Osteoporosis International

, Volume 15, Issue 7, pp 535–540

Radial bone density and breast cancer risk in white and African-American women

Authors

    • Department of Internal MedicineWayne State University
    • Karmanos Cancer Institute
    • Rheumatology, UHC 4-HWayne State University
  • L. L. Darga
    • Department of Internal MedicineWayne State University
    • Karmanos Cancer Institute
  • M. S. Simon
    • Department of Internal MedicineWayne State University
    • Karmanos Cancer Institute
  • R. K. Severson
    • Department of Family MedicineWayne State University
    • Karmanos Cancer Institute
Original Article

DOI: 10.1007/s00198-003-1576-z

Cite this article as:
Nelson, D.A., Darga, L.L., Simon, M.S. et al. Osteoporos Int (2004) 15: 535. doi:10.1007/s00198-003-1576-z

Abstract

A number of different models for assessing individual risk of breast cancer use known risk factors such as age, age at menarche, age at first live birth, previous breast biopsies, and family history. High bone mass in white women is also associated with an increased breast cancer risk; however, bone mass as a risk factor has not been studied in African-American women. We conducted a case-control study to evaluate bone mineral density as a risk factor for breast cancer in white and African-American women. We recruited 221 women with newly diagnosed breast cancer from a comprehensive breast cancer center at a large university hospital, and 197 control women who were frequency matched for ethnicity and age. Odds ratios were based on proximal and distal radial bone density measured by peripheral bone densitometry (Norland pDEXA) and expressed as a standardized “Z-score” (age and ethnicity specific). Logistic regression models were fitted controlling for body mass index, menopausal status, age, and HRT use (ever/never and duration). With proximal bone density Z-score included in the model as a continuous variable, a one-unit increase in radial shaft bone density increased the risk of breast cancer by 25% (p=0.02). When proximal bone density Z-score was analyzed as a dichotomous variable (≤0, >0) the odds ratio was 1.98 (95% CI, 1.32 to 2.97); that is, having an above average proximal bone density (age-specific) doubles the risk of breast cancer. There were no significant interactions with, and no appreciable confounding effects by, other covariates. An above-average radial shaft Z-score is a significant risk factor for breast cancer in both white and African-American women. The present study extends the association between bone mass and breast cancer risk to African-Americans, and suggests another potential application for bone density testing.

Keywords

Bone densityBreast cancer riskEthnicity

Introduction

After nonmelanomatous skin cancer, breast cancer has the highest age-adjusted incidence rate among women in the United States. Because it is a critical health issue for most women, a number of different models have been developed to assess risk at the individual level. These models use algorithms based on known risk factors such as age, age at menarche, age at first live birth, number of previous breast biopsies, and number of first-degree relatives with breast cancer [1, 2, 3]. The Gail model, for example, employs ethnicity, current age, age at menarche, age at first live birth, first-degree relatives with breast cancer, number of previous biopsies, and presence or absence of atypical hyperplasia [3].

In addition to these well-known risk factors, there have been several reports that white women with a high bone mass are at a significantly increased risk for breast cancer [4, 5, 6, 7, 8]. In 8,905 white women in the Study of Osteoporotic Fractures (SOF), Zmuda et al. [4] reported that the risk of breast cancer in the highest quartile of bone mineral density (BMD) was 2.7 (95% CI, 1.4 to 5.3), relative to the lowest quartile. BMD was measured in the wrist, forearm, and heel by single photon absorptiometry (SPA). In the Framingham Study, the relative risk was 3.5 (95% CI, 1.8 to 6.8) for the highest quartile of metacarpal cortical area, measured by radiogrammetry [5]. Buist et al. reported an association between higher hip bone mass and increased breast cancer risk in a multicenter cohort study of 8,203 postmenopausal women who participated in the Fracture Intervention Trial (FIT) [6, 9]. In this study, BMD was measured by dual-energy X-ray absorptiometry (DXA) in the total hip region of the proximal femur. The relative risk for the highest quartile was 1.8 (95% CI, 1.0 to 3.1) [6]. In the Rotterdam Cohort Study [7] of 3,107 women, the hazard ratio for breast cancer in the highest tertile of BMD was 2.1 (1.1–3.7) for the spine. BMD measurements based on the femoral neck were not significant. In the Dubbo Osteoporosis Epidemiology Study (a case-control study in Australia), each 0.1 g/cm2 increase in BMD (as measured in the lumbar spine and femoral neck) was associated with odds ratios of 2.1 (95% CI, 1.3 to 3.4) and 1.5 (95% CI, 1.0 to 2.4) for breast cancer, respectively [8]. Differences among these studies might be explained by different bone mass measurement methods, different skeletal regions measured, and/or by differences in the characteristics of the samples. The SOF, Framingham, Rotterdam, and Dubbo studies comprised white women only, while the study by Buist comprised 97% white women [10].

