Breast Cancer Research and Treatment

, Volume 105, Issue 2, pp 195–207

Breast cancer risk factors and second primary malignancies among women with breast cancer

Authors

    • Paul P. Carbone Comprehensive Cancer CenterUniversity of Wisconsin
    • Department of Population Health Sciences, School of Medicine and Public HealthUniversity of Wisconsin
  • Polly A. Newcomb
    • Paul P. Carbone Comprehensive Cancer CenterUniversity of Wisconsin
    • Cancer Prevention ProgramFred Hutchinson Cancer Research Center
  • Hazel B. Nichols
    • Paul P. Carbone Comprehensive Cancer CenterUniversity of Wisconsin
  • John M. Hampton
    • Paul P. Carbone Comprehensive Cancer CenterUniversity of Wisconsin
Epidemiology

DOI: 10.1007/s10549-006-9446-y

Cite this article as:
Trentham-Dietz, A., Newcomb, P.A., Nichols, H.B. et al. Breast Cancer Res Treat (2007) 105: 195. doi:10.1007/s10549-006-9446-y

Abstract

Purpose

To examine the association between breast cancer risk factors and second primary cancers (independent diagnoses occurring at least 12 months after the initial breast cancer diagnosis) among breast cancer survivors.

Methods

In this population-based study, cancer outcomes among breast cancer survivors first diagnosed during 1987–2000 were investigated. Invasive breast cancer cases were identified from the statewide tumor registry and interviewed regarding their pre-diagnosis risk factors, including reproductive and lifestyle characteristics, approximately 1 year after diagnosis. Data on second primary cancers (not recurrences) and deaths were obtained by linkage with tumor registry reports and death certificates through December 31, 2002. Hazard ratios (HR) were estimated using proportional hazards regression stratified by age and adjusted for stage and other factors.

Results

Among the 10,953 breast cancer cases, 10.8% experienced a second cancer diagnosis within an average of 7 years (including 488 breast, 132 colorectal, 113 endometrial, and 36 ovarian cancers). Risk of a second primary breast cancer increased according to low parity (P = 0.002), older age at menopause (P = 0.08), greater body mass index (P = 0.003) and adult weight gain (P = 0.02), and a family history of breast cancer-particularly among women with 2 or more first-degree affected relatives (HR = 1.8, 95% CI: 1.1–2.9). Reduced risk of colorectal cancer after breast cancer was observed in relation to older ages at menarche (P = 0.05), younger age at menopause (P = 0.04), postmenopausal hormone use (HR = 0.4, 95% CI: 0.3–0.7), normal body mass index (P = 0.07), and infrequent alcohol consumption (P = 0.01). Second endometrial cancer risk was associated with increasing body mass index (P < 0.01) and adult weight gain (P = 0.03). Risk of second ovarian cancer appeared related to recent alcohol intake and family history of breast cancer. Women who reported consuming any alcohol appeared to have a 55% reduction in ovarian cancer risk (95% CI: 0.2–1.0) compared to non-drinkers, while having 2 or more first-degree relatives with breast cancer was associated with an increased risk of ovarian cancer (HR = 4.3, 95% CI: 1.3–14.6).

Conclusion

This study suggests that family history of breast cancer as well as potentially modifiable characteristics including body weight, alcohol intake, and postmenopausal hormone use may be associated with risk of a second cancer diagnosis among breast cancer cases.

Keywords

NeoplasmsSecond primaryBreast neoplasmsBody mass indexBody weightWeight gainBody heightAlcoholic beverages

Introduction

The diagnosis of more than one primary tumor in a woman’s lifetime is still a relatively rare event, although it is becoming more common as the number of cancer survivors in the US population increases [1]. Women with a personal history of cancer have a 2- to 6-fold increased risk of (additional) cancer compared with women without a personal history of malignancy [25]. Suspected reasons for this increased risk include shared risk factors and consequences of treatment.

Several studies have reported a potential increase in risk of a second primary (an independent diagnosis occurring at least 12 months after the initial breast cancer diagnosis) breast, colorectal, endometrial and/or ovarian cancer among women diagnosed with breast cancer [511]. Many previous investigations have utilized large tumor registries, often with available treatment information [1]. Indeed, much research among breast cancer survivors has evaluated second cancers as a consequence of treatment for the first diagnosis. Studies have focused on concerns regarding colorectal [12] and endometrial cancer [13] after tamoxifen use, lung cancer after radiation therapy, and leukemia after chemotherapy [14]. Investigations have also evaluated suspected shared etiologic factors between the first and second cancers, such as with contralateral breast cancer due to genetic predisposition [3, 15].

Women who develop a second primary cancer after breast cancer may have increased mortality [1, 9, 10]; identifying factors that put women at greatest risk may impact discussions of behavior modifications and/or surveillance to reduce risk of a second cancer. While large population registries or clinic databases may be well-suited for monitoring late effects of treatment, population-based epidemiologic studies that involve interviews with breast cancer survivors are better positioned to evaluate reproductive and lifestyle factors in relation to risk of second cancers.

Few previous investigations have described risks or benefits associated with breast cancer risk factors including parity, postmenopausal hormone use and/or weight gain, or alcohol consumption in relation to breast, colorectal, endometrial or ovarian cancer after a breast cancer diagnosis [2, 3]. Many risk factors are shared between these cancer sites because of the central role of hormones, in particular estrogen, in the etiology of cancer in women. The bi-directional nature of the increased risk of a second additional cancer (women with breast cancer are at increased risk of colorectal, endometrial, and/or ovarian cancer and vice-versa) lends support to the hypothesis that changes in risk are due to shared risk factors, rather than treatment [5]. Here we describe an investigation of pre-diagnosis risk factors among breast cancer survivors in relation to risk of a second asynchronous cancer diagnosis.

