|Home | About | Journals | Submit | Contact Us | Français|
Sleep has been suggested to influence breast cancer risk, however the evidence is mixed. Black women have higher prevalence of both short (<6 hr) and long (≥9 hr) sleep duration and are more likely to develop more aggressive, hormone receptor negative breast cancer. No study has examined the relationship between sleep and breast cancer in blacks. We focused on race-specific associations among the blacks.
In the Southern Community Cohort Study (SCCS), a prospective study of which 2/3 of the population were black, we prospectively investigated self-reported sleep duration in relation to overall breast cancer risk and by estrogen (ER) and progesterone receptor (PR) status in all women and in black women alone.
Sleep duration was not associated with risk of total or hormone receptor positive breast cancer. However, we found a suggestion of an inverse relationship between sleep duration and risk of ER− and PR− breast cancer among all women and black women. Compared to the reference group (8 hr), black women who reported shorter sleep had an increased risk of ER−PR− breast cancer (odds ratios (95% confidence intervals): 2.13 (1.15, 3.93), 1.66 (0.92, 3.02) and 2.22 (1.19, 4.12) for <6, 6 and 7 hr, respectively, p for trend, 0.04).
Short sleep may be a risk factor for hormone receptor negative breast cancer among black women.
Night time sleep, a critical component of the 24 hour circadian rhythm, has been hypothesized to play a role in breast cancer development (1). Mechanistic studies have shown that melatonin, a key hormone involved in circadian regulation, can interact with the estrogen signaling pathway and may act as a tumor suppressor (2). Moreover, sleep deficit is associated with various breast cancer risk factors, including obesity (3), metabolic dysfunction (4) and chronic inflammation (5). Previous studies have found an elevated breast cancer risk among night shift workers (6), further supporting a potential carcinogenic effect of circadian disruption.
Several epidemiologic studies have examined sleep duration in relation to breast cancer risk, but their findings are inconsistent (7–13). Two prospective studies found an inverse association, with one reporting an elevated risk among short sleepers (<6 hour) (10), and the other reporting reduced risk among long sleepers (≥9 hour) (7), comparing to women with 7–8 hours of sleep. In contrast, a case-control study suggested a positive trend of modestly increased risk with longer sleep (9). Four studies, one case-control (11) and three large cohort studies (8, 12, 13), had largely null findings.
In addition to these conflicting findings, another deficit in the literature is lack of examinations of sleep duration and breast cancer risk in black women. In the United States, blacks have the highest prevalence of both long (≥9 h) and short (<6 h) sleep durations (14), making it particularly important to understand the health consequences of insufficient and excessive sleep in this population. Moreover, black women are more likely to develop more aggressive, hormone receptor negative breast cancer than white women (15). Causes of this disparity remain unclear and may involve biological, social and lifestyle factors (16). Little is known about whether the associations between sleep duration and breast cancer differ by tumor subtypes, and whether racial differences in sleep duration contribute to racial disparities in breast cancer. To date, only two studies have examined the effect of sleep duration on breast cancer risk by estrogen receptor (ER) status and both reported null findings for ER+ and ER− subtypes (11, 13).
In this study, we investigated sleep duration in relation to breast cancer risk in the Southern Community Cohort Study (SCCS), a prospective study that enrolled a large number of blacks. We placed special emphasis on tumor subtypes by ER and progesterone receptor (PR) status and examined the race-specific associations among the blacks.
The SCCS is a prospective study initiated in 2002 focusing on racial and socioeconomic disparities in the risk of cancer and other chronic diseases (17). Between 2002 and 2009, over 85,000 men and women, aged 40 to 79, were recruited from 12 southeastern states. The majority of the participants (86%) were enrolled from 71 Community Health Centers (CHC), institutions providing basic health and preventive services mainly to low income, underinsured and uninsured persons. The rest of the study participants (14%) enrolled in 2004–2006 by responding to a mailed questionnaire sent to randomly selected residents of the same 12 states. Informed consent was obtained from each participant upon enrollment into the SCCS. Institutional Review Boards at Vanderbilt University (Nashville, TN) and Meharry Medical College (Nashville, TN) approved the study.
