Breast Cancer Research and Treatment

, Volume 122, Issue 1, pp 251–257

Current evidence on the relationship between polymorphisms in the COX-2 gene and breast cancer risk: a meta-analysis

Authors

  • Ke-Da Yu
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Breast SurgeryCancer Hospital/Cancer Institute, Breast Cancer Institute, Fudan University
  • Ao-Xiang Chen
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Breast SurgeryCancer Hospital/Cancer Institute, Breast Cancer Institute, Fudan University
  • Chen Yang
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Breast SurgeryCancer Hospital/Cancer Institute, Breast Cancer Institute, Fudan University
  • Li-Xin Qiu
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Medical Oncology, Cancer HospitalFudan University
  • Lei Fan
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Breast SurgeryCancer Hospital/Cancer Institute, Breast Cancer Institute, Fudan University
  • Wen-Huan Xu
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Breast SurgeryCancer Hospital/Cancer Institute, Breast Cancer Institute, Fudan University
    • Department of Oncology, Shanghai Medical CollegeFudan University
    • Department of Breast SurgeryCancer Hospital/Cancer Institute, Breast Cancer Institute, Fudan University
    • Institutes of Biomedical ScienceFudan University
Epidemiology

DOI: 10.1007/s10549-009-0688-3

Cite this article as:
Yu, K., Chen, A., Yang, C. et al. Breast Cancer Res Treat (2010) 122: 251. doi:10.1007/s10549-009-0688-3

Abstract

The association between single-nucleotide polymorphisms (SNPs) in the COX-2 gene and breast cancer risk is still ambiguous. We here try to derive a more precise estimation of the relationship by performing a meta-analysis based on currently available evidence from literature. More than 15 SNPs have been studied, and the most studied genetic variants were rs5275, rs5277, and rs20417. Crude odds ratios (ORs) with 95% confidence intervals (CIs) were used to assess the strength of association between each polymorphism and breast cancer risk under the codominant model, dominant model, and recessive model, respectively (nine studies with 6,968 cases and 9,126 controls for rs5275; three studies with 2,901 cases and 3,463 controls for rs20417; two studies with 5,551 cases and 6,208 controls for rs5277). No overall significant associations were observed in single-locus analysis between the three polymorphisms of COX-2 and breast cancer risk, though a borderline significant increased risk of breast cancer was detected with rs5277 in a recessive model (OR: 1.217, 95% CI: 0.958–1.547, P = 0.107). The results were not changed when studies were stratified by ethnicity. In conclusion, the present meta-analysis suggests that none of the most studied three SNPs (rs5275, rs20417, and rs5277) in the COX-2 gene is a conspicuous low-penetrant risk factor for developing breast cancer. There is a need for further large studies into the role of these polymorphisms (especially rs5277) and other potentially functional polymorphisms/haplotypes in the COX-2 gene as breast cancer risk modifiers.

Keywords

COX-2PolymorphismBreast cancerSusceptibilityMeta-analysis

Introduction

For most sporadic breast cancers and more than half of familial breast cancers, a substantial component of risk may be determined by the effect of multiple low-penetrance susceptibility genes [1]. Recently, several common low-penetrance loci, such as rs2981582 in FGFR2, rs3817198 in LSP1, rs1053485 in CASP8, and rs2046210 at 6q25 located upstream of ESR1 have been identified through genome-wide association studies [2, 3]. It has also been widely accepted that common variants within genes involving breast carcinogenesis-related pathways are candidate loci for breast cancer susceptibility [4]. Numerous association studies regarding genes that are involved in estrogen metabolism pathway, carcinogen metabolism pathway, DNA repair pathway, and immunosurveillance pathway have been performed. Genes involved in the prostaglandin pathway could play potential roles in breast cancer susceptibility [58]. Accumulating evidence has showed that the inducible prostaglandin synthase cyclooxygenase-2 (COX-2, the gene encoding this protein is PTGS2) is implicated in breast cancer tumorigenesis. COX-2 is overexpressed in approximately 40% of human breast cancers [9], and its expression is particularly positively associated with human epidermal growth factor receptor-2 (HER2), a pivotal breast cancer oncogene protein [10, 11]. HER2-induced mammary tumorigenesis and angiogenesis are reduced in COX-2 knockout mice [12].

