Introduction

Trace elements are essential micronutrients for many physiological processes and are involved in many pathological changes in tissues [1]. The concentration of these elements in the thyroid gland is higher than in any other tissues [2]. More than 20 chemical elements influence the normal physiology of the thyroid gland [2, 3]. Previous meta-analysis study provides evidence on cancer-specific tissue zinc level alteration and suggests association between low zinc levels with most tumors, especially thyroid cancer [4].

Normal thyroid function depends on the presence of many trace elements for both synthesis and metabolism of thyroid hormones [5]. It is frequently assumed that trace element disturbances, such as selenium (Se), copper (Cu), and magnesium (Mg), are important risk factors for thyroid cancer [6–8]. Se incorporates into selenoproteins within cells, which is vital for the removal of damaging peroxides, reduction of oxidized proteins and membranes, regulation of intracellular reduction–oxidation signaling, and thyroid hormone metabolism [9]. Se concentration is higher within the thyroid than in other tissues, and Se is important for thyroid hormone metabolism [10, 11]. Cu has many important physiological functions as redox active element in maintaining thyroid activity and lipid metabolism. Cu is incorporated in the production of hemoglobin, myelin, and melanin and is essential for thyroid gland functioning [12], by stimulating production of thyroxin hormone (T4) and preventing over absorption of T4 by controlling the calcium levels. Cu can act both as antioxidant and pro-oxidant. As an antioxidant, Cu scavenges or neutralizes free radicals and may reduce or prevent some of the damage caused by them [13–15]. A high concentration of Cu can induce growth proliferation and cancer by damaging deoxyribonucleic acid (DNA) with toxic free hydroxyl radicals [16]. Experimental and clinical data have suggested that Mg deficiency might be associated with inflammation and/or increased levels of free radicals, which might lead to oxidative DNA damage and cancer formation [17, 18]. Mg is known to stabilize the structure of nucleic acids and is a vital cofactor of enzymes involved in DNA replication, repair, and gene expression; thus, any Mg deficiency may contribute to defect in these systems and appearance of DNA mutations, which may thereby lead to tumor-genesis [19, 20]. Some clinical studies show that there is a significant relationship between serum levels of Se, Cu, and Mg with thyroid cancer, indicating that patients with thyroid cancer have lower serum levels of Se and Mg and higher Cu levels than healthy control [15, 21–26]. However, other clinical studies find no association between serum level of Se and Cu with thyroid cancer [27]. Although serum Se, Cu, or Mg disturbance is plausibly linked to thyroid cancer, the inconsistency among the findings of previous studies precludes definitive recommendations at present.

Meta-analysis is an important tool for revealing trends that might not be apparent. Therefore, we performed a comprehensive and critical review of the studies, in order to draw a more clear and evidence-based conclusion on the association between serum levels of selenium, copper, and magnesium with thyroid cancer.

Methods

Search Strategy

We searched all English written articles indexed in PubMed published up to January 2015. Literature searches were performed using medical subject heading (MeSH) or free text words. The searching keywords were (“serum selenium” OR selenium OR “serum copper” OR copper OR “serum magnesium” OR magnesium) AND thyroid cancer. Reference lists of all eligible studies were screened to identify potentially eligible studies. Emails were sent to the authors of identified studies for additional information if necessary.

Selection Criteria

Three authors (Fei Shen, Jie Cao, and Bo Xu) conducted the search independently. Titles and abstracts were screened for subject relevance. Studies that could not be definitely excluded based on abstract information were also selected for full text screening. Two authors (Fei Shen and Jie Cao) independently selected eligible studies for inclusion possibility. Where there was a disagreement for study inclusion, a discussion was held (with Bo Xu) to reach a consensus. The included studies should meet the following criteria: (1) human study; (2) case–control study or cohort study; (3) studies focusing on the association between serum levels of Se, Cu, or Mg and thyroid cancer; (4) studies providing serum levels of Se, Cu, or Mg for both subjects with thyroid cancer and healthy controls; and (5) subjects with no other diseases and no drugs intake which might influence the serum levels of Se, Cu, or Mg. Exclusion criteria included the following: (1) in vitro or laboratory study; (2) animal study; (3) review or case report; (4) studies not providing serum levels of Se, Cu, or Mg for both subjects with thyroid cancer and healthy controls; (5) subjects with diseases/drugs (for Se: Keshan disease, Kashin–Beck disease, heart disease, high blood pressure, and breast cancer; for Cu: hepatocirrhosis, hepatitis, obstructive jaundice, anemia, leukemia, kidney disease, Wilson’s syndrome, Addison disease, curly hair syndrome, oral zinc intake, oral Cu intake, and contraceptive medicine intake; for Mg: kidney disease, acute myocardial infarction, hypofunction of thyroid gland, hypoparathyroidism, hyperthyroidism, hyperparathyroidism, Addison’s disease, multiple myeloma, acute viral hepatitis, amoebic liver abscess, liver cirrhosis, arthritis, and oral Mg intake), which might influence the serum levels of Se, Cu, or Mg; and (6) sample size <10.

