Abstract
Atrial fibrillation (AF) is a common arrhythmia that poses a significant risk of stroke. Cross-sectional and case-control studies have shown evidence of associations between AF and breast or colorectal cancer, but there have been no longitudinal studies in which this has been assessed. We prospectively examined a cohort of 93,676 postmenopausal women enrolled in the Women's Health Initiative from 1994 to 1998 to determine whether there are relationships between baseline AF and the development of invasive breast or colorectal cancer. The prevalence of self-reported physician diagnosis of AF at baseline was 5.1%. Over approximately 15 years of follow-up, the incidence of invasive breast cancer was 5.7%, and the incidence of colorectal cancer was 1.6%. Adjusted hazard ratios and 95% confidence intervals were obtained using Cox proportional hazards models. We found no significant association between AF and incident colorectal cancer, but we did see a 19% excess risk of invasive breast cancer among those with AF (adjusted hazard ratio (HR) = 1.19, 95% confidence interval (CI): 1.03, 1.38). Additional adjustment for baseline use of cardiac glycosides attenuated the association between AF and invasive breast cancer (HR = 1.01, 95% CI: 0.85, 1.20). Cardiac glycoside use was strongly associated with incident invasive breast cancer (HR = 1.68, 95% CI: 1.33, 2.12) independent of AF and other confounders. Mechanisms of the associations among breast cancer, AF, and cardiac glycosides need further investigation.
Keywords: atrial fibrillation, breast cancer, colorectal cancer, cardiac glycosides, digoxin
Atrial fibrillation (AF) is the most common clinically significant cardiac arrhythmia. It occurs in approximately 1% of the general population, and the prevalence increases substantially with age to approximately 7% in persons 60–69 years of age and 13% in persons older than 80 years (1). AF is a significant risk factor for stroke and heart failure. Primarily associated with cardiovascular conditions such as hypertension, coronary artery disease, myocardial infarction, hearth failure, and valvular disease, AF shares some characteristics and co-exists with many noncardiac conditions (2). Very few data currently exist on possible relationships between AF and malignant diseases.
There have been 2 studies thus far in which investigators have reported a higher risk of AF diagnosis in cancer patients. In 1994, Müller et al. (3) published a case-control study of nonsteroidal antiinflammatory drug treatment among 12,304 veterans with a primary diagnosis of colon cancer; they described as a secondary finding that AF and atrial flutter were associated with an increased occurrence of colon cancer after 5–10 years. More recently, in 2008, Guzzetti et al. (4) reported an at least 2- to 3-fold higher prevalence of AF among subjects with either colorectal or breast cancer cases than among controls who did not have cancer. Several mechanisms for this association have been postulated. For example, inflammation is thought to play a role because of the observed elevation of inflammatory markers in AF and carcinogenesis (5–7). There have been no prospective studies of the relationships of AF with incident cancers.
The Women's Health Initiative (WHI) Observational Study, an ongoing prospective study of 93,676 postmenopausal women with long-term follow-up, addresses the roles of biologic and lifestyle factors in the common causes of morbidity, mortality, and impaired quality of life in postmenopausal women (8). This study provides an excellent opportunity to examine the prospective associations of AF with breast cancer and colorectal cancer after controlling for potential confounders.
METHODS
The WHI Observational Study, which was sponsored by National Heart, Lung, and Blood Institute, enrolled 93,676 women 50–79 years of age in 40 centers throughout the United States during the years 1994–1998. The study design and baseline characteristics of participants have been described in detail elsewhere (8). In brief, participants were recruited through mass mailings to voter registration, motor vehicle registration, and commercial lists or were women who did not wish to join or were not eligible for the WHI clinical trials of hormone therapy or dietary modification. Participants completed multiple questionnaires about their physical and mental health and comorbid conditions and had a baseline clinic visit during which they had physical measurements (weight, height, waist-hip measurements, blood pressure) taken and a fasting blood draw. Participants brought in their medications in original pill bottles, and the labels were scanned and entered into a medications database. Three years after the baseline visit, women had another clinic visit, at which time the same measurements were obtained and questionnaires administered. All research activities were approved by the institutional review boards of all involved institutions, and all participants in the WHI provided written informed consent.
Outcomes ascertainment
Annual follow-up was conducted using mailed questionnaires and telephone calls to determine hospitalizations and potential outcome events. When a hospitalization occurred, medical records were obtained from the hospital and outside providers. Outcomes packets were then prepared for adjudication of events by local study physicians and subsequently sent to the WHI coordinating center in Seattle, Washington, for central adjudication and coding of breast cancer stage, size, nodal status, grade, histology, and estrogen receptor and progesterone receptor status. Of the 93,676 women in the observational study, 471 (0.5%) had missing data on follow-up time and were excluded from these analyses. Incident breast cancer was defined as no history of any type of breast cancer at baseline, no evidence of in situ breast cancer on follow-up, and a diagnosis of breast cancer (invasive) during follow-up. Similarly, incident colorectal cancer was defined as no history of colorectal cancer at baseline and a diagnosis of colorectal cancer during follow-up.
Definition of AF
In the WHI Observational Study, information on AF was ascertained from annual self-reports of AF diagnosis. One problem in research on AF is that because it may be paroxysmal or intermittent, relying on diagnoses from electrocardiograms (ECGs) results in underestimation of the prevalence. Risks of cerebrovascular events associated with AF are similar for those with persistent or intermittent AF, and many patients who initially present with paroxysmal AF often progress to persistent or more recurrent AF (9–11). Further, based on the results of the Reasons for Geographic and Racial Differences in Stroke (REGARDS) Study (12), self-reported AF is a strong predictor of stroke and can be used interchangeably with an ECG diagnosis of AF in models predicting risk. The sensitivity of ECG diagnosis is lower than that for self-report, but the specificity is higher. Thus, using self-reported histories of physician-diagnosed AF might lead to better estimates of prevalence than using results from individual ECGs. Although only self-reported data are available in the WHI Observational Study, we can draw inferences about the sensitivity of self-report versus ECG diagnosis from the clinical trial of the WHI, in which there were both ECG and self-reported data on AF. In the clinical trial, the prevalence of AF on baseline ECG was 0.3% (184 of 66,777 women in the clinical trial with nonmissing ECGs). Among those who had an ECG diagnosis of AF, 80.2% self-reported a physician's diagnosis compared with 3.7% of those who did not have AF on baseline ECG, indicating that the vast majority of women with ECG-documented AF were aware of it. Among 2,589 women who self-reported a diagnosis of AF in the clinical trial, 146 (5.6%) also had a baseline ECG diagnosis of AF compared with 36 of 63,020 (0.06%) of those who did not self-report AF, indicating that a single screening ECG did not capture the large majority of women with AF. In the present study, use of the self-reported physician diagnosis of AF may tend to lead to inclusion of women without AF in the AF group, but it is not likely to include women with AF in the non-AF group, thus providing conservative estimates of relationships.
