Serum and Tissue Zinc in Epithelial Malignancies: A Meta-Analysis

Background and Objectives Current studies give us inconsistent results regarding the association of neoplasms and zinc(II) serum and tissues concentrations. The results of to-date studies using meta-analysis are summarized in this paper. Methods Web of Science (Science citation index expanded), PubMed (Medline), Embase and CENTRAL were searched. Articles were reviewed by two evaluators; quality was assessed by Newcastle-Ottawa scale; meta-analysis was performed including meta-regression and publication bias analysis. Results Analysis was performed on 114 case control, cohort and cross-sectional studies of 22737 participants. Decreased serum zinc level was found in patients with lung (effect size = −1.04), head and neck (effect size = −1.43), breast (effect size = −0.93), liver (effect size = −2.29), stomach (effect size = −1.59), and prostate (effect size = −1.36) cancers; elevation was not proven in any tumor. More specific zinc patterns are evident at tissue level, showing increase in breast cancer tissue (effect size = 1.80) and decrease in prostatic (effect size = −3.90), liver (effect size = −8.26), lung (effect size = −3.12), and thyroid cancer (effect size = −2.84). The rest of the included tumors brought ambiguous results, both in serum and tissue zinc levels across the studies. The association between zinc level and stage or grade of tumor has not been revealed by meta-regression. Conclusion This study provides evidence on cancer-specific tissue zinc level alteration. Although serum zinc decrease was associated with most tumors mentioned herein, further – prospective - studies are needed.


Introduction
Zinc(II) plays a role in several intracellular signalling pathways. It is also a cofactor of numerous enzymes [1]. Its dysregulation is present in various cancers. Imbalance of zinc transporters causing intracellular and serum zinc(II) levels alteration was described in prostate and breast cancers. [2][3][4][5]. Questions were raised whether these associations have clinical applications. Studies focusing on zinc content in biological materials in cancer patients provide inconsistent results. Zinc levels in tumor tissues of prostate [6], liver [7], and lung [8] and its serum levels in breast, lung, stomach, and prostate cancer patients were reviewed previously [9].
We investigated the associations of serum and cellular zinc(II) levels with carcinomas via meta-analysis.

Literature Search
Search was performed in Web of science (Science citation index expanded 1945 to April 2013), PubMed (Medline 1968 to April 2013), Embase (1977 to April 2013), and Cochrane Library (CENTRAL 1953 to April 2013); keywords are shown in Figure 1. Moreover, cited references of found articles were analyzed.

Selection Criteria
Diagram (Fig. 1) shows acquisition process. Among published articles, the search was done for clinical trials, case-control and cohort studies investigating the associations between carcinoma and tissues and serum zinc levels. Because no difference in zinc(II) level between serum and plasma was found [10], both materials were referred as ''serum''. Studies with full texts available were included only. Only the studies where the data were displayed in the following ways were accepted: (1) sample size, means and standard deviations, or (2) sample size, means, P values and statistical test type (one-or two-tailed). If similar data were found in more studies by the same group, study with most data was included.
The eligibility of the studies for meta-analysis was evaluated by J.G. and V.A., discrepancies were discussed with R.K and M.M.

Assessment of Methodological Quality
The quality of studies was assessed by the Newcastle-Ottawa Scale (NOS) [11]. NOS ranges from 0 to 9 stars. The studies with ,4 stars were excluded, with .6 stars were considered as high quality; the mean was 5.6 stars.

Main and Subgroup Analyses
First, differences of the serum and tissue zinc levels between overall tumors and controls were analysed (Table 1). Then, analyses by tumor type, histology, and methodological quality were performed (Table 2). To meet conditions of between-study independence, zinc level was averaged in studies with multiple tumors, forming age groups or detecting gender separately in case of the summary. For comparison of individual tumors, the tumor was taken as a unit of analysis unless violated between-study independence.

Statistical Analyses
To express the differences in serum and tissue concentrations, standardized difference in means (Cohen's d) was used. To assess heterogeneity across studies, Higgins I 2 (describes percentage of variability) was calculated [12]. Random effects model metaanalysis was employed when I 2 .50%; otherwise, fixed model was used. Publication bias was evaluated using funnel plots and twosided Egger tests in groups with .3 studies. Funnel plots of studies with Egger's test p,0.05 are asymmetric (Table S1). There was performed meta-regression using unrestricted maximum likehood method of studies reporting stage or grade if number of studies with corresponding moderators was .10 [13,14]. Comprehensive Meta-analysis Version 2 software (Biostat, Englewood, NJ) was used for analysis.

