1 Örebro University
School of Medicine Degree Project, 15 ECTS May 2017
Lithium in drinking water and suicide: a
systematic review of the evidence
Version
21
Author: Axel Wänghammar
Supervisor: Dr. Mats Humble
Adjunct Lecturer/Örebro University School of Medical Sciences
2
Abstract
Introduction: Suicide is worldwide a major public health issue. Research has shown that lithium has preventive capabilities on suicide and even on a trace level, it can be therapeutic on other mental health issues. Multiple studies have tried to find an inverse association between suicide and high trace-levels of lithium in tap water.
Objective: This paper aimed to be a systematic review over such researches.
Method: Two databases (PubMed and PsychInfo) were searched through to find articles matching the criteria. Results were presented in complementing text and table-forms. Results: 12 articles matched the criteria. 9 of them had a majority of their outcome indicators showing a significant association; 2 partially so and 1 in none.
Conclusion: Most evidence suggests that the negative association is present, but the heterogeneity of the studies and the lack of essential considerations in many of them makes it hard to conclude such a connection is present. More needs to be done to shed light on the lack of knowledge around lithium and its effect in a low dose, as well.
3
Contents
1. INTRODUCTION………4
1.1 Low-dose lithium and its association with suicide….……….………4
1.2 Lithium…….………...4
2 OBJECTIVE…….…….………5
2.1 Aim………….………..5
2.2 Hypothesis………6
3. METHOD………...6
4. RESULTS………...7
5. DISCUSSION………..…….15
6. CONCLUSION…...………..17
7. ACKNOWLEDGEMENTS………..17
8. REFERENCES……….………18
4 1.1 Low-dose lithium and its association with suicide
Suicide is one of the largest issues that the healthcare systems worldwide faces, especially among men and in the developed world [1]. The World Health Organization estimates that a total of 800,000 suicides are committed annually [2]. To confront this major issue, states all over the world are running preventive programs with different approaches. In the field of medicine, the anti-suicidal effect of lithium [3] is something that has increasingly gained interest of researchers. Even though most of the evidence on lithium’s effect has been gained through research on its properties at a clinical dose concentration, the proposition that it can have a substantial health impact on a trace-dose level has garnered substantial momentum. In 1970, Dawson and others conducted a study [4] in which they found significant associations between re/admissions into psychiatric care hospitals (among patients with mental disorders) and concentrations of lithium in regional tap water; an inverse relationship. A 2014 review published in the Australian and New Zealand Journal of Psychiatry by researchers Mauer, Vergne and Ghaemi [5] compiled evidence on what possible benefits there are with a low-dose exposure to lithium (more specifically, in terms of prevention of dementia, and other psychiatric and medical indications). In the study, suicide was one of the outcomes on which lithium seemed to have a beneficial action. Multiple studies have further investigated the effect of trace-dose lithium on the outcome suicide isolated, with the latest systematic review of the evidence having been processed in early 2014 and published the following year [6].
1.2 Lithium
While lithium may be one of the most classical and widely used psychiatric medicines to this day, there is still a lot of uncertainty regarding its mechanisms of actions to be elucidated. Currently, there are a few postulated mechanisms with differing degrees of evidence supporting them. It has been proposed that the effect of lithium is on G-protein coupled receptors and on secondary messengers downstream (the phosphoinositide or adenylyl systems, for example) [7, 8]. Also, there are speculations considering the potential inhibiting effect lithium may have on glycogen synthase kinase-3 (GSK-3) [9], with one of its functions being to phosphorylate the protein tau (an often implicated "culprit" in the development of Alzheimer's disease). Another interesting
5 revelation has come through animal tests, which have shown an elevated grade of the
anti-apoptotic protein Bcl-2 [10]. To summarize, there seems to be indications that lithium has a neuroprotective and mood-stabilizing impact on neurons that would partly explain its
effectiveness as a psychotropic medicine.
With the mechanisms at a cell and molecule level in mind, a few speculations have arisen on how lithium acts to prevent suicidal behavior. According to the GSK-3-based theory, the inhibiting effect of lithium works indirectly as an anti-inflammatory substance. Increased levels of stress-related cytokines, mediating inflammation has been proven to be connected to an amplification of risk factors for suicide: impulsivity, aggression and depression (11). To further strengthen the connection between physiological stress and suicide: the administration of inflammation factors for therapeutic reasons has resulted in an increase in suicidal risk in a few published cases (12, 13, 14). On a larger level, we also find lithium being modulatory on an anatomical and physiological level in the brain, with anti-suicidal effects. It seems to have a partial effect in decreasing the "bottom-up" drive that promotes aggressive actions and increasing the "top-down" deactivation path, acting (15). Lithium increases the volume of the preorbital cortex and anterior cingulate gyrus (part of the "top-down" function). The limbic system (of which the "bottom-up" system is consisted of) is also altered by lithium when it comes to activity and size. Tests on male rats haves also shown that the metal element has a decreasing effect on testosterone levels, which may very well be applicable on humans (16). Higher levels of the hormone could possibly increase the risk of the aggressive act that is suicide.
2. Objective
2.1 AimThe aim of this study is to give an updated overview on the evidence regarding the hypothesis that low-dose lithium in drinking water has an inverse association with suicide rate. Since the last systematic review (Vita et al, 2015) was executed, there have been plenty of more studies published that is of interest to include in a systematic review.