In order to further reduce potential ascertainment bias, we considered a woman to have AF if she self-reported a physician diagnosis of AF at baseline regardless of her self-reports on follow-up questionnaires. We considered a woman to be in the no-AF group if she did not report AF at baseline and did not report it at any of the follow-up visits/questionnaires. Those with missing responses to the AF question (n = 1,531) and those who did not report AF at baseline but did report it in 1 of the follow-up visits (n = 5,628) were excluded from all analyses, resulting in a total sample size of 86,046 women with valid AF data.
Statistical analysis
Comparisons of baseline characteristics between women with and without AF, between with incident invasive breast cancer and those with no breast cancer, and between women with incident colorectal cancer and those with no colorectal cancer were done using χ2 tests for categorical variables and Student's t test for continuous variables. Cox proportional hazards models were run to obtain hazard ratios and 95% confidence intervals for the associations of self-reported AF with incident invasive breast cancer and colorectal cancer. Variables that were associated with both AF prevalence and incident invasive breast cancer were considered potential confounders in breast cancer regression analyses and were used as adjustment factors. Similarly, variables associated with both AF prevalence and incident colorectal cancer were used as adjustment variables in analyses pertaining to colorectal cancer. For the outcome of invasive breast cancer, parity and age at first birth were both potential confounders; however, because of collinearity, only one could be retained in multivariable analyses. Serial adjustment for either parity or age at first birth resulted in similar estimates of risk, so we report models adjusted for parity because parity data were missing for fewer participants. Patients with a history of breast cancer at baseline or a missing response to that question (n = 5,400), as well as those who were diagnosed with in situ breast cancer (n = 1,016) or whose type of breast cancer was unknown (n = 43), were excluded from the breast cancer analyses, and patients with a history of colorectal cancer or missing data at baseline (n = 1,338) were excluded from the colorectal cancer analyses. For both outcomes separately, we ran Cox regression models that were unadjusted, adjusted for age and race, and adjusted for age, race, and cancer-specific potential confounders. Missing data were minimal (<1.8%) for all adjustment variables except for age at menopause and income, which were missing on approximately 7%–10% of the sample. However, estimates from unadjusted and minimally adjusted models were virtually identical regardless of whether complete case analysis for the fullest model was applied or not; therefore, we report results from all models restricted to participants with complete covariate data based on the fullest adjustment to allow for direct comparison of estimates from unadjusted to adjusted models. We also conducted several sensitivity analyses in which we excluded women who self-reported congestive heart failure at baseline (for both invasive breast cancer and colorectal cancer); evaluated AF and cardiac glycoside use as time-varying variables in models in which invasive breast cancer was the outcome; and stratified models of invasive breast cancer by estrogen receptor status. In the time-varying analyses, we incorporated information about self-reported AF and cardiac glycoside use at baseline and at year 3. There was no evidence of violations of the assumption of proportional hazards in any of the regression models reported. All reported P values are 2-sided, and P < 0.05 was considered statistically significant. Statistical analyses were conducted with STATA, version 14 (StataCorp LP, College Station, Texas).
RESULTS
The prevalence of self-reported physician diagnosis of AF at baseline in the analytic sample was 5.1% (4,376 of 86,046). The incidence of invasive breast cancer during follow-up was 5.7% (4,497 of 79,587). Median follow-up time for breast cancer analyses was 15.3 years (interquartile range, 8.1–17.1). The incidence of colorectal cancer was 1.6% (1,373 of 84,708), with a median follow-up time of 15.9 years (interquartile range, 8.5–17.4). Tables 1–3 compare baseline characteristics of participants with AF, breast cancer, or colorectal cancer with characteristics of those without these conditions. Because of the large numbers, many (but not all) baseline variables were statistically significantly different among the AF, breast cancer, and colorectal cancer groups.
Table 2.
