Introduction

Working outside the hours with daylight from about 7 a.m. to 7 p.m. has become increasingly normal during the last 100 years. The invention of electrical lighting and the industrialization of Western countries in the late 1800s have facilitated this change of working life from solely the period with naturally lighted day period to also include evening, night, and early morning by use of artificial electrical light. Consequently, most work has moved from outdoor with bright natural light exposure to indoor with dim light conditions, including a narrower spectrum than natural daylight [1]. Currently, in the EU, about 22% of men and 11% of women work on shifts that include night work. Similarly in the USA, about 17% of full-time salaried men and 12% of women worked on shifts that included nights [2]. The most frequent sectors with night work are hospitals, hotels, transportation, security, and parts of industry that depend on a 24-h production 7 days a week, as well as modern societies in general expect 24-h activity [3, 4]. Shift and night work are complex exposures due to many different aspects that may characterize such exposure, including the direction of shifts (e.g., forwards (day, evening, night) and backwards (day, morning, night)), time of start and end, sequence of non-day shifts, work hours per shift, and rest periods between shifts [5]. Thus, Hall et al. in a recent study from Canada based on a survey of 88 companies observed over 400 different shift-systems [6].

From ancient time, the light on Earth has been determined by the planet’s 24-h rotation about itself (morning, day, evening and night) and about the 365-days circulation around the sun (seasons). Consequently, over the 3 billion years evolutionary past, virtually all life on Earth from cyanobacteria to humans has adapted to the 24 h circle of light and dark. Thus, the timing and duration of the daily exposure to light is known as one of the most important determinants for circadian (24 h) rhythms (daily oscillation) in humans, animals, and other living species on the Earth [7, 8]. In humans and other mammals, the pineal gland hormone melatonin is the main biological signal which synchronizes the main time keeper, the suprachiasmatic nucleus (SCN) in the anterior hypothalamus, with local clocks in cells and tissue, and aligns the entire circadian system to the local environmental dark-light time of the 24-h day [9]. Normally, melatonin is only produced from dusk to dawn, and the amplitude and total daily production is strongly influenced by age, sex, chronotype (morning or evening preference), individual sensitivity, timing, duration, and history and spectrum of light exposure on retina during the day. In the event of exposure to light-at-night, e.g., during social activities or in particular work, melatonin is normally suppressed, and a phase shift may occur after several consecutive light-at-night exposures [10,11,12,13]. Melatonin has oncostatic properties in animal assays and has been shown to mediate pathways involved in cancer [14], including estrogens involved in breast cancer [15, 16]. Further, it has been shown in some prospective studies that women with the highest levels of metabolites of melatonin have a lower risk of breast cancer compared to women with the lowest levels [17, 18].

Three decades ago Richard Stevens was the first to hypothesize that exposure to light-at-night may suppress the pineal gland production of melatonin and in turn increase the risk of breast cancer [19].

In 2007, the International Agency for Research on Cancer (IARC), which is a part of the WHO, evaluated the evidence of shift work in relation to cancer. Due to the normal procedure by IARC, the evidence from epidemiology, animal experiments, and mechanistic studies were evaluated separately, and an overall synthesized conclusion was established [20]. Only eight epidemiologic studies (three cohorts and five case-control studies) on shift work and breast cancer were available for the 2007 evaluation. The strongest evidence came from two independent prospective cohort studies on nurses from the USA, showing significantly increased breast cancer risk after over 20–30 years of rotating night shift work [21, 22]. Additional support came from four case-control studies and several studies on flight attendants with increased breast cancer risk. The latter group may both have had night work and have crossed time zones, but results may at least partly be confounded from exposure to cosmic radiation [4]. A crude meta-analysis based on the longest category of night work exposure from six studies found a relative risk of breast cancer of 1.51 (95% CI, 1.36–1.68) [23]. Overall, the evidence for an association between night work and breast cancer was credible, but bias, confounding, and chance could not be excluded (i.e., limited evidence according to IARC criteria). Over 50 animal experiments were available representing different biological aspects of melatonin and light exposure, including animal “jet lag models,” rather than “shift work models.” In particular, Blask et al. conducted a novel experiment where nocturnal light exposure was shown to decrease melatonin levels and to progress the growth of human breast cancer cells in rat models [24, 25]. Overall, there was sufficient evidence in experimental animals for the carcinogenicity of light during the daily dark period (biological night). Furthermore, experimental data supported that similar biological mechanisms occur both in animals and humans on molecular, cellular, and systemic level [26]. Taking all the evidence together, it was concluded that shift work that includes circadian disruption is probably carcinogenic to humans [27]. Subsequently, it became possible in Denmark as the only country so far to get breast cancer after long-term night shift work acknowledged as an occupational disease and get economical compensation [28].

