SEARCH

SEARCH BY CITATION

Keywords:

  • breast cancer;
  • sleep duration;
  • meta-analysis

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Studies on the association of short or long sleep duration with breast cancer risk have reported inconsistent results. We quantitatively assessed this association by conducting a meta-analysis based on the evidence from observational studies. In April 2013, we performed electronic searches in PubMed, EmBase and the Cochrane Library to identify studies examining the effect of sleep duration on breast cancer incidence. The odds ratio (OR) was used to measure any such association in a random-effects model. The analysis was further stratified by confounding factors that could bias the results. A total of six studies (two case–control and four cohort studies) involving 159,837 individuals were included in our meta-analysis. Our study did not show an association between either short or long sleep duration and breast cancer risk (short sleep duration data: pooled OR = 1.01, 95% confidence interval (CI) = 0.90–1.14, p = 0.853; long sleep duration data: pooled OR = 0.95, 95% CI = 0.86–1.04, p = 0.251). Moreover, we did not identify any statistically significant association between sleep duration and breast cancer risk in all the subgroup analyses. In conclusion, our findings indicate that sleep duration has no effect on breast cancer risk.

Breast cancer has become one of the most common cancers and is a leading cause of morbidity and mortality in women worldwide.[1-3] Lifestyle factors, including alcohol consumption, weight gain and night-shift work, are reportedly associated with breast cancer risk.[2, 4] With societal development, sleeping patterns of the population have changed, which may affect the quality of life. Recently research has suggested a link between sleep duration and long-term health status.[5-11] Two systematic reviews have indicated that both short and long sleep duration are significantly associated with an increased risk of all-cause mortality.[7, 8] Moreover, several epidemiological studies have shown that sleep duration is associated with risk of diseases such as cardiovascular disease, obesity and cancer.[5, 6, 9-11] However, the association between sleep duration and breast cancer incidence has not been confirmed by observational studies or meta-analysis.

A possible reason for these inconsistent findings could be that individual studies did not have sufficient power to indicate any benefit or harm, particularly if event rates were lower than expected, thus resulting in broad confidence intervals (CIs). Wang et al. conducted a meta-analysis on the association of long sleep duration and breast cancer risk compared with a short sleep duration reference group, and found no strong evidence for any relationship. The result of this meta-analysis was published as a conference abstract.[12] To gain a greater understanding of the effect of both short and long sleep duration on breast cancer risk, we performed a quantitative meta-analysis of observational studies to updates the previous meta-analysis and to evaluate the association between short and long sleep duration and breast cancer risk.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Search strategy and selection criteria

Observational studies (case–control and cohort studies) examining sleep duration and breast cancer risk were eligible for inclusion in our meta-analysis, without any restriction on language or publication status. We electronically searched the PubMed (from 1965 to April 2013), EmBase (from 1965 to April 2013), and Cochrane Library databases using the search terms “sleep” and “breast cancer.” We also conducted manual searches of reference lists from relevant original and review articles.

Eligible studies had to meet the following inclusion criteria: (i) published as an original article; (ii) a case–control study, nested case–control study, or cohort study; (iii) reported the relative risk (RR) or odds ratio (OR), and 95% CI for the association between the sleep duration and breast cancer risk. Animal experiments, mechanistic research and studies involving cancers other than breast cancer were excluded. If more than one article reported data from one study, the most recent and complete articles were included.

Data extraction and quality assessment

The following information was extracted from each eligible study: first author's surname; study design; publication year; patient numbers; regimen details; sleep reference category; category for “short” and “long” sleep duration; the RR, OR or hazard ratio (HR) for breast cancer risk for both short and long sleep duration; corresponding 95% CI; and covariates adjusted in the statistical analysis.

Data extraction and quality assessment were conducted using a standardized data-recording form and the Newcastle-Ottawa Scale (NOS).[13] The NOS is quite comprehensive and has been partially validated for assessing the quality of nonrandomized studies in meta-analysis.[14, 15] The NOS is judged on three broad subscales: selection of the study group (four items), comparability of the groups (one item), and elucidation of the exposure or outcome of interest for case–control or cohort studies, respectively (three items). A “star system” (range, 0–9) has been developed for assessment. Each study is awarded a maximum of one star for each numbered item within the selection and exposure categories, whereas a maximum of two stars are assigned for comparability. In this study, we considered a study that is awarded ≥7 stars as a high-quality study, as standard criteria have not been established.

The search and study selection (Q.Y.Y. and Z.X.), data extraction and quality assessment (Q.Y.Y. and Z.Y.H.) were conducted independently by two investigators. Information was examined and adjudicated independently by an additional investigator (H.J.), who referred to the original articles.

