← Back
NeuroTrials.ai
Neurology Clinical Trial Database

LATTICE

Low-Dose Lithium for Mild Cognitive Impairment: A Pilot Randomized Clinical Trial

Year of Publication: 2026

Authors: Gildengers AG, Ibrahim TS, Anderson SJ, et al.

Journal: JAMA Neurology

Citation: JAMA Neurol. 2026;83(4):310-319

Link: https://doi.org/10.1001/jamaneurol.2026.0072


Clinical Question

Does low-dose lithium slow cognitive decline and preserve brain volume in adults with mild cognitive impairment?

Bottom Line

Low-dose lithium (150-300 mg daily) did not meet any of its 6 coprimary endpoints over 2 years in adults with MCI. A nominally significant verbal memory benefit (P=0.05) and a numerical trend toward hippocampal preservation were observed but fell short of the prespecified threshold. Larger effects in amyloid-positive participants warrant further investigation in enriched populations.

Major Points

  • None of the 6 coprimary endpoints reached the prespecified P<0.01 significance threshold -- the trial's primary result is negative.
  • CVLT-II verbal memory showed the strongest signal: lithium reduced annual decline by ~49% (0.73 vs 1.42 pts/year, diff 0.69, 95% CI 0.01-1.37, P=0.05) -- nominally significant but below threshold.
  • Hippocampal volume loss was numerically less with lithium (59 mm3/year difference, P=0.09) -- a trend without significance.
  • Amyloid-positive subgroup showed larger effect sizes: Hedges g=0.74 for verbal memory and g=0.82 for hippocampal volume, suggesting potential benefit in those with underlying Alzheimer pathology.
  • SAEs occurred in 29% lithium vs 23% placebo; none were considered definitely treatment-related, supporting tolerability of low-dose lithium.
  • As a single-site pilot trial (n=80), the study was underpowered -- results are hypothesis-generating and should not change clinical practice.
  • Findings support a larger RCT enriched for amyloid-positive MCI participants to adequately test lithium as a disease-modifying agent.

Design

Study Type: Single-center, randomized, double-blind, placebo-controlled pilot trial

Randomization: 1

Blinding: Double-blind (participants and investigators blinded to treatment assignment)

Allocation: 1:1

Enrollment Period: February 2018 - August 2022

Follow-up Duration: 2 years

Centers: 1

Countries: USA

Sample Size: 83

Analyzed: 80

Analysis: Linear mixed-effects trajectory models with random intercept and slope; treatment × time interaction as primary comparison

Power Calculation: Powered to detect medium effect sizes (d=0.54-0.72 depending on correlation between repeated measures) for each coprimary outcome at P<0.01

Registration: NCT03185208


Inclusion Criteria

  • Age 60 years or older
  • Diagnosis of mild cognitive impairment (MCI) per Petersen criteria
  • Preserved activities of daily living
  • Ability to complete neuropsychological testing

Exclusion Criteria

  • Major psychiatric illness per MINI structured interview
  • Major neurologic illness
  • Contraindications to lithium
  • Nonremediable sensory or motor impairments (e.g., blindness)

Baseline Characteristics

CharacteristicLithium (n=41)Placebo (n=39)
Age, mean (SD)72.93 (8.77) years71.22 (6.47) years
Female, %56% (23/41)56% (22/39)

Arms

FieldLithiumControl
N4139
InterventionLithium carbonate 150 mg or 300 mg daily (dose adjusted to maximum tolerated), taken orally for 2 yearsMatching placebo capsule taken orally daily
Duration2 years2 years

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
6 coprimary outcomes at 2 years: (1) CVLT-II delayed recall (verbal memory), (2) BVMT-R delayed recall (visual memory), (3) PACC (global cognitive composite), (4) hippocampal volume, (5) cortical gray matter volume, (6) plasma BDNF -- assessed via linear mixed-effects trajectory modelsPrimaryNone of 6 outcomes met thresholdNone of 6 outcomes met thresholdNone reached P<0.01 prespecified thresholdAll P>0.01
CVLT-II verbal memory -- annual rate of decline (treatment × time interaction)Secondary1.42 pts/year decline0.73 pts/year decline0.05
Hippocampal volume -- annual rate of change (treatment × time interaction)SecondaryGreater annual loss59 mm3/year less loss than placebo0.09
Cortical gray matter volume -- annual rate of change (treatment × time interaction)Secondary-352 mm3/year difference0.78
BVMT-R visual memory compositeSecondary0.78
PACC global cognitive compositeSecondary0.80
Plasma BDNF (log-transformed)Secondary0.93
Amyloid-positive completers -- CVLT-II effect sizeSecondaryHedges g=0.74 (exploratory)
Amyloid-positive completers -- hippocampal volume effect sizeSecondaryHedges g=0.82 (exploratory)
Serious adverse events (SAEs)Safety23% (9/39)29% (12/41)
Increased creatinineSafety31% (12/39)29% (12/41)
DiarrheaSafety15% (6/39)29% (12/41)
TirednessSafety15% (6/39)29% (12/41)
TremorSafety15% (6/39)24% (10/41)
DeathSafety1 (not study related)0
Serious Adverse EventsAdverse23% (9/39)29% (12/41)

Subgroup Analysis

Amyloid-positive completers showed larger effect sizes vs amyloid-negative completers: Hedges g=0.74 vs 0.32 for CVLT-II verbal memory, g=0.82 vs 0.09 for hippocampal volume, and g=0.81 vs 0.12 for hippocampal percentage change. Exploratory analysis with small subgroup sizes; suggests potential greater neuroprotective benefit in individuals with underlying Alzheimer amyloid pathology.


Criticisms

  • Single-site pilot trial with limited statistical power and generalizability
  • 6 coprimary endpoints tested at prespecified P<0.01 threshold to address multiple comparisons; still susceptible to false-positive risk in exploratory analyses
  • Small sample size (n=80) means most individual endpoints are severely underpowered
  • Sample predominantly amyloid-negative (~68%), likely diluting any true disease-modifying effect
  • Amyloid subgroup analysis was exploratory with small subgroup sizes -- highly susceptible to type I error
  • 36% overall medication discontinuation may limit interpretation
  • The 2-year follow-up may be insufficient to detect disease modification effects in MCI
  • Conducted during COVID-19 pandemic which affected assessments and possibly retention

Funding

Primary support: National Institute on Aging grant R01 AG055389. Additional support: K23 AG076663 (Weinstein) and R01 AG083874 (Karikari); 7 Tesla Bioengineering Research Program; imaging developments supported by R01MH111265, R01AG063525, and T32MH119168; University of Pittsburgh Center for Research Computing and Data (H2P cluster) supported by National Science Foundation award OAC-2117681; open access publication supported by the Dale and Alvin Filstrup Charitable Gift Fund.

Based on: LATTICE (JAMA Neurology, 2026)

Authors: Gildengers AG, Ibrahim TS, Anderson SJ, et al.

Citation: JAMA Neurol. 2026;83(4):310-319

Content summarized and formatted by NeuroTrials.ai.