Open Access

Metallomics study using hair mineral analysis and multiple logistic regression analysis: relationship between cancer and minerals

  • Hiroshi Yasuda1Email author,
  • Kazuya Yoshida1,
  • Mitsuru Segawa1,
  • Ryoichi Tokuda1,
  • Toyoharu Tsutsui1,
  • Yuichi Yasuda2 and
  • Shunichi Magara3
Environmental Health and Preventive Medicine200914:92

https://doi.org/10.1007/s12199-009-0092-y

Received: 25 December 2008

Accepted: 14 May 2009

Published: 26 June 2009

Abstract

Objectives

The objective of this metallomics study is to investigate comprehensively some relationships between cancer risk and minerals, including essential and toxic metals.

Methods

Twenty-four minerals including essential and toxic metals in scalp hair samples from 124 solid-cancer patients and 86 control subjects were measured with inductively coupled plasma mass spectrometry (ICP-MS), and the association of cancer with minerals was statistically analyzed with multiple logistic regression analysis.

Results

Multiple logistic regression analysis demonstrated that several minerals are significantly correlated to cancer, positively or inversely. The most cancer-correlated mineral was iodine (I) with the highest correlation coefficient of r = 0.301, followed by arsenic (As; r = 0.267), zinc (Zn; r = 0.261), and sodium (Na; r = 0.190), with p < 0.01 for each case. In contrast, selenium (Se) was inversely correlated to cancer (r = −0.161, p < 0.05), followed by vanadium (V) (r = −0.128). Multiple linear regression value was highly significantly correlated with probability of cancer (R 2 = 0.437, p < 0.0001), and the area under the receiver-operating characteristic (ROC) curve was calculated to be 0.918. In addition, using contingency table analysis and the chi-square test, the precision of discrimination for cancer was estimated to be 0.871 (chi-square = 99.1, p < 0.0001).

Conclusions

These findings suggest that some minerals such as arsenic, selenium, and probably iodine, zinc, sodium, and vanadium contribute to regulation of cancer and also that metallomics study using multiple logistic regression analysis is a useful tool for estimating cancer risk.

Keywords

CancerMetallomicsArsenicSeleniumMultiple logistic regression analysis

Introduction

Cancer has been the first cause of death in developed countries, and worldwide there are more than 10 million new cancer cases each year [1, 2]. Therefore, a large number of experimental and epidemiological studies have been undertaken to identify potential risk factors for cancer, among which the association with some carcinogenic metals such as cadmium, nickel, beryllium or arsenic has been reported [25]. However, there is no metallomics study comprehensively examining the relationship between cancer risk and multiple biometals.

For the last few years, we have measured scalp hair minerals in over 60,000 subjects, ranging from infant to elderly, in order to assess exposure to toxic metals, deficiency/excess profile of essential minerals, and some relationships between minerals and physical or mental disorders [68].

The purpose of this study is to examine comprehensively some roles of toxic and essential minerals in cancer risk. Twenty-four kinds of minerals in hair samples were measured for cancer and control subjects aged 20–80 years, and the association between cancer and these multiple minerals was examined with multiple logistic regression analysis.

Materials and methods

Materials

On the basis of informed consent, scalp hair samples (about 0.2 g) were collected from 124 solid-cancer patients suffering from breast (n = 28), stomach (n = 22), lung (n = 11), colon (n = 10), prostate (n = 9), liver (n = 7), pancreas (n = 5), uterus (n = 5), ovary (n = 4), esophagus (n = 4), malignant lymphoma (n = 4), kidney (n = 3), thyroid (n = 2) or other tumors who had been cared for and treated in Somon Hachioji Clinic (Table 1). In addition, control hair samples from healthy subjects aged 20–80 years old from the Tokyo metropolitan district were collected. The study protocol was approved by the ethical committee of this laboratory. All of the data obtained are held securely in such a form as to ensure anonymity.
Table 1

Tested subjects (n = 210)

 

No.