There are known differences in the incidence and outcome of breast cancer in white and African-American women [11, 12], along with well-documented ethnic differences in bone density [13, 14, 15]. Therefore, we undertook a case-control study that included both white and African-American women with newly diagnosed breast cancer to evaluate the association between bone mass and breast cancer risk in each ethnic group. We chose to use peripheral densitometry to measure the radius because of the wider availability of this method compared with the more expensive instruments.

DXA is the state-of-the-art method for measuring bone density [16], but central DXA instruments are large and expensive. In contrast, peripheral DXA (pDXA) machines are less expensive, portable, and easier to operate than central DXAs. These features have allowed a wider deployment of bone densitometry to communities and smaller medical practices that do not have access to full-size DXA machines. Two types of pDXA instruments are currently being used in the National Osteoporosis Risk Assessment study (>180,000 healthy postmenopausal women) of peripheral bone density and fracture risk. This study has reported an excellent fracture risk prediction capability of these instruments based on 1 year of follow-up [17]. Because of the availability and reliability of pDXA, we chose this modality for our study.

Materials and methods

Subjects were eligible for inclusion in the study if they were aged 40–85, had no previous history of cancer and had never previously taken any medication that affects bone density (steroids, bisphosphonates) for more than 1 month. However, past or present hormone replacement therapy (HRT) was not an exclusion criterion. Patients were recruited over a 2.5-year period from a diagnostic and surgical comprehensive breast center that serves the Detroit (USA) metropolitan area and were included if they had received an initial diagnosis of stages 0–III breast cancer, confirmed by pathologic assessment of breast tissue, within the last 6 months; and if they had not yet started chemotherapy or were ≤1 month post–chemotherapy initiation. Controls were recruited concurrently and included outpatients who were attending the clinic for other reasons and who did not have breast cancer; controls also included employees at the institution and other women from the Detroit metropolitan area who were free of any cancer diagnosis. Cases and controls were frequency matched for ethnicity, as determined by self-report, and age by decade. This study was approved by the institutional review board at the study institution, and written informed consent was obtained from each subject.

We approached 471 women to participate in the study. The number of refusals was evenly distributed between white and African-American women (8.3% versus 10.1%, respectively). There were more refusals in the case group than the control group (16.5% versus 2%, respectively).

Out of 120 white cases, 119 were clinic patients and 1 responded to an advertisement about the study that was placed in community outreach centers associated with the cancer center. One of 111 African-American cases also responded to the advertisement, and the rest were clinic patients. All controls were free of a cancer diagnosis. Out of 96 white controls, 29 had presented for mammography and had negative results; 16 accompanied patients or were friends of either patients or employees; 45 were employees of the institution; and 6 responded to the advertisement. Out of 102 African-American controls, 70 had negative mammograms, 10 accompanied patients or were friends of either patients or employees; 21 were employees, and 1 responded to the advertisement.

Women were interviewed in person by one of two trained study staff regarding information on demographics, gynecologic and reproductive history, and family history of breast cancer. A standardized questionnaire was used for all interviews, and data were recorded directly onto each form by the interviewer. Radial BMD (in g/cm2) was measured in the radial shaft and the distal radius with a Norland pDEXA bone densitometer according to the manufacturer’s recommendations. The accuracy of DXA measurements of the forearm, expressed as a correlation coefficient for the measured bone mineral content and the ash content, is r>0.97 [18]. The precision of BMD measurements is also excellent, with errors (expressed as coefficients of variation) ranging from 0.7% to 1.3% [19, 20, 21]. Because of the wide age range and the inclusion of two ethnic groups in our study, we utilized standardized scores (Z-scores) that express BMD in standard deviation units based on age-, sex-, and ethnicity-specific reference data. These scores were based on comparison of individual patient data to reference data provided by the manufacturer.