Methods

Identification of breast cancer cases

Subjects for the present analysis consisted of the case series from three consecutive population-based case–control studies of breast cancer [16, 17]. Eligible cases were female residents of Wisconsin aged 18–79 years diagnosed with an invasive breast cancer (excluding carcinoma in situ) and reported to Wisconsin’s statewide tumor registry during 1987–2000. All primary malignant neoplasms are reported by statutory mandate to the statewide tumor registry except basal cell and squamous cell carcinoma of the skin.

Study protocols were performed with approval by the University of Wisconsin Health Sciences Institutional Review Board.

After identification by the registry, the physician of record for each eligible subject was contacted by mail and provided an opportunity to actively refuse participation on the case patient’s behalf. Eligibility was limited to women with listed telephone numbers and drivers’ licenses by self-report (if less than 65 years of age). We identified 13,213 eligible invasive breast cancer cases. Of these, physicians refused contact for 507 (3.8%), 615 (4.7%) were deceased, 150 (1.1%) could not be located, and 933 (7.1%) refused to participate. Thus, interview data for 11,008 women (83.3%) were collected. Data for 7 subjects were considered unreliable according to the interviewers, and 37 women were excluded from analysis due to missing information regarding potential second primary cancer diagnoses. Finally, 11 women died within 1 year after the initial breast cancer diagnosis. After these exclusions, data for 10,953 women were available for inclusion in the analysis.

Case interviews

Cases were interviewed after providing informed consent by telephone from September 1988 through May 2001 approximately one year after diagnosis. The 25–45 min interview (depending on the time period of the study) elicited information on reproductive experiences, physical activity, alcohol and tobacco use, exogenous hormone use, height and weight, medical history, and demographic factors. Exposure information was truncated at the diagnosis date. Questionnaire items utilized in this analysis remained similar in each of the case–control studies; variation in the average interview duration was based, for the most part, on a greater or lesser number of additional items (e.g. lactation history, physical activity). Some slight variations in risk factor definitions existed between years of data collection; for example, exposures termed “recent” (e.g. weight, alcohol consumption, etc.) occurred 2–5 years before diagnosis depending on the questionnaire utilized in the original case–control study. Similarly, “early” exposures were self-reported for ages 18–20 depending on era of data collection.

Cancer and vital status outcomes

Data on first and subsequent tumor diagnoses and cancer-directed treatment were obtained by linkage with tumor registry reports. As specified by Wisconsin tumor registry rules to determine whether more than one independent primary tumor existed [18], only second primary cancer diagnoses that occurred at least twelve months (>365 days) after the initial breast cancer diagnosis were included in this analysis. Since laterality was not collected by the tumor registry until 1996, ipsilateral breast cancer diagnoses were not considered to be second primaries unless the histologic types of the initial and subsequent diagnoses were different and diagnosed at least twelve months apart. Cancer treatment information included planned first-course treatment with surgery, chemotherapy, radiotherapy, and/or hormone therapy. Other data for each diagnosis included date of diagnosis, extent of disease (in situ, localized, regional or distant spread), and cell type and site codes classified according to the International Classification of Diseases for Oncology, Third Edition [19]. Vital status for cases was obtained through linkage with Wisconsin death certificates and the National Death Index through December 31, 2002.

Exposure definitions and statistical analysis

Menopausal status was defined as postmenopausal if the woman reported a natural menopause or a bilateral oophorectomy prior to the initial breast cancer diagnosis date. Women reporting hysterectomy alone were classified as postmenopausal if the age at diagnosis was greater than or equal to 54 years for smokers and 55 for nonsmokers (the 90th percentiles of age at natural menopause for the control group in the original case–control studies), and premenopausal if the age at diagnosis was less than 42 years (the 10th percentile of age at natural menopause for the control group in the original case–control studies). Menopausal status was considered to be unknown for women with hysterectomy without bilateral oophorectomy if the age at diagnosis was between 42 and 54 years (or 55 years for nonsmokers).

Postmenopausal hormone use was defined as the use of oral or transdermal noncontraceptive hormones, including estrogens and/or progestins, for three cumulative months or more. A woman was defined as a recent user if she reported use within approximately 2 years of diagnosis. Women were considered former or current smokers if they had smoked at least 100 cigarettes in their lifetimes; current smokers were smoking cigarettes at the time of the initial breast cancer diagnosis. Pack-years of cigarette smoking were defined as the number of years smoked multiplied by the average number of packs of cigarettes per day.

Body mass index was defined as recent weight (kg) divided by maximum adult height (m) squared. Adult weight gain was defined as the difference in body weight between early life (ages 18–20) and recent weight. Total recent alcohol intake was defined as the average total number of drinks of beer, wine, and hard liquor consumed. Over 95% of the cases identified themselves as white (N = 10,580), as such, analysis of other racial or ethnic groups was not feasible due to small numbers.

Cumulative incidence of a second cancer diagnosis was estimated to correctly account for death from other causes as a competing risk [20]. Incidence calculations for the tables and figures exclude the first 12 months (≤365 days) after the initial breast cancer diagnoses.

