Suicide is a leading cause of mortality worldwide, responsible for over 800 000 deaths per year, and is the second most common cause of death in people aged 15–29 years.1 In 2018, there were 6507 suicides registered in the UK, an age-standardised rate of 11.2 deaths per 100 000 population, with a male:female ratio of 3:1.2 This male predominance is observed in almost all countries.Reference Bertolote and Fleischman3 People with mental health conditions, especially those with mood disorders such as depression and bipolar disorder (labelled manic–depressive illness until 1980) are at substantially increased risk of suicide.Reference Isometsä4 The prevalence of mental health conditions and national suicide rates are increasing in many countries. Findings from population-based surveys suggest that suicidal ideation, suicide planning and suicide attempts have especially increased over the past 10 years among young people (aged ≤25 years) and that these increases have co-occurred with increasing prevalence of mental health conditions (especially mood disorders such as depression and bipolar disorder) and chronic substance misuse.5 Bipolar disorder affects about 1 in 100 people globallyReference Isometsä4 – without treatment, it can become a relentless cycle of emotional highs and lows.
Lithium as a medication
Since the discovery of its therapeutic effect (and dose) by John Cade (an Australian psychiatrist) in 1949, lithium is widely and effectively used in pharmacological doses in psychiatry for the treatment and prevention of manic and depressive episodes, stabilising mood and reducing the risk of suicide in people with mood disorders.Reference Cipriani, Hawton, Stockton and Geddes6,Reference Lewitzka, Severus, Bauer, Ritter, Müller-Oerlinghausen and Bauer7 Mood disorders are characterised by a 30–50 times increased risk of suicide that can be reduced with lithium treatment. Lithium also appears to have an anti-suicidal effect that might be independent of its mood stabilising property – there is some evidence that it decreases aggression and possibly impulsivity, which might be another mechanism mediating the anti-suicidal effect5. The recommended serum levels for lithium range between 0.6 and 1.0 mmol/L for maintenance therapy of bipolar disorder. However, the optimal blood level at which lithium exerts a possible preventive effect against suicide has not been confirmed and the mechanism of action by which this is achieved is complex and not fully characterised,Reference Malhi, Tanious, Das, Coulston and Berk8 in particular the potential effect of trace doses of lithium is not clear. In one small randomised controlled trial, micro-doses of lithium (400 μg daily) taken by former drug users showed an improvement in mood when compared with placebo,Reference Schrauzer and de Vroey9 suggesting that doses of lithium considerably lower than those generally used in psychiatry may have the potential to influence mood and possibly reduce suicide risk.
Naturally occurring lithium
Lithium, sometimes referred to as the ‘magic ion’, is a naturally occurring element and is found in variable amounts in vegetables, grains, spices and drinking water. It is present in trace amounts in virtually all rocks and is mobilised by weathering into soils, ground and standing water, and thus into the public water supply in varying concentrations.Reference Schrauzer10 In certain areas, particularly those close to sources of briny water such as northern Argentina, concentrations of up to 1000 μg/L have been reported.Reference Concha, Broberg, Grander, Cardozo, Palm and Vahter11 In the first ecological study on the subject, Schrauzer & Shrestha (1990) reported that the average incidence rates of suicide and violence (i.e. homicide and rape) in 27 counties of Texas, USA, over a 10-year period were consistently lower in counties with relatively high natural lithium levels in the drinking water compared with those with medium or low levels. On the basis of these findings, the authors hypothesised that lithium may exert a moderating effect on suicidal and violent criminal behaviour at levels that may be found in public water supplies. Since the publication of this report in 1990, a number of ecological studies from the USA, Japan and Europe have tested the hypothesis that trace doses of naturally occurring lithium in drinking water may have a protective effect against suicide in the general population.
Natural lithium is a mixture of two stable isotopes, lithium-6 and lithium-7. Lithium-7 accounts for over 92% of the natural abundance of the element. The health benefits and curative powers of naturally occurring lithium in water have been known for centuries. For example, the Lithia Springs (in Georgia, USA), an ancient Native American sacred medicinal spring, with its natural lithium-enriched water, is reputed for its health-giving properties and is the source of the brand called Lithia Spring Water. Lithium drinks were so much in demand in the early 20th century that, when the 7-Up commercial drink was created in 1929, it contained the element and was called Bib-Label Lithiated Lemon-Lime Soda. The US Food and Drug Administration banned the use of lithium in soft drinks and beer in 1948 and 7-Up was reformulated. It has been suggested that the 7 in 7-Up referred to the atomic mass of lithium and the Up referred to ‘mood or lithium lift’.
