Breast Cancer

, Volume 22, Issue 4, pp 399–405

Association of body mass index with risk of luminal A but not luminal B estrogen receptor-positive and HER2-negative breast cancer for postmenopausal Japanese women


  • Yoshimasa Miyagawa
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Tomohiro Miyake
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Ayako Yanai
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Keiko Murase
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Michiko Imamura
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Shigetoshi Ichii
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Yuichi Takatsuka
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
  • Takashi Ito
    • Surgical PathologyHyogo College of Medicine
  • Seiichi Hirota
    • Surgical PathologyHyogo College of Medicine
  • Masaru Saito
    • Maikodai Hospital
  • Yoshinao Kotoura
    • Maikodai Hospital
  • Keisuke Miyauchi
    • Miyauchi Clinic
  • Yasuhisa Fujimoto
    • Tachibana Hospital
  • Takuya Hatada
    • Uminosato Clinic
  • Mitsunori Sasa
    • Tokushima Breast Care Clinic
    • Division of Breast and Endocrine, Department of SurgeryHyogo College of Medicine
Original Article

DOI: 10.1007/s12282-013-0493-z

Cite this article as:
Miyagawa, Y., Miyake, T., Yanai, A. et al. Breast Cancer (2015) 22: 399. doi:10.1007/s12282-013-0493-z



The impact of body mass index (BMI) on the risk of postmenopausal estrogen receptor (ER)-positive breast cancers has been well documented. However, the mechanism for the impact of BMI on the etiology of luminal A and luminal B subtypes has not yet been identified.


We analyzed associations between BMI and breast cancers stratified by immunohistochemically defined intrinsic subtypes, and 1,297 Japanese women (615 breast cancer patients and 682 healthy women from a breast cancer screening program) were enrolled in a case–control study. ER-positive/human epidermal growth factor receptor 2 (HER2)-negative breast cancers were classified into luminal A and B subtypes according to Ki67 expression levels.


Higher BMI was significantly positively associated with postmenopausal breast cancer risk for one-unit increase in BMI (adjusted odds ratio (aOR) 1.09, 95 % confidence interval (CI) 1.04–1.15; P = 0.0008). Analyses of postmenopausal women revealed that BMI was consistently and exclusively associated with luminal A incidence (aOR 1.18, 95 % CI 1.10–1.26; P < 0.0001). When BMI was divided into three categories corresponding to those of controls, among postmenopausal women, the observed positive association was confined to luminal A (high vs low, aOR 2.98, 95 % CI 1.53–5.80; P < 0.005), but not luminal B (aOR 0.95, 95 % CI 0.47–1.91) subtypes.


We observed that BMI was significantly positively associated with increased risk of postmenopausal breast cancer for Japanese women with luminal A, but not with luminal B tumor subtype.


Body mass indexBreast cancer riskEstrogen receptorLuminal A


The impact of both genetic and environmental risk factors on breast cancer etiology has been well established. In particular, positive associations of reproductive risk factors such as early menarche, late menopause, nulliparity, and postmenopausal obesity are reportedly stronger for estrogen receptor-positive (ER+) than for ER-negative (ER−) breast cancers [1, 2]. These findings suggested that the incidence of breast cancers differs depending on ER status. Since evidence linking circulating estrogen levels to breast cancer risk has been established in prospective studies [35], higher levels of estrogens as well as longer estrogen exposure are considered to be associated with higher risks of ER+ breast cancers. Considering the fact that the estrogen status of premenopausal and postmenopausal women differs, the etiology of these breast cancers also might be different depending on menopausal status.

Recently, breast cancer incidence has been assessed with the focus on body mass index (BMI) and taking menopausal status as well as ER status into consideration. According to these assessments, an increase in BMI was positively associated with breast cancer risk in postmenopausal women [6, 7], whereas this null or inverse association was found in premenopausal women [6, 8]. Phipps et al. [9] reported that for postmenopausal women, BMI was associated with the risk of ER+ breast cancers [odds ratio (OR) for highest versus lowest quartiles 1.7, 95 % confidence interval (95 % CI) 1.2–2.4] and possibly associated with risk of triple-negative breast cancer (OR 2.7, 95 % CI 1.0–7.5). Suzuki et al. [10] conducted a meta-analysis and found that ER+ and progesterone receptor-positive (PR+) breast cancers showed a 33 % higher risk for postmenopausal women for each five-unit increase in BMI, but found no such association for ER+ and PR− breast cancers. These results suggested that the etiology of ER+/PR+ and ER+/PR− breast cancers is different.