Hazard ratios (HR) were estimated using proportional hazards regression stratified by age at the initial breast cancer diagnosis and adjusted for year of diagnosis, initial cancer extent of disease, family history of breast cancer, pack-years of cigarette smoking, recent alcohol intake, parity, postmenopausal hormone therapy, menopausal status, and body mass index. Preliminary models additionally adjusted for planned first course treatment (radiation, chemotherapy, or hormone therapy). However, as effect estimates remained unchanged, we did not retain treatment variables in final models. Hazard ratios for weight gain were adjusted for height and weight at age 18–20 instead of body mass index, whereas hazard ratios for height were adjusted for weight residuals (calculated as the difference between weight and the predicted weight after regression on height) instead of body mass index. A trend variable was computed for risk factors (including body weight, weight gain, height, and alcohol intake) using these factors parameterized as continuous linear variables in separate regression models. Follow-up started at 1 year after the initial diagnosis until a subsequent cancer diagnosis, censoring at death, or December 31, 2002, whichever occurred first. Women reporting a hysterectomy and/or bilateral oophorectomy prior to her initial breast cancer diagnosis were not included in analyses evaluating a second primary diagnosis of uterine/endometrial or ovarian cancer.

Reliability

To assess the reliability of the questionnaire, a sequential sample of cases from the second case–control study was reinterviewed during March–July 1995 [21]. Intraclass correlation coefficients (ICC) and 95% lower confidence limits (LCL) were estimated to evaluate the reproducibility of questionnaire items [22]. We used Cohen’s κ with 95% LCLs to measure the reliability of the subjects’ responses regarding variables coded never/ever (e.g. family history of breast cancer, postmenopausal hormone use) [22]. After an average of 3 months (range, 2–6 months), 186 cases (71%) were successfully recontacted and reinterviewed. Among women interviewed a second time, reproducibility was high: recent body mass index (ICC = 0.92; LCL = 0.90), total recent alcohol intake (ICC = 0.73; LCL = 0.67), family history of cancer (κ = 0.99; LCL = 0.95), and postmenopausal hormone use (κ = 0.92; LCL = 0.86) [23].

Results

Follow-up lasted an average of 7.1 years (range 1–19 years). The majority of these breast cancer cases were initially diagnosed with localized disease (63.0%; Table 1). Women were, on average, 59.4 years of age at diagnosis. Overall, 10.8% (N = 1,188) of breast cancer cases experienced additional cancer diagnoses; 1,070 women developed one additional cancer after breast cancer, 105 women developed two, 12 women developed three, and one woman had four additional cancer diagnoses after breast cancer. Second breast cancer diagnoses were the most common (N = 488, 4.5%), followed by colorectal (N = 132, 1.2%), endometrial (N = 113, 1.4%), and ovarian cancer (N = 36, 0.4%).
Table 1

Descriptive characteristics of 10,953 female breast cancer cases diagnosed in Wisconsin, 1987–2000

 

N

Percent (%)

5-Year cumulative incidence per 1,000a

Characteristics at the initial diagnosis (all cases)

Year of diagnosis

   

    1987–1991

4,616

42.1

74.1

    1992–1995

2,649

24.2

77.0

    1996–2000

3,688

33.7

45.8

Age (years)

   

    18–39

550

5.0

56.1

    40–49

1,661

15.2

49.4

    50–59

2,970

27.1

54.7

    60–69

3,860

35.2

73.2

    70–79

1,912

17.5

91.5

Stage of disease

   

    Local

6,898

63.0

66.5

    Regional

3,168

28.9

66.6

    Distant

251

2.3

86.3

    Unknown

636

5.8

76.5

Site of the second primary diagnosesb(ICD-O code)

    Anyc

1,188

10.8

67.7

    Breast (C50)

488

4.5

27.6

    Colorectal (C18, C19, C20)

132

1.2

7.3

    Endometrial (C54)d

113

1.4

8.4

    Ovarian (C56)d

36

0.4

2.1

aCumulative incidence does not include the 12 months immediately following the initial breast cancer diagnosis for all subjects

bCategories are not exclusive since women had 1–4 cancer diagnoses after their breast cancer diagnosis

cSites other than breast, colorectal, endometrial and ovarian included lung (C34, N = 122), skin (C44, N = 22) excluding basal cell and squamous cell carcinoma, hematopoietic (C42, N = 34) excluding AML, and lymph nodes (C42, N = 32). Less frequent (total N = 90) site codes included C17, 21, 22, 23, 24, 31, 37, 41, 47, 48, 49, 51, 55, 57, 69, 71, 73, 76, and 80

dAfter exclusions for hysterectomy or bilateral oophorectomy, N = 8,385 breast cancer cases in endometrial models and N = 9,356 in ovarian cancer models. Abbreviation: ICD-O, International Classification of Diseases for Oncology

Extent of disease at the initial breast cancer diagnosis was associated with risk of a second primary diagnosis. Diagnosis with distant breast cancer was associated with an increased risk of a second breast primary as compared with women initially diagnosed with local disease (HR 2.76, 95% CI: 1.65–4.61). Regionally-spread disease at the initial diagnosis was not associated with an increased risk of a second breast cancer (HR 0.94, 95% CI: 0.75–1.17) (data not shown). Restricting the analysis to breast cancer cases initially diagnosed with local or regional disease did not materially change the associations with reproductive and lifestyle characteristics.

