Lithium and Alzheimer’s disease: what to weave of this LATTICE?

Scientists have been investigating a possible therapeutic role for lithium in Alzheimer’s disease (AD) for two decades now. Last fall, we wrote about an intriguing paper in Nature by Aron et al.¹ that suggested that using a specific salt of lithium (lithium orotate) might make lithium dramatically more effective at preventing and treating AD than the lithium carbonate that is usually used in psychiatry.

We were careful to frame our enthusiasm in conditional terms: the paper reported promising preclinical science, a novel and intriguing mechanism, and a good reason to watch the space closely—but at the end of the day, it was only animal and mechanistic data, not something that should automatically lead every person concerned about their long-term cognitive health to start supplementing lithium.

For that, we should look for clinical trials in humans—and in that vein, we mentioned that a small randomized controlled trial (RCT) testing low-dose lithium’s ability to slow cognitive decline in people on the verge of AD dementia had just wrapped up, and that we were awaiting the published results.

Now that trial—Lithium as A TreatmenT to prevent Impairment of Cognition in Elders (LATTICE)—has been published in JAMA Neurology.² The headline story is that low-dose lithium failed to move any of the trial’s six prespecified primary endpoints. But a closer reading suggests that it’s premature to discount lithium’s potential in AD. What the trial actually tells us is more nuanced than a straight-up failure, and—combined with the Nature paper¹ and previous evidence—gives us a roadmap for exciting future research.

To understand why, let’s first retrace the evidence that brought us to this trial in the first place.

The leadup to LATTICE

Evidence for lithium as a potential Alzheimer’s therapy has accumulated across multiple lines of evidence over more than two decades—none of it offering definitive conclusions, but collectively pointing in a compelling direction.

Something in the water

First was the observational data. A series of observational studies has found that populations living in areas with higher lithium levels in the drinking water tend to have lower rates of dementia and AD. A 2024 meta-analysis³ found evidence of a threshold effect: AD risk was unaffected by rising lithium levels from the lowest levels up to 0.002 mg/L, but once that threshold was passed, increasing lithium levels were associated with progressively lower rates of AD in the community.

Such whole-population (ecological) comparisons are the weakest kind of observational study, and observational studies cannot establish causation and are subject to confounding, but these studies provide one of several layers of evidence that low doses of lithium might prevent AD.

Observational data in bipolar disorder

Lithium has been the mainstay of treatment for bipolar disorder (BPD) since the 1940s, making patients with BPD a natural group in whom to probe this question. Observational studies show that people who are on long-term lithium are at reduced risk of dementia⁴ and show less atrophy of the hippocampus (an area of the brain that is badly afflicted in AD) and less damage to the structures that insulate neurons (white matter) than do BPD patients treated with other medications or who go unmedicated—with genetic support for the white matter finding.⁵

But caution is due. In addition to the usual limitations of observational studies, extrapolating from what happens when you partially normalize the function of the brain of a person with BPD to a benefit that might accrue to people with normal brain function given the same treatment is fraught: this is the same logical error that led to premature enthusiasm about metformin in normoglycemic individuals. Still, these data fit with the mechanistic picture we get from other lines of evidence.

Tantalizing small trials

LATTICE is the latest RCT of lithium to treat or prevent AD, but it is not the first. Forlenza and colleagues at the University of São Paulo found that low-dose lithium slowed cognitive decline compared to placebo over one year in 45 people with so-called amnestic mild cognitive impairment (MCI).⁶ (”Mild” cognitive impairment is a misnomer. MCI constitutes a considerable level of cognitive impairment: much more severe than ordinary forgetfulness, though not yet full-on dementia. We’ll get back to that “amnestic” qualifier later on). During an additional two years of followup, during which volunteers knew whether they’d been assigned to lithium or placebo but investigators were still kept blind, people in the placebo group continued to suffer cognitive and functional decline, while lithium-treated subjects remained stable.⁷

An average of 13 years after the original trial started—which was 9 years after subjects in the trial had stopped receiving more pills—there were still hints that the lithium group’s earlier four years of treatment might have had a longer-term benefit.⁸ The differences were not statistically significant, but there were several factors that would have diluted any signal that was there in the data.

Crucially, this trial also measured lithium’s effect on the molecular damage that is the hallmark of AD: beta-amyloid and an abnormal form of the protein tau. People in the lithium group had increased levels of beta-amyloid 42 in their cerebrospinal fluid (CSF—the fluid that bathes the brain and spinal cord), suggesting improved amyloid clearance out of the brain and into the surrounding fluid, and also had a reduction in abnormal tau at the 12 month mark,⁶ though the tau effect had faded one year after all patients stopped receiving their pills (which was 3 years from the initiation of the original trial).⁷

Lithium also reduced the activity of the enzyme glycogen synthase kinase 3-beta (GSK-3β) in the subjects’ platelets by about 50%. GSK-3β is abnormally active in people with AD: it can convert normal tau protein into one abnormal form, and may also accelerate the production of beta-amyloid. Moreover, lithium has been shown to inhibit GSK-3β in cell culture models and in the brains of animal models of AD, providing protection against molecular damage and cognitive deficits. 