With the aid of gene expression profiling, breast cancers with ER+ and human epidermal growth factor receptor 2-negative (HER2−) status have been further classified into luminal A and luminal B subtypes [11]. Although many ER+/PR+ cancers are likely to be luminal A rather than luminal B cancers, these two types of cancers cannot be differentiated on the basis of PR expression. Furthermore, response to preoperative treatment with aromatase inhibitors was found to be associated with ER expression but not with PR expression [12]. These results suggested that PR expression does not play a role as a biomarker for response to preoperative treatment with aromatase inhibitors. Instead, classification based on intrinsic subtypes may be more effective for estimating estrogen dependency. Therefore, elucidation of the etiology of intrinsic subtypes appears to be of vital importance.

Since there is a direct correlation between postmenopausal obesity and endogenous estrogen levels [13, 14], it is reasonable to speculate that postmenopausal BMI is closely related to risk of ER+ breast cancers mediated through estrogens. On the other hand, Furberg et al. [15] reported that serum high-density lipoprotein cholesterol was inversely associated with risk of postmenopausal breast cancers. In addition, insulin levels have been found to be positively associated with postmenopausal breast cancer risk (hazard ratio of highest vs lowest quartile 2.40, 95 % CI 1.30–4.41; P < 0.01) [16]. Since BMI is significantly positively associated with serum levels of cholesterol and insulin, BMI may be involved in the incidence of postmenopausal breast cancer mediated through not only estrogens, but also through these other factors.

Although biological differences between luminal A and luminal B breast cancers are prominent, etiological differences have yet to be elucidated. In the study presented here, we therefore examined the association of BMI with breast cancer incidence especially in terms of intrinsic subtypes.

Materials and methods

Cases and controls

Eligible cases were 615 Japanese women with invasive breast cancers who were consecutively treated with mastectomy or breast-conserving surgery at the Hyogo College of Medicine between February 2005 and November 2012 and at the Tokushima Breast Clinic between May 2009 and April 2011. For this study, 223 premenopausal and 392 postmenopausal women with a mean age of 57.9 years (standard deviation (SD) 13.2, range 20–92) were enrolled. Histological diagnosis of breast cancer was confirmed in each case and identified 578 invasive ductal carcinomas, 16 invasive lobular carcinomas and 21 other types. Patients with non-invasive carcinoma were not included. Controls were 682 Japanese healthy women who participated in the breast cancer screening program between April 2008 and March 2010 at affiliated institutes in Hyogo Prefecture. Controls comprised 353 premenopausal and 329 postmenopausal women with a mean age of 50.5 years (SD 11.3, range 25–82). All controls were confirmed to be free from breast cancer by means of mammography and physical examination. All anthropometric and epidemiological data were collected by self-administered questionnaire at the first visit to an institute. This study was approved by the Ethics Committee of Hyogo College of Medicine.

Immunohistochemical staining and classification of subtypes

Subtypes were identified by means of immunohistochemical staining of ER, HER2, and Ki67 expression levels. Formalin-fixed, paraffin-embedded tissues were sliced and used for immunohistochemical staining. Expression levels of ER (1D5; Dako, Glostrup, Denmark), HER2 (Hercep Test; Dako), and Ki67 (MIB1; Dako) were determined immunohistochemically in terms of the percentage of positive cancer cells in the nuclei for ER and Ki67, and by means of membrane staining for HER2 using automated immunostainers (BOND-MAX; Leica Microsystems, Tokyo, Japan for ER. Autostainer; Dako, Tokyo, Japan for HER2 and Ki67). When nuclear stained cells accounted for 10 % or more of all cells, we considered the tissue to be positive for ER, whereas HER2-positive was defined as HER2 (3+), or HER2 (2+) and FISH-positive.

Since the biology of tumors with ER <10 % and ER ≥10 % seems to be different [17], we used a cutoff value not of 1 %, but of 10 %. Luminal A and luminal B ER-positive and HER2-negative breast cancers were classified in terms of Ki67 expression levels as reported by Cheang et al. [18]. The subtypes were defined as luminal A, i.e., ER+/HER2−, Ki67 <14 %; luminal B, i.e., ER+/HER2−, Ki67 ≥14 %; HER2, i.e., ER+/HER2+ and ER−/HER2+; and triple negative, i.e., TN, ER−/HER2−. Since only a small number of HER2-positive breast cancers (109 cases) were recruited for this study, ER+/HER2+ and ER−/HER2+ cancers were analyzed in combination with the HER2 subtype.

