Lithium is a well-known mood stabilizer & suicide preventive; some research suggests lithium may be a cognitively-protective nutrient and on population levels chronic lithium consumption through drinking water predicts mental illness, violence, & suicide
14 October 2010–7 Dec 2019 · in progress · certainty: unlikely · importance: 9
The metal lithium is a well-known mood stabilizer & suicide preventive widely used in psychiatry. It is also a trace mineral present to various levels in all drinking water and much food. A long-running but obscure vein of research speculates on whether lithium is beneficial and a nutrient, specifically, cognitively-protective. Epidemiological research has correlated chronic lithium consumption through drinking water with a number of population-level variables like rates of mental illness, violence, & suicide. If causal, lithium should be regarded as a vital nutrient for mental health and added to drinking water to substantially improve population-wide outcomes.
However, the evidence is weak. Most of this research is cross-sectional, only some is longitudinal, none offers particularly strong causal evidence using natural experiments or other designs, there are questions about confounding with autocorrelated spatial properties such as altitude, and some of the best research, using Scandinavian population registries, offers more mixed evaluations of claimed correlates.
It is unlikely that further such correlational research will resolve the debate, despite the mounting opportunity cost. I suggest that formal experimentation is required, and concerns about harms from lithium supplementation making experiments ‘unethical’ can be circumvented by instead removing lithium or looking for natural experiments with cause changes (such as changes or upgrades to water treatment plants or plumbing modify lithium concentration).
The elemental metal lithium (review; FDA adverse events) is a pretty unusual substance and like caffeine and the amphetamines, questionably classified as a nootropic. As a metal, lithium is dangerous at many doses. It’s famously used for manic-depression and some other disorders, but the doses are large and verge on the point where ‘the cure is worse than the disease’. Most lithium research focuses on these larger doses, so one has to parse citations carefully to see whether it is telling one something useful about the low levels one might use for supplements or just reinforcing what one already knew (‘large doses are double-edged swords’).
So, on the positive side:
Discover’s “The Metal Marvel That Has Mended Brains for 50 Years” covers some of the neurogenesis effects of lithium (see also Shiotsuki 2008), which may be related to its possible antioxidant effect
in mice, lithium helps prevent Parkinson’s
The results on dementia are also mixed: Terao et al 2006 found patients prescribed lithium had less, while Dunn et al 2005 had found the exact opposite. A Danish population registry study found a correlation in Denmark of groundwater lithium with Alzheimer’s & dementia (Kessing et al 2017b).
Tsaltas et al 2008 (see also Vo et al 2015), is a review of lithium studies. There may be some long-term benefits related to the neuroprotective effects12, which counterbalance the negative or null effects discussed.
natural drinking water levels have been correlated with
decreased crime3 4 5, & Giotakos et al 2015; but Dawson’s data was criticized as confounded with distance by Pokorny et al 1972, and lithium/suicide seems related to & possibly confounded by altitude (Helbich et al 2013/Helbich et al 2015 found lithium decreased with altitude in Austria and affected suicide rates; but Huber et al 2014 finds in the USA that lithium increases with altitude) and altitude itself has suggested to cause suicide/mental illness. Gonzalez et al 2008, as expected, found r=-0.35, but its sample (1 year of violent crime in 11 counties) was too small to make that correlation statistically-significant; the larger Servello 2008 and Blüml et al 2013 tried to correct for the criticisms & found a statistically-significant reduction in Texan suicides. This is especially interesting given reduced lithium in criminals6 These studies are reviewed in Mauer et al 2014, Vita et al 2014, Mischley 2014 (media op-ed), & Barjasteh-Askari et al 2020.