Together, showing that low-dose lithium inhibited GSK-3β activity in human MCI patients’ platelets, potentially increased amyloid clearance, and acutely lowered aberrant tau in their CSF all support the idea that lithium had not just boosted cognitive function like a stimulant, but may have worked by targeting the underlying AD disease process.

In another small trial, Leyhe et al. randomized 27 people in the early stages of existing AD dementia to either low-dose lithium or placebo for ten weeks.⁹ Over the course of the trial, the lithium group’s performance actually improved compared to what it had been at the start of the study, while the placebo group worsened as expected, with a significant time-by-treatment interaction.

Leyhe et al. also looked at the effect of lithium on brain-derived growth factor (BDNF), a protein that supports the survival of existing neurons and the growth and differentiation of new ones. As is typical for people with AD, the subjects in this trial started off with abnormally low levels of BDNF. But after 10 weeks of low-dose lithium, BDNF levels had normalized in the lithium group, while they remained depressed in people on placebo.⁹ This was another sign that lithium might actually be impacting the disease process in AD.

In a larger and more surprising trial, another group assigned 113 people—who again already had AD dementia—to either a microdose of lithium or placebo for 15 months.¹⁰ Where standard lithium carbonate doses used for bipolar disorder range from 900 to 1200 milligrams per day, and the other AD trials used in the range of 150–450 milligrams, this group tested 300 micrograms (0.3 milligrams)—an amount that is only a fraction of people’s typical dietary intake of lithium.¹¹ Cognitive scores remained stable in people assigned to microdose lithium, while those of people on placebo continued to decline.¹⁰

But the picture leading into LATTICE was not uniformly positive. One early open-label low-dose study was negative, but that isn’t a big surprise: it was only ever intended to assess the feasibility of running a proper trial, and its primary outcome was adverse events.¹² Indeed, it was dominated by tolerability problems.¹² Another early open-label study found no effects on cognitive function, but it similarly isn’t strong counter-evidence: it only lasted 5 weeks, it involved just 14 people, and it does not appear to have escalated the dosing appropriately.¹³ Short duration, the lack of a control group, and, in the case of the latter study, accelerated dosing in an older population make these studies less evidential than the previous ones we’ve reviewed.

The Lit-AD trial raises some more specific questions. It found no effect of lower-dose lithium on cognitive scoring,¹⁴ but improved psychiatric symptoms in people with AD.¹⁵ Lithium also had no effect on BDNF in the followup to this trial,¹⁴ which you might have expected to start to move over the course of 12 weeks. We’ll return to this point in discussing LATTICE.

So, to reprise: LATTICE was founded on a significant but far from conclusive evidence base for the idea that lithium might prevent dementia in people on their way to AD. We have observational associations of protection from lithium in drinking water and its use by people with BPD, and animal studies that support the potential causality of those associations.

And we have several small clinical trials, some of which show protection in people with amnestic MCI or even AD. Trials that used higher doses also yoke those clinical results to mechanistic data, including inhibition of GSK-3β (which is the molecular mechanism expected to prevent the malformation of tau and possibly the aggregation of beta-amyloid) and actual reductions in downstream molecular damage. There’s also a signal of increased neuroprotective BDNF. By contrast, shorter-term trials and trials using lower doses tend to show null results on cognitive outcomes, and no effect on BDNF in the one case where it was tested.

To move on from this uncertain signal, we clearly needed more trials. And LATTICE was an attempt to fill in our evidence gap. So what did it test, and what did it find?

How they wove the LATTICE

The LATTICE trial, led by Ariel Gildengers at the University of Pittsburgh, randomized 80 adults aged 60 and older with MCI to either low-dose lithium carbonate or placebo for two years.² The mean daily dose was approximately 195 mg—roughly one-fifth of the dose range typically used to treat bipolar disorder—producing mean serum levels of about 0.17 mEq/L, which is well below the 0.6–1.0 mEq/L range targeted in psychiatry, and also substantially lower than nearly all of the MCI and AD trials we reviewed above.

The investigators tracked six coprimary outcomes: verbal memory (CVLT-II delayed recall), visuospatial memory (BVMT-R), a composite cognitive score (PACC), hippocampal volume, cortical gray matter volume, and plasma BDNF.

To cut to the anticlimax: Not one of these outcomes reached the prespecified significance threshold of p < 0.01.

