaging therapies might induce cytokine release syndrome post-treatment initiation, which can be mitigated through an incremental within-patient dosing strategy.

To address these complexities, we propose a Bayesian multi-cycle time-to-event (TTE) safety model, which extends existing frameworks such as the Bayesian Logistic Regression Model (BLRM) and Escalation With Overdose Control (EWOC). This TTE model efficiently leverages trial data to predict adverse event rates at various time points across multi-cycle therapies, beyond just the initial cycle. It also facilitates rapid cohort enrollment by accommodating partially-observed data due to ongoing enrollment or dropout.

We present considerations for developing these models through concrete case studies, illustrating model structures, priors, key data scenarios, and selected operating characteristics. Particular emphasis is placed on the impact of informative dropout on model performance, accurate MTD determination, patient safety during ongoing trials, and efficiency in terms of trial duration. Our findings demonstrate that the proposed TTE model serves as a viable alternative to the BLRM, particularly under conditions of patient dropout. Last, but not least, we hint at how these TTE models can further be extended to consider between-patient heterogeneity by including partially available pharmaco-kinetics data.



40-onco-dose-escalation: 3

Potential Responses and the Order of Patient Inclusion in Early-Phase Sequential Trials.

Meliha Akouba1, Matthieu Clertant1, Alexia Iasonos2, John O'quigley3

1Université Sorbonne Paris Nord, France; 2Memorial Sloan-Kettering Cancer Center, U.S.A; 3University College London, U.K

Phase I clinical trials are the first tests of a new agent on human patients. The primary objective during this phase is to determine the maximum tolerated dose (MTD). Various statistically guided designs are used to sequentially allocate patients across dose levels. Traditionally, each dose is associated with a toxicity rate, defined by the proportion of patients experiencing dose-limiting toxicity (DLT). This conventional approach assumes a large, homogeneous patient population, summarized by an increasing vector of toxicity probability $beta$, but overlooks the heterogeneity of the actual trial population, leading to potential issues in accuracy and reproducibility.

To address this limitation, we treat the trial population as fixed, meaning we condition on the specific patients enrolled in the trial. Each patient's potential responses are represented as a vector of binary treatment effects across dose levels, assuming monotonicity assumption. The overall toxicity rate in the trial population is then summarized by an increasing vector $R$. By eliminating random variability, this approach ensures full reproducibility of results, unlike analyses based on superpopulations ($beta$). Furthermore, variance decomposition allows us to assess the impact of patient inclusion order, enabling a more precise evaluation of method performance.

Our experiments indicate that, while traditional analyses might suggest only minor differences between methods, significant quantitative differences emerge when using $R$. We show both numerically and theoretically that comparisons based on the superpopulation $beta$ introduce nuisance variability, which tends to obscure competitive differences. For example, the mean squared error (MSE) difference between two designs might appear to be 15% under Beta but rises to approximately 25% under R, highlighting more pronounced disparities.

Our methodology provides a comprehensive assessment of interval-based designs, including the Bayesian Optimal Interval (BOIN) method, which is the only Phase I design officially recognized as fit-for-purpose by the FDA and is now widely used in early-phase oncology trials. Our findings reveal that BOIN underperforms compared to dose-response model-based designs and even the historitical "3+3" design. Moreover, its dose exclusion rule—intended to prevent the reassessment of highly toxic doses—exhibits considerable variability depending on the order of patient inclusion. This instability results in the exclusion of the true MTD in approximately one in five trials with a standard sample size of 36 patients, ultimately increasing the probability of recommending suboptimal doses in terms of efficacy.



