Dietary Lithium: A critique of the claimed effect on depression and suicide
Many research groups continue to publish papers claiming to suggest relationships between (inter tot alia) depression, suicide, aggressive behaviour, and dementia, with the (ultra-low) levels of lithium (Li) in reticulated drinking-water supplies (municipal ‘tap-water’). The notion of adding Li to drinking water supplies has been mooted, written about in medical journals, and covered in the mass media. However, Li in tap-water is only a small proportion of the total dietary Li intake. There are no data in these papers addressing the crucial question of whether differences even exist in the serum levels of Li, and 24 hr Li excretion, between subjects from the different areas that have different Li tap-water levels. This commentary clarifies these fundamental omissions and logical errors with relevant data from the scientific literature which have not been cited previously in this context. Such data clarify key facts about Li intake, and the regulation of the usual, but very low, serum Li levels that result from usual dietary intake. These data indicate clearly and unequivocally that such ‘associative epidemiology’ is conceptually flawed, methodologically unsound, and over-interpreted. It constitutes another example of science that would never have been published if editorial and refereeing standards of journals were adequate.
There are many publications that can be described as ‘associative epidemiology’ trying to link lithium (Li) in reticulated tap-water with various aspects of nervous system functioning, such as depression, suicide, aggressive behaviour, dementia, and onychoschizia.
I refer to these studies as ‘associative epidemiology’ because the scientific and logical observations and requirements that might indicate even a slight chance of an actual mechanism and causal link are in some parts weak, and in other parts absent. The old trope of ‘association is not causation’ endures because each new generation of researchers repeat the same mistakes, with dispiriting and tedious regularity. This expanding area of investigation is an example of ignorance of basic science, of methodology, and of misunderstanding of statistics, especially P-values . A Bayesian would want to think about the ‘prior probability’ and require rather better evidence (particularly, a mechanism detailing a plausible causal link). See:
I have previously discussed concepts related to establishing cause-and-effect relationships (the Bradford Hill ‘criteria’, et seq.), which are not well-observed in psychiatry . The above link contains a summary about this — my previous discussion  was related to an analysis of the weak cause-effect nexus between neuroleptic drugs and neuroleptic malignant syndrome (NMS).
The principles enumerated by Sir Austin Bradford Hill, that are often referred to as the Bradford Hill criteria* (NB, he never suggested they should be treated as criteria), have been used simplistically. They have been progressively refined [4-6], especially by the distinguished work on causality by Judea Pearl [7, 8], for which he received the Turing award, which I have discussed elsewhere.
*Incidentally, although he is frequently referred to as ‘Sir Austin Bradford Hill’, which seems to have made many people think he had a double-barrelled surname, he was in fact generally referred to as Tony Hill — so I suppose his name should formally be expressed as Sir Austin Hill, but since most of the literature refers to him as Bradford Hill I tend to follow that convention to avoid confusion.
This commentary argues that there are fundamental misunderstandings in the published studies concerning the ingestion and regulation of the usual trace-amounts (<0.5 mg/day) of dietary Li intake in humans, and the sources of the ingested Li. It discusses how these errors and misconceptions indicate that the epidemiological studies thus far are based on incorrect data and invalid assumptions. Their inferences and conclusions are unsustainable.
Lithium: Trace levels and their regulation
Serum levels: the best replicated and most recent data indicate that typical human serum Li levels (i.e., serum levels arising from obligate-ingestion from food and water, sometimes referred to as ‘endogenous’) are around 0.0003 mmol/L and show low variation in different populations (~ 0.0001 -0.0005) in relation to dietary ingestion amounts.
Dietary intake: current best estimates of dietary daily intake are 0.1-0.5 mg/day, of which only ~35% is from ‘water’ (see below).
The therapeutic oral-dose range for Li, when used to treat bipolar disorder, is approximately 500 to 1,500 mg daily (of Li2CO3), producing serum levels of approximately 0.4 – 1.2 mmol/L.
In clinical psychiatry and clinical chemistry human serum Li levels are usually given in mmol/L — one mmol equals 0.7 mg/100ml.
Most people probably think that, in those not taking any form of Li treatment, or supplement, there is no Li the body. That is almost true. It is, of course, a question of limits of detection/quantification (LODs/LOQs): usual clinical chemistry laboratory methods are insufficiently sensitive to measure Li levels below 0.1 mmol/L.
