Association of TGF-β1 Polymorphisms with Breast Cancer Risk: A Meta-Analysis of Case–Control Studies †

Reports on the association of TGF-β1 polymorphisms with breast cancer (BC) have been conflicting, inconsistent, inconclusive, and controversial. PubMed, EMBASE, and Google Scholar were used to identify studies on TGF-β1 polymorphisms and BC risk. Data were extracted independently, and of the initial 3043 studies, 39 case-control studies were eligible for inclusion in the meta-analysis. Information from these studies was extracted, and the overall associations of three TGF-β1 polymorphisms (TGF-β1 29>T/C, TGF-β1-509 C/T, and TGF-β1*6A) with BC risk were analyzed using overall allele, homozygous, heterozygous, recessive, and dominant models. None of the three TGF-β1 polymorphisms studied had a significant influence on the development of BC. However, stratified analysis revealed a positive correlation between the TGF-β1 29T>C polymorphism and BC risk according to a heterozygous model of the Asian population (odds ratio (OR) = 1.115, 95% confidence interval (CI) = 1.006–1.237, p = 0.039). Interestingly, this polymorphism was associated with lower odds of BC according to a heterozygous model of the Middle Eastern population (OR = 0.602, 95% CI = 0.375–0.966, p = 0.035). Thus, our analysis of large datasets indicates that the TGF-β1 29T>C polymorphism is significantly associated with BC risk in the Asian population. In contrast, the TGF-β1*6A and TGF-β1-509 C/T polymorphisms failed to show an association with BC.


Assessment of Heterogeneity among the Studies and Publication Bias
Heterogeneity was observed in the allelic models of all three TGF-β1 polymorphisms, as well as all genetic models of the TGF-β1 29T>C polymorphism and the homozygous model of the TGF-β1-509 C/T polymorphism. In contrast, none of the genetic models of the TGF-β1*6A polymorphism were heterogeneous. Heterogeneity among the studies of all three polymorphisms in all ethnic groups is represented in Supplementary Table S1. Information regarding the presence of potential publication bias in ethnic studies is illustrated in Supplementary Figure S2

Stratified Analysis
Studies were categorized based on the ethnicities of the individuals studied. Studies covering the TGF-β1 29T>C polymorphisms were divided into four groups: Caucasian, European, Asian, and Middle Eastern. The TGF-β1 29T>C polymorphism was found to be associated with BC risk in the heterozygous model of the Asian population (OR = 1.115, 95% CI = 1.006-1.237, p = 0.039). In contrast, the same polymorphism was associated with lower odds of BC in the heterozygous model of the Middle Eastern population (OR = 0.602, 95% CI = 0.375-0.966, p = 0.035) ( Figure 5). No correlation was found between the TGF-β1 29T>C polymorphism and BC risk in the Caucasian and European populations. Similarly, results of other variants of TGF-β1 polymorphism have shown no association with BC risk in ethnic groups. These findings suggested that the TGF-β1 29T>C polymorphism alone is associated with BC risk in the Asian population. The ORs and 95% CIs of all stratified ethnic group analyses are provided in Table 2.

Sensitivity Analysis
One-way sensitivity analysis was performed to analyze the effect of each study on the combined OR. During the sensitivity analysis for each polymorphism, each relevant study was deleted iteratively from the dataset prior to the analysis. No single study had a significant influence on any of the combined ORs, suggesting that our meta-analysis is relatively robust and credible ( Supplementary Figures S5-7).

Trial Sequential Analysis (TSA)
TSA was used to investigate whether the number of samples included in the present study was sufficient for detecting a possible role of each TGF-β1 polymorphism in BC. The Z curve touched the trial monitoring boundaries or reached the required information line, indicating that a sufficient number of studies were included in the meta-analysis for overall and ethnic group investigations (Supplementary Figure S8).