Age (mean ± SD), years

Male

 Cancer

52

60.9 ± 12.2

 Control

53

51.6 ± 11.9

Female

 Cancer

72

54.9 ± 10.6

 Control

33

53.7 ± 16.0

Hair mineral analysis

Hair sample of 75 mg was weighed into 50 ml plastic tube, and washed twice with acetone and then with 0.01% Triton solution, in accordance with the procedures recommended by the Hair Analysis Standardization Board [9]. The washed hair sample was mixed with 10 ml 6.25% tetramethylammonium hydroxide (TMAH, Tama Chemical) and 50 μl 0.1% gold solution (SPEX Certi Prep.), and then dissolved at 75°C with shaking for 2 h. After cooling of the solution to room temperature, internal standard (Sc, Ga, and In) solution was added and, adjusting its volume gravimetric, the obtained solution was used for mineral analysis. The mineral concentrations were measured with inductively coupled plasma mass spectrometry (ICP-MS; Agilent-7500ce) by the internal standard method [68, 10], and are expressed as ng/g hair (ppb). For quality control of the mineral analysis, the human hair certified reference material supplied from the National Institute for Environmental Studies of Japan (NIES CRM no. 13) [11] was used.

Statistical analysis

The hair mineral concentrations were log-normal distributed, and geometric rather than arithmetic means were used as representative of hair mineral levels. For statistical analysis, the values of mineral levels were converted to logarithms. The relation between cancer risk and 24 minerals was investigated with multiple logistic regression analysis (JMP6; SAS Institute).

Results

Twenty-four kinds of mineral concentrations in the scalp hair of 124 solid-cancer patients and 86 control subjects were measured, and the association between cancer and each mineral was statistically examined with multiple logistic regression analysis. Some minerals were found to be significantly correlated to cancer, positively or inversely, and the multiple coefficient of determination (R 2) for estimating the probability of cancer risk was estimated to be 0.465 (Table 2). The most cancer-correlated mineral with the highest partial correlation coefficient was iodine (I; r = 0.301), followed by arsenic (As; r = 0.267), zinc (Zn; r = 0.261), and sodium (Na; r = 0.190), with p < 0.01 for each element. These correlation coefficients were higher than that of ageing, a well-known risk factor of cancer. Regression charts for the best three minerals exhibiting significantly high positive correlation with cancer are shown in Fig. 1a–c. In contrast, the most inverse-correlated mineral to cancer was selenium (Se; r = −0.161) (Fig. 1d), followed by manganese (Mn) and vanadium (V) (Table 2).
Table 2

Relationship between cancer risk and minerals (multiple logistic regression analysis)

 

Part. coeff.

Corr. coeff.

p value

Significance

 

Part. coeff.

Corr. coeff.

p value

Significance

I

0.3012

0.3484

0.0000

****

Se

−0.1614

−0.1114

0.0282

*

As

0.2668

0.2319

0.0002

***

Mn

−0.1533

0.1337

0.0372

*

Zn

0.2612

0.2754

0.0003

***

V

−0.1279

−0.1237

0.0827

 

Na

0.1903

0.2172

0.0095

**

K

−0.0722

0.0747

0.3291

 

Br

0.1839

0.2213

0.0122

*

Mg

−0.0661

0.0793

0.3711

 

Fe

0.1275

0.2623

0.0837

 

Ni

−0.0456

0.0003

0.5377

 

Al

0.0873

0.0189

0.2375

 

B

−0.0372

−0.0204

0.6149

 

Cr

0.0849

0.0821

0.2504

 

Mo

−0.0227

−0.1123

0.7595

 

Co

0.0804

0.1554

0.2767

 

P

−0.0140

0.0918

0.8499

 

Ca

0.0766

0.0947

0.3000

      

Cu

0.0617

0.1186

0.4043

      

Ge

0.0551

0.2917

0.4566

      

Pb

0.0072

0.0486

0.9225

      

Cd

0.0051

0.1087

0.9446

      

Hg

0.0016

-0.0396

0.9833

      

Age

0.1491

0.1955

0.0428

*

     

Determination coefficient: 0.465

The relationship between cancer and minerals was analyzed with multiple logistic regression analysis of data for 24 minerals from 124 cancer patients and 86 control subjects