Identification of Relevant Studies and their Characteristics
Total of 3201 articles were found. After excluding articles not meeting the criteria (Fig. 1) and duplicates, 114 articles studying 130 tumors were included (several studies studied more tumors, Table S2).

Overall Zinc Level in Sera and Tumors
Serum level. As shown on Fig. 2 and Table 1, serum zinc level is significantly decreased in patients with tumors (effect size = 21.08; 95% confidence interval, CI, 21.33 to 20.82) using random effects model of 58 studies (6223 cases, 10364 controls). This is consistent with subgroup of 26 high-quality studies (effect size = 21.30; 95% CI, 21.72 to 20.88). High level of heterogeneity is observed (Higgins I 2 = 96.71%). Meta-regression did not reveal that stage, grade, and age or publication year affect effect size (Table S1). Six studies analyzed serum zinc level in group of patients with malignant tumor without other specification (516 cases, 3871 controls), and significant decrease was found (Fig. 2) [15][16][17][18][19][20].

Gynecological Tumors (Uterine Corpus, Cervix and Ovarian)
Serum. Among ovarian and cervical cancers, there was a significant decrease found (20.39; CI20.60 to 20.17) using fixed effects model of three studies (164 cases, 171 controls) [20,48,49]. No publication bias was observed. When subgroup meta-analysis by histological type was done, significant decrease was found in cervical [48], insignificant changes in ovarian cancer [49]. There were no significant trends in relation to stage and grade [49].
Tissue level. No significant difference between uterine corpus and cervix cancers was determined using random-effects model of 4 studies (80 cases, 123 controls) [37,48,50,51]. No publication bias was determined. Significant decrease was reported in two studies [48,51]. Results of high-quality studies did not show a significant trend.
Colorectal cancer, serum. No significant difference was determined. No study fulfilled criteria of ''high-quality''. Five studies were included (313 cases, 216 controls) [28,30,[61][62][63] and no publication bias was detected. However, two studies showed significant decrease [28,62], one study showed significant elevation [61], and thus, high serum zinc level variances were associated with colorectal cancer. Of studies reporting stage, one showed significantly decreasing trend on Dukes stage [62], whereas other showed no significant changes related to TNM stage [61].
Gallbladder, tissue. Significant decrease was determined in two studies [68,69]   Liver, serum. No significant change was observed in three studies [70][71][72] (149 cases, 121 controls) and no publication bias was observed. Two high-quality studies [71,72] are in accordance, showing fixed serum zinc level. One study showed no significant trend in relation to stage [70].
Insignificant changes were observed in large sample studiescohort from French SuViMax study (4961 cases and controls) [82] and Multiethnic Cohort study (1175 cases and controls) [85]. Subgroup meta-analysis found significant decrease in high quality studies (22. [86]. Most significant decrease was observed in study by Guntupalli et al. [89]. High level of publication bias was observed among studies (p = 0.0007). Results are in agreement with more distinct effect size in high-quality studies (26.03; 95% CI, 29.39 to 22.67). Trend was not observed in one study relating to stage and grade [6].
Tissue. Two studies (45 cases, 27 controls) [103,104] were included in the analysis. Findings of these studies were contradictory: one showed significant elevation [103] while the othersignificant decrease [104]. Random effects model did not show any trend. No significant trend between grade and zinc level was observed in one study [104].