Unconditional logistic regression was used to evaluate the risk of breast cancer in relation to BMD while controlling for the potential confounding effects of other covariates. Interaction terms were employed to assess whether other covariates modified the effect of BMD on the subsequent development of breast cancer.

Results

Demographic and risk factor data for cases and controls appear in Table 1. There was a similar distribution of African-Americans and whites in cases (48.8% vs 51.9%, respectively) and controls (51.5% vs 48.5%, respectively). The distribution by age groups was similar in cases and controls, although cases appeared to be slightly older than controls. While there was no difference in cases and controls by weight, it appears that cases tended to have a smaller stature, higher body mass index, lower age at menarche, higher age at menopause, and lower age at first pregnancy. Thirty-eight percent of the cases and 35% of the controls reported being premenopausal. A total of 38% of the cases and 47% of the controls reported ever using HRT.
Table 1

Distribution of selected characteristics among cases and controls

Variable

Cases

Controls

Number

%

Number

%

Total

231

100.0

198

100.0

Ethnicity

  White

120

51.9

96

48.5

  African-American

111

48.8

102

51.5

Age

  <49

66

28.6

68

34.3

  50–59

73

31.6

67

33.8

  60–69

56

24.2

39

19.7

  70–79

29

12.6

22

11.1

  80+

7

3.0

2

1.0

Weight (kg)

  30–59

35

15.2

33

16.7

  60–69

50

21.6

50

25.3

  70–79

57

24.7

50

25.3

  80–99

61

26.4

45

22.7

  100+

28

12.1

20

10.1

Height (cm)

  130.0–154.9

32

13.9

19

9.6

  155.0–159.9

33

14.3

24

12.1

  160.0–164.9

73

31.6

52

26.3

  165.0–169.9

51

22.1

51

25.8

  170.0+

42

18.2

52

26.3

Body mass index

  <18.0

3

1.3

3

1.5

  18.0–24.9

68

29.4

72

36.4

  25.0–29.9

68

29.4

65

32.8

  30.0+

92

39.8

58

29.3

Age at menarche

  <12

65

28.1

43

21.8

  12

55

23.8

56

28.4

  13

56

24.2

47

23.9

  14+

55

23.8

51

25.9

  Unknown

0

1

Age at menopause

  <45

26

18.1

31

24.0

  45–49

40

27.8

34

26.4

  50–54

56

38.9

56

43.4

  55+

22

15.3

8

6.2

  Premenopausal

87

69

Age at first pregnancy

  <20

67

34.2

39

23.9

  20–24

71

36.2

70

42.9

  25–29

37

18.9

34

20.9

  30+

21

10.7

20

12.3

  Unknown

5

2

  Nulliparous

30

33

No. of full term pregnancies

  0

32

14.2

35

17.8

  1

36

15.9

28

14.2

  2

65

28.8

56

28.4

  3

46

20.4

34

17.3

  4+

47

20.8

44

22.3

  Unknown

5

1

Hormone replacement therapy

  Unknown

44

35

  No

116

62.0

86

52.8

  Yes

71

38.0

77

47.2

HRT duration (years)

  <1

17

25.4

21

28.0

  1–4.9

22

32.8

29

38.7

  5.0+

28

41.8

25

33.3

  Unknown

48

37

Bone density data for cases and controls are shown in Table 2. As expected, older subjects tended to have lower bone density values, and younger subjects had higher bone density values. When comparing the distributions in cases and controls, the cases tended to have greater representation in the highest quartiles of BMD in the proximal radius. The mean proximal radial Z-score was significantly higher in cases than controls (0.45 ± 1.07 vs 0.19 ± 1.08 g/cm2, p=0.015), but there was no significant difference in the distal measurement (p=0.832).
Table 2

Quartiles of bone mineral density (BMD) by age groups

Age

40–49

50–59

60–69

70+

N

%

N

%

N

%

N

%

Distal radius BMD (g/cm2)