Second cancer at any site

Cumulative incidence of a second cancer (all sites combined) increased according to later ages at menopause relative to premenopausal women (Fig. 1). Conversely, women who reported using postmenopausal hormones at the time of the initial breast cancer diagnosis (recent use) were generally at lower, and former users at intermediate, risk of any second cancer relative to never users of postmenopausal hormone therapy (Fig. 2). Overall, leaner postmenopausal breast cancer cases were less likely than cases with greater body mass index to experience a second primary cancer diagnosis at any site (Fig. 3). The increased risk of second cancer after breast cancer according to positive family history of breast cancer was also apparent for all second cancer sites combined (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs10549-006-9446-y/MediaObjects/10549_2006_9446_Fig1_HTML.gif
Fig. 1

Estimated cumulative incidence curve for a second primary diagnosis of any cancer by age at menopause

https://static-content.springer.com/image/art%3A10.1007%2Fs10549-006-9446-y/MediaObjects/10549_2006_9446_Fig2_HTML.gif
Fig. 2

Estimated cumulative incidence curve for a second primary diagnosis of any cancer by postmenopausal hormone use

https://static-content.springer.com/image/art%3A10.1007%2Fs10549-006-9446-y/MediaObjects/10549_2006_9446_Fig3_HTML.gif
Fig. 3

Estimated cumulative incidence curve for a second primary diagnosis of any cancer by quartiles of body mass index in postmenopausal women

https://static-content.springer.com/image/art%3A10.1007%2Fs10549-006-9446-y/MediaObjects/10549_2006_9446_Fig4_HTML.gif
Fig. 4

Estimated cumulative incidence curve for a second primary diagnosis of any cancer by family history of breast cancer

Second breast cancer

Risk of second primary cancer of the breast did not appear to be associated with age at menarche or age at first birth. Breast cancer cases reporting high parity (4 or more full-term pregnancies) had a lower risk of second breast cancers compared with nulliparous women (HR 0.63, 95% CI: 0.47–0.86). Menopausal status was not related to risk of subsequent breast cancer. However; cases with older ages at menopause (55+ years) had an approximate 50% increase in risk of subsequent breast cancer (95% CI: 1.07–2.31) compared to cases who reached menopause at younger ages (<45 years). Oral contraceptives and postmenopausal hormone use were not strongly associated with risk of second breast cancer among the breast cancer cases enrolled in our study (Table 2).
Table 2

Hazard ratiosa and 95% confidence intervals of a second primary cancer diagnosis according to reproductive and hormonal risk factors

Risk factor

Site of second primary

Breast

Colorectal

Endometrial

Ovarian

N

HR (95% CI)

N

HR (95% CI)

N

HR (95% CI)

N

HR (95% CI)

Age at menarche

    <12

83

1

24

1

25

1

8

1

    12

116

1.01 (0.76–1.34)

35

0.95 (0.56–1.61)

31

0.90 (0.53–1.53)

6

0.53 (0.18–1.55)

    13

158

1.18 (0.90–1.55)

38

0.83 (0.50–1.40)

26

0.65 (0.37–1.13)

7

0.50 (0.18–1.41)

    14+

120

0.92 (0.69–1.23)

34

0.66 (0.39–1.13)

28

0.67 (0.39–1.16)

14

1.05 (0.43–2.55)

    P-value

 

0.37

 

0.05

 

0.19

 

0.58

Age at first birth (parous only)

    <25

65

1

17

1

18

1

4

1

    25–29

192

0.93 (0.70–1.24)

53

0.91 (0.52–1.58)

47

0.85 (0.49–1.48)

13

0.97 (0.31–3.05)

    30–34

113

1.01 (0.74–1.39)

30

0.85 (0.46–1.56)

25

0.86 (0.45–1.61)

7

1.09 (0.30–3.91)

    35+

48

1.07 (0.72–1.59)

16

1.05 (0.51–2.16)

7

0.64 (0.25–1.62)

5

2.32 (0.55–9.72)

    P-value

 

0.63

 

0.98

 

0.42

 

0.10

Parity

    Nulliparous

69

1

16

1

15

1

5

1

    1

59

1.06 (0.75–1.50)

19

1.45 (0.74–2.83)

11

0.96 (0.44–2.11)

3

0.82 (0.20–3.47)

    2

140

0.99 (0.74–1.32)

30

1.01 (0.55–1.86)

18

0.65 (0.33–1.29)

8

0.82 (0.27–2.55)

    3

104

0.87 (0.64–1.18)

29

1.10 (0.60–2.04)

30

1.15 (0.62–2.15)

10

1.19 (0.40–3.53)

    4 +

115

0.63 (0.47–0.86)

38

0.78 (0.43–1.40)

38

0.91 (0.50–1.65)

9

0.67 (0.22–2.03)

    P-value

 

0.002

 

0.08

 

0.77

 

0.28

Menopausal status

    Postmenopausal

360

1

128

1

91

1

29

1

    Premenopausal

113

0.96 (0.64–1.45)

3

1.29 (0.23–7.14)

22

1.70 (0.78–3.74)

5

0.82 (0.16–4.12)

Age at menopauseb

    <45

56

1

16

1

9

1

4

1

    45–49

75

1.16 (0.81–1.64)

18

0.88 (0.45–1.73)

17

0.75 (0.33–1.70)

7

1.37 (0.39–4.77)

    50–54

122

1.15 (0.83–1.60)

55

1.53 (0.87–2.68)

29

0.61 (0.28–1.30)

10

0.89 (0.27–2.90)

    55+

55

1.57 (1.07–2.31)

19

1.44 (0.74–2.82)

25

1.61 (0.74–3.51)

5

1.31 (0.35–4.99)

    P-value

 

0.08

 

0.04

 

0.13

 

0.98

Oral contraceptive use

    Never

311

1

116

1

81

1

29

1

    Ever

174

0.90 (0.72–1.14)

15

0.60 (0.33–1.07)

34

0.85 (0.53–1.36)

12

0.76 (0.34–1.65)

Postmenopausal hormone use

    Never

199

1

95

1

59

1

14

1

    Ever

91

0.85 (0.66–1.09)

20

0.43 (0.26–0.70)

17

0.85 (0.48–1.48)