Aims and objective
The objective of the present study was to determine the association between lithium levels/concentration in public drinking water and local/regional suicide rates in the general population. To our knowledge, this is the most comprehensive synthesis of the epidemiological evidence on the subject.
Method
We conducted a systematic review and meta-analysis of the association between naturally occurring lithium content in publicly accessible sources of drinking water and local/regional suicide rates. This report is in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.Reference Moher, Liberati, Tezlaff and Altman12 The study protocol was registered with PROSPERO (CRD42016041375, https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=41375).
Search strategy and selection criteria
We searched four electronic databases (MEDLINE, Embase, Web of Science and PsycINFO), without any language restrictions, for articles published between 1 January 1946 and 10 September 2018. The search terms included: lithium AND drinking water OR public water OR water OR tap water AND suicide OR mortality OR violent OR violence. The full search strategy used for Embase in Ovid is given in supplementary Fig. 1, available at https://doi.org/10.1192/bjp.2020.128. Articles were eligible for inclusion if they were based on original ecological, population-based studies (i.e. analysing aggregate group data defined by geopolitical boundaries rather than individuals) that evaluated the association between lithium levels/concentrations in publicly accessible sources of drinking water and local/regional suicide mortality rates. Articles were excluded if the water samples for lithium measurement were obtained more than 10 years before or after the time period for which suicide rates were measured.
Titles and abstracts of the identified articles were screened by S.F. or I.R. for potentially relevant studies, and the full text was retrieved for articles identified at this screening stage. Two independent reviewers (either S.F. and A.M. or I.R. and A.M.) carried out assessment of the full-text articles for inclusion in the synthesis, and any disagreements were resolved by discussion. The references of all full-text articles and relevant review articles were also searched for additional studies. Authors of ongoing studies that fit the selection criteria were contacted to request unpublished data.
Data extraction and analysis
The relevant data from individual articles were extracted using a preconceived and standardised data extraction form. Information extracted included: first author's name, year of publication, country and region, size of the population studied, methodology of lithium sampling and laboratory analysis, average lithium levels/concentrations in drinking water, mean total suicide mortality rate and/or standardised mortality ratio (SMR), time period of the suicide data, list of covariates that were adjusted for in the analysis, and outcome measure (statistical methods and effect size and s.e.). Corresponding authors were contacted for any data not presented in the published article. For two of the included studies,Reference Palmer, Cates and Gorman13,Reference Kabacs, Memon, Obinwa, Stochl and Perez14 standardised regression coefficients were not included in the published articles, but were calculated from data supplied by the authors. Where only the standardised regression coefficient and P-value or t-value were reported, the s.e. of the standardised regression coefficient was estimated using the method of Altman and BlandReference Altman and Bland15 (s.e. calculated by this method were adjusted for one studyReference Sugawara, Yasui-Furukori, Ishii, Iwata and Terao16 to account for the small sample size).
Quality assessment of included studies
To assess the quality of epidemiological ecological studies, we adapted the checklists produced by Tu & KoReference Tu and Ko17 and Betran et alReference Betran, Torloni, Zhang, Ye, Mikolajczyk and Deneux-Tharaux18 to give five evaluation criteria, which were used to assess each study (supplementary Table 1). The evaluation criteria assessed whether the study participants were representative of the conclusions being drawn, whether statistical methods were used appropriately, whether confounders were adjusted for appropriately, whether key elements of the study design were presented and justified, and discussion of limitations such as the ecological fallacy. Quality assessments were performed by two independent reviewers (either S.F. and A.M. or I.R. and A.M.) and any discrepancies resolved by consensus. The results of the quality assessment were not used to decide on inclusion or exclusion of studies.