So far, so null. But the details matter.

The signal that wasn’t—unless it was

The one suggestive finding sits just outside the threshold of statistical significance. On the CVLT-II delayed recall, verbal memory declined at 1.42 recalled words per year in the placebo group versus 0.73 words per year in the lithium group, from a baseline of approximately 8 words.² Thus, nominally, lithium was associated with roughly a halving of the rate of decline. And the treatment-by-time interaction reached p = 0.05, which would normally squeak by the conventional significance test.

But in this study, the threshold had to be made more restrictive, because the fact that the investigators were testing so many different endpoints at once increased the chance that one of them would hit the 0.05 threshold simply by virtue of the researchers having taken so many shots on goal.

Thus, while the direction of this finding is consistent with an effect, a p-value of 0.05 for one of six tests in an underpowered pilot cannot be concluded to be real. Lon Schneider of USC put it bluntly in commentary on the trial: “This kind of profile could come out if you used a wholly inactive ingredient … If you did the study again with a different group of 80 people, there’s roughly a 50% chance you just wouldn’t see this effect.”¹⁶

But it’s also the kind of statistical fragility we would expect from a 40-person treatment arm. Something may be there. We just don’t know yet.

How the LATTICE may have come undone

No target, no benefit

If lithium really is an effective treatment to slow down or delay AD dementia, one potentially major factor that might have diluted any signal that might have come out of LATTICE is that only about one in four participants actually had beta-amyloid in their brains.² In other words, the great majority of the subjects in the trial had cognitive impairment for some reason other than AD: vascular disease, dementia with Lewy bodies, frontotemporal dementia, or something more obscure.

If lithium’s mechanism is specific to Alzheimer’s pathology—correcting a local brain deficiency of lithium created by its sequestration in amyloid, thereby inhibiting GSK-3β hyperactivity, slowing beta-amyloid aggregation, and blocking changes in tau protein—then testing it in a population that is mostly free of beta-amyloid is like testing a blood pressure drug in a cohort whose members mostly have normal blood pressure. You’ve diluted away the signal before the study begins, and you might even harm more people than you benefit.

By contrast, more than 70% of the patients in Forlenza et al. had beta-amyloid and abnormal tau in their CSF (reference 6, and personal communication, Orestes Forlenza). While this wasn’t an admission criterion, it also wasn’t a fluke: Forlenza et al. exclusively enrolled people with amnestic MCI (aMCI).⁶ Compared to people with other MCI variants, people with aMCI are much more likely to have beta-amyloid deposits in their brains¹⁷ and to go on to develop AD rather than some other kind of dementia.^17,18,19

This isn’t the first time that a trial of an amyloid-targeting therapy has enrolled substantial numbers of people with little or no beta-amyloid in their brains (though it may be the first time that investigators went into an AD trial knowing they had many such participants). Back when PET scan tracers to detect beta-amyloid in the brain were a newer, more expensive technology that hadn’t yet proven its worth—and well before we had the blood-based tests that are accelerating AD research today—the researchers behind the Phase III trials for the beta-amyloid-targeting antibodies bapineuzumab and solanezumab discovered too late that about a quarter of the volunteers they had enrolled into their trials did not have meaningful levels of beta-amyloid plaque.^20,21 Obviously, no matter how effective it is, a drug that works by clearing beta-amyloid can’t succeed if there’s no amyloid to clear!

In turn, one of the things that allowed the later anti-amyloid antibodies lecanemab/Leqembi^® and donanemab/Kinsula^® to succeed was that would-be volunteers were tested for beta-amyloid as part of the screening process, and subjects were only admitted into the trial if there was enough of it around for the drugs to have something to work on.

So what happens if we look separately at subjects in LATTICE according to whether they had or did not have beta-amyloid on their brain scans? Suddenly the effects of lithium look impressive: Hedges’ g=0.74 for verbal memory and 0.82 for hippocampal volume in people whose PET scans were positive for beta-amyloid, compared with g=0.32 and 0.09, respectively, in amyloid-negative completers.² These numbers reflect medium-to-large effect sizes, concentrated precisely in the subgroup with AD pathology, and largely absent in those without it.

These numbers sound compelling, but because only a subgroup of people underwent PET scans for beta-amyloid, they come from only 10 and 9 people (respectively), so we can’t take them to the bank. But they are a signal that there might still be something to lithium for AD, and a clear pointer for how to design the next trial.

False negatives?