The study populations aimed to be covered in this work are the normal populations in the given countries, divided into different districts (municipalities, cities, prefectures, etc.) to be able to draw comparisons. It can range from the entire population of a country to a small subdivision of
6 it. Data on both genders combined and not separated is preferred, but the alternative is acceptable as well. No active intervention is to be observed in the studies: the intention is to include studies who have measured lithium levels in the water of the different districts. It is not an exclusion criteria, however, if they personally have not sampled and tested the water that they use in their researches. There will not either be any specific control groups used as standards in the studies, as the comparisons will be drawn within the study population, between the different districts. They must also be studies of which suicide is a measured outcome, naturally.
2.2 Hypothesis
The hypothesis for this work is that consumption of water with high lithium levels has an inverse association with suicidality. Consequently, it is also hypothesized that there is a significantly lower rate of suicidality in regions with higher lithium concentrations in the water than in those areas where the opposite is true.
3. Method
Thanks to access from the library of Örebro University, literature searches were conducted mainly through the database of PubMed (https://www.ncbi.nlm.nih.gov/pubmed). Likewise, searches were also run in the PsychInfo database
(http://web.b.ebscohost.com.db.ub.oru.se/ehost/search), but yielded no additional relevant
material.
During the searches, two combinations of keywords were used: “suicide AND lithium AND water”, as well as “lithium AND drinking water”. The first combination rendered 35 hits; the second 388. In order to decrease the high number of hits from the second search, the articles that remained after selecting the filter “Species: Human” were examined (96 were left). In the end, 12 articles were deemed suitable for this review and are presented below.
During the searches, two combinations of keywords were used: “suicide AND lithium AND water”, as well as “lithium AND drinking water”. The first combination rendered 35 hits; the second 388. In order to decrease the high number of hits from the second search, the articles that remained after selecting the filter “Species: Human” were examined (96 were left). In the end, 12
Ändrad fältkod Ändrad fältkod
7
articles were deemed suitable for this review and are presented below.
publications by the same teams. This intention gave rise to the “Exclusion criteria” of the flow-chart below.. The exception being presented in this review is the article by Helbich et al (2015) in which the main aim is markedly different from the original work. Other follow-up works, like that of Helbich et al in 2012 ([7]) or Shiotsuki et al, 2016 ([11]) were brought up in the discussion instead, as they only exhibited aims to control the results of their previous original researches.
When assessing the quality of the studies, a few main criteria were considered: the amount of confounders controlled for, the sampling quality (amount of samples and width of lithium concentration span) and the size of the study population (number of districts and individuals). In Table 1 of the result section, these criteria are referred to as a, b, and c respectively; a being the most and c being the least important criteria when assessing the quality. Since suicidality is affected by a wide range of factors, adjusting for confounders is the most essential consideration. If the number of samples for a given study is low and the range between their respective concentrations are narrow, the validity goes down for the result. Lastly, the same goes for a study with a smaller population sample and fewer districts included in the analysis. This might only affect the statistical confidence slightly as the studies all handle large population pools.
Language was not a limitation when it came to selecting references.
The selected articles were examined by both the author of this paper and the responsible supervisor, Dr. Mats Humble.
Ethically, no special considerations were assessed to be necessary in a systematic review like this. No data used in this review represented any specific individual, as it handled information only on a population level. All the interventions conducted by the research teams were passive (sampling water).
8
Figure 1.
3 separate searches (2 on PubMed, 1 on PsychInfo) led to a total of 447 article hits. 364 of them were non-duplicate. 14 remained after application of the inclusion criteria and lastly, 12 after the exclusion criteria
4. Results:
The main findings are presented in Table 1 and each study is complemented below with important additional information that were provided in the articles. As the studies differed in strength by means of the quality criteria, their levels of validity are rated as “lower”, “medium” or “higher”. Criterium a is considered of “lower” strength when the amount of confounders is
Formaterat: Teckensnitt:16 pt, Fet, Kursiv
Formaterat: Teckensnitt:10 pt, Kursiv Formaterad tabell
9 from 0-2, medium from 3-5 and “higher” when there are 6 or more. For b, the three different ratings represent concentration spans of ≤50 µg, 50-100 µg and >100 µg, alongside considerations of the number of samples per district. Also, c is split into the three categories which weigh the relative size of each study.
One of the trailblazing pioneering studies on this subject was executed performed in 1989 and published the year after by the duo Gerhard N. Schrauzer and Krishna P. Shrestha [17]. Whilst Tthe study investigated an array of different outcomes (aggressive actions such as robberies, thefts and homicides; addictive behavior measured by), it and managed to find a correlationve bond between suicide and areas with relatively meager lower amounts of lithium in the drinking water. The aim was to examine whether such a connection was to be found in a population pool spanning multiple counties in the American state Texas. Counties were separated into three main categories regarding lithium concentrations - seemingly arbitrarily so – called A, B and C (“High Li”, “Medium Li” and “Low Li” respectively, as well). An additional group was added, called D, which only included low population areas with low lithium concentration, to eliminate the confounding factor population density and its effect on criminality. Also, the statistics in use for estimates on population sizes and crime/suicide levels were extracted from databases of Texas Department of Health and US Department of Justice, respectively. Calculations on statistical significance was were done with Student's t-test, adjusted with a Bonferroni correction, or with a paired t-test.