Data synthesis
Although we acknowledge the concerns with pooling non-randomised study designs,Reference Reeves, Deeks, Higgins, Wells, Higgins and Sally19 where the comparisons were deemed reasonably homogeneous the standardised regression coefficients and their corresponding s.e. were pooled using random-effects meta-analyses. Heterogeneity was assessed with I 2 and Cochran's Q. Separate meta-analyses were conducted for total, male and female suicide rates. Where both adjusted and unadjusted regression coefficients were presented, the unadjusted regression coefficient was used in preference. The exposure to lithium was determined according to the lithium levels/concentration in drinking water, or the log lithium level; and the outcome variable was either a standardised regression coefficient of SMR or suicide mortality rate. All analyses were carried out in R 3.5.0 (run on the Windows platform)20 using the metafor package.Reference Khazaei, Armanmehr, Nematollahi, Rezaeian and Khazaei21 Subgroup analyses were conducted to assess heterogeneity according to whether the study took place in a country where the mean age-adjusted total suicide mortality rate in the World Health Organization (WHO) 2015 data22 was either ≥10 per 100 000 (high suicide rate countries, i.e. Japan, Austria, Lithuania and the USA), or <10 per 100 000 (low suicide rate countries, i.e. Greece, Italy and the UK), and according to the highest observed lithium levels in the sampled drinking water (≥80 μg/L or <80 μg/L). Quality assessment was also used to explain heterogeneity between studies. In the primary meta-analyses (all eligible studies of total, male and female suicide rates), we also performed a sensitivity analysis to evaluate robustness and stability by sequentially omitting one study at a time.
Results
A flowchart describing the study selection process is given in Fig. 1. The literature search identified 415 articles – after exclusion of duplicate titles and abstracts, 260 articles were screened and 41 were selected for full-text evaluation. Of these, 15 articles fulfilled the selection criteria and were included in the synthesis, 4 studies were conducted in Japan, 4 in Austria, 3 in USA, and 1 each in England, Greece, Italy and Lithuania (including lithium levels in drinking water samples and corresponding suicide mortality rates for 1286 regions/counties/cities).
Fig. 1 PRISMA flow diagram for study selection.
In the included ecological studies, drinking water samples were collected from publicly available tap-water sources in a variety of locations or from public wells. Most water samples were analysed by mass spectrometry or inductively coupled plasma emission spectroscopy. Three studies used water supply company's data on water composition. Eleven studies controlled for one or more confounding variables in the analyses; and six studies controlled for a measure of socioeconomic status (e.g. proportion of population with a college degree, unemployment rate, average income). A variety of other possible confounders were adjusted for in different studies, including measures of local climate, altitude and density of medical professionals. Suicide mortality data were obtained from government statistics in all the studies and covered time periods of between 1 yearReference Ishii, Terao, Araki, Kohno, Mizokami and Shiotsuki23 and 11 years.Reference Giotakos, Nisianakis, Tsouvelas and Giakalou24,Reference Pompili, Vichi, Dinelli, Pycha, Valera and Albanese25 All studies included the overall suicide mortality rate across all age groups in the geographical areas covered, 14 studies presented data on total suicide rates, and 10 each presented data on male and female suicide rates. All but three studiesReference Kabacs, Memon, Obinwa, Stochl and Perez14,Reference Giotakos, Nisianakis, Tsouvelas and Giakalou24,Reference Schrauzer and Shrestha26 obtained lithium levels/concentration data over a range of years that overlapped with the population-based aggregate suicide mortality data. The number of drinking-water samples taken ranged from 22 (ref.Reference Liaugaudaite, Mickuviene, Raskauskiene, Naginiene and Sher27) to 6460 (ref.Reference Kapusta, Mossaheb, Etzersdorfer, Hlavin, Thau and Willeit28). The mean lithium levels in the drinking-water samples ranged from 3.8 μg/L (ref.Reference Shiotsuki, Terao, Ishii, Takeuchi, Kuroda and Kohno29) to 46.3 μg/L (ref.Reference Bluml, Regier, Hlavin, Rockett, Konig and Vyssoki30). The total study populations ranged from 1 109 261 (ref.Reference Liaugaudaite, Mickuviene, Raskauskiene, Naginiene and Sher27) to 22 097 948 (ref.Reference Bluml, Regier, Hlavin, Rockett, Konig and Vyssoki30) and the total suicide mortality rate per 100 000 per year ranged from 7.53 (ref.Reference Pompili, Vichi, Dinelli, Pycha, Valera and Albanese25) to 27 (ref.Reference Liaugaudaite, Mickuviene, Raskauskiene, Naginiene and Sher27).