Statistical analysis

The relationship between BMI, calculated as weight (kilograms)/[height (meters)]2, and breast cancer risk was determined with the logistic regression method to obtain the OR and 95 % CI. We adjusted for age (continuous), age at menarche (continuous), parity (yes or no), family history within second-degree relatives of breast cancer (yes or no), and age at menopause for postmenopausal women (continuous). Statistical significance was set at P < 0.05 and JMP10 (SAS Institute Japan, Tokyo, Japan) was used for all the analyses.


Epidemiological characteristics in cases and control subjects according to breast cancer subtypes

Breast cancer cases were divided into luminal A (ER+/HER2− and Ki67 <14 %; n = 263), luminal B (ER+/HER2− and Ki67 ≥14 %; n = 161), HER2-positive (ER+/HER2+ or ER−/HER2+; n = 109), and triple-negative (TN, ER−/HER2−; n = 82). The distribution of epidemiological risk factors for controls and patients by breast cancer subtype is shown in Table 1. Multivariate analyses of these epidemiological factors indicated that one-unit increase in BMI was positively associated with an increased risk of breast cancer after adjustment for age, age at menarche, parity, and family history of breast cancer in all women (the multivariable adjusted odds ratio (aOR) 1.08, 95 % CI 1.04–1.12; P < 0.0001) (Fig. 1a). We further analyzed the association between BMI and breast cancer risk stratified by menopausal status (Fig. 1b, c). For premenopausal women, higher age at menarche, but not BMI was significantly associated with incidence of breast cancer (aOR 0.82, 95 % CI 0.71–0.95; P = 0.009) (Fig. 1b). On the other hand, for postmenopausal women, BMI (aOR 1.09, 95 % CI 1.04–1.15; P = 0.0008) and family history (aOR 0.49, 95 % CI 0.31–0.79; P = 0.003) were significantly associated with breast cancer risk (Fig. 1c).
Table 1

Characteristics of cases and control subjects by breast cancer subtype



Luminal A

Luminal B



All women

n = 682

n = 263

n = 161

n = 109

n = 82

BMI (kg/m2)a

21.67 (3.1)

23.2 (4.2)

22.72 (3.9)

22.41 (3.5)

21.47 (4.0)

Age (years)a

50.5 (11.3)

58.1 (12.6)

57.3 (14.2)

58.1 (12.6)

58.2 (13.7)

Age at menarche (years)a

13.1 (1.5)

13.3 (1.9)

13.1 (1.7)

13.2 (1.8)

13.5 (1.9)


 Yes (%)

548 (80.4)

200 (76.0)

127 (78.9)

84 (77.1)

63 (76.8)

 No (%)

134 (19.6)

63 (24.0)

34 (21.1)

25 (22.9)

19 (23.2)

Family history

 Yes (%)

148 (21.7)

34 (12.9)

34 (21.1)

16 (14.7)

13 (15.9)

 No (%)

534 (78.3)

229 (87.1)

127 (78.9)

93 (85.3)

69 (84.1)

Premenopausal women

n = 353

n = 100

n = 64

n = 36

n = 23

BMI (kg/m2)a

21.2 (3.0)

22.2 (4.8)

22.6 (4.1)

22.4 (3.7)

21.0 (3.4)

Age (years)a

41.2 (4.9)

45.7 (5.7)

43.2 (5.4)

44.9 (6.8)

42.1 (6.4)

Age at menarche (years)a

12.7 (1.3)

12.5 (1.3)

12.4 (1.3)

12.3 (1.3)

12.8 (2.1)


 Yes (%)

282 (79.9)

33 (61.1)

32 (69.6)

19 (73.1)

11 (65.7)

 No (%)

71 (20.1)

21 (38.9)

14 (30.4)

7 (26.9)

6 (35.3)

Family history

 Yes (%)

71 (20.1)

20 (20.0)

13 (20.3)

9 (25.0)

7 (30.4)

 No (%)

282 (79.9)

80 (80.0)

51 (79.7)

27 (75.0)

16 (69.6)

Postmenopausal women

n = 329

n = 163

n = 97

n = 73

n = 59

BMI (kg/m2)a

22.2 (3.1)

23.9 (3.6)

22.8 (3.9)

22.4 (3.4)

23.1 (4.0)

Age (years)a

60.5 (6.8)

65.7 (9.2)

66.6 (9.9)

