Herpes Zoster Vaccination and Dementia Risk: A Systematic Review of Natural Experiments and Cohort Evidence

Background/Objectives: More than 55 million people live with dementia, and no treatment halts its progression, which makes prevention the priority. Herpes zoster (HZ) results from reactivation of varicella-zoster virus (VZV) and has been linked to a higher dementia risk. Vaccination is already recommended for HZ and postherpetic neuralgia; whether it also protects cognition remains unclear. We critically appraise the evidence linking HZ vaccination to dementia risk across study designs. Methods: Following the PRISMA 2020 statement, we searched PubMed/MEDLINE and Europe PMC from 29 October 2026 with forward citation searching and identified 36 eligible reports: 25 primary investigations (four natural experiments using date-of-birth eligibility cutoffs, one quasi-experimental analysis of a vaccine transition, 17 cohort studies, two nested case–control analyses, one cross-sectional survey; 20 with vaccination as the exposure; and five with infection as the exposure) and 11 secondary syntheses (10 systematic reviews or meta-analyses and one secondary quantification). Six narrative reviews served only as contextual references and were not counted. Results: Natural experiments report a consistent reduction in incident dementia, 8.5% (95% CI 1.9–15.1) on an eligibility basis and about 20% when scaled to vaccine receipt. Cohort analyses show larger, more variable reductions (24% to 51%), part of which reflects healthy vaccinee bias; the association persists, attenuated, against negative-control vaccines. Candidate mechanisms include VZV-specific immunity and broader immune modulation. Conclusions: The association between HZ vaccination and lower subsequent dementia risk is consistent in direction, biologically plausible, and modest in size, but every dementia outcome was graded very low certainty, so causality and magnitude both remain uncertain, and the evidence does not support a dementia-prevention indication. Randomized trials, mechanistic studies, and designs separating an effect on onset from an effect on the timing of diagnosis are needed. Retrospective registration: OSF Registries (10.17605/OSF.IO/MBXWE).