Data Extraction and Quality Assessment

The following information was extracted from each included study: first author’s family name; year of publication; type of study; country; demography of subjects (age and number of patients); data on serum levels of Se, Cu, or Mg; and type of trace elements measurement. Two authors (Wen-Song Cai and Jiang-Lin Li) independently extracted data using a standard form.

Two authors (Fei Shen and Zhe Feng) assessed the quality independently. The qualities of all included studies were assessed using the Newcastle–Ottawa Scale (NOS) (Supplement Material Figure S1) [28]. Using the tool, each study is judged on eight items, categorized into three groups: the selection of the study groups, the comparability of the groups, and the ascertainment of either the exposure or outcome of interest for case–control or cohort studies, respectively. Stars awarded for each quality item serve as a quick visual assessment. Stars are awarded such that the highest quality studies are awarded up to nine stars. Studies were graded as good quality if they awarded six to nine stars, fair if they awarded three to five stars, and poor if they awarded less than three stars.

Statistical Analysis

The extracted data were used to perform meta-analysis to obtain the standardized mean difference (SMD) and 95% confidence intervals (CI). The SMDs were calculated using either fixed-effects models or, in the presence of heterogeneity, random-effects models. Heterogeneity between studies was tested through the chi-square and I-square tests. If the I 2 value was >50%, and the p < 0.05, the meta-analysis was considered as homogeneous. Subgroup analyses involve splitting all the participant data into subgroups, often so as to make comparisons between them, which can be done as a means of investigating heterogeneous results. Subgroup analyses stratified by the ethnicity/country and type of measurement were used to identify associations between serum levels of Se, Cu, or Mg and other relevant study characteristics as possible sources of heterogeneity. Publication bias was defined as the publication or nonpublication of studies depending on the direction and statistical significance of the results and was measured using Begg’s test and visualization of funnel plot. The stability of the study was also detected by sensitivity analysis, through re-meta-analysis with one involved study excluded each time. All statistical analyses were performed with Stata version 11.0 (StataCorp, College Station, TX, USA).

Results

Literature Search

The literature search yielded a total of 243 primary articles. These articles were included for full-text assessment, of which 235 were excluded for one of the following reasons: (1) irrelevant to our topic (n = 102), (2) non-original studies (reviews, etc.) (n = 67), (3) non-human studies (n = 63), (4) articles not providing serum level of Se, Cu, or Mg for both subjects with thyroid cancer and healthy controls (n = 3). Overall, 8 eligible articles with 1291 subjects met the inclusion criteria for meta-analysis: 4 articles with 6 case–control studies for Se, 4 articles with 5 case–control studies for Cu, and 2 articles with 3 case–control studies for Mg [15, 21–27]. Of note, two articles involved more than one trace element and were included in more than one group. A flow diagram of the study selection process is presented in Fig. 1.

Fig. 1
figure 1

Flow diagram of screened and included papers

Full size image

Study Characteristics and Quality Assessment

The characteristics of the included studies and the results of the quality assessment are listed in Table 1. The earliest study was published in 1989, and the latest in 2012. By geographic location, 8 articles with 14 case–control studies were conducted in 6 different countries (Norway, Poland, Austria, Turkey, China, and Kuwait). The number of subjects in each study ranged from 33 to 627. Five articles with ten case–control studies measured Se, Cu, or Mg concentration by atomic absorption spectrometry (AAS); two articles with two case–control studies measured Se or Cu concentration by total reflection X-ray fluorescence (TRXRF); while one article with two case–control studies measured Cu or Mg concentration by atomic emission spectrometry (AES). The overall study quality averaged six stars on a scale of 0–9.