Study quality
Of the nine studies included in the meta-analysis, three fulfilled all five of the evaluation criteria, four fulfilled four of the criteria, one fulfilled three and one fulfilled two. Failure to adjust for covariates was the most common methodological omission. Studies fulfilling at least four of the criteria were considered to be of high quality. Details of the quality assessment are given in supplementary Table 1.
Meta-analysis of association between lithium levels in drinking water and suicide mortality rates
The studies included in the meta-analysis are summarised in Table 1. Eight studies that reported the association between lithium levels in drinking water and suicide mortality as a standardised regression coefficient of either SMR or suicide rates on log lithium levels were eligible for inclusion in the meta-analysis. One further studyReference Giotakos, Nisianakis, Tsouvelas and Giakalou24 reported the standardised regression coefficient of age-standardised suicide mortality rate on untransformed lithium levels – the meta-analysis was repeated with and without the inclusion of this study. Four of the articles identified in the literature searchReference Kapusta, Mossaheb, Etzersdorfer, Hlavin, Thau and Willeit28,Reference Helbich, Leitner and Kapusta31–Reference Helbich, Leitner and Kapusta33 were multiple analyses carried out on the same Austrian data – only the original studyReference Kapusta, Mossaheb, Etzersdorfer, Hlavin, Thau and Willeit28 was included in the meta-analysis. Three studies conducted in Japan had overlapping drinking-water sample areasReference Ishii, Terao, Araki, Kohno, Mizokami and Shiotsuki23,Reference Shiotsuki, Terao, Ishii, Takeuchi, Kuroda and Kohno29,Reference Ohgami, Terao, Shiotsuki, Ishii and Iwata34 – only the study with the largest sample sizeReference Shiotsuki, Terao, Ishii, Takeuchi, Kuroda and Kohno29 was included in the meta-analysis. One study examined suicide rates over three 10-year periodsReference Pompili, Vichi, Dinelli, Pycha, Valera and Albanese25 – only the results for the last period were included in the meta-analysis, as this overlapped with the time when the drinking-water samples were collected. Two of the included studies were conducted in Japan, two in the USA and one each in England, Austria, Greece, Italy and Lithuania. A range of different population sizes and municipalities were evaluated, from entire countries to individual provinces and prefectures. Two studiesReference Liaugaudaite, Mickuviene, Raskauskiene, Naginiene and Sher27,Reference Shiotsuki, Terao, Ishii, Takeuchi, Kuroda and Kohno29 excluded suicides that were not registered in cities, otherwise all suicides in the regions studied were included.
Table 1 Characteristics of studies included in the systematic review and meta-analysis
n.r., not reported; SMR, standardised mortality ratio; GP, general practitioner; ICP-OES, inductively coupled plasma optical emission spectrometry; PWLS, population-weighted least squares; T, total (i.e. both genders combined); M, male; F, female; Li, lithium.
a. Personal communication from Nikolett Kabacs.
b. Calculated from data supplied by Nikolett Kabacs.
c. Personal communication from Nestor Kapusta.
d. SMR has reference 1.00 not 100.
e. s.e. from personal communication from Vilma Liaugaudaite.
f. Calculated from data supplied by Greg Gorman.
Total (i.e. both genders combined) suicide mortality rate
Eight eligible studies examined the association between lithium levels in drinking water and total suicide mortality rates – seven of these studies reported a protective (i.e. inverse or negative) association between lithium levels and total suicide rates, which was statistically significant in five of the seven studies (Fig. 2). The random-effects meta-analysis showed a statistically significant protective (or inverse) association between lithium levels and total suicide rates (pooled β = −0.27, 95% CI −0.47 to −0.08; P = 0.006, I2 = 83.3%). To assess the impact of relatively high heterogeneity between the studies, we conducted a sensitivity analysis. The heterogeneity was reduced most on excluding the study by Liaugaudaite et al,Reference Liaugaudaite, Mickuviene, Raskauskiene, Naginiene and Sher27 and the meta-analysis yielded a pooled β of −0.17 (95% CI −0.25 to −0.09; P<0.0001, I 2 = <0.01%). The results of one studyReference Giotakos, Nisianakis, Tsouvelas and Giakalou24 were derived from a standardised regression of SMR on untransformed lithium values rather than log lithium – omitting this study had no effect on the results (pooled β = −0.29, 95% CI −0.52 to −0.06; P = 0.01, I 2 = 85%). We obtained a similar result (β = −0.25, 95% CI −0.50 to −0.01; P = 0.04, I 2 = 86.0%) when we repeated the meta-analysis including only the studies considered to be of high quality.