64.7 (9.2)

64.5 (10.2)

Age at menarche (years)a

13.5 (1.6)

13.9 (2.0)

13.8 (1.8)

13.7 (1.3)

13.8 (2.1)


 Yes (%)

266 (80.9)

73 (80.2)

49 (79.0)

38 (76.0)

33 (86.8)

 No (%)

63 (19.1)

18 (19.8)

13 (21.0)

12 (24.0)

5 (13.2)

Family history

 Yes (%)

77 (23.4)

14 (8.6)

21 (21.6)

7 (9.6)

6 (10.2)

 No (%)

252 (76.6)

149 (91.4)

76 (78.4)

66 (90.4)

53 (89.8)

Age at menopause (years)a

50.2 (4.7)

50.1 (4.6)

49.2 (5.2)

50.1 (4.6)

49.6 (4.3)

aMean (standard deviation)
Fig. 1

Associations of body mass index (BMI) with breast cancer incidence by menopausal status. Odds ratios (closed solid circles) and 95 % confidence intervals (horizontal lines) are shown. BMI was calculated per one-unit increase. a All women, b premenopausal women, c postmenopausal women

Association between BMI and breast cancer risk by subtype

Since BMI was significantly positively associated with postmenopausal breast cancer risk, epidemiological risk factors, including BMI, were evaluated for postmenopausal women stratified by breast cancer subtype. BMI was significantly associated with a higher risk of luminal A breast cancer (aOR 1.18, 95 % CI 1.10–1.26; P < 0.0001) but not associated with an increased risk of other subtypes (Fig. 2).
Fig. 2

Associations of body mass index (BMI) with postmenopausal breast cancer incidence by breast cancer subtype. Odds ratios (closed solid circles) and 95 % confidence intervals (horizontal lines) are shown. BMI was calculated per one-unit increase. a Luminal A, b luminal B, c HER2-positive, d triple negative

For further evaluation of the significance of BMI for luminal breast cancer risk, controls were divided into three groups corresponding to the BMI levels [low, i.e., ≤20.00 kg/m2, n = 226 (33.1 %); intermediate, i.e., >20.0, ≤22.37 kg/m2, n = 228 (33.4 %); high, i.e., >22.37 kg/m2, n = 228 (33.4 %)] and the same criteria were used as for BMI of luminal A and luminal B breast cancer patients. The aORs and 95 % CIs were calculated after adjustment for epidemiological risk factors as shown in Table 2. As expected, women in the highest tertile of the BMI group showed a significantly higher risk of luminal A breast cancer than women in the lowest tertile of BMI (aOR 2.98, 95 % CI 1.53–5.80; P < 0.001). However, there was no significant association between BMI and breast cancer incidence for luminal B subtype (aOR 0.95, 95 % CI 0.47–1.91; P = 0.87).
Table 2

Relationship between body mass index (BMI) and risk of luminal A or luminal B breast cancer



Luminal A

OR (95 % CI)a

Luminal B

OR (95 % CI)a

Premenopausal women

 BMI tertiles (kg/m2)


147 (41.6 %)

34 (34.0 %)


21 (32.8 %)


  >20.00, ≤22.37

109 (30.9 %)

32 (32.0 %)

1.59 (0.84–3.02)

18 (28.1 %)

1.10 (0.54–2.25)


97 (27.5 %)

34 (34.0 %)

1.35 (0.70–2.59)

25 (39.1 %)

1.49 (0.76–2.93)

Postmenopausal women

 BMI tertiles (kg/m2)


79 (24.0 %)

23 (14.1 %)


24 (24.7 %)


  >20.00, ≤22.37

119 (36.2 %)

35 (21.5 %)

1.22 (0.59–2.52)

26 (26.8 %)

0.47 (0.21–1.05)


131 (39.8 %)

105 (64.4 %)

2.98 (1.53–5.80)b

47 (48.5 %)

