Annual immunogenicity measurements in randomized clinical trials help reveal heterogeneity in long-term vaccine efficacy: a pooled logistic regression modeling study

Vaccine development requires understanding the duration of vaccine-induced protection and whether it differs across demographic groups. Phase 2b and phase 3 efficacy trials are rarely powered to assess annual efficacy or subgroup differences in long-term protection, so these attributes can remain unevaluated until years after licensure, when observational effectiveness studies become available. However, late-stage clinical trials typically collect longitudinal immunogenicity data (for established or putative correlates of protection), which could be useful for predicting vaccine efficacy over time. To address this, a weighted cross-sectionally pooled logistic regression model was applied to immunogenicity substudies from two phase 3 trials to predict the durability of protection in specific subgroups: baseline dengue seronegative and seropositive individuals for dengue vaccine, and younger (≤ 69 years) and older (≥ 70 years) individuals for zoster vaccine. Incorporating time-dependent immunogenicity data increased the precision of time-varying efficacy estimates compared with traditional case-count methods. The modeling framework extends prior immune correlate analyses that used immunogenicity measured at a single timepoint post vaccination (without requiring an absolute or mechanistic correlate of protection). When time-dependent immune markers are sufficiently predictive of protection, the framework helps reveal trends in efficacy that would otherwise be obscured by sparse case counts or limited follow-up. Collecting and modeling annual immunogenicity data in clinical trials (whether from all participants, an immunogenicity cohort, or a case-cohort substudy) can thus accelerate understanding of the persistence of vaccine-induced protection (across and within subgroups) and guide efficient development of vaccines that are protective for years.