Association of Herpes Simplex Virus Type-1 With Dementia Outcomes: Longitudinal Retrospective Cohort Study Using Real-World Electronic Health Record Data

Abstract Background Global dementia cases, currently exceeding 55 million, are projected to triple by 2050. In the absence of disease-modifying therapies, identifying modifiable risk factors is critical. Preclinical studies show that herpes simplex virus type-1 (HSV-1) is neurotropic and can drive amyloid-beta accumulation and tau hyperphosphorylation. Epidemiologic findings remain inconsistent, partly because many studies lack standardized designs for real-world data. Objective To quantify the association between clinically coded HSV-1 diagnosis and incident cognitive impairment or dementia among patients receiving care in a health-system electronic health record (EHR) network. Methods We assembled a retrospective cohort within an Observational Medical Outcomes Partnership (OMOP)–mapped EHR data lake (2010‐2024), comparing patients with a first HSV-1 diagnosis (n=6274) to those without HSV-1 codes but with parallel baseline criteria (n=379,975). Eligibility required at least 365 days of prior observation and absence of baseline cognitive, HIV, selected neurotropic viral, or recent transplant codes. The outcome combined mild cognitive impairment and Alzheimer disease and related dementias (ADRD). A high-dimensional propensity score (PS) incorporating approximately 17,000 baseline covariates was estimated with regularized logistic regression; 4 strata weights were applied in a Cox model. Sensitivity analyses examined equipoise trimming, doubly adjusted models, fixed 1-, 5-, and 10-year censoring horizons, and a dementia-only outcome subset. Negative-control outcomes (n=271) were prespecified to assess empirical calibration. Results The HSV-1 cohort contributed 23,186 person-years (PY) and 469 events; the non-HSV-1 cohort contributed 1,519,827 PY and 24,764 events. Crude incidence was 20.23 (95% CI 18.44‐22.14) per 1000 PY in the HSV-1 cohort and 16.29 (95% CI 16.09‐16.50) in the non-HSV-1 cohort, an absolute difference of 3.94 events per 1000 PY. PS-stratified analysis yielded a hazard ratio (HR) of 1.14 (95% CI 1.03‐1.25; P =.007). Estimates were comparable after equipoise trimming (HR 1.12, 95% CI 1.01‐1.25, P =.04), doubly adjusted modeling (HR 1.13, 95% CI 1.02‐1.24, P =.01), and fixed 5-year ( P =.008) and 10-year follow-up ( P =.004) (both HR 1.15). The 1-year censored model showed HR 1.00 (95% CI 0.84‐1.20; P =.96). The dementia-only endpoint mirrored the primary analysis (HR 1.14, 95% CI 1.03‐1.25; P =.008). No postindex events occurred for any negative-control outcome, so empirical calibration could not be performed and all estimates are uncalibrated. Conclusions Within an OMOP-standardized EHR cohort, an HSV-1 diagnosis was associated with a small elevation in the hazard of subsequent cognitive impairment or dementia compared with individuals with no recorded HSV-1 diagnosis. Given the ubiquity of HSV-1, even a modest elevation in individual hazard could translate to a meaningful population-level increment if the association is causal. Confirmation in external data sets with laboratory viral typing, antiviral treatment records, and mortality linkage is warranted.