There are different polymorphism sites in the COX-2 gene. Some of them are suggested to be functional and indicated to be associated with COX-2 expression level. For instance, a previous study showed that the −765C allele of the single-nucleotide polymorphism (SNP) rs20417 (−765G>C) was associated with significantly reduced COX-2 expression compared with the −765G allele, and this effect was postulated to be mediated by the loss of Sp1 transcription factor binding to its cognate element [13], although Szczeklik et al. further demonstrated that −765G>C eliminates an Sp1-binding site but creates an E2F-binding site [14]. Another potential functional SNP is rs689466 (−1195G>A), Zhang et al. reported that reporter gene expressions driven by −1195A-containing COX-2 promoters were 4- to 6-fold greater than those driven by the −1195G-containing counterparts [15]. Thus far, some association studies regarding the contribution of SNPs in the COX-2 gene to breast cancer risk have been performed. To the best of our knowledge, rs5275 (8473T>C), rs20417 (−765G>C) and rs5277 are the most extensively studied polymorphisms. However, the small effect of each single polymorphism on breast cancer risk and the relatively small sample size in each published study made the results inconclusive and controversial.

In consideration of the putative role of COX-2 in breast carcinogenesis, we conduct the present meta-analysis to evaluate the association between genetic variants in the COX-2 gene and breast cancer risk. Although, a meta-analysis regarding this issue has been reported by Zhu et al. recently [16], it is regrettably that, in that report, some eligible studies [17, 18] were not included by the authors, thereby might bias the results and make their conclusions questionable. Moreover, since then, additional studies with larger sample sizes have been reported [19], which warrants the current meta-analysis to be a more precise estimation.

Methods

Publication search

We searched Medline, PubMed, and Web of Science (updated to November 1, 2009) using the following search terms: “COX2” or “COX-2’’ or “Cyclooxygenase-2” or “Cyclooxygenase 2” or “PTGS2” or “prostaglandin endoperoxide synthase 2” and “genetic variant” or ‘‘polymorphism’’ and “breast’’. All the searched studies were retrieved, and their references were checked as well for other relevant publications. Review articles were also searched to find additional eligible studies. Only those published studies in English language with full text articles were included. For overlapping studies, only the first published one was selected; for republished studies, only the one with the largest sample numbers was included.

Eligible studies and data extraction

The inclusion criteria were: (a) evaluation of the polymorphisms in the COX-2 gene and breast cancer risk, (b) retrospective case–control studies or prospective cohort studies, (c) with sufficient available data to estimate an odds ratio (OR) with its 95% confidence interval (95% CI), (d) in English language, and (e) fulfilling Hardy–Weinberg equilibrium in the control group (P > 0.01 was eligible).

The following variables were extracted from each study if available: first author’s surname, publication year, country, ethnicity, numbers of cases and controls, and numbers of cases and controls of different genotypes, respectively. Different ethnicity descents were categorized as Caucasian, Asian, African American, or mixed. We did not define any minimum number of patients to include a study in this meta-analysis. Information was carefully extracted from all eligible publications independently by two of the authors (Yu, K. D. and Chen, A. X.). Disagreement was resolved by discussion between the two authors. If they could not reach a consensus, a third investigator (Shao, Z. M.) adjudicated the disagreements.