The measurement of nanogram levels (10,000 times lower than treatment levels) of serum Li (and other trace metals), and Li in biological tissues and fluids, involves technical challenges and difficulties. Some older reported measurements are probably less accurate. The present benchmark measurement technique is inductively coupled plasma mass spectrometry (IC-PMS); for recent reviews see Meermann  and Heitland . Previous reviews include [11-13].
Fractional Li excretion
The use of the measurement of fractional Li excretion in hypertension research — made possible by the reduced difficulty and cost of IC-PMS — which involves measuring nanogram levels of serum Li and 24 hr. urinary Li excretion (resulting from trace levels of Li normally present in food and excreted in the urine), has therefore expanded over the last couple of decades.
These fractional Li excretion studies have produced a large set of patient samples of serum Li levels in humans — well in excess of 2,000 subjects from all over the globe — these are subjects whose source of Li is their usual diet. As a result of their studies, the group at the Studies Coordination Centre, Laboratory of Hypertension, Leuven, Belgium [14, 15], have concluded; ‘These observations suggest that serum lithium is tightly regulated (at ~0.0003 mmol/l), despite large variations in dietary intake’. All subjects also had 24 hr urine Li estimated, which had, in the Belgium sample, an average of 5 µmol of Li, equating to a daily intake of ~0.035 mg of Li, and in a South African sample 1 µmol, equating to a daily intake of ~0.007 mg: the mean serum Li levels in these two groups were almost identical, and within a narrow range.
Cwynar et al. found almost identical serum Li results in 130 subjects in Poland .
Heitland  measured a mean serum Li of 0.0009 mmol/l in 102 subjects in Germany.
Folkerd  found a mean serum Li of 0.00027 +/- 0.02 mmol/l (n = 25, range 0.00013-0.00055 mmol/l).
New data on dietary lithium intake
Li intake can be calculated from the 24-hour urine samples; therefore data in hypertension studies also represent the largest body of evidence directly measuring (rather than estimating) typical dietary Li intake, and they indicate an intake of 0.05 – 0.5 mg/day): the best estimate previously available (derived by estimating the amount of lithium in food and beverages consumed) was the French total diet study (TDS)  which gave an estimate of 0.05 mg daily.
The estimate of Li intake from 24-hour Li urine assays does assume a steady state condition, which may not always be the case. Nevertheless, it seems reasonable to assume that the majority of the subjects in such studies will have been ingesting Li at a more or less constant rate, therefore the steady-state assumption is a valid approximation. However, in particular studies, such as that by Shiotsuku, below, such subjects would definitely not be in steady state.
Incidentally, out of 1278 subjects (Bochud) 36 were excluded because of a ‘very high Li concentration in serum (>0.001 mmol/L)’ which they considered indicated ‘external contamination’ (i.e., ingestion of supplements).
In view of the size of the above sample, discussion of other results (especially those using older assay methods), is somewhat superfluous. However, Miller found serum Li levels around 0.00016 mmol/L for subjects dwelling in the Denver metropolitan area. The mean 24-hr excretion rate was 0.005 mmol/day , identical to the value found by the Leuven group.
At usual levels of dietary intake (~0.05 – 0.5 mg/day), which are now rather more reliably established from the above hypertension studies, serum Li appears to be tightly regulated between about 0.0001-0.0003 mmol/L. Serum Li is not proportional to daily intake in this range.
Relevance to research protocols
These data are informative vis-à-vis how significant scientific mistakes compromise epidemiology research, probably to the point of making it invalid — all the epidemiological studies have accounted only for Li in tap-water and not for Li from other sources (which comes from outside the area in which the statistics on suicide etc. are being considered).
This also applies to interventional studies such as Nunes et al.; they gave 0.3 mg/day to try to treat subjects with dementia. Since this is within the normal dietary intake range, it is a misconceived trial, since it is unlikely to have significantly altered pre-existing serum Li levels .
Shiotsuku’s study  of subjects drinking large amounts of Li-rich mineral-water at a spa, although methodologically weak and lacking key data, seems to be the only study with serum Li estimations in subjects (n=43) ingesting Li in doses of around 20-50 mg/day — despite the poor description and methodology the serum Li averaged 0.07 mmol/L, which is ~200 times higher than usual dietary serum Li levels.
