|Home | About | Journals | Submit | Contact Us | Français|
An outcome of considerable concern among breast cancer survivors is the development of second primary breast cancer. However, evidence regarding how potentially modifiable lifestyle factors modulate second breast cancer risk is limited. We evaluated the relationships between obesity, alcohol consumption, and smoking on risk of second primary invasive contralateral breast cancer among breast cancer survivors.
Utilizing a population-based nested case-control study design, we enrolled 365 patients diagnosed with an estrogen receptor–positive (ER+) first primary invasive breast cancer and a second primary contralateral invasive breast cancer, and 726 matched controls diagnosed with only an ER+ first primary invasive breast cancer. Obesity, alcohol use, and smoking data were ascertained from medical record reviews and participant interviews. Using conditional logistic regression we evaluated associations between these three exposures and second primary contralateral breast cancer risk.
Obesity, consumption of ≥ 7 alcoholic beverages per week, and current smoking were all positively related to risk of contralateral breast cancer (odds ratio [OR], 1.4; 95% CI, 1.0 to 2.1; OR, 1.9; 95% CI, 1.1 to 3.2; and OR, 2.2; 95% CI, 1.2 to 4.0, respectively). Compared with women who consumed fewer than seven alcoholic beverages per week and were never or former smokers, women who consumed ≥ 7 drinks per week and were current smokers had a 7.2-fold (95% CI, 1.9 to 26.5) elevated risk of contralateral breast cancer.
Our population-based study adds to the limited available literature and suggests that obesity, smoking, and alcohol consumption influence contralateral breast cancer risk, affording breast cancer survivors three means of potentially reducing this risk.
High incidence rates of invasive breast cancer, coupled with the disease's now greater than 90% 5-year survival rate in the United States, have resulted in a large and ever growing number of breast cancer survivors. Second primary contralateral breast cancer is an outcome of particular concern to breast cancer survivors since they have a two to six times greater risk of developing a contralateral breast cancer than women in the general population have of developing a first breast cancer.1 While adjuvant hormone therapy can reduce contralateral breast cancer risk by 47%,2 little information on other factors within a woman's control that can influence this risk is known.
Obesity is a potential risk factor for contralateral breast cancer. While two early studies found no relationship between obesity and second primary breast cancer risk,3,4 four more recent studies reported positive associations.5–8 Data with respect to alcohol consumption and smoking are also mixed with most studies finding modest increases in risk that did not reach statistical signficance.3,4,7–9 These studies have been limited by relatively small sample sizes, although a large recently published case-control study found that regular alcohol consumption was associated with a 30% increase in risk of contralateral breast cancer, but that smoking was not.9
We evaluated the relationship between three potentially modifiable lifestyle factors, body mass index (BMI), alcohol use, and smoking, and risk of second primary contralateral breast cancer among survivors of invasive estrogen receptor (ER) –positive breast cancer. Given the growing number of breast cancer survivors, identification of potentially modifiable risk factors for second contralateral tumors could have broad clinical and personal relevance to this population.
We conducted a population-based nested case-control study of the risk of second primary invasive contralateral breast cancer that was approved by the Fred Hutchinson Cancer Research Center's institutional review board. Patients and controls were identified from the underlying cohort of 17,628 women diagnosed with a first primary invasive, stage I to IIIB, ER+ breast cancer at age 40 to 79 years in the four-county Seattle-Puget Sound region from January 1, 1990, to September 30, 2005. Data from the population-based cancer registry that serves western Washington, the Cancer Surveillance System (a participant in the National Cancer Institute's Surveillance, Epidemiology and End Results program since 1974) was used to assemble this cohort.
Cancer Surveillance System was used to identify patients with second primary invasive contralateral breast cancer diagnosed in our cohort. Eligible participants were included regardless of vital status. All women who were alive provided informed consent, and deceased women were enrolled through an institutional review board–approved waiver of consent. Patients were defined as women who developed invasive cancer in the breast contralateral to their first breast cancer 6 months or longer after their first breast cancer diagnosis from July 1, 1990, to March 31, 2007, in our four county catchment area. A total of 446 eligible patients were identified of which 369 (83%) were enrolled. Controls were individually matched 2:1 to patients on age, year of diagnosis, county, race/ethnicity, and stage (localized v regional). In addition, controls had to be alive and reside in their county of diagnosis from their breast cancer diagnosis to at least the duration between their matched patient's first and second breast cancer diagnoses. A total of 982 eligible controls were identified of which 734 (75%) were enrolled.