Characteristic | Incident Invasive Breast Cancer (n = 4,497) | No Incident Invasive Breast Cancer (n = 75,090) | P Value | ||||
---|---|---|---|---|---|---|---|
Mean (SD) | No.a | % | Mean (SD) | No.a | % | ||
Age, years | 63.3 (7.1) | 63.3 (7.4) | 0.47 | ||||
BMIb | 27.3 (5.7) | 27.2 (5.9) | 0.27 | ||||
Age at menopause, years | 48.4 (6.6) | 47.4 (6.7) | <0.001 | ||||
Resting pulse (30 seconds) | 34.8 (6.1) | 34.6 (6.0) | 0.24 | ||||
Ethnicity | <0.001 | ||||||
Non-Hispanic white | 4,015 | 89.5 | 62,076 | 82.9 | |||
Black | 226 | 5.0 | 6,247 | 8.3 | |||
Hispanic/Latino | 102 | 2.3 | 3,023 | 4.0 | |||
Other | 144 | 3.2 | 3,534 | 4.7 | |||
Educational level | <0.001 | ||||||
<College degree | 2,309 | 51.7 | 43,422 | 58.3 | |||
≥College degree | 2,161 | 48.3 | 31,050 | 41.7 | |||
Income | <0.001 | ||||||
<$20,000 | 514 | 12.3 | 11,191 | 16.1 | |||
$20,000–$49,999 | 1,775 | 42.4 | 30,067 | 43.2 | |||
$50,000–$74,999 | 925 | 22.1 | 14,064 | 20.2 | |||
≥$75,000 | 976 | 23.3 | 14,264 | 20.5 | |||
Marital status | 0.001 | ||||||
Never married | 231 | 5.2 | 3,417 | 4.6 | |||
Divorced/separated | 667 | 14.9 | 11,948 | 16.0 | |||
Widowed | 691 | 15.4 | 12,762 | 17.1 | |||
Married/living such | 2,891 | 64.5 | 46,595 | 62.4 | |||
Smoking status | <0.001 | ||||||
Never smoker | 2,141 | 48.1 | 38,080 | 51.4 | |||
Past smoker | 2,035 | 45.7 | 31,344 | 42.3 | |||
Current smoker | 274 | 6.2 | 4,735 | 6.4 | |||
Physical activity level | 0.03 | ||||||
No activity | 552 | 12.4 | 10,095 | 13.6 | |||
Mild | 1,663 | 37.4 | 28,519 | 38.3 | |||
Moderate | 864 | 19.4 | 13,697 | 18.4 | |||
Strenuous | 1,372 | 30.8 | 22,086 | 29.7 | |||
Age at menarche, years | 0.11 | ||||||
≤10 | 323 | 7.2 | 4,880 | 6.5 | |||
11–13 | 3,189 | 71.1 | 52,889 | 70.6 | |||
14–15 | 796 | 17.8 | 14,088 | 18.8 | |||
≥16 | 175 | 3.9 | 3,057 | 4.1 | |||
Age at first birth, years | <0.001 | ||||||
No term pregnancy | 657 | 15.8 | 9,349 | 13.8 | |||
<20 | 413 | 9.9 | 8,651 | 12.8 | |||
20–29 | 2,672 | 64.3 | 43,991 | 65.1 | |||
≥30 | 411 | 9.9 | 5,606 | 8.3 | |||
Parity | <0.001 | ||||||
No term pregnancy | 657 | 14.7 | 9,349 | 12.5 | |||
1 | 415 | 9.3 | 6,794 | 9.1 | |||
2–3 | 2,313 | 51.7 | 37,718 | 50.5 | |||
≥4 | 1,088 | 24.3 | 20,819 | 27.9 | |||
Hormone therapy use | <0.001 | ||||||
Never | 1,541 | 34.3 | 29,883 | 39.8 | |||
Past | 565 | 12.6 | 10,349 | 13.8 | |||
Current | 2,384 | 53.1 | 34,792 | 46.4 | |||
Hysterectomy | <0.001 | ||||||
No | 2,829 | 63.0 | 43,844 | 58.4 | |||
Yes | 1,664 | 37.0 | 31,176 | 41.6 | |||
Hypertension | 0.31 | ||||||
Nonhypertensive | 3,022 | 68.1 | 49,860 | 67.3 | |||
Untreated | 323 | 7.3 | 5,840 | 7.9 | |||
Treated | 1,093 | 24.6 | 18,353 | 24.8 | |||
Diabetes (treated) | <0.001 | ||||||
No | 4,359 | 97.0 | 71,881 | 95.9 | |||
Yes | 135 | 3.0 | 3,113 | 4.2 | |||
High cholesterol requiring medication | 0.15 | ||||||
No | 3,791 | 86.0 | 62,877 | 85.3 | |||
Yes | 615 | 14.0 | 10,874 | 14.7 | |||
History of CVD | <0.02 | ||||||
No | 4,130 | 91.8 | 68,169 | 90.8 | |||
Yes | 367 | 8.2 | 6,921 | 9.2 | |||
History of CHF | 0.64 | ||||||
No | 4,455 | 99.1 | 74,435 | 99.1 | |||
Yes | 42 | 0.9 | 651 | 0.9 | |||
Warfarin use | 0.22 | ||||||
No | 4,438 | 98.7 | 74,254 | 98.9 | |||
Yes | 59 | 1.3 | 835 | 1.1 | |||
Anti-arrhythmic use | 0.29 | ||||||
No | 4,474 | 99.5 | 74,783 | 99.6 | |||
Yes | 23 | 0.5 | 306 | 0.4 | |||
Aspirin use | 0.18 | ||||||
No | 3,428 | 76.2 | 57,897 | 77.1 | |||
Yes | 1,069 | 23.8 | 17,192 | 22.9 | |||
Statin use | 0.25 | ||||||
No | 4,157 | 92.4 | 69,054 | 92.0 | |||
Yes | 340 | 7.6 | 6,035 | 8.0 | |||
β-blocker use | 0.35 | ||||||
No | 4,107 | 91.3 | 68,876 | 91.7 | |||
Yes | 390 | 8.7 | 6,214 | 8.3 | |||
Calcium channel blocker use | 0.95 | ||||||
No | 4,065 | 90.4 | 67,853 | 90.4 | |||
Yes | 432 | 9.6 | 7,237 | 9.6 | |||
ACE inhibitor use | 0.44 | ||||||
No | 4,159 | 92.5 | 69,207 | 92.2 | |||
Yes | 338 | 7.5 | 5,883 | 7.8 | |||
Angiotensin II receptor blocker use | 0.80 | ||||||
No | 4,465 | 99.3 | 74,530 | 99.3 | |||
Yes | 32 | 0.7 | 560 | 0.7 | |||
Cardiac glycoside use | <0.001 | ||||||
No | 4,372 | 97.2 | 76,654 | 98.1 | |||
Yes | 125 | 2.8 | 1,435 | 1.9 |
Abbreviations: ACE, angiotensin-converting enzyme; BMI, body mass index; CHF, congestive heart failure; CVD, cardiovascular disease; SD, standard deviation.
a Subgroup totals may not sum to the column total because of missing data.
b Weight (kg)/height (m)2.
Table 1.
