Sleep duration and the risk of breast cancer: the Ohsaki Cohort Study

In a prospective study of 23 995 Japanese women, short sleep duration was associated with higher risk of breast cancer (143 cases), compared with women who slept 7 h per day, the multivariate hazard ratio of those who slept ⩽6 h per day was 1.62 (95% confidence interval: 1.05–2.50; P for trend=0.03).

Breast cancer is the commonest cancer in women, worldwide (Parkin et al, 2005). In Japan, the incidence rate age-standardised to the world population was 28.3 per 100 000 in 1991 and 39.5 in 2001 (The Research Group for Population-based Cancer Registration in Japan, 1998;Marugame et al, 2007).
Melatonin, which is secreted mainly from the pineal gland and plays a role in sleep duration, is suggested as an agent in the association between sleep duration and breast cancer (Brzezinski, 1997;Schernhammer and Schulmeister, 2004). This is because melatonin suppresses the synthesis and secretion of sex hormones by promoting the release of gonadotropin-releasing hormone (Martin and Klein, 1976;Aleandri et al, 1996). In relation to melatonin secretion, there have been several observational studies on night work or visual impairment and breast cancer (Feychting et al, 1998;Verkasalo et al, 1999;Kliukiene et al, 2001;Schernhammer et al, 2001;Megdal et al, 2005;Pukkala et al, 2006;Schwartzbaum et al, 2007). In addition, there have been three prospective cohort studies of sleep duration and the risk of breast cancer, although with inconsistent findings (Verkasalo et al, 2005;Pinheiro et al, 2006;Wu et al, 2008).
We therefore examined the association between sleep duration and the risk of breast cancer in a population of Japanese women.

MATERIALS AND METHODS
Details of the Ohsaki National Health Insurance (NHI) Cohort Study have been described previously (Tsuji et al, 1998;Kuriyama et al, 2006). Briefly, this prospective cohort study, started in 1994, included 28 515 women aged 40 -79 years living in the 14 municipalities of Miyagi Prefecture, northeastern Japan. The response rate was 95.0% (N ¼ 27 134) for the questionnaire, including items on sleep duration and other health-related lifestyle factors. The study protocol was reviewed and approved by the ethics committee of Tohoku University School of Medicine.
After exclusion of participants who had withdrawn from the NHI before follow-up, those who had history of cancer, those who had omitted responses for sleep duration, and those who had reported sleep duration of less than 4 h or more than 12 h, 23 995 participants remained. To follow up participants for mortality and migration, we reviewed the NHI withdrawal history files for 1995 -2003. Through the Miyagi Prefectural Cancer Registry, we identified 143 incident cases of breast cancer.
With regard to the sleep duration, participants answered the mean integer number of hours of sleep per day during the last year. Because of the small number who slept for less than 7 h and less than 8 h, we categorised sleep duration into four groups: p6, 7, 8, and X9 h per day. We estimated hazard ratios (HRs) and 95% confidence intervals (CIs) of breast cancer incidence according to sleep duration, using the Cox proportional hazards model, with adjustment for age and potential confounders. The continuous P for trend was calculated by treating sleep duration as a continuous variable, and the categorical P for trend by treating each category as a continuous variable. Interactions between the risk and all confounders were tested through the addition of crossproduct terms to multivariate model. All statistical analyses were performed using SAS statistical software, version 9.1 (SAS Institute Inc, Cary, NC, USA), and all those reported were two-sided; differences at P-values of o0.05 were accepted as significant. Table 1 shows the baseline characteristics of participants according to sleep duration. Participants who slept 6 h or less were more likely to have a family history of cancer, to have used oral contraceptive drugs, and to be premenopausal. Participants who slept 9 h or more were older, had a smaller total caloric intake, lower educational level, were more likely to have a history of diseases, and were less likely to be employed, married, and premenopausal. Using women who slept 7 h as the reference group, we found an inverse association between sleep duration and breast cancer risk. The HR of women who slept 6 h or less was 1.62 (95% CI: 1.05 -2.50), of those who slept 8 h was 1.14 (95% CI: 0.75 -1.73), and of those who slept 9 h or more was 0.72 (95% CI: 0.36 -1.43) (P for trend ¼ 0.03). This result did not change substantially when participants whose event occurred within 3 years of baseline (N ¼ 49) were excluded and stratified analysis by age and menopausal status (Table 2). In addition, we examined in detail confounding and effect modification by other covariates on the associations between sleep duration and the risk of breast cancer. No statistically significant interaction was observed between sleep duration and other confounding factors for the risk of breast cancer on a multiplicative scale (data not shown).