Data on demographic, epidemiologic, and clinical factors were ascertained from two sources—structured interviewer-administered telephone questionnaires and a detailed medical record review. Interviewers and record abstractors were blinded to our study's primary hypotheses, but not to case/control status. The telephone interview queried women on a variety of topics. Participant recall was enhanced through the use of showcards with common responses and individualized life-events calendars. Women were asked about breast cancer surgeries, radiation therapy, chemotherapy, and adjuvant hormone therapy. For hormone therapy data on specific medications used, patterns of use, and start and stop dates were collected. Women were also asked to recall their history of various established breast cancer risk factors. Self-reported data on height and weight at multiple time points including at the time of their first breast cancer diagnosis and at their assigned reference date (which for patients with contralateral breast cancer was the date of their contralateral breast cancer diagnosis and for controls was the date of their matched patient's contralateral breast cancer diagnosis) were collected. Lifetime alcohol consumption through reference date was collected through asking about the number of alcohol containing beverages women consumed at different times in their lives and when their patterns of use changed. Smoking history through reference date was assessed by determining whether women ever smoked, and then asking ever smokers about their smoking patterns, when they started smoking, and if they ever stopped.
Medical records were sought from multiple sources including oncology and primary care practices so that complete data on breast cancer treatments, clinical and pathologic tumor characteristics, and breast cancer risk factors could be abstracted. For the 103 patients with contralateral breast cancer and 161 controls who were deceased at enrollment, data were only collected from medical records.
For quality control purposes, the study coordinator evaluated a random 10% sample of all recorded telephone interviews to ensure study protocols were correctly followed. Similarly, a random 10% of medical records were rereviewed by our medical record coordinator to ensure that data were abstracted consistently.
BMI data from medical records were prioritized over self-reported data. Medical record data on BMI at first breast cancer diagnosis and reference data were available for 86% and 63% of all participants, respectively, and in total BMI data were available at first breast cancer diagnosis and reference date for 98% (355 patients with contralateral breast cancer and 712 controls) and 88% (309 patients with contralateral breast cancer and 649 controls) of participants, respectively. BMI categories were based on WHO criteria (reference category = BMI < 25.0 kg/m2; overweight = BMI between 25.0 to 29.9 kg/m2; obese = BMI ≥ 30 kg/m2). With respect to BMI at reference date, there was 75% agreement (κ = 0.63) between self-reported and medical record-based BMI data that did not differ by case-control status (77% for patients and 74% for controls). In general, self-reported BMI was lower than BMI abstracted from medical records. Among those with data from both sources the mean self-reported BMI at reference date was 26.4 (standard deviation, 5.8), and the mean BMI from medical records was 28.2 (standard deviation, 6.2).
Data on alcohol use and smoking history relative to both first breast cancer diagnosis and reference date were ascertained only from self-report because of an inability to obtain reliable data on these exposures from medical records. Alcohol consumption was analyzed as average number of alcoholic beverages consumed per week at two time points: at the time of first breast cancer diagnosis and during the interval between first breast cancer diagnosis and reference date. Alcohol consumption was grouped into four categories (nondrinkers, < 3 drinks/week, 3 to 6.9 drinks/week, and ≥ 7 drinks/week), and smoking was grouped into three categories (never, former, and current).
Associations between BMI, smoking, and alcohol consumption and risk of second primary contralateral breast cancer were estimated by conditional logistic regression using Stata SE (Stata Corp, College Station, TX). As a result, all statistical models were implicitly adjusted for each of the matching variables. All analyses were also adjusted for adjuvant hormone therapy and chemotherapy, so the four patients with contralateral breast cancer and eight controls with missing data for one or both of these variables were excluded from all analyses, leaving 365 patients with contralateral breast cancer and 726 controls. Odds ratios (ORs) and 95% CIs were calculated as estimates of the relative risk. All statistical tests were two sided. We systematically assessed a series of potential confounders and effect modifiers (those listed in Table 1). Factors that changed our risk estimates by greater than 10% when adjusted for were considered to be confounders and included in our final statistical models. Based on this approach, ORs for BMI were additionally adjusted for use of menopausal hormone at first breast cancer diagnosis; ORs for alcohol use were additionally adjusted for BMI at reference date; and ORs for smoking were additionally adjusted for first-degree family history of breast cancer. Effect modification was assessed using likelihood ratio testing. None of the factors listed in Table 1, including adjuvant hormone therapy, had a statistically significant (P < .05) interaction with any of the main effects we assessed.
We compared patients and controls with respect to several factors (Table 1). Other than controls being more likely to have received adjuvant hormone therapy compared with patients (P < .0001), there were no other statistically significant differences between these two groups. Of our 365 contralateral patients, 303 were ER+, 54 were ER-negative, and 12 had an unknown ER status.
Compared with women with a BMI lower than 25.0 kg/m2, those with a BMI ≥ 30.0 kg/m2 had an elevated risk of contralateral breast cancer (Table 2). This was observed both for BMI at first breast cancer diagnosis and at reference date, though the latter risk estimate was within the limits of chance. The relationship between BMI and contralateral breast cancer was unchanged when the analysis was restricted only to BMI data abstracted from medical records (OR, 1.5; 95% CI, 1.0 to 2.2).
