Can Lithium Slow Alzheimer's?

What LATTICE actually tested

LATTICE was a single-site, double-blind, randomized, placebo-controlled pilot trial in adults 60 and older with mild cognitive impairment (MCI), a decline in memory or thinking short of dementia that can precede it. Of 83 randomized, 80 started treatment, 41 on low-dose lithium carbonate and 39 on placebo, over two years. Six prespecified coprimary outcomes: verbal memory, visual memory, a composite cognitive score, hippocampal volume (a memory-related region), cortical gray matter volume, and brain-derived neurotrophic factor (BDNF), a blood growth factor tied to nerve-cell health. A result had to clear p less than 0.01.

None cleared it, so on its face this is negative. Two features complicate that. First, the dose was very low: among completers, mean daily dose was 195 mg lithium carbonate, mean blood lithium 0.17 mEq/L, well under bipolar targets and below the 0.25 to 0.5 mEq/L used in earlier positive Forlenza MCI trials. Second, LATTICE was a feasibility study meant to estimate effect sizes, not test efficacy. At about 40 per arm, it was underpowered to detect a modest effect.

The one signal worth noticing

The largest cognitive effect appeared on verbal memory, measured by delayed recall on the California Verbal Learning Test-II (CVLT-II), a word-list test. Placebo declined 1.42 points per year, lithium 0.73, a difference of 0.69 (95% CI 0.01 to 1.37), roughly halving decline from a baseline near eight points. A lead to chase, not proof.

One more wrinkle: several measures, including visual memory, the cognitive composite, and BDNF, stayed relatively stable rather than declining on placebo, so lithium had no decline to slow. That makes the null less clean than 0-for-6 and means fewer real chances to detect an effect than the headline suggests.

Who was actually in the trial matters

Only about one in four participants had amyloid-beta plaques (protein clumps that are a core sign of Alzheimer's biology) on brain scans. The rest were amyloid-negative or unknown, so many may have had MCI from non-Alzheimer causes such as vascular disease, Lewy body disease, or frontotemporal degeneration. If lithium works mainly on Alzheimer-specific biology, a mostly amyloid-negative sample would dilute any signal. In the small amyloid-positive subgroup, exploratory differences looked larger for verbal memory and hippocampal volume, but came from very few people and cannot be relied on.

Prior human evidence is mixed. Two Forlenza randomized trials in amnestic MCI used lithium carbonate at 0.25 to 0.5 mEq/L, one over 12 months, the later a two-year double-blind phase. In both, lithium patients tended to stay cognitively stable while placebo declined, with favorable cerebrospinal fluid changes: reduced phosphorylated tau (linked to tangles) in the 12-month trial and higher amyloid-beta 42 (a plaque-related fragment) in the later. Shorter trials in established dementia disappointed: a 10-week multicenter trial and the 12-week Lit-AD trial showed no clear cognitive benefit. Observational studies link lithium to lower dementia risk, but findings vary and cannot prove cause. Bipolar disorder itself carries roughly triple the dementia risk and lithium goes preferentially to those patients, making confounding hard to untangle. One Taiwan study showing higher Alzheimer's risk in the general population found no association after restricting to bipolar patients.

The orotate question, and why it is still unanswered

Much of the recent excitement traces to a 2025 Nature study reporting lithium was reduced in the brains of people with MCI and Alzheimer's, apparently trapped inside amyloid plaques, while blood lithium looked normal. The researchers proposed lithium orotate, a less ionized organic salt sold as a supplement, may be less trapped by amyloid than carbonate. In Alzheimer's mouse models, low-dose orotate sharply cut plaque burden and restored memory measures, while carbonate did less or nothing. This is a mechanistic hypothesis from tissue, cell, and mouse data, not a treatment shown in people. The small positive human MCI trials used carbonate, and no completed trial has compared orotate with carbonate head to head.

That gap matters because orotate products are already sold online. U.S. labels for lithium carbonate list bipolar I disorder, not Alzheimer's, MCI, or dementia prevention, and FDA warned at least one company in 2018 over Alzheimer's and neurodegenerative claims for an orotate product. As of July 2026, a human trial of orotate in biomarker-confirmed early Alzheimer's is registered but not recruiting; its aim is feasibility, safety, tolerability, and cerebrospinal fluid lithium over nine weeks, not efficacy. Even low-dose lithium needs monitoring: longer-term exposure, especially at higher blood levels, is linked to chronic kidney disease, hypothyroidism (an underactive thyroid), and elevated calcium.

What this means for you

LATTICE neither closes nor opens the door on lithium for Alzheimer's. Its finding is narrow: very low-dose lithium carbonate, in a mostly amyloid-negative population, produced no statistically significant benefit over two years, though it was generally tolerated and left a faint verbal-memory signal worth testing properly. A larger trial using amyloid-confirmed participants at an adequate, monitored dose has not been run.

For a prevention-minded reader, the evidence does not support taking lithium, carbonate or orotate, to preserve cognition. The animal orotate data are no basis for self-experimentation, and supplement doses delivering tens of milligrams of elemental lithium (the actual lithium in the salt) differ from trace dietary or drinking-water exposure. If you worry about your risk, the more actionable step is knowing whether Alzheimer's biology is present. In specialty care, validated blood-based biomarkers such as plasma p-tau217 (a tau signal in blood) can help estimate whether amyloid or Alzheimer-type pathology is present in people with cognitive symptoms, interpreted with a clinician as part of a full workup. That tells you whether Alzheimer's biology is even the relevant question, which LATTICE could not test cleanly.