Speaker

Biography

Abstract 

Seth Seegobin
Executive Director of Biostatistics and Head of the Statistics Department of Vaccine and Immune Therapies, AstraZeneca.

Seth Seegobin, PhD, is currently an Executive Director of Biostatistics and Head of the Statistics Department of Vaccine and Immune Therapies at AstraZeneca. His team, which is spread across various AstraZeneca sites globally, provides statistical leadership and consultation to the therapeutic area global head of Biometrics, various quantitative groups within AstraZeneca, and senior management in discovery, clinical development, and commercial functions. With over 18 years of industry experience, he has designed and analysed clinical trials in a variety of therapeutic areas.

A Threshold of Protection Model for Monoclonal Antibody Efficacy Against SARS-CoV-2 Variants
Clinical development of monoclonal antibodies (mAbs) against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is challenging due to rapid changes in the variant landscape. Using efficacy data from the phase 3 PROVENT pre-exposure prophylaxis trial of tixagevimab–cilgavimab (NCT04625725), individual nAb titres were predicted by dividing serum mAb concentration by prevalence-adjusted tixagevimab–cilgavimab potency (from in vitro IC50 values combined with viral surveillance data) and related to efficacy with a Cox model. The Threshold of Protection (ToP) Cox model was externally validated using data from the phase 3 SUPERNOVA trial (NCT05648110), which assessed sipavibart efficacy against symptomatic COVID-19 in immunocompromised participants. This novel approach integrates predicted nAb titres against multiple SARS-CoV-2 variants into a ToP Cox model that can be applied across different variants and could serve as a surrogate endpoint in immunobridging studies to expedite clinical evaluation and regulatory approval for mAbs targeting SARS-CoV-2.

Tom White
Head of Clinical Data Science, Vaccine and Immune Therapies, AstraZeneca.

Tom White is the Head of Clinical Data Science for the Vaccine and Immune Therapies division at AstraZeneca. In this role, he leads initiatives in data science and artificial intelligence to support the development and evaluation of vaccines and immunotherapies. He has been involved in several significant studies related to COVID-19 vaccines, including the AZD1222 (ChAdOx1 nCoV-19) vaccine. His work has contributed to understanding the durability of protection and immunogenicity of the vaccine over time. Additionally, he has co-authored research on estimating vaccine efficacy during open-label follow-up of COVID-19 vaccine trials, utilizing population-level surveillance data. Dr. White's expertise lies at the intersection of clinical data science and immunotherapy development.

Winter is coming; forecasting the next season’s wave in advance.
The incidence of many infectious diseases varies seasonally, which complicates recruitment planning and study monitoring activities. An accurate, time-resolved forecast of incidence which reacts to changes in the present season would be highly valuable.

We will describe a methodology for monitoring publicly available global surveillance data to forecast incidence in target countries, and show how that forecast can be integrated with live study data to continually refine expectations of events in the study. Our approach employs the discovery of leading indicators, building time series regression models, simulation-based modelling, and the automation of this cycle using cloud infrastructure.

A real example will be shown in which we accurately forecasted the RSV season in the United States six months in advance, primarily using leading indicators identified in South America. 

Scott Patterson
Senior expert statistical scientist at Sanofi Vaccines, Sanofi.

Dr. Scott Patterson PStat® earned his PhD in Statistics from the University College of London and has over 30 years experience as a professional statistician.  Currently, he is the senior expert statistical scientist at Sanofi Vaccines.  Scott has a proven record of successfully developing, understanding, applying, and communicating complex statistical and mathematical concepts and algorithms for multi-national biopharmaceutical corporations.  He is an expert in mixed-effects modelling and Bayesian statistics with extensive specialization and concentration in Vaccines and Biologics following on from Anti-infective, Cardiovascular, and Metabolic Clinical Development in addition to Clinical Pharmacology.  He also has a proven track record of successful collaboration with international colleagues and has authored two editions of his book with Professor Byron Jones and numerous chapters, papers, and presentations with current h-index of 30.  He has spoken and taught at venues ranging from universities to professional meetings to high schools to elementary schools.  Scott and his family reside in Pennsylvania.