The 2007 evaluation by IARC apparently stimulated epidemiologic research on shift work and cancer. Although there have been some studies on shift work and other cancer sites than breast cancer, in particular prostate and colorectal cancers, relatively few studies have addressed these cancers compared to breast cancer. Therefore, an overall evaluation of other cancers than breast cancer is somewhat in its early phase due to the lack of epidemiological studies.

The main aim of the present review is to describe and evaluate studies on night work and risk of breast cancer that have been published since the IARC evaluation in 2007.

Assessment of the Night Shift Work Exposure

Accurate assessment of exposure is normally the Achilles’ heel in all epidemiologic studies and inaccurate exposure may bias results. All studies on shift and night work and breast cancer risk have used different definitions of the exposure, and partly adjusted for different potential confounders, which was one of the limitations of the epidemiological studies noted at the 2007 IARC evaluation [2]. Therefore, an international group of epidemiologists gathered for a workshop at IARC in 2009 and suggested standardization of variables used in future studies [29]. In particular, it was suggested that studies should capture at least “1) shift system (start time of shift, number of hours per day, rotating or permanent, speed and direction of a rotating system, regular or irregular); 2) years on a particular non-day shift schedule (cumulative exposure to the shift system over the subject’s working life); and 3) shift intensity (time off between successive work days on the shift schedule)” [29]. Despite these recommendations, all of the recent studies on night work and cancer have used different definitions of night work concerning both the period of work within the 24-h day and the intensity of night work, i.e., consecutive working nights and periods of rest in between. Thus, many studies, including all existing cohorts use only number of years of night shift as a surrogate for dose, irrespectively that most persons do not have night work every day of the year. This inaccurate assessment of the extent of night work may underestimate risks because of the obvious different impact, within the span from a few to 31 nights per month, on physiology and thereby potentially the breast cancer risk. One exception is a large register-based cohort from Denmark based on all employees in public hospitals where objective payroll data, including time of start and end of each shift, is available from 2007 to 2012 on an individual level [30]. Because of the potential for linkage with many health-related registries in Denmark [31], this cohort will be very valuable in future decades when sufficient follow-up time for cancer becomes available.

The intention in most definitions of night work that are used in epidemiological studies of cancer is to focus on exposure to electrical light during the period of the night where melatonin normally peeks, i.e., during about 2–3 a.m. [32, 33]. A recent study, including employees mainly at hospitals showed that the proportion of shifts classified as night shifts in practice differs little in Denmark when night shifts are based on definitions including a period during the night (i.e., working time after midnight) [30]. Thus, the critical issue concerning definition of night shifts may rather be the number of consecutive night shift, e.g., during a week, which have different physiological impact on circadian disruption and cancer risk [34,35,36].

Differences in exposures to intensity and wavelength of the light, which may be very different in different occupational settings, also influence the level of melatonin suppression during the night [37, 38••]. Furthermore, individual differences in diurnal preference, which seems partly under genetic influence, result in melatonin peaking earlier for people with morning preference and later than average for people with evening preference [39,40,41,42,43,44]. Thus, the genetic component of diurnal preference or chronotype seems associated with breast cancer risk [32] via a polymorphism in the circadian gene PER3 [45] [46]. One relatively small case-control study has shown that both women with morning and evening preference have an increased risk of breast cancer associated with night shift work, but the risk is highest for women with morning preference [47]. This has, however, yet to be confirmed in larger studies.

