Arthur Hughes - Meta-analysis of High-Dimensional Trial-Level Surrogates of Vaccination

by Arthur Hughes (Université de Bordeaux, Inserm Bordeaux Population Health U1219, BIOSTAT)

Thursday 19 Mar 2026, 13:00 → 14:00 Europe/Paris

amphi Louis (ISPED)

Speaker: Arthur Hughes from BPH
Title: Meta-analysis of High-Dimensional Trial-Level Surrogates of Vaccination

Abstract: Measuring clinically relevant primary endpoints in clinical trials can sometimes be costly, time-consuming, unethical, or otherwise impractical. Surrogate endpoints are markers that capture or predict the treatment effect on a clinical endpoint. When the estimand of interest is the average treatment effect in a population, such markers are referred to as trial-level surrogates, predicting the average treatment effect on the primary outcome using only the surrogate. High-dimensional molecular data, such as early post-vaccination transcriptomic profiles, are promising candidates for trial-level surrogate markers in vaccine trials. We recently proposed RISE, a methodology for identifying and evaluating surrogate markers in high-dimensional settings that accounts for the complexity of these data, but this method operates within the single-trial framework relying on assumptions regarding the external generalisability of the derived surrogate.
Here, we introduce RISE-meta, an extension to the meta-analytic evaluation of high-dimensional surrogate markers across clinical trials, leveraging the heterogeneity in treatment effects across multiple trials to assess the transportability of candidate surrogates. It has 3 stages: i) RISE algorithm is applied within each study to obtain study-specific measures of surrogacy for each marker; ii) a random-effects inverse-variance meta-analysis estimates pooled effects across studies for testing each marker surrogacy strength iii) significant markers after multiplicity correction are combined into a composite signature using weights reflecting both their surrogacy strength and variability, and the pooled surrogate effect is assessed across trials. 
We applied RISE-meta to five inactivated influenza vaccine trials with both antibody and transcriptomic measurements collected pre- and post-vaccination. The analysis identified a day-1 post-vaccination transcriptomic signature comprising innate immune-related genes that strongly predicted the treatment effect on subsequent antibody responses across trials.

 

This seminar will be in English

 

Calendar subscription link for the complete seminar series:
https://indico.math.cnrs.fr/category/711/events.ics

Program of the Biostatistics seminars:
https://indico.math.cnrs.fr/category/711/

Subscribe to the seminar mailing list:
https://diff.u-bordeaux.fr/sympa/subscribe/seminaire.biostat.bph

 

Biostatistics seminar series from the Department of Public Health from the University of Bordeaux and the Bordeaux Population Health UMR 1219 research center