Article Text

Short report
Shift work that involves circadian disruption and breast cancer: a first application of chronobiological theory and the consequent challenges
  1. Lin Fritschi1,
  2. J Valérie Groß2,
  3. Ursula Wild2,
  4. Jane S Heyworth3,
  5. Deborah C Glass4,
  6. Thomas C Erren2
  1. 1 School of Public Health, Curtin University, Perth, Australia
  2. 2 Institute and Policlinic for Occupational Medicine, Environmental Medicine and Preventive Research, University of Cologne, Cologne, Germany
  3. 3 School of Population and Global Health, The University of Western Australia, Perth, Australia
  4. 4 Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia
  1. Correspondence to Dr Lin Fritschi, School of Public Health, Curtin University, GPO Box U1987, Perth, Western Australia, 6845, Australia; lin.fritschi{at}


Objectives In 2007, the International Agency for Research on Cancer classified shift work involving circadian disruption (CD) as probably carcinogenic to humans. Circadian disruption could be conceptualised as the overlap of activity, such as work, with an individual’s biological night. The latter can be approximated from a worker’s chronotype (or morning/evening preference). Few previous studies have taken chronotype into account when assessing CD caused by shift work. Our objective was to test the hypothesis that women working during their biological night would be at increased risk of breast cancer.

Methods We used data from our case–control study of breast cancer to investigate associations between shift work involving CD and breast cancer risks. Previously, we had assumed that everyone working in jobs which involved work for two or more shifts between midnight and 05:00 hours was equally exposed to CD. In the present analyses, we reclassified as unexposed those who had a late chronotype in which their preferred bedtime was 2 hours after the end of their shift.

Results Only 30 of 1385 night jobs changed classification and the overall finding (OR 1.17, 95% CI 0.98 to 1.41) was not different to the original finding when chronotype was not considered.

Conclusions We found virtually no difference between our new and old classifications of exposure. However, we were not able to calculate the total number of chronodisrupted shifts over a lifetime in order to assess dose and nor were we able to determine how many women were exposed to CD when doing shifts which began before midnight. Our first practical application highlights challenges for future chronobiology-based research.

  • breast cancer
  • night work
  • chronodisruption
  • case-control

Statistics from

What this paper adds

  • The International Agency for Research in Cancer has classified ‘shift work that involves circadian disruption’ as probably carcinogenic. Circadian disruption may be assessed by taking into account the chronotype (morning/evening type) of the person doing the shifts.

  • We aimed to reclassify our exposure data from a case–control study of breast cancer by incorporating circadian disruption into the definition of exposure to shift work.

  • Only a few jobs changed category and we were limited by the data we collected. Future studies should collect data on all shift times not just night shifts and also obtain the number of shifts per year.


In 2007, the International Agency for Research on Cancer (IARC) concluded that ‘shift work that involves circadian disruption (CD)’ is probably carcinogenic to humans based on ‘sufficient evidence in experimental animals for the carcinogenicity of light during the daily dark period (biological night)’ and ‘limited evidence’ of increased risks of breast cancer in shift workers in epidemiological studies.1 The working group did not define the term CD.

The body clock governs circadian rhythms with daily cycles of sleeping, wakefulness, eating, etc. Humans vary in the timing of their genetically determined body clock (chronotype); some people prefer to be awake in the mornings, others prefer the evening.2 3 This implies that not all workers with the same shifts will experience CD to the same extent.4 For example, a shift which ends at 02:00 will result in CD for an early chronotype but not for a late chronotype.

Studies examining shift work and breast cancer have had inconsistent results,5 and one potential reason for this inconsistency may be that the shift-work assessment does not always consider chronotype. To do this, one can estimate each person’s biological night and then determine whether the shift they do overlaps with this time window. We would expect that people who work during their biological night would experience CD.6 In this analysis, we have refined our measure of shift work used in our previously reported case–control study of breast cancer7 to include CD.


The methods for the Breast Cancer Environment and Employment Study (BCEES) have been described previously7 and are summarised here. Approval for the study was obtained from the Human Research Ethics Committee of The University of Western Australia (WA) and the WA Department of Health. Informed consent was obtained from all participants.