9

1.14 (0.47–2.79)

    Recent

57

0.88 (0.64–1.20)

5

0.21 (0.08–0.52)

13

1.28 (0.66–2.47)

7

1.36 (0.49–3.73)

aRegression models conditional on age; hazard ratios adjusted for year of diagnosis, stage of disease at the initial breast cancer diagnosis, family history of breast cancer, pack-years of cigarette smoking, recent alcohol intake, parity, postmenopausal hormone therapy, menopausal status, and body mass index

bAmong postmenopausal women only. N = 8,020 in models for breast and colorectal. N = 6,513 for ovarian cancer. N = 5,724 in models for endometrial cancer

Abbreviations: HR, hazard ratio; CI, confidence interval

Note: Among all women: N = 10,953 in models for breast and colorectal. N = 9,356 for ovarian cancer. N = 8,385 in models for endometrial cancer

Among postmenopausal women, second primary breast cancer risk was elevated by about 60% for the fourth versus first quartile categories of both body mass and weight gain, with a statistically significant positive trend for increasing risk of second breast cancer associated with greater weight gain (P = 0.02) (Table 3). We did not observe an association with height, smoking, or alcohol intake and second breast cancer risk. A positive family history of breast cancer was associated with risk of a second primary breast cancer diagnosis (HR 1.35, 95% CI: 1.09–1.66). The association was stronger for women with two or more female relatives with breast cancer (HR 1.79, 95% CI: 1.12–2.85).
Table 3

Hazard ratiosa and 95% confidence intervals of a second primary cancer diagnosis according to lifestyle risk factors and family history of breast cancer

Risk factor

Site of second primary

Breast

Colorectal

Endometrial

Ovarian

N

HR (95% CI)

N

HR (95% CI)

N

HR (95% CI)

N

HR (95% CI)

Body mass index quartileb (kg/m2)

    1 (<22.5)

64

1

24

1

17

1

8

1

    2 (22.5–25.0)

86

1.26 (0.91–1.75)

24

0.91 (0.51–1.60)

18

0.98 (0.50–1.90)

7

0.81 (0.29–2.26)

    3 (25.1–28.8)

103

1.57 (1.15–2.16)

39

1.54 (0.92–2.59)

19

1.07 (0.55–2.07)

6

0.67 (0.23–1.96)

    4 (≥28.9)

98

1.56 (1.13–2.16)

40

1.67 (0.99–2.82)

36

2.23 (1.23–4.05)

7

0.80 (0.28–2.29)

    P-value

 

0.003

 

0.07

 

<0.0001

 

0.45

Adult weight gain quartileb,c (kg)

    Weight loss

7

0.56 (0.24–1.24)

5

0.69 (0.25–1.87)

5

2.04 (0.63–6.55)

2

1.49 (0.26–8.47)

    1 (0–9.06)

44

1

21

1

8

1

5

1

    2 (9.07–16.78)

65

1.31 (0.89–1.93)

18

0.67 (0.36–1.27)

13

1.57 (0.65–3.83)

5

0.88 (0.25, 3.08)

    3 (16.79–36.28)

79

1.71 (1.17–2.49)

25

0.96 (0.52–1.74)

18

2.40 (1.02–5.61)

5

0.81 (0.23–2.90)

    4 (≥36.29)

91

1.63 (1.03–2.60)

26

0.65 (0.29–1.47)

21

3.14 (1.18–8.38)

6

1.35 (0.31–5.84)

    P-value

 

0.02

 

0.81

 

0.03

 

0.54

Tallest adult height quartiled (m)

    1 (< 1.60)

93

1

30

1

23

1

6

1

    2 (1.60–1.62)

130

1.09 (0.83–1.42)

34

0.96 (0.59–1.58)

23

0.78 (0.43–1.39)

11

1.43 (0.53–3.88)

    3 (1.63–1.67)

151

1.26 (0.97–1.63)

40

1.18 (0.74–1.91)

43

1.55 (0.93–2.57)

9

1.01 (0.35–2.92)

    4 (≥1.68)

110

1.09 (0.82–1.44)

28

1.05 (0.63–1.77)

23

0.93 (0.52–1.66)

9

1.30 (0.46–3.69)

    P-value

 

0.46

 

0.33

 

0.77

 

0.71

Smoking status

    Never

251

1

63

1

60

1

22

1

    Former

140

1.09 (0.89–1.35)

51

1.58 (1.08–2.31)

35

1.14 (0.74–1.74)

10

0.83 (0.38–1.79)

    Current

92

1.00 (0.78–1.28)

18

1.05 (0.61–1.81)

18

0.93 (0.54–1.61)

3

0.33 (0.10–1.13)

Recent alcohol intake (drinks/week)

    None

75

1

23

1

23

1

10

1

    Any

405

1.09 (0.85–1.41)

107

1.27 (0.81–2.02)

88

0.85 (0.53–1.36)

25

0.45 (0.21–0.98)

    1–7

336

1.09 (0.85–1.41)

81

1.15 (0.72–1.84)

74

0.86 (0.53–1.39)

20

0.44 (0.20–0.97)

    >7

69

1.09 (0.78–1.53)

26

1.92 (1.07–3.43)

14

0.84 (0.42–1.69)

5

0.55 (0.18–1.72)

    P-value

 

0.91

 

0.01

 

0.47

 

0.87

Family history of breast cancer

    No

358

1

99

1

87

1

25

1

    Yes

118

1.35 (1.09–1.66)

30

1.08 (0.72–1.63)

23

1.10 (0.69–1.75)

9

1.50 (0.70–3.25)