A 2009 study was conducted by a Japanese team, where lithium levels measured in tap water between the years 2002 and 2006 were used [18]. Hirochika Ohgami, Takeshi Terao and others conducted this research, partly on basis of further exploration of the findings in the
Schrauzer/Shrestha article. To balance the discrepancy population-wise between the municipalities and their influence on the result, a weighted least squares (WLS) regression analysis was carried out. Also, the differences between the regions in age and male/female ratio within the populations was were compensated by utilizing the standardized mortality ratio (SMR). A single measurement of the lithium level for each region was used in the investigation, although the differences in measurements made the following year were negligible. This was interpreted to be indicative that the degree of lithium in the water was not fluctuating enough to raise any questions on the initial measurements. Calculations were presented with the entire
10 population as a group, but also with two additional categories, separating women and men into their own subdivisions.
Published in April of 2011, an Austrian team headed by Nestor D. Kapusta, performed a study similar to that of the aforementioned Japanese group two years prior [19]. The water collecting process went on from 2005 to 2010. In order to minimize the risk of confounding, several methods were used and multiple factors were taken into account. Among these factors were population density, accessibility to mental and general health providers, Roman Catholic background, unemployment and per capita income. These data were provided by official Austrian institutions and the effective year for each of them varied from between 2001 and 2009. They resorted to different methods whilst processing the data: SMR regarding age and gender distribution, log-transformation on differences in population and health service density and lithium level and lastly a WLS regression analysis to balance the population size disparities. The same month, another work on the same topic was published by a team based in the United Kingdom; Nikolett Kabacs et al [20]. 2010 was the year in which all the water sampling took place. The SMR for suicide in use was based on official statistics for years 2006 to 2008. Little else information was provided in the text regarding methodology, other than their utilization of Pearson's correlation coefficient while examining the association between lithium level and suicidality. When it came to presenting the results, statistics on eventual significance were demonstrated in groups separate groups for males, females and one for both genders combined. Another study on the counties of Texas was carried out in 2012. Blüml et al amassed data on lithium levels that were in the possession of the Texas Water Development Board Groundwater Database [21]. The samples were collected from the years 1999 to 2007. Collected data on suicide mortality rates were also regarding the same year span. This information consisted of both a crude and an age-adjusted rate, extracted from the Texas Department of Health. Texas
Department of State Health Services provided the study with statistics on population levels and the US Census Bureau with population density figures, for the year of 2000. Data on male/female ratio, racial composition and socioeconomic circumstances in the counties were obtained from the American Community Survey of 2007. To analyze the data, both a linear and a Poisson rate regression model were used with all the previously mentioned factors accounted for by an adjustment. The crude suicide rate was chosen in favor the age adjusted one (they were deemed
11 to be interchangeable to a substantial degree, however).
Norio Sugawara and his team went on to further investigate the subject of tap water lithium and its connection to suicidality through their 2013 study on the Japanese prefecture Aomori [22]. Population statistics were obtained from an official government database, valid for the year 2010. A SMR for suicide was determined with age and gender disparities in mind and was done so for each municipality respectively. With a weighted least-squares regression analysis, size
differences between the different areas were dealt with statistically. The results were also presented with two different models: one with only lithium as a variable and a second with both medical institution density and unemployment as covariates.
In September of 2013, another study on the topic was published by a Greek team (Giotakos et al) [23]. Theirysamples were sampled collected during the prior year, 2012. From the database of the Greek Statistic Authority, the numbers on suicides per gender were received and used. The statistical analysis was performed with the IBM SPSS 20 package.
An Italian research team led by Maurizio Pompili published in January 2015 their own findings on the eventual connection between tap water lithium levels and suicidality in Italy [24]. Collection of water was conducted during the years 2009 and 2010. The sizes of the communities varied greatly, with village Cutigliano and Italian capital Rome at each end of the spectrum. Figures on population sizes and suicidality wereas taken from the Italian National Institute of Statistics. Mortality data was accessed to from the database of the institute and it encompassed the years 1980 to 2011, with 2004 and 2005 being the exceptions. To standardize the suicide rate to gender and age, the data was compared with population statistics of 2001, as in the amount of people living in each affected region (municipalities were water was sampled). Suicide rates were also determined for each region. The SMR was assigned into three groups (split up according to which decade each data represented) and data on men and women were separately handled. With a WLS regression analysis, population size variations were adjusted for. The results were presented in both a basic form with only lithium level accounted for and in a secondary multivariable model. In the latter model, factors that has a provenpreviously shown to correlateion with elevated levels of suicidality were included. These variables were: regions being "totally mountainous" or "highly urbanized" and the southern part of Rome.
12 Ishii, Terao and others that were also part of the Japanese publication in 2009 on the Oita prefecture returned to the subject matter by conducting a new, bigger study that spanned the entire Kyushu island of Japan [25]. Each region was assigned its own SMR for suicide, also in regard to both genders. Results were ordered in three different models: one crude (without taking confounders into account), one adjusted (adjusting for degree of elderly people, single person household, rate of high education and ratio of people in primary industry) and one twicely further adjusted (factors added on top of the first adjusted model were employment and marriage rate, local annual temperature and savings through postal services per person and year). Respective data were extracted from the Statistics Bureau, Ministry of Internal Affairs and Communication and Japan Meteorological Agency, for the years 2010 and 2009. Data point influence was balanced using a WLS regression analysis.