Part. coeff. partial correlation coefficient by logistic multiple regression analysis, Corr. coeff. simple correlation coefficient by simple regression analysis

Statistically significant association is shown with asterisk marks: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001

Fig. 1

Positive or inverse relationship between cancer and minerals in multiple logistic regression analysis (a iodine, b arsenic, c zinc, d selenium). The relationship between minerals and probability of cancer was obtained by logistic multiple regression analysis of data for 24 minerals from 124 cancer (open square) and 86 control (dots) subjects. The abscissa and ordinate express the logarithmic values of hair mineral (a iodine, b arsenic, c zinc, d selenium) level and probability of cancer, respectively. The blue and red marks show male and female subjects, respectively

Figure 2 exhibits a significantly high correlation of cancer with multiple linear regression value obtained by the following equation (R 2 = 0.447, p < 0.0001):
Fig. 2

Significantly high relationship between cancer and multiple linear regression value in multiple logistic regression analysis. The relationship between multiple linear regression value and probability of cancer was obtained by multiple logistic regression analysis of data for 24 minerals from 124 cancer (open square) and 86 control (dots) subjects. The abscissa and ordinate express the calculated multiple linear regression value of the tested subjects and the probability of cancer, respectively. The blue and red marks show male and female, respectively

$$ {\text{Multiple linear regression}} = \pm 0.70\,\left( {\text{Sex}} \right) + 0.023 \times {\text{Age}} + 1.75 \times {\text{Log I}} + 3.59 \times {\text{Log As}} + 5.38 \times {\text{Log Zn}} + 6.67 \times {\text{Log Fe}} + 2.04 \times {\text{Log Na}}-6.47 \times {\text{Log Se}}-1.48 \times {\text{Log K}}-1.47 \times {\text{Log Mn}}-50.88 $$
Thus, the area under the receiver-operating characteristic (ROC) curve for discriminating cancer was estimated as 0.918 for multiple linear regression, and significantly greater than 0.704 for iodine, 0.647 for arsenic, and 0.662 for zinc. The accuracy of the discrimination of cancer was also assessed using contingency table analysis and the chi-square test (Table 3). From the contingency table, the precision of discriminating cancer patients was estimated to be 0.871 (108/124 = 87.1%; chi-square = 99.1, p < 0.0001).
Table 3

Contingency table

 

No.

Judgment

Negative

Positive

Cancer

124

  

 Count

 

16

108

 Row %

 

0.129

0.871

Control

86

  

 Count

 

68

18

 Row %

 

0.791

0.209

 

210

84

126

Using contingency table analysis and the chi-square test, the accuracy and precision of discrimination of cancer patients was assessed

chi-square = 99.1 (p < 0.0001)

The mean hair arsenic level in the patient group was significantly higher than that in the control group, in both sexes, and gender difference was observed (Table 4).
Table 4

High arsenic concentrations in cancer patients

 

No.

Hair arsenic concentration

p value

Geomean

Log As (mean ± SD)

Male

 Control

53

61.9

1.7914 ± 0.2425

 

 Cancer

52

123.7

2.0924 ± 0.4173

0.0000****

Female

 Control

33

31.4

1.4975 ± 0.4324

 

 Cancer

72

52.9

1.7231 ± 0.4058

0.0141*

The hair arsenic concentrations in the 124 cancer patients are compared with those in the 86 control subjects. Each value of arsenic concentration was converted to its logarithm, and the geometric mean was used as the average hair arsenic level. Statistical significance was determined using Welch’s t test

Statistically significant difference is shown with asterisk marks: *p < 0.05 and ****p < 0.0001

Discussion

Hair mineral analysis has been used in forensic medicine, in screening populations for toxic metal poisoning, and in monitoring environmental pollutants [6, 1214]. Furthermore, its clinical use has been tried for diagnosis of some diseases and symptoms [7, 8], but reliability remained to be confirmed [15, 16]. Our previous studies demonstrated that hair levels of some toxic metals such as lead, cadmium, and aluminum are extraordinarily high in infants and children [6, 17]. In addition, we reported that autistic children are suffered from global mineral deficiency of various essential trace elements and also that, in autistic children, there are some subgroups with high accumulation of mercury, cadmium, and lead, or aluminum and iron [7].