decreased suicides in
- Japan: Ohgami et al 2009 (see discussion in Terao et al 2009), Sugawara et al 2013 (criticized in Chandra & Babu 2009; see also Schöpfer & Schrauzer 2011’s finding of unmeasurably low lithium levels in many Tokyoites’ hair) & Ishii et al 2015 & Shiotsuki et al 2016
- Austria: Kapusta et al 2011/Helbich et al 2012/Helbich et al 2015
- Greece: Giotakos et al 2013
- Chile: König et al 2017
- Lithuania: Liaugaudaite et al 2017/Liaugaudaite et al 2019
- Alabama, USA: Palmer et al 2018
- possibly Portugal, although Oliveira et al 2019 finds no correlation
- but not quite in England: Kabacs et al 2011
- nor in Italy: Pompili et al 20157
- nor in Denmark, in probably the largest & best correlational study yet and strong evidence against there being a correlation: Knudsen et al 2017 (apparently Knudsen et al 2016 didn’t pan out)
This correlation would make sense given that lithium prevents suicide in patients with mood disorders8
on the other hand, a 1975 survey of Maryland counties found no correlation between lithium & mood
and increased schizophrenia/bipolar disorder in Denmark (Schullehner et al 2019)
weakly correlates with fewer self-reported mental health problems in Japanese students (Ando et al 2017/Shimodera et al 2018) and weakly predicts depressive symptoms (Ishii et al 2017), extraversion along with latitude (Matsuzaki et al 2017); causes improved mood in former drug-users and visitors to Japanese springs (Shiotsuki 2008); a lithium-water experiment in 2012 was terminated for logistical reasons.
One of the main problems with inferring that lithium causes these reductions is that it seems difficult to reconcile with how large the doses must be to treat mental illness:
- most dose-response relationships tend to have relatively simple curves which look like U-curves or V-curves or straight lines, where the effect diminishes fast when you move away from the best dose;
- the psychiatric-useful doses are something like 100x the higher groundwater doses;
- so for most such curves, if the peak is at X mg, then a dose at X/100 or X/1000 mg will do little;
- any effects in the population should be ~0, and thus nearly impossible to detect,
- but the correlates are often found, and if causal, would be large reductions in crime/suicide/mental-illness rates / large effects in the population;
- 4 & 5 seem to be contradictory.
The best responses seem to be that either lithium’s effects diminish quite gradually so that small groundwater doses can still have a meaningful population effect (negate #1), that groundwater doses are more effective than one would expect comparing to psychiatric doses of lithium carbonate (perhaps due to chronic lifelong exposure; negate relevance of #2/3), or that lithium may have multiple mechanisms one of which kicks in at psychiatric dose levels and the other at groundwater levels (somewhat supported by some psychiatric observations that depressives seem to benefit from lower doses but in different ways; negate #1 in a different way).
Ken Gillman, echoing the earlier criticisms of the Ohgami et al 2009 correlation by Chandra & Babu 2009, criticizes the correlations as generally invalid due to the smallness of the drinking water dose compared to the dietary doses of lithium; I disagree inasmuch as lithium doses are cumulative, Schrauzer 2002 reports an FDA estimate of daily American lithium consumption 1mg, points out that natural levels can reach as high as 0.34mg via drinking water, Dawson 1991 finds Texans’ lithium excretion to “vary inversely with rainfall, reflecting the dilution of drinking water supplies” (see also Dawson 1970), and Dawson et al 1972 directly compared lithium levels in county waters with urine lithium measurements and found a clean linear relationship as expected. These points suggest strongly that Gillman is wrong to think that consumption of bottled water or imported vegetables would swamp any contribution from drinking water - random noise cancels out, and small correlations can be detected using very large samples like state or nation level samples. (One commenter has suggested that the darkness caused by rain is what increases suicide rates, not the dilution of natural lithium in the drinking water; this seems unlikely as it would be inconsistent with the known peaking of suicide rates in spring when there is little darkness, and with the lack of correlation of life satisfaction with daily weather.)