Characteristic | Atrial Fibrillation at Baseline (n = 4,376) | No Atrial Fibrillation at Baseline (n = 81,670) | P Value | ||||
---|---|---|---|---|---|---|---|
Mean (SD) | No.a | % | Mean | No.a | % | ||
Age, years | 66.9 (7.1) | 63.2 (7.3) | <0.001 | ||||
BMIb | 27.4 (6.0) | 27.2 (5.8) | 0.04 | ||||
Age at menopause, years | 47.1 (7.1) | 47.5 (6.6) | <0.01 | ||||
Resting pulse (30 seconds) | 34.1 (6.3) | 34.7 (6.1) | <0.001 | ||||
Ethnicity | <0.001 | ||||||
Non-Hispanic white | 3,687 | 84.5 | 67,895 | 83.4 | |||
Black | 393 | 9.0 | 6,658 | 8.2 | |||
Hispanic/Latino | 115 | 2.6 | 3,195 | 3.9 | |||
Other | 170 | 3.9 | 3,688 | 4.5 | |||
Educational level | <0.001 | ||||||
<College degree | 2,852 | 65.6 | 46,448 | 57.3 | |||
≥College degree | 1,498 | 34.4 | 34,552 | 42.7 | |||
Income | <0.001 | ||||||
<$20,000 | 949 | 23.6 | 11,687 | 15.4 | |||
$2,000–$49,999 | 1,867 | 46.4 | 32,580 | 43.0 | |||
$50,000–$74,999 | 658 | 16.4 | 15,564 | 20.6 | |||
≥$75,000 | 551 | 13.7 | 15,924 | 21.0 | |||
Marital status | <0.001 | ||||||
Never married | 205 | 4.7 | 3,828 | 4.7 | |||
Divorced/separated | 649 | 14.9 | 12,919 | 15.9 | |||
Widowed | 1,027 | 23.6 | 13,667 | 16.8 | |||
Married/living such | 2,476 | 56.8 | 50,858 | 62.6 | |||
Smoking status | 0.19 | ||||||
Never smoker | 2,224 | 51.7 | 41,135 | 51.0 | |||
Past smoker | 1,837 | 42.7 | 34,417 | 42.7 | |||
Current smoker | 244 | 5.7 | 5,116 | 6.3 | |||
Physical activity level | <0.001 | ||||||
No activity | 690 | 16.0 | 10,813 | 13.4 | |||
Mild | 1,795 | 41.7 | 30,858 | 38.1 | |||
Moderate | 777 | 18.0 | 15,005 | 18.6 | |||
Strenuous | 1,048 | 24.3 | 24,230 | 30.0 | |||
Age at menarche, years | 0.02 | ||||||
≤10 | 296 | 6.8 | 5,322 | 6.5 | |||
11–13 | 2,991 | 68.6 | 57,656 | 70.8 | |||
14–15 | 886 | 20.3 | 15,205 | 18.7 | |||
≥16 | 188 | 4.6 | 3,282 | 4.0 | |||
Age at first birth, years | 0.04 | ||||||
No term pregnancy | 524 | 13.3 | 10,476 | 14.2 | |||
<20 | 538 | 13.7 | 9,121 | 12.4 | |||
20–29 | 2,516 | 64.0 | 47,758 | 64.9 | |||
≥30 | 352 | 9.0 | 6,278 | 8.5 | |||
Parity | <0.001 | ||||||
No term pregnancy | 524 | 12.0 | 10,476 | 12.9 | |||
1 | 378 | 8.7 | 7,477 | 9.2 | |||
2–3 | 2,114 | 48.6 | 41,146 | 50.7 | |||
≥4 | 1,337 | 30.7 | 22,121 | 27.2 | |||
Hormone therapy use | <0.001 | ||||||
Never | 1,755 | 40.1 | 33,167 | 40.7 | |||
Past | 824 | 18.9 | 11,838 | 14.5 | |||
Current | 1,793 | 41.0 | 36,590 | 44.8 | |||
Hysterectomy | <0.001 | ||||||
No | 2,176 | 49.8 | 48,188 | 59.1 | |||
Yes | 2,193 | 50.2 | 33,413 | 41.0 | |||
Hypertension | <0.001 | ||||||
Nonhypertensive | 2,243 | 52.7 | 54,836 | 68.1 | |||
Untreated | 441 | 10.4 | 6,183 | 7.7 | |||
Treated | 1,570 | 36.9 | 19,533 | 24.3 | |||
Diabetes (treated) | <0.001 | ||||||
No | 4,045 | 92.6 | 78,370 | 96.1 | |||
Yes | 325 | 7.4 | 3,199 | 3.9 | |||
High cholesterol requiring medication | |||||||
No | 3,366 | 78.8 | 68,583 | 85.5 | |||
Yes | 905 | 21.2 | 11,612 | 14.5 | |||
History of CVD | <0.001 | ||||||
No | 3,029 | 69.2 | 75,065 | 91.9 | |||
Yes | 1,347 | 30.8 | 6,605 | 8.1 | |||
History of CHF | <0.001 | ||||||
No | 4,104 | 93.8 | 81,179 | 99.4 | |||
Yes | 272 | 6.2 | 486 | 0.6 | |||
Warfarin use | <0.001 | ||||||
No | 3,785 | 86.5 | 81,269 | 99.5 | |||
Yes | 591 | 13.5 | 400 | 0.5 | |||
Anti-arrhythmic use | <0.001 | ||||||
No | 4,091 | 93.5 | 81,594 | 99.9 | |||
Yes | 285 | 6.5 | 75 | 0.1 | |||
Aspirin use | <0.001 | ||||||
No | 2,923 | 66.8 | 63,358 | 77.6 | |||
Yes | 1,453 | 33.2 | 18,311 | 22.4 | |||
Statin use | <0.01 | ||||||
No | 3,859 | 88.2 | 75,212 | 92.1 | |||
Yes | 517 | 11.8 | 6,457 | 7.9 | |||
β-blocker use | <0.001 | ||||||
No | 3,335 | 76.2 | 75,567 | 92.5 | |||
Yes | 1,041 | 23.8 | 6,103 | 7.5 | |||
Calcium channel blocker use | |||||||
No | 3,425 | 78.3 | 74,285 | 91.0 | |||
Yes | 951 | 21.7 | 7,385 | 9.0 | |||
ACE inhibitor use | <0.001 | ||||||
No | 3,804 | 86.9 | 75,468 | 92.4 | |||
Yes | 572 | 13.1 | 6,202 | 7.6 | |||
Angiotensin II receptor blocker use | <0.001 | ||||||
No | 4,314 | 98.6 | 81,090 | 99.3 | |||
Yes | 62 | 1.4 | 580 | 0.7 | |||
Cardiac glycoside use | <0.001 | ||||||
No | 3,137 | 71.7 | 81,175 | 99.4 | |||
Yes | 1,239 | 28.3 | 494 | 0.6 |
Abbreviations: ACE, angiotensin-converting enzyme; BMI, body mass index; CHF, congestive heart failure; CVD, cardiovascular disease; SD, standard deviation.
a Subgroup totals may not sum to the column total because of missing data.
b Weight (kg)/height (m)2.
