SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References

The incidence of breast cancer in Japanese women has doubled in all age groups over the past two decades. We have recently shown that this marked increase is mostly due to an increase in the estrogen receptor (ER)-positive subtype. It is necessary to establish risk factors capable of predicting the risk of ER-positive breast cancer that will enable the efficient selection of candidates for preventive therapy. We analyzed genetic factors, including 14 single nucleotide polymorphisms (SNPs), environmental risk factors (body mass index, age at menarche, pregnancy, age at first birth, breastfeeding, family history of breast cancer, age at menopause, use of hormone replacement therapy, alcohol intake, and smoking), serum hormones and growth factors (estradiol, testosterone, prolactin, insulin-like growth factor 1 [IGF1] and IGF binding protein 3 [IGFBP3]), and mammographic density in 913 women with breast cancer and 278 disease-free controls. To identify important risk factors, risk prediction models for ER-positive breast cancer in both pre- and postmenopausal women were created by logistic regression analysis. In premenopausal women, one SNP (CYP19A1-rs10046), age, pregnancy, breastfeeding, alcohol intake, serum levels of prolactin, testosterone, and IGFBP3 were considered to be risk predictors. In postmenopausal women, one SNP (TP53-rs1042522), age, body mass index, age at menopause, serum levels of testosterone, and IGF1 were identified as risk predictors. Risk factors may differ between women of different menopausal status, and inclusion of common genetic variants and serum hormone measurements as well as environmental factors might improve risk assessment models. Further validation studies will clarify appropriate risk groups for preventive therapy. (Cancer Sci 2011; 102: 2065–2072)

Breast cancer is the most common cancer among women, not only in North America and Europe but also in Japan. Its incidence in Japanese women has doubled in all age groups over the past two decades, and we have recently shown that this marked increase in breast cancer incidence is mostly due an increase in the ER-positive subtype, especially in women aged 50 years or less.(1) Generally, the incidence remains only one-third of that seen in women in Western countries. However, the age-specific incidence in women less than 50 years of age is similar to that in the USA and the UK, and the peak age of incidence in Japanese women is approximately 45 years.(2)

Breast cancer subtypes, especially those defined by ER status, likely reflect etiologic differences.(3) Genome-wide association studies have identified genetic susceptibility loci for breast cancer according to ER status.(4–7) However, the frequencies of these variants differed markedly between ethnicities, and the common SNPs identified in Europeans were not associated with breast cancer risk in Japanese Americans.(5) We previously compared genetic polymorphisms of ERα, estrogen metabolism genes, and p53 between ER-positive and ER-negative Japanese breast cancer patients, and showed that polymorphisms of ERα, CYP19A1, COMT, and p53 frequently occurred in ER-positive breast cancer.(8) However, genetic factors such as SNPs have been considered to be associated with only a small to moderate increase in the risk of breast cancer,(9) and it has been suggested that changes in environmental factors, such as dietary and reproductive patterns, might be having a more significant effect on the increase in breast cancer incidence in Japanese women.(1)

The risk assessment tools have been used to predict the risk of breast cancer in North America and Europe, and prevention trials have shown that selective ER modulators lower ER-positive breast cancer incidence in women determined to be at increased risk based on the Gail model.(10) It was reported that most of the well-established reproductive risk factors, such as parity-related factors and age at menarche, seem more strongly associated with ER-positive breast cancer compared with ER-negative disease.(11,12) Moreover, recent analyses indicated that the Gail model has identified populations at increased risk of ER-positive breast cancer in postmenopausal women.(13) The penetration of risk factors may vary by age, menopausal status, and race. It has been reported that there were bimodal premenopausal and postmenopausal breast cancer populations divided by Clemmesen’s menopausal hook,(14) and the etiology of pre- and postmenopausal breast cancers is likely to be different.

Establishment of risk factors, both genetic and environmental, capable of predicting the risk of ER-positive breast cancer, which will then enable the efficient selection of candidates for hormone receptor-targeted preventive therapy, is urgently needed for Japanese women. In this study, we examined 14 genetic, 11 reproductive, and 5 circulating hormones as well as mammographic density, all of which had been determined by previous research to be risk factors for breast cancer. We estimated the effect of these factors on breast cancer risk stratified by ER and menopausal status, and created risk prediction models for ER-positive breast cancer separately for pre- and postmenopausal Japanese women to identify important risk factors.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References

Patients and samples.  The study included 913 consecutive Japanese women with breast cancer, newly diagnosed at Nagoya City University Hospital (Nagoya, Japan) between January 1992 and June 2010, and 278 control Japanese women who visited our hospital because of abnormalities found on breast screening or breast pain, and were confirmed to be without disease. The study protocol was approved by the institutional review boards and conformed with the guidelines of the 1996 Declaration of Helsinki. All breast cancer patients except for those with stage IV disease underwent surgical treatment (mastectomy or lumpectomy). Tumor samples from patients with stage IV disease were taken by core needle biopsies. Blood samples were taken before treatment. Family history of breast cancer was defined as positive if a first-degree relative had had breast cancer.