In the present metallomics study for cancer patients, the multiple logistic regression analysis exhibited a significant association between some minerals and cancer. In particular, carcinogenic arsenic and its antagonistic element, selenium, was confirmed to correlate to cancer risk positively and inversely, respectively (Fig. 1b, d, Table 2).

A significantly high close relationship between probability of cancer and multiple linear regression value was demonstrated (R 2 = 0.447, p < 0.0001) (Fig. 2), and the corresponding area under the receiver-operating characteristic (ROC) curve for diagnosis of cancer was estimated to be 0.918. Furthermore, this close relationship was confirmed by the results of contingency table analysis: the precision of discrimination for cancer patients was estimated to be 0.871 (108/124 = 87.1%) and the accuracy to be 0.838 [(108 + 68)/210 (chi-square = 99.1, p < 0.0001)] (Table 3).

These findings suggest that there is a significantly high close relationship between probability of cancer and trace bioelements, including toxic and essential minerals.

The multiple coefficient of determination (= R 2) for estimating the probability of cancer risk was calculated to be 0.465 (p < 0.0001). This means that the cancer risk of an individual case can be evaluated with 46.5% probability by using hair mineral analysis and multiple logistic regression analysis.

From the regression curve shown in Fig. 1b, the probability of cancer in a subject with a logarithmic arsenic value of 2.5 was estimated to be about 80%, and about 30% increment of the probability was estimated to be associated with a tenfold increase in hair arsenic level from 10 to 100 ppb (odds ratio 36.2; 95% CI 7.1–234). Arsenic has been known to be toxic and induce chromosomal aberrations and carcinogenesis via various cellular changes including alterations in cell differentiation and proliferation [4, 18, 19]. Cells exposed to arsenic have been shown to increase cellular tyrosine phosphorylation, which is related to the aberrant cell signaling and uncontrolled cell growth associated with cancer development. Some ecologic studies in areas known for high levels of arsenic in drinking water have suggested that this element is associated with increased risks of lung and bladder cancer [4, 2023]. In Japan, much of the arsenic accumulated in the human body is known to derive from dietary intake of arsenic-containing foods, such as seaweeds and shellfishes, and also from environmental exposure.

The most strongly correlated mineral, iodine, possibly makes some contribution to cancer pathogenesis, although its association with cancer remains to be studied further.

Zinc has an important role in cell-mediated immune functions, and zinc deficiency was reported to be associated with increased risk of cancer [2426]. However, in this metallomics study, zinc level in hair was found to be positively correlated with cancer risk (Table 2; Fig. 1c). This discrepancy may be explained by the hypothesis that hair zinc level behaves paradoxically when deficient in the human body. Further studies are needed for clarifying the role of zinc in cancer risk.

Selenium, a representative competitive mineral against arsenic, correlated inversely with cancer risk; the higher the level of this element, the lower the probability of cancer risk (Fig. 1d; Table 2). This finding is consistent with the hypothesis of anticancer activity of selenium. Selenium is the key element of glutathione peroxidase and other selenium proteins, and is essential for counteracting reactive oxygen species, protecting from oxidative stress, and regulating the redox status in cells [2]. This essential mineral also appears to correlate with stimulation of metabolic detoxification and excretion of arsenic [27, 28]. Thus, selenium is expected to reduce the toxicity and cancer risk of arsenic [29, 30], although high body burden of selenium is toxic. Furthermore, selenium is known to have an ability to compete with other toxic metals such as mercury, cadmium, and lead.

The present study suggests that some minerals such as not only arsenic and selenium but also iodine, zinc, sodium, and vanadium play a pivotal role in cancer pathogenesis, indicating that metallomics analysis using multiple logistic regression analysis seems to be a good tool for estimating cancer risk. Further studies for each type/nature of solid cancer are needed for certifying the usefulness of this analysis.