These data indicate that the tight regulation of serum Li breaks down once intake reaches around 5-20 mg/day, when it becomes proportional to the daily dose, as it is when Li is used therapeutically for BPD at doses of hundreds of mg per day.
Tap-water: A small proportion of daily intake
The above data indicate that serum Li is closely regulated so that usual dietary Li intake, below ~1 mg/day, will not much affect serum Li levels.
The best recent data has produced a figure for daily Li intake of 0.05 – 0.5 mg/day. In the most extensive study of its kind, the French Total Diet Study, water contributed only 35% of total dietary Li intake , say ~ 0.01 mg. Moreover, a substantial percentage of that water was not ‘tap-water’ — it was bottled water which contains a higher Li concentration (which, of course, does not come from the area where the subject lives), see below.
NB, I corresponded with the authors of the French study who confirmed that they have the data showing what percentage of water was mineral water, but they did not present that data — the data they presented was the total of both tap-water and mineral water.
These epidemiological studies all relate exclusively to average levels in the reticulated water supply (viz. tap-water) — none of them take account of Li from other sources. Thus, tap-water constitutes only a minor proportion of typical daily dietary Li intake; the remainder of Li intake will come from foodstuffs and liquids neither of which originate in the area where the statistics are gathered.
It is reasonable to deduce from these more recent and more reliable data that tap-water constitutes only about 10% of total daily Li intake in a large proportion people sampled in studies.
The influence of bottled water
The consumption of bottled ‘spring-water’ is a multi-billion-dollar industry throughout the western world and bottled waters have a mean of five times higher average Li concentrations: viz. median level of bottled waters is 0.010 mg/L vs. 002 mg/L for tap-water [22-26]*. Sales figures indicate that about one thousand bottles of ‘mineral’ water are sold per person per year (USA: 30 billion bottles in 2008). In Europe the consumption figure varies between 100-200 L per person/year (in 2019 the EU average was 118 l/yr). Source (European Federation of Bottled Waters): http://efbw.eu/bwf.php?classement=07 .
*Lithium comes from igneous pegmatite-type rocks: it is present in greater concentrations when the water has been in contact with such rocks for longer periods of time. That is generally the case in deep aquifers — thus where such rocks exist, the deeper the aquifer, and the longer the water has been in the aquifer, the higher the Li concentration is likely to be. Therefore, deep spring waters tend to have more lithium, whereas shallow spring waters generally have less. Surface waters have least of all. The exception among surface waters being mountain waters that run over exposed pegmatite gravels in places like the Andes; however, such high Li-content surface water gets rapidly diluted as it joins the larger streams and rivers that flow into the lowlands, from where town supplies are generally drawn.
The extensive data that now exist on Li levels in reticulated water (which usually comes from surface water, rather than deep subterranean water) is detailed in the references given herein.
It is notable that few of ‘psychiatric’ publications are aware of these data — the seminal citations herein (from the geological survey organisations in Europe and the USA) do not appear in the ‘Psych’ literature.
Levels in surface water are generally extremely low, usually a few thousandths of a mg/L (<0.002 mg/L) and often as low as millionths of a mg/L (0.000,005 mg/L) [22-24, 27], but levels in bottled mineral water from springs and underground supplies can be quite high: this has been assessed at 884 different European sites which had a median level of 0.010 mg/L. Thus, a typical bottled mineral water is equivalent to 5 L of tap-water. That indicates even low levels of consumption of bottled water would substantially alter the Li intake from ‘water’. A large proportion of subjects live in areas where the concentration in tap-water is only a few millionths of mg/L, and they therefore do not ingest a significant amount of Li from tap-water.
These data demonstrate that it is certain a substantial proportion of the subjects in these samples will have been ingesting more Li from bottled mineral/spring-water than from tap-water.
That is, with little doubt, enough to invalidate and nullify the statistical analyses and inferences in all of this published ‘epidemiological’ data.
In summary, there are fatal factual flaws invalidating and nullifying all research concerning endogenous Li in tap-water and its effect on humans.
First, at normal ‘obligate’ levels of dietary intake (0.05-0.5 mg per day) serum Li is not proportional to intake, rather, it is regulated within a narrow range.
Second, even if the total ingested amount of Li did make a difference to serum Li concentrations, it would not be as a result of varying Li levels in the reticulated tap-water supply, because that is only ~10% of the total dietary Li intake.