Likewise, Kapusta (along Marco Helbich and Michael Leitner) returned to further investigation of the topic in their May 2015 study, that had plenty of common elements with the Austrian publication four years prior (Kapusta et al, 2011) [26]. Once again, all the 99 districts that Austria is made up of were covered and the same number of water samples (handled by the same company) were was used. The type of controls used to eliminate confounders were was also seemingly identical. The aim, however, was a bit different in this work as it also tried to observe whether any conclusion could be drawn from data on degrees of lithium medication prescriptions and suicidality within the regions. Also, additional plots were drawn that were separated from the other graphs by the exclusion of the capital cities for each region. In this study, spatial biases were also taken care of through a calculation of Moran's I statistic. Spatial autocorrelation was thereby controlled for for regarding the suicide SMR that differed regionally. Further modelling of this was presented with a multivariate spatial Bayesian hierarchical statistical tool.
A new team to study this issue came from Lithuania, Liaugaudaite et al, and in March 2017 they went on to publish their observations [27]. Their investigation would include nine Lithuanian cities, with the motivation of picking these large settlements being to "decrease as much as possible the impact of the heterogeneity of economic and cultural background". A population ratio comparing the rate of men versus women in the different cities were provided by the Department of Statistics, covering the years 2009 to 2013. They compiled the statistics on suicide rates from the Lithuania Database of Health Indicators. These numbers represented the total
13 amount suicides committed during the years 2009 to 2013. An age standardized suicide SMR were calculated using statistics from the World Health Organization. A WLS regression analysis handled the population differences, cities in-between, and their respective statistical influence. Lastly, resulting data from the study was analyzed in the SPSS-program, version 17.0.
Kapusta, Blüml and others returned with another study on the subject and this time Chile was the country of interest [28]. This study was also published in March 2017, containing an analysis of suicide rates in all of Chile and how it regionally associates with local natural lithium levels in drinking water. This work had a hypothesis that it aimed to investigate: if the suicide rates in the Atacama (northern Chile) is significantly lower than in the rest of the country. With this in mind, the researchers divided the results into an Atacama and a non-Atacama category. Atacama is an area with exceptionally high levels of lithium in the ground and natural water sources. Excluded from the study were was the Chilean capital Santiago;, supposedly too statistically aberrant regarding compared to the other parts of the country and therefore incomparable. The last decade, 2000 to 2009, is the time span from which this work aimed to extract population data; for example, the suicide statistics, obtained from the Chilean Ministry of Health, included in the text are for these years. Statistics on the levels of lithium in the water seem to have been taken from a 2012 study by Figueroa et al, referenced to in the introduction of the paper. No other
specification is given on if any tap water measurements were executed or if other sources of data for lithium levels were used. Data on eventual confounder were mainly provided by the National Institute of Chile: the variables were employment status, median income, urbanization level and the degree of indigenous population. Additionally, they used a Moran I analysis to tackle the spatial auto-correlation bias that could skew the results.
14
St
ud
y
st
re
ngt
h
(qu
a
li
ty
c
ri
te
ri
a
)
a*, b ** * , c *** A ltogth e r of : m e di u m qua li ty a*, b*, c ** Low e r qu ali ty a*** , b **, c *** H ig h e r qu a li ty a*, b *, c ** * Low e r qu ali ty a**, b*** , c * M e di um qua li ty a*, b *, c ** Low e r qu ali ty * = Lo w er q u al it y , * * = m ed iu m q u al it y , * * * = h ig h er q u al it ya
b
le
1
.
St
udy
d
eta
il
s
a
n
d
r
e
sult
s
R
e
sult
s,
sta
tistical
si
g
n
if
ica
n
ce
*
S ig n ifi ca n ce fo u n d f o r g ro u p A (co n ce n trati o n ra n g e s 6 0 -1 7 0 ), P < 0 .0 0 5 . S ig n ifi ca n ce fo u n d f o r th e e n ti re p o p u la ti o n a n d m e n (P = 0 .0 0 4 an d 0 .0 0 8 res p ec ti v el y ), wo m en b ei n g m ar g in al ly si g n ifi ca n t (0 .0 6 > P > 0 .0 5 ) as a g ro u p . S ig n ifi ca n ce fo u n d , P = 0 .0 0 0 0 7 3 fo r o v erall s u ic id e, P = 0 .0 0 0 0 3 0 fo r su ic id e S M R . No sig n ifi c an c e fo u n d , P = 0 .8 3 8 fo r b o th g e n d er s, 0 .7 7 7 fo r fem al es a n d 0 .7 1 5 fo r m al es. S ig n ifi ca n ce fo u n d , P < 0 .0 5 fo r al l m o d el s (cr u d e P o iss o n rate reg res si o n , P o iss o n rate reg res si o n wi th p o ly n o m ia ls a n d W ei g h te d q u asi -P o iss o n reg res si o n ). S ig n ifi ca n ce fo u n d f o r fem al es b efo re a d ju stme n ts (P < 0 .0 5 ), wi th a t re n d a ft er a d ju st m en ts (P = 0 .1 0 ) a n d n o n e wh at so e v er fo r m al es . * In d ic at in g a n a sso ci at io n b et we e n h ig h er l e v el s o f li th iu m in d ri n k in g wa te r wi th l o we r su ic id e rat es . L it hi um l evels (r ange or m ean) (L i/ L ) 0 -120 µ g < 0,7 -59 µg 11. 3 µ g (S D = 27) <1 -27 µ g < 2,8 -219 µ g ,0 -12. 9 µ gSam