We mentioned earlier that the headline finding of LATTICE was that lithium had no statistically reliable effect on any of the six cognitive tests the investigators ran.² However, two out of those six tests (PACC for overall cognitive function and BVMT-R delayed recall for visual memory) didn’t decline in the controls over the course of the two-year trial. While you could imagine lithium being so effective that it would actually make people’s cognitive function better than it had been before the trial started, the more reasonable hypothesis—and the one for which the trial was actually powered—was that lithium would moderately slow the rate of decline in cognitive functioning. Obviously, if two of the six tests didn’t change in the controls, you can’t detect a slowing of that change. So perhaps LATTICE’s record is 0 for 4 (or even 1 for 4), not 0 for 6.

Underdosing

A third possible reason for the failure of LATTICE is that the dose was simply too low. The average LATTICE patient’s achieved serum lithium level was 0.17 mEq/L,² which is less than half the level achieved in Forlenza’s positive trial (0.39 mEq/L during the double-blind phase, and 0.41 mEq/L during the open-label extension⁷), and less than a quarter as high as in Leyhe et al.’s 10-week AD trial (0.782 mEq/L⁹).

The biomarker data offer a further, if indirect, signal that the dose was too low to engage its targets. Unfortunately, while the LATTICE investigators had intended to measure GSK-3β activity, they were unable to do so, because their assays failed quality control.² However, we know that Forlenza’s positive trial reported only partial GSK-3β inhibition at the dose they used,⁶ and LATTICE effectively used half as much lithium as Forlenza et al. did.

Similarly, there’s some reason to think that underdosing may be responsible for the lack of an effect on BDNF in LATTICE. The positive signal on BDNF came from Leyhe et al.’s trial, where the mean serum lithium levels were higher than in any of these trials: 0.782 mEq/L. By contrast, lithium had no effect on BDNF in the followup to the Lit-AD trial, in which the serum lithium levels were 0.35 mEq/L¹⁴—half of those in the Leyhe trial, while still coming in two times higher than serum concentrations observed in LATTICE.

We are thus left without some of the key target-engagement data that would tell us whether lithium was modulating its intended molecular targets at the doses achieved, and with some suggestion that the dose may have been too low to do so.

The trouble with Goldilocks

But the efficacy question is only one side of the dosing puzzle. There is also the safety side, and a plausible J-shaped dose-response relationship to contend with. LATTICE did find that very low-dose lithium carbonate was tolerable in this population, with no serious adverse events judged definitely related to study medication.² Common adverse events—diarrhea, tiredness, and tremor—were meaningfully more frequent in the lithium arm than placebo, but serious adverse event rates were only modestly higher (29% vs. 23%).

But as we discussed above, the dosing of lithium that avoided side effects seems likely to have been too low to achieve a therapeutic benefit in these patients. And it’s not that the LATTICE team intentionally planned to use a subtherapeutic dose: the LATTICE clinicians adjusted the participants’ doses every week until they reached the maximum dose that people could tolerate, and then decreased the dose as needed to achieve tolerability. The investigators wrote that “importantly, we observed that older adults have substantial difficulty tolerating doses greater than 300mg daily,” and suggested that future trials should stay within the 150–300 mg daily range. But this potentially sets them up for another failure due to inadequate dosing.

If we look at the opposite side of the dosing spectrum, toxicity from chronic excess lithium in older adults is predominantly neurocognitive in character: its known central nervous system effects include tremor, difficulty concentrating, and (in more severe cases) disorientation and cognitive impairment.²² Even in patients with BPD—in whom long-term lithium is associated with preserved hippocampal volume and reduced dementia risk—there is evidence of on-drug cognitive impairment in some lithium-treated patients.²³ A drug that is simultaneously neuroprotective at the population and structural levels and cognitively impairing at the individual functional level is not contradictory, but it does suggest that precision dosing will be challenging.

This not only suggests that a therapeutic window for cognitive benefit, if it exists, may be narrow. It creates a troubling interpretive ambiguity: in a population already experiencing cognitive decline, low-grade lithium neurotoxicity might be indistinguishable from disease progression. Participants experiencing minor lithium toxicity would be hard to distinguish from those undergoing the unmodified disease process, masking any neuroprotective benefit and potentially generating the appearance of a null result in aggregate, even if a beneficial signal exists at lower doses or in some patients.

This is the new stuff

This is where Aron et al.’s Nature paper¹ becomes especially compelling. The form of lithium used in LATTICE was lithium carbonate, the same form most often used to treat BPD. But as we discussed in our September newsletter, the Nature paper introduced a fascinating mechanistic story and animal evidence suggesting that the carbonate form may be one of the least effective lithium salts for preventing or treating AD.

