Causal Inference
Herpes Zoster Vaccine for the Prevention of Dementia: A Target Trial Emulation Using Electronic Health Records Guilin Li* Guilin Li Michael J. Figueroa Muñiz Hanna Gerlovin William Robb Alejandro Szmulewicz Miguel A. Hernán Barbra A. Dickerman
Several observational studies have reported a lower risk of dementia among recipients of the herpes zoster (shingles) vaccine compared with non-recipients. Although the prospect of preventing dementia with a vaccine already recommended for use among adults aged ≥60 years for the prevention of herpes zoster is very appealing, observational studies are susceptible to selection bias (when restricting the analysis to persons who sought health care during follow-up) and confounding due to insufficient adjustment (e.g., for health care-seeking behaviors). As an alternative approach to confounding adjustment, some observational studies have implemented a regression discontinuity design (based on a birth date threshold), but the generalizability of these findings to broader age ranges is unclear.
In this study, we used nationwide electronic health records from the Department of Veterans Affairs, the largest integrated healthcare system in the United States, to emulate a target trial of Zostavax (zoster vaccine live) vaccination and dementia among ~2.5 million eligible individuals (median age 65 years, IQR 12 years) over up to 10 years of follow-up. First, we used inverse-probability weighted pooled logistic regression models to estimate dementia risks, adjusting for baseline and time-varying confounders. Second, we explored the influence of closely matching individuals in each vaccine group according to key confounders. We also evaluated the potential for residual or unmeasured confounding using negative controls and by benchmarking against a randomized trial.