Genotyping.  Genomic DNA for genotyping was extracted from the whole blood samples using the Wizard SV Genomic DNA purification system (Promega, Madison, WI, USA) according to the manufacturer’s instructions. A total of 14 SNPs were analyzed: ESR1 (ERα)-rs6905370,(8)ESR1 (ERα)-rs827 421,(8)CYP2C19-rs4917623,(15)CYP17A1-rs743572,(8,16)CYP19A1 (aromatase)-rs10046,(8,17)HSD17B1-rs676387,(8)COMT-rs4680,(8)TP53 (p53)-rs1042522,(8)FGFR2-rs2981579,(18)MAP3K1-rs889312,(19)TNRC9 (TOX3)-rs3803662,(5) and chromosomes 5p12-rs10941679,(6) 6q25-rs3757318,(20) and 2q35-rs13387042.(5) All genotyping was carried out using TaqMan PCR assays (Applied Biosystems, Warrington, UK) in 96-well arrays that included blank wells as negative controls, according to the manufacturer’s instructions as described previously.(8) TaqMan Pre-Designed SNP Genotyping assays and TaqMan MGB probes were used. TaqMan PCR and genotyping analyses were carried out on the Applied Biosystems 9600 Emulation System.

Measurement of serum samples.  Blood samples were centrifuged at 1300g for 10 min, and the separated sera were stored in aliquots at −20°C. Serum samples were obtained from 372 breast cancer patients and 262 controls. Concentrations of estradiol, IGF1, and IGFBP3 were measured using commercially available direct radioimmunoassays: RIACOAT Estradiol US (Cisbio International, Gif-sur-Yvette, France) for estradiol; IGF-1 IRMA “Daiichi” (TFB, Tokyo, Japan) for IGF1; and IGFBP-3 “Cosmic” (Bioclone Australia, Sydney, Australia) for IGFBP3. Serum testosterone levels were measured by electrochemiluminescence immunoassay using Ecrusis Testosterone (Roche Diagnostics, Tokyo, Japan) and serum prolactin levels were measured by chemiluminescent immunoassay using Architect Prolactin (Abbott Japan, Matsudo, Japan), according to the manufacturers’ instructions.

Mammographic measurements.  Standard four-view bilateral mammograms were obtained and stored as digitized data from 543 breast cancer patients and 273 controls between 2004 and 2010. Percent density of each mammogram was measured as 100 × absolute dense area/total breast area in craniocaudal mammograms of right breasts or, if these were not available, of left breasts by a single trained expert (N.Y.) using Centricity imaging software (GE Healthcare, Hino, Japan).

Immunohistochemical evaluation of ER status.  The ER status of the breast cancer tissue samples was assessed by immunohistochemistry as described previously.(21) The primary antibody was a monoclonal mouse anti-human ERα antibody (1D5; Dako, Glostrup, Denmark) at 1:100 dilution. The Dako EnVision system (Dako EnVision labeled polymer, peroxidase) was used for detection. Tumors with 1% or more positive cells were considered positive.(21,22)

Statistical analysis.  Differences in environmental factors and mammographic density between breast cancer patients and controls were analyzed by Student’s t-test and the Chi squared test. The Mann–Whitney U-test was used to compare serum hormone levels between breast cancer patients and controls. To assess the strength of the associations between SNPs and breast cancer risk, odds ratios with 95% confidence intervals were estimated using logistic models. The genotype frequencies of all SNPs among the controls were found to occur in accordance with the Hardy–Weinberg law. Multivariate logistic-regression analyses were first carried out for all risk factors in a model, then stepwise selection variables were used based on the AIC, where AIC = −2 × (maximum log likelihood) + 2 × (total number of parameters). The probability of developing cancer was calculated for each woman using the different models, and ROC curves were generated. The estimated AUC was calculated for all models. Backward stepwise analyses were carried out until the AIC was increased by the removal of any variable in a particular step. All analyses were carried out using R statistical software (version 2.9.0) with the caTools package (http://www.updatestar.com/ja/publisher/R+Development+Core+Team). A two-tailed level of 5% was chosen for the type I error rate. The Kruskal–Wallis rank test and Spearman’s rank correlation test were used to analyze correlations among environmental factors, genotypes, serum biomarkers, and mammographic density. Statistical significance was set at a P-value of less than 5%, or set at ρ of less than −0.4 or more than 0.4 with Spearman’s rank correlation test.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References