Information on working time is normally obtained by self-reports from study participants, which potentially may be inaccurately obtained by questionnaires or interviews both in cohort and in case-control studies. Härmä et al. has recently compared objective information on working time from payroll data with questionnaire-based data on working time. Overall, there was a good correlation between self-reported shift work with night work, and permanent night shift (sensitivity and specificity over 90%), whereas shift work without night work had moderate validity (sensitivity 62%; specificity 87%). If this can be generalized to other studies, this means that self-reported information on night work in general may be valid, but may underestimate the effects in shift work without night shifts [48].

New Epidemiologic Evidence After the 2007 IARC Evaluation

In total, nine case-control studies [47, 49,50,51,52,53,54,55,56], one case-cohort study [57], and eight cohort studies have been published since the IARC evaluation [58,59,60,61,62,63]. Results and descriptive characteristics of the 18 new studies are shown in Tables 1 and 2. Including the eight studies available at the IARC evaluation [21, 22, 64,65,66,67,68,69], in total, 26 studies of night shift work and breast cancer were published in scientific journals by the end of 2016. A recent update, including 14 years of extended follow-up of two independent studies based on Nurses Health cohorts [21, 22] originally issued in 2001 and 2006, were published in early 2017 [70••].

Table 1 Cohort studies of night work and breast cancer risk published after the IARC evaluation in 2007
Full size table
Table 2 Case-control studies of night work and breast cancer risk published after the IARC evaluation in 2007
Full size table

Cohort Studies

All of the existing cohort studies, including those published prior to the 2007 IARC evaluation, were designed for other purposes than studying the association between night work and breast cancer. Thus, information on night work in all these studies is limited to relatively crude and imprecise questions on night work, which most often have been obtained several years after the establishment of the cohorts. Thus, at its best, they have information on duration of night work, but not on intensity, e.g., number of shifts during a week or month. Typically, the main shift work related question is “How many years have you worked during the night?”

Many cohort studies [21, 22, 59, 61, 70••], even without information on night shift intensity, have shown increased breast cancer risk, however, only after long-term night work.

In contrast, a recent series of three large UK cohorts on night work and breast cancer found no association [62]. The main limitations in the UK cohorts are short follow-up time (3 years) and an aging survivor population (68 years old at baseline) where women may have stopped working years before start of follow-up and consequently may be negatively biased due to severe left-truncation [71,72,73,74,75]. Thus, it has recently been shown from the large American Nurses’ Health Study (baseline 1988) that the risk of breast cancer attenuates and disappears years after cessation of night shifts. Thus, an estimated relative risk of 1.36 (95% CI, 1.07–1.78) among the nurses with at least 30 years rotation night work (mean age 60 years old at baseline) changed to a relative risk of 0.95 (95% CI, 0.77–1.17) when follow-up was continued from 1998 to 2012, thus including only nurses with post-retirement time [21, 70]. In contrast, the increased relative risk for nurses with at least 20 years of night shift work in the equivalent 19 years old younger cohort of Nurses’ Health study II (41 years old at baseline in 1989) remained increased after the similar extended follow-up period (relative risk of 2.15 (95% C.I, 1.23–3.73)). Another recent study is based on payroll information from primary hospital employees in Denmark, where females are followed up for breast cancer during 2008–2012, i.e., for a maximum of 5 years [63]. There was no evidence of an association between cumulative night shift work and breast cancer risk in this study. In addition to the relatively short follow-up time, the major limitation of this study is lack of information on working time prior to 2007, because it is highly likely that the reference group of dayworkers in the period after 2007 have been exposed to night work at earlier ages before 2007 [76], since virtually all health professionals in Denmark have night work early in their career. Null results have also been found in one cohort with crude exposure assessment [60] based on Dutch Labor Force Surveys, which did not even have information on duration of night work [60]. Also a relative large cohort from Shanghai, China (mean age 53 years old at baseline in 1996–2000) reported null results [58], both based on applying a job exposure matrix at baseline and based on questionnaires on night work obtained during 2002–2004. Cohort members were followed up for breast cancer until the end of 2007, thus, only between 3 and maximum 7 years for the sub-cohort, including self-reported night work information, which limits the statistical power of the study, including only 73 women with night work and breast cancer. Further, it has been suggested that Asian ethnicity like the Chinese is less prone to circadian disturbance [77, 78]. Finally, in contrast, two independent cohorts from Sweden that included younger participants than the null studies and have longer follow-up time reported increased relative risks for long-term night shift workers [59, 61].