Discussion
The TGF-β1 signaling pathway has been studied extensively in several cancers, including BC. TGF-β1 plays an important role in cell differentiation, migration, invasion, and tumor growth [4]. Both tumorigenic and tumor-suppressive roles of TGF-β1 have been well established [5]. Elevated plasma levels of TGF-β1 have been associated with cancer development, and TGF-β1 polymorphisms

Sensitivity Analysis
One-way sensitivity analysis was performed to analyze the effect of each study on the combined OR. During the sensitivity analysis for each polymorphism, each relevant study was deleted iteratively from the dataset prior to the analysis. No single study had a significant influence on any of the combined ORs, suggesting that our meta-analysis is relatively robust and credible (Supplementary Figures S5-S7).

Trial Sequential Analysis (TSA)
TSA was used to investigate whether the number of samples included in the present study was sufficient for detecting a possible role of each TGF-β1 polymorphism in BC. The Z curve touched the trial monitoring boundaries or reached the required information line, indicating that a sufficient number of studies were included in the meta-analysis for overall and ethnic group investigations (Supplementary Figure S8).

Discussion
The TGF-β1 signaling pathway has been studied extensively in several cancers, including BC. TGF-β1 plays an important role in cell differentiation, migration, invasion, and tumor growth [4]. Both tumorigenic and tumor-suppressive roles of TGF-β1 have been well established [5]. Elevated plasma levels of TGF-β1 have been associated with cancer development, and TGF-β1 polymorphisms have been found to cause high transcription and expression of TGF-β1 [9,11]. The prominent role of TGF-β1 in cancer progression underscores the importance of studying the association between TGF-β1 polymorphisms and BC risk.
The most common TGF-β1 polymorphism is a substitution of cytosine with thymine at the 29th nucleotide (TGF-β1 29 T>C), which results in the substitution of the proline at codon 10 in exon 1 with leucine [8][9][10][11]. TGF-β1 29T>C has been associated with elevated levels of TGF-β1 in plasma [9]. Studies have also suggested that TGF-β1 29T>C is associated with BC risk; however, contrasting results were reported. Dunning et al. suggested that the TGF-β1 29T>C polymorphism is associated with BC risk; they also found that this polymorphism is associated with elevated levels of the TGF-β1 protein [9]. In contrast, Ziv et al. reported no such associations [12]. Marchand et al. reported similar results in their multi-ethnic study, which suggested no association between the TGF-β1 29T>C polymorphism and BC [13]. Interestingly, Lee et al. demonstrated that the TGF-β1 29T>C polymorphism was associated with an increased risk of BC in postmenopausal women [32,33]. However, Hishida et al. observed a negative correlation between the TGF-β1 29T>C polymorphism and BC risk in premenopausal women [27]. Interestingly, Shin et al. demonstrated that no association was found in the initial stages of BC, whereas a significant correlation was observed in later stages [30].
A systematic meta-analysis was performed to provide conclusive evidence for the association of TGF-β polymorphism with breast cancer risk. Interestingly, neither the overall allele nor genotypic models showed an association between the TGF-β1 29T>C polymorphism and BC risk. However, the TGF-β1 29T>C polymorphism was found to be associated with an increased risk of BC in the heterozygous model of the Asian population. In contrast, the TGF-β129T>C polymorphism favors lower odds of BC in the heterozygous model of the Middle Eastern population.
The TGF-β1-509 C/T polymorphism is located in the promoter region of TGF-β1. Recent studies demonstrated the association between the TGF-β1-509 C/T polymorphism and elevated levels of the TGF-β protein. It has also been well established that the TGF-β1-509 C/T polymorphism is highly associated with BC risk. Both Dunning et al. [9] and Parvizi et al. [52] suggested that the polymorphism is associated with increased risk of invasive BC, whereas Cox et al. showed no such association [35]. Vinod et al. demonstrated that heterozygosity of the TGF-β1-509 C/T polymorphism was associated with BC susceptibility in an Indian population [51]. However, Babyshkina et al. reported that homozygosity of the T allele of the TGF-β1-509 C/T polymorphism was not associated with BC in a Russian population [49]. These inconclusive and controversial results suggest the necessity of a cumulative and robust analysis to arrive at a definitive conclusion regarding the association between this polymorphism and cancer risk in multiple ethnic groups. Six meta-analyses were performed to address these issues; three investigated the association between the TGF-β1-509 C/T polymorphism and multiple cancer subtypes. Huang et al., Qi et al., and Woo et al. performed meta-analyses examining the association between the TGF-β1-509 C/T polymorphism and BC specifically [14,17,58]. However, these studies were performed using a limited number of studies, and Huang et al. included a study that was not related to BC risk [58]. These drawbacks can influence the analysis and affect the results. In order to definitively determine the association between the TGF-β1-509 C/T polymorphism and BC, we performed a meta-analysis including additional studies and excluding studies not related to BC risk. Our robust analysis demonstrated that the TGF-β1-509 C/T polymorphism is not associated with BC in the overall population or within specific ethnic groups.
Researchers have identified a polyalanine variant in exon 5 of TGF-β1 and analyzed its association with BC risk. Recent studies have also suggested that the TGF-β1*6A polymorphism is associated with BC risk. Specifically, Kaklamani et al., Pasche et al., and Song et al. reported that the polymorphism is associated with an increased incidence of BC [19,31,55]. Pasche et al. also found that the TGF-β1*6A polymorphism is associated with risk for other cancers, including colorectal and ovarian cancer [19]. However, Chen et al., Colleran et al., and Cox et al. suggested that the TGF-β1*6A variant is not associated with BC risk. Meta-analyses have been performed; however, they investigated the association between the TGF-β1*6A polymorphism and all cancer types; no meta-analyses have yet correlated this polymorphism with BC specifically [20,59]. Thus, in the present study, we analyzed for the first time the association between the TGF-β1*6A polymorphism and BC risk. Our cumulative analysis suggests that the TGF-β1*6A variant is not associated with BC risk. In our analysis of the European ethnic group, the TGF-β1*6A variant also showed no association with BC risk.
Although various meta-analyses have already been reported in the literature, the present study has several advantages over them. A recent investigation by Alqumber et al. included reports published until the year 2013; in contrast, we screened for studies published up to May 2019, resulting in the inclusion of three additional studies. Furthermore, we assessed the possible association of three TGF-β1 polymorphisms (TGF-β1 29>T/C, TGF-β1-509 C/T, and TGF-β1*6A) with the risk of BC [60]. However, multiple comparisons failed to show a significant correlation between these polymorphisms with BC risk: the present study warrants the non-importance of TGF-β1 polymorphisms on the pathogenesis of BC. Finally, TSA revealed that our meta-analysis included a sufficient number of studies, performed worldwide, to dissect the possible association of TGF-β1 variants with BC risk, and further investigation is not required.