The criticisms of the trace lithium correlation seem weak to me, and even keeping in mind the meta results that the overwhelming majority of correlations disappear when experimentally tested (a 1% chance is a reasonable guess at the true a priori odds), the potential benefits seem so overwhelming that I am puzzled that, in the 42+ years since the correlation was first noted, no one has done a simple experiment of randomizing some counties and increasing their trace lithium concentrations within the normal range of natural variations in trace lithium concentrations. A quick estimate for what I mean. In Dawson et al 1972, the counties in the highest lithium category had 30% of the mental hospital admissions that the lowest lithium counties did; the USA spends something like $60b annually on mental health issues (the NYT quotes $150b government expenditure & $500b society-wide costs). If the reduction in admissions was equivalent to a reduction in the underlying disorders and expenditures scale exactly with number of disorders, then lithizing the USA would reap gains of something like $35b annually (NPV at 5%: ~$615b); however, this is only the direct expenditures, arguably the field is underfunded (reportedly, less than half the sufferers of mental illness may be treated), and of course the losses to society is far larger than that as peoples’ lives are destroyed, crime increases (with its massive negative externalities), careers abandoned, etc. At 1% prior odds and $615b total payoff, the expected value is ~$600m; a conclusive experiment ought to be trivially cheap to run (a few millions?), since it requires only supplementation of lithium at centralized water facilities, presents minimal ethical concerns due to remaining strictly within natural variation & regulatory limits10 (and vastly less than people voluntarily consume in bottled mineral waters), and can be analyzed using only statistics already being collected by police or health departments.
(The potential gain is large enough that, even if one objected that we don’t know for sure that psychiatric lithium does not impede creativity and so it is hypothetically possible that lithization would reduce societal creativity, the benefit probably outweighs the costs. An example: suppose lithization cost us one Nikola Tesla a year; as it happens, an academic researcher can be funded for life with ~$4m (fully loaded cost: ~$200k I’ve read (some relevant figures), and a ~20y career - assuming no outside earnings or grants or payments), and using the previous savings of $30b, one could fully fund 7500 people each year to research and work on whatever they want for life; is it plausible that one potential Tesla outweighs 7500 independent effectively-tenured researchers?)
On the null and negative sides:
- although the lithium-Parkinson’s research with its relatively low doses is a reminder to avoid lithium doses anywhere near what is used for psychiatric disorders: “In at least two known cases, toxic levels of the drug have actually caused Parkinson’s.”
- In one small human trial (149 experimental, 447 total selected from >1000) on PatientsLikeMe investigating a 2008 paper’s finding that lithium might delay ALS, only the null effect was found.
- There are serious negative effects to taking a lot of lithium - 2-4 grams will trash your long-term memory and similar doses have been linked with many cognitive issues.
- Tsaltas et al 2008 says many studies can’t be generalized to healthy populations; for every study finding damage to performance or memory, there seems to be a study finding the null result. But whichever is true, it is not encouraging11.
- Of 22 Alzheimer’s patients taking 100mg of lithium carbonate, 3 stopped due to side-effects (Macdonald et al 2008); but 100mg carbonate is substantially more than 5mg orotate, and one might guess that old ill people would report more side-effects in general
- Broberg et al 2011 documents extreme daily consumption of 2-30mg from water & food in Andean villages, correlating with changes in some thyroid-related biomarkers, suggestive of potential thyroid issues
- Aprahamian et al 2014 was an RCT of 61 patients with “mild cognitive impairment” given ~150mg of lithium carbonate (targeting blood level of 0.25-0.5 mmol/L); it found minimal damage to liver function, but a few worrisome secondary outcomes such as somewhat higher complaints of side-effects (4.07 vs 4.98 symptoms)
- Kessing et al 2017 finds no correlation between bipolar disorder & lithium in a population registry study of Denmark (despite bipolar disorder being one of the primary uses for psychiatric doses of lithium)
The cost of doing lithium supplementation, like many other kinds of supplementation such as iodization, is minimal in terms of raw materials. A back of the envelope suggests that the cost in 2018 of USA-wide lithium would be <$75m for the raw materials, and the current cost less than half that. So the true cost is almost surely dominated by however much work it is to actually add said materials.
Lithium is an element commercially mined at scale (something like 40,000 tons/year), and even with escalating demand, costs are still in the thousands of dollars per ton. Claims of ‘lithium shortages’ are relative and more of a concern for car manufacturers using potentially hundreds of kilograms per car (who will have to invest heavily in efficiency & sourcing lithium from new mines to resolve any shortages—much like how the rare earth element embargo quickly resulted in them being neither rare nor particularly embargoed) than supplementation. This should be unsurprising, as the levels correlated with benefits are reachable in many places by drinking ordinary untreated water, after all: so either there is a lot of lithium out there, the necessary levels are tiny, or both.