Case-Control Studies

As shown in Table 2, most case-control studies observe increased relative risks for breast cancer, though they are not all significantly increased [47, 49, 50, 52,53,54, 64,65,66,67, 79, 80].

A case-control study on nurses from Denmark attempted to differentiate between associations of breast cancer and evening shifts, rotating night shifts, and permanent night shifts, i.e., normal consecutive night work, respectively [80]. Results from this study indicates no association with breast cancer and evening shifts from 3 p.m. to 11 p.m. (OR = 0.9; 95% CI 0.4–1.9), but increased risk for both rotating (OR = 1.8; 1.2–2.8) and permanent night shifts from 11 p.m. to 7 p.m. (OR = 2.9; 95% CI 1.1–8.0). This was partly replicated in a study from Spain where the category of over 1800 permanent night shifts, indicated higher OR than the similar number of rotating night shifts, 1.48; 95% CI 0.81–2.68 versus OR = 1.08; 95% CI 0.66–1.79 [55].

It has been suggested that circadian disruption and misalignment of internal clocks normally occurs only after over three consecutive night shifts [26, 34]. Two case-control studies from Norway and Denmark have focused on the importance of number of consecutive night shifts and association with breast cancer risk in nurses and find only increased risk after at least 3–4 consecutive night shifts during at least 5 years [47, 50]. This was, however, not replicated in a more recent study from France [52].

The French study hypothesized that the risk of breast cancer should be more pronounced if night work starts before the first pregnancy, when mammary gland cells are incompletely differentiated than after first full-term pregnancy. Interestingly, the relative risk for these two situations were 1.47 (95% CI 1.02–2.12) and 1.09 (95% CI 0.77–1.55), respectively [52].

Subtypes of Breast Cancer

Despite the fact that breast cancer is a group of heterogeneous diseases, it has been treated as one entity in most, but not all studies on night work. This may also have contributed to the heterogeneity in results unless night work is associated with all subtypes, which is less likely. Different ages of study participant’s that have different distributions of pre- and postmenopausal breast cancers have been suggested to have potential different etiology [81], thus may influence results on studies of night work and breast cancer. A recent study using pooled data from five case-control studies showed that it is in particular premenopausal breast cancer is increased after night work, rather than postmenopausal breast cancer [49, 52,53,54, 79, 82]. The same findings were seen in a recent study from China [56], and in the updated Nurse’s Health studies [70••]. It can be generalized that this may further contribute to explain the heterogeneity in results of studies of night work and cancer where the age distribution and menopausal status is diverse. In particular, the recent negative UK cohorts consist in particular of women, who are almost all postmenopausal [62].

The distribution of different hormonal subtypes in breast cancer may also influence results [83, 84]. This has been studied in a number of case-control studies and in a case-control study nested within cohorts, where some heterogeneity in results has been observed [21, 53, 55, 56, 63, 85,86,87]. The most consistent findings is between night work and the human epidermal growth factor positive (HER2+) breast cancers [55, 56, 63, 86], which also has shown to be associated with melatonin suppression in experimental studies [88]. Thus, four such recent studies showed OR’s in the range of 1.3–1.9, though results were only significant in one study from France [86].

Meta-analyses

Meta-analysis is a useful tool in order to make overall risk estimates of existing studies. A main requirement for conducting such analyses is, however, the existence of similar definitions of both exposure and outcome in included studies, which is not the case for any of the existing studies on night work and breast cancer [89]. Further, different criteria have been used for inclusion and exclusion of studies and each meta-analysis is thereby subject to different lists of studies. For instance, the most recent meta-analysis by Travis et al., which despite violation of normal criteria for conduction proper meta-analysis found no association when they included their own studies, was solely based on cohort studies [62], and did not include the most recent update of Nurses’ Health studies [70••]. Overall, a number of recent meta-analyses have been published from authors without previous research within this field, and with somewhat different results and conclusions [62, 90,91,92,93].