Statistical methods

Crude ORs with their 95% CIs were used to assess the strength of association between each studied polymorphism and breast cancer risk. Four different ORs were calculated: (i) heterozygous versus common homozygous carriers, (ii) rare homozygous versus common homozygous carriers, (iii) rare allele carriers versus common homozygous carriers (dominant model), and (iv) rare homozygous versus common allele carriers (recessive model). Heterogeneity assumption was checked by the Q-test and a P-value greater than 0.10 indicates a lack of heterogeneity among studies. The pooled OR was calculated by a fixed-effects model (the Mantel–Haenszel method) or a random-effects model (the DerSimonian and Laird method) according to the heterogeneity [20]. The potential publication bias was examined visually in a funnel plot of log [OR] against its standard error (SE), and the degree of asymmetry was tested using Egger’s test (P < 0.05 considered representative of statistical significance). We also performed sensitivity analysis by omitting each study to find potential outliers [21]. All of the statistical analyses were performed using Stata/SE version 10.0 (Stata Corporation, College Station, TX, USA).

Results

A total of 10 publications (10,482 cases and 12,732 controls) met the inclusion criteria [8, 1719, 2227] (Table 1). In these reports, more than 15 SNPs were studied, and the most studied genetic variants were rs5275, rs5277, and rs20417. Therefore, our meta-analysis focused on the three SNPs and three publications [8, 25, 26] were excluded from further analysis since none of the three polymorphisms was presented in these articles. Of note, Cox et al. validated their primary results in other two independent populations [22]. Each validation group was considered separately in pooling analyses. Thus, there were seven publications containing nine studies including 9,379 cases and 11,649 controls entered into the subsequent meta-analysis [1719, 2224, 27]. Of those, nine studies [1719, 2224, 27] were with regard to the association between rs5275 and breast cancer risk (6,968 cases and 9,126 controls), three studies [22, 24, 27] for rs20417 (2,901 cases and 3,463 controls), and two studies [19, 22] for rs5277 (5,551 cases and 6,208 controls). Table 2 lists the main characteristics of studies identified for three SNP-based meta-analyses. No studied had a deviation from Hardy–Weinberg equilibrium in controls at a statistical significance level of 0.01.
Table 1

Characteristics of all the association studies regarding polymorphisms in the COX-2 gene and breast cancer risk

Study number

Author

Year

Country

Ethnicity

Cases (n = 10,482)

Controls (n = 12,732)

Studied SNPs

1

Abraham

2009

UK

Caucasian

4,470

4,560

rs4648310; rs689467; rs2206593; rs5275; rs5277; rs20424; rs4648276

2

Cox-1

2007

USA

Caucasian

1,270

1,762

rs20417; rs5277; rs20432; rs5275; rs4648298

3

Cox-2

2007

USA

Caucasian

317

634

rs5275

4

Cox-3

2007

USA

Caucasian

702

703

rs5275

5

Gallicchio

2006

USA

Caucasian

91

1,376

rs2143416; rs2745557; rs2206593; rs5275

6

Gao

2007

China

Asian

601

643

rs689466; rs20417; rs5275

7

Kang

2005

Korea

Asian

30

95

“−1166C>G”; “−1186T>G”

8

Langsenlehner

2006

Austria

Caucasian

500

500

rs5275

9

Li

2009

China

Asian

310

310

“169C>G”

10

Moorman

2005

USA

African American

763

678

rs5273 (Val511Ala)

11

Shen

2006

USA

Mixed

1,067

1,110

rs20417; rs20432; rs5275

12

Vogel

2006

Denmark

Caucasian

361

361

rs5275

Table 2

Characteristics of studies included for meta-analyses of the three SNPs

SNP

Study number

Author

Cases

Controls

HWE

MAF

AA*

Aa*

Aa*

AA

Aa

aa

rs5275

1

Abraham

927

985

260

996

1,010

259

0.903

0.337

2

Cox-1

541

567

141

699

808

213

0.383

0.359

3

Cox-2

140

131

30

270

259

81

0.134

0.345

4

Cox-3

281

296

67

278

294

79

0.925

0.347

5

Gallicchio

38

31

11

559

583

133

0.293

0.333

6

Gao

404

179

18

429

194

20

0.733

0.182

8

Langsenlehner

214

224

62

234

232

33

0.014

0.299

11

Shen

475

585a

467

635a

N.A.

N.A.