The distribution of environmental factors among ER-positive and ER-negative breast cancer cases and controls stratified by menopausal status is shown in Tables 1 and 2. Of the premenopausal women included in these analyses, 300 had ER-positive breast cancer, 57 had ER-negative disease, and 178 were without breast cancer (Table 1). Unexpectedly, an increased number of pregnancies was associated with increased risk of both ER-positive (P = 0.0058) and ER-negative (P = 0.023) breast cancer. A lower number of children who had breastfed tended to correlate with increased risk of ER-positive breast cancer (P = 0.052). Older age (P < 0.0001) was significantly correlated with increased risk of ER-positive, but not ER-negative disease. Alcohol intake tended to be correlated with increased risk of ER-positive breast cancer (P = 0.055), but this was not statistically significant. Interestingly, higher BMI was correlated with increased risk of ER-negative disease (P = 0.0041). Of the postmenopausal women in these analyses, 430 had ER-positive breast cancer, 126 had ER-negative disease, and 100 were without breast cancer (Table 2). Older age (P = 0.013) and lower height (P = 0.02) were associated with increased risk of ER-positive, but not ER-negative breast cancer. A smaller number of breastfed children was significantly correlated with increased risk of both ER-positive (P = 0.0012) and ER-negative (P = 0.045) diseases. Higher BMI tended to be associated with increased risk of ER-positive disease (P = 0.055). Age at menopause was not statistically correlated with either ER-positive or ER-negative breast cancer risk.

Table 1.   Distribution of environmental factors among estrogen receptor (ER)-positive and ER-negative breast cancer cases and controls in premenopausal women
 ER+ cases (n = 300)ER− cases (n = 57)Controls (n = 178)P-values ER+ cases versus controlsP-values ER− cases versus controls
  1. *< 0.05 considered significant. Data are shown as the mean (SD) or n (%), and percentages exclude women with missing values.

Age (years)44.3 (5.6)43.2 (6.4)41.8 (6.2)<0.0001*0.1200
Height (cm)157.3 (5.3)156.7 (5.3)158.0 (5.1)0.15000.0840
Weight (kg)53.0 (8.2)55.0 (9.0)52.6 (7.3)0.58000.0440*
Body mass index (kg/m2)21.4 (3.2)22.5 (3.9)21.1 (2.9)0.23000.0041*
Age at menarche (years)12.7 (1.2)12.6 (1.2)13.0 (1.0)0.46001.0000
Pregnancy (number)1.8 (1.2)2.0 (1.3)1.4 (1.0)0.0058*0.0230*
Age at first birth (years)27.4 (4.6)26.9 (4.2)28.3 (4.0)0.15000.1500
Breastfeeding (number of children)1.3 (1.0)1.4 (1.1)1.6 (1.0)0.05200.5000
Family history of breast cancer
 Yes29 (12.6%)5 (11.9%)31 (17.7%)0.0150*0.3600
 No202 (87.4%)37 (88.1%)144 (82.3%)
 Unknown69153
Alcohol intake
 Current or former83 (55.0%)7 (33.3%)78 (44.3%)0.05500.3400
 Never68 (45.0%)14 (66.7%)98 (55.7%)
 Unknown149362
Smoking
 Current or former38 (25.5%)2 (9.5%)31 (17.6%)0.08300.3500
 Never111 (74.5%)19 (90.5%)145 (82.4%)
 Unknown151362
Table 2.   Distribution of environmental factors among estrogen receptor (ER)-positive and ER-negative breast cancer cases and controls in postmenopausal women
 ER+ cases (n = 430)ER− cases (n = 126)Controls (n = 100)P-values ER+ cases versus controlsP-values ER− cases versus controls
  1. *< 0.05 considered significant. Data are shown as the mean (SD) or n (%), and percentages exclude women with missing values.

Age (years)64.7 (9.5)62.9 (7.7)62.1 (8.2)0.0130*0.470
Height (cm)152.0 (6.1)153.3 (5.8)154.1 (5.8)0.0200*0.270
Weight (kg)53.5 (8.3)52.9 (8.3)53.3 (8.2)0.82000.700
Body-mass index (kg/m2)23.2 (3.6)22.5 (3.5)22.4 (3.2)0.05500.820
Age at menarche (years)13.8 (1.7)13.7 (1.5)13.5 (1.5)0.14000.600
Pregnancy (number)2.0 (1.2)2.3 (1.1)2.3 (1.6)0.11000.910
Age at first birth (years)25.4 (3.3)24.3 (2.9)25.5 (3.6)0.78000.088
Breastfeeding (number of children)1.5 (1.1)1.6 (1.0)1.9 (0.9)0.0012*0.045*
Age at menopause (years)50.4 (3.3)51.4 (2.9)51.2 (3.9)0.09400.620
Family history of breast cancer
 Yes35 (10.8%)5 (5.6%)18 (18.6%)0.042*0.068
 No290 (89.2%)85 (94.4%)79 (81.4%)
 Unknown105363
Alcohol intake
 Current or former52 (29.1%)10 (30.3%)30 (30.6%)0.7900.970
 Never127 (70.9%)23 (69.7%)68 (69.4%)
 Unknown251932
Smoking
 Current or former30 (17.2%)3 (9.4%)16 (16.2%)0.8200.340
 Never144 (82.8%)29 (90.6%)83 (83.8%)
 Unknown256951
Hormone replacement therapy use
 Current or former10 (5.8%)1 (3.1%)12 (12.4%)0.0580.130
 Never163 (94.2%)31 (96.9%)85 (87.6%)
 Unknown257943