Third, the failure to measure serum levels of Li in a sample of subjects in different areas (exposed to different levels of Li in their tap-water) and to (attempt to) establish a correlation between the two variables, is an obligatory research step, the omission of which fatally weakens any remaining vestige of credence in a possible causal link.
The crowning irony is that researchers did not even need to do those measurements for themselves, because, as explained above, it has already been done, in more than 1,000 subjects [14, 15]. But none of these publications appear to have recognized that fact.
One hardly needs to comment on the perspicacity of reviewers who approved the publication of these papers, or the grants for these projects, or the fact that it is likely few, if any, of these papers underwent expert statistical review. Some of these projects/papers must have cost a great deal of money. They have certainly wasted a great deal of time, of the researchers, of the people involved in the publication process, and of the readers.
Therefore, trying to correlate Li in tap-water with depression and suicide, or any other state or condition, in different geographical areas with different water supplies is, a priori, without scientific basis.
Further publications about Lithium and suicide
There are now so many publications (subsequent to my initial posted comment a few years ago) about Li and depression etc. that is has become pointless to list them all — I refer readers to the meta-analyses mentioned, which give the references [28-30].
The only critical and negative study I recall seeing was Parker et al. . Pompili et al. was neutral .
Sequelae: Public health proposals
The benefits of Li have been taken so seriously that there are now several meta-analyses of these studies [28-30] and a number of publications, some from eminent academics e.g., [33, 34] that discuss adding Li to public drinking water supplies: also there have been many media articles, e.g., New York Times; Should We All Take a Bit of Lithium? .
These will have the ethicists and the social conspiracy theorists in hot debate for the foreseeable future (cf. fluoride). Considering the extremely poor quality of the research and the weak association claimed, pontificating about this can only make psychiatry look amateurish and foolish, yet again.
It is important that a more informed and critical analysis of the doubtful logic and rationality of such research is recognised and promulgated.
1. Greenland, S., et al., Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. Eur J Epidemiol, 2016. 31(4): p. 337-50.
2. Phillips, C.V. and K.J. Goodman, The missed lessons of Sir Austin Bradford Hill. Epidemiol Perspect Innov, 2004. 1(1): p. 3.
3. Gillman, P.K., Neuroleptic Malignant Syndrome: Mechanisms, Interactions and Causality. Movement Disorders, 2010. 25(12): p. 1780-1790.
4. Hernan, M.A. and J.M. Robins, Estimating causal effects from epidemiological data. J Epidemiol Community Health, 2006. 60(7): p. 578-86.
5. Muthen, B. and H.C. Brown, Estimating drug effects in the presence of placebo response: causal inference using growth mixture modeling. Stat Med, 2009. 28(27): p. 3363-85.
6. Maldonado, G. and S. Greenland, Estimating causal effects. Int J Epidemiol, 2002. 31(2): p. 422-9.
7. Pearl, J., M. Glymour, and N.P. Jewell, Causal inference in statistics: A primer. 2016: John Wiley & Sons.
8. Pearl, J., Reasoning with cause and effect. AI Magazine, 2002. 23(1): p. 95-95.
9. Meermann, B. and V. Nischwitz, ICP-MS for the analysis at the nanoscale–a tutorial review. Journal of Analytical Atomic Spectrometry, 2018. 33(9): p. 1432-1468.
10. Heitland, P. and H.D. Koster, Human biomonitoring of 73 elements in blood, serum, erythrocytes and urine. J Trace Elem Med Biol, 2021. 64: p. 126706.
11. Subramanian, K.S., Determination of metals in biofluids and tissues: sample preparation methods for atomic spectroscopic techniques. Spectrochimica Acta Part B: Atomic Spectroscopy, 1996. 51(3): p. 291-319.
12. Lu, Y., et al., Alkali dilution of blood samples for high throughput ICP-MS analysis—comparison with acid digestion. Clinical Biochemistry, 2015. 48(3): p. 140-147.
13. Clarke, W.B., et al., Endogenous lithium and boron red cell-plasma ratios: normal subjects versus bipolar patients not on lithium therapy. Biol Trace Elem Res, 2004. 97(2): p. 105-16.
14. Seidlerova, J., et al., Association between arterial properties and renal sodium handling in a general population. Hypertension, 2006. 48(4): p. 609-15.