Hair mineral analysis can provide global information for human metallomics study, i.e., the deficiency/excess profile of multiple trace bioelements that contribute to various physiological and/or adverse events, and therefore will be useful as a biomarker for monitoring health and risk assessment.

Conclusion

This metallomics study for cancer patients demonstrated that there is a close relationship between cancer risk and trace bioelements including toxic and essential minerals, and that many patients have some mineral imbalance in their body. These findings suggest the possibility that metallomics study using hair mineral analysis and multiple logistic regression analysis will be a useful tool for estimating cancer risk and also for cancer prevention.

Notes

Declarations

Acknowledgments

The authors sincerely appreciate Emeritus Prof. Toshiro Haga for his useful advice regarding the statistical analysis based on multiple logistic regression analysis in this study.

Authors’ Affiliations

(1)
Research Laboratory, La Belle Vie Inc.
(2)
Health Science Laboratory
(3)
Somon-Hachioji Clinic

References

  1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin. 2005;55:74–108.PubMedView ArticleGoogle Scholar
  2. Silvera SAN, Rohan TE. Trace elements and cancer risk: a review of the epidemiologic evidence. Cancer Causes Control. 2007;18:7–27.View ArticleGoogle Scholar
  3. Drasch G, Schopfer J, Schrauzer GN. Selenium/cadmium ratios in human prostates: Indicators of prostate cancer risk of smokers and nonsmokers, and relevance to the cancer protective effects of selenium. Biol Trace Elem Res. 2005;103:103–7.PubMedView ArticleGoogle Scholar
  4. Cantor KP, Lubin JH. Arsenic, internal cancers, and issues in inference from studies of low-level exposures in human populations. Toxicol Appl Pharmacol. 2007;222:252–7.PubMedView ArticleGoogle Scholar
  5. Boffetta P. Carcinogenicity of trace elements with reference to evaluations made by the International Agency for Research on Cancer. Scand J Work Environ Health. 1993;19(Suppl 1):67–70.PubMedGoogle Scholar
  6. Yasuda H, Yonashiro T, Yoshida K, Ishii T, Tsutsui T. High toxic metal levels in scalp hairs of infants and children. Biomed Res Trace Elem. 2005;16:39–45.Google Scholar
  7. Yasuda H, Yonashiro T, Yoshida K, Ishii T, Tsutsui T. Mineral imbalance in children with autistic disorders. Biomed Res Trace Elem. 2005;16:285–91.Google Scholar
  8. Yasuda H, Yonashiro T, Yoshida K, Ishii T, Tsutsui T. Relationship between body mass index and minerals in male Japanese adults. Biomed Res Trace Elem. 2006;17:316–21.Google Scholar
  9. Cranton EM, Bland JS, Chatt A, Krakovitz R, Wright JV. Standardization and interpretation of human hair for elemental concentrations. J Holistic Med. 1982;4:10–20.Google Scholar
  10. Yasuda H, Yoshida K, Fukuchi K, Tokuda R, Tsutsui T, Yonei Y. Association between aging and minerals in male Japanese adults. Anti Aging Med. 2007;4:38–42.View ArticleGoogle Scholar
  11. Yoshinaga J, Morita M, Okamoto K. New human hair certified reference material for methylmercury and trace elements. Fresenius J Anal Chem. 1997;357:279–83.View ArticleGoogle Scholar
  12. Batzevich VA. Hair trace element analysis in human ecology studies. Sci Total Environ. 1995;164:89–98.PubMedView ArticleGoogle Scholar
  13. Chlopicka J, Zachwieja Z, Zagrodzki P, Frydrych J, Slota P, Krosniak M. Lead and cadmium in the hair and blood of children from a highly industrial area in Poland. Biol Trace Elem Res. 1998;62:229–34.PubMedView ArticleGoogle Scholar