Aron et al.’s initial finding was that lithium levels were abnormally low in the post-mortem human brain tissue of people with MCI—and were further depleted in Alzheimer’s disease.¹ It wasn’t that people with AD were systemically deficient: serum lithium levels were the same among cognitively intact, MCI, and AD subjects. Instead, the lithium in the MCI and AD brains was being sequestered—it was trapped inside their amyloid plaques, leaving the surrounding brain tissue deficient. None of this was true for the other 27 metals they analyzed.¹

When the investigators screened 16 different lithium salts for therapeutic potential, they found that inorganic salts like lithium carbonate dissociate into lithium ions too easily, rendering the positively-charged lithium ions susceptible to the “magnetic” pull of the negatively-charged amyloid plaques, which then trap the lithium and make it unavailable to protect the brain. Lithium orotate, being less polar and remaining bound to its carrier molecule longer, delivers more lithium to neurons in plaque-free tissue.¹

Layered onto these tissue data were Aron et al.’s studies in mouse models of AD and “normal” aging. When treatment was begun early, lithium orotate almost completely prevented beta-amyloid plaque formation and the early stages of tau damage in AD mouse models, and—more remarkably—it nearly fully restored cognitive function, even when treatment started after plaques and memory impairment were already established. By contrast, lithium carbonate either had no effect on some of these parameters, and much weaker effects on others.¹

And Aron et al.’s Nature paper¹ is not the first to suggest that lithium orotate may have superior effects in the brain relative to carbonate and other inorganic salts. Lithium orotate was first proposed as a more effective form of lithium in the 1970s²⁴ and was supported by some early animal research. A day after administering either lithium orotate or carbonate to rats, brain lithium levels in animals administered lithium orotate were triple those achieved after administering the same amount of lithium as the carbonate.²⁵ Other researchers reported that lithium orotate was more toxic than carbonate,^26,27 but these studies high doses of lithium equivalent to the dose of lithium carbonate used in human BPD patients, when the whole point of using lithium orotate was that one could use much lower amounts of lithium in this form to achieve therapeutic levels specifically in the brain.

In recent years, prior to Aron et al.’s report,¹ researchers at the University of Saskatchewan in Canada revisited this question in an animal model of BPD that responds to lithium. They found that lithium orotate was more effective than lithium carbonate at just one-tenth the dose.²⁸ And thanks to those lower effective doses, lithium orotate was safer: using the doses that were effective in treating the animals’ BPD, lithium carbonate caused the usual problems that plague human lithium users (kidney problems in both sexes and signs of hypothyroidism in females)—but lithium orotate did not.²⁸

Separately, a standard toxicology report was prepared by a contract lab with funding from a lithium orotate supplement vendor, which found no toxicity from lithium orotate to the kidney or any other organ, except in two animals in an extremely high dose group (400 mg elemental lithium per kilogram of body weight). However, no measurements of thyroid hormones were performed.²⁹

If the rodent studies were to translate to humans, it would offer a way out of the J-shaped dosing dilemma that we discussed earlier: we could use low levels of lithium orotate and achieve high brain concentrations that would actively engage its targets instead of being enmeshed in people’s amyloid plaque—and meanwhile, serum levels would remain low, so fewer safety problems would emerge.

But it’s critical to emphasize that this is all still animal data. Every positive human trial of lithium for MCI or AD has used lithium carbonate or another standard formulation, not orotate: that includes Forlenza,⁶ the Nunes microdosing study,¹⁰ and the Leyhe et al. study that reported improved cognitive function and increased BDNF.⁹ It also includes the observational data in people with BPD.^4,5

The potential orotate advantage, however compelling mechanistically and in animal models, has not yet been demonstrated in any human trial. Clinical trials of lithium orotate for AD are, as of this writing, still in the planning phase,^30,31 and one of them is just to test safety, tolerability, and to see if it reaches the CSF.³⁰ Until those (and later) trials are completed, orotate’s superiority in humans remains an hypothesis, not a finding.

What should come next

Disappointing as it may be, what we really need is the perpetual cry of the scientist: “more research.”

The most important step to ensure that future trials can actually give us answers about lithium as an AD prophylactic is to make sure that the people enrolled in such trials actually have AD-driving molecular damage in their brains. Fortunately, we have much cheaper and more convenient ways to do that today than we did when LATTICE began: a simple blood test for plasma ptau217, alone or in combination with other plasma markers.

ptau217 is an abnormal form of the protein tau that reflects the earliest stages of damaged tau in the CSF, before the late-stage, insoluble forms of aberrant tau appear on brain scans.³² Because this early tau damage is driven by beta-amyloid, ptau217 can be used to determine whether people have beta-amyloid in their brains and to track the time-course of its accumulation^32,33,34 and it identifies cognitively-intact people at risk of cognitive decline or MCI.^34,35,36

If the mechanistic and animal data are telling a true story about how the mineral works, then using ptau217 to ensure that participants have beta-amyloid in their brains would ensure that the trial is testing lithium in people who would actually benefit from it.