Thyroid
Serum. No significant difference was observed using random effects model of three studies (131 cases, 93 controls) [105][106][107]. One study showed significant elevation [107], the other showed significant decrease [105] and third, ranked as high quality [106], found no significant differences. No publication bias was observed. Subgroup analysis by histological type does not highlight significant difference between papillary and follicular cancer. Medullar carcinoma was not included in meta-analysis.
Tissue. Statistically significant decrease was observed using random effects model (effect size = 22.84; 95% CI, 25.39 to 2 0.29) of three studies (109 cases, 123 controls) [105,108,109]. However, one study reported insignificant results [105], another included papillary, follicular, medullar cancers and reticulosarcoma [108] and the third did not specify histological types. No publication bias was present. One study determined no significant difference between papillary and follicular cancers [105].
Tissue. Significant decrease was determined using random effects model (23.12; 95% CI, 24.57 to 21.67) of six studies, all ranked as low-quality [34,46,115,[123][124][125] (470 cases, 1820 controls). However, publication bias was observed (p = 0.03). Insignificant change was observed in one study [123]. Significant decrease was determined in small cell lung cancer (20.52; 95% CI, 20.80 to 20.23) using fixed model and no significant decrease was identified in non-small cell lung cancer using random effects model. No significant trend was observed in squamous and adenocarcinomas. Large cell cancer showed significant decrease in one study [115].

Discussion
Decreased serum zinc level was found in patients with lung, head and neck, breast, liver, stomach, and prostate cancers. The elevation was not proven in any tumor. More specific zinc patterns are evident in tumors. Unequivocal increase was observed in breast cancer tissue only and decrease in prostatic, liver, lung, and thyroid cancer. The rest of the studied tumors brought ambiguous results, both in serum and tissue zinc levels across the studies. It cannot be confirmed that the serum zinc level does not change except of the abovementioned tumors. Serum and tissue zinc level reduction was evident to certain extent in majority of tumors. Although insignificant differences were found, the analysis indicates that none of the tumors clearly disproves that the zinc levels remained unchanged. Variation of serum zinc level were found in esophageal cancer patients, in cell zinc content in liver cancer and both in serum and tissue zinc level in stomach, colorectal, and thyroid cancers.
Number of studies point to decreasing trend in tumors of higher grades or stages. Nevertheless, meta-regression could not be performed on the majority of tumors due to limited number of studies reporting stage/grade or to inconsistency in the scale used. Regression analysis of all tumors, however, did not show dependence on these parameters. Thus, this meta-analysis fails to explain the sources of high heterogeneity between the studies.
Although serum zinc level decrease in lung, head and neck, and breast carcinomas was shown by meta-analysis, it is unclear, whether hypozincaemia is a consequence of tumor, chronic stress or of a combination of both these effects. Stress, infection or chronic diseases lead to redistribution of zinc(II) between body compartments, and thus reduce zincaemia [126]. In addition, chronic inflammation is a common hallmark of cancer, and thus might be important mechanism of serum zinc level decrease. The association of tissue zinc level and prostate [6], liver [7], and lung [8] cancers serum zinc level and risk of breast, lung, stomach, and prostate cancers [9] were in scope of several reviews. Decrease in prostate cancer tissue zinc level is well-evidenced [127]. Also review by Zaichick et al. show decrease of zinc in prostate cancer tissue as compared with benign hyperplasia [6]. A review by Catalani, focusing on zinc content in lung tumors, is the only to date meta-analysis. However, its results did not allow summarizing the significance of tissue metals. No relationship among tissue zinc level and histotype or stage was found. Zinc decrease in liver cancer tissues were reviewed by Gurusamy et al. They declared that meta-analytic approach is impossible because of heterogeneity of analyzed studies. All mentioned reviews concluded that there is poor data agreement between studies determining tissue zinc level. This fact -combined with the low metal concentrations -calls for the standardization of methods. Catalani et al. propose standardization of sample collection, storage, and analysis. Previous reviews were performed only on specific tumors, with limited number of studies and/or statistic approaches were missing. Our meta-analytical analysis was done on all identified carcinomas, serum and tissue levels were analyzed together, publication bias was assessed and meta-regression was performed when case sufficient data were present. To reduce selection and publication biases, prospective cohort study with defined conditions separating the influence of inflammation is needed. Interest should be focused on the relation of zinc level in each histological type, stage, and grade.
There are limitations in this study caused by features of individual studies: sample sizes, subjects' characteristics, sampling, storage and detection methods, and different tumors classification.
Serum zinc level has a limited predictive value, because it is particularly intracellular ion and it fluctuates in circadian rhythm.
This meta-analysis shows a decrease of zincaemia in lung, head and neck, and breast carcinoma, increase of tissue zinc in breast cancer and its decrease in prostate, liver, and lung cancers. However, this analysis does not provide conclusive data with regard to stage and grade, and thus does not clarify heterogeneity in values between the studies.