Fourteen SNPs previously reported to be associated with ER-positive breast cancer were genotyped in breast cancer cases and controls (Table 3). The data reported from the HapMap project showed that significant differences were observed in the genotypes of rs6905370, rs10046, rs4680, rs1042522, rs889312, rs3803662, rs10941679, rs3757318, and rs13387042 between Europeans and Japanese. The frequencies of alleles of each SNP did not differ between controls in this study and Japanese in the HapMap project. The results of SNP genotyping, indicating differences between ER-positive breast cancer cases and controls, are shown in Tables 4 and 5. In premenopausal women, CYP19A1-rs10046 (= 0.016 in the codominant model, = 0.019 in the recessive model) and TNRC9 (also known as TOX3)-rs3803662 (= 0.027 in the dominant model) had significant associations with ER-positive breast cancer (Table 4), whereas ESR1-rs6905370 (= 0.033 in the recessive model) and TP53-rs1042522 (= 0.045 in the recessive model) had significant associations with ER-positive breast cancer in postmenopausal women (Table 5).

Table 3.   Fourteen genomic loci associated with breast cancer risk
Genomic lociGene/location ER+ cases (%)ER− cases (%)Controls (%)HapMap
Japanese% (n = 90)European% (n = 120)P-values
  1. *< 0.05 considered significant. The genotype frequencies of all SNPs among the controls were in accordance with the Hardy-Weinberg law. –, no gene identifier.

rs6905370ESR1 (ERα)/6q25AA171 (26.4)44 (25.9)71 (27.2)20.08.30.0021*
AG321 (49.5)88 (51.8)138 (52.9)48.938.3
GG156 (24.1)38 (22.4)52 (19.9)31.153.3
rs827421ESR1 (ERα)/6q25CC101 (17.0)24 (21.2)38 (14.4)15.621.70.5700
CT294 (49.4)47 (41.6)138 (52.3)51.146.7
TT200 (33.6)42 (37.2)88 (33.3)33.331.7
rs4917623CYP2C19/10q23CC143 (21.8)44 (31.4)70 (25.4)35.623.30.1200
CT345 (52.6)60 (42.9)138 (50.0)46.751.7
TT168 (25.6)36 (25.7)68 (24.6)17.825.0
rs743572CYP17A1/10q24AA207 (30.0)42 (24.4)75 (28.5)36.437.90.4000
AG334 (48.5)91 (52.9)137 (52.1)52.344.8
GG148 (21.5)39 (22.7)51 (19.4)11.417.2
rs10046CYP19A1/5q21CC210 (30.5)29 (21.0)97 (35.0)37.818.20.0070*
CT353 (51.3)74 (53.6)120 (43.3)40.052.7
TT125 (18.2)35 (25.4)60 (21.7)22.229.1
rs676387HSD17B1/17q21GG205 (30.0)53 (30.8)65 (24.8)35.148.10.1200
GT324 (47.4)93 (54.1)133 (50.8)40.536.5
TT155 (22.7)26 (15.1)64 (24.4)24.315.4
rs4680COMT/22q11AA55 (8.9)15 (9.1)26 (10.1)4.525.0<0.0001*
AG284 (46.0)66 (40.2)119 (46.3)38.653.3
GG279 (45.1)83 (50.6)112 (43.6)56.821.7
rs1042522TP53 (p53)/17p13GG253 (40.7)51 (36.4)111 (43.0)40.961.70.0024*
CG290 (46.7)66 (47.1)106 (41.1)36.430.0
CC78 (12.6)23 (16.4)41 (15.9)22.78.3
rs2981579FGFR2/10q26CC181 (28.3)47 (34.6)80 (30.3)46.237.30.4300
CT328 (51.2)59 (43.4)143 (54.2)35.940.7
TT131 (20.5)30 (22.1)41 (15.5)17.922.0
rs889312MAP3K1/5p15AA109 (17.0)19 (13.9)37 (14.0)13.646.7<0.0001*
AC300 (46.9)72 (52.6)140 (53.0)63.645.0
CC231 (36.1)46 (33.6)87 (33.0)22.78.3
rs3803662TNRC9/TOX3/16q12TT235 (35.5)50 (36.0)75 (27.6)33.311.7<0.0001*
CT309 (46.7)67 (48.2)150 (55.1)44.436.7
CC118 (17.8)22 (15.8)47 (17.3)22.251.7
rs10941679–/5p12AA127 (19.9)34 (25.4)57 (21.4)22.240.0<0.0001*
AG327 (51.3)68 (50.7)142 (53.4)55.055.0
GG183 (28.7)32 (23.9)67 (25.2)22.25.0
rs3757318–/6q25AA42 (6.6)8 (5.9)14 (5.3)6.81.70.0079*
AG219 (34.5)57 (41.9)98 (37.1)25.011.7
GG374 (58.9)71 (52.2)152 (57.6)68.286.7
rs13387042–/2q35AA11 (1.7)6 (4.3)5 (1.8)4.440.0<0.0001*
AG137 (20.7)30 (21.6)53 (19.6)13.346.7
GG513 (77.6)103 (74.1)213 (78.6)82.213.3
Table 4.   Associations between single nucleotide polymorphism genotyping and estrogen receptor-positive breast cancer risk in premenopausal women
  Cases (%)Controls (%)OR (95% CI)P-values
  1. *P < 0.05 considered significant. CI, confidence interval; OR, odds ratio; ref., reference.