  14. Samanta G, Sharma R, Roychowdhury T, Chakraborti D. Arsenic and other elements in hair, nails, and skin-scales of arsenic victims in West Bengal, India. Sci Total Environ. 2004;326:33–47.PubMedView ArticleGoogle Scholar
  15. Wang CT, Chang WT, Zeng WF, Lin CH. Concentrations of calcium, copper, iron, magnesium, potassium, sodium, and zinc in adult female hair with different body mass indexes in Taiwan. Clin Chem Lab Med. 2005;43:389–93.PubMedView ArticleGoogle Scholar
  16. Seidel S, Kreutzer R, Smith D, McNeel S, Gilliss D. Assessment of commercial laboratories performing hair mineral analysis. J Am Med Assoc. 2001;285:67–72.View ArticleGoogle Scholar
  17. Yasuda H, Yoshida K, Segawa M, Tokuda R, Yasuda Y, Tsutsui T. High accumulation of aluminum in hairs of infants and children. Biomed Res Trace Elem. 2008;19:57–62.Google Scholar
  18. Bates MN, Smith AH, Hopenhayn-Rich C. Arsenic ingestion and internal cancers: a review. Am J Epidemiol. 1992;135:462–76.PubMedGoogle Scholar
  19. Hayes RB. The carcinogenicity of metals in humans. Cancer Causes Control. 1997;8:371–85.PubMedView ArticleGoogle Scholar
  20. Chen CJ, Chuang YC, You SL, Lin HY. A retrospective study on malignant neoplasms of bladder, lung and liver in black foot disease endemic area in Taiwan. Br J Cancer. 1986;53:399–405.PubMedGoogle Scholar
  21. Chiou HY, Hsueh YM, Liaw KF, Horng SF, Chiang MH, Pu YS, et al. Incidence of internal cancers and ingested inorganic arsenic: a seven-year follow-up study in Taiwan. Cancer Res. 1995;55:1296–300.PubMedGoogle Scholar
  22. Ferreccio C, Gonzalez PC, Milosavjlevic SV, Marshall GG, Sancha AM, Smith AH. Lung cancer and arsenic concentrations in drinking water in Chile. Epidemiology. 2000;11:673–9.PubMedView ArticleGoogle Scholar
  23. Chen Y, Ahsan H. Cancer burden from arsenic in drinking water in Bangladesh. Am J Public Health. 2004;94:741–4.PubMedView ArticleGoogle Scholar
  24. Prasad AS. Zinc: mechanisms of host defense. J Nutr. 2007;137:1345–9.PubMedGoogle Scholar
  25. Yan M, Song Y, Wong CP, Hardin K, Ho E. Zinc deficiency alters DNA damage response genes in normal human prostate epithelial cells. J Nutr. 2008;138:667–73.PubMedGoogle Scholar
  26. Memon AU, Kazi TG, Afridi HI, Jamali MK, Arain MB, Jalbani N, et al. Evaluation of zinc status in whole blood and scalp hair of female cancer patients. Clin Chim Acta. 2007;379:66–70.PubMedView ArticleGoogle Scholar
  27. Christian WJ, Hopenhayn C, Centeno JA, Todorov T. Distribution of urinary selenium and arsenic among pregnant women exposed to arsenic in drinking water. Environ Res. 2006;100:115–22.PubMedView ArticleGoogle Scholar
  28. Chen Y, Hall M, Graziano JH, Slavkovich V, Geen A, Parvez F, et al. A prospective study of blood selenium levels and the risk of arsenic-related premalignant skin lesions. Cancer Epidemiol Biomark Prev. 2007;16:207–13.View ArticleGoogle Scholar
  29. Reid ME, Duffield-Lillico AJ, Garland L, Turnbull BW, Clark LC, Marshall JR. Selenium supplementation and lung cancer incidence: an update of the nutritional prevention of cancer trial. Cancer Epidemiol Biomark Prev. 2002;11:1285–91.Google Scholar
  30. Duffield-Lillico AJ, Dalkin BL, Reid ME, Turnbull BW, Slate FH, Jacobs ET, et al. Selenium supplementation, baseline plasma selenium status and incidence of prostate cancer: an analysis of the complete treatment period of the Nutritional Prevention of Cancer Trial. Br J Urol Int. 2003;91:608–12.Google Scholar

Copyright

© The Japanese Society for Hygiene 2009