Bayesian Statistics for Age Expansion in Vaccines with Example 
Noninferiority (NI) designs are often used in vaccine clinical development to demonstrate that a new vaccine is similar in terms of immunogenicity to a registered vaccine. To demonstrate NI in a pediatric population based on traditional frequentist NI analysis usually requires large sample sizes, raising ethical and feasibility issues. Bayesian approaches bring an opportunity to increase the success and efficiency of pediatric vaccine development.

The methods are illustrated with a vaccine NI trial in pediatrics as a case study with two independent age cohorts (denoted here groups 1 and 2), where the objective is to demonstrate NI of the new vaccine against the control, for each age cohort separately. Enrollment of the 2 age groups is performed within the same study, same countries and sites, making borrowing ‘internally’ within the same study a unique feature.  The focus is on the borrowing of data from the age group 1 as prior for the age group 2.

Two Bayesian priors are explored for information borrowing across age groups. Simulations are performed for each borrowing approach under the same set of clinical scenarios to assess the operating characteristics of each approach. The pre-specified borrowing weight was then determined based on the simulations for the planned statistical analysis.

With a Bayesian design in this example, the sample size could be reduced by 15%. Type I error was better controlled with a robust mixture prior than the power prior approach.  The use of Bayesian approach enhanced trial efficiency by exposing less numbers of children and potentially allowed faster access for children.

Margherita Annaratone,
Principal Statistician GSK Vaccines.

Margherita Annaratone holds a Master's degree in Engineering Mathematics from Politecnico of Turin. Her professional career began with an internship at BioMarin in London. Here, she concentrated on propensity score methods to utilize historical data as control arms in clinical studies. Following her internship, she has been working in GSK Vaccines in Siena, Italy as study lead statistician since 2019.

Robust (exchangeability) Bayesian methods for vaccines targeting multiple serotypes
Combination vaccines have been extensively used in recent decades (such as pneumococcal vaccination, human papillomavirus, rotavirus, influenza, meningococcus, poliomyelitis…) as they offer protection against multiple diseases or several subtypes of bacteria or viruses and help simplify the current immunization schedule. These vaccines are often extended by including additional components. To demonstrate the efficacy of these improved vaccines, regulatory agencies often require immunological studies showing superiority of the new components and non-inferiority of the common components. The sample size of these trials is usually large because it depends on the number of coprimary outcomes. With an application to a case study, we consider some Robust (exchangeability) Bayesian methods and compare them to the frequentist approach - considered as benchmark - in order to evaluate the potential increase in power of these studies

Prof. Xinxue Liu,
Associate Professor of Epidemiology and Health Statistics at the University of Oxford.

Xinxue Liu, Associate Professor of Epidemiology and Health Statistics at the University of Oxford, works at the Oxford Vaccine Group, Department of Paediatrics, and serves as the group lead of statistics and epidemiology. During the COVID-19 pandemic, he designed and led several large clinical trials, providing critical data and scientific evidence to inform COVID-19 vaccination policies in the UK and worldwide. His research findings have been cited in the vaccination guidelines of the World Health Organisation, the European Medicines Agency, and the UK's Medicines and Healthcare Products Regulatory Agency.

As an investigator in the Typhoid Vaccine Acceleration Consortium (TyVAC), Dr. Liu led the design of a cluster-randomised controlled trial in Bangladesh in 2017 to evaluate the effectiveness of a single-dose typhoid conjugate vaccine (TCV). This trial, which will continue until 2027, aimed to assess short- and long-term vaccine protection by TCV. Data from this study have played a vital role in informing WHO policy decisions regarding TCV scheduling. Dr. Liu currently serves as a member of the WHO Strategic Advisory Group of Experts (SAGE) typhoid Working Group and the WHO Technical Advisory Group on Salmonella Vaccines.

In addition to his applied research, Dr. Liu has a strong interest in methodological advancements, particularly in enhancing the efficiency of controlled human infection models (CHIMs) for vaccine development. His primary focus is on utilising Bayesian design in vaccine CHIMs, incorporating historical data to reduce sample size.

