New data on ecological analysis of possible relationship between magnesium in drinking water and liver cancer

ARTICLE

Auteur(s) : Andrzej Tukiendorf¹, Zbigniew Rybak²

¹ Technical University, ul. Mikolajczyka 5, 45-271 Opole, Poland. E-mail: antupo.opole.pl;
² Medical University, ul. Poniatowskiego 2, 50-326 Wroclaw, Poland.
E-mail: zrybak@chirn.am.wroc.pl

Introduction

Numerous studies in all the continents suggest an inverse relationship between various types of heart diseases and the drinking water hardness. Drinking soft water increases the cardiovascular risk and this effect is reciprocally reduced by the hard water consumption (see e.g. [1-3]). Among many variables involved in the “water story”, and the hardness of water especially, magnesium (Mg) appears preeminent [4] (a wide review of evidence of the influence of Mg on health can be found in [4, 5].

Magnesium research is moving rapidly and it is clear that in the next few years there will be many advances in understanding the role of magnesium in the human organism [6]. Up to now it has been documented that over 300 enzymes that influence the metabolism of carbohydrate, amino acids, nucleic acids and protein, and ion transport, require Mg [7] and its functions in the fatty acid and phospholipid acid metabolism that affect permeability and stability of membranes are being elucidated. It has been proposed that Mg is central in the cell cycle and its deficiency is an important conditioner in the precancerous cell transformation. In addition, immunocompetence (that eliminates transformed cells) is Mg-dependent. Moreover, the magnesium supplementation of those who are Mg deficient might decrease emergence of some malignancies. However, Mg deficiency can paradoxically protect against oncogenesis or increase this risk [4-9]. In several experimental animal models of benign or premalignant tumorogenesis, the element inhibits the development of the cellular defect, while it is very important that the preventive effect of Mg exists only at the early stage of tumorogenesis. However, indirectly through mediation of growth factors and directly through its effects on the cell (essential growth factor), Mg stimulates normal and proliferative cell growth. This action constitutes one of the main mechanisms of its carcinogenic property [8, 9].

Although in dozens of studies – in particular in those done on the largest geographical scale, the effects of drinking water Mg on morbidity and mortality – predominantly from cardiovascular disease – have been recognized, until now not many reports have confirmed the cancer epidemiology of water magnesium. Important findings in this field were provided very recently by Taiwanese scientists who indicated negative statistical associations between colon, rectal, pancreatic, gastric as well as prostatic cancer morbidity/mortality and the hardness of water, calcium and magnesium [10-15]. In these reports the authors suggested that these factors (among which water magnesium) may be protective against the neoplasms. The earlier epidemiologic findings confirmed high cancer rates within areas with low soil Mg [16, 17].

Since magnesium has been judged likely to contribute to human carcinogenesis, it is worth testing whether drinking water Mg is a protective geochemical agent against liver cancer. Moreover, an evaluation of what it is in different geographic areas provides an interesting approach to age disturbances of the health of patients, since a depressed immunologic function occurs with aging (see e.g. [18]). Thus, the main objective of this study was examining whether living in a magnesium rich water area might protect against cell transformation and might reduce the risk of prevalence that increases with age.

The latest data on ecological analysis of a possible relationship between Mg in drinking water and liver cancer incidence gives new light on magnesium cancer epidemiology [19]. Because the findings have not been supported by any scientific reports yet, they may imply apparent controversies and should undergo scientific verification or critics.

Materials and Methods

In the work presented, extensive data were used to study possible effects of the water quality on the human health status.

Information on magnesium concentrations in drinking water originated from the Provincial Sanitary-Epidemiological Station in Opole and its ten dependant Local Sanitary-Epidemiological Stations that were the legally obligated institutions to control drinking water. In the analyzed period 1980-1985, the laboratory results from 1,040 dug wells and 307 municipal water mains were taken into account – but only from those for which the number of analyses performed a year was a few per dug well and about a dozen per water main. It is of note that the period examined was chosen because of a stagnation in water supply construction that, in turn, had a purely economical background (this inactive period was confirmed by only incidental cases of new water mains or dug wells constructed at that time in the region).