p
le
am
o
u
n
tt
, lit h iu m l ev els (r an g e o r m ea n ) (L i/ L ) N ot sp e cifie d , 0 -120 µ g N ot sp e cifie d , < 0,7 -59 µ g 6460 , 11.3 µ g (S D = 27) 47 , < 1 -27 µ g 3123 , < 2,8 -219 µ g N ot sp e cifie d , 0 -12. 9 µ g Lith iu mPo
p
u
lat
io
n
(
spa
n
,
lar
g
es
t-sm
alle
st)
N ot s p ec ifie d (2,7 82, 000 -11,000) 1,206,174 (463 ,97 3 -2,408) 8,297,964 (1,6 67, 878 -1,714) 5,700,000 (188 ,80 0 -62,200) N ot s pecif ied ( not specif ied) 1,373,339 (299 ,52 0 -1,594)R
e
g
io
n
s,
per
io
d
27 coun ti es , Te x as st a te , U S A , 1978 -1987 18 m uni c ip ali ties , O ita pre fe ctu re , Ja p a n, 2002 -2006 99 di stri c ts , A u stri a , 2005 -2009 46 subd iv isi on s, Ea st of Engla nd re gi on, U ni te d K in gdo m , 20 06 -2008 226 c oun ti es , Te x as st a te , U S A , 19 99 -2007 40 m uni c ip ali ties , A om ori pre fe c tu re , Ja pa n, 2 008 -2010
d
y
auze r es ht a, et 200 9 st a et 201 1 ac s et 201 1 m l, e t 201 3 w ar a 013 Formaterat: Teckensnitt:14 ptFormaterat: Teckensnitt:Inte Fet Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Radavstånd: enkelt
Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Radavstånd: enkelt
Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Teckensnitt:Inte Fet
Formaterat: Teckensnitt:9 pt
Formaterat: Teckensnitt:11 pt
Formaterat: Teckensnitt:16 pt
15
St
ud
y
st
re
ngt
h (
qu
a
li
ty
cr
it
er
ia
)
a*, b*** , c* A ltoge th er of: l owe r qu ali ty a* * , b** , c *** M e di um qu a li ty a***, b** *, c *** H ig h e r qu a li ty a***, b**, c *** H ig h e r qu a li ty a*, b *, c * Low e r qu a li ty a**, b*, c *** M e di um qu a li ty * = Lo w er q u al it y , * * = m ed iu m q u al it y , * * * = h ig h er q u al it ya
b
le
1
, c
o
ntinu
e
d
R
e
sult
s,
sta
tistical
si
g
n
if
ica
n
ce
*
S ig ni fic a n c e found, P < 0 .05 . S ig n ifi ca n ce (P = < 0 .0 5 ) fo u n d i n c ru d e u n iv ari at e m o d el fo r 1 9 8 0 -1 9 8 9 , fo r m en i n th e sa m e p eri o d a n d wo m en t h ro u g h a ll d ec a d es . In t h e ad ju sted m o d el , si g n ifi ca n ce wa s o n ly f o u n d a m o n g w o m e n 1 9 8 0 -1 9 8 9 . Si g n if ica n ce ( P < 0 .0 5 ) fo u n d i n c ru d e m o d el an d ad ju st ed m o d el 1 , an d f o r m en i n ad ju st ed m o d el 1 an d 2 . Si gni fi canc e (P < 0 .05) f ound in t he Spear m an m odel and in t he M or an m odel f or genders com bi ne d and m al es , but not f o r fe m ales a lo ne . Si gni fi canc e (P < 0 .005) f o und in acc or dan ce to Pear son co rr e lat ion coeff ici ent am ong bot h gen der s and m en. It w as al so f ound w it h w ei ght ed regr es si on m odel s f or s ize am ong the genders com bi ned in m odel 2 and 3 and am ong m en i n m odel 2 and 3. S ig ni fic a n c e found for the ge nde rs c om bi n e d (P < 0.0 01) a nd am ong m e n (P = 0.0 01). * In d ica ti n g an a ss o ci at io n b etw ee n h ig h er l ev els o f lith iu m i n d ri n k in g w at er w it h l o w er s u ic id e ra tes . Li th iu m le v el s (r an g e o r m ea n ) (L i/ L) 0.1 -121 µg 0.11 -60.8 µg 0-130 µ g 1 0 µ g (SD =1 0 ), m in im u m = 3 µ g 0.48 -35.5 3 µ g, 10.9 (SD = 9.1 ) A s u rf ac e wa te r m ax im u m o f 2 0 7 m g i n t h e At ac am aSam
p
le
am
o
u
n
t
, lit h iu m lev els ( ra n g e o r m ea n ) (L i/ L ) 149 , 0.1 -121 µg 157 , 0.11 -60.8 µg 43 4 4, 0 -130 µ g 646 0, 10 µ g (S D = 10), m in im um = 3 0 µg 22 , 0.48 -35.53 µ g, 10.9 (SD = 9.1 ) N ot sp e cifie d , a su rf ac e w at er m ax im u m o f 2 0 7 m g i n th e A tac am a Lith iu m . * Ass u m ed sa m e va lu es a s in K a pu st a e t a l Po p u la tio n (s p an , lar g es t -sm all est) N o t sp ec if ied (n o t sp ec if ied ) 1 7 ,2 0 0 ,0 0 0 (2 ,6 0 0 ,0 0 0 -1, 561) 14 ,6 4 6 ,1 2 1 (1 ,4 0 9 ,2 9 7 -3 6 6 ) N o t sp ec if ied (n o t sp ec if ied ) * 1,190,261 (526 ,35 6 -2 525) No t sp ec if ied (n o t sp ec if ied )R
e
g
io
n
s,
p
er
io
d
32 pre fe c tu re s, G re ec e , 1999 -2010 145 com m uni ti es , It al y , 1980 -2011 2 7 4 m u n ic ip al it ie s, Ky u sh u Ii sla n d , Ja p a n , 2 0 1 1 99 di stri c ts , A u stri a , 2005 -2009 9 c it ie s, Lith ua ni a , 2009 -2013 A ll 12 re gi on s of , C hi le . 2000 -2009
d
y
ta ko s et 201 3 pi li e t 201 5 i et al, ic h et 201 5 udai te , 2017 g e t al,Formaterat: Radavstånd: enkelt
Formaterat: Teckensnitt:Inte Fet Formaterat: Teckensnitt:12 pt, Inte Fet Formaterat: Teckensnitt:Inte Fet Formaterat: Teckensnitt:Inte Fet Formaterat: Teckensnitt:12 pt, Inte Fet