The validity of test-negative design for assessment of vaccine protection
The test-negative design (TND) is a useful tool to evaluate vaccine protection following deployment. We compared different observational and experimental study designs for assessing VE by re-analysing data from the TyVAC Bangladesh trial, a participant- and observer-blinded cluster randomised controlled trial (CRCT). We compared estimates of VE from the CRCT analysis, which assessed the risk of blood-culture-confirmed typhoid fever among TCV recipients compared with JE recipients, to estimates from the cohort and TND analyses, which compared TCV recipients and non-vaccinees in the TCV clusters. We further conducted a negative-control exposure (NCE) and a negative-control outcome (NCO) analyses as bias indicator.  The VE estimates were 89% (95% CI: 81-93) in the CRCT analysis, 79% (95% CI: 70-86) by the cohort design, and 88% (95% CI: 79-93) and 90% (95% CI: 75-96) by the TND with two definitions of test-negative controls. Using NCE analysis, JE vaccination was associated with an increased risk of typhoid fever in the cohort design (IRR: 1.98, 95% CI: 1.56-2.52), but no significant association was found in the TND. Similarly, an increased risk of non-typhoid infections was observed in the cohort NCO analyses when comparing vaccinees with non-vaccinees in both JE and TCV clusters, but not in the TND NCO analyses.  The TND provides reliable estimates of TCV VE, while the cohort design can bias the VE estimates likely due to unmeasured confounding effects, such as healthcare-seeking behaviours. NCE and NCO approaches are useful tools for identifying such biases.  

Irene Garcia-Fogeda, Steven Abrams, Stijn Vanhee, Maha Salloum, Benson Ogunjimi, Niel Hens - University of Antwerp, Belgium
Statistical Researcher

Irene García-Fogeda, born in Madrid, Spain, is a 30-year-old researcher specializing in biostatistics and within-host mathematical modeling of infectious diseases. Irene completed her undergraduate studies in Applied Statistics in Madrid and pursued a Master's in Biostatistics with a focus on epidemiology at UHasselt, Belgium. Currently based in Antwerp, she has been conducting her PhD research for the past four years, exploring within-host mathematical models applied to various infectious diseases. Her work primarily focuses on antibody kinetics, their role in protecting against infections, and their dynamics following vaccination or infection.
Outside of her academic pursuits, Irene is passionate about traveling and experiencing new cultures. A lifelong tennis enthusiast and finds joy out of sports.

Within-host mathematical models to study antibody kinetics after the prophylactic Ebola vaccine in the Democratic Republic of the Congo
Ebola virus disease remains a threat in the Democratic Republic of Congo, where outbreaks persist due to animal reservoirs. While vaccines like Ad26.ZEBOV and MVA-BN-Filo are safe and immunogenic, the mechanisms driving antibody responses after the two-dose regimen and a booster dose are not fully understood. Within-host mathematical models offer valuable insights into disease dynamics and waning immunity processes, but data-driven mechanistic models of antibody kinetics remain scarce.
The present study seeks to elucidate the mechanisms involved in the antibody production after the two-dose vaccine regimen with and Ad26.ZEBOV and MVA-BN-Filo vaccines, followed by a booster dose with Ad26.ZEBOV, addressing challenges in inferring and implementing within-host approaches.
By drawing on existing theoretical frameworks and referencing recent empirical findings on Ebola vaccines, we demonstrate how mechanistic modelling can extend and strengthen our understanding of antibody dynamics. Particular emphasis was placed on identifying the distinction in antibody half-lives, and the understanding of the role of vaccine antigens in generating an immunological signal through germinal centers. Steps were taken to ensure a feasible and interpretable model.
Antibody half-lives were longer after the booster compared to the second dose, indicating steadier decay. Neutralizing antibodies showed a significant role in antigen binding during the vaccination schedule.
This study has raised important questions about steps to implement within-host mechanistic models and robust data to inform estimation of model parameters. Further research is required to understand the decay of memory B cells and long-lived plasma cells responses, as they are key processes in sustaining antibody-mediated immunity.

