Using joint models in phase I dose-finding designs in oncology: considerations for frequentist approaches

2025-09-12

AI Summary

Conventional phase I dose-finding trials rely on safety-based designs that struggle to incorporate multipoint predictive biomarkers for accurate binary toxicity modeling; likelihood-based continual reassessment methods (CRM) suffer from estimation instability in small samples—a long-overlooked issue. Method: We propose a joint modeling framework that simultaneously characterizes longitudinal biomarker dynamics and the dose-toxicity relationship by decomposing their joint distribution to enhance toxicity prediction accuracy. Contribution/Results: This work is the first to systematically embed dynamic biomarker information into the CRM paradigm, mitigating its small-sample bias. Simulation studies demonstrate that the proposed method significantly improves the correct identification rate of the maximum tolerated dose (MTD) and enhances robustness in dose selection. However, it also reveals an inherent limitation—persistent data sparsity in early-phase trials under the frequentist paradigm—highlighting the need for further methodological advances.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Dose-finding trials for oncology studies are traditionally designed to assess safety in the early stages of drug development. With the rise of molecularly targeted therapies and immuno-oncology compounds, biomarker-driven approaches have gained significant importance. In this paper, we propose a novel approach that incorporates multiple values of a predictive biomarker to assist in evaluating binary toxicity outcomes using the factorization of a joint model in phase I dose-finding oncology trials. The proposed joint model framework, which utilizes additional repeated biomarker values as an early predictive marker for potential toxicity, is compared to the likelihood-based continual reassessment method (CRM) using only binary toxicity data, across various dose-toxicity relationship scenarios. Our findings highlight a critical limitation of likelihood-based approaches in early-phase dose-finding studies with small sample sizes: estimation challenges that have been previously overlooked in the phase I dose-escalation setting. We explore potential remedies to address these challenges and emphasize the appropriate use of likelihood-based methods. Simulation results demonstrate that the proposed joint model framework, by integrating biomarker information, can alleviate estimation problems in the the likelihood-based continual reassessment method (CRM) and improve the proportion of correct selection. However, we highlight that the inherent data limitations in early-phase dose-finding studies remain a significant challenge that cannot fully be overcomed in the frequentist framework.

Problem

• Improving dose-toxicity assessment using predictive biomarkers
• Addressing estimation challenges in small sample size trials
• Enhancing accuracy of dose selection in oncology studies

Innovation

• Joint model integrates biomarker for toxicity prediction
• Compares to CRM using binary toxicity data
• Improves correct selection via biomarker integration