For this analysis, of the 48,970 women with information on sleep duration, we excluded those who reported a previous diagnosis of cancer except for nonmelanoma skin cancer (N=3,872), and those who resided outside the 11 states (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia) where cancer registry data was available (N=2,145). The final analytic sample included 42,953 women.
Incident breast cancer cases were identified via linkage to state cancer registries. Because information on follow-up times for cancer incidence varied from state to state and censoring dates were not rigid, we used a nested case-control design. All incident invasive breast cancers were included as cases (N=518) and controls were all the other female participants who were free of breast cancer at the most recent registry linkage (N=42,435). Data on ER status and PR status were obtained from the cancer registries and supplemented by pathology reports and medical records. ER and PR status was available for 438 and 436 women, respectively (status for both ER and PR was available for 430 cases).
In the baseline questionnaire, participants were asked how many hours they typically slept in a 24-hour period, on weekdays and weekends separately. In addition, we calculated a weighted average sleep duration per 24 hours ((weekday sleep duration x 5)+(weekend sleep duration x 2)/7). We grouped weekday and weekend into 5 categories: <6 hr, 6 hr, 7 hr, 8 hour and ≥9 hr. Average sleep duration was rounded and grouped into the same 5 categories. The largest group (8 hr) was considered as the reference. In analysis of tumor subtypes among white women, we combined the 8 and 9+ hour groups to form the reference group to preserve statistical power. To examine the effects of different weekday vs. weekend sleep patterns, we also categorized participants as consistently normal/long sleepers (8+ hour of sleep on both weekdays and weekends), consistently short sleepers (<8 hour of sleep on both weekdays and weekends), weekday-only short sleepers (<8 hour of sleep on weekdays but 8+ hour of sleep on weekends), and weekend-only short sleepers (<8 hour of sleep on weekends but 8+ hour of sleep on weekdays).
The baseline questionnaire also collected comprehensive information on demographic characteristics, height and weight, medical history, reproductive history, hormone therapy use, family history of cancer, and diet and other lifestyle factors such as physical activity, alcohol drinking, and smoking.
Participants in different sleep categories were compared using χ2 statistics for categorical characteristics and the nonparametric Kruskal Wallis test for continuous variables. Logistic regression models were used to estimate odds ratios (OR) and 95% confidence intervals (CI) for incident breast cancer for each category of sleep duration, with 8 hour of sleep or consistent normal/long sleepers as the reference groups. In the multivariable models we adjusted for potential confounders and risk factors for breast cancer, including age (continuous), enrollment year (continuous), race (black, white, other, or missing), enrollment state (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia), education (less than high school, high school or GED, some college or vocational training, college graduate or higher, or missing), marital status (single, married, separated, divorced or widowed, or missing), income (<$15,000,$ 15,000–<$25,000, $25,000–<$50,000, ≥$50,000, or missing), body-mass index (<25, 25–<30, 30+ kg/m2 or missing), moderate-to-vigorous physical activity (quartiles, metabolic equivalent), overall sitting (quartiles, hours/day), smoking status (current, former, never or missing), pack-years of smoking (0, 0–20, 21–40, 40+, or missing), number of live births (0, 1, 2+, or missing), age at first birth (nulliparous, <20, 20–<30, 30 years, or missing), length of breast feeding (nulliparous, 0–<1 year, ≥1 year, or missing), age at menarche (≤12, >12 years of age, or missing), post-menopausal (yes, no, or missing), ever use of menopausal hormone therapy (yes, no, or missing), current use of multivitamins (yes, no, or missing), current use of aspirin (yes, no, or missing), history of diabetes (yes, no, or missing), family history of cancer among first degree relatives (yes, no, or missing), average number of alcoholic drinks consumed per day (0, >0–1, >1, or missing), and daily intake of total fat (continuous), fiber (continuous), folate (continuous) and total calories (continuous). For all dietary intakes, missing values (N=2,297) were set at mean, and were adjusted for total energy intake using the density method (dividing intake amount by total calories). Tests for linear trend were performed by modeling a numeric value (1 through 5) for each sleep category.