Statistical analysis

The adjusted RR, HR, OR and corresponding 95% CI were extracted from the selected studies and used to evaluate the association between sleep duration and breast cancer risk. Both fixed-effect and random-effects models were used to evaluate the pooled OR for the association between sleep duration and breast cancer incidence. Although both models yielded similar findings, results from the random-effects model presented here assume that the true underlying effect varies among the included studies.[16] Furthermore, in the random-effects model, we added the prediction interval (PI) to illustrate the degree of heterogeneity in forests plots, which could also provide a predicted range for the true treatment effect in an individual study.[17, 18] Compared with the reference category of sleep, the pooled OR and 95% CI of breast cancer risk for short and long sleep duration were calculated, separately. Heterogeneity between the studies was evaluated by the chi-square (χ2) test and I-squared (I2) statistic.[19] These indices assess the percentage of variability across studies that is attributable to heterogeneity rather than chance. Statistical heterogeneity was considered significant when p < 0.10 for theχ2 test or I2 > 50%.

Several methods were used to check for potential publication bias. Visual inspection of funnel plots for breast cancer was performed, and the Egger's regression test[20] and Begg test[21] were used to statistically assess publication bias. Subgroup analysis was performed according to menopausal status (premenopause and postmenopause), study design (retrospective and prospective) and population region (Asian and Non-Asian) to minimize heterogeneity among the included studies. In each specific population, the effects of short or long sleep duration on breast cancer risk were evaluated. We also conducted a sensitivity analysis by removing each individual study from the meta-analysis. Finally, in a cumulative meta-analysis, outcome data for breast cancer from all available studies were included sequentially according to the year in which they first became available, to assess the evolution of the observed effects over time.[22] All reported p values were two-sided and p values < 0.05 were regarded as statistically significant. Statistical analyses were performed using STATA 11.0 (Stata Corporation, Lakeway, TX) and the STATA programs are shown in Supporting Information Table S1.

Role of the funding source

The funding sources had no role in the study design, data collection, analysis, preparation of the manuscript or decision to publish. The corresponding author had full access to all the data in the study and had the final responsibility for the decision to submit for publication.

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Search results and characteristics of the studies

Six studies (159,837 individuals) were included in our meta-analysis based on our selection criteria (Fig. 1).[23-28] One study was reported in two separate published articles,[23, 29] and only the recent one was included in our analysis.[23] Of these studies, four studies were designed as cohort studies,[23, 25, 26, 28] and two as case–control studies.[24, 27] According to the nine-point NOS scale, four studies had NOS scores of 7,[24-27] whereas two had scores of 8.[23, 28]

image

Figure 1. Flow diagram of the studies search and selection process.

Download figure to PowerPoint

The sleep reference categories in each study were classified as 7 hr,[25, 26] 7 to 8 hr,[24, 27] 8 hr[28] and ≤6 hr of sleep/night.[23] Short sleep duration was defined as ≤4 hr,[28] ≤5 hr[26, 27] and ≤6 hr of sleep/night.[24, 25] Long sleep duration was defined as >8h,[24] ≥9 hr[23, 25-27] and ≥10 hr of sleep/night.[28] For analysis of the study by Wu et al, we identified 7 hr of sleep/night as the sleep reference category when compared to short sleep,[23] and the OR was recalculated by the exponential of negative ln (OR). Further details of each study are listed in Table 1.

Table 1. Characteristic of studies included in the meta-analysis
AuthorStudyYearCountryStudy typeAge at baselineSample sizeReference categoryShort duration of sleepLong duration of sleepAdjusted variableNOS score
  1. a

    The authors defined the short sleep as the reference category. We identified 7 hr as reference category when compared short sleep analysis. And the OR was recalculated by the exponential of negative ln(OR).

  2. Abbreviations: BMI: body mass index; HRT: hormone replacement therapy; NHI: National Health Insurance; SCHS: Singapore Chinese Health Study.