Formaterat: Teckensnitt:Inte Fet Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:10 pt Formaterat: Teckensnitt:11 pt Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:10 pt Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:16 pt Formaterat: Teckensnitt:12 pt Formaterat: Radavstånd: enkelt Formaterat: Teckensnitt:10 pt Formaterat: Teckensnitt:10 pt Formaterat: Teckensnitt:10 pt Formaterat: Teckensnitt:10 pt Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:10 pt Formaterat: Radavstånd: Flera 1,15 li Formaterat: Radavstånd: enkelt Formaterat: Teckensnitt:10 pt Formaterat: Radavstånd: Flera 1,15 li Formaterat: Radavstånd: enkelt Formaterat: Teckensnitt:11 pt Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:9 pt Formaterat: Teckensnitt:10 pt
16 Table 1. Data on each included study. The evaluations in “Study strength” are based on the quality criteria elaborated on in section 3, Method.
5. Discussion
To summarize the results: in terms of the studies observed, evidence slightly suggests that there is an inverse relationship between lithium levels in drinking water and suicidality, overall.
Significance in most the outcome indicators were found in 9 of 12 studies, partial evidence was established in 2 of them (Pompili et al and Sugawara et al) and none in one research (Kabacs et al).
Comparing the researches, however, is not as straightforward as one who has limited familiarity with the studies may believe. There is a great deal of heterogeneity between the studies considering their designs: differing outcome variables (usage of suicide SMR, categorizations (such as by gender), etc.), exclusion criteria (cities/regions excluded), inclusion criteria (adjacent regions, regions with diverse characteristics, all regions of a country, arbitrarily chosen regions), water sampling methodology (dozens of samples per district, few/a single sample/s per district, no original sampling), confounders adjusted for and statistical tools for result calculations. The most striking difference between the studies are is perhaps how each research team chose their set of confounders to standardize their subsequent results. It ranges from not incorporating any types of adjustments (Giotakos et al) to including a wide and plentiful number of
confounders (Ishii et al, Kapusta et al) into the respective studies’ final calculations. Also, whilst a research certain study may in fact take plenty of confounders into account, drawing parallels between it and the other studies can be problematic. This is regarding due to the diverse nature of the confounders used in all the differentthe included studies: one used annual local temperature (Ishii et al), another one the degree of Roman Catholics in the population (Kapusta et al) and a third one the ethnic composition of the inhabitants (Blüml et al) as variables, to name a few. AlthoughHowever, to be fair, some basic variables were controlled for repeatedly across the researchesstudies, like that ofsuch as population density and employment status. This was also true for how the suicide SMR was implemented, often adjusted for age and sex-ratio differences
Formaterad tabell
17 between districts. Still, more homogeneity with the use of confounders would have made it much easier to draw accurate conclusions when comparing the studies.
However, there have been a few further studies that some of the research teams have engaged in, having the intent to test their results with new sets of confounders included in their final calculations. One such adjustment was the spatial autocorrelation control exercised by Helbich, Kapusta and Leitner in their 2012 follow-up work to their 2011 original study [29]. They employed a few new methods, like Moran’s I and local G*-static, to explore whether spatial autocorrelation existed and to what extent. Statistical significance remained after these adjustments and the conclusion from their previous study was in their perspective still valid, hence. The same group of researchers, with the addition of Victor Blüml, continued their research by publishing another study in 2013 [30]. This time, altitude was the factor controlled for as it has been postulated that an association with suicide frequency can be found [31, 32]. In the end, SMR for suicide and its inverse relation to suicide rates stood strong with altitude adjusted for and it also concluded that lithium levels in the water were lower the higher up you measure,
altitude-wise. Also, the team behind the study on Kyushu iIsland (Ishii et al, 2015) went on to test the strength of their earlier results in a new effort the year after [33]. In this instance, weather factors were weighed into the calculations (annual rainfall, snowfall, mean temperature and sunshine hours). Associations between these meteorological factors and suicidality haves been extensively discussed in previous studies, as in [34] and [35]. Statistical significance for lithium emerged only for the suicide SMR of males, however,and for maleswith suicide correlated directly with the weather factors “annual total rainfall”, and negatively with annual total sunshine and mean“temperature”isolated.