12

Vogel

167

150

44

155

165

41

0.770

0.342

 

Total

6,968

9,126

  

rs20417

2

Cox-1

865

336

42

1,185

485

45

0.580

0.167

6

Gao

526

73

2

582

59

2

0.699

0.049

11

Shen

670

387a

691

414a

N.A.

N.A.

 

Total

2,901

3,463

  

rs5277

1

Abraham

3,042

1,169

108

3,132

1,263

98

0.025

0.162

2

Cox-1

875

322

35

1,217

464

34

0.181

0.155

 

Total

5,551

6,208

  

Abbreviations: HWE Hardy–Weinberg equilibrium, MAF minor allele frequency, N.A. not available

* “A” denotes a major allele; “a” denotes a minor allele

aFor these just presenting the information for genotypes of AA and Aa + aa, dominant model was calculated only

Table 3 presents, in detail, the results of meta-analyses. For rs5275, the women harboring rare alleles were not associated with altered breast cancer risk either in a dominant model (pooled OR = 0.972, 95% CI: 0.911–1.037, fixed-effects) or in a recessive model (OR = 1.032, 95% CI: 0.856–1.244, fixed-effects). Furthermore, neither heterozygous carriers (OR = 0.979, 95% CI: 0.909–1.055) nor rare homozygous carriers (OR = 1.013, 95% CI: 0.830–1.235) had significantly different risk compared with common homozygous carriers. These results were not changed, when we stratified the studies by ethnicity (data not shown). Similar results were observed in terms of rs20417 and this polymorphism seemed to be not associated with breast cancer risk in any model (codominant, dominant, or recessive model). For rs5277, although, we again found this variant had no relationship with breast cancer risk in the dominant model, a trend of borderline significance was observed in the recessive model (OR = 1.217, 95% CI: 0.958–1.547, P = 0.107).
Table 3

Pooled ORs of the three SNPs in different models

Analysis model

rs5275 (n = 9; 6,968:9,126*)

rs20417 (n = 3; 2,901:3,463)

rs5277 (n = 2; 5,551:6,208)

OR (95% CI)

OR (95% CI)

OR (95% CI)

Codominant (Het. versus Common Hom.)

0.979 (0.909–1.055)

1.098 (0.773–1.561)

0.956 (0.881–1.037)

P/Pha

0.581/0.768

0.601/0.071

0.28/0.896

Codominant (Rare Hom. versus Common Hom.)

1.013 (0.830–1.235)

1.270 (0.835–1.932)

1.202 (0.945–1.529)

P/Ph

0.902/0.039

0.264/0.888

0.133/0.411

Dominant (Rare Hom. + Het. versus Common Hom.)

0.972 (0.911–1.037)

1.004 (0.898–1.122)

0.973 (0.899–1.054)

P/Ph

0.391/0.546

0.949/0.214

0.507/0.738

Recessive (Rare Hom. versus Het. + Common Hom.)

1.032 (0.856–1.244)

1.286 (0.848–1.953)

1.217 (0.958–1.547)

P/Ph

0.741/0.042

0.237/0.851

0.107/0.417

Abbreviations: Het. heterozygous, Hom. homozygous, OR odds ratio, CI confidence interval

* Study number and number of cases and controls

aPh denotes P-value of Q-test for heterogeneity test

In addition, for the most studied polymorphism rs5275, we evaluated the influence of any individual study on the overall OR (Fig. 1a). No individual study affected the overall OR dominantly, since omission of any single study made no materially huge difference. Moreover, Begg’s funnel plot and Egger’s test were performed to evaluate the publication bias of literatures. The shapes of the funnel plots did not reveal any evidence of obvious asymmetry (Fig. 1b), and the Egger’s test suggested the absence of publication bias (P = 0.499).
https://static-content.springer.com/image/art%3A10.1007%2Fs10549-009-0688-3/MediaObjects/10549_2009_688_Fig1_HTML.gif
Fig. 1