The liver cancer data concerned all morbidity cases (219 in males and 273 in females – all ages) of the disease (code 155 following the International Classification of Diseases, Ninth Revision), registered in the province in the decade of 1985-1994. The information originated from the Opole Cancer Registry (OCR), which has been providing a highly reliable service following Finnish Cancer Registry instructions since 01.01.1985. The data collection relied on an active program of cancer registration, i.e. follow-back/follow-up verifications and a set of eight other disposable medical methods to confirm the disease. It is of note that the OCR's collection was used by the International Agency for Research on Cancer [20].

The referred male and female populations in age groups in the analyzed administrative units were obtained from the National Census 1988. In the study, the total numbers of 462,511 males and 485,153 females, respectively, were taken into account.

In the study, some simple and complex methodological approaches were applied to analyze this ecological relationship.

The quality of the water supplied was estimated as follows:

(i) for the rural administrative units, the monitored Mg concentration in drinking water was attributed to the whole population inhabiting a particular village. In case of villages with more than one monitored dug well, these contents were averaged;

(ii) for the urban administrative units, the synthetic Mg contents (weighted-average concentrations) were estimated based upon the sizes of populations supplied by particular water providers.

To show spatial distribution of the concentrations, the results were presented geographically in thematic maps.

Complementarily, to indicate differences in morbidity and in density in age groups, histograms of the liver cancer prevalence and of the population structure were created.

To test a relation between the magnesium exposure and the liver cancer morbidity, a Bayesian modelling using random effects logistic regression [21-23] was conducted although some other statistical approaches may be of use in this example. To justify the choice, the suitability and simplicity of this method to test the trend in data must be principally emphasized. From the model (after [22, 23])

r_i ~ Binomial (π_i, n_i)
logit(π_i) = α +  ρ(x) + b_i

where r_i is the number of cancer cases within the age group and within the magnesium exposure category, n_i – the respective population “at risk”, ρ(x) – the regression function of x (i.e. Mg/age two-dimensional categorical variable), and b_i – the random effects (unexplained variation), the true prevalence probabilities (PPs) – π_i were calculated for the chosen 0-29, 30-59, 60-89 age bands and (,10], (10,20], (20,30], (30,40], (40,) mg/dm³ Mg drinking water levels. This ordering of categories (with the first as baseline) provides the lower ρ(x) as baseline. In the supposed regression function of x.

ρ(x) = β₁ x_i1 + β₂ x_i2 + β₃ x_i3,

x_i1 denotes the following Mg exposure categories while x_i2 and x_i3 the two higher age groups (similarly to [23]). In the model, no censoring for the β parameters was assumed.

For the purpose of this study, the cancer prevalence was understood as a measure of the burden of cancer in a population in an age group at a particular exposure category.

The computation was performed in WinBUGS version 1.4 [24] relying on a simulation technique known as Markov Chain Monte Carlo (MCMC) [25]. To achieve the convergence, three parallel chains were run and the first 1,000 samples of each were discarded as a burn-in while the following 10,000 cycles of the Gibbs sampler were used to estimate each quantity of interest. An equilibrium state of streams of values was established via an examination of within chain autocorrelation and a comparison of the results of the chains started with overdispersed initial values, including the use of the Gelman-Rubin statistics available within the software (see [24] web site for details).

The posterior analyses of the relations were set in a table and exposed graphically in scatterplots.

Results

The spatial distribution of the drinking water magnesium exposure in Opole province in the years 1980-1985 (see also [19]) is shown in figure 1 (areas indexed as “NA” were not taken into account due to an unrepresentative sample of chemical analyses – see the Materials and Methods section for details).

The map in figure 1 shows a geographical differentiation of the magnesium exposure across the province. The highest Mg concentration in drinking water was noticed in the middle east of the region, while the element's lowest contents were supplied in the northern and the southwestern parts of the province.

The empirical probability densities of patients (males and females) in the analyzed age trimesters 0-29, 30-59, and 60-89 years old are presented in figure 2.

The bars in figure 2 provide evidence of an apparent age effect in liver cancer prevalence. The most diseased group (both male and female) consisted of the oldest patients while the youngest group represented the extreme low cancer risk.

The population structure in age groups for the analyzed administrative units is given in figure 3.

Figure 3 testifies very similar population structures of males and females within the age groups considered except for the oldest persons.

The posterior estimates of the slope regression parameters are set in table I.

Table I. Posterior analysis of the slope regression parameters

The posterior analysis of the β parameters testifies a reducing effect of the magnesium contents on the liver cancer prevalence probability (negative β₁) and a considerable increase of the disease risk with age (positive β₂ and β₃) both in males and females. The contributions of the drinking water exposure to endemic male and female incidences are presented in figures 4 and 5.