DISCUSSION
This study revealed an inverse association between sleep duration and the risk of breast cancer in Japanese women, participants who slept 6 h or daily having a significantly increased risk of breast cancer.
There have been three prospective cohort studies of breast cancer in relation to sleep duration (Verkasalo et al, 2005;Pinheiro et al, 2006;Wu et al, 2008), of which the last two reported a significantly decreased risk in long sleepers and our results are consistent with these. By contrast, another study reported no such association (Pinheiro et al, 2006), possibly studied because residential nurses were studied with rotating-shift work and varying timing of sleep, so that generalising from their results may be inappropriate.
Melatonin is suggested to be involved in this relationship with sleep duration, a decrease that results in a shorter duration of nocturnal melatonin secretion (Wehr, 1991). A lower melatonin level was associated with an increased risk of breast cancer (Schernhammer and Hankinson, 2005;Schernhammer et al, 2008). Melatonin may have an inhibitory effect on gonadal function, including the synthesis and secretion of sex hormones, by promoting the release of gonadaotropin-releasing hormone (Martin and Klein, 1976;Aleandri et al, 1996); it also exerts an antiproliferative effect on breast cancer cell lines (Blask et al, 1997).
Our study had several strengths. First, we recruited participants from the general population, allowing possible generalisation of our results. Second, the Miyagi Prefectural Cancer Registry is one of the earliest and most accurate population-based cancer .5 -25.0, X25.0 kg m À2 ); history of diseases (having history of stroke, hypertension, myocardial infarction, or diabetes mellitus); family history of cancer (presence or absence in first-degree relatives); job (employed or unemployed); marital status (married or unmarried); education (junior high school or less, high school, or college/university or higher); cigarette smoking (never smoker, ex-smoker, current smoker 1 -19 cigarettes per day, or current smoker X20 cigarettes per day); alcohol consumption (never drinker, ex-drinker, or current drinker); time spent walking (less than 1 h per day, or longer than 1 h per day). b Multivariable HR2 was adjusted for above plus total caloric intake (continuous variable, kcal day À1 ); menopausal status (premenopausal or postmenopausal); age at menarche (p13 years, 14 -15 years, or X16 years); age at first delivery (p21 years, 22 -25 years, or X26 years); number of deliveries (0 births, 1 -2 births, or X3 births); using of oral contraceptive drugs (yes or no); using of hormone drugs except for oral contraceptive drugs (yes or no). c Multivariable HR3 denotes the HR2 with cases diagnosed in the first 3 years of follow-up excluded from the analysis.
registries in Japan (Takano and Okuno, 1997), with only 2.7% of breast cancer cases ascertained by death certificate only (DCO) in -2002(Curado et al, 2007. Our study also had several methodological limitations. First, we used self-reported sleep duration, and the assessment was done only once. Second, we had no information on such factors as sleep quality, the timing of sleep, the use of sleep medication, or the presence of sleeping disorders that can influence sleep duration and thereby might affect breast cancer risk. Finally, we had no information about rotating-shift work or night work, but since 23% of our participants were housewives, 19.0% farmers, and 15.7% retired, such details would have been unlikely to have changed the result substantially.
In conclusion, we have found a significant inverse association between sleep duration and breast cancer risk in Japanese women, those who slept 6 h or less having a significantly increased risk.