Wu et al.[23]SCHS2013SingaporeCohort45–7434,028≤6 hrNAa≤9 hrAge, parity, menopausal status, education, BMI, dialect group and year8
Girschik et al.[24]The Breast Cancer Environment and Employment Study2013Western AustraliaCase–control study18–802,8277–8 hr<6 hr>8 hrAge, age at first birth, number of children, menopausal status, breastfeeding, use of HRT, comparative weight at age 30 years, duration of use of HRT, ever use of melatonin, alcohol consumption and physical activity7
Kakizaki et al.[25]Ohsaki NHI Cohort Study2008JapanCohort40–7923,9957 hr≤6 hr≤9 hrAge, age at first birth, age at menarche, number of deliveries, menopausal status, education, job, marital status, history of diseases, family history of cancer, using of oral contraceptive drugs, use of HRTBMI, alcohol consumption, smoking, time spent walking and total caloric intake7
Pinheiro et al.[26]The Nurses' Health Study2006USCohort30–5577,4187 hr≤5 hr≤9 hrAge, age at first birth, age at menarche, age at menopause, parity, history of benign breast disease, family history of breast cancer, BMI, postmenopausal hormone use, alcohol and caloric intake, smoking and physical activity7
McElroy et al.[27]Population-based US case–control study2006USCase–control study20–69 cases9,3477–8 hr<5 hr≤9 hrAge, age at first birth, age at menopause, parity, menopausal status, state, education, family history of breast cancer, marital status, BMI, postmenopausal hormone use and alcohol consumption7
Verkasalo et al.[28]Finnish Twin Cohort2005FinlandCohortNA12,2228 hr≤4 hr≤10 hrAge, number of children, social class, BMI, use of oral contraceptives, alcohol use, smoking, physical activity and zygosity8

Short sleep duration and breast cancer risk

The relationship between short sleep and breast cancer risk is shown in Figure 2a. The pooled analysis revealed that the short sleep was not associated with breast cancer risk (OR = 1.01; 95% CI = 0.90–1.14; p = 0.853; 95% PI = 0.80–1.27) using a random-effects model, and significant heterogeneity was not observed among individual studies (p = 0.326; I2 = 13.8%). Sensitivity analyses showed that the OR and 95% CI did not alter substantially after removing any one study (data not shown). A review of funnel plots could not exclude the potential for publication bias for breast cancer (Egger test p value = 0.410; Begg test p value = 0.452; Fig. 3a). Furthermore, a cumulative meta-analysis of short sleep duration did not show any trend affecting breast cancer risk (Fig. 4a).

image

Figure 2. Forest plot of the risk of breast cancer associated with sleep duration. (a) Forest plot for short duration of sleep; (b) forest plot for long duration of sleep.

Download figure to PowerPoint

The point estimate for the pooled OR in the majority of the subgroup analyses was >1, except for the non-Asian population (OR = 0.96; 95% CI = 0.84–1.09) and postmenopausal population (OR = 0.99; 95% CI = 0.88–1.12). However, there was no statistically significant association between short sleep duration and breast cancer risk (Table 2).

Table 2. Subgroup analyses for the effect of sleep duration on risk of breast cancer
    Heterogeneity
SubgroupStudyOR (95% CI)pχ2I2p
Menopausal status      
Premenopause      
Short sleep24–261.06 (0.64–1.75)0.8243.1436.3%0.0715
Long sleep24–260.85 (0.65–1.11)0.2250.160.0%0.692
Postmenopause      
Short sleep23–260.99 (0.88–1.12)0.8503.175.2%0.367
Long sleep23–260.95 (0.84–1.07)0.3991.300.0%0.730
Study design      
Retrospective      
Short sleep24,271.02 (0.81–1.29)0.8510.220.0%0.640
Long sleep24,270.97 (0.85–1.11)0.6910.280.0%0.597
Prospective      
Short sleep23,25,26,281.04 (0.86–1.26)0.6895.5445.9%0.136
Long sleep23,25,26,280.87 (0.69–1.10)0.2584.2729.8%0.233
Population regions      
Asian      
Short sleep23,251.22 (0.77–1.94)0.4034.0975.6%0.043
Long sleep23,250.59 (0.21–1.70)0.3313.1668.3%0.076
Non-Asian      
Short sleep24,26–280.96 (0.84–1.09)0.5330.660.0%0.883
Long sleep24,26–280.96 (0.87–1.06)0.4130.930.0%0.818

Long sleep duration and breast cancer risk

The relationship between long sleep duration and breast cancer risk is shown in Figure 2b. The results indicated that long sleep duration was not associated with breast cancer risk (OR = 0.95; 95% CI = 0.86–1.04; p = 0.251; 95% PI = 0.83–1.08) using a random-effects model, and no significant heterogeneity was noted (p = 0.427; I2 = 0%). Sensitivity analyses also showed that the OR and 95% CI did not alter substantially. We found no evidence of publication bias by pooled analysis of long sleep duration using the funnel plots and the Begg–Mazumdar test (p = 0.060), but Egger's regression test results indicated potential publication bias (p = 0.029; Fig. 3b). However, the result of the trim and fill method showed that no trimming was performed and the pooled result had not been changed. A cumulative meta-analysis of long sleep duration and breast cancer risk did not show any statistically significant association. A potential trend towards the protective effect of long sleep duration could be observed according to the cumulative meta-analysis in Figure 4b; however, this trend may not be statistically significant.

image

Figure 3. Funnel plot of log relative risk vs. standard error of log relative risks. (a) Funnel plot for short duration of sleep; (b) Funnel plot for long duration of sleep.