François Beckers, Global Head of Vaccines Biostatistical Sciences, Sanofi.

François Beckers is an engineer in agronomy by background and obtained a PhD in biometry from the université catholique de Louvain in 1997 while working as lecturer at university. After working few years in a biometrics CRO, he joined GSK Vaccines in 2000 where he was heading the Global Data Management and Biostatics group. Between late 2013 and mid 2022, he was the Global Head of Biostatistics, Epidemiology and Medical Writing at Merck KGaA, Germany with main focus on Oncology, Neurology, Fertility and General Medicine. Since then, he joined Sanofi as Global Head of Vaccines Biostatistical Sciences. Since 2008, he is an invited Professor at the katholieke universiteit leuven.

Adopting Estimand in Prophylactic Vaccine Trials
While the estimand framework has been quiet broadly discussed and implemented in various therapeutical areas, limited literature and consensus has been reached on the application of the framework in prophylactic vaccine trials. A vaccine estimand implementation group has brought together people from industry, academia and regulatory into a joint effort and provided a comprehensive review of the application of the framework, acknowledging the many unique characteristics of preventive vaccine development, including emphasis on estimating the “biological” vaccine effect as target. Beside discussing various considerations in identifying relevant intercurrent events amongst the other estimand attributes, it discusses various considerations in choosing strategies to handle them in terms of their utility in addressing the scientific questions. 

Prof. Stijn Vansteelandt
University of Gent, Belgium.

Assumption-Lean Strategies for Handling Intercurrent Events in Randomized Trials
Intercurrent events—such as treatment noncompliance, rescue medication, or truncation by death—pose major challenges in the analysis of randomized trials. The ICH E9 (R1) addendum has motivated methodological work aimed at defining and estimating treatment effect estimands that capture a treatment’s net effect despite the presence of such events. Standard causal inference approaches often focus on hypothetical estimands that remove intercurrent events or on principal stratum estimands that condition on unobservables, both of which rely heavily on untestable assumptions such as unconfoundedness. Composite endpoints offer more assumption-lean alternatives but often yield effects that are difficult to interpret clinically.

In this talk, I will survey recent work on assumption-lean strategies for handling intercurrent events, each leveraging repeated measures in trial designs to sidestep unconfoundedness assumptions. I will first describe new instrumental variable approaches to handle noncompliance and estimate the effect of actually taking semaglutide versus placebo on body weight over 68 weeks in the STEP 1 trial, as well as similar evaluations in the REDEFINE study. I will then present a novel approach for testing whether a treatment has an effect on outcomes that may be truncated by death, by targeting a more modest estimand, identification of which avoids the strong assumptions of hypothetical or principal stratum strategies.

Prof. Geert Molenberghs
I-BioStat Universiteit Hasselt & KU Leuven, Belgium

Geert Molenberghs was born in Antwerp, Belgium, on February 5, 1965. He is Professor of Biostatistics at the Universiteit Hasselt and KU Leuven in Belgium. He received the B.S. degree in mathematics (1988) and a Ph.D. in biostatistics (1993) from the Universiteit Antwerpen. Dr Molenberghs published methodological work on surrogate markers in clinical trials, categorical data, longitudinal data analysis, and on the analysis of non-response in clinical and epidemiological studies. He served as Joint Editor for  Applied Statistics, Co-editor for Biometrics, Co-editor for Biostatistics, Series Editor of Wiley Probability & Statistics, and Wiley StatsRef. He is currently Executive Editor of Biometrics. He acted and acts as Associate Editor for several journals and undertook numerous refereeing tasks (for journals, faculty member promotion, faculty member appointments, etc.). He was President of the International Biometric Society. He was elected Fellow of the American Statistical Association and received the Guy Medal in Bronze from the Royal Statistical Society.  He has held visiting positions at the Harvard School of Public Health (Boston, MA). He is founding director of the Center for Statistics at Hasselt University and currently the director of the Interuniversity Institute for Biostatistics and statistical Bioinformatics, I-BioStat, a joint initiative of the Hasselt and Leuven universities. He published, as editor and author of several books on longitudinal data analysis, possibly subject to missingness (with Geert Verbeke) and surrogate endpoints. He has (co-)taught nearly 200 short courses on the topic in universities as well as industry, in Europe, North America, Latin America, and Australia. He received research funding from FWO, IWT, the EU (FP7), U.S. NIH, U.S. NSF, UHasselt, and KU Leuven. He is member of the Belgian Royal Academy of Medicine. Since the beginning of the SARS-CoV-2 induced pandemic, he has served as an advisor to the Belgian government and has been a member of several official scientific boards in his home country. He has also taken up roles in science communication to the general public in the context of the pandemic. He is currently member of the Risk Assessment Group of the Belgian Federal Government, and of the Advisory Committee for Public Health Emergencies of the EU. He is involved in several Population Health Management Projects, under the auspices of the Faculties of Medicine of the Leuven and Hasselt Universities.