rs10046 (CYP19A1)
 CodominantCC90 (32.0)63 (35.8)1 (ref.)0.016*
CT154 (54.8)75 (42.6) 1.44 (0.94–2.20)
TT37 (13.2)38 (21.6) 0.68 (0.39–1.19)
 DominantCC90 (32.0)63 (35.8)1 (ref.)0.410
CT+TT191 (68.0)113 (64.2) 1.18 (0.80–1.76)
 RecessiveCC+CT244 (86.8)138 (78.4)1 (ref.)0.019*
TT37 (13.2)38 (21.6) 0.55 (0.33–0.91)
rs3803662 (TNRC9)
 CodominantAA102 (38.1)48 (27.9)1 (ref.)0.083
AG130 (48.5)99 (57.6) 0.62 (0.40–0.95)
GG36 (13.4)25 (14.5) 0.68 (0.37–1.25)
 DominantAA102 (38.1)48 (27.9)1 (ref.)0.027*
AG+GG166 (61.9)124 (72.1) 0.63 (0.42–0.95)
 RecessiveAA+AG232 (86.6)147 (85.5)1 (ref.)0.740
GG36 (13.4)25 (14.5) 0.91 (0.53–1.58)
Table 5.   Associations between single nucleotide polymorphism genotyping and estrogen receptor-positive breast cancer risk in postmenopausal women
  Cases (%)Controls (%)OR (95% CI)P-values
  1. CI, confidence interval; OR, odds ratio; ref., reference. *P < 0.05 considered significant.

rs6905370 (ESR1)
 CodominantAA94 (24.2)28 (28.9)1 (ref.)0.100
AG191 (49.2)53 (54.6) 1.07 (0.64–1.81)
GG103 (26.5)16 (16.5) 1.92 (0.98–3.77)
 DominantAA94 (24.2)28 (28.9)1 (ref.)0.350
AG+GG294 (75.8)69 (71.1) 1.27 (0.77–2.09)
 RecessiveAA+AG285 (73.5)81 (83.5)1 (ref.)0.033*
GG103 (26.5)16 (16.5) 1.83 (1.02–3.27)
rs1042522 (TP53)
 CodominantGG152 (40.9)41 (43.2)1 (ref.)0.062
CG173 (46.5)34 (35.8) 1.37 (0.83–2.27)
CC47 (12.6)20 (21.1) 0.63 (0.34–1.19)
 DominantGG152 (40.9)41 (43.2)1 (ref.)0.690
CG+CC220 (59.1)54 (56.8) 1.1 (0.70–1.73)
 RecessiveGG+CG325 (87.4)75 (78.9)1 (ref.)0.045*
CC47 (12.6)20 (21.1) 0.54 (0.30–0.97)

Table 6 shows correlations between circulating hormone levels and breast cancer risk stratified by ER and menopausal status. Serum testosterone levels were significantly higher in both ER-positive and ER-negative breast cancer cases than in controls among premenopausal women (P < 0.0001 and P = 0.026, respectively) and postmenopausal women (P < 0.0001 and P = 0.0003, respectively). Moreover, serum prolactin levels were significantly higher in both ER-positive and ER-negative breast cancer cases than in controls in premenopausal women (P < 0.0001 and P = 0.0033, respectively), whereas serum prolactin levels were significantly higher in ER-positive (P = 0.007), but not ER-negative (P = 0.75), breast cancer cases than in controls in postmenopausal women. Serum estradiol levels, differences in which are difficult to assess in premenopausal women because of the menstrual cycle, were significantly higher in both ER-positive and ER-negative breast cancer cases than in controls in postmenopausal women (P = 0.0005 and P = 0.0059, respectively). Serum IGF1 levels tended to be higher in both ER-positive and ER-negative breast cancer cases than in controls in postmenopausal women (P = 0.072 and P = 0.058, respectively), but they were not statistically significant.