The results displayed in figures 4 and 5 confirm the morbidity trends estimated via the slope regression parameters table I. The prevalence probability was smaller and directly correlated with the greater magnesium exposure. The modeled data show that PP for the depleted Mg concentration in drinking water is about three times as high as far for its elevated contents in males and about twice in females. Moreover, the highest PPs are observed for the older patients while the youngest group is at the lowest cancer risk. Due to a very small number of prevalent cases figure 1, the youngest patients have the widest credible 95% intervals. The remaining 30-59 and 60-89 age groups have very similar PPs in relation to Mg categories.

Discussion

The discussion over the research conducted can be confined to some major points:

(i) evaluating the potential impacts of environmental exposures on human health is an increasingly important component of chronic disease epidemiology. However, assessing exposure levels to many environmental components has proven difficult. For example, studies of the influence of drinking water ingredients on health may sometimes be hampered by inadequate historical data on their concentrations in water supplies on both spatial and temporal scales. While personal exposure levels are the most desirable measure for epidemiologic research, historical ingredient levels at the tap' for individual households are usually not available. Therefore, epidemiologic investigations of drinking water contents and health outcomes have used average or aggregate measures of water ingredients, generally at the administrative unit level [4];

(ii) the potentially highly controversial nature of study in this field makes it imperative that results are based upon good data. Despite the convenience of working on highly reliable statistics, due to the small number of cases involved, caution is certainly necessary. Since research was centered on a particular rare disease, the distribution of the population over categories may be often highly skewed with similarly small number of persons exposed at one of the concentration levels. Consequently, standard tests of trend which rely only on asymptotic approximations may give exaggerated significance levels and thus more advanced methods are required;

(iii) in this context, the chosen model (logistic regression [21-23]) as well as the computation methodology (MCMC/WinBUGS [25, 24]) have been commonly employed and already acknowledged by the statistical/epidemiological community. Thus, their application seems to be fairly justified in such an ecological analysis and should provide satisfactory results from the statistical point of view;

(iv) at the present stage of scientific knowledge, however, the given findings become new data in the epidemiology of water magnesium. Therefore, from the epidemiological perspective, the results might be worth taking into consideration in the field of the organ's cancer pathogenesis that is still being scrutinized. Such a new focus will not be unproblematic since recent scientific reports have included chronic liver infection (hepatitis) as one of the most important etiological factors of liver cancer [26]. Moreover, they accordingly confirm that there is a strong association between chronic hepatitis B and C viral infection and the development of hepatocellular carcinoma. However, people with cirrhosis also are at an increased risk of liver cancer. Other possible hepatocarcinogens include aflatoxin, nitrosamines, oral estrogen compounds, and numerous other chemicals [26]. Thus, by reason of a very wide range of the mentioned risk factors, it is very difficult to suggest any possible biochemical mechanism which could explain the phenomenon considered in this research without controversy;

(v) the problem relies on contributions of magnesium in drinking water to the total amount ingested. In case of myocardial infarction (MI), for instance, it has been suggested that the contribution of Mg in water to total dietary intake might be critical in soft water areas, leading to subclinical Mg deficiency, with increased risk of MI [27, 28]. Others [29] pointed out that marginal magnesium shortage is a problem in developed countries, and hypothesized that Mg in water might thus be a determining factor in whether the element's intake is adequate. A survey by [30] disclosed that hard water contributes an average of 12% of the Mg ingested. It has been also reported [31] that hard water could contribute up to a fifth of the daily Mg needs, which might be enough to prevent deficiency in those with marginal dietary magnesium. It has been found that some hard waters could provide up to 100 mg of Mg daily. Moreover, in some areas and for those drinking much water, these waters might actually provide the recommended daily requirement of magnesium [1, 32]. These evidences corroborate a significant contribution of Mg to its total intake;

(vi) because alcohol causes abnormalities in magnesium metabolism and results in severe liver diseases (including cirrhosis [33, 34]), one of the hypothesized reasons for the association between the Mg exposure and liver cancer could be the direct/indirect effects of alcohol consumption, since it decreases the element's concentration in body and can be provided back to organism via water ingestion. Then, linking the higher intake of the element from the richer water supplies together with its preventive properties in neoplasmatic processes might be a reasonable justification why the liver cancer morbidity diminishes with the increase of drinking water magnesium exposure, and vice versa;

(vii) the hypothesis presented could be true with the assumption of a similar amount of water ingested by the population at different exposures to the element. However, it is less reasonable to assume that these people had other needs in water consumption since they represented very similar geographical and socio-economic conditions.