Using Schrauzer as an example of what it’d take, Schrauzer defines ‘high’ as a concentration of >70μg /l (ie 70 millionths of a gram per liter). The USGS estimates total use of all water for any purpose in the USA at 322 billion gallons/day, of which ~43 billion gallons/day or 162 billion liters/day are ‘domestic’+‘public supply’.
Assuming the worst case, that all of this must be supplemented and that all of it has 0μg /liter lithium and must be supplemented by 70μg/liter, then the total lithium required per day is kilograms, or 4,141,935 kilograms per year, or 4,565 tons per year.
At its 2018 peak, which triggered alarmism over shortages, high-quality battery-grade lithium cost ~$16500/ton (it has since fallen as low as $7000 in 2019), so the total cost per year comes to $75,322,500; or to put it another way, [$0.23]($2018)/year per capita. For further perspective, the US federal government uses values of life which are >$10.37m, and the lifetime costs of diseases like schizophrenia or the cost of crimes like murder typically come in at orders of magnitude around $1m or higher.
This assumes lithium prices don’t fall further, that all water must be fully supplemented, etc, and is a loose upper bound. A more reasonable cost would use long-range forecasts extrapolating with experience curve & economies of scale, the ability to use the lowest-quality & cheapest lithium sources (minute differences in purity or type are critical to batteries but not supplementation) and target water supplies most likely to be drunk, and realistic estimates of how much water actually needs to be supplemented; it will likely be a small fraction of [$0.23]($2018)/year, and quite possibly down in the <[$0.05]($2018) range. (Just using a more recent price per ton would more than halve the estimate, after all.)
And of course, just because the existing epidemiological evidence focuses on drinking water (due to the relative ease of measurement) doesn’t mean that supplementing drinking water is the only or best way to do it. Great flexibility is possible. Given the small quantities involved, many other approaches could be cheaper: delivery via salt like iodine, fortification of flour like iron, fortification of milk like vitamin D, long-lasting injections like iodine or vitamin D for cases where regular supplementation is infeasible, addition to multi-vitamins like vitamin A or the B vitamins…
So, cost is not a major objection to lithium supplementation.
From Tsaltas 2008, pg 15:
Tentative conclusions from studies in normal subjects are that acute lithium does not affect short-term memory; subchronic administration spares basic short-term memory of ongoing events but higher task demands (as in neuropsychological testing) occasionally reveal mild deficits. As do learning deficits, these too appear transient. A similar picture emerges with respect to lithium effects on human long-term recall. In animal studies, subchronic and chronic lithium with clinically relevant serum levels does not affect spatial reference or object recognition memory and actually enhances working memory under certain conditions. This is consistent with recent clinical MRI findings noting improved immediate verbal memory after a 4-year period of lithium treatment, along with MRI evidence of increased hippocampal volume over the same period.