Table 6.   Associations between circulating hormone levels and breast cancer risk
 ER+ cases (n = 142)ER− cases (n = 21)Controls (n = 167)P-values versus controls
Mean (SD)Mean (SD)Mean (SD)ER+ casesER− cases
  1. *< 0.05 considered significant. †Data are expressed as the median (interquartile range). ER+, estrogen receptor-positive; ER−, estrogen receptor-negative; IGF1, insulin-like growth factor 1; IGFBP3, IGF binding protein 3.

Premenopausal women
 Estradiol (pg/mL) 62.0 (56.9) 46.3 (57.7) 69.5 (58.0)0.08600.0058*
 Testosterone (ng/mL)   0.39 (0.22)   0.36 (0.16)   0.28 (0.16)< 0.0001*0.0260*
 Prolactin (ng/mL)   31.39 (41.16)   21.69 (12.44)  14.23 (7.56)< 0.0001*0.0033*
 IGF1 (ng/mL)150.0 (52.0)157.0 (56.0)150.0 (43.0)0.80000.6500
 IGFBP3 (μg/mL)   2.65 (0.40)   2.79 (0.48)   2.60 (0.35)0.19000.1500
 (n = 169) Mean (SD)(n = 40) Mean (SD)(n = 95) Mean (SD) 
Postmenopausal women
 Estradiol (pg/mL)†    6.2 (4.4–8.0)     6.1 (4.8–7.4)    4.8 (2.7–6.9)0.0005*0.0059*
 Testosterone (ng/mL)   0.33 (0.30)   0.33 (0.23)   0.19 (0.11)< 0.0001*0.0003*
 Prolactin (ng/mL)   13.56 (12.77)   11.89 (17.49)   11.84 (12.53)0.0070*0.7500
 IGF1 (ng/mL)116.0 (51.0)117.0 (43.0)101.0 (42.0)0.07200.0580
 IGFBP3 (μg/mL)   2.50 (0.45)   2.60 (0.52)   2.48 (0.44)0.86000.3000

Mammographic density was compared between ER-positive and ER-negative breast cancer cases and controls stratified by menopausal status (Table 7). There was no difference in mammographic density between ER-positive cases and controls in both pre- and postmenopausal women. In contrast, it was significantly lower in ER-negative cases than in controls in premenopausal women (P = 0.035).

Table 7.   Associations between mammographic density and breast cancer risk
 ER+ cases (n = 209)ER− cases (n = 22)Controls (n = 174)P-values versus controls
Mean (SD)Mean (SD)Mean (SD)ER+ casesER− cases
  1. *P < 0.05 considered significant. ER+, estrogen receptor-positive; ER−, estrogen receptor-negative.

Premenopausal women
 Whole breast (mm2)6458 (2266)8364 (4624)6840 (2637)0.1300.022
 Dense area (mm2)3610 (1550)3861 (2186)3899 (1621)0.0760.920
 Percent density (%)56.8 (16.4)50.4 (19.5)59.3 (18.3)0.1600.035*
 (n = 265) Mean (SD)(n = 47) Mean (SD)(n = 99) Mean (SD) 
Postmenopausal women
 Whole breast (mm2)8438 (3237)8484 (3723)8619 (3686)0.650.84
 Dense area (mm2)2996 (1471)2985 (1427)3166 (1658)0.350.52
 Percent density (%)37.6 (16.2)37.9 (17.0)39.8 (18.0)0.260.54

We next analyzed correlations between the various risk factors measured in this study. ESR1-rs827421, CYP17A1-rs743572, and MAP3K1-rs889312 were associated with BMI (P = 0.027, P = 0.038, and P = 0.011, respectively), whereas CYP2C19-rs4917623 was correlated with height (P = 0.022). 5p12-rs10941679 was associated with age at menarche (P = 0.019) and 6q25-rs3757318 was associated with age at menopause (P = 0.004). In premenopausal women, ESR1-rs6905370 was associated with serum IGF1 levels (P = 0.032), whereas CYP17A1-rs743572 was correlated with serum testosterone levels (P = 0.015). In addition, HSD17B1-rs676387 and TP53-rs1042522 were associated with serum prolactin levels (P = 0.039 and P = 0.035, respectively). TNRC9-rs3803662 was associated with serum estradiol levels in postmenopausal women (P = 0.027). In premenopausal women, the presence of a family history of breast cancer was positively correlated with serum testosterone levels (P = 0.004), whereas alcohol intake was positively correlated with serum testosterone and IGF1 levels (P = 0.008 and P = 0.031, respectively). Alcohol intake was positively correlated with serum IGFBP3 levels (P = 0.002), and smoking was positively correlated with serum testosterone levels (P = 0.033) in postmenopausal women. Serum estradiol and testosterone levels were positively correlated in postmenopausal women (P < 0.0001). Mammographic density was inversely correlated with BMI in both pre- and postmenopausal women (Spearman’s correlation coefficient; −0.532 and −0.410, respectively).