40-onco-dose-escalation: 4

How best to allocate backfill patients in dose-finding oncology trials: a methodological review & simulation studies assessing best performance

Elli Bourmpaki1, Helen Barnett2, Oliver Boix3, Hakim-Moulay Dehbi1

1Comprehensive Clinical Trials Unit, University College London, United Kingdom; 2School of Mathematical Sciences, Lancaster University, United Kingdom; 3Bayer AG, Leverkusen, Germany

Background

Dose-finding oncology trials (DFOTs) are a crucial step in research detecting the optimum dose of potentially effective anticancer therapies. Patients’ allocation in DFOTs vary depending on the trial designs used. Novel approaches have been adopted recently by including additional patients at lower doses, these patients are referred to as backfill patients. Several statistical methodologies have been developed considering different ways of allocating backfill patients, however it is not clear which allocation approaches are better than others and under which clinical scenarios. Therefore, a methodological review and further exploration on the impact of various allocation schemes for backfill patients is required.

Methods

We will conduct a literature review using MEDLINE to explore the uptake of various allocation schemes for backfill patients in published DFOTs’ results and developed methodologies between 2014 to 2024. Our aim is to assess: i) how many methods are developed for allocating backfill patients, ii) how many trials have used backfilling, iii) what allocation schemes have been used for backfill patients, iv) what trial designs were used, v) how was the recommended phase 2 dose selected using dose-finding and backfill patients’ data. All possible ways of allocating backfill patients will be reported based on the developed methods and published trial results. The performance of selected allocation schemes will be explored using a simulation study. Patient and public involvement and engagement have contributed in this research by discussing potential allocation schemes for backfill patients.

Results

For all eligible DFOTs we will report the proportion of various allocation schemes used for backfill patients, trial designs used, and developed methods for selecting the recommended phase 2 dose.

Conclusions

Our review will show how widespread is the use of backfilling in DFOTs and what approaches are currently used to allocate backfill patients. Following our review, we will list all possible allocation schemes and use simulation studies to investigate the performance of selected schemes under specific trial designs and clinical scenarios. This body of work may be used to shape future trial design, conduct and analysis guidance for DFOTs considering backfilling.



40-onco-dose-escalation: 5

Dose optimisation in early phase oncology trials - backfill and expansion cohorts

James Willard1, Thomas Jaki1,2, Burak Kürsad Günhan3, Christina Habermehl3, Anja Victor3, Pavel Mozgunov1

1MRC Biostatistics Unit, University of Cambridge, Cambridge, United Kingdom; 2Faculty of Informatics and Data Science, University of Regensburg, Regensburg, Germany; 3Merck Healthcare KGaA, Darmstadt, Germany

Historically, early phase dose finding trials in oncology focused on identifying the correct doses of cytotoxic chemotherapies, where more benefit was expected from higher doses. Therefore, these trials were traditionally designed to find a maximally tolerated dose (MTD), defined as the largest dose which satisfies specific toxicity constraints. Recently, the FDA’s Project Optimus highlighted how modern targeted therapies can provide benefit at doses lower than the MTD and so identifying these doses via dose optimisation has become a major objective of early phase trials. Unfortunately, the small sample sizes and short observation periods of these trials make dose optimisation challenging, since it is difficult to collect comprehensive information on the dose response curves under these settings. This may result in suboptimal doses being recommended for future development, adversely affecting patients and all later phase studies. To help remedy this and collect more information on the dose response curves before recommending doses for further study, the use of backfilling and expansion cohorts has been proposed. During dose escalation, backfilling cohorts are assigned to doses lower than the current estimate of the MTD. After dose escalation, expansion cohorts are assigned to a small number of the most promising doses. In this work, we examine the relationship between the escalation and expansion components of dose optimisation. We compare the performance of a variety of adaptive stopping rules which determine when to terminate escalation and transition to expansion. Furthermore, we investigate how the timing and number of patients used for backfilling may impact the selection of the doses used in expansion. Findings from an extensive simulation study will be discussed and recommendations for performing dose optimisation with backfilling and expansion cohorts will be provided.



 
Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: ISCB46
Conference Software: ConfTool Pro 2.6.154+TC
© 2001–2025 by Dr. H. Weinreich, Hamburg, Germany