  Controls

    0–0.28

1

1.5

8

12.1

11

28.2

16

66.7

    0.29–0.33

20

29.4

25

37.9

16

41.0

2

8.3

    0.34–0.37

22

32.4

19

28.8

7

17.9

4

16.7

    0.38+

25

36.8

14

21.2

5

12.8

2

8.3

  Cases

    0–0.28

4

6.3

13

18.3

23

43.4

20

58.8

    0.29–0.33

14

22.2

15

21.1

17

32.1

8

23.5

    0.34–0.37

21

33.3

17

23.9

7

13.2

5

14.7

    0.38+

24

38.1

26

36.6

6

11.3

1

2.9

Proximal radius BMD (g/cm2)

  Controls

    0–0.72

6

8.8

6

9.1

16

41.0

16

66.7

    0.73–0.80

18

26.5

23

34.8

10

25.6

5

20.8

    0.81–0.87

25

36.8

19

28.8

9

23.1

3

12.5

    0.88+

19

27.9

18

27.3

4

10.3

0

0.0

  Cases

    0–0.72

2

3.2

10

14.1

25

47.2

19

55.9

    0.73–0.80

12

19.0

17

23.9

15

28.3

7

20.6

    0.81–0.87

25

39.7

20

28.2

5

9.4

5

14.7

    0.88+

24

38.1

24

33.8

8

15.1

3

8.8

The univariate odds ratios for breast cancer are shown in Table 3. The risk of breast cancer increased significantly with increasing age and decreasing height. There were no significant trends in any of the other variables, although there was some weak evidence of an increased risk with increasing body mass index, increasing age at menopause, and decreasing age at first pregnancy. When the odds ratios are adjusted for age and ethnicity, three results change: the increased risk with increasing body mass index becomes statistically significant (p=0.019), as does the OR for HRT use. The relationship with age at first pregnancy is marginally significant at p=0.05 (Table 3). With proximal bone density Z-score included as a continuous variable in a logistic regression model, a one-unit increase in bone density in the radial shaft increased the risk of breast cancer by 25% (p=0.02, see Table 4). With proximal bone density Z-score included as a dichotomous variable (≤0, >0) in a logistic regression model, the odds ratio for breast cancer was twofold for higher than average bone density: OR=1.98 (95% CI, 1.32 to 2.97). When the ethnic groups were evaluated separately, the odds ratios for the dichotomized Z-score were similar and statistically significant for both groups (OR=2.02 for whites and 1.95 for African-Americans, see Table 4). No significant interactions were found by menopause status, age group, body mass index, ever/never use of HRT, or duration of HRT on the relationship between BMD and breast cancer risk. In addition, no appreciable confounding effects by these variables were observed.
Table 3

Univariate regression odds ratios for breast cancer, adjusted and unadjusted

Variable

Univariate regression ORs

Controlling for age & ethnicity

OR

95% CI

Trend p

OR

95% CI

Trend p

Ethnicity

  White

1.00

  African-American

0.87

0.60–1.27

Age

  <49

1.00

  50–59

1.12

0.70–1.80

  60–69

1.48

0.87–2.52

  70–79

1.36

0.71–2.60

  80+

3.61

0.72–18.0

0.023

Weight (kg)

  30–59

1.00

1.00

  60–69

0.94

0.51–1.75

0.95

0.51–1.78

  70–79

1.08

0.59–1.98

1.10

0.59–2.06

  80–99

1.28

0.69–2.36

1.36

0.72–2.56

  100+

1.32

0.63–2.78

0.390

1.71

0.78–3.74

0.138

Height (cm)