Table 3.
Characteristic | Incident Colorectal Cancer (n = 1,373) | No Incident Colorectal Cancer(n = 83,335) | P Value | ||||
---|---|---|---|---|---|---|---|
Mean (SD) | No.a | % | Mean (SD) | No.a | % | ||
Age, years | 66.1 (7.0) | 63.3 (7.3) | <0.001 | ||||
BMIb | 27.9 (5.8) | 27.2 (5.8) | <0.001 | ||||
Age at menopause, years | 47.8 (6.7) | 47.4 (6.6) | 0.06 | ||||
Resting pulse (30 seconds) | 35.1 (6.6) | 34.7 (6.1) | 0.02 | ||||
Ethnicity | <0.01 | ||||||
Non-Hispanic white | 1,171 | 85.4 | 69,332 | 83.4 | |||
Black | 122 | 8.9 | 6,786 | 8.2 | |||
Hispanic/Latino | 32 | 2.3 | 3,220 | 3.9 | |||
Other | 46 | 3.4 | 3,760 | 4.5 | |||
Educational level | 0.07 | ||||||
<College degree | 820 | 60.0 | 47,633 | 57.6 | |||
≥College degree | 546 | 40.0 | 35,029 | 42.4 | |||
Income | <0.001 | ||||||
<$20,000 | 235 | 18.3 | 12,126 | 15.7 | |||
$20,000–$49,999 | 623 | 48.5 | 33,275 | 43.1 | |||
$50,000–$74,999 | 229 | 17.8 | 15,765 | 20.4 | |||
≥$75,000 | 197 | 15.3 | 16,119 | 20.9 | |||
Marital status | <0.001 | ||||||
Never married | 69 | 5.0 | 3,894 | 4.7 | |||
Divorced/separated | 216 | 15.8 | 13,140 | 15.8 | |||
Widowed | 297 | 21.7 | 14,082 | 17.0 | |||
Married/living such | 787 | 57.5 | 51,817 | 62.5 | |||
Smoking status | 0.04 | ||||||
Never smoker | 652 | 48.1 | 42,102 | 51.2 | |||
Past smoker | 602 | 44.4 | 35,037 | 42.6 | |||
Current smoker | 101 | 7.5 | 5,173 | 6.3 | |||
Physical activity level | <0.01 | ||||||
No activity | 219 | 16.1 | 11,095 | 13.4 | |||
Mild | 540 | 39.7 | 31,553 | 38.2 | |||
Moderate | 250 | 18.4 | 15,311 | 18.5 | |||
Strenuous | 353 | 25.9 | 24,628 | 29.8 | |||
Age at menarche, years | 0.19 | ||||||
≤10 | 109 | 8.0 | 5,432 | 6.5 | |||
11–13 | 946 | 69.0 | 58,768 | 70.7 | |||
14–15 | 261 | 19.0 | 15,570 | 18.7 | |||
≥16 | 55 | 4.0 | 3,366 | 4.1 | |||
Age at first birth, years | 0.82 | ||||||
No term pregnancy | 178 | 14.3 | 10,634 | 14.2 | |||
<20 | 153 | 12.3 | 9,351 | 12.4 | |||
20–29 | 797 | 64.1 | 48,768 | 64.9 | |||
≥30 | 115 | 9.3 | 6,403 | 8.5 | |||
Parity | 0.03 | ||||||
No term pregnancy | 178 | 13.0 | 10,634 | 12.8 | |||
1 | 129 | 9.4 | 7,589 | 9.2 | |||
2–3 | 642 | 47.0 | 42,008 | 50.7 | |||
≥4 | 418 | 30.6 | 22,660 | 27.3 | |||
Hormone therapy use | <0.001 | ||||||
Never | 674 | 49.2 | 33,616 | 40.4 | |||
Past | 226 | 16.5 | 12,217 | 14.7 | |||
Current | 471 | 34.4 | 37,428 | 45.0 | |||
Hysterectomy | 0.52 | ||||||
No | 794 | 57.8 | 48,869 | 58.7 | |||
Yes | 579 | 42.2 | 34,390 | 41.3 | |||
Hypertension | 0.01 | ||||||
Nonhypertensive | 862 | 63.7 | 55,459 | 67.5 | |||
Untreated | 111 | 8.2 | 6,400 | 7.8 | |||
Treated | 381 | 28.1 | 20,355 | 24.8 | |||
Diabetes (treated) | 0.01 | ||||||
No | 1,298 | 94.6 | 79,857 | 96.0 | |||
Yes | 74 | 5.4 | 3,374 | 4.1 | |||
High cholesterol requiring medication | 0.08 | ||||||
No | 1,130 | 83.6 | 69,799 | 85.3 | |||
Yes | 222 | 16.4 | 12,068 | 14.7 | |||
History of CVD | 0.01 | ||||||
No | 1,220 | 88.9 | 75,173 | 90.9 | |||
Yes | 153 | 11.1 | 7,622 | 9.2 | |||
History of CHF | 0.58 | ||||||
No | 1,363 | 99.3 | 82,606 | 99.1 | |||
Yes | 10 | 0.7 | 724 | 0.9 | |||
Warfarin use | 0.58 | ||||||
No | 1,355 | 98.7 | 82,376 | 98.9 | |||
Yes | 18 | 1.3 | 958 | 1.2 | |||
Anti-arrhythmic use | 0.48 | ||||||
No | 1,369 | 99.7 | 82,989 | 99.6 | |||
Yes | 4 | 0.3 | 345 | 0.4 | |||
Aspirin use | 0.31 | ||||||
No | 1,074 | 78.2 | 64,221 | 77.1 | |||
Yes | 299 | 21.8 | 19,113 | 22.9 | |||
Statin use | 0.83 | ||||||
No | 1,264 | 92.1 | 76,589 | 91.9 | |||
Yes | 109 | 7.9 | 6,745 | 8.1 | |||
β-blocker use | 0.83 | ||||||
No | 1,257 | 91.6 | 76,427 | 91.7 | |||
Yes | 116 | 8.5 | 6,908 | 8.3 | |||
Calcium channel blocker use | 0.33 | ||||||
No | 1,230 | 89.6 | 75,306 | 90.4 | |||
Yes | 143 | 10.4 | 8,029 | 9.6 | |||
ACE inhibitor use | 0.98 | ||||||
No | 1,265 | 92.1 | 76,794 | 92.2 | |||
Yes | 108 | 7.9 | 6,541 | 7.9 | |||
Angiotensin II receptor blocker use | 0.56 | ||||||
No | 1,361 | 99.1 | 82,720 | 99.3 | |||
Yes | 12 | 0.9 | 615 | 0.7 | |||
Cardiac glycoside use | 0.40 | ||||||
No | 1,341 | 97.7 | 81,662 | 98.0 | |||
Yes | 32 | 2.3 | 1,672 | 2.0 |
Abbreviations: ACE, angiotensin-converting enzyme; BMI, body mass index; CHF, congestive heart failure; CVD, cardiovascular disease; SD, standard deviation.