Alcohol consumption was also positively related to contralateral breast cancer risk when assessed at both first breast cancer diagnosis and over the interval between first breast cancer diagnosis and reference date (Table 2). Similarly, compared with never smokers, current smokers at both first breast cancer diagnosis and at reference date had elevated risks of contralateral breast cancer. The relationship with current smoking did not vary by pack years of smoking (data not shown). Though the number of women who were current smokers at the time of their first breast cancer diagnosis but had quit by their reference date was small (14 patients and 29 controls), among these women current smoking at first breast cancer diagnosis was not related to contralateral breast cancer risk (OR, 0.8; 95% CI, 0.2 to 3.3). These latter two observations suggest that recency of smoking is most relevant to risk.
Compared with women who consumed 0 to 6.9 drinks/week and were never smokers, those who consumed ≥ 7 drinks/week and were current smokers at either first breast cancer diagnosis or reference date had substantially higher risks of contralateral breast cancer (Table 3).
All three potentially modifiable risk factors we investigated were positively related to risk of second primary contralateral breast cancer. The 40% to 50% elevation in risk associated with obesity we observed is consistent with the 58% elevation in contralateral breast cancer risk women with a BMI ≥ 30 kg/m2 had compared with women with a BMI lower than 25 kg/m2 enrolled in the National Surgical Adjuvant Breast and Bowel Project B-14 randomized trial of adjuvant tamoxifen in patients with node-negative invasive ER+ breast cancer.5 It is also consistent with the 56% elevation in risk of second primary breast cancer observed for women in the highest BMI quartile compared with women in the lowest quartile in a large population-based cohort of breast cancer survivors.7 Thus, our study adds to the growing body of evidence from recent studies that obesity may be an important risk factor for second primary contralateral breast cancer. The mechanisms through which obesity confers an increased risk of contralateral breast cancer are likely to be quite similar to those which govern its well established relationship with risk of first primary breast cancer.10 In postmenopausal women, adipose tissue is the primary producer of endogenous estrogens, and this is thought to be the primary pathway through which obesity confers an elevation in breast cancer risk. In addition, there is growing data to suggest that hyperinsulinemia may also be an important contributor to the relationship between obesity and breast cancer.11
We observed that consumers of ≥ 7 alcoholic beverages per week had a 90% increased risk of contralateral breast cancer. Our result is broadly consistent with the increased risk of contralateral breast cancer observed among women who had ever regularly consumed alcohol in a large recently published multicenter patient-control study.9 Our risk estimate is higher though than the 20% increase in risk this prior study observed among women who on average consumed ≥ 1 drink per day. Other smaller studies have also observed modest positive associations that were not statistically significant.3,7,8 While the reasons for this difference in the magnitude of the association are unclear, one potential explanation is that our study was restricted to women whose first breast cancer was ER+. Similar to obesity, the relationship between alcohol use and risk of first breast cancer is well established.12 The primary mechanism through which it is thought to elevate risk is hormone since alcohol consumption directly increases endogenous estrogen levels in postmenopausal women.13,14 There is also some evidence that alcohol use is more strongly related to risk of ER+ versus ER-negative breast cancer.15,16 Thus, based on the available biologic and epidemiologic evidence, the relationship between alcohol use and contralateral breast cancer may be stronger among women whose first breast cancer was ER+.
Our observation that current smokers have a substantial 2.2-fold increased risk of contralateral breast cancer is somewhat consistent with the results of two previous reports, although the magnitude of the association we observed is again higher. One of these studies observed a 70% increased risk among women who smoked more than a pack a day,4 and the other reported a 24% increased risk among ever smokers3 (compared with never smokers), but neither result was statistically significant. In contrast, the previously mentioned large multicenter case-control study that found a positive association with alcohol, did not observe a relationship between ever having smoked during the interval between first breast cancer diagnosis and reference date and contralateral breast cancer risk.9 However, it did not separate former and current smokers, making it somewhat difficult to directly compare our results. Lastly, a study evaluating risk of second primary breast cancer regardless of laterality found that current smoking was not related to risk.7
Unlike obesity and alcohol use, smoking is not a well established risk factor for postmenopausal breast cancer despite biologic plausibility for a positive association. Tobacco smoke contains carcinogens that are genotoxic to the rodent mammary gland,17 and in humans tobacco carcinogen DNA adducts have been found in breast tissue18,19 and these carcinogens can be metabolically activated in breast epithelial cells.20,21 Inconsistency in published results from epidemiologic studies may due to evolving smoking patterns in women. Several recent studies indicate that smoking initiation at a young age and/or before a first full-term pregnancy may be more strongly related to breast cancer risk than is initiation at older ages.22–25 In addition, greater intensity and duration of smoking appears to be positively associated with risk in several recent cohort and patient-control studies.24–28 Thus, earlier studies of smoking and breast cancer may have failed to find an association with risk because too few women included started smoking at a young age or had smoked at sufficient intensities or durations to detect an association. The interaction between alcohol consumption and smoking in relation to contralateral breast cancer observed here has not been previously reported. While our study had limited statistical power to assess interactions, our results suggest that smoking may confer an increased risk of contralateral breast cancer independent of alcohol use, alcohol's effect may depend on concurrent smoking, and women with higher levels of alcohol consumption who are current smokers may have particularly high risks of contralateral breast cancer. However, these relationships require confirmation.
