Highlights
- A 2-year pilot RCT (PMID: 41770546) suggests that low-dose lithium carbonate is feasible and safe for older adults with Mild Cognitive Impairment (MCI).
- While primary outcomes did not reach significance, a trend toward slower decline in verbal memory (CVLT-II) was observed (P = .05), suggesting a potential cognitive signal.
- Lithium’s role is being re-evaluated alongside emerging disease-modifying therapies like lecanemab and novel multisystem targets such as GPR3 and the gut-brain axis.
- Advanced machine learning and multimodal modeling (PMID: 41376794) are essential for identifying MCI subgroups likely to benefit from neuroprotective agents like lithium.
Background
Mild Cognitive Impairment (MCI) represents a critical clinical window for intervention, standing at the precipice between normal age-related cognitive decline and Alzheimer disease (AD) dementia. Despite the recent approval of amyloid-targeting monoclonal antibodies, there remains a profound unmet need for safe, accessible, and pleiotropic pharmacological agents that can stabilize cognitive function and modulate neurodegeneration. Lithium, traditionally used at high doses for bipolar disorder, has long been hypothesized to possess neuroprotective properties at sub-therapeutic levels. Preclinical and epidemiological evidence suggests that lithium may inhibit glycogen synthase kinase-3 beta (GSK-3β), enhance brain-derived neurotrophic factor (BDNF) expression, and potentially reduce amyloid-beta (Aβ) and tau phosphorylation. However, robust clinical trial data in the MCI population have been scarce until the completion of the University of Pittsburgh pilot feasibility trial.
Key Content
The Pilot Randomized Clinical Trial: Gildengers et al. (2026)
The primary focus of recent clinical attention is the single-site, randomized, double-blind, placebo-controlled pilot trial (NCT03185208) conducted by Gildengers and colleagues. This trial aimed to assess whether daily low-dose lithium carbonate could delay cognitive decline over a 2-year period in adults aged 60 and older with MCI.
Methodology and Participants: The study randomized 83 participants (41 lithium, 42 placebo). The intervention aimed for low-dose exposure, balancing the need for neuroprotective potential with the safety concerns inherent in an elderly population. The trial’s methodology utilized linear mixed-effects models in an intention-to-treat (ITT) framework, focusing on six coprimary outcomes spanning cognitive performance (CVLT-II, BVMT-R, PACC), neuroimaging (hippocampal and cortical gray matter volume), and plasma BDNF levels.
Cognitive and Imaging Outcomes: After 24 months, none of the six coprimary outcomes met the prespecified significance threshold for efficacy. However, a clinically interesting signal emerged in the California Verbal Learning Test-II (CVLT-II) delayed recall. The placebo group declined by 1.42 points annually, whereas the lithium group declined by only 0.73 points (difference 0.69 points/year; 95% CI, 0.01-1.37; P = .05). Volumetric analyses showed steady declines in hippocampal and cortical volumes in both groups, without a significant treatment-by-time interaction.
Safety and Tolerability: In an elderly MCI cohort, safety is paramount. The trial confirmed that low-dose lithium is generally well-tolerated. Common adverse events included increased creatinine levels (29% lithium vs. 31% placebo), diarrhea (29% vs. 15%), and tremors (24% vs. 15%). Importantly, serious adverse events were comparable between groups (29% vs. 23%), and none were definitively linked to the treatment. This establishes a clear safety profile for larger Phase II/III trials.
Lithium in the Context of Modern AD Therapeutics
To understand the clinical relevance of lithium, it must be compared with the evolving landscape of AD therapies:
Biologics and Modeling: Modeling studies on lecanemab (PMID: 41603883) suggest that disease-modifying therapies (DMTs) can delay the progression to moderate and severe AD by nearly 2 years. While lecanemab targets Aβ clearance, lithium’s proposed mechanism is more upstream or adjunctive, focusing on neuronal resilience and synaptic plasticity. The integration of lithium with biologics could represent a future multi-target strategy.
Metabolic and Inflammatory Modulators: Recent research into the microbiota-gut-brain axis (PMID: 41527932) highlights how Mediterranean diet-derived bioactives (Neurosyn240) reduce hippocampal amyloid deposits and microglial activation in animal models. Similarly, the study of gut-derived metabolites like TMAO (PMID: 41459734) shows that systemic metabolic products can exacerbate tau phosphorylation via HIF1α signaling. Lithium may interact with these pathways by modulating the systemic inflammatory milieu, as seen in cross-disorder analyses of CXCL8 in PTSD and AD (PMID: 41510671).