a Subgroup totals may not sum to the column total because of missing data.
b Weight (kg)/height (m)2.
Women with AF at baseline had a significantly higher risk of invasive breast cancer (hazard ratio (HR) = 1.23, 95% confidence interval (CI): 1.06, 1.42) in unadjusted analyses (Table 4). The excess risk remained after adjustment for age and race (HR = 1.19, 95% CI: 1.03, 1.37) and further adjustment for educational level, income, marital status, physical activity level, parity, age at menopause, hormone therapy use, hysterectomy, diabetes, and history of cardiovascular disease (myocardial infarction, stroke, transient ischemic attack, angina, or revascularization) (HR = 1.19, 95% CI: 1.03, 1.38). Use of warfarin, anti-arrhythmic drugs, β-blockers, aspirin, statins, calcium channel blockers, or angiotensin II receptor blockers was not related to breast cancer; however, baseline use of cardiac glycosides was associated with a 68% higher risk of invasive breast cancer (HR = 1.68, 95% CI: 1.33, 2.12) after adjustment for other potential confounders. When we considered only women with estrogen receptor–positive tumors (n = 2,994), the hazard ratio for cardiac glycosides was 1.75 (95% CI: 1.34, 2.27). The hazard ratio among women with estrogen receptor–negative tumors (n = 511) was not significant (HR = 1.33, 95% CI: 0.68, 2.62). There were 231 women in our analytic cohort with invasive breast cancer who had missing assays. When we included cardiac glycosides in the Cox model, the association of AF with breast cancer was attenuated to a hazard ratio of 1.01 (95% CI: 0.85, 1.20). There was no significant association of baseline AF with colorectal cancer in either unadjusted or adjusted analyses. Cardiac glycoside use was not associated with the risk colorectal cancer (adjusted HR = 1.08, 95% CI: 0.71, 1.64).
Table 4.
Adjustment | Incident Invasive Breast Cancer (n = 65,352)a | Incident Colorectal Cancer (n = 76,252)b | ||||
---|---|---|---|---|---|---|
HR | 95% CI | P Value | HR | 95% CI | P Value | |
None | 1.23 | 1.06, 1.42 | <0.01 | 1.17 | 0.91, 1.50 | 0.22 |
Age and race | 1.19 | 1.03, 1.37 | 0.02 | 0.94 | 0.73, 1.21 | 0.66 |
Age, race, and potential confoundersc | 1.19 | 1.03, 1.38 | 0.02 | 0.91 | 0.71, 1.18 | 0.49 |
Variables above and cardiac glycoside use | 1.01 | 0.85, 1.20 | 0.89 | 0.89 | 0.67, 1.19 | 0.43 |
Abbreviations: CI, confidence interval; HR, hazard ratio.
a There were 3,736 events over a median follow-up time of 15.3 years.
b There were 1,251 events over a median follow-up time of 15.9 years.
c In the invasive breast cancer model, potential confounders were those baseline characteristics that were associated with both atrial fibrillation prevalence and incident breast cancer: educational level, income, marital status, physical activity level, parity, age at menopause, hormone therapy use, hysterectomy, diabetes, and history of cardiovascular disease (myocardial infarction, stroke, transient ischemic attack, angina, or revascularization). In the colorectal cancer model, potential confounders were baseline characteristics that were associated with both atrial fibrillation prevalence and incident colorectal cancer: income, marital status, physical activity level, parity, hormone therapy use, hypertension, diabetes, resting pulse rate, and history of cardiovascular disease (myocardial infarction, stroke, transient ischemic attack, angina or revascularization).
We further investigated the association of AF with breast cancer in relation to cardiac glycoside use by categorizing women into 4 mutually exclusive groups based on their self-reported baseline AF status and self-reported cardiac glycoside medication use at baseline. Compared with women who did not have AF and were not taking cardiac glycosides, women with AF who were not taking cardiac glycosides did not have a higher risk of breast cancer (HR = 1.01, 95% CI: 0.85, 1.20) in adjusted analyses (Table 5). Women taking cardiac glycosides at baseline, regardless of whether they reported AF or no AF at baseline, had a significantly increased risk of breast cancer on follow-up (for AF, HR = 1.70, 95% CI: 1.35, 2.13; for no AF, HR = 1.69, 95% CI: 1.13, 2.50.
Table 5.
Mutually Exclusive Categories of Atrial Fibrillation and Cardiac Glycoside | Invasive Breast Cancerb | Colorectal Cancerc | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Total Sample Size | No. of Events | HR | 95% CI | P Value | Total Sample Size | No. Events | HR | 95% CI | P Value | |
No atrial fibrillation, no cardiac glycosides | 61,794 | 3,513 | 1.00 | Referent | 72,062 | 1,178 | 1.00 | Referent | ||
Atrial fibrillation, no cardiac glycosides | 2,281 | 121 | 1.01 | 0.85, 1.20 | 0.89 | 2,693 | 44 | 0.91 | 0.67, 1.23 | 0.54 |
Atrial fibrillation, cardiac glycosides | 932 | 77 | 1.70 | 1.35, 2.13 | <0.001 | 1,088 | 20 | 0.93 | 0.60, 1.46 | 0.76 |
No atrial fibrillation, cardiac glycosides | 345 | 25 | 1.69 | 1.13, 2.50 | 0.01 | 409 | 9 | 1.18 | 0.61, 2.29 | 0.62 |
Abbreviations: CI, confidence interval; HR, hazard ratio.
a The invasive breast cancer model was adjusted for baseline characteristics that were associated with both atrial fibrillation prevalence and incident breast cancer: age, race, educational level, income, marital status, physical activity level, parity, age at menopause, hormone therapy use, hysterectomy, diabetes, and history of cardiovascular disease (myocardial infarction, stroke, transient ischemic attack, angina, or revascularization). The colorectal cancer model was adjusted for baseline characteristics that were associated with both atrial fibrillation prevalence and incident colorectal cancer: age, race, income, marital status, physical activity level, parity, hormone therapy use, hypertension, diabetes, resting pulse rate, and history of cardiovascular disease (myocardial infarction, stroke, transient ischemic attack, angina, or revascularization).
b The median follow-up time was 15.3 years.
c The median follow-up time was 15.9 years.