Fig. 2 Random-effects meta-analysis of the association between lithium concentration in drinking water and total suicide rates.
Male suicide mortality rate
Seven eligible studies examined the association between lithium levels in drinking water and male suicide mortality rates – five of these studies reported a protective (or inverse) association between lithium levels and male suicide rates, which was statistically significant in three of the five studies (Fig. 3). On the other hand, two studies reported a non-significant positive association. The random-effects meta-analysis showed a protective (or inverse), but statistically non-significant, association between lithium levels and male suicide rates (pooled β = −0.26, 95% CI −0.56 to 0.03; P = 0.08, I 2 = 91.9%). In the sensitivity analyses, the heterogeneity was reduced most on excluding the study by Liaugaudaite et al,Reference Liaugaudaite, Mickuviene, Raskauskiene, Naginiene and Sher27 and the meta-analysis yielded a pooled β of −0.12 (95% CI 0.28–0.03; P = 0.13, I 2 = 64.0%). We obtained a similar result (β = −0.21, 95% CI −0.53 to 0.10; P = 0.19, I 2 = 93.0%) when we repeated the meta-analysis including only the studies considered to be of high quality.
Fig. 3 Random-effects meta-analysis of the association between lithium concentration in drinking water and male suicide rates.
Female suicide mortality rate
Seven eligible studies examined the association between lithium levels in drinking water and female suicide mortality rates – five of these studies reported a protective (or inverse) association between lithium levels and female suicide rates, which was statistically significant in two of the five studies (Fig. 4). On the other hand, two studies reported a non-significant positive association. The random-effects meta-analysis showed a statistically significant protective (or inverse) association between lithium levels and female suicide rates (pooled β = −0.13, 95% CI −0.24 to −0.02; P = 0.03, I 2 = 28.5%). In the sensitivity analyses, the heterogeneity was reduced most on excluding the study by Shiotsuki et al,Reference Shiotsuki, Terao, Ishii, Takeuchi, Kuroda and Kohno29 and the meta-analysis yielded a pooled β of −0.17 (95% CI −0.28 to −0.07; P = 0.001, I 2 = 0.02%). We obtained a similar result (pooled β = −0.11, 95% CI −0.22 to −0.001; P = 0.05, I 2 = 26.0%) when we repeated the meta-analysis including only the studies considered to be of high quality.
Fig. 4 Random-effects meta-analysis of the association between lithium concentration in drinking water and female suicide rates.
Meta-analysis by higher/lower total suicide mortality rates, according to the WHO data
The random-effects meta-analysis, including five studies from countries with higher suicide rates, showed a statistically significant protective (or inverse) association between lithium levels and total suicide rates (pooled β = −0.40, 95% CI −0.68 to −0.12; P = 0.005, I2 = 86.0%); including only the three studies from countries with lower suicide rates yielded a pooled β of −0.11 (95% CI −0.23 to 0.01; P = 0.08, I2 = 0.0%), with no heterogeneity (supplementary Fig. 2).
Meta-analysis by higher/lower observed lithium levels in drinking water
There were three studies of total suicide mortality rate where the highest observed lithium level in drinking water was ≥80 μg/L (supplementary Fig. 3). The random-effects meta-analysis showed a statistically significant protective (or inverse) association with a pooled β of −0.20 (95% CI −0.31 to −0.10; P = 0.0002, I 2 = 0.0%), with no heterogeneity. For the five studies conducted in regions with lithium levels <80 μg/L, the pooled β was −0.33 (95% CI −0.68 to 0.01; P = 0.06, I 2 = 87.0%).