    2 or more relatives

19

1.79 (1.12–2.85)

7

1.75 (0.81–3.79)

4

1.79 (0.65–4.92)

3

4.28 (1.25–14.6)

aRegression models conditional on age; hazard ratios adjusted for year of diagnosis, stage of disease at the initial breast cancer diagnosis, family history of breast cancer, pack-years of cigarette smoking, recent alcohol intake, parity, postmenopausal hormone therapy, menopausal status, and body mass index

bAmong postmenopausal women only. N = 8,020 in models for breast and colorectal. N = 6,513 for ovarian cancer. N = 5,724 in models for endometrial cancer

cAdjusted for adult height and weight at age 18–20 instead of body mass index

dAdjusted for weight residuals instead of body mass index. Abbreviations: HR, hazard ratio; CI, confidence interval

Note: Among all women, 10,953 in models for breast and colorectal. N = 9,356 for ovarian cancer. N = 8,385 in models for endometrial cancer

Colorectal cancer after breast cancer

We observed an inverse trend between increasing age at menarche and decreasing risk of a second primary colorectal cancer diagnosis (P = 0.05). Women who reported reaching menarche at age 14 or older had 0.66 times (95% CI: 0.39–1.13) the risk of colorectal cancer compared to women younger than 12 at menarche. Age at first birth, parity and menopausal status did not appear to be associated with risk of subsequent colorectal cancer. We observed a positive trend between increasing age at menopause and increasing risk of colorectal cancer after breast cancer (P = 0.04). Cases with older ages at menopause (55+ years) had an approximate 50% increase in risk of subsequent colorectal cancer (95% CI: 0.74–2.82) compared to cases who reached menopause at younger ages (<45 years). Results suggested that ever use of oral contraceptives (HR 0.60, 95% CI: 0.33–1.07) and/or postmenopausal hormones (HR 0.43, 95% CI: 0.26–0.70) prior to the first breast cancer diagnosis was related to decreases in subsequent colorectal cancer risk, compared to never users (Table 2).

We observed a suggested increase in risk of second colorectal cancer (P = 0.07) according to greater body mass, but not weight gain (P = 0.81). We did not observe an association with height and subsequent colorectal cancer risk. Former smokers (HR = 1.58: 95% CI: 1.08–2.31), but not current smokers (HR = 1.05: 95% CI: 0.61–1.81), appeared to have an increased risk of subsequent colorectal cancer compared to never smokers. Greater recent consumption of alcoholic beverages prior to the first breast cancer diagnosis was associated with increasing risk of colorectal cancer (HR = 1.92; 95% CI: 1.07–3.43 for >7 drinks per week versus 0; P-trend = 0.01). The risk estimate for colorectal cancer after breast cancer was elevated (HR = 1.75) among women with two or more first-degree relatives with breast cancer, although the limited sample size resulted in wide confidence intervals (95% CI: 0.81–3.79) (Table 3).

Endometrial cancer after breast cancer

Risk of second primary cancer of the endometrium did not appear to be strongly associated with age at menarche, age at first birth, parity, menopausal status, age at menopause, oral contraceptives, or postmenopausal hormone use among the breast cancer cases enrolled in our study (Table 2). Risk of second primary endometrial cancer for the fourth versus first quartile categories of both body mass and weight gain was elevated two to three-fold among the postmenopausal women enrolled in our study (95% CI: 1.23–4.05 and 1.18–8.38 for body mass index and weight gain, respectively), and both greater body mass index and adult weight gain prior to the initial breast cancer diagnosis showed significant positive trends in risk (P<0.001 and 0.03, respectively). Height, smoking, and alcohol intake were not strongly related with risk of subsequent endometrial cancer. The risk estimate for endometrial cancer after breast cancer was elevated but imprecise (HR = 1.79; 95% CI: 0.65–4.92) among women with two or more first-degree relatives with breast cancer (Table 3).

Ovarian cancer after breast cancer

Risk of second primary ovarian cancer was not associated with age at menarche. Our results suggested a possible increase in risk of subsequent ovarian cancer according to older ages at first birth with an increase in the hazard ratio from 0.97 (95% CI: 0.31–3.05) among women age 25–29 at first birth to 2.32 (95% CI: 0.55–9.72) among women older than 35, compared to women younger than 25 at first birth; however, the P-value for age at first birth as a continuous term did not approach statistical significance (P = 0.1). Parity, menopausal status, age at menopause, oral contraceptives, or postmenopausal hormone use did not appear to be associated with risk of subsequent ovarian cancer risk (Table 2).

Risk of ovarian cancer after breast cancer was not related to body mass index, adult weight gain, or height among the breast cancer cases enrolled in our study. While the association was not statistically significant, results suggested a reduced risk of ovarian cancer among breast cases who smoked at the time of the breast cancer diagnosis (HR 0.33, 95% CI: 0.10–1.13). Breast cancer cases who reported consuming alcoholic beverages had a 55% reduction in risk (95% CI: 0.21–0.98) of subsequent ovarian cancer compared to non-drinking cases. Second-primary ovarian cancer risk was elevated according to family history of breast cancer (HR 1.50, 95% CI: 0.70–3.25), especially among women who reported two or more female relatives with breast cancer (HR 4.28, 95% CI: 1.25–14.6) (Table 3).