Conclusions

Researchers in hospitals and medical centers around the world are working to learn more about what causes liver cancer. At this time, no one knows its exact etiology and new aspects of the disease are still being considered. Recently, magnesium research has released many scientific signals on the subject of cancer epidemiology.

In this study, based on extensive data and professional statistical methods, the underlying effect of the magnesium exposure has been included among the protective factors against liver cancer. Despite the findings that suggest that drinking water rich in magnesium reduces the risk of the disease, there are many unknowns concerning whether the assessed cancer tendency has only a regional or a global nature, or about the possible biochemical mechanism of this relationship, etc. Thus, to investigate these questions, firstly, a verification of the findings with the results of other epidemiologic studies (conducted in a different geographic location) is needed. Secondly, it is necessary to move on from studies on aggregated populations to studies on individuals (from the epidemiological to a clinical one).

As one of the possible backgrounds of the results presented here, an alcohol hypothesis has been discussed, since its consumption both harms the liver and decreases magnesium concentration in the body. Because the organism regains the Mg content via water ingestion rich with the element, some unknown more intensive protection mechanisms of the organ related to magnesium against alcohol damage may exist.

Certainly, these new data and the suggested hypothesis require further corroboration in order to assess any scientific speculations and to conclude what vital role in the protection against liver cancer, or in the antineoplastic effects is really played by the drinking water Mg. It is believed, however, that this report will be helpful to indicate the proper investigation direction and will give new light both in water magnesium epidemiology and oncology of liver cancer.

Acknowledgement

We would like to thank the two anonymous Referees for the very constructive comments on an earlier draft.

References

1. Feder GL, Hopps HC. Variations in drinking water quality and the possible effects on human health. Trace Subst Environm Health 1981; 15: 96-103.

2. Pocock SJ, Cook DG, Shaper AG. Analysing geographic variation in cardiovascular mortality: methods and results. J Royal Stat Soc 1982; A, 145: 313-41 (with Discussion).

3. Sauvant MP, Pepin D. Drinking water and cardiovascular disease – A review. Food Chem Toxic 2002; 40: 1311-25.

4. Durlach J, Bara M, Guiet-Bara A. Magnesium level in drinking water: its importance in cardiovascular risk. In: Magnesium in health and disease, eds. Y. Itokawa, J. Durlach, 1989; pp. 173-82. London: John Libbey.

5. Seelig MS. Epidemiology of water magnesium evidence of contributions to health. The Magnesium Web Site, http://www.mgwater.com/epidem.shtml, 2002.

6. Rayssiguier Y, Mazur A, Durlach J, eds. Advances in magnesium research: nutrition and health. 2001; London: John Libbey.

7. Seelig MS. Magnesium in oncogenesis and in anti-cancer treatment: interaction with minerals and vitamins. The Magnesium Web Site, http://www.mgwater.com/cancer.shtml, 1999.

8. Durlach J, Rinjard P, Bara M, Guiet-Bara A, Collery P. Données nouvelles sur les rapports entre magnésium et cancer. In: Magnesium – Physiologische Aspekte für die Praxis, eds. B. Lassere, 1987; pp. 26-45. Hedingen/Zrich: Pascentia Verlag.

9. Durlach J. Magnesium and its relationship to oncology. In: Metal ions in biological systems. Vol. 26. Compendium on magnesium and its role in biology, nutrition and physiology, eds. H. Sigel, A. Sigel, 1990; pp. 549-78. New York – Basel: Marcel Dekker, Inc.

10. Yang CY, Chiu HF. Calcium and magnesium in drinking water and risk of death from rectal cancer. Intl J Cancer 1998; 77: 528-32.

11. Yang CY, Chiu HF, Chiu JF, Tsai SS, Cheng MF. Calcium and magnesium in drinking water and risk of death from colon cancer. Japn J Cancer Res 1997; 88: 928-33.

12. Yang CY, Cheng MF, Tsai SS, Hsieh YL. Calcium, magnesium, and nitrate in drinking water and gastric cancer mortality. Japn J Cancer Res 1998; 89: 124-30.