We then created risk prediction models for ER-positive breast cancer for pre- and postmenopausal women separately by logistic regression analysis. First, all parameters such as environmental factors, genotypes, serum hormone levels, and mammographic density were included to create the full models, of which the AUCs were 0.848 for premenopausal women and 0.772 for postmenopausal women. The best models were established using stepwise selection variables, calculated as follows: +0.0362 (age) +0.499 (pregnancy) −0.124 (breastfeeding) +0.110 (alcohol intake) +0.110 (rs10046 CT+TT) −0.00578 (rs10046 CC) +0.811 (testosterone) +0.00337 (prolactin) +0.134 (IGFBP3) defined as 0.842 of AUC for premenopausal women (Fig. 1), and +0.120 (age) +0.0141 (BMI) −0.0195 (age at menopause) −0.198 (rs1042522 CC) +0.395 (testosterone) +0.00277 (IGF1) defined as 0.770 of AUC for postmenopausal women (Fig. 2).

image

Figure 1.  Receiver operating characteristic curve of the best risk prediction model for estrogen receptor-positive breast cancer in premenopausal women. The area under the curve (AUC) is estimated as 0.842. 1-specificity, 1 minus specificity.

Download figure to PowerPoint

image

Figure 2.  Receiver operating characteristic curve of the best risk prediction model for estrogen receptor-positive breast cancer in postmenopausal women. The area under the curve (AUC) is estimated as 0.770. 1-specificity, 1 minus specificity.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References

Recent gene expression-based molecular classification has revealed that there are large-scale molecular differences between ER-positive and ER-negative breast cancers that reach far beyond the presence or absence of ER.(23,24) In addition, Clemmesen’s hook, shown by the bimodal dip of the age frequency histogram, suggests there may be distinct mechanisms for breast cancer development between pre- and postmenopausal women.(14) In this study, we analyzed genetic and environmental factors, including 14 SNPs, serum levels of circulating hormones, and mammographic density among Japanese breast cancer patients and controls, and created risk prediction models for ER-positive breast cancer separately in pre- and postmenopausal women to identify important risk factors.

As expected, factors that were included in our risk prediction models differed between pre- and postmenopausal women, with the exceptions of serum testosterone levels and age. Of the reported SNPs, CYP19A1 (aromatase)-rs10046 for premenopausal women and TP53 (p53)-rs1042522 for postmenopausal women were included in our models. Zhang et al.(25) reported that the rs10046 T carriers were associated with premenopausal cancer risk, particularly for ER-positive, PgR-positive tumors in Chinese women. Our results also showed that rs10046 T carriers and TT genotype increased ER-positive breast cancer risk in premenopausal women. It was reported that the T allele of this gene is associated with the level of aromatase mRNA in breast tumors and it is considered a high activity genotype.(26) We showed that TP53-rs1042522 CC genotype (Pro72) decreased ER-positive breast cancer risk in postmenopausal women. It has been shown that the Arg72 variant (GG genotype) of TP53 induces apoptosis much more strongly than does the Pro72 variant (CC genotype),(27) and that Arg72 alleles have the greatest effect on apoptosis and fertility.(28) Interestingly, sharp ethnic differences are observed for these two SNPs, the CC genotype of CYP19A1 and the Pro72 variant (CC genotype) of TP53 being less prevalent in Europeans than in Japanese according to the HapMap data (Table 3). These polymorphisms might affect breast cancer incidence and different SNPs could contribute to breast cancer risk among different ethnic groups.