Discussion

Results from this study demonstrate associations between breast cancer and other cancer sites with several shared reproductive and lifestyle risk factors; in many instances, these factors are reflective of hormonal influences. Reproductive factors such as ages at menarche and menopause, parity, and oral contraceptive use may initiate life-long changes to the hormonal milieu [24, 25]. The combination of early age at menarche, low parity, and late age at menopause increases the number of ovulatory menstrual cycles and associated hormonal fluctuations. Oral contraceptive use, conversely, suppresses ovulation. In our study we observed patterns of decreased risk of second cancer associated with later ages at menarche (in relation to colorectal and endometrial cancer risk), high parity (in relation to second breast cancer risk), and earlier ages at menopause (in relation to second breast and colorectal cancer risk) among breast cancer survivors.

Cumulative incidence for all second cancers combined was higher among postmenopausal women, particularly those with later ages at menopause (Fig. 1). This finding is in agreement with one previous investigation [26] but not another [5]; in the former, estimates were presented in relation to all non-breast cancer malignancies and in the latter, age was used as a surrogate for menopausal status. Our results are also in agreement with reports that suggest an increased risk of second cancer (all sites) among breast cancer survivors with greater body mass [27]. The putative detrimental effects of greater body weight and adult weight gain may continue to influence a woman’s cancer risk even in the presence of breast cancer therapy.

Previous investigations of contralateral breast cancer risk have reported weak effects, including null or statistically non-significant estimates (often in conflicting directions) associated with age at menarche, age at first birth, and parity [3, 2832]. Premenopausal breast cancer has been associated with non-significant decreases in risk of subsequent contralateral breast [29, 30]. Our findings with respect to second breast cancer were null for menopausal status; however, we did observe increases in risk among postmenopausal women with older ages at menopause (55 years or older) compared to younger (<45 years). In an earlier study, Horn and Thompson observed a similar, but statistically non-significant, pattern of greatest risk among women with older ages at menopause [29].

In this analysis, we observed a suggested, but not statistically significant, reduction in risk of second breast cancer associated with oral contraceptive use. Previous findings of ever use of oral contraceptives in relation to contralateral breast cancer have been null [28]; however, only 16 of 292 cases and 13 of 264 controls reported ever use of oral contraceptives in one study [29], while another presented only categories of 0, <10, and 10+ years of use [32]. Our study agrees with others by finding no additional risk of second primary breast cancer among breast cancer survivors associated with postmenopausal hormone use [29].

Several studies have evaluated body mass in relation to risk of contralateral breast cancer among breast cancer survivors, and the results are fairly divided with four of eight studies showing significant increased risk estimates equal to 1.6–2.2 associated with greater body mass or body weight [27, 32, 33]. Inconsistencies in study results may be due to several reasons, including limited power, reliance on self-reported body weight, cancer or treatment-related weight changes, and inadequate control for confounding factors [26, 2830, 34]. The inclusion of premenopausal breast cancer cases and cases of carcinoma in situ of the breast may also have contributed to null results in previous studies.

Greater weight gain is a well accepted as a risk factor for postmenopausal breast cancer among women without a personal history of cancer [35]. Breast cancer cases diagnosed as part of the Nurses’ Health Study cohort who gained greater amounts of weight after diagnosis were more likely to die from breast cancer or other causes and to experience a recurrence, although results were strongest for never-smokers and premenopausal cases [36]. We are unaware of any other published findings regarding subsequent breast cancer risk among breast cancer survivors according to weight gain prior to the initial breast cancer diagnosis, although Bernstein [37] reported an increased risk of mortality among women with two breast cancer diagnoses and modest gains in body mass.

Numerous studies have reported increased risks for first primary breast cancer among taller women [38], whereas only one other study has reported on the association between height and risk of contralateral breast cancer; results, like those of our study, were not statistically significant [30]. Tobacco exposure at a young age may be associated with shorter stature or correlated with other factors that are observed among women who do not achieve their maximum height potential, such as inadequate early life nutrition or lower socioeconomic status [39]. In general, smoking is not strongly associated with first primary breast cancer risk although a few studies have described an increased risk associated with smoking of long durations begun in adolescence [40]. Two of three previous studies suggested that smoking increased risk of contralateral breast cancer, although the results were not strong and did not demonstrate dose-response relations [26, 29, 34]. We found no increased risk of second primary breast cancer associated with cigarette smoking. Similarly, other studies have reported null associations between alcohol intake and risk of second breast cancer [32, 34].

Family history has often been investigated in relation to risk of second primary breast cancer among breast cancer survivors [3]. Among 11 studies of contralateral breast cancer, all have reported elevated risks associated with family history. Among the seven studies with statistically significant findings, hazard ratios range from 1.5 to 3.2 [3, 15, 26, 2833, 41, 42]. As with risk of primary breast cancer, risk estimates are generally stronger when two or more first-degree relatives have been diagnosed [15, 29, 30]. Our study suggests that risk of a second breast diagnosis is increased by about 80% among breast cancer survivors with two affected first-degree relatives.

One previous study reported finding no difference in risk of colorectal cancer after breast cancer among nulliparous women relative to parous women [43]. Premenopausal breast cancer was associated with non-significant decreases in risk of subsequent colorectal cancer in one study [43], while another reported higher standardized incidence ratios for second primary colorectal cancer among women with premenopausal (≤45 years; SIR = 1.44, 95% CI: 1.30–1.58), relative to postmenopausal (≥56 years; SIR = 1.20, 95% CI: 1.17–1.24), breast cancer [5]. In our study, the hazard ratio for colorectal cancer after breast cancer was elevated among women with premenopausal breast cancer; however, the confidence interval was wide and non-significant (HR = 1.29, 95% CI: 0.23–7.41).