Articles included in the narrative synthesis only
Six articles provided an estimate of the association between lithium levels in drinking water and suicide mortality rates but were not eligible for inclusion in the meta-analysis (supplementary Table 2). Three were additional analyses/reports on the Austrian data-set, which controlled for a variety of additional confounding variables, including altitude and rates of lithium prescriptions. Two studies from JapanReference Ishii, Terao, Araki, Kohno, Mizokami and Shiotsuki23,Reference Ohgami, Terao, Shiotsuki, Ishii and Iwata34 were conducted in areas that overlapped with the study area for Shiotsuki et al.Reference Shiotsuki, Terao, Ishii, Takeuchi, Kuroda and Kohno29 There was also a study that compared suicide rates in Texas, USA, in counties with relatively high, medium or low lithium concentrations in drinking water.Reference Schrauzer and Shrestha26 All of these studies found statistically significant protective (or inverse) associations between lithium levels and total suicide rates; where data on gender-specific rates were presented, protective associations were also found with male suicide rates but not with female suicide rates.
Discussion
This is the first meta-analysis of the ecological association between the lithium levels/concentration in publicly available drinking water and the incidence of suicide. We found a consistent protective (or inverse) association between lithium levels and total, male and female suicide mortality rates, which was statistically significant for total and female suicide rates. Similar protective association was observed in the six studies included in the narrative synthesis, and in the subgroup meta-analyses based on the higher/lower suicide mortality rates (≥10 per 100 000, <10 per 100 000) and lithium levels (≥80 μg/L, <80 μg/L). These findings, which are consistent with the finding in clinical trials that lithium reduces suicide and related behaviours in people with a mood disorder, suggest that naturally occurring lithium in drinking water may have the potential to reduce the risk of suicide and may possibly help in mood stabilisation, particularly in populations with relatively high suicide rates and geographical areas with a greater range of lithium concentration in the drinking water.
Strengths and limitations
To our knowledge, this is the most comprehensive synthesis of the ecological association between lithium levels/concentrations in publicly available drinking water and suicide mortality rates. The main limitations of epidemiological ecological studies are that they are observational and are subject to the ecological fallacy (or ecological bias). Ecological studies are essentially conducted to generate hypotheses. They compare aggregate exposure (e.g. population exposed to lithium in drinking water) and disease/health outcome (e.g. suicide mortality rate in the exposed population) across different populations over the same time period or within the same population over time. They are subject to confounding as information on potential confounder(s) may not be available and associations at the population level do not necessarily represent associations at the individual level (ecological fallacy). Populations may also differ in terms of ethnic, religious and social class distribution, prevalence and management of mental disorders, and mobility patterns. As with all systematic reviews and meta-analyses, our study may also be subject to reporting/publication bias. These biases arise when the dissemination of research findings is influenced by the nature and direction of results – statistically significant ‘positive’ results are more likely to be published and cited, whereas non-statistically significant results may be filtered, manipulated or presented in such a way that they become/seem positive. We were unable to conduct a formal assessment of possible publication bias, as our meta-analyses did not meet a key assumption (i.e. a minimum of 10 studies) for the tests of funnel plot asymmetry.
Although most of the studies were assessed to be of good quality using our adapted criteria, and our results were unaltered on excluding the two studies judged to be of lower quality, no standardised criteria for quality assessment of ecological studies were available. The studies varied widely in the range of years for which the suicide rates were measured, and in the number and timing of drinking-water samples taken. Some of the studies utilised drinking-water samples and suicide data from different time periods, thus relying on the assumption that lithium levels do not fluctuate substantially over time. In a study from Japan, Ohgami et alReference Ohgami, Terao, Shiotsuki, Ishii and Iwata34 found negligible change in lithium levels in drinking water when they repeated the measurements after 1 year. A study from Denmark also found little variability over time in lithium levels collected from ground water between 1947 and 2012.Reference Knudsen, Schullehner, Hansen, Jorgensen, Kristiansen and Voutchkova35 However, additional evidence on stability over time of lithium levels in drinking water is needed from other geographical areas.
