Article Text

Download PDFPDF

Shift work and breast cancer: do we need more evidence and what should this be?
  1. Kyriaki Papantoniou1,2,
  2. Manolis Kogevinas1,2
  1. 1Centre for Research in Environmental Epidemiology (CREAL), Barcelona, Spain
  2. 2CIBER Epidemiología y Salud Pública (CIBERESP), Madrid, Spain
  1. Correspondence to Professor Manolis Kogevinas, Centre for Research in Environmental Epidemiology (CREAL), Doctor Aiguader 88, Barcelona 08003, Spain; kogevinas{at}

Statistics from

Research on circadian disruption and cancer was stimulated by the 2007 evaluation by the International Agency for Research on Cancer of light at night and circadian rhythm disruption as probably carcinogenic to humans.1 Circadian rhythm disruption ultimately affects a large part of the population in modern societies and not only those in shift work. The experimental evidence on light at night and breast cancer is very strong, clearly showing the potential for light at night and circadian rhythm disruption to cause cancer. The evidence from epidemiological studies is mostly positive notwithstanding inconsistencies and lack of a clear dose–response. Since the first meta-analysis by Megdal et al,2 several systematic reviews and meta-analyses have been performed, three of them published in 2013, attempting to summarise the association between night shift work and breast cancer risk. Two3 ,4 of the three recent reviews concluded that the epidemiological evidence is overall weak or insufficient, whereas the third review5 supported the hypothesis. Even though inconclusive, all reviews highlighted the methodological limitations of the existing studies and the heterogeneity between studies in methods including basic definitions of exposure. So, do we need more evidence and what should this be?

The paper by Grundy et al6 provides more evidence supporting the association between shift work and breast cancer risk, as most case–control studies have done so far. In this well-conducted large case–control study, a twofold risk of breast cancer was found in women doing more than 30 years’ shift work, while no association was found for shorter periods of exposure. Exposure assessment was carefully carried out and was based on individual data with lifetime occupational history. Shift work was characterised using different classifications based on increasing percentages of time worked in the evening and/or night period ranging from 20% to 100%, and additionally based on time schedules including jobs between 23:00 and 07:00 hours. Results from both classifications were similar among those with over 30 years of work, although they differed for workers with a shorter duration of shift work. The classification based on time schedule, which may be more relevant, indicated an increased risk even for shift workers with less than 15 years’ duration. The study did not distinguish between permanent and rotating night shift work and included both shift patterns under a single definition. Only a few studies have been able to compare exposures to permanent versus rotating shifts and night versus evening shifts with cancer risk. Shift work characteristics (regular/irregular shift schedules, permanent/rotating night shifts, time schedules of each shift, intensity, direction and speed of rotation) need to be captured in more detail in future studies.7 The inclusion of a wide range of occupations in this study, both health and non-health related, rather than a single profession, increases the external validity of the study and addresses one of the earlier doubts given that early data were based mostly on two occupational groups.

Breast cancer is a heterogeneous disease and risk factors may vary, for example, according to menopausal status at diagnosis or tumour characteristics. Therefore, in addition to exposure misclassification we may also have outcome misclassification. In the study by Grundy et al6 a higher risk was observed for postmenopausal breast cancer while no interaction was observed with oestrogen receptor status. Only a few studies have evaluated subtypes of disease, and this is an interesting aspect given that hormonal pathways are involved in circadian rhythm disruption. Furthermore, periods in life with higher susceptibility have been understudied (eg, reproductive years), and should be evaluated to detect exposure windows in which the female breast might be most vulnerable to night work.

Anyone who has intruded into the mysteries of circadian rhythm disruption in population studies will understand the complexity of the evaluation of shift work and circadian rhythm disruption in epidemiological studies. Reading the epidemiological evidence carefully one finds very different exposure classifications that frequently affect the magnitude of the risk estimates observed. Exposure assessment varies across studies, with very different definitions and/or metrics for shift work and night work as well as the use of different cut-offs in quantitative assessments of the exposure, both for duration and intensity. A large part of the problem is the difficulty in evaluating work patterns and light at night and this goes for both the ‘exposed’ and the ‘unexposed’, many of whom are exposed to light at night. Another part of the problem is the uncertainty regarding which aspects of shift work are related to biological effects relevant to cancer. Is a permanent night shift more dangerous for cancer than a rotating shift, or vice versa?

Confounding is generally considered as a potential important factor in studies on shift work, and was one of the reasons why the International Agency for Research on Cancer working group classified the epidemiological findings available at the time of the evaluation in 2007 as limited. As frequently happens in epidemiological studies, control for confounding tends not to be systematic across studies, also because patterns of exposure and co-exposure differ between societies. It is noteworthy that only one paper, a review,4 discussed the selection of confounders using directed acyclic graphs (DAGs). It is also noteworthy and surprising that several studies did not identify major confounding by factors such as socioeconomic status, number of children, age at first birth, or body mass index, alcohol and smoking. Notwithstanding these results on the lack of major confounding in some studies, adjustment for confounding is important, and analyses on confounders, mediators and effect modifiers need to be carried out carefully in future studies.

Several researchers have stressed the importance of individual characteristics on the adaptability to circadian rhythm disruption and shift work, including chronotype (an individuals’ preference for the morning or evening period) or genetic and epigenetic mechanisms.7–10 How much these factors may act as effect modifiers in the night shift work–cancer association remains to be seen, and clearly we need more data.

Let us assume that there is a causal association between shift work and breast cancer and that the ‘real’ relative risk for some, say 10, years’ night shift work is somewhere between 1.1 and 1.3. Is it surprising that we have contradictory findings both within and between studies? Hardly surprising would be our response, and some of the reasons are outlined above. So, what next? The answer may seem a ‘cliché’, but basically we need more epidemiological studies in the working and general population that advance our knowledge: this means studies with as good as possible exposure assessment, and with as much as possible evaluation of hypotheses based on the understanding of mechanisms through which light at night and circadian rhythm disruption may cause cancer. If this is feasible, we would gain by having a pooled analysis (not a meta-analysis) of all studies on breast cancer examining the raw data of the studies and applying common definitions for exposure.


View Abstract


  • Contributors KP and MK have contributed to this paper by drafting and revising this article. They have both aproved the final version of the manuscript

  • Funding KP has a fellowship from the Instituto Carlos III (PFIS/550306370), Madrid, Spain.

  • Competing interests None.

  • Provenance and peer review Commissioned; internally 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.

Linked Articles