15. Bochud, M., et al., Context dependency of serum and urinary lithium: implications for measurement of proximal sodium reabsorption. Hypertension, 2007. 49(5): p. e34.
16. Cwynar, M., et al., Blood pressure, arterial stiffness and endogenous lithium clearance in relation to AGTR1 A1166C and AGTR2 G1675A gene polymorphisms. J Renin Angiotensin Aldosterone Syst, 2016. 17(2): p. 1470320316655669.
17. Kalonji, E., et al., Nutritional Risk Assessment of Eleven Minerals and Trace Elements: Prevalence of Inadequate and Excessive Intakes from the Second French Total Diet Study. European Journal of Nutrition & Food Safety, 2015. 5(4): p. 281-296.
18. Miller, N.L., J.A. Durr, and A.C. Alfrey, Measurement of endogenous lithium levels in serum and urine by electrothermal atomic absorption spectrometry: a method with potential clinical applications. Anal Biochem, 1989. 182(2): p. 245-9.
19. Folkerd, E., et al., Clearance of endogenous lithium in humans: altered dietary salt intake and comparison with exogenous lithium clearance. Am J Physiol, 1995. 268(4 Pt 2): p. F718-22.
20. Nunes, M.A., T.A. Viel, and H.S. Buck, Microdose lithium treatment stabilized cognitive impairment in patients with Alzheimers disease. Curr Alzheimer Res, 2012.
21. Shiotsuki, I., et al., Drinking spring water and lithium absorption: a preliminary study. German Journal of Psychiatry, 2008. 11: p. 103-106.
22. Krachler, M. and W. Shotyk, Trace and ultratrace metals in bottled waters: Survey of sources worldwide and comparison with refillable metal bottles. Science of The Total Environment, 2009. 407: p. 1089–1096.
23. Demetriades, A., C. Reimann, and M. Birke, European Ground Water Geochemistry Using Bottled Water as a Sampling Medium in Clean Soil and Safe Water. 2012, Springer Netherlands: Dordrecht. p. 115-139.
24. Reimann, C. and B. M, Geochemistry of European bottled water. 268 pp. Available online at:
http://www.schweizerbart.de/publications/detail/artno/001201002. 2010, Stuttgart: Borntraeger Science Publishers.
25. Levei, E., M. Hoaghia, and R. Savastru, Quality assessment of Romanian bottled mineral water and tap water. Environmental monitoring and assessment, 2016. 188(9): p. 521.
26. González-Weller, D., et al., Dietary intake of barium, bismuth, chromium, lithium, and strontium in a Spanish population (Canary Islands, Spain). Food and chemical toxicology, 2013. 62: p. 856-868.
27. Salminen, R., et al., Geochemical atlas of Europe. Part 1 – Background information, methodology and maps. Geological Survey of Finland, Espoo, Finland, 2005: p. http://www.gtk.fi /.
28. Memon, A., et al., Association between naturally occurring lithium in drinking water and suicide rates: systematic review and meta-analysis of ecological studies. Br J Psychiatry, 2020. 217(6): p. 667-678.
29. Eyre-Watt, B., et al., The association between lithium in drinking water and neuropsychiatric outcomes: A systematic review and meta-analysis from across 2678 regions containing 113 million people. Aust N Z J Psychiatry, 2021. 55(2): p. 139-152.
30. Del Matto, L., et al., Lithium and suicide prevention in mood disorders and in the general population: A systematic review. Neurosci Biobehav Rev, 2020. 116: p. 142-153.
31. Parker, W.F., et al., Association Between Groundwater Lithium and the Diagnosis of Bipolar Disorder and Dementia in the United States. JAMA Psychiatry, 2018.
32. Pompili, M., et al., Relationships of local lithium concentrations in drinking water to regional suicide rates in Italy. World J Biol Psychiatry, 2015: p. 1-8.
33. McGrath, J.J. and M. Berk, Could Lithium in Drinking Water Reduce the Incidence of Dementia? JAMA Psychiatry, 2017. 74(10): p. 983-984.
34. Ghaemi, N., Lithium: The gift that keeps on giving in psychiatry. Medscape. available from: https://www.medscape.
35. Fels, A., Should we all take a bit of Lithium? New York Times. New York: SR6., 2014.