In a nationwide closed (or non-dynamic) historical (or retrospective) cohort study published in 2017, Knudsen et alReference Knudsen, Schullehner, Hansen, Jorgensen, Kristiansen and Voutchkova35 linked individual-level register-based data on the entire Danish adult population (3.7 million individuals) from 1991 to 2012 with a moving 5-year time-weighted average lithium exposure level from drinking water. Cases of suicides were identified through the Danish Register of Causes of Death and lithium levels in drinking water were ascertained from 158 water samples obtained from 151 public waterworks supplying approximately 42% of the Danish population. Of these, 139 samples were collected via a drinking-water sampling campaign during April–June 2013 (spatially covering the entire country) and 19 samples were collected via a separate campaign at the Greater Copenhagen Utility between October 2009 and June 2010. The lithium levels were measured at a single point in time (i.e. cross-sectional) and it was assumed that they had remained constant over the 22-year study period (from 1991 to 2012). The mean lithium level in the drinking-water samples was 11.6 μg/L (range 0.6–30.7 μg/L). The overall national suicide rate decreased by 38% during the study period – from 29.7 per 100 000 in 1991 to 18.4 per 100 000 in 2012. The spatial regression analysis was adjusted for confounding factors (including gender, age, employment/civil status and calendar year). The study found no statistically significant association between increasing 5-year time-weighted average lithium levels and decreasing suicide rate – although all the incidence rate ratios were ≤1. The authors concluded that there did not seem to be a protective effect of exposure to lithium on the incidence of suicide with levels below 31 μg/L in drinking water. They also noted that in the previous (i.e. ecological) studies that found a significant protective association, the lithium exposure levels were relatively much higher than those found in their study, and the lack of variation in lithium levels in their study may have ‘challenged’ their analyses. The authors acknowledged that the study had several limitations, as it was based on a single cross-sectional measurement of lithium levels from 151 waterworks supplying approximately half of the Danish population; and suggested that future studies could prospectively determine lithium levels from more waterworks on regular basis for a number of years. It is, nevertheless, challenging that the only published study linking exposure to lithium in drinking water with suicide on an individual level was negative.
It has been suggested that the association between lithium in drinking water and suicide might be modified by the rate of lithium prescriptions in the population, as excreted lithium might find its way into the water supply. However, Helbich et alReference Helbich, Bluml, Leitner and Kapusta32 found no evidence that the rate of lithium prescriptions in an area affected the association between lithium levels in drinking water and suicide rates (either directly or via an effect on the mean lithium level in the water).Reference Helbich, Leitner and Kapusta33 It is also noteworthy that some vegetables, grains/nuts, meat and spices are also rich in lithium, but it is difficult to measure dietary lithium as the content is variable and relates to lithium content in ground water/soil. It is therefore likely that lithium intake from food will be relatively higher in areas that have high lithium content in both ground water and the public water supply.Reference Schrauzer10,Reference Figueroa, Razmillic, Zumeata, Aranda, Barton and Schull36,Reference Gonzalez-Weller, Rubio, Gutierrez, Gonzalez, Caballero Mesa and Revert Girones37 It has been estimated that the average daily intake of lithium from food by adults in the USA ranges from 650 to 3100 μg, whereas in areas of Texas, which have unusually high lithium levels in the public water supply, the contribution from drinking water might be around 340 μg.Reference Schrauzer10 The association between exposure to dietary lithium and incidence of suicide has not been investigated. Furthermore, bottled drinking water (processed/treated or natural mineral water from springs) often has a much higher lithium content than tap water – the association between exposure to lithium via bottled water and suicide has not been studied.