Download figure to PowerPoint

image

Figure 4. Cumulative meta-analysis of the risk of breast cancer associated with sleep duration. (a) Forest plot for short duration of sleep; (b) forest plot for long duration of sleep.

Download figure to PowerPoint

We found that the point estimates for the pooled OR for all of the long sleep duration subgroup analysis were <1. Moreover, we did not identify any statistically significant result in the subgroup analyses (Table 2).

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

This study incorporated six available published observational studies and provided a qualitative estimate of the association between sleep duration and breast cancer risk. After integrating all the available evidence, we found that neither short nor long sleep duration was associated with breast cancer risk. Furthermore, the results obtained for long sleep duration was consistent with the previous meta-analysis (aggregate RR = 0.96; 95% CI = 0.86–1.07).[12]

Of the six studies examined, the majority indicated no association between sleep duration and breast cancer incidence, but two reported conflicting results.[25, 27] McElroy et al. demonstrated that there was a significant trend for increased sleep duration to be associated with an elevated risk of breast cancer (p for trend is 0.03).[27] However, Kakizaki et al. reported that there was inverse association between sleep duration and breast cancer risk (p for trend is 0.03), and the multivariate HR for the people who slept ≤6 hr was 1.62 (95% CI = 1.05–2.50) when compared with those who slept 7 hr.[25] The pooled results of our meta-analysis were consistent with most of studies analyzed, and no evidence of an association between short or long sleep duration and breast cancer risk was noted. Moreover, we discovered that all the pooled OR estimate points were <1 and had a potential trend to move to the left in the cumulative meta-analysis of long sleep duration. Moreover, the range of corresponding 95% CI became narrower when the number of studies and sample size increased. We suggest that there might be a potential protective effect of long sleep on breast cancer incidence, however, this trend may not be obvious and should be validated by further research. The potential trend associating long sleep with breast cancer risk was consistent with that noted by Kakizaki et al.[25]

In subgroup analyses of menopausal status, only short sleep duration resulted in a potential risk effect for the premenopausal population (OR = 1.06; 95% CI = 0.64–1.75). Stevens et al. considered that a case–control design is probably not useful for studies on sleep duration and health due to potential flaws in exposure misclassification and bias by indication. He also suggested that prospective studies are more appropriate for examination of sleep duration and disease, as bias is not a problem in this type of study.[30] However, this criticism of case–control studies should not be universally applied, and the case–control design is useful for addressing many questions. In the subgroup analyses of study design, we found that the results of retrospective and prospective studies are similar in relation to short and long sleep duration, respectively. In subgroup analyses of population regions, only short sleep duration presented a potential risk for the Asian population (OR = 1.22; 95% CI = 0.77–1.94). However, we did not identify any statistically significant association between sleep duration and breast cancer risk in all subgroup analyses.

Most of the studies on sleep duration and cancer risk focused on its relationship with melatonin production. Several experimental studies suggested that melatonin may be able to regulate the initiation, promotion, and progression of cancer.[31, 32] Moreover, other research indicated that melatonin levels and night-shift work are associated with cancer risk.[4, 33-37] People with a short sleep duration might be exposed to light for a longer period at night-time by waking up earlier or staying up later, and low melatonin production was associated with exposure to light at night.[38] Accordingly, a hypothesis linking sleep duration and cancer risk has been proposed. However, several studies have suggested that there is no relationship between self-reported sleep duration and melatonin levels.[39, 40] This may possibly explain why there is no association between sleep duration and breast cancer risk.

Our meta-analysis has several potential limitations. First, our study was based on reported data, which may not provide robust estimation for the association of sleep duration and breast cancer. The quality of our study was determined by the quality of the individual studies included. Second, the studies that were included in our meta-analysis were observational studies. Exposure misclassification and bias by indication may be potential flaws in these studies and may affect the result of our pooled analysis. Third, the sleep duration classification criteria were not consistent among the included studies, which could cause heterogeneity in our meta-analysis. Therefore, all analyses in our study were conducted under a random-effects model. Fourth, we identified and assessed only six studies describing the association between sleep duration and breast cancer risk. This limitation could affect the results of our meta-analysis. Therefore, we could only relatively evaluate the association between different sleep durations and breast cancer incidence.

In conclusion, our study suggested that there is no association between either short or long sleep duration and breast cancer risk. Furthermore, no statistically significant effect were observed when based on predefined subsets The association between sleep duration and breast cancer risk and more reasonable criteria for the classification of sleep duration warrant more research.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We thank Xiaofei Ye and Zhichao Jin for his generous assistance with this study.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ijc28452-sup-0001-suppinfo01.doc42KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.