A potential limitation of all patient-control studies is recall bias. Our BMI results were not impacted by this bias because our primary source of anthropometric data was medical records and risks were unchanged when analyses were restricted to data only from medical records. In contrast, data on smoking and alcohol use could not be reliably obtained from medical records and so analyses are based only on self-reported data. Consequently, data from enrolled deceased women could not be included. The extent of the potential resulting biases is unknown. Encouragingly, when we restricted our BMI analyses to self-reported data from alive women, the same association was observed when women with a BMI ≥ 30 kg/m2 at first breast cancer diagnosis were compared with women with a BMI lower than 25 kg/m2 (OR based on medical record data including both alive and deceased women: 1.6; 95% CI, 1.1 to 2.2 v OR based on self-reported data from alive women only: 1.6; 95% CI, 1.0 to 2.5). Given the considerable correlation within our three exposures between the two time points of interest, date of first breast cancer diagnosis and reference date (particularly for BMI and alcohol use), it was difficult to evaluate how changes in these exposures over time influenced contralateral breast cancer risk. However, in a subanalysis of women who were current smokers at the time of their first breast cancer diagnosis, but had quit smoking by reference date, the fact that they were current smokers at their first breast cancer diagnosis did not impact their risk of contralateral breast cancer. This suggests that recency of smoking may be most relevant to contralateral breast cancer risk. With respect to the generalizability of our results, this study was limited to women with ER+ first primary breast cancers, so the extent to which the risk factors studied influence risk of contralateral breast cancer among women with ER-negative first breast cancers could not be assessed. Also, this study included women diagnosed with their first breast cancer as long ago as 1990 when patterns or hormone therapy use were different then they are now. As a result fewer women ever used hormone therapy, or used if for a duration conferring maximal clinical benefit, than women today. While our analyses were adjusted for use of hormone therapy, interactions with hormone therapy could not be assessed given the size of our study.
Few studies, most with relatively small sample sizes, have evaluated the influence that potentially modifiable lifestyle factors have on risk of second primary contralateral breast cancer. Identifying such factors could be both of broad public health relevance and of individual importance to the growing number of breast cancer survivors. Our population-based study adds to the limited available literature and suggests that obesity, smoking, and alcohol are all positively related to risk affording breast cancer survivors three means of potentially reducing their risk of contralateral breast cancer.
The authors acknowledge the substantial contributions of Sarah Taylor and Heather O'Brien in the conduct of this research study; other staff member who made important contributions to this work were: Elisabeth Beaber, Nancy Blythe, Ann Bradshaw, Kay Byron, Fran Chard, Lora Cox, Diane DeHart, Sue Ellingson, Carolyn Howard, Dick Jacke, Jean Jue, Eileen Louie, Karen Lunna, Charlotte Palmberg, Amanda Phipps, Patty Pride, Babette Siebold, Camille Taylor, Loni Tipton, and Vicky Tran. Lastly, we want to acknowledge the time and generosity of all of the women who participated in this research.
See accompanying editorial on page 5301
Supported by Grant No. R01-CA097271 from the National Cancer Institute. The National Cancer Institute did not play a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: Christopher I. Li, Janet R. Daling, Peggy L. Porter, Kathleen E. Malone
Financial support: Janet R. Daling, Peggy L. Porter, Kathleen E. Malone
Provision of study materials or patients: Christopher I. Li, Janet R. Daling
Collection and assembly of data: Christopher I. Li, Janet R. Daling, Peggy L. Porter, Kathleen E. Malone
Data analysis and interpretation: Christopher I. Li, Peggy L. Porter, Mei-Tzu C. Tang, Kathleen E. Malone
Manuscript writing: Christopher I. Li, Janet R. Daling, Peggy L. Porter, Mei-Tzu C. Tang, Kathleen E. Malone
Final approval of manuscript: Christopher I. Li, Janet R. Daling, Peggy L. Porter, Mei-Tzu C. Tang, Kathleen E. Malone