  130.0–154.9

1.00

1.00

  155.0–159.9

0.82

0.38–1.77

0.78

0.36–1.70

  160.0–164.9

0.83

0.43–1.63

0.83

0.42–1.63

  165.0–169.9

0.59

0.30–1.18

0.61

0.31–1.22

  170.0+

0.48

0.24–0.96

0.021

0.51

0.25–1.02

0.040

Body mass index

  <18.0

1.00

1.00

  18.0–24.9

0.94

0.18–4.84

0.97

0.18–5.14

  25.0–29.9

1.05

0.20–5.37

1.08

0.20–5.74

  30.0+

1.59

0.31–8.13

0.072

1.81

0.34–9.59

0.019

Age at menarche

  <12

1.00

1.00

  12

0.65

0.38–1.11

0.65

0.38–1.12

  13

0.79

0.46–1.36

0.81

0.46–1.40

  14+

0.71

0.42–1.23

0.389

0.70

0.41–1.22

0.348

Age at menopause

  <45

1.00

1.00

  45–49

1.40

0.70–2.81

1.40

0.69–2.84

  50–54

1.19

0.63–2.26

1.10

0.57–2.12

  55+

3.28

1.25–8.59

0.072

2.70

1.00–7.26

0.224

Age at first pregnancy

  <20

1.00

1.00

  20–24

0.59

0.35–0.99

0.49

0.28–0.86

  25–29

0.63

0.34–1.17

0.54

0.27–1.06

  30+

0.61

0.30–1.27

0.072

0.54

0.25–1.17

0.050

No. full term pregnancies

  0

1.00

1.00

  1

1.41

0.71–2.80

1.56

0.78–3.14

  2

1.27

0.79–2.31

1.42

0.77–2.62

  3

1.48

0.77–2.84

1.54

0.80–2.99

  4+

1.17

0.62–2.20

0.957

1.10

0.58–2.09

0.648

HRT

  No

1.00

1.00

  Yes

0.70

0.47–1.04

0.57

0.37–0.87

Table 4

Odds ratios for proximal radius Z-scores and breast cancer risk, for all subjects and for ethnic groups separately

All subjects

Whites

African-Americans

OR

CI

OR

CI

OR

CI

Z-score dichotomized

  <0

1.00

1.00

1.00

  >0

1.98

1.32–2.98

2.02

1.13–3.60

1.95

1.10–3.46

Z-score as continuous variable

 

1.25

1.04–1.51

1.18

0.90–1.54

1.32

1.03–1.70

Discussion

Our study may be the first one designed specifically to investigate bone density and breast cancer risk in both white and African-American women. Previously reported studies of bone density and breast cancer risk utilized data from large epidemiologic studies of white women [4, 5, 7, 8], and in one case, from a clinical trial comprising 97% white women [6]. These studies employed a variety of methods of bone mass measurement. In the SOF study, BMD was measured by SPA in the forearm, wrist, and heel [4]. In the Framingham study, radiogrammetry of hand radiographs was used to assess metacarpal bone mass[5]. In contrast with these earlier studies, Buist et al. measured the proximal femur with DXA in a cohort of women who were screened for a fracture intervention trial [6]. During 3.7 years of follow-up, 131 incident breast cancer cases occurred. The risk of breast cancer, by increasing quartile of hip bone density, was 1.0 (reference), 1.9, 1.5, and 1.5, but the risk estimates for the two highest quartiles were not statistically significant. The Rotterdam and Dubbo studies utilized DXA of the lumbar spine and femoral neck, but found significant associations with breast cancer risk for the spine only. Thus it appears from these studies that peripheral skeletal sites and the lumbar spine are better predictors of breast cancer risk than the hip. It is possible that the hip, as a major load-bearing site, may be more affected by lifestyle differences among groups or individuals. Another possible reason for differences among the studies is that there are differences in the populations studied, particularly in the inclusion criteria. For example, in FIT (used for the Buist study, [6]), women were excluded if they had used estrogen or calcitonin in the previous 6 months or other osteoporosis drugs at any time in the past [10]. In contrast, the epidemiologic studies did not exclude these medications, but there may be significant regional or cultural differences in HRT use. Age, body size, and lifestyle factors that affect both bone mass and breast cancer risk also varied among the study populations. One limitation of our study is that we may not have had a sufficient number of subjects in each of the individual strata to completely evaluate effect modification.

We found that an above average radial shaft Z-score appears to be an important risk factor for breast cancer in white and African-American women. This finding is consistent with the previous reports and can now be extended to include African-Americans. Peripheral DXA is inexpensive, noninvasive, and portable, and can therefore be readily utilized in a clinical setting. We suggest that, along with other risk factors, it may be an important additional tool in the evaluation of the risk of breast cancer in both African-American and white women. Our findings contribute to our understanding of breast cancer etiology, but further investigations are needed to elucidate the assumed role of endogenous and exogenous estrogens in mediating the bone mass / breast cancer risk association.

Acknowledgements

This research was supported by grant DAMD 17-98-1-8354 from the United States Department of Defense. We also acknowledge the Barbara Ann Karmanos Cancer Institute for additional support of this project.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2004