A modest reduction in colorectal cancer risk has been associated with oral contraceptive use among women without breast cancer [4446], we are unaware of other published estimates of colorectal cancer risk in relation to oral contraceptive use among breast cancer survivors. In this analysis, we observed a suggested, but not statistically significant, reduction in risk of secondary cancer, including second colorectal cancer, associated with oral contraceptive use. Colorectal cancer risk was reduced by 55–79% among breast cancer survivors who reported use of postmenopausal hormones, depending on the time since use. This is consistent with observed reductions in colorectal cancer risk associated with postmenopausal hormone use in women without a personal history of cancer [47]. One other study reported no significant association with second primary colorectal cancer for either estrogen alone or combined estrogen–progestin use among breast cancer cases [43].

Our results are also in agreement with reports that suggest an increased risk of colorectal cancer among breast cancer survivors with greater body mass [43]. We are unaware of any other published findings regarding subsequent colorectal cancer risk among breast cancer survivors according to weight gain. Our results suggested an increased risk of colorectal cancer after breast cancer associated with higher alcohol intake (P-value 0.01). Kmet reported a two-fold increased risk of colorectal cancer among breast cancer cases with a positive family history of breast cancer; our results were not inconsistent with an increased risk [43] (Table 3).

In our study, risk of endometrial cancer after breast cancer increased according to greater body mass index and adult weight gain. Our results are in agreement with a previous report that suggested an increased risk of endometrial cancers among breast cancer survivors with greater body mass [27]. We also observed a suggested, but not statistically significant, reduction in risk of endometrial cancer associated with oral contraceptive use. We are unaware of any other published findings regarding subsequent endometrial cancer risk among breast cancer survivors according to oral contraceptive use or weight gain prior to the initial breast cancer diagnosis.

A recent study reported significantly higher standardized incidence ratios for ovarian cancer after premenopausal (≤45 years; SIR = 2.84, 95% CI: 2.61–3.09) breast cancer, compared to postmenopausal (≥56 years; SIR= 1.12, 95% CI: 1.06–1.19) breast cancer [5]. We did not observe an increase in risk of ovarian cancer after breast cancer according to premenopausal status, relative to postmenopausal. In this analysis, we observed a suggested, but not statistically significant, reduction in risk of secondary ovarian cancer associated with oral contraceptive use. We are unaware of other published estimates of ovarian cancer risk in relation to oral contraceptive use among breast cancer survivors. We also found that risk of ovarian cancer was increased almost 4-fold among breast cancer survivors with a family history of two first-degree relatives with breast cancer, possibly reflecting influence of major susceptibility genes such as BRCA1 [48].

Some limitations must be considered in interpreting our results. Tumor registry treatment information is limited to planned first course therapy; we found even this limited information to be reported incompletely. Changes in reporting requirements over time, including the inclusion of clinic reporting starting in 1992, contributed to incomplete treatment reports (e.g., tamoxifen use) when only hospitals submitted reports to the registry. We performed all analyses both adjusted and unadjusted for planned first course therapy (radiation, chemotherapy, hormone therapy). As results remained unchanged, final models did not adjust for the first course therapy variables. We were unable to further evaluate associations between breast cancer treatment and second primary cancer. However, the strength of this population-based study was in the availability of lifestyle factor information which is often lacking in patient series or clinical trials.

Other limitations should be noted. Lifestyle factor information was available only before the initial breast cancer diagnosis, so changes after diagnosis were unknown. Since treatment information may have been incomplete, as discussed above, we were unable to eliminate all women who were not at risk of certain types of additional cancer diagnoses, such as women who received bilateral mastectomies or hysterectomies. Information regarding reproductive surgery after diagnosis was not available. Since information on type of breast cancer surgery also was not available, we were unable to identify which women received bilateral mastectomies at the time of their initial breast cancer diagnoses to exclude them from evaluations of second breast cancer diagnoses. Inclusion of women with bilateral mastectomies in subsequent breast cancer analyses likely attenuated our effect estimates by increasing the number of women presumed at risk. We were able to eliminate women with pre-diagnosis hysterectomies and/or bilateral oophorectomies from analyses of endometrial and ovarian cancer. All breast cancer survivors may undergo more intense screening including mammograms and colonoscopies after diagnosis. We were not able to account for these screening efforts after the initial breast cancer diagnosis. Screening bias of this nature may have resulted in overestimates of second primary cancers among these breast cancer survivors.

According to the National Cancer Institute’s Office of Cancer Survivorship, over two million breast cancer survivors are estimated to be living in the US, making up the largest group of cancer survivors in the country [49]. Breast cancer survival has been increasing for the past several decades, likely due to improvements in early detection and therapy. These study results suggest that over 10% of breast cancer survivors—at a 5-year incidence of approximately 68 per 1,000—may experience another cancer diagnosis. While most research to date has focused on contralateral breast cancer among survivors, over half of second cancer diagnoses among breast cancer survivors occur at other sites. The long-term health of breast cancer survivors would benefit from additional research to evaluate effective cancer prevention strategies.

Acknowledgements

A preliminary version of this work was presented at the 37th Annual Meeting of the Society for Epidemiologic Research, June 17, 2004 in Salt Lake City, Utah, USA.

We thank Drs. Patrick Remington, Henry Anderson, and Jane McElroy for their advice and assistance during this project; Laura Stephenson, Robert Borchers and the rest of the staff of the Wisconsin Cancer Reporting System for providing data and technical support; and Jerry Phipps, Mary Pankratz, and the staff of the Wisconsin Women’s Health Study for their dedication. We are indebted to the generous participation of breast cancer survivors throughout Wisconsin.

This project was supported in part by grant CA47147 from the National Cancer Institute and faculty startup funds from the University of Wisconsin School of Medicine & Public Health.

Copyright information

© Springer Science+Business Media, LLC 2006