Conclusively, it is fair to say that adding these potential confounders in the other works surely would have strengthened the validity of their results.
Another important aspect pertaining to these types of studies is how patterns of consumption in the populations affect and possibly skew the results. It is not investigated in any of the researches how and to what amount the different populations consume the tap water in their homes, etc. Whilst intake of lithium in the diet varies with a multitude of factors (location, diet composition, lithium concentration in the food and water), an estimate in the United States, for example, is that the daily consumption of lithium ranges between 650 and 3100 µg, of which 66 to 90 percent is
18 estimated to come from grains and vegetables in the diet (the rest is mainly from animal
products) [36]. Consequently, evidence suggests that water plays only a minor part in the daily dietary intake to provide the individual with lithium. To further speculate, it is not clear how much tap water the average individual in any given district consume (daily, for example) and how that trend differs between countries and the districts within them.
For future research, there are plenty of other limitations in these types of studies that need to be addressed. Firstly, a more complete understanding on what effects lithium has physiologically at a trace level dosage would be very instructive. Secondly, it would be helpful to know at what different lithium concentration intervals one can find clear patterns of these effects; also, epidemiologically. This presupposes that it is known how much the intake of lithium/tap water is for a given individual or population district. Thirdly, one has to identify how wide the range of lithium concentrations in a multi-district spanning study must be to yield significant results. In the study by Kabacs et al., where significance was not found at all, this range was quite narrow. Indeed, the generally low levels in this study may be below the putative threshold for trace-level lithium to be relevant. Fourthly, it could be misleading to assume that one sample or the median value of a number of samples in a certain district are accurately representative of the households and public buildings drinking water. The results from the repeated measurements made by e.g. Ohgami et al, however, suggests otherwise.
6. Conclusion
Most of the evidence presented in the results suggests that lithium, at a trace-level amount, has an inverse relationship with suicide. Differences between the studies in terms of execution,
thoroughness and size makes it difficult to assess whether this assumption is valid, however. Also, in many of the studies, the significance was only found partially and not across-the-board in the outcome measures. This demonstrates the need of more uniformity among upcoming studies on this issue so that more comprehensive conclusions can be drawn. Also, as many as possible of the uncertainties brought up in the discussion need to be addressed, too. Indeed, it is quite feasible that the beneficial effects of trace amounts of lithium can be obscured by the many disturbing elements possibly affecting manipulating the results of thisese kinds of studies.
19 Special thanks go out to supervisor Dr. Mats Humble and the Örebro University Medical Library.
1. Hawton K, van Heeringen K. Suicide. Lancet 2009 Apr;373:1372-81
2. WHO. Preventing suicide: a resource for media professionals. 2000 Geneva: World Health Organization
3. Cipriani A, Hawton K, Stockton S, Geddes JR. Lithium in the prevention of suicide in mood disorders: updated systematic review and meta-analysis. BMJ 2013;346:f3646 4. Dawson EB, Moore TD, McGanity WJ. The mathematical relationship of drinking water
lithium and rainfall to mental hospital admission. Dis Nerv Syst. 1970 Dec;31(12):811-20 5. Mauer S, Vergne D, Ghaemi SN. Standard and trace-dose lithium: A systematic review of
dementia prevention and other behavioral benefits. Aust N Z J Psychiatry. 2014 Sep;48(9):809-18
6. Vita A, De Peri L, Sacchetti E. Lithium in drinking water and suicide prevention: a review of the evidence. Int Clin Psychopharmacol. 2015 Jan;30(1):1-5
7. Lenox RH, Hahn CG. Overview of the mechanism of action of lithium in the brain: fifty-year update. J Clin Psychiatry. 2000;61 Suppl 9:5-15
8. Farhy Tselnicker I, Tsemakhovich V, Rishal I, Kahanovitch U, Dessauer CW, Dascal N. Dual regulation of G proteins and the G-protein-activated K+ channels by lithium. Proc Natl Acad Sci U S A. 2014 Apr 1;111(13):5018-23
9. Serretti A, Drago A, De Ronchi D. Lithium pharmacodynamics and pharmacogenetics: focus on inositol mono phosphatase (IMPase), inositol poliphosphatase (IPPase) and glycogen sinthase kinase 3 beta (GSK-3 beta). Curr Med Chem. 2009;16(15):1917-48 10. Manji HK, Moore GJ, Chen G. Lithium up-regulates the cytoprotective protein Bcl-2 in