We conducted additional subgroup analyses stratified by race (white/black), as well as by ER and PR status. In sensitivity analyses, we also excluded cases diagnosed within 2 years after enrollment. Additionally, we tested for interactions between sleep duration and age, education, BMI, and menopausal status using the likelihood ratio test comparing models with the cross-product terms to those without. All analyses were performed using SAS (SAS 9.3; SAS Institute, Cary, North Carolina).
At baseline, 28% of the women reported sleeping 8 hours per night (our analytic referent) on weekdays, 59% reported <8 hours and 13% reported ≥9 hours. On weekends, 29%, 50% and 21% reported 8, <8 and ≥9 hours of sleep, respectively. Baseline characteristics by weekday sleep duration are presented in table 1 (race-specific baseline characteristics by weekday sleep duration, supplementary table 1). Compared to women who reported 8 hours of sleep, both <6 hr and ≥9 hr groups were less likely to have a college education, or to be married, but they were more likely to be current smokers and report household income less than $15,000. Moreover, sleep duration was inversely associated with early menarche and physical activity.
We did not observe an association between overall breast cancer risk and weekday sleep duration, weekend sleep duration, or the weighted average of weekday and weekend sleep duration, in combined analyses of the entire cohort (table 2), or in analyses separately conducted among white or black women (supplementary table 2). Adjusting for potential confounders had little impact on the results, and the null associations were unchanged after we excluded cases diagnosed within the first 2 years of follow up. We also found no interaction between sleep duration and age, education, income, BMI or menopausal status on the risk of breast cancer (data not shown).
Because people who had more variation in sleep duration over a week’s time may experience more circadian disruption, we examined the relationship between weekday/weekend sleep patterns and breast cancer risk. Compared to consistent normal/long sleepers (who reported sleeping 8 hour or more on both weekdays and weekends), we did not find a significant difference in overall breast cancer risk among women who were consistent short sleepers (OR (95% CI): 1.08 (0.89, 1.31)), weekday-only short sleepers (1.16 (0.87, 1.54)) or weekend-only short sleepers (0.98 (0.59, 1.62)). These results were similar for black and white women separately (data not shown).
We further examined the relationship between sleep duration and breast cancer by hormone receptor status. Among black women, we did not find an association between weekday sleep duration and ER+, PR+ and ER+PR+ breast cancers (table 3). However, we found a suggestion of elevated risks for ER−, PR− and ER−PR− breast cancers among black women who reported <8 hours of sleep on weekdays. When compared to the reference group the ORs (95% CI) for ER−PR− breast cancer were 2.13 (1.15, 3.93), 1.66 (0.92, 3.02) and 2.22 (1.19, 4.12) for <6 hr, 6 hour and 7 hours of sleep, respectively (p for trend, 0.04). In contrast, we did not observe a similar association with hormone receptor-negative breast cancer among white women, but the number of cases was small (supplementary table 3); and the results for all women were largely similar to those for black women (supplementary table 4). Finally, we observed no association between weekend sleep duration and breast cancer risk by hormone receptor status, and the results for average sleep duration were similar to those for weekday sleep.
In this cohort comprised of predominantly black women, we did not observe an association between sleep duration and overall breast cancer risk, either in the total study population or among black or white women when examined separately. However, we found a suggestion of an elevated risk of ER− and PR− breast cancer subtypes among the subgroup of black women who reported less than 8 hours of sleep on weekdays.
The lack of association between sleep duration and overall breast cancer risk is consistent with the findings from some previous studies, particularly the two large cohort studies, the Nurses’ Health Study (NHS) (8) and the Women’s Health Initiative (WHI) (13). Both studies included thousands of breast cancer cases (4,223 in NHS and 5,149 in WHI). Compared to the reference group (7 hours of sleep), the multivariate relative risk (95% CI) associated with ≤5 hour and ≥9 hour were 0.93 (0.79–1.09) and 0.95 (0.82–1.11) in the NHS and 0.95 (0.85, 1.07) and 1.03 (0.90, 1.18) in the WHI. In contrast, the two cohort studies that reported an inverse relationship between sleep duration and overall breast cancer risk were smaller (242 for the Finnish Twin cohort (7) and 143 for the Japanese Ohsaki cohort (10)). A recent meta-analysis that summarized six of the seven studies (without the WHI study) reported no evidence for a relationship between sleep duration and breast cancer risk (18).
