TABA Networking Event: Thursday, October 03, 2024 18:00 EST
Agenda:
Presenter: Kevin Hou, Biostatistician, Eversana
Topic: Enhancing Reliability in Automated Record Screening: A Resampling Algorithm
Abstract: Record screening is a critical aspect of systematic review and meta-analysis, involving the meticulous task of identifying relevant paper records from a pool of candidate papers. This process is widely acknowledged for its time-consuming nature, significant costs, and susceptibility to human error. Despite the proliferation of automatic literature screening methods leveraging machine learning and AI in recent years to alleviate this burden, there remains a noticeable absence of methods that guarantee performance. In this presentation, I will introduce a flexible resampling algorithm capable of collaborating with any existing automatic literature screening methods to ensure consistent performance. I will also delve into the mathematical and probabilistic foundations of this algorithm and discuss its real-life implementation in record screening assisted by AI large language models.
Presenter: Vanessa Breton, Data Scientist, Roche
Topic: Navigating the Dose Maze: Innovative dose-finding strategies for Phase 1 Oncology Trials
Abstract: In oncology, Phase 1 clinical trials are pivotal for determining the optimal dosing of new therapeutic agents, balancing efficacy and safety. Traditional dose-finding methods, such as the 3+3 design, often fall short in addressing study complexities, including patient heterogeneity and intricate toxicity profiles. This presentation will explore and evaluate modern approaches that enhance dose-finding precision and patient safety. We will begin by reviewing the significance of dose-finding in oncology and the methods used to determine the maximum tolerated dose (MTD). Through a detailed review of a case study from patients with advanced malignancies, and including a study comparing the 3+3 design with novel methods like the Continual Reassessment Method (CRM) and the Bayesian Optimal Interval (BOIN) design, we will critically evaluate whether the MTD would have differed depending on the applied approach. Additionally, we will discuss the need for shifting from the Maximum Tolerated Dose (MTD) paradigm in light of FDA’s Project Optimus, which advocates for a more holistic approach to dose-finding. The session will conclude with an exploration of whether and how we can feasibly transition out of the traditional dose-finding maze, leveraging emerging methodologies to enhance the accuracy and efficiency of Phase 1 clinical trials in oncology.
Presenter: Mahmood Gohari, University of Waterloo
Topic: Exploring the Impacts: Statistical Methods for Evaluating Natural Experiments
Abstract: In many areas of public health, conducting randomized clinical trials, the gold standard in research, is often impractical or unethical. Instead, natural experiment designs offer valuable insights, especially in unpredictable scenarios such as the COVID-19 pandemic. This talk will explore alternative research designs and statistical methods commonly used in natural experiment settings, focusing on difference-in-difference methods and regression discontinuity analysis. We will review recent studies that examine the impact of the pandemic on youth mental health and substance use, highlighting how these methods can effectively evaluate public health interventions and policies.