the CNS in vivo: a role for neurotrophic and neuroprotective effects in manic depressive illness. J Clin Psychiatry. 2000;61 Suppl 9:82-96
20 11. Beurel E, Jope RS. Inflammation and lithium: clues to mechanisms contributing to
suicide-linked traits. Transl Psychiatry. 2014 Dec 16;4:e488
12. Janssen HL1, Brouwer JT, van der Mast RC, Schalm SW. Suicide associated with alfa-interferon therapy for chronic viral hepatitis. J Hepatol. 1994 Aug;21(2):241-3 13. Fragoso YD, Frota ER, Lopes JS, Noal JS, Giacomo MC, Gomes S et al. Severe
depression, suicide attempts, and ideation during the use of interferon beta by patients with multiple sclerosis. Clin Neuropharmacol. 2010 Nov-Dec;33(6):312-6
14. Dieperink E, Ho SB, Tetrick L, Thuras P, Dua K, Willenbring ML. Suicidal ideation during interferon-α2b and ribavirin treatment of patients with chronic hepatitis C. Gen Hosp Psychiatry. 2004 May-Jun;26(3):237-40
15. Terao T, Goto S, Inagaki M, Okamoto Y. Even very low but sustained lithium intake can prevent suicide in the general population. Med Hypotheses. 2009 Nov;73(5):811-2 16. Sher L. Suicide in men. J Clin Psychiatry. 2015 Mar;76(3):e371-2
17. Schrauzer GN, Shrestha KP. Lithium in drinking water and the incidences of crimes, suicides, and arrests related to drug addictions. Biol Trace Elem Res. 1990
May;25(2):105-13
18. Ohgami H, Terao T, Shiotsuki I, Ishii N, Iwata N. Lithium levels in drinking water and risk of suicide. Br J Psychiatry. 2009 May;194(5):464-5; discussion 446
19. Kapusta ND, Mossaheb N, Etzersdorfer E, Hlavin G, Thau K, Willeit M et al. Lithium in drinking water and suicide mortality. Br J Psychiatry. 2011 May;198(5):346-50
20. Kabacs N, Memon A, Obinwa T, Stochl J, Perez J. Lithium in drinking water and suicide rates across the East of England. Br J Psychiatry. 2011 May;198(5):406-7
21. Blüml V, Regier MD, Hlavin G, Rockett IR, König F, Vyssoki B et al. Lithium in the public water supply and suicide mortality in Texas. J Psychiatr Res. 2013
Mar;47(3):407-11
22. Sugawara N, Yasui-Furukori N, Ishii N, Iwata N, Terao T. Lithium in tap water and
Formaterat: Svenska (Sverige)
21 suicide mortality in Japan. Int J Environ Res Public Health. 2013 Nov 12;10(11):6044-8 23. Giotakos O, Nisianakis P, Tsouvelas G, Giakalou VV. Lithium in the public water supply
and suicide mortality in Greece. Biol Trace Elem Res. 2013 Dec;156(1-3):376-9
24. Pompili M, Vichi M, Dinelli E, Pycha R, Valera P, Albanese S et al. Relationships of local lithium concentrations in drinking water to regional suicide rates in Italy. World J Biol Psychiatry. 2015;16(8):567-74
25. Ishii N, Terao T, Araki Y, Kohno K, Mizokami Y, Shiotsuki I et al. Low risk of male suicide and lithium in drinking water. J Clin Psychiatry. 2015 Mar;76(3):319-26 26. Helbich M, Leitner M, Kapusta ND. Lithium in drinking water and suicide mortality:
interplay with lithium prescriptions. Br J Psychiatry. 2015 Jul;207(1):64-71
27. Liaugaudaite V, Mickuviene N, Raskauskiene N, Naginiene R, Sher L. Lithium levels in the public drinking water supply and risk of suicide: A pilot study. J Trace Elem Med Biol. 2017 Mar 24. pii: S0946-672X(16)30288-7
28. König D, Baumgartner J, Blüml V, Heerlein A, Téllez C, Baus N et al. Impact of natural lithium ressources on suicide mortality in Chile 2000-2009: a geographical analysis. Neuropsychiatr. 2017 Mar 29 (pages notspecified)
29. Helbich M, Leitner M, Kapusta ND. Geospatial examination of lithium in drinking water and suicide mortality. Int J Health Geogr. 2012 Jun 13;11:19
30. Helbich M, Blüml V, Leitner M, Kapusta ND. Does altitude moderate the impact of lithium on suicide? A spatial analysis of Austria. Geospat Health. 2013 May;7(2):209-18 31. Brenner B, Cheng D, Clark S, Camargo CA Jr. Positive association between altitude and
suicide in 2584 U.S. counties. High Alt Med Biol. 2011 Spring;12(1):31-5
32. Kim N, Mickelson JB, Brenner BE, Haws CA, Yurgelun-Todd DA, Renshaw PF. Altitude, gun ownership, rural areas, and suicide. Am J Psychiatry. 2011 Jan;168(1):49-54
33. Shiotsuki I, Terao T, Ishii N, Takeuchi S, Kuroda Y, Kohno K et al. Trace lithium is inversely associated with male suicide after adjustment of climatic factors. J Affect Disord.
Formaterat: Svenska (Sverige)
Formaterat: Svenska (Sverige)
Formaterat: Svenska (Sverige)
22 2016 Jan 1;189:282-6
34. Terao T, Soeda S, Yoshimura R, Nakamura J, Iwata N. Effect of latitude on suicide rates in Japan. Lancet. 2002 Dec 7;360(9348):1892
35. Tsai JF. Socioeconomic factors outweigh climate in the regional difference of suicide death rate in Taiwan. Psychiatry Res. 2010 Sep 30;179(2):212-6
36. Schrauzer GN. Lithium: Occurrence, Dietary Intakes, Nutritional Essentiality. J Am Coll Nutr. 2002 Feb;21(1):14-21
