A Meta-Analysis on the Relationship between Exposure to ELF-EMFs and the Risk of Female Breast Cancer

Objective To comprehensively analyze the relationship between exposure to extremely low frequency electromagnetic fields (ELF-EMFs) and the development of female breast cancer. Methods Reports of case-control studies published from 1990 to 2010 were analyzed. The quality effect model was chosen to calculate total odds ratio (OR) depending on the data in studies and quality scores. Subgroup analyses were also performed by the situation of menopause, estrogenic receptor and exposure assessment respectively. Results For all 23 studies the OR was 1.07, 95% CI = 1.02–1.13, for estrogen receptor positive subgroup,OR = 1.11, 95% CI = 1.03–1.20; for premenopausal subgroup, OR = 1.11, 95% CI = 1.00–1.23. The results of other subgroups showed no significant association between ELF-EMF and female breast cancer. Conclusion ELF-EMFs might be related to an increased risk for female breast cancer, especially for premenopausal and ER+ females. However, it's necessary to undertake better epidemiologic researches to verify the association between ELF-EMF and female breast cancer due to the limits of current study, especially the one on exposure assessment.


Introduction
Extremely low frequency electromagnetic fields (ELF-EMFs) are 0-300 Hz electromagnetic fields that are mainly generated by power transmission lines, power equipment and appliances. The possible association between ELF-EMFs and different cancers such as brain tumors, leukemia, and breast cancer [1][2][3] has been discussed widely since 1979 when an epidemiologic study by Wertheimer and Leeper suggested a possible link between ELF-EMFs and childhood leukemia [4]. Breast cancer is the most prevalent malignant disease in women, and its incidence continues to increase. In 1987, Stevens suggested that ELF-EMFs and visible light at night (around 1015 Hz) may increase the long term risk of breast cancer [5]. To date, numerous scholars around the world have examined the correlation between ELF-EMFs and the development of breast cancer.
The methods on choosing the objects of a study were similar in terms of previous case-control studies on ELF-EMFs and breast cancer. The cases were often chosen from the registration systems related to cancer or hospital case reports. There were various ways of choosing objects to form control groups, the controls of some studies were chosen from other cancer patients and some were chosen from a group of people who live in the same area by random sampling. Meanwhile age and gender were often taken into consideration to form a control group. The studies on the association between ELF-EMF and development of female breast cancer have not reached a consensus yet. The accuracy of study on the correlation between ELF-EMF and development of female breast cancer was mainly attributed to the assessment of exposure and the identification between exposure group and un-exposure group. The electromagnetic fields were everywhere, therefore it's inevitable that everyone was exposed to electromagnetic fields to a certain degree. Although some studies have taken life and occupations into consideration, most of the studies only focused on one aspect of life habits, resident environment or work environment in terms of exposure assessment due to its complexity. The exposure groups were often confirmed in the following ways: 1.using electric heating equipment such as electronic blanket or not; 2. exposure level of living environment; 3. work duty or measuring and assessment of working environment; 4. distance from the high voltage power lines; 5. combining some above factors.
Currently there were two teams who had conducted the metaanalysis in regard to the correlation between ELF-EMF and development of female breast cancer. At first, T.C. Erren metaanalysed the case-control studies and cohort studies before 2000 and the results showed the pooled RR from studies in women was 1.12 (95% CI = 1.09-1.15), but variations between the contributing results are not easily attributable to chance (P = 0.0365). Meanwhile T. C. Erren indicated that some research methods might lead to the incorrect assessment of exposure or cases, which has caused the discrepancy of many study results [6]. A total of 15 case-control studies published over the period 2000 to 2009 including 24,338 cases and 60,628 controls were involved in the meta-analysis of Chunhai Chen etc. [7]. The results showed no significant association between ELF-EMF exposure and female breast cancer risk in total analysis (OR = 0.988, 95% CI = 0.898-1.088) and in all the subgroup analyses by exposure modes, menopausal status, and estrogen receptor status.
A large number of scholars around the world have undertaken plenty of studies on the association between ELF-EMF and development of female breast cancer. The method of metaanalysis used in those studies was remarkable. However, Chunhai Chen only summarized the literature from 2000 to 2009 and failed to incorporate many significant studies before 2000. Our study firstly would go through the literature from 1990 to 2012 using the method of meta-analysis. Secondly, because the confounding factors adjusted in the statistical analysis and the methods dividing the exposure groups were different in different literature, our study would undertake statistical analysis through searching the number of exposed and un-exposed individuals in the original cases or controls instead of adopting the adjusted OR or RR value in the literature. The number of samples in different groups was missing in most of cohort studies so that we have just taken all case-control studies into consideration for calculating the total OR value in this study. At last, if using traditional fixed effect model or random effect model, the important role of literature with excellent study design would not be displayed in the Meta analysis study. Considering the diversity of different study design quality, our study would use quality effect model to conduct meta-analysis in which literature with better study design would be prioritized to produce more accurate results.