Paradigms for the evaluation of surrogate endpoints and correlates-of-protection: vaccines, the pandemic preparedness, and beyond. 
Biostatistics has played a major role during the COVID-19 pandemic, along with a large array of clinical disciplines and the humanities. From a scientific perspective, this included, especially at the earlier stages, epidemiological and mathematical modeling work regarding the virus and its characteristics, and the non-pharmaceutical interventions (NPIs) taken in response.

Resulting from an unparalleled concerted effort, in less than one year several highly efficacious and effective vaccines were approved by the various regulators. This is especially remarkable given that SARS-CoV-2, especially variants such as Alpha and Delta, had extremely high basic reproduction numbers. Meticulous post-marketing information enabled the identification very rare but serious side effects, such as the thrombosis with thrombocytopenia syndrome (adeno-26 vector vaccines) and myocarditis (mRNA vaccines) See Kurz et al. (2022) and Dorta et al. (2025). The way in which sponsors, regulators, and academic scientists collaborated was exemplary and provides a model also for non-emergency settings, for rare and common diseases, for communicable and non-communicable diseases. This was already noticed in 2020 (Molenberghs et al. 2020).

Important lessons should be – and are being – learned, in terms of preparedness for future pandemics as well as paradigms for more routine clinical trial and epidemiological work. First, the timely availability of high-quality data is crucial. Perfect epidemiological data do not exist, but we should make every effort to have them as qualitative as humanly possible and then use the most appropriate methodology to draw nuanced conclusions from imperfect and incomplete data. The pandemic has catalyzed the evolution towards combining experimental (clinical trial) data with epidemiological data (natural history data, real-world evidence) and with representative survey data (e.g., sentinel surveys, serological surveys, public opinion surveys, etc.). This is by no means restricted to infectious diseases, but is seen in many areas, ranging from rare diseases to common neurodegenerative diseases such as Alzheimer’s Disease and Parkinson.

Second, we should reflect on which endpoints to use when conducting, in particular, vaccine trials. While vaccines may not provide high levels of protection against (symptomatic) infections, because of waning, ultimately the way in which they protect from severe infection, requiring hospitalization or ICU admission, and mortality, is what matters. Related to this, concerted efforts are needed to evaluate correlates-of-protection that not only relate to vaccine effectiveness in the short term (humoral immunity) but also in the long run, against the background of viral mutation and regular boosting (likely cellular immunity). Important work has been done in this area (Gilbert et al. 2023), but a systematic approach at all levels is welcome.

Third, in non-pandemic times, it is crucial to invest in health literacy. Misconceptions arising from Simpson’s paradox in vaccine efficacy (relating to their age dependence), over genuine hesitancy to outright conspiracy theory may be the undoing of brilliant clinical and epidemiological research. Arguably, biostatisticians, properly trained in science communication and outreach, are well placed to explain complicated and potentially confusing concepts (Grieve et al. 2023). This is especially important when potentially invasive measures are considered or implemented, such as COVID-19 passes (Natalia et al.  2023), curfews, or vaccination mandates.