Table 1 Characteristics of subjects in eligible studies
Full size table

Serum Se and Thyroid Cancer

The random-effects meta-analysis results indicated that patients with thyroid cancer had lower serum levels of Se than the healthy controls [SMD = −0.485, 95%CI = (−0.878, −0.092), p = 0.016]. The six sets of results showed a statistically significant amount of heterogeneity (I 2 = 98.3%, p < 0.001) (Fig. 2a).

Fig. 2
figure 2

Forest plots of studies in Se, Cu, and Mg levels for subjects with thyroid cancer versus healthy controls. The combined SMD and 95 % confidence intervals (CIs) were calculated using the random-effects model. a For serum Se, b for serum Cu, and c for serum Mg

Full size image

The subgroup analysis showed that geographical location and type of Se measurement had an influence on the serum levels of Se in thyroid cancer and healthy controls. Further subgroup analysis stratified by geographical location indicated that subjects with thyroid cancer had higher serum Se levels than healthy controls in Norway [SMD = −0.410, 95%CI = (−0.758, −0.062), p = 0.021] and Austria [SMD = −0.549, 95%CI = (−0.743, −0.355), p < 0.001], but not in Poland [SMD = −0.417, 95%CI = (−1.724, 0.891), p = 0.532] (Fig. 3a). The difference of serum Se levels between thyroid cancer and healthy controls measured by TRXRF [SMD = −1.707, 95%CI = (−2.290, −1.123), p < 0.001] was higher than that by AAS [SMD = −0.331, 95%CI = (−0.612, −0.050), p = 0.021] (Fig. 3b). The further subgroup analysis stratified by type of measurement can successfully reduce the heterogeneity of studies (Table 2).

Fig. 3
figure 3

Subgroup analyses of studies in Se and Cu levels for subjects with thyroid cancer versus healthy controls. a Serum Se stratified by country. b Serum Se stratified by type of measurement. c Serum Cu stratified by country. d Serum Cu stratified by type of measurement

Full size image
Table 2 Differences between studies by subgroup analysis
Full size table

Serum Cu and Thyroid Cancer

The random-effects meta-analysis results indicated that patients with thyroid cancer had higher serum levels of Cu than the healthy controls [SMD = 2.372, 95%CI = (0.945, 3.799), p = 0.001]. The five sets of results showed a statistically significant amount of heterogeneity (I 2 = 96.2%, p < 0.001) (Fig. 2b).

The subgroup analysis showed that geographical location and type of Cu measurement had an influence on the serum levels of Cu in thyroid cancer and healthy controls. Further subgroup analysis stratified by geographical location indicated that subjects with thyroid cancer had higher serum Cu levels than healthy controls in China [SMD = 1.571, 95%CI = (1.121, 2.020), p < 0.001] and Turkey [SMD = 0.977, 95%CI = (0.521, 1.432), p < 0.001], but not in Poland [SMD = 3.471, 95%CI = (−0.056, 6.997), p = 0.054] (Fig. 3c). The difference of serum Cu levels between thyroid cancer and healthy controls measured by TRXRF [SMD = 9.645, 95%CI = (7.957, 11.333), p < 0.001] was higher than that by AAS [SMD = 0.785, 95%CI = (0.461, 1.109), p < 0.001] and AES [SMD = 1.571, 95%CI = (1.121, 2.020), p < 0.001] (Fig. 3d). The further subgroup analysis stratified by type of measurement can successfully eliminate the heterogeneity of studies (Table 2).

Serum Mg and Thyroid Cancer

The fixed-effects meta-analysis results indicated that patients with thyroid cancer had lower serum levels of Mg than healthy controls [SMD = −0.795, 95%CI = (−1.092, −0.498), p < 0.001]. The three sets of results showed no significant amount of heterogeneity (I 2 = 0, p = 0.725) (Fig. 2c).

Publication Bias and Sensitivity Analysis

Publication bias was measured using Begg’s tests and visualization of funnel plots. There was no evidence of publication bias for Se (Begg’s test, p = 0.604), Cu (Begg’s test, p = 0.806), and Mg (Begg’s test, p = 0.296) (Fig. 4). Sensitivity analysis showed that excluding the studies performed by Kucharzewski in 2002 and in 2003 can reduce the heterogeneity of studies for serum Se and Cu, respectively (Table 3).