Similar results were found for both outcomes in sensitivity analysis in which we restricted the sample to those without prevalent congestive heart failure (n = 663 for breast cancer analyses and n = 779 for colorectal cancer analyses; data not shown). AF was associated with an increased risk of breast cancer that was attenuated after adjustment for cardiac glycoside use, and no significant difference in risk was seen for colorectal cancer in adjusted or unadjusted analyses. In analyses of invasive breast cancer in which both AF and cardiac glycoside use were treated as time-varying variables, the results were similar. Findings for women with estrogen receptor–positive tumors (n = 2,994) were similar to those for women with all types of invasive breast cancer. There were only 511 women with estrogen receptor–negative tumors, so the nonsignificant findings in that group may reflect low power. The remainder of the breast cancer case patients had missing assays.
DISCUSSION
In a large prospective study of postmenopausal women 50–79 years of age, we found that AF in older women was associated with a 19% higher risk of incident breast cancer after adjustment for multiple variables; however, this association was explained by use of cardiac glycosides at baseline. In our study, there was no relationship between AF and colorectal cancer incidence.
Studies of AF can be challenging because AF can be paroxysmal (self-terminating), persistent (sustained greater than 7 days), or permanent (typically greater than 1 year and when cardioversion has failed or is foregone). For the purposes of our study, we considered self-reported clinical diagnoses of AF to provide a more relevant estimate of the true prevalence of all types of AF than a 1-time ECG diagnosis. In a retrospective case-control study by Guzzetti et al. (4) in which AF was determined using a presurgical ECG, the authors observed a prevalence of AF that was at least 2 times higher in both patients with breast cancer and those with colorectal cancer than in control subjects, which raises the interesting question of whether AF precedes or follows cancers. Analyses of our data suggest that there is no association of incidence of breast cancer or colorectal cancer following a baseline diagnosis of AF after controlling for important confounders.
Although our results indicated a higher risk of breast cancer in women with prevalent AF even after adjustment for age, race, and cardiovascular comorbid conditions or risk factors for heart diseases, further adjustment for use of cardiac glycosides resulted in an attenuation of that risk, indicating that cardiac glycoside use is a partial mediator because cardiac glycosides may be prescribed for AF. However, it is also a confounder because not all AF is treated with glycosides. In our sample, 28% of those with AF at baseline were taking cardiac glycosides compared with 0.6% of those without AF. Thus, the excess risk of breast cancer observed with AF is explained by use of cardiac glycosides. The significant association of cardiac glycosides with incident breast cancer in our cohort is consistent with other research (13–15). However, it is also possible that other factors associated with use of cardiac glycosides result in a bias by indication.
Our results also differ from those from a study by Müller et al. (3) in which the authors showed a positive association of AF and atrial flutter with increased occurrence of colon cancer after 5–10 years. The authors examined discharge International Classification of Diseases codes to determine the frequency of individual diseases preceding colon cancer in the patient population. Although our sample population had a lower number of incident cases of colorectal cancer (n = 1,373) than did the population in the study by Müller et al. (n = 12,304), the latter study was a retrospective review of patient treatment files and discharge diagnoses. The authors indicated that although AF and atrial flutter did not represent any known risk factor, the usage of anticoagulants, such as warfarin or heparin, for the treatment of these arrhythmias may lead to chronic gastrointestinal bleeding and may be a possible mechanism for the association.
In conclusion, cardiac glycoside use was associated with a 68% higher risk of breast cancer. The association of risk of breast cancer with AF at baseline in postmenopausal women is explained by the use of cardiac glycosides. There was no association of AF with colorectal cancer incidence.
ACKNOWLEDGMENTS
Author affiliations: Department of Epidemiology and Population Health, Albert Einstein College of Medicine, Bronx, New York (Sylvia Wassertheil-Smoller, Aileen P. McGinn); Division of Cardiology, School of Medicine, George Washington University Washington, District of Columbia (Lisa Martin); Geriatric Medicine, John A. Burns School of Medicine, University of Hawaii at Manoa, Honolulu, Hawaii (Beatriz L. Rodriguez); Stanford Prevention Research Center, School of Medicine, Stanford University, Stanford, California (Marcia L. Stefanick); and Division of Cardiovascular Medicine, Stanford University, Stanford, California (Marco Perez).
The Women's Health Initiative program is funded by the National Heart, Lung, and Blood Institute of the National Institutes of Health, United States Department of Health and Human Services, through contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, HHSN268201100004C, and HHSN271201100004C.