In addition to choosing the right participants for future trials, we need to determine the doses and serum levels of lithium that actually engage its target(s). Therefore, a study measuring how GSK-3β inhibition, serum BDNF, and plasma ptau217 respond to different doses of lithium should ideally be performed to select a dose that will have a fair chance of succeeding in a larger, longer-term efficacy trial.

Third, no randomized, controlled, head-to-head comparison of lithium carbonate and orotate has ever been done in humans, and is overdue. The first such study could compare the two forms in the kind of dose-ranging target-engagement study we just described.

And once dose ranging studies are completed, researchers should ideally open up the treatment window much earlier with a trial to prevent MCI in biomarker-positive but cognitively intact individuals. This is a stretch goal, but one informed by the available evidence on lithium and the insidious pathological processes behind AD. The association between lithium in drinking water and lower dementia incidence³ and the reduced hippocampal atrophy and other degenerative changes in BPD patients on long-term lithium^4,5 all involve exposures that begin decades earlier than an MCI diagnosis. Beta-amyloid begins accumulating in the brains of people who go on to develop AD roughly 15–20 years before the onset of cognitive symptoms,^37,38 and by the time a patient carries a diagnosis of MCI, substantial neurodegeneration has already occurred. Their brain is riddled with beta-amyloid and aberrant tau; they have lost neurons in key areas of the brain, and others have become disconnected from their networks; their hippocampus has already begun to atrophy.

To really get a handle on diseases of aging as terrible as AD, we would ideally want to intervene before substantial cognitive impairment is apparent, in the long preclinical window when pathology is accumulating but plenty of intact brain structure remains to be rescued. This is exactly what’s happening now with prevention trials for the two approved amyloid-clearing antibodies,^39,40 and there is no scientific reason why it couldn’t be done for lithium once the dose-ranging target-engagement studies we described above are completed.

The real impediments will be financial and logistical. None of this would be fast or cheap. Lithium is generic, and there is no commercial sponsor with a financial stake in proving it works for AD. Thus, the funding burden falls entirely on public sources, and the infrastructure demands of a large multi-site prevention trial are substantial, as Forlenza has noted: “We really need a multicentric study, as we would do with any new pharmaceutical compound. But who will pay for that?”⁴¹

Our takeaway

LATTICE does not close the door on lithium for Alzheimer’s disease. What it does demonstrate is that low-dose lithium carbonate, tested in a mostly amyloid-negative MCI population, does not produce statistically significant improvements in specific cognitive and structural outcomes over two years. That is a far cry from “lithium doesn’t work,” and the distinction matters.

The signal in verbal memory, the larger effect sizes in the amyloid-positive subgroup, the coherent mechanistic framework supplied by decades of preclinical work, the clinical and pharmacodynamic evidence from Forlenza that the biology can be engaged with higher doses,⁶ and the consistent direction across a decades-long body of observational and experimental evidence—none of this constitutes proof of efficacy. But it constitutes a case for continued investigation—and for hope.

The most defensible synthesis of where things stand is this: lithium may have a role in slowing Alzheimer’s disease, but that role may be largely invisible when tested late in the disease course, in people who largely don’t actually have Alzheimer’s-related pathology, at serum levels that may not engage its pharmacodynamic targets, and with a formulation that some provocative mechanistic science suggests is suboptimal. A well-designed next trial—amyloid-positive enrollment via plasma ptau217, dosing based on a pharmacodynamic study, serious assessment of orotate, and ideally intervention earlier in the disease course—would give us the best shot at getting a definitive answer to the questions left open by LATTICE.

Lithium orotate is available over-the-counter as a supplement, but the question of long-term safety in otherwise-healthy older adults at doses likely to be used in a prevention trial remains open. The animal data showed no toxicity except at very high doses, and the pharmacokinetic profile suggests that lower systemic lithium exposure and higher target engagement in the brain can be achieved at lower doses of elemental lithium when using lithium orotate than with the go-to lithium carbonate.

But we need to test all of that in humans. According to one market report, the lithium orotate supplement market was worth $411 million in 2024⁴² and is likely to be growing in light of the Aron et al. Nature paper¹ and the media coverage around it. Many people are self-experimenting with over-the-counter lithium orotate products without medical supervision. This is probably mostly harmless given the low elemental lithium content in standard lithium orotate supplements, but there are adverse event reports related to lithium orotate products in various international regulators’ databases²⁹ and a case report of a woman who showed up in the emergency department with nausea and tremors after taking 18 tablets of lithium orotate, providing about 83 mg of elemental lithium.⁴³

The evidence does not yet support prescribing lithium widely for cognitive preservation, or for self-experimentation beyond the 1-5 mg doses that reflect typical dietary intake.¹¹ But neither does it support writing it off. We need more science, not more supplement ad copy.