Cases were women aged between 18 and 80 years with an incident breast cancer (2009–2011) identified from the State Cancer Registry. Of 2084 presumably eligible cases, 1205 consented, 334 refused and 545 did not respond. Controls were randomly selected from the State electoral roll, frequency age matched to the expected distribution of cases. Of 4356 presumably eligible controls, 1789 consented to participate, 939 refused and 1628 failed to respond. Three cases and four controls did not complete the occupational section of the questionnaire.

Participants completed a questionnaire regarding demographic characteristics, reproductive history, lifestyle factors, and family history of breast cancer. We estimated the start and end of each woman’s biological night from two items of the Horne-Ostberg questionnaire8 which asked: ‘Considering only your own ‘feeling best’ rhythm, at what time would you get up (go to bed) if you were entirely free to plan your day?’

Participants also provided details for jobs they had held for at least 6 months including whether the job involved shift work. For shift work jobs, interviewers asked whether they worked between midnight and 05:00 hours (graveyard shift), whether they worked a shift that started between 05:00 and 07:00 hours and start and finish times of each shift.

Classification of circadian disruption

For each job, we added 2 hours before the shift start time and after the shift end to take into account preparation and travel time. CD was defined as occurring if the woman worked for 1 hour or more during her preferred hours of sleep (ie, her biological night). Late CD occurred if 1 hour or more of the evening work day was after the start of the woman’s biological night. Early CD occurred if the start of the morning work day was before the end of the woman’s biological night.


We combined jobs for each person to determine if a woman had ever worked in a job which involved late, early or any CD and the total number of years with each type of CD. These six variables were the predictor variables of interest in logistic regression models with the outcome of breast cancer, adjusting for the frequency matching variable of age.

We repeated the models adjusting for socioeconomic score and remoteness of residence,9 10 highest level of education, country of birth, family history of breast cancer, number of children, age at menarche, number of months of breastfeeding, alcohol intake, physical activity, smoking and body mass index (BMI) in their 30s.

Analyses were repeated stratifying by menopausal status at the time of the cancer diagnosis or, for controls, their interview. Sensitivity analyses were conducted including jobs where only one night shift in a row was worked.

If women self-select into shifts which suit their chronotype, we would expect to see more evening preference women doing night shifts and more morning preference women doing early shifts. We therefore compared womens’ Horne-Ostberg categorisation with whether they had ever worked rosters involving night and/or morning shifts and the number of years of those rosters.


For this analysis, we dropped three subjects who had improbable preferred sleep and wake times leaving 1201 cases and 1783 controls. As in our previous publications,7 controls were slightly younger than cases and the known pattern of associations with reproductive factors was seen, cases tended to be more likely than controls to be nulliparous, to have a family history of breast cancer and have an earlier age at menarche.

There were 1595 jobs (8.3% of all jobs) in which the woman had a roster involving work between the 00:00 and 05:00 hours. Of these, 210 jobs involved only one night shift in a row, so in the primary analysis that job was not considered to induce CD. In a further 30 jobs, although the shift involved work between midnight and 05:00 hours, the working hours (+2 hours each side) did not include the woman’s biological night. That is, the woman was working a graveyard shift, but was not disrupting her innate sleep/wake time. This left 1355 jobs involving late CD.

There were 928 jobs where the woman worked a shift which started between 05:00 and 07:00 hours. Of these, 83 involved only one early morning in a row and another 19 did not include the woman’s biological night, leaving 826 jobs which resulted in early CD.

There were 377 jobs which involved both early and late CD and a total of 1804 (9.4% of all jobs) which had any CD.

When we combined all jobs, there were 585 women with late CD (see table 1 in the online Supplementary file 1). That is, they had ever worked in a job with two or more graveyard shifts in a row and 1 hour or more of the evening work day (including 2 hours start up and wind-down times) was after the start of the woman’s biological night (table 1). After adjusting for age, the risk of breast cancer associated with late CD was 1.17 (95% CI 0.98 to 1.41) and for early morning CD was 1.00 (95% CI 0.82 to 1.21). Those who had jobs involving late CD for shorter periods had higher risks than those who had longer term CD. In the fully adjusted models, the number of subjects reduced by 441, due primarily to missing BMI data, and none of the ORs were statistically significant.