Identification of studies and eligibility criteria
Medline, PubMed, EMBASE and Hirewire databases were used to search for studies assessing the relationship between ELF-EMFs and female breast cancer from January 1990 to December 2012 using synonyms and combinations of the terms 'breast cancer', 'breast neoplasm', and 'electromagnetic fields'. At first, two reviewers (Q.S. Chen, W.Z. Wu) picked up the case-control studies and cohort studies from the identified articles by reviewing the titles and abstracts. Then they read the full text and chose the studies as objects of study according to the following criteria: (1) the publication was a population epidemiology study on the association between ELF-EMF exposure and breast cancer in females; (2) the papers must offer the size of the samples, number of exposed and non-exposed individuals in cases and controls; and (3) publication language was confined to English. When multiple publications reported on the same or overlapping data, we used the most recent or largest population as recommended by Little et al.

Data collection and quality assessment
Assessment and division of exposure often was the most important reason leading to difference in the results of studies of correlation between ELF-EMF exposure and breast cancer in females. In the articles covered in this study, the ELF-EMF exposure has been assessed through investigating the usage of electric blanket or position title, divided through measuring and assessing of the electrometric field level in the living and working environment. For the studies assessing exposure level by measuring or calculation, the researchers usually used the group with the lowest exposure as the referent to calculate the OR value and different intensity factors such as 0.1 mT,0.2 mT are used to decide the lowest exposure group in different studies. In this metaanalysis, the group with the lowest exposure was selected as the referent and other groups were put together as exposure group to calculate the total OR value. The all literature mentioned in our study has taken age and region into consideration in choosing cases and controls. The confounding factors adjusted in the statistical analysis (such as age, habit of smoking and drinking, race, menopausal status, and estrogen receptor status) were quite differential in different study. So, the number of exposed and unexposed individuals in cases and controls were abstracted for metaanalysis instead of copying the adjusted OR or RR value from the literature.
In this study an extracting list on the important information of the literature was made. Two researchers read and analysed the literature individually. They abstracted the key information that includes: the author, the year of publishing, the country of the objects, the number and selecting method of cases and controls, the assessing methods of exposure, matching factors of cases and controls, the year of research, OR value and 95%CI. If the results of the analysis conducted by the two researchers are inconsistent, it will be solved through discussion.
The studies on the correlation between ELF-EMF exposure and breast cancer in females used different research methods, especially on the assessment and dividing of the exposed and unexposed individual. Considering some studies are more accurate and reliable, the quality assessment method of cases and controls study has been developed based on Newcastle-Ottawa Scale [8]. All studies were assessed from three aspects that included ten indicators in terms of choosing method of cases and controls, comparability of cases and controls and exposure assessment. The Quality scoring criteria was showed in table 1 in which each indicator was given one score and the total score of every study was ten. The quality assessment and scoring of all literature would be conducted by two experts in this field independently. If the results of the analysis are inconsistent, it would be solved through discussion.

Statistical analysis
Microsoft Excel was used to organize the initial data and build a database. The quality index of every study was equal to the quality total score of every study divided by ten. Then a quality effect model of MetaXL version 1.3 was applied to analyze the data and calculate the total OR and 95% confidence interval to assessing the relationship between ELF-EMFs and breast cancer in females [9][10][11]. MetaXL implements a process called quality effects model to explicitly address study heterogeneity caused by differences in study quality. This model is a modified version of the fixed-effects inverse variance method that additionally allows giving greater weight to studies of high quality. Considering the difference of exposure assessment method, subgroup meta-analyses by quality effect model were performed according to exposure modes (occupational exposure, residential exposure, blanket exposure and multiple exposures). Considering the effect of hormone to breast cancer, the subgroup analyses also were performed by menopausal status and estrogen receptor (ER) status. Heterogeneity assumption was assessed by Chi-square based Q-test and Isquared test. If P value for Q test ,0.10 then refuse Ho, the heterogeneity is significant. An inverted funnel plot was drawn with an abscissa of each study's OR value and an ordinate of the standard error (SE) for publication bias examination.
From figure 1, the results of sixteen studies, occupied 69.57% of all 23 case control studies, showed OR value was over one and the other OR values of seven studies was under one. Except the results of Kliukiene J and Loomis DP's studies [13,30] showed a statistic      As shown in table 3, the P value was less than 0.1 and the I2 value was equal to 39% by heterogeneity test for total comparison that indicated the heterogeneity was significant, though the difference was small. For all subgroups, except multiple exposures subgroup in which the heterogeneity was significant (P,0.1, I2 = 82%) the homogeneities of other subgroups were great, P.0.1. MetaXL version 1.3 software was used to draw a funnel plot and to perform a linear regression analysis. Figure 2 presents the data distribution, which was almost bilaterally symmetrical, demonstrating that the bias was small.