13. Yang CY, Hung CF. Colon cancer mortality and total hardness levels in Taiwan's drinking water. Arch Environm Contam Toxicol 1998; 35: 148-51.

14. Yang CY, Tsai SS, Lai TC, Hung CF, Chiu HF. Rectal cancer mortality and total hardness levels in Taiwan's drinking water. Environm Res 1999; 80: 311-6.

15. Yang CY, Chiu HF, Tsai SS, Cheng MF, Lin MC, Sung FC. Calcium and magnesium in drinking water and risk of death from prostate cancer. J Toxicol Environm Health 2000; 60: 17-26.

16. Bazikian KL. The significance of magnesium salts in oncology. (Epidemiological, experimental and clinical observations). Proc of First Intl Sympos on Magnesium 1971, ed. J. Durlach, 1973; pp. 593-606. Vittel, France.

17. Bolviken B, Flaten TP, Zheng C. Relations between nasopharyngeal carcinoma and magnesium and other alkaline earth elements in soils in China. Med Hypotheses 1997; 48: 21-25.

18. Walter SD, Miller CT, Lee JAH. The use of age-specific mean cohort slopes in the analysis of epidemiological incidence and mortality data. J Royal Stat Soc 1976; A, 139: 227-45.

19. Tukiendorf A. Magnesium in drinking water and liver cancer morbidity – a possible relation ? Centr Europ J Publ Health 2002; 4: 157-62.

20. Parkin DM, Muir CS, Whelan SL, Gao Y-T, Ferlay J, Powell J, eds. Cancer Incidence in Five Continents. Lyon: IARC Scientific Publications, VI, 1992; 120: 690-3.

21. Bernardinelli L, Pascutto C, Montomoli C, Komakec J, Gilks W. Ecological regression with errors in covariates: an application In: Disease mapping and risk assessment for public health, eds. A. Lawson, A. Biggeri, D. Böhning, E. Lesaffre, J.-F. Viel, R. Bertollini, 1999; pp. 329-48. Chichester: Wiley.

22. Spiegelhalter D, Thomas A, Best N, Gilks W. BUGS: Examples, Volume 1, 1996; pp. 10-4. Cambridge: Medical Research Council – Biostatistics Unit, http://www.mrc-bsu.cam.ac.uk/bugs/

23. Congdon P. Bayesian Statistical Modelling, 2001; pp. 125-6. Chichester: Wiley.

24. Spiegelhalter D, Thomas A, Best N, Lunn D. WinBUGS Version 1.4. Cambridge: Medical Research Council – Biostatistics Unit, http://www.mrc-bsu.cam.ac.uk/bugs/, 2002.

25. Smith AFM, Roberts GO. Bayesian computation via the Gibbs sampler and related Markov chain Monte Carlo methods. J Royal Stat Soc 1993; B, 55: 3-23.

26. National Cancer Institute. What You Need To Know About Liver Cancer, http://www.nci.nih.gov/cancerinfo/wyntk/liver, 2003.

27. Anderson TW, Neri L, Schreiber GB, Talbot F, Zdrojewski A. Ischemic heart disease, water hardness and myocardial magnesium. Can Med Assoc J 1975; 113: 199-203.

28. Anderson TW, Leriche WH, Hewitt D, Neri LC. Magnesium, water hardness, and heart disease. In: Magnesium in Health and Disease, eds. M. Cantin, M.S. Seelig, 1980; pp. 565-71. New York: Spectrum.

29. Durlach J, Bara M, Guiet-Bara A. Magnesium level in drinking water and cardiovascular risk factor: a hypothesis. Magnesium 1985; 4: 5-15.

30. Hankin JH, Margen S, Goldsmith NF. Contribution of hard water to calcium and magnesium intakes of adults. J Am Dietet Assoc 1970; 56: 212-24.

31. Sharrett AR. The role of chemical constituents of drinking water in cardiovascular diseases. Am J Epidemiol 1979; 110: 401-19.

32. Hopps HC, Feder GL. Chemical qualities of water that contribute to human health in a positive way. Sci Tot Environm 1986; 54: 207-16.

33. Flink EB, Omar M, Shane SR. Alcoholism, liver disease and magnesium. Magn Bull 1981; 3: 209-16.

34. Duffy JC, Latcham RW. Liver cirrhosis mortality in England and Wales compared to Scotland: an age-period-cohort analysis 1941-1981. J Royal Stat Soc 1986; A, 149: 45-59.