Syphilis infection and incident dementia in two real-world cohorts: A harmonized propensity-matched analysis in Taiwan NHIRD and TriNetX Asia-Pacific.

BACKGROUND: Syphilis is a treatable infection with established neurotropism and broad systemic inflammatory consequences, but population-level evidence linking a history of syphilis to later dementia remains limited. We examined this association across two independent real-world data sources using harmonized analytic methods.

METHODS: We conducted retrospective cohort analyses in Taiwan's National Health Insurance Research Database (NHIRD, 2001-2020) and the TriNetX Asia-Pacific electronic health record network (2001-2020). Adults aged 20 years or older with syphilis were propensity-score matched to participants without syphilis (NHIRD 1:4; TriNetX 1:1) on age, sex as recorded in each data source, index year, and prespecified comorbidities. Individuals with pre-existing dementia, HIV infection, major psychiatric disorders (bipolar disorder, anxiety disorders, schizophrenia, depression), and other conditions likely to complicate dementia ascertainment were excluded at baseline. Cox proportional-hazards models estimated hazard ratios (HRs) for incident all-cause dementia and dementia subtypes. Cohort-specific estimates were pooled using random-effects meta-analysis. Age- and sex-stratified analyses were performed, and a prespecified one-year lag sensitivity analysis was conducted in NHIRD.

RESULTS: After matching, NHIRD included 9,234 individuals with syphilis and 36,936 controls, and TriNetX included 14,491 participants in each group. A history of syphilis was associated with increased risk of incident all-cause dementia in NHIRD (HR 1.35, 95% CI 1.23-1.48) and TriNetX (HR 1.64, 95% CI 1.42-1.89), yielding a pooled HR of 1.48 (95% CI 1.22-1.79; I2 = 80%). Associations were present in both age strata and in both sexes. The prespecified one-year lag sensitivity analysis in NHIRD remained directionally consistent (HR 1.34, 95% CI 1.22-1.48). Dementia subtype estimates were heterogeneous across data sources and are interpreted as exploratory.

CONCLUSIONS: Across two independent real-world cohorts, a history of syphilis was associated with higher long-term risk of incident all-cause dementia. The association remained positive in the prespecified sensitivity analysis. Subtype findings were inconsistent across data sources and should be interpreted cautiously. These results support further study of infection-related pathways to cognitive decline while not establishing causality or disease-specific mechanisms.