Outlier analysis and publication bias analysis for rs5275 a shows the influence of individual studies on the summary OR. The vertical axis indicates the overall OR and the two vertical axes indicate its 95% CI. Every hollow round indicates the pooled OR when the left study is omitted in this meta-analysis. The two ends of every broken line represent the respective 95% CI. b shows the Begg’s funnel plot of studies included in the meta-analysis. The vertical axis represents log [OR] and the horizontal axis means the standard error of log [OR]. Horizontal line and sloping lines in funnel plot represent fixed-effects summary OR and expected 95% CI for a given standard error, respectively. Area of each circle represents contribution of the study to the pooled OR

Discussion

The present meta-analysis systematically evaluated the association between the genetic variants in the COX-2 gene and breast cancer risk. Our results indicate that, among the most studied three SNPs (rs5275, rs20417, and rs5277), no one is a conspicuous low-penetrant risk factor for developing breast cancer, although rs5277 reached a borderline significance. Having analyzed a 2-fold larger number of studies than a previous meta-analysis [16], our results provide the most powerful conclusion thus far. It is important to point out that, the results of the present study are not in accordance with the previous Zhu’s report [16], which meta-analyzed the relationships between two SNPs (rs5275 and rs20417) and breast cancer, and concluded that the CC genotype of rs5275 was associated with a decreased risk compared with the TT or TC/TT genotypes. The potential explanation of discrepancy might be that the Zhu’s meta-analysis did not include recent large sample size studies [19] as well as ignored some previously eligible reports [17, 18]. Our current meta-analysis including all eligible studies warrants the reliability of outcomes.

The pathways of breast carcinogen metabolism are complex, mediated by the activities of multiple susceptibility genes. Therefore, it is plausible that the effect of a single gene might have more limited impact on breast cancer than that has been anticipated. The failure of demonstrating remarkable association between each of the three polymorphisms in the COX-2 gene and breast cancer risk does not rule out the possibility that other polymorphisms (or combinations of alleles) in the same gene could be substantially relevant to breast carcinogenesis. For instance, Li et al. [26] reported a novel SNP in exon 2 of the COX-2 gene associated with breast cancer. Gao et al. [24] also observed that no overall significant associations were detected in the single-locus analysis between three polymorphisms (−1195G>A, rs20417 (−765G>C), and rs5275 (8473C>T)) of the COX-2 and breast cancer risk; however, a significantly increased risk of breast cancer was associated with the haplotypes constructed by these three SNPs. Therefore, a more comprehensive haplotype-based approach rather than single polymorphism-based strategy may provide more precise information on genetic contribution of the COX-2 to breast cancer etiology.

Some limitations of this meta-analysis should be acknowledged. First, though most controls were selected mainly from healthy populations, some had benign breast disease [23], which might lead to misclassification bias because those controls have potential risks of developing breast cancer [28]. Second, the overall outcomes were based on individual unadjusted ORs, while a more precise evaluation should be adjusted by other potentially suspected factors. Third, although the available data on rs5275 and rs5277 comprise at least 5,500 cases and at least 6,000 controls, the present data on rs20417 are insufficient (approximately 3,000 cases and 3,500 controls) to enable reliable assessment yet. Relatively limited study number also made it impossible to perform subgroup analysis for rs5277 and rs20417. Finally, lack of individual data of each study prevents more detailed analyses, such as joint effects of SNP–SNP or gene–environment factors.

In summary, this meta-analysis suggests a lack of association between three SNPs (rs5275, rs5277, and rs20417) in the COX-2 gene and breast cancer risk. Further studies regarding other SNPs (or haplotypes) in the COX-2 gene and breast cancer risk are necessary to better understand the role of COX-2 in breast carcinogenesis. We also encourage further evaluations on the contribution of rs5277 to breast cancer risk in larger, more comprehensive, and well-designed association studies.

Acknowledgments

This research is supported by grants from the National Basic Research Program of China (2006CB910501) and the National Natural Science Foundation of China (30971143, 30972936).

Copyright information

© Springer Science+Business Media, LLC. 2009