Supplementary file 1

Table 1

Association between breast cancer and measures of circadian disruption, Breast Cancer Environment and Employment Study, Western Australia

There were no major changes to the findings when we stratified for menopausal status, omitted the 2-hour preparation and wind-down times or assumed that even one shift in a row induced CD (see table 2 in the online Supplementary file 1).

There was a tendency for morning types to do early shifts (8.9% of morning types vs 6.8% of evening types) and for evening types to do graveyard shifts (12.8% of evening types vs 9.7% of morning types) (see table 3 in the online Supplementary file 1). However, there was no difference by chronotype in the time spent in the different shift rosters.


This study is the first investigation of shift work and breast cancer to incorporate chronotype into the definition of exposure. We found a small, non-statistically significant association similar to that when shift work was measured without taking into account individual biological cycle.7 This is not surprising as only 30 jobs changed status from being exposed to the graveyard shift to being not exposed to late CD.

However, although we lost exposed subjects when their biological night started after their shift finished, we did not gain any subjects whose biological night started before midnight and they worked, say, until 23:30. Unfortunately, it was not possible to quantify this CD as we only had asked details on shift times for those who worked after midnight. We do know that 853 women stated that they did shift work in their questionnaire but they did not do shifts involving work between midnight and 05:00 hours. Of these, 258 women were morning types and so may have been chronodisrupted when working shifts which ended before midnight.

As well as the general limitations of this case–control study,7 there were specific limitations in calculating CD. Most importantly, we lack information to quantify CD that may have resulted from work in time windows beyond the graveyard shift. This lack disallowed testing our hypothesis of CD doses being accumulated by different chronotypes during different work time windows over 24 hours.

In addition, we used the Horne-Ostberg questionnaire to obtain preferred time to go to bed and to get out of bed. These times are not necessarily the same as the preferred sleep and wake times as people have different preferences as to whether they go to sleep immediately or not.11 In addition, we only had one measure of preferred times which may have changed over the woman’s lifetime.

Our sensitivity analysis did not show strong selection effects into shift work by chronotype, although since the chronotype was measured only after the jobs were completed there is a possibility of reverse causation—in that doing shift work changed reported chronotype.

Several groups have stratified the association between night shift work and breast cancer by chronotype.7 12 13 Results have been inconsistent with the highest association between shift work and risk of breast cancer being reported in women with neutral preference,7 morning preference12 and evening preference.13 However, stratified analyses are not the same as incorporating chronotype into the definition of exposure. In stratified analyses, all night shifts are considered equally exposed, whereas we were able to individually examine the overlap between shift times and preferred sleep and wake times. For example, in our analyses a person with a preferred bed time of 23:00 and a graveyard shift start of 23:00 would be classified as chronodisrupted irrespective of self-reported chronotype, whereas in stratified analyses they would be classified as chronodisrupted if they were a morning person and not if they were an evening person.

Ultimately, we would like to calculate the total number of CD hours during a lifetime.6 Unfortunately, we had data on number of night shifts a month for only 28% of the 1388 jobs.

As our data were not originally collected to test IARC’s definition of ‘shift work involving CD’,6 we were missing important variables, such as preferred sleep/wake times and times of all shifts as the basis for quantifying the individuals’ total ‘dose’ of CD. For future studies of shift workshift work, we recommend that data are collected on preferred sleep and wake times for all shift times rather than just the graveyard shift and on the numbers of each shift per month.


This manuscript was developed during LF’s work as a Visiting Professor at the University of Cologne, Germany and we would like to thank the Medical Faculty of the University for supporting the visit. The Breast Cancer Environment and Employment Study was funded by a National Health and Medical Research Council Australia (NHMRC) project grant #572530 and by a grant from the Cancer Council Western Australia (CCWA). LF is supported by fellowships from the NHMRC and CCWA.


View Abstract


  • Contributors LF, JSH, DCG and TCE designed and carried out the original study. LF, JVG, UW and TCE developed the measures of chronodisruption. LF analysed the data and wrote the first draft of the manuscript. All authors contributed in a meaningful way to the final manuscript.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval University of Western Australia HREC.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.