Model-based evaluation of influenza vaccination outcomes

andreashandel.com/presentations

Andreas Handel

2026-09-08

Talk Overview

  • Part 1
    Heterologous immunogenicity of high-dose (HD) influenza vaccines
  • Part 2
    Estimating improved protection of HD vaccines
  • Part 3
    The role of prior immunity on immunogenicity
  • Part 4
    Software, courses and other resources

Overall Motivation

  • Influenza vaccines work, but could be better.
  • Older individuals often don’t respond well to vaccines.
  • Influenza virus keeps evolving (strain mismatch).
  • Pre-existing immunity often leads to poor induction of immune responses.

A better understanding of existing vaccines can help design better future ones.

Part 1: Heterologous immunogenicity of high-dose (HD) vaccines

Billings et al 2025 JID

The impact of dose for influenza vaccination

  • One of the currently available enhanced vaccines is Fluzone High-Dose (HD).
  • The HD vaccine protect(ed) better compared to the standard dose (SD) vaccine against matching strains.
  • How broad is this protection if strains are mis-matched?

The data

Immunogenicity (antibody) data from a large vaccination cohort. All individuals ≥65y.

Vaccine Strains

“Raw” data — homologous responses

“Raw” data — heterologous responses

The model (brms code)

Bayesian, hierarchical model. Nesting of strains/individuals/seasons.

See paper supplement for all the math.

H1N1 antibodies after HD/SD vaccine

H3N2 antibodies after HD/SD vaccine

Overall response

Summing across all strains and subtypes.

Part 1 Summary

  • Overall, the HD vaccine seems to induce a somewhat better response compared to the SD vaccine.
  • The effect is not strong, and varies by vaccine strain/season.
  • It might be worth further tweaking/increasing the dose.

Part 2: Estimating improved protection of HD vaccines

Hammerton et al 2025 JID

Introduction

  • In 2011-13, HD vaccine showed better protection compared to standard dose (SD).
  • Strains are being updated each year, we don’t know if HD is still better.
  • How do different antibody levels map to protection?

Coudeville et al, 2010 BMC Med Res Methodol

Approach Overview (1/2)

Approach Overview (2/2)

VE Estimates

VE Comparisons: OASD vs. YASD

VE Comparisons: OAHD vs. YASD

VE Comparisons: OAHD vs. OASD

Part 2 Summary

  • The HD vaccine is predicted to improve protection somewhat compared to the SD vaccine.
  • The effect is not strong, and varies by vaccine strain/season.
  • It might be worth further tweaking/increasing the dose or giving HD to all ages.

Part 3: The role of prior immunity on immunogenicity

Introduction

  • Influenza vaccines are given in the context of prior immunity.
  • Pre-existing antibodies have been associated with lower response (ceiling effect, response blunting).
  • (Almost) all models of antibody vaccine responses include pre-existing antibodies in a linear model.

Data

Models

  • Fit linear, sigmoid and spline models to each subset (season/strain/vaccine).
  • Compare quality of fits with cross-validated RMSE.
  • Perform (Bayesian) resampling to obtain distributions and uncertainty estimates.

Individual Results Example

Resampling Schematic

HAI Results

MN and ELISA Results

Part 3 Summary

  • Linear models seem to be sufficient in most cases.
  • Disentangling statistical effects from a real signal requires further investigation.
  • More details: github.com/ahgroup/hemme-pare-public

Part 4: Software, courses and other resources

Jorge Cham · phdcomics.com

R packages

  • DSAIRM teaching · stable
    Dynamical Systems Approach to Immune Response Modeling — teaches modeling for within-host infection and immunology.
  • DSAIDE teaching · stable
    Dynamical Systems Approach to Infectious Disease Epidemiology (Ecology/Evolution) — teaches model-based infectious disease epidemiology in a user-friendly way.
  • modelbuilder research · less stable
    Build and analyze simulation models without the need to write code.
  • flowdiagramr research · less stable
    Easy creation of high-quality flow diagrams.

Courses

Other Resources

Acknowledgements

Zane Billings

Savannah Miller/Hammerton

Hayley Hemme

Co-authors, collaborators, NIH

Questions?

Jorge Cham · phdcomics.com

Slides: andreashandel.com/presentations