0.95 (0.47–1.91)

aOdds ratio and 95 % confidence interval adjusted for age, age at menarche, parity, age at menopause, and family history

bP < 0.005


The observations of the study presented here showed that BMI is significantly associated with breast cancer incidence of luminal A, but not of luminal B in postmenopausal breast cancers. As far as we know, ours is the first study to analyze breast cancer incidence stratified by luminal A and luminal B subtypes. It has been speculated that luminal A breast cancers are highly dependent on estrogen, whereas both estrogen and growth factor signaling are involved in the development of luminal B breast cancers [19]. This hypothesis seems to be supported, at least in part, by the finding that proliferative marker Ki67 expression levels after aromatase inhibitor treatment were significantly higher in luminal B than in luminal A cancers [20]. Since the response to endocrine therapy can be evaluated in terms of reductions in Ki67 expression levels, the efficacy of this treatment seems to be partially diminished in luminal B cancers by cross talk with growth factor signaling. On the basis of observations of ER and PR expression levels in luminal A and luminal B cancers while taking menopausal status into consideration, Murase et al. [21] concluded that the biological characteristics of luminal A cancers are strongly affected by the estrogen environment, but that its influence on luminal B cancers may be limited. These considerations may support the notion that a richer estrogenic environment leads to an increase in the incidence of luminal A cancers rather than of luminal B cancers. It is thus conceivable that BMI may affect luminal A breast cancer susceptibility mediated through estrogens among postmenopausal women.

Ritte et al. [22] conducted a prospective cohort study and reported that BMI was positively associated with ER+/PR+ breast cancer risk for women ≥65 years [hazard ratio (HR) 1.25, 95 % CI 1.16–1.34], but not for those with ER+/PR− (HR 0.92, 95 % CI 0.78–1.08). Suzuki et al. [23] found that increases in BMI (increments of 5 kg/m2) from the age of 20 were significantly positively associated with the development of ER+/PR+ cancers (relative risk (RR) 2.24, 95 % CI 1.50–3.34), but not of ER+/PR− cancers (RR 0.63, 95 % CI 0.31–1.27). In addition, Yang et al. analyzed pooled samples and found that among older women (>50 years) the frequency of highest BMI (≥30 kg/m2) compared with lowest (<25 kg/m2) was significantly lower for patients with ER+/PR− than for those with ER+/PR+ cancers (OR 0.70, 95 % CI 0.62–0.79) [24]. Considering the fact that PRs can be induced by estrogen signaling and are likely to be associated with tumors classified as luminal A [21], the above findings seem to support our data. However, we also believe that there is a need for epidemiological analyses of estrogen-related factors for breast cancer incidence stratified by luminal subtypes rather by PR status.

Furthermore, the involvement of obesity-associated factors other than estrogens such as cholesterols, triglyceride, glucose, insulin, c-peptide, adiponectin, leptin, and insulin-like growth factors in susceptibility for breast cancers needs to be investigated. Gunter et al. [16] reported that obesity (BMI ≥ 30 kg/m2) is positively associated with the risk of postmenopausal breast cancer (HR for BMI ≥30 kg/m2 vs 18.5 to <25 kg/m2 = 2.12, 95 % CI 1.26–3.58; P = 0.003). However, this significance was diminished by adjustment for insulin (P = 0.40), which seems to suggest that serum insulin levels enhance postmenopausal breast cancer risk more than BMI. Since these factors seem to be associated with breast cancer risk irrespective of luminal subtype, the positive association between higher BMI and increased risk of postmenopausal luminal A breast cancers may be induced by higher estrogen levels, rather than by other serum factors associated with obesity including insulin.

Controls in this study showed a significantly higher frequency of family history than did the cases. Since controls were recruited not from the general population, but from participants in a breast cancer screening program, we cannot deny the possibility that there was a tendency for women who worried about breast cancer because of their family history to participate. Cases were recruited from Hyogo College of Medicine (422 patients) and Tokushima Breast Clinic (193 patients). Ideally, the same percentage of controls should have been recruited from the Tokushima area, but since we observed the same results when cases were restricted to patients treated at Hyogo College of Medicine, we believe the difference in institute probably had no effect on these results. In addition, the controls were significantly younger than the cases. However, since no association was observed between age and BMI in either premenopausal or postmenopausal women in the present study (data not shown), the difference in age distribution may not have influenced the results. One other limitation of this study is that the conclusion was based on a retrospective study. Ideally, serum estrogen levels and obesity-associated factors also need to be assessed, but this was not feasible in this study because there were not enough blood samples.

In conclusion, we have demonstrated that higher BMI is significantly associated with an increased risk of postmenopausal luminal A, but not of luminal B breast cancers. Prospective studies including determination of serum estrogen and obesity-associated factors and with a larger number of cases are needed to validate our results.

Conflict of interest

Y. Miyoshi has received honoraria from Sanofi, AstraZeneca K.K., and GlaxoSmithKline K.K. The other authors declare that they have no conflict of interest.

Copyright information

© The Japanese Breast Cancer Society 2013