General Limitations and Confounding

In industrialized societies, most people suffer from some circadian disruption originating from social obligations, so called social jet lag. Thus, most people normally delay bedtime and advance their wake up time, which is then often compensated during the days off [94]. In a population-based study in Germany, the average level of social jet lag was about 1.5 h [95]. This means that even a group of dayworkers is not fully appropriate as a “clean” reference group, which is ideally required in epidemiologic studies, which may dilute a true increased risk from night work, if such an association exists. In the majority of studies, adjustment for potential confounders seems without major influence. If any, the confounding seems mostly weakly negative (data not shown).

In general, the more recent case-control studies have a more detailed assessment of working time in comparison with both previous cases-control studies and most of the existing cohort studies. Almost all of the case-control studies were designed with the main purpose of investigating night work and breast cancer risk. Thus, most case-control studies have information on both duration of night work and intensity, e.g., number of lifetime shifts (Table 2). Furthermore, the case-control studies have normally obtained lifetime history of working time, job-by-job for a broad range of ages, whereas cohort studies often are prone to left-truncation bias because the age at baseline of disease-free cohort members is usually relatively high in order to capture as many cases in the near future. This is particularly problematic for studies of working time because complex mechanisms of selection in and out of night work exist, which may be better captured in case-control studies than in prospective cohort studies. On the other hand, case-control studies may be subject to recall and participation bias. Therefore, evidence based on results from different well-conducted study designs is also needed in the future.

New Study Directions

Animal models included in the IARC 2007 focused on the effect of ill-timed exposure to light and aspects of melatonin, including pinealectomy. Van Dycke et al. has recently reported on an animal model of rotating shift work where a group of control mice where one group of mice were exposed to 12 h of light and 12 h of darkness circles over about 70 weeks. Another group had the same cycle of exposure to light and darkness, but each week, the period of light and darkness was reversed. Almost all mice in both groups developed breast cancer. However, the group of mice that were exposed to simulated rotating night work had a significantly decreased latency of breast cancer of about 8 weeks compared to the control group [96•]. Timing and quality of meals in shift workers is significantly different from dayworkers, and breast cancer risk is a new avenue of research which should be included in future studies [97,98,99]. Moreover, some pioneering results concerning long-term night work and epigenetic changes should be explored further [16, 100].

Prevention

Although night work yet remains to be confirmed as a cause of breast cancer, initiatives for preventive actions may be prudent [101]. In 2012, an international group suggested a few guidelines in order to prevent potential breast cancer after night work, especially by using only forward rotating shifts, limiting the period with night shift work and by limiting the number of consecutive night shifts [102]. In general, however, there is a lack of evidence-based studies on this topic, in particular concerning night work and potential health outcomes in general [103]. Thus, future in-depth understanding of mechanisms may improve evidence-based prevention [104].

Conclusion

The number of new studies on night work and breast cancer has increased over threefold during the last 10 years after the IARC evaluation in 2007. In general, the case-control studies are better in capturing at least some of the complexity of night shift work, including intensity of night work than the cohort studies. Further, all recent studies include information on most potential confounders for breast cancers, although confounding appears to be relatively low, and in most studies in a negative direction. Four recent cohort studies from the UK and Denmark, despite advances in relatively large study populations, have not provided evidence of an association between night work and breast cancer risk. These null findings may be due to limitations in the design of these studies. Overall, the observed heterogeneity in results from epidemiologic studies may at least partly be attributed to wide differences in the definitions of night work, study design, length of follow-up, left-truncation in cohort studies, lack of information on chronotype, social jet lag, and differences in the investigated populations’ menopausal status and breast cancer subtypes. Future studies on night work and cancer should use standardized definition of working time, including information on number of consecutive shifts, diurnal preference, menopausal status, and meal intake patterns. Cohort studies should include lengthy follow-up of relative young people. Overall, evidence based on results from different well-conducted study designs is also needed in the future. Finally, there is a need for studies of sites of other cancers than breast cancer.

Overall, the epidemiological evidence of an association between shift work that includes night work has increased since the IARC evaluation in 2007, and it might soon be time for a reevaluation by IARC.