For a list of all previous weekly emails, click here. 

podcast | website | ama

References

1. Aron L, Ngian ZK, Qiu C, et al. Lithium deficiency and the onset of Alzheimer’s disease. Nature. 2025;645(8081):712-721.

2. Gildengers AG, Ibrahim TS, Anderson SJ, et al. Low-dose lithium for mild cognitive impairment: A pilot randomized clinical trial: A pilot randomized clinical trial. JAMA Neurol. 2026;83(4):310-319.

3. Fraiha-Pegado J, de Paula VJR, Alotaibi T, Forlenza O, Hajek T. Trace lithium levels in drinking water and risk of dementia: a systematic review. Int J Bipolar Disord. 2024;12(1):32.

4. Velosa J, Delgado A, Finger E, Berk M, Kapczinski F, de Azevedo Cardoso T. Risk of dementia in bipolar disorder and the interplay of lithium: a systematic review and meta-analyses. Acta Psychiatr Scand. 2020;141(6):510-521.

5. Benedetti F, Bollettini I, Barberi I, et al. Lithium and GSK3-β promoter gene variants influence white matter microstructure in bipolar disorder. Neuropsychopharmacology. 2013;38(2):313-327.

6. Forlenza OV, Diniz BS, Radanovic M, Santos FS, Talib LL, Gattaz WF. Disease-modifying properties of long-term lithium treatment for amnestic mild cognitive impairment: randomised controlled trial. Br J Psychiatry. 2011;198(5):351-356.

7. Forlenza OV, Radanovic M, Talib LL, Gattaz WF. Clinical and biological effects of long-term lithium treatment in older adults with amnestic mild cognitive impairment: randomised clinical trial. Br J Psychiatry. 2019;215(5):668-674.

8. Damiano RF, Loureiro JC, Pais MV, et al. Revisiting global cognitive and functional state 13 years after a clinical trial of lithium for mild cognitive impairment. Rev Bras Psiquiatr. 2023;45(1):46-49.

9. Leyhe T, Eschweiler GW, Stransky E, et al. Increase of BDNF serum concentration in lithium treated patients with early Alzheimer’s disease. J Alzheimers Dis. 2009;16(3):649-656.

10. Nunes MA, Viel TA, Buck HS. Microdose lithium treatment stabilized cognitive impairment in patients with Alzheimer’s disease. Curr Alzheimer Res. 2013;10(1):104-107.

11. Schrauzer GN. Lithium: occurrence, dietary intakes, nutritional essentiality. J Am Coll Nutr. 2002;21(1):14-21.

12. Macdonald A, Briggs K, Poppe M, Higgins A, Velayudhan L, Lovestone S. A feasibility and tolerability study of lithium in Alzheimer’s disease. Int J Geriatr Psychiatry. 2008;23(7):704-711.

13. Brinkman SD, Pomara N, Barnett N, Block R, Domino EF, Gershon S. Lithium-induced increases in red blood cell choline and memory performance in Alzheimer-type dementia. Biol Psychiatry. 1984;19(2):157-164.

14. Deliyannides DA, Graff JA, Niño I, et al. Effects of lithium on serum Brain-Derived Neurotrophic Factor in Alzheimer’s patients with agitation. Int J Geriatr Psychiatry. 2023;38(9):e6002.

15. Devanand DP, Crocco E, Forester BP, et al. Low dose lithium treatment of behavioral complications in Alzheimer’s disease: Lit-AD randomized clinical trial. Am J Geriatr Psychiatry. 2022;30(1):32-42.

16. Brzostowicki D. A Verdict on Low-Dose Lithium for Cognitive Preservation? Medscape. May 7, 2026. Accessed July 13, 2026. https://www.medscape.com/viewarticle/verdict-low-dose-lithium-cognitive-preservation-2026a1000esx

17. Yeung MK, Chau AKY, Chiu JYC, Shek JTL, Leung JPY, Wong TCH. Differential and subtype-specific neuroimaging abnormalities in amnestic and nonamnestic mild cognitive impairment: A systematic review and meta-analysis. Ageing Res Rev. 2022;80(101675):101675.

18. Ferman TJ, Smith GE, Kantarci K, et al. Nonamnestic mild cognitive impairment progresses to dementia with Lewy bodies. Neurology. 2013;81(23):2032-2038.

19. Busse A, Hensel A, Gühne U, Angermeyer MC, Riedel-Heller SG. Mild cognitive impairment: long-term course of four clinical subtypes. Neurology. 2006;67(12):2176-2185.

20. Degenhardt EK, Witte MM, Case MG, et al. Florbetapir F18 PET amyloid neuroimaging and characteristics in patients with mild and moderate Alzheimer dementia. Psychosomatics. 2016;57(2):208-216.