Human attention is quite consistently reported normal under lithium. Some older animal studies report narrowing of attention onto high-salience cues and compromised latent inhibition, but these results are challenged by more recent data indicating normal function. Finally, information on lithium effects on executive functions is sparse and cannot be evaluated at present. More basic research is definitely needed with respect to lithium effects on attention and executive functions. Recent reports on lithium effects on the cognitive-behavioral deficits induced by various challenges to the nervous system in animal models are quite promising. Lithium protects against neuroanatomical and neurochemical effects and also moderates cognitive deficits induced by stress or CNS trauma such as irradiation or anoxia. In some cases, such deficits are not simply prevented but appear to be reversed post facto. This combined neuroprotective- and cognitive-enhancing action of lithium is noted primarily with respect to hippocampally related spatial memory tasks. It appears to involve protection against the reduced cell proliferation and increased apoptotic rate noted mainly in the hippocampus under these challenges…
Tsaltas 2008, continued:
A similar picture emerges in relation to lithium effects on the cognitive compromises induced by neurodegenerative disorders. Lithium reduces the prevalence of Alzheimer’s disease in bipolar patients, and there is evidence suggesting that this is associated with reduced GSK-3beta expression. Evidence of lithium’s moderating action on hippocampally related cognitive deficits also comes from transgenic animal models of Alzheimer’s disease…In cognitive dysfunction associated with psychiatric conditions, beneficial effects of lithium have emerged on the neuroanatomical level from imaging studies. Lithium treatment of bipolar patients has been associated with hippocampal volume increase and appears to entail concomitant cognitive improvements. These neuroimaging findings are not limited to bipolar patients, but involve people at ultrahigh risk of developing a psychotic disorder where lithium appears to arrest neuroanatomical and neurochemical changes associated with the onset of psychosis. In conclusion, increasing neuroanatomical and neurochemical evidence from both in vitro and in vivo studies supports that lithium has neuroprotective properties, mainly involving hippocampal cells (Moore et al. 2000; Manji et al. 2001; Sassi et al. 2002; Kim et al. 2004; Chuang 2004; Chuang and Priller 2006).
A closely-related Texas study found similar inverse correlations for mental hospitals, but I haven’t been able to find fulltext: “The relationship of tap water and physiological levels of lithium to mental hospital admission and homicide in Texas”, Dawson, in Lithium in Biology and Medicine 1991. An odd result is decreased lithium levels in autistic children and also their mothers, Adams et al 2006.↩︎
A mouse study found lithium reduced aggression in one mouse breed: “Effects of nutritional lithium deficiency on behavior in rats”, Klemfuss & Schirauzer 1995↩︎
Two earlier studies: “The mathematical relationship of drinking water lithium and rainfall to mental hospital admission”, “Relationship of lithium metabolism to mental hospital admission and homicide” (which used directly measured lithium levels via urine samples).↩︎
Informally, the chemist William Walsh found by 1983 that among his other results from hair analysis, lithium was low in his studied inmates as well. Some studies cite an unpublished Walsh manuscript, “Chemical imbalance and criminal violence: results of two controlled studies of California institutions”, held at the “Health Res. Institute, Chicago”.↩︎
Kabacs et al 2011 turns in the expected negative point value (r = -0.03), but it is not statistically-significant. It’s not clear how much of a counter-example this is; the correlation is simple without any adjustments for other factors (like Kapusta et al 2011 did), and there’s an issue of range restriction: the authors write “In the East of England, there was relatively little variation in population size across the 47 subdivisions. Also, the lithium levels in drinking water in Texas and in the Oita prefecture ranged from 0 to 160 μg/l, and 0.7 to 59 μg/l, respectively. These values represent a much wider range and higher top level than those found in the East of England (<1-21 μg/l).” Pompili has a similar probable power issue: they don’t reach statistical-significance overall with their 145 sites but they do in one subgroup and the overall relationships are always inverse as predicted (in the 3 decades, the overall r was -0.081, -0.099, & -0.039); suggesting to me that here again we may have an issue of insufficient power to detect a fairly small effect and that a meta-analysis may confirm the correlation.↩︎
Fajardo et al 2018 notes that the all-cause mortality reduction loses statistical-significance after controlling for suicide rates, but do not do a formal mediation test.↩︎
A randomized experiment could be conducted either by adding additional lithium to drinking water, remaining below the regulatory limits, or alternately, by preventing increases in lithium levels somehow. The former is probably much easier in practice. Ethically, there should be no problem: if it’s unethical to add any lithium to water (anywhere up to the FDA-approved safe levels), then by symmetry & the reversal test doesn’t that imply there is an ethical duty to control currently-naturally-varying lithium levels to even lower levels than currently allowed?↩︎
Tsaltas et al 2008, pg 15:
The effects of lithium on learning in clinical populations appear to be mildly detrimental, possibly attributable to lithium’s generalized dampening effect on performance. They appear most pronounced in the initial stages of lithium administration, as corroborated by animal studies. Therefore, results produced by subchronic regimes should be treated cautiously, as perhaps reflecting general influences on arousal and mood.