We found a suggestion of increased risk of hormone receptor negative tumors among women reporting less than 8 hours of sleep. In contrast, previous studies that examined ER− tumors reported no association with sleep duration (11, 13). One of the key differences between our study and the other studies is that two-thirds of our study population are blacks, while the two earlier studies were of predominantly European ancestry populations. Although we had small numbers of ER− and PR− tumors and the findings could be due to chance alone, especially given the multiple comparisons made, the elevated risk among black women with short sleep may deserve further investigation. Compared to their white counterparts, black women are more likely to develop more aggressive, hormone receptor negative tumors (15), and have short sleep duration (14). Therefore it would be important to examine whether insufficient sleep contributes to the racial disparities of aggressive breast cancer. Additionally, circadian disruption has been associated with factors that may influence breast cancer risk independent of estrogen pathways, such as metabolic dysfunction (4) and chronic inflammation (5), suggesting that a link between short sleep and hormone receptor negative breast cancer is biologically plausible.
We found no association between sleep duration and hormone receptor positive tumors. An Australian case-control study and the WHI also showed no relationship between sleep duration and ER+ tumors (11, 13). These findings are contradictory to the melatonin hypothesis, which suggests that short sleep duration is associated with decreased levels of melatonin, a molecule with anti-estrogenic effects that may protect against ER+ tumors (1). Although studies have linked low levels of melatonin with increased breast cancer risk, supporting a role of melatonin in breast carcinogenesis (19–21), there is little evidence showing a link between self-reported sleep duration and melatonin level and previous studies on this topic showed mixing results (12, 22, 23). The secretion of melatonin is promoted by the absence of light at night and does not require the occurrence of sleep (1). Therefore, self-reported sleep duration may be a poor marker of nighttime light exposure and melatonin production, which may partially explain the lack of association between sleep duration and hormone receptor positive breast cancer in the literature.
Our study has several limitations. First, sleep duration was self-reported. In another population, a validation study comparing self-reported and actigraphy-measured sleep duration among adults found a moderate correlation (correlation coefficient ~0.45) between the two (24), and sleep duration reported in questionnaire tended to be longer than the measured duration, particularly among people with short sleep duration and among blacks. Therefore, such measurement error in sleep duration reporting may lead to a misclassification of short sleepers as normal/long sleepers, which would have attenuated the results in our study. Second, we only had a single report of sleep duration, and were not able to capture a long term sleep patterns among the study population. As argued by a recent paper, to better understand the relationship between sleep and cancer risk, it may be important to evaluate the cumulative exposure of short or long sleep in critical time windows of cancer development (25). Third, we lacked information on other important aspects of sleep and circadian rhythm which may also influence breast cancer risk, such as sleep quality, snoring, sleep disorder, and chronotype (2). Moreover, we did not have information on shift work, another potentially important confounder of the relationship between sleep and breast cancer risk. Lastly, we were underpowered to examine ER− and PR− tumors, thus our findings should be interpreted cautiously. Particularly, we had a small number of hormone receptor negative cases among the white women, and thus are limited in commenting on the true likelihood of effect modification by race..
Our study does have several important advantages. This is a large prospective study of a well characterized population, with sleep duration measured before the development of breast cancer. Also, we were able to perform sensitivity analyses in which cases diagnosed within two years after sleep assessment were excluded, which should reduce the likelihood of reverse causation (e.g., that undiagnosed breast cancer influenced reported sleep patterns). We controlled for a large number of potential confounders, including demographics, socioeconomic status, reproductive factors, smoking, alcohol, diet and physical activity. Finally, we had a large population of black women included in this analysis, which allowed us to conduct analyses specific to black women for the first time.
In conclusion, our findings suggest that sleep duration has no effect on hormone receptor positive breast cancers, but we could not exclude the possibility of increased risks for ER− and PR− tumors among black women with short sleep. Our study is the first to examine sleep and breast cancer risk among black women. Given the high prevalence of insufficient sleep in this population, more studies are needed to investigate its potential health consequences.
Funding: This research was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, National Institutes of Health, Department of Health and Human Services.
Potential competing interests
The authors declared no conflict of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.