Possible mechanism(s) behind the association between lithium in drinking water and suicide
A possible mechanism by which lithium in drinking water might prevent suicide is by ameliorating the symptoms of mood disorders. In a randomised controlled trial involving former drug users a 400 μg daily dose of lithium was found to improve mood, with peak effect after 4 weeks of administration.Reference Schrauzer and de Vroey9 There may also be a cumulative effect of sustained low-dose exposure over longer time periods, as lithium levels in drinking water have been shown to correlate with serum levels of lithium in areas of high drinking-water lithium concentrations.Reference Harari, Åkesson, Casimiro, Lu and Vahter38 There is little information on the association between lithium in drinking water and the prevalence and severity of mood disorders. In a population-based nested case–control study from Denmark, higher long-term lithium exposure from drinking water was not associated with a lower incidence of mania/bipolar disorder – suggesting that long-term exposure to micro-doses of lithium does not modulate the risk of these conditions.Reference Kessing, Gerds, Knudsen, Jørgensen, Kristiansen and Voutchkova39 On the other hand, in another population-based nested case–control study by the same research group, higher long-term lithium exposure from drinking water was associated with a lower incidence of dementia in a non-linear way.Reference Kessing, Gerds, Knudsen, Jørgensen, Kristiansen and Voutchkova40
It is also possible that trace doses of lithium might reduce suicide rates via its anti-aggressive effects. Therapeutic/pharmacological doses of lithium have been found to reduce aggressive/violent behaviour in a variety of populations,Reference Muller-Oerlinghausen and Lewitzka41 and inverse associations have been observed between lithium levels in drinking water and rates of violent crimes.Reference Schrauzer and Shrestha26,Reference Giotakos, Tsouvelas, Nisianakis, Giakalou, Lavdas and Tsiamitas42 Similarly, a recent cross-sectional study of adolescents in Kochi prefecture in Japan found an inverse association between the lithium content of drinking water available to schools and interpersonal violence and depressive symptoms among adolescents.Reference Ando, Koike, Shimodera, Fujito, Sawada and Terao43 As violent methods of suicide are more likely to be lethal, lithium ingestion might be expected to reduce the lethality of suicide attempts. This was observed in a recent meta-analysis of the effects of long-term lithium treatment of people with major affective disorder, where the incidence ratio of attempted:completed suicides increased 2.5-fold with lithium treatment, indicating a considerably reduced lethality.Reference Baldessarini, Tondo, Davis, Pompili, Goodwin and Hennen44 The meta-analysis also showed that the risks of attempted and completed suicide were consistently lower (by about 80%) in people with bipolar and other major affective disorders treated with pharmacological doses of lithium for an average of 18 months.
Recommendations for future research
Given that our results suggest a possible protective (or inverse) association between lithium levels in drinking water and suicide mortality at the population level, randomised community trials of lithium supplementation of the water supply might be a possible means of testing the hypothesis, particularly in communities (or settings) with demonstrated high prevalence of mental health conditions, violent criminal behaviour, drug dependency and chronic substance misuse and risk of suicide. This may provide further evidence to support the hypothesis that lithium could be used at the community/population level to reduce or combat the risk of these conditions. It may also be possible to measure lithium levels in people who report to emergency departments with self-harm/suicidal ideation and at inquest for suicide. The synthesised evidence suggests that the protective (or inverse) association between lithium levels in drinking water and suicide mortality rates is likely to be stronger in populations with relatively higher suicide mortality rates. Future studies might benefit by concentrating on areas with a wide range of lithium levels/concentration in drinking water and relatively high suicide mortality rates and would ideally also consider the effect and levels of dietary lithium from food sources.
Supplementary material
Supplementary material is available online at http://doi.org/10.1192/bjp.2020.128.
Data availability
The data that support the findings of this study are available from the corresponding author (A.M.) on reasonable request. The corresponding author had full access to all the data in the study.
Author contributions
A.M. conceived and designed the study. I.R., S.F. and A.M. did the literature search and quality assessment of included studies. I.R. and S.F. selected the studies and extracted the relevant information. I.R. and A.M. synthesised the data. A.M. and I.R. wrote the manuscript. B.C., R.S., D.H.-M. and A.H.Y. critically reviewed the manuscript for intellectual content. All authors approved the final version of the manuscript for publication. A.M. guided and supervised the overall work, and had final responsibility for the decision to submit for publication.
Funding
This work was supported by Brighton and Sussex Medical School (BSMS) and part-funded by the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London. The views expressed are those of the author(s) and not necessarily those of the above sponsors. The study sponsors had no role in the study design, in the collection, analysis and interpretation of data, or in the writing of the report and the decision to submit the paper for publication.
Declaration of interest
A.H.Y. has received payment for lectures and advisory board membership from AstraZeneca, Eli Lilly, Lundbeck, Sunovion, Servier, Livanova and Janssen; he is the lead investigator for the Embolden Study (AZ), BCI Neuroplasticity Study and Aripiprazole Mania Study, and has led investigator-initiated studies from AstraZeneca, Eli Lilly, Lundbeck, Wyeth and Janssen. R.S. has received an honorarium for speaking at a Lundbeck-sponsored event.
ICMJE forms are in the supplementary material, available online at https://doi.org/10.1192/bjp.2020.128.
eLetters
No eLetters have been published for this article.