Discussion
The association between the exposure to ELF-EMFs and the development of female breast cancer has always been a controversial topic in the scientific field. Animal testing has indicated that exposure to ELF-EMFs could increase the risk of cancer. Loscher and others used DMBA to induce breast cancers in mice and used 50 Hz,0.2-1 mT, 10 mT, 50 mT, and 100 mT magnetic fields to irradiate mice 24 hours a day for 13 weeks [33]. A dose-response relationship between cancer rates and magnetic fields was observed in that study. However, Owing to the uncertainties of exposure and confounding factors such as the information on the exposure levels of the case and the control group in most epidemic studies is uncertain, research reports on ELF-EMFs have often demonstrated differences in results. In this studies, two studies showed a relation between exposure to ELF-EMFs and female breast cancer [11,28], while other 21 studies did not indicate this conclusion [3,10,[12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]29,30]. This meta-analysis comprehensively analyzed twenty three items of case-control studies to examine the relationship between the exposure to ELF-EMFs and breast cancer. As shown in table 3, except multiple exposures subgroup in which the heterogeneity was significant because of the limit of sample size, the homogeneities of other groups were great. The funnel plot also demonstrated that the  publication bias of the selected articles was small. The results of this study showed the pooled OR = 1.07,95% CI = 1.02-1.13 by the quality effect model, which indicated that the development of breast cancer might be potentially related to exposure to ELF-EMFs. The result of our study is close to the one of Meta-analysis conducted by T.C. Erren [6] in 2001 in which the pooled RR from studies in women was 1.12 and 95% CI = 1.09-1.15. However, because results from individual studies were very variable and in part contradictory, T.C. Erren concluded the paramount methodological problem inhibiting valid conclusions about an association between EMF and breast cancer was the probable misclassification of exposure and the possible misclassification of the disease itself. Another meta analysis about an association between EMF and breast cancer was published by Chunhai Chen etc. in 2010 and the results showed no significant association between ELF-EMF exposure and female breast cancer risk in total analysis (OR = 0.988, 95% CI = 0.898-1.088). Compared with the above two studies, this study have covered literature from 1990 to 2012 and taken the lowest exposure group (most nearly zero exposure) as the un-exposed groups. This study was also different from the previous studies in which fixed effect model and random effect model were adopted in terms of methods on meta-analysis. This study has used quality effect model to do meta-analysis in which studies with better methods are prioritized. Of course, judging from the quality scoring of the selected 23 studies, exposure assessment was still the most important factor that affected the accuracy and reliability of study. It is generally accepted that, to date, no research has demonstrated that EMFs directly affect the body to exert a chronic influence on health. The relationships between the EMFs and the environment, body, and other factors have become the major focus for research on health-related issues, especially cancer, caused by EMFs. Many studies have suggested that the occurrence and development of breast cancer were closely related to estrogen, melatonin, and other hormones. As previously discussed, Girgert et al. [34] confirmed the changes in expression of cofactors of the estrogen receptors in human breast cancer cells exposed to low frequency EMFs by in vitro study. This meta-analysis analyzed nine items of premenopausal case-control studies and nine items of postmenopausal studies to examine the relationship between exposure to ELF-EMFs and breast cancer. Through stratification analysis, the OR of the premenopausal subgroup was 1.11, 95% CI = (1.00-1.23) and that of the postmenopausal group was OR = 1.02, 95% CI = (0.95-1.09). The seven studies with estrogen receptor information were analyzed, the ER+ (estrogen receptor positive) subgroup OR = 1.11, 95% CI = (1.03-1.20) and ER2 (estrogen receptor negative) subgroup OR = 0.96, 95% CI = (0.84-1.10). These results indicated that, for the premenopausal group and ER+ group, the occurrence of breast cancer may be related to exposure to ELF-EMFs. However, for the postmenopausal group and ER2 group, no relationship was observed. The results of this study were in accordance with the results of Girgert et al's studies. This will give us a big reason to make further study to assure whether ELF-EMFs affect the development of women's breast cancer by affecting hormones and what is the specific mechanism(s) behind this phenomenon.
In the epidemiologic studies, the result of correlation between ELF-EMF exposure and breast cancer in females was mainly affected by exposure assessment. There were only two studies taking living and working environment into consideration in the 23 studies. The rest studies only focused on one aspect of life or work such as usage of electric heating equipment (eg. electronic blanket), work duty and distance from the high voltage power lines. It was well known that the electrometric field was everywhere due to the wide-spread application of electric equipment. It was really hard to take all the exposures in living and working environment into consideration, however it was obvious that the exposure assessment was insufficient if only taking one aspect into account. The result of this meta-analysis suggested a significant difference (OR = 1.08, 95% CI = 1.00-1.15) for the occupational exposure subgroup was consistent with the result from T.C. Erren. Our study has especially conducted a metaanalysis on two studies with more comprehensive exposure assessment, OR = 1.35, 95% CI = 0.97-1.89,although statistical difference was not significant, the level of OR value indicated that more comprehensive studies need to be done on exposure assessment to discover the correlation between ELF-EMF exposure and breast cancer in females on top of the accuracy and reliability on case and control group selection. Our study has recommended the direction for future studies in this field.
From the results of this study, we determined that exposure to ELF-EMFs might be a risk factor contributing to the development of breast cancer, especially for premenopausal and ER+ females. Currently most of the studies have some defects in assessing and dividing exposure, therefore more studies with comprehensive and accurate exposure assessment are in need to further confirm the correlation between ELF-EMF exposure and breast cancer in females, Especially the combined effects of ELF-EMFs and estrogen, melatonin, or other hormones on the development of breast cancer.