Chair: Wenqiong Xue, PhD (Boehringer Ingelheim Pharmaceuticals, Inc.)
Co-Chair: Dooti Roy, PhD (Boehringer Ingelheim Pharmaceuticals, Inc.)
Abstract: Design of CNS clinical trials has been facing various challenges due to many factors, including the heterogeneity of the diseases, the uncertainty of the outcome assessments, high placebo response, variability of the patient population, etc. Those challenges often lead to inconclusive or interpretable study results. Thus, there is a need to develop and implement innovative statistical methodologies and study designs to enhance the efficiency as well as the robustness of CNS studies. In this session, we will hear from both industry and regulatory agency regarding the challenges in designing CNS trials, but more importantly the opportunities to leverage statistical methods addressing and mitigating the risks in CNS studies. We will discuss methods including Bayesian borrowing of external data in the pediatric setting, adaptive dose-finding design which enables early signal detection, and also the use of AI/ML in predicting placebo response and beyond in this field.
Speaker: Mandy Jin, PhD (AbbVie)
Title: Enhancing Machine Learning Methods to Predict and Control Confounding Placebo Effect in Randomized Clinical Trials
Abstract: In randomized clinical trials (RCTs) with subjective outcomes, the placebo effect is one of the major challenges in evaluating possible mechanisms related to the true therapeutic effect and is highly associated with the failure of many RCTs for many clinical trials. To address the issue, we propose a few innovative strategies to predict and control placebo effects and machine leaning (ML) predictive models have been enhanced to improve the prediction. These strategies are comprehensively evaluated by simulation to assess bias, mean squared error (MSE), power and Type I error rate, and applied in a real clinical study. These findings suggest that incorporating machine learning and weighting techniques can enhance efficiency to control placebo effect.
Speaker: Shyla Jagannatha, PhD (Johnson & Johnson Innovative Medicine)
Title: Bayesian Borrowing in a Phase 3 Pediatric Major Depressive Disorder Trial with Propensity Score Weighting
Abstract: One of the main barriers in pediatric or adolescent mental health clinical trials is poor recruitment rate. One of the available strategies to mitigate this is to consider Bayesian borrowing to possibly reduce sample size and enable completion of the program requirements in a timely manner. A case study of a Phase 3 pediatric clinical trial design where borrowing was conducted on treatment effect from a previously conducted phase 2 trial in a similar population. A robust mixture prior using weighted combination of an informative prior and an uninformative prior was used. To account for potential baseline characteristic imbalances between the phase2 study used for borrowing and the phase3 study, the informative prior was constructed by first weighting the phase2 treatment effect using propensity scores. Simulations were performed to evaluate the operating characteristics of the Bayesian borrowing design. Healthy authority feedback for the proposed design will also be mentioned.
Speaker: Shuli Li, PhD (Takeda)
Title: A Dose-Finding Adaptive Design with Flexibility of Early Efficacy Evaluation
Abstract: We present an adaptive, double-blinded, placebo-controlled study design within the dose-finding framework. The design also offers the flexibility to incorporate early assessments of efficacy and safety, guiding future patient allocation and enabling faster overall treatment evaluation. The availability and relevance of short-term outcome data in this particular setting for one endpoint, significantly empowered the integration of the adaptive component. Further, the design is particularly well-suited to our setting, where extensive data from earlier drug development is already available, which allows us to adopt a slightly more aggressive approach, prioritizing efficiency while maintaining flexibility, and our simulation also demonstrates improved design performance by leveraging data from the placebo arms of earlier internal trials. We also consider alternative design options within this context and simulations are used to evaluate the design properties. We outline their respective strengths and limitations and provide practical guidance for implementation in different scenarios.
Speaker: Xiang Ling, PhD (FDA)
Title: Challenges in CNS Clinical Trials and Potential Innovative Applications of AI/ML
Abstract: Central Nervous System (CNS) clinical trials face numerous challenges, including limited sample sizes, difficulties in patient enrollment, and complexities in endpoint selection, particularly for rare or heterogeneous CNS disorders. These challenges highlight the need for innovative and robust statistical and methodological approaches to enhance trial efficiency and interpretability. This presentation will explore examples such as using prognostic covariate adjustment (PROCOVA) to improve precision and study power, developing master protocol platform trials to optimize efficiency, and employing Bayesian borrowing to incorporate external data. Additional strategies include integrating mortality into the primary analysis of functional outcomes, developing composite endpoints and refining endpoint selection. The discussion will also emphasize the potential of AI/ML applications to enhance trial design and analysis, ultimately driving progress in CNS drug development.