Criteria for the Inclusion and Exclusion of Studies
The present meta-analysis included studies that evaluated the association between TGF-β1 polymorphisms and BC risk, were published in English, presented original data, and provided the genotypic frequency of both case and control samples or had odds ratios (ORs) with 95% confidence interval (CI) values. The analysis excluded reviews, abstracts, duplicate or overlapping studies, studies not published in English, studies that correlated TGF-β1 polymorphisms with non-breast cancers, and studies that did not provide genotypic or allele frequencies for both case and control samples.

Data Extraction
Each publication was assessed thoroughly, and data were extracted independently by three researchers, following the same pattern and extracting the same data essential for meta-analyses that have been previously described [22,23]. The researchers conducted group discussions to resolve any discrepancies in extraction.

Statistical Analysis
Statistical analysis was performed using Comprehensive Meta-Analysis Software (CMA version 3) (Biostat, 14 North Dean Street, Englewood, NJ 07631 USA). In order to appraise the associations between TGF-β1 polymorphisms and BC risk, the combined OR, 95% CIs, and their respective p-values were calculated by comparing the cancer patient samples vs. respective healthy controls. Heterogeneity among the studies and publication bias were calculated using Begg's funnel plots and chi-squared-based Cochran's Q tests, respectively. Egger's regression tests were performed to analyze and measure the symmetry of funnel plots, as described previously [22]. Asymmetrical funnel plots were made symmetrical using the "Trim and Fill" method. Random models were used to calculate the combined ORs for studies showing heterogeneity [24]. In contrast, fixed models were used to calculate the combined ORs for studies with homogeneity [25].

