Lithium in ground-water and well-being

Lithium is a well-known mood stabilizer & suicide preventive; some research suggests lithium may be a cognitively-protective nutrient and at the population level, chronic lithium consumption (through drinking water) predicts lower levels of mental illness, violence, & suicide.
nootropics, psychology, statistics, meta-analysis, bibliography
2010-10-142019-12-07 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 (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:

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:

  1. most s 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;
  2. the psy­chi­atric-use­ful doses are some­thing like 100x the higher ground­wa­ter dos­es;
  3. 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;
  4. any effects in the pop­u­la­tion should be ~0, and thus nearly impos­si­ble to detect,
  5. but the cor­re­lates are often found, and if causal, would be large reduc­tions in crime/suicide/mental-illness rates / large effects in the pop­u­la­tion;
  6. 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 , 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 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 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 ( at 5%: ~$615b); how­ev­er, this is only the direct expen­di­tures, arguably the field is under­funded (re­port­ed­ly, less than half the suffer­ers of men­tal ill­ness ), 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 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
  • 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 , 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 —much like how the 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 which are >$10.37m, and the life­time costs of dis­eases like schiz­o­phre­nia or the cost of crimes 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 & 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 . 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.

  1. 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…

  2. 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).

  3. 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.↩︎

  4. A mouse study found lithium reduced aggres­sion in one mouse breed: “Effects of nutri­tional lithium defi­ciency on behav­ior in rats”, Klem­fuss & Schi­rauzer 1995↩︎

  5. 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).↩︎

  6. 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”.↩︎

  7. 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.↩︎

  8. Reviews: , Baldessarini et al 2006, & Guzzetta et al 2007, Lewitzka et al 2015.↩︎

  9. 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.↩︎

  10. 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 does­n’t that imply there is an eth­i­cal duty to con­trol cur­rent­ly-nat­u­ral­ly-vary­ing lithium lev­els to even lower lev­els than cur­rently allowed?↩︎

  11. 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.