The Women's Health Initiative Investigators are as follows. Program Office: National Heart, Lung, and Blood Institute, Bethesda, Maryland (Elizabeth Nabel, Jacques Rossouw, Shari Ludlam, Joan McGowan, Leslie Ford, and Nancy Geller). Clinical Coordinating Center: Fred Hutchinson Cancer Research Center, Seattle, Washington (Ross Prentice, Garnet Anderson, Andrea LaCroix, Charles L. Kooperberg, Ruth E. Patterson, Anne McTiernan); Medical Research Labs, Highland Heights, Kentucky (Evan Stein); and University of California at San Francisco, San Francisco, California (Steven Cummings). Clinical Centers: Albert Einstein College of Medicine, Bronx, New York (Sylvia Wassertheil-Smoller); Baylor College of Medicine, Houston, Texas (Aleksandar Rajkovic); Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts (JoAnn E. Manson); Brown University, Providence, Rhode Island (Charles B. Eaton); Emory University, Atlanta, Georgia (Lawrence Phillips); Fred Hutchinson Cancer Research Center, Seattle, Washington (Shirley Beresford); George Washington University Medical Center, Washington, District of Columbia (Lisa Martin); Los Angeles Biomedical Research Institute at Harbor, University of California at Los Angeles Medical Center, Torrance, California (Rowan Chlebowski); Kaiser Permanente Center for Health Research, Portland, Oregon (Yvonne Michael); Kaiser Permanente Division of Research, Oakland, California (Bette Caan); Medical College of Wisconsin, Milwaukee, Wisconsin (Jane Morley Kotchen); MedStar Research Institute/Howard University, Washington, DC (Barbara V. Howard); Northwestern University, Chicago/Evanston, Illinois (Linda Van Horn); Rush Medical Center, Chicago, Illinois (Henry Black); Stanford Prevention Research Center, Stanford, California (Marcia L. Stefanick); State University of New York at Stony Brook, Stony Brook, New York (Dorothy Lane); The Ohio State University, Columbus, Ohio (Rebecca Jackson); University of Alabama at Birmingham, Birmingham, Alabama (Cora E. Lewis); University of Arizona, Tucson/Phoenix, Arizona (Cynthia A Thomson); University at Buffalo, Buffalo, New York (Jean Wactawski-Wende); University of California at Davis, Sacramento, California (John Robbins); University of California at Irvine, Irvine, California (F. Allan Hubbell); University of California at Los Angeles, Los Angeles, California (Lauren Nathan); University of California at San Diego, LaJolla/Chula Vista, California (Robert D. Langer); University of Cincinnati, Cincinnati, Ohio (Margery Gass); University of Florida, Gainesville/Jacksonville, Florida (Marian Limacher); University of Hawaii, Honolulu, Hawaii (J. David Curb); University of Iowa, Iowa City/Davenport, Iowa (Robert Wallace); University of Massachusetts/Fallon Clinic, Worcester, Massachusetts (Judith Ockene); University of Medicine and Dentistry of New Jersey, Newark, New Jersey (Norman Lasser); University of Miami, Miami, Florida (Mary Jo O'Sullivan); University of Minnesota, Minneapolis, Minnesota (Karen Margolis); University of Nevada, Reno, Nevada (Robert Brunner); University of North Carolina, Chapel Hill, North Carolina (Gerardo Heiss); University of Pittsburgh, Pittsburgh, Pennsylvania (Lewis Kuller); University of Tennessee Health Science Center, Memphis, Tennessee (Karen C. Johnson); University of Texas Health Science Center, San Antonio, Texas (Robert Brzyski); University of Wisconsin, Madison, Wisconsin (Gloria E. Sarto); Wake Forest University School of Medicine, Winston-Salem, North Carolina (Mara Vitolins); Wayne State University School of Medicine/Hutzel Hospital, Detroit, Michigan (Michael Simon).
Conflict of interest: none declared.
REFERENCES
- 1. Go AS, Hylek EM, Phillips KA, et al. . Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study. JAMA. 2001;285(18):2370–2375. [DOI] [PubMed] [Google Scholar]
- 2. Lainscak M, Dagres N, Filippatos GS, et al. . Atrial fibrillation in chronic non-cardiac disease: where do we stand. Int J Cardiol. 2008;128(3):311–315. [DOI] [PubMed] [Google Scholar]
- 3. Müller AD, Sonnenberg A, Wasserman IH. Diseases preceding colon cancer. A case-control study among veterans. Dig Dis Sci. 1994;39(11):2480–2484. [DOI] [PubMed] [Google Scholar]
- 4. Guzzetti S, Costantino G, Vernocchi A, et al. . First diagnosis of colorectal or breast cancer and prevalence of atrial fibrillation. Intern Emerg Med. 2008;3(3):227–231. [DOI] [PubMed] [Google Scholar]
- 5. Guzzetti S, Costantino G, Fundarò C. Systemic inflammation, atrial fibrillation, and cancer. Circulation. 2002;106(9):e40. [DOI] [PubMed] [Google Scholar]
- 6. Coussens LM, Werb Z. Inflammation and cancer. Nature. 2002;420(6917):860–867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Chung MK, Martin DO, Sprecher D, et al. . C-reactive protein elevation in patients with atrial arrhythmias: inflammatory mechanisms and persistence of atrial fibrillation. Circulation. 2001;104(24):2886–2891. [DOI] [PubMed] [Google Scholar]
- 8. Design of the Women's Health Initiative clinical trial and observational study. The Women's Health Initiative Study Group. Control Clin Trials. 1998;19(1):61–109. [DOI] [PubMed] [Google Scholar]
- 9. Allessie MA, Boyden PA, Camm AJ, et al. . Pathophysiology and prevention of atrial fibrillation. Circulation. 2001;103(5):769–777. [DOI] [PubMed] [Google Scholar]
- 10. Rizos T, Rasch C, Jenetzky E, et al. . Detection of paroxysmal atrial fibrillation in acute stroke patients. Cerebrovasc Dis. 2010;30(4):410–417. [DOI] [PubMed] [Google Scholar]
- 11. Lombardi F, Tarricone D, Tundo F, et al. . Autonomic nervous system and paroxysmal atrial fibrillation: a study based on the analysis of RR interval changes before, during and after paroxysmal atrial fibrillation. Eur Heart J. 2004;25(14):1242–1248. [DOI] [PubMed] [Google Scholar]
- 12. Soliman EZ, Howard G, Meschia JF, et al. . Self-reported atrial fibrillation and risk of stroke in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study. Stroke. 2011;42(10):2950–2953. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Ahern TP, Lash TL, Sørensen HT, et al. . Digoxin treatment is associated with an increased incidence of breast cancer: a population-based case-control study. Breast Cancer Res. 2008;10(6):R102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Ahern TP, Tamimi RM, Rosner BA, et al. . Digoxin use and risk of invasive breast cancer: evidence from the Nurses’ Health Study and meta-analysis. Breast Cancer Res Treat. 2014;144(2):427–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Biggar RJ, Wohlfahrt J, Oudin A, et al. . Digoxin use and the risk of breast cancer in women. J Clin Oncol. 2011;29(16):2165–2170. [DOI] [PubMed] [Google Scholar]