Monday, Aug 4: 8:30 AM - 10:20 AM
4032
Contributed Papers
Music City Center
Room: CC-209B
Main Sponsor
Biopharmaceutical Section
Presentations
Externally controlled trials (ECTs) play a crucial role in evaluating the safety and effectiveness of medical products when randomized controlled trials are not feasible. Despite recent recommendations, there are still significant gaps in methodological consistency and regulatory evaluation criteria for ECTs. These inconsistencies lead to varied applications that impede decision-making, and the limited regulatory experience with ECTs may create hesitation among sponsors. This presentation reviews the existing guidance on ECTs, identifies methodological gaps, and proposes practical solutions for their design, analysis, conduct, and evaluation. We conducted a comprehensive search of electronic databases and grey literature up to 2025 to identify relevant guidance documents, white papers, and peer-reviewed articles. Each document was analyzed for its scope, key recommendations, and methodological considerations. Our findings emphasize best practices, identify inconsistencies, and suggest standardized evaluation criteria to ensure uniformity and reliability in assessing ECTs. These criteria will help sponsors navigate the complexities of regulatory submissions and generate evidence to inform treatment guidelines.
Keywords
external controls
clinical trial design
real-world data
target trial emulation
bias reduction
causal inference
In rare diseases limited numbers of patients present several challenges to clinical research, including statistical issues. In this presentation, we start by providing an overview of previous initiatives resulting in a range of statistical approaches to clinical research in rare diseases. These include adaptive designs and evidence synthesis, both in frequentist and Bayesian frameworks (Friede et al 2018 Orphanet). To provide perspectives on the current practice, these concepts will be introduced using clinical trial examples including paediatric settings (Gross et al 2020 Kidney International; Papchristofi et al 2024 Pharmaceutical Statistics). Particular emphasis will be given to the integration of randomized controlled trials with clinical registries, also referred to as real world data (Friede et al 2023 Prävention und Gesundheitsförderung). Finally, an outlook for future research methods in rare diseases is provided. Specifically, we will comment on the role in-silico clinical trials might play in small populations including rare diseases and paediatrics.
Keywords
Adaptive designs
Evidence synthesis
Bayesian statistics
Clinical trials
Model-based drug development
In-silico clinical trials
First Author
Tim Friede, University Medical Center Goettingen
Presenting Author
Tim Friede, University Medical Center Goettingen
Permutation tests offer a robust and non-parametric approach to hypothesis testing, particularly when traditional assumptions of normality or homoscedasticity are violated. This paper explores the application of permutation testing in vaccine research, focusing on critical measures such as immunogenicity and vaccine efficacy profiles. Immunogenicity, typically assessed through antibody titers or cellular immune responses, often exhibits skewed or non-normal distributions, making traditional parametric methods less reliable. Similarly, vaccine efficacy, expressed as the reduction in disease incidence between vaccinated and unvaccinated groups, presents challenges due to complex data structures. Through simulation studies and the analysis of real-world vaccine trial data, we demonstrate how permutation tests and studentized permutation method provide a flexible and robust alternative for comparing immune response distributions and efficacy rates across treatment groups. The study also contrasts permutation tests with traditional statistical methods, highlighting the competitive performance in various scenarios and non-standard data settings.
Keywords
permutation test
studentized permutation method
exact and asymptotic inference
First Author
Lei Li, Sanofi
Presenting Author
Lei Li, Sanofi
Parkinson's disease (PD) is characterized by long-term degeneration of neurons that leads to debilitating impairments. Assessing PD progression often focus on clinical scales, which can be limited due to variability. Integrating other measurements, such as clinical imaging, can help define alterative metrics of progression that better represent the underlying disease. Novel modeling approaches can lead to discovery of new disease states or improve the statistical efficiency of clinical trials. To approximate biological timing of PD progression, we leveraged the Parkinson's Progression Markers Initiative (PPMI) data set and applied pseudotime approaches. We derived lower-dimensional representations of the data to identify cluster centroids that serve as anchor points in the disease trajectory. We inferred pseudotime values through a curve fitting method. We found that the inferred pseudotime has a good association with the progression of calendar time, as well as existing clinical measurements. Particularly, we have identified that clinical imacharacteristics have strong correlation with pseudotime, suggesting potential of this measurement modality in defining disease progression.
Keywords
Parkinson’s Disease
Pseudotime
Disease Progression Modeling
Clinical Trials
Multi-modal data analysis
Timely characterization of the safety profile of a pharmaceutical product under development is imperative for assessing the overall benefit-risk relationship of the product and for making key development decisions. For ongoing clinical development, a robust safety monitoring and safety signal detection program which is based upon quantitative statistical reasoning is critical. Methods presented here can be applied to safety signal detection and periodic safety monitoring. Various statistical properties, distributions, and models, all utilizing a Bayesian framework are considered and further examined in order to identify robust methods applicable to a broad set of scenarios and situations. Methods developed for incidence counts (including those with under-dispersed distributions) with variable time-at-risk and with underlying constant or non-constant hazard rates, are proposed and compared to traditional methods designed to assess adverse event incidence rates or binomial incidence proportions (which assume an underlying constant hazard rate and subsequent Poisson distribution for modeling event counts).
Keywords
Clinical trials
Safety Signal detection
Under-dispersion
Right-censored Poisson Model
Log-logistic distribution
Bayesian
External control arms (ECAs) using real-world data (RWD) offer an alternative to randomized controlled trials for assessing efficacy. Common in single-arm oncology trials with binary or time-to-event outcomes, ECAs face challenges with continuous outcomes due to differences in outcome data collection timing and frequency between clinical trials and real-world settings. We propose an innovative spline-based weighted approach to estimate the treatment effect in these ECA studies, addressing biases introduced by these differences.
Our method adjusts the ECA to align with clinical trial data through propensity score-based weighting, incorporated into a Bayesian spline model to estimate treatment effects. This approach effectively handles the irregular timing of RWD outcome collection, facilitating comparisons with scheduled trial assessments. We performed simulations to compare our approach to linear interpolation and the window approach, exploring various scenarios with different outcome distributions, levels of confounding, and degrees of irregularity in data collection. Our results demonstrate the utility of this method to minimize bias in the context of an ECA study.
Keywords
External Control Arm
Continuous Outcomes
Spline Regression
Propensity score
Bayesian modeling
Real-world Evidence
The Adaptive COVID-19 Treatment Trial found that remdesivir hastened recovery times compared to placebo for adults hospitalized with COVID-19. Subgroup analysis by baseline ordinal score (OS), defined by oxygen therapy use, found that people receiving supplemental oxygen at enrollment benefitted most from remdesivir. A post-hoc analysis developed the "ACTT risk score", combining OS, ANC, ALC, and platelets. The ACTT risk score better predicts disease outcomes than OS alone, and the highest ACTT risk quartile benefitted most from remdesivir. In contrast, we find that the 4C mortality score (a broadly validated COVID-19 mortality risk score) is more prognostic than the ACTT risk score but did not identify a subgroup benefitting most. We also applied SIDES, an algorithm explicitly focused on defining subgroups benefitting most from treatment. SIDES found a larger remdesivir benefit for people enrolling within 15 days from symptom onset, consistent with other work, and for those enrolling with both low hemoglobin and high blood pressure- a novel finding. We compare subgroup identification performance of SIDES to prognostic score-based approaches in simulation.
Keywords
subgroup
SIDES
prognostic score
individualized medicine