Background: Lithium is a naturally occurring trace element in food and drinking water and is widely used therapeutically for bipolar disorder. Experimental and clinical evidence suggests potential neuroprotective effects, but epidemiologic findings on lithium and dementia remain inconsistent. We examined the association between drinking water lithium exposure and incident dementia in the U.S.

Methods: Drinking water lithium concentrations were obtained from the U.S. EPA’s Fifth Unregulated Contaminant Monitoring Rule (UCMR 5). Regression kriging was applied to estimate nationwide lithium concentrations using UCMR 5 data, U.S. Geological Survey’s groundwater levels and primary water sources. Concentrations were averaged at the 3-digit ZIP code level and linked to participants in the All of Us Research Program by residential location. Dementia outcomes (all-cause dementia, Alzheimer’s disease [AD], and vascular dementia [VD]) were ascertained from electronic health records. Lithium exposure was categorized into four groups. Cox proportional hazards models estimated hazard ratios (HRs), with stratified analyses by sex, age, APOE ε4 carrier status, obesity status, and insurance status.

Results: Among participants residing in the same ZIP code for >5 years, the highest lithium exposure category was associated with a lower hazard of AD (HR = 0.287, 95% CI: 0.097–0.855). Within this group, inverse associations were stronger among APOE ε4 carriers (HR = 0.097, 95% CI: 0.017–0.557) and those with obesity (HR = 0.070, 95% CI: 0.007–0.689). For VD, higher lithium exposure was associated with lower risk among participants living in areas where ≥80% of drinking water was supplied by community water systems (p-trend = 0.027).

Conclusion: Higher estimated drinking water lithium exposure was associated with lower hazards of select dementia outcomes, with stronger inverse associations in specific subgroups. Replication with improved individual-level exposure assessment is warranted.