21. Salloway S, Sperling R, Fox NC, et al. Two phase 3 trials of bapineuzumab in mild-to-moderate Alzheimer’s disease. N Engl J Med. 2014;370(4):322-333.

22. Sun M, Herrmann N, Shulman KI. Lithium toxicity in older adults: A systematic review of case reports. Clin Drug Investig. 2018;38(3):201-209.

23. Rybakowski JK. Effect of lithium on neurocognitive functioning. Curr Alzheimer Res. 2016;13(8):887-893.

24. Nieper HA. The clinical applications of lithium orotate. A two years study. Agressologie. 1973;14(6):407-411.

25. Kling MA, Manowitz P, Pollack IW. Rat brain and serum lithium concentrations after acute injections of lithium carbonate and orotate. J Pharm Pharmacol. 1978;30(6):368-370.

26. Smith DF, Schou M. Kidney function and lithium concentrations of rats given an injection of lithium orotate or lithium carbonate. J Pharm Pharmacol. 1979;31(3):161-163.

27. Smith DF. Lithium orotate, carbonate and chloride: pharmacokinetics, polyuria in rats. Br J Pharmacol. 1976;56(4):399-402.

28. Pacholko AG, Bekar LK. Different pharmacokinetics of lithium orotate inform why it is more potent, effective, and less toxic than lithium carbonate in a mouse model of mania. J Psychiatr Res. 2023;164:192-201.

29. Murbach TS, Glávits R, Endres JR, et al. A toxicological evaluation of lithium orotate. Regul Toxicol Pharmacol. 2021;124(104973):104973.

30. Morrow C. LiO-AD: Lithium Orotate in Alzheimers Disease Feasibility, Biomarker Engagement, and Clinical Response (LiO-AD). Clinicaltrials.gov. June 8, 2026. Accessed July 8, 2026. https://clinicaltrials.gov/study/NCT07459959

31. Boles S. An Alzheimer’s breakthrough 10 years in the making. Harvard Gazette. January 27, 2026. Accessed July 12, 2026. https://news.harvard.edu/gazette/story/2026/01/an-alzheimers-breakthrough-10-years-in-the-making/

32. Cogswell PM, Lundt ES, Therneau TM, et al. Modeling the temporal evolution of plasma p-tau in relation to amyloid beta and tau PET. Alzheimers Dement. 2024;20(2):1225-1238.

33. Yun J, Lee J, Shin D, et al. Plasma phosphorylated tau 217 cutoffs for amyloid pathology and kidney function, body mass index, and anemia. JAMA Neurol. 2026;83(3):269-279.

34. Kasuga K, Kikuchi M, Kikkawa-Saito E, et al. Evaluation of plasma p-tau217 biomarkers in detecting amyloid pathology and predicting cognitive outcomes: Observations from Japanese Alzheimer’s disease neuroimaging initiative cohort. J Prev Alzheimers Dis. 2026;13(4):100502.

35. Shadyab AH, Zhang B, LaCroix AZ, et al. Plasma phosphorylated tau 217 and incident mild cognitive impairment and dementia in older women. JAMA Netw Open. 2026;9(3):e261295.

36. Buckley RF, Townsend DL, Birkenbihl CJ, et al. The prognostic value of blood-based p-tau217 levels on progression to clinical impairment. medRxiv. Published online February 9, 2026:2026.02.06.26345770. doi:10.64898/2026.02.06.26345770

37. Li Y, Yen D, Hendrix RD, et al. Timing of biomarker changes in sporadic Alzheimer’s disease in estimated years from symptom onset. Ann Neurol. 2024;95(5):951-965.

38. Jack CR Jr, Andrews JS, Beach TG, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimers Dement. 2024;20(8):5143-5169.

39. Yaari R, Holdridge KC, Williamson M, et al. Donanemab in preclinical Alzheimer’s disease: Screening and baseline data from TRAILBLAZER-ALZ 3. Alzheimers Dement. 2025;21(9):e70662.

40. Rafii MS, Sperling RA, Donohue MC, et al. The AHEAD 3-45 Study: Design of a prevention trial for Alzheimer’s disease. Alzheimers Dement. 2023;19(4):1227-1233.

41. Heaton G. Lithium Trial Misses Cognitive Endpoints. Alzforum. https://www.alzforum.org/news/research-news/lithium-trial-misses-cognitive-endpoints. March 5, 2026. Accessed June 13, 2026.

42. Lithium Orotate Supplement Market, Global Outlook and Forecast 2025-2032. 24 Market Reports. July 25, 2025. Accessed June 20, 2026. https://www.24marketreports.com/life-sciences/global-lithium-orotate-supplement-forecast-2025-2032-715

43. Pauzé DK, Brooks DE. Lithium toxicity from an Internet dietary supplement. J Med Toxicol. 2007;3(2):61-62.