Not all immune responses leave the same mark. In our latest work, we present a new model to better understand how antibody levels change over time and what this can tell us about past infections. Most public health analyses flatten questions of immunity into a binary: you’re either immune or you’re not. But immunity is not a switch but a spectrum. In our new study, we detail a statistical approach that embraces this complexity and allows us to detect more subtle patterns in immunity and reinfection risk.
We applied a Bayesian mixture model to cross-sectional serological data collected in Belgium for two common viruses: varicella-zoster virus (VZV) and parvovirus B19 (PVB19). While VZV is thought to induce lifelong immunity, there is growing evidence that immunity to PVB19 may wane over time. Using data from 2,382 people, the model tracked antibody levels against VZV and PVB19 and mapped them against age. The approach incorporated individual variability and allowed for changing immunity over time. Crucially, the model didn’t rely on a fixed threshold but instead inferred patterns from the raw data using Bayesian techniques, such as Markov-chain Monte-Carlo (MCMC) sampling.
Stable or variable
Our results show a key difference between the two pathogens. For PVB19, we found that while antibody levels rise steeply in childhood, they begin to decline between the ages of 20 and 40, before increasing again in older age groups. This pattern could indicate a loss of detectable immunity in some individuals, or possible reinfection at later ages. In contrast, for VZV, antibody levels remain relatively stable among seropositive individuals, with only a small decrease with age, consistent with long-term immunity.
The added value of our model lies in its ability to work with continuous antibody data rather than relying on arbitrary cut-offs. This means we can better capture how immunity varies between individuals and over time. It also allows us to distinguish between different possible causes of changes in population-level immunity, such as waning antibody levels versus changes in exposure risk.
Nuanced immune dynamics
This work contributes to a more nuanced understanding of immune dynamics in populations, especially for infections where immunity is not lifelong. We hope this approach will be useful in future sero-epidemiological studies and in better interpreting what antibody levels really tell us about immunity, exposure, and susceptibility.
It has been great to work together with our collaborator, Adelino Martins at the Eduardo Mondlane University in Mozambique.
Read the paper here