Of the possible environmental factors, pregnancy (number of times), breastfeeding (number of children), and alcohol intake, for premenopausal women, and BMI and age at menopause, for postmenopausal women, were included in our models. Surprisingly, our results showed that an increased number of pregnancies increased the ER-positive breast cancer risk in premenopausal women. Moreover, age at menopause was lower in women with ER-positive breast cancer than in controls. Iwasaki et al.(29) from the Japan Public Health Center-based prospective study reported that early age at menarche for premenopausal women, late age at natural menopause for postmenopausal women, and nulliparity and low parity for both premenopausal and postmenopausal women were significantly associated with an increased risk of breast cancer. The Life Span Study of the Radiation Effects Research Foundation in Hiroshima and Nagasaki showed an inverse association between the number of births and total period of breastfeeding, and breast cancer risk in Japanese women.(30) Although it is difficult to access which factors are important at present, reproductive factors that affect ER-positive breast cancer risk seem to be different between pre- and postmenopausal women. Furthermore, the total fertility rate has decreased since the 1970s from an estimated 2.16 in 1971 to an estimated 1.32 in 2006 in Japan. This phenomenon should be considered when evaluating reproductive factors and breast cancer risk. In addition, controls in this study were women who visited our hospital because of abnormalities found on breast screening or breast pain, and were confirmed to be without disease. Thus, women with a relatively high risk of breast cancer might preferentially be recruited as controls. The medical background of the controls may have introduced bias into the study, although this would have had only limited impact. It will be necessary to analyze healthy women who visit medical examination centers for breast screening.

High mammographic density is an established risk factor for breast cancer in Western countries.(31) Our results indicated that there was no correlation between breast cancer risk and mammographic density in either pre- or postmenopausal women. However, IGFBP3 in premenopausal women and IGF1 in postmenopausal women were included in our risk prediction models. Components of the IGF axis are considered to influence the tissue composition of the breast.(32) It has been shown that high levels of circulating IGF1, or higher IGF1 in relation to IGFBP3, correlated with higher premenopausal mammographic density, similarly to their association with breast cancer risk.(32) It has also been reported that an IGF1 gene polymorphism (rs6220) correlated with higher mammographic density and higher serum IGF1 levels in premenopausal women.(33) We showed that ESR1-rs6905370 polymorphism and alcohol intake are associated with serum levels of IGF1 in premenopausal women. Renehan et al.(34) and the Endogenous Hormones and Breast Cancer Collaborative Group(35) reported meta-analyses of associations with IGF1 and IGFBP3 and breast cancer risk, and revealed associations between circulating IGF1 with breast cancer risk in both pre- and postmenopausal women, whereas no evidence of interactions between IGFBP3 and breast cancer risk was found.

Serum levels of prolactin as well as IGFBP3 were included in our models for premenopausal women. Tworoger et al.(36,37) showed that high prolactin levels were associated with a 60% increased risk of ER-positive breast cancer, and that epidemiologic data suggested prolactin is involved in breast cancer etiology in both pre- and postmenopausal women. They also reported a confirmed correlation between prolactin levels and nulliparity and oral contraceptive use, and a possible association between prolactin levels and family history of breast cancer and increased mammographic density. Our results indicated that HSD17B1-rs676387 and TP53-rs1042522 polymorphisms were associated with serum prolactin levels, but there were no correlations between epidemiologic factors and serum prolactin levels.

We found that high serum testosterone levels increased the risk of ER-positive and ER-negative breast cancer in both pre- and postmenopausal women, and that serum testosterone levels were correlated with CYP17A1-rs743572 polymorphism and with alcohol intake in premenopausal women, and with smoking in postmenopausal women. A prospective study within the Nurses’ Health Study II reported that higher levels of serum estrogens and testosterone levels were associated with risk of breast cancer, especially ER-positive and PgR-positive breast cancer in premenopausal women.(38) Three other prospective studies also showed that premenopausal women with elevated serum testosterone levels were at an increased risk of breast cancer.(39–41) In postmenopausal women, a strong positive association between breast cancer risk and higher levels of serum estrogens and androgens has now been confirmed.(42) Preclinical data indicate that testosterone has dual effects on breast tumorigenesis: a proliferative effect mediated by the ER; and an antiproliferative effect mediated by the androgen receptor.(43) Therefore, testosterone has the potential to regulate the growth of both ER-positive and ER-negative tumors.

In conclusion, in the present study, we created risk prediction models for ER-positive breast cancer for both pre- and postmenopausal Japanese women to identify important risk factors. Our results suggest that risk factors differ between women of different menopausal status, and that inclusion of common genetic variants (SNPs) and serum hormone measurements as well as environmental factors might improve risk assessment models. Further validation studies will clarify appropriate risk groups for preventive therapy.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References

This work was supported in part by a Grant-in Aid for Scientific Research from the Japan Society for the Promotion of Science.

Disclosure Statement

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References

The authors have no conflict of interest to declare.

Abbreviations
AIC

Akaike information criterion

AUC

area under the ROC curve

BMI

body mass index

ER

estrogen receptor

IGF1

insulin-like growth factor 1

IGFBP3

IGF binding protein 3

PgR

progesterone receptor

ROC

receiver operating characteristic

SNP

single nucleotide polymorphism

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgment
  7. Disclosure Statement
  8. References