Mechanistic Insights and Translational Implications
Lithium’s neuroprotective potential likely stems from its ability to restore redox homeostasis and attenuate neuroinflammation. This aligns with findings from novel thieno-oxazine hybrids (PMID: 41481412) and selenium-based compounds like 4-PSQ (PMID: 41697767), which normalize corticosterone levels and restore antioxidant enzyme activity.
Furthermore, the role of gasotransmitters (Nitric Oxide, Carbon Monoxide, and Hydrogen Sulfide) as modulators of adult neurogenesis (PMID: 41396689) provides a biological rationale for why even low-dose lithium might support cognitive reserve, despite failing to drastically alter bulk brain volumes in the short term. The focus is shifting from simply clearing Aβ to protecting the neurons from the resulting oxidative stress and inflammatory cascades.
Methodological Advances in MCI Research
The Gildengers trial underscores the difficulty of reaching significance in pilot MCI trials due to heterogeneity. Recent advances in computational medicine offer solutions:
Multimodal Forecasting: The use of Attention-based LSTM and multimodal frameworks (PMID: 41376794) allows for the stratification of patients into “Low,” “Mild,” and “Fast” progressors. Applying such models to lithium trials could help identify a subgroup of “Fast” progressors who might show a more significant response to intervention.
Missing Data and Biomarkers: High missingness in neuroimaging studies can be mitigated by machine learning imputation techniques like MissForest (PMID: 41684835). Additionally, tools like CEAM for analyzing brain DNA methylation (PMID: 41428862) provide a more granular view of cell-type specific changes, which could serve as more sensitive endpoints than gross volumetric MRI.
Expert Commentary
The pilot RCT of low-dose lithium in MCI provides a vital “proof of concept” for feasibility and safety rather than a definitive statement on efficacy. The P-value of .05 for CVLT-II delayed recall is tantalizing but requires cautious interpretation. In the context of clinical medicine, this signal suggests that lithium might preserve specific aspects of episodic memory, which is often the first domain affected in the prodromal phase of AD.
Controversy remains regarding the “lithium deficiency” hypothesis. Is lithium an essential micronutrient for cognitive health, or is its effect purely pharmacological? The strong correlation between cancer and dementia incidence worldwide (PMID: 41427617) suggests shared cellular determinants, such as DNA repair mechanisms and apoptosis regulation (Bcl-2/Bax pathways), both of which are modulated by lithium and other novel agents like thieno-oxazine hybrids (PMID: 41481412).
From a guideline perspective, lithium is not currently recommended for MCI. However, for patients who cannot access or afford monoclonal antibodies, or those with comorbidities like depression (where 4-PSQ and lithium both show promise), low-dose lithium represents a compelling area for further research. The primary limitation of the Gildengers study was its sample size; a larger, multicenter trial is now justified, potentially using multimodal biomarkers to enrich the participant pool.
Conclusion
The 2-year pilot randomized clinical trial of low-dose lithium carbonate successfully established the feasibility and safety of this intervention in an older MCI population. Although the coprimary outcomes did not reach statistical significance, the favorable trend in verbal memory decline provides a foundation for future Phase III investigations. Progress in the field is increasingly defined by multi-target approaches—integrating lithium’s neuroprotective effects with amyloid-clearing biologics, gut-microbiota modulation, and advanced AI-driven patient stratification. The future of MCI therapy lies in precision medicine, where low-dose lithium may yet find its place as a cornerstone of cognitive stabilization.
References
- Gildengers AG, et al. Low-Dose Lithium for Mild Cognitive Impairment: A Pilot Randomized Clinical Trial. JAMA neurology. 2026. PMID: 41770546.
- Lopresti BJ, et al. Estimating the long-term health outcomes of treatment with lecanemab in early Alzheimer’s disease: a modelling study. J Med Econ. 2026. PMID: 41603883.
- Ibrahim TS, et al. Utilizing multimodal models to forecast Alzheimer’s disease progression and clinical subtypes. Health Inf Sci Syst. 2025. PMID: 41376794.
- Santini T, et al. A novel Mediterranean diet-inspired supplement reduces hippocampal amyloid deposits and microglial activation. Gut Microbes. 2026. PMID: 41527932.
- Diaz JL, et al. Trimethylamine-N-oxide exacerbates Alzheimer’s disease progression via targeting HIF1α signaling. Gut Microbes. 2026. PMID: 41459734.
- Royse SK, et al. Machine learning for missing data imputation in Alzheimer’s research: predicting medial temporal lobe dynamic flexibility. Cogn Neurodyn. 2026. PMID: 41684835.