Trial Sequential Analysis (TSA)
An appropriate meta-analysis must include all eligible studies published to date to draw definitive conclusions. The number of studies available in the literature in the studied areas should be sufficient for decisive concluding remarks. TSA is a statistical tool developed by the Copenhagen Trial Unit, Centre for Clinical Intervention Research, Denmark, to estimate the sample size required to reach significance with definitive power. In TSA, a Z curve analysis is performed to check whether a sufficient number of samples is included in the study. If the Z curve intersects the TSA monitoring boundary before reaching the required information size or if the total number of samples exceeds the required information line, then the total number of studies included in the investigation is sufficient, and additional trials are not required. On the other hand, if the Z curve fails to touch the TSA monitoring boundaries or cross the required information line, the number of studies is limited and more are required to draw any definitive conclusions. TSA software version 0.9 (http://www.ctu.dk/tsa) was used for this analysis.

Conclusions
TGF-β1 plays an important role in BC progression, metastasis, stemness, and chemo-resistance. TGF-β1 has both pro-oncogenic and tumor-suppressive roles during cancer development. High levels of TGF-β1, which are influenced by TGF-β1 polymorphisms, have been associated with cancer risk. TGF-β1polymorphisms have been associated with BC risk, but conflicting results have also been reported recently. Previous meta-analyses either included studies on TGF-β1 polymorphisms that did not evaluate associations with BC or did not include new studies with huge datasets that could greatly influence their results. To address these shortcomings, a stringent and highly robust analysis was conducted to provide conclusive evidence for the association between TGF-β1 polymorphisms and BC risk. The current meta-analysis included large datasets from all recent eligible studies and excluded studies that did not evaluate the associations between these polymorphisms and BC. This study suggests that TGF-β1 polymorphisms are not associated with BC risk in the overall population. Both TGF-β1-509 C/T and TGF-β1*6A also showed no association with BC risk in stratified analyses.
However, the TGF-β1 29T>C polymorphism was found to be associated with BC risk in the Asian ethnic group.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/2/471/s1, Figure S1: PRISMA flowchart: Search, screening and selection of eligible studies included in meta-analysis for evaluating the association of TGF-β polymorphisms with breast cancer risk. Figure S2: Funnel plot: Assessment of publication bias using Begg's funnel plot and Egger's regression test. Asymmetric Begg's funnel plot was made symmetric using "Trim and Fill" method. Figure S3: Forest plot: Analyzing the influence of detection methods on association of TGF-β1 29T>C polymorphism with breast cancer risk. Studies were separated based on the methods used for the detection of polymorphism [(A) PCR-RFLP and (B) Taqman] and were analyzed individually. Figure S4: Forest plot: Studies were separated based on the methods used for the detection of TGF-β1 -509 C/T polymorphism [(A) PCR-RFLP and (B) Taqman] and were analyzed individually to assess its association with increased risk of breast cancer. Figure S5: Sensitivity analysis: One individual study of TGF-β1 29T>C was excluded each time and was analyzed to assess the effect of individual study on combined OR. Figure  S6: Sensitivity analysis: Each study was assessed for the potential influence on combined OR by deleting single study each time and performing the analysis for TGF-β1 -509 C/T polymorphism. Figure S7: Sensitivity analysis of each study representing TGF-β *6A polymorphism in the meta-analysis by one study removal method. Figure  S8: Trial sequence analysis of all the included studies. TSA figure for TGFβ-1 29T>C overall (A), Caucasian (B), Asian (C) showed enough number of samples enrolled for the analysis except for the American (D). TSA analysis for other two SNPs TGF-β1 -509 C/T (E) and TGF-β*6A (F) also revealed similar results. Table S1: Statistics for heterogeneity test. Funding: The present work was supported in part by grants from the United States Department of Defense (W81XWH-16-1-0641) and the National Cancer Institute of the National Institutes of Health (P30CA33572). Funding from the Beckman Research Institute of City of Hope is also acknowledged.