Fig. 4
figure 4

Funnel plots of studies in Se, Cu, and Mg levels for subjects with thyroid cancer versus healthy controls. a For serum Se. b For serum Cu. c For serum Mg

Full size image
Table 3 The heterogeneity of the included studies through sensitivity analysis
Full size table

Discussion

The result of random-effects meta-analysis indicated that the low serum levels of Se was associated with thyroid cancer. This may reflect the increased risk of thyroid cancer in patients with low serum Se levels and confirm the results of other studies. In a review of an article conducted by Rayman, the relationship between selenium and cancer was examined. This review showed that selenium has anticancer activities [29]. Many studies have shown that there is an association between selenium diet and cancer risk [6, 10, 23, 30–32]. Selenium nutritional supplements may reduce incidence of cancer [33]. Jonklaas et al. [9] has shown that serum selenium concentrations were inversely correlated with thyroid cancer stage. Although the specific selenium anticarcinogenic mechanisms are not yet fully known, multiple mechanisms have been proposed to explain these selenium characteristics. Actually, antioxidant properties of selenoenzymes are relevant in carcinogenesis and tumor progression [34]. Selenium is present in glutathione peroxidase, which protects the DNA and main cellular components from the damage of the free radicals by decreasing ROS generation; in this way, levels of dietary antioxidant vitamins and carotenoids that affect antioxidant selenoproteins modify the effect of selenium on cancer risk [35–37]. In addition, selenium enhances tumor-suppressor protein p53 activates that inhibits proliferation, increases DNA repair, and promotes apoptosis [38, 39].

The outcome of this meta-analysis also suggested that patients with thyroid cancer had higher serum levels of Cu than the healthy controls. Cu is believed to be the switch that turns on the angiogenesis process in tumor cells. Cu is cofactor of superoxide dismutase 1, which prevents the onset and progression of tumors through mechanisms of cell protection against free radicals production [40]. A high concentration of Cu can induce growth proliferation and cancer by damaging DNA with toxic free hydroxyl radicals [16]. Abnormally high serum Cu levels are found in the patients with many types of progressive tumors, making Cu an obligatory cofactor in angiogenesis process [41]. In a review by Blazewicz et al. [42], the concentration of Cu was significantly higher in the healthy thyroid in comparison with the group of benign thyroid disease. Another study found that the postoperative serum Cu levels were significantly decreased compared to those of preoperative in the benign thyroid disease [25].

The present study still found that low serum level of Mg was associated with thyroid cancer. It has been found that magnesium could impact carcinogenesis by two mechanisms [20]. Mg deficiency might be associated with inflammation and/or increased levels of free radicals. Mg-deficient animals show an increased susceptibility to in vivo oxidative stress, and their tissues are more susceptible to in vitro peroxidation [18]. There is convincing evidence from other animal studies that Mg could exert a protective effect in the early stages of carcinogenesis [17]. Mg inhibits lead- and nickel-induced carcinogenesis in the rat kidney and 3-methylcholantrene-induced fibro-sarcomas in rats [43, 44]. It has been reported that Mg supplementation reduces the incidence of experimentally induced colon cancer in animals, which might be related to a decrease in colonic epithelial cell proliferation [45, 46]. Some studies found that a diet poor in magnesium increases the incidences of thymic tumors and leukemias [47, 48].

To the best of our knowledge, this is the first meta-analysis to estimate the association between serum levels of Se, Cu, and Mg with thyroid cancer. However, the possible limitations of our study must be considered. First, our results showed strong heterogeneity among the studies for Cu and Se. Heterogeneity indicates differences in results across the studies. In present study, the further subgroup analysis indicated that the type of Cu and Se measurement was the source of heterogeneity. We found that the further subgroup analysis stratified by type of measurement can successfully eliminate the heterogeneity of studies for Cu, and reduce the heterogeneity of studies for Se. Second, only 1291 subjects from 8 articles and no randomized clinical trial included in the meta-analysis might weaken the quality of the results. Despite these limitations, our findings point out new directions for future research, like what is the compound effect of multiple risk factors on thyroid cancer. For instance, what is the risk of thyroid cancer with both Se and Cu disturbances? To answer this question, several well designed studies with adequate control for confounding factors should be considered. In addition, our results showed the ethnic/geographical paradox; therefore, a trans-regional multicenter study is needed for the investigation of the interrelationship between Se, Cu, and Mg with thyroid cancer of different human races or regions.

In conclusion, this meta-analysis supports a significant association between serum levels of Se, Cu, and Mg with thyroid cancer.