Article Text
Statistics from Altmetric.com
In 2007, the International Agency for Research on Cancer (IARC) convened an expert working group who examined all relevant information and concluded that “Shiftwork that involves circadian disruption is probably carcinogenic to humans (Group 2A).”1 As a key element of their classification, IARC judged that there was “sufficient evidence in experimental animals for the carcinogenicity of light during the daily dark period (biological night) [emphasis added].”1 However, while IARC identified a key role for ‘biological night’ in animals, it is more difficult to define—and work with—a biological night in humans. This editorial suggests ‘how’ we can do this in practice to arrive at an epidemiological measure of circadian disruption.
Remarkably, the key link in the ‘probable’ chain of causation between shift work and cancer, that is, circadian disruption, is nowhere defined in the IARC monograph.1 However, we would expect it to occur in individuals who do work—usually associated with light—at times when their body anticipates—and is prepared for—sleep, that is, at their biological night. The possibly underlying cause of cancer may be conceptualised as consequences of chronodisruption, that is, a relevant disturbance of the circadian organisation of physiology, endocrinology, metabolism and behaviour.2 To understand the mechanisms by which disruptions of activity and rest and of wake and sleep states may contribute to disease will lie in the domain of laboratory studies. However, epidemiologists should also consider the concept of chronodisruption by comparing how much shift work-associated times overlap with workers’ biological nights.3
Biological night has been described as the time when the circadian clock promotes sleep.4 Importantly, humans vary as to when they tend to be awake and asleep in the course of 24 hours. This is evinced and captured by the chronotype which depends on factors such as genes, sex, age and environmental light. The biological night, therefore, differs for people of different chronotypes and cannot be delineated by civil time alone. To exemplify why this distinction is important, note that for a late chronotype with biological nights of 03:00–11:00,3 a work shift from 18:00 to 02:00 may not be during their biological night at all. For the general population, ∼5% early (biological night: 22:00–06:00), 75% intermediate (biological night: 01:00–09:00) and 20% late chronotypes have been observed.5
No epidemiological study thus far has defined the biological night on an individual basis, taking into account chronotype. As one example, in a recent paper in the OEM, Gyarmati and colleagues investigated night shift work and stomach cancer risk in the MCC-Spain study.6 The authors defined night shift work when individuals worked, at least partly, between 24:00 and 06:00 civil time. At a first glance, this constitutes a reasonable time focus as IARC experts concluded that “among the many different patterns of shiftwork, those including nightwork are the most disruptive for the circadian clock.”7 However, depending on workers’ chronotype, work outside the commonly assumed nighttime window may lead to significant circadian disruption.3 Note that Gyarmati et al wrote that MCC-Spain lacked chronotype information for stomach cancer cases. Their closing sentence reads “Further analysis including data on chronotype may contribute to better understanding of the relationship between night shiftwork and stomach cancer risk.” We certainly agree. From the view of chronobiology, we cannot readily equate work during nights which are defined by civil time alone with ‘shiftwork that involves circadian disruption’. Instead, we should focus on work during an individual's biological night.8
In practice, to study ‘shiftwork that involves circadian disruption’, we need to consider both when individuals start and end their work and when individuals start and end their biological night. Capturing both sets of information appears straightforward but may be challenging nevertheless.8 For example, with regard to shift timings, there is a wide range of shift systems,9 and many of these are irregular from week to week. In addition, we need to capture shift work over many years, and shift work systems will change over time even in one job, let alone in the different jobs individuals might have during their working life. In fact, no important prospective cohort study—including the Nurses Health Study, the Million Women Study or the EPIC studies—has collected this level of detail of shift timing information. Ultimately, only industry-based or community-based studies may provide shift details of appropriate completeness and accuracy over years or decades in an independent fashion.
Similarly, it is not easy to identify study individuals’ biological night as we need to capture chronotype information. While the Morningness–Eveningness Questionnaire10 and the Munich-ChronoType Questionnaire (MCTQ)5 are generally accepted tools, they are quite long and their use is difficult within large-scale epidemiological studies which already include complex and overlong materials. In search of a simplified approach, Hansen and Lassen11 assessed diurnal preference (morning, evening, neither) from a single question. The authors wrote “that answers to a similar question on diurnal preference correlated well with answers on more comprehensive questionnaires.” But justifying the Danish single question approach with reference to the MCTQ512 can be disputed. It is not known how that single MCTQ question would have been answered without the MCTQ's preceding questions at the time. In addition, chronotype may change over time.13 By how much in how many appears empirically unclear. In effect, assessments at one point in time which do not take age into account may be insufficient. Ultimately, obtaining necessary details regarding chronotype may only be possible in cohort studies which collect data over time.
Lastly, we must quantify the overlap between shift work-associated times with a worker's biological night to determine how much chronodisruption a person of a particular chronotype experiences when working at particular civil times. Every shift should be weighed by the individual chronotype and summed up to arrive at individual cumulative doses of chronodisruption. This implies, for example, that a shift from 20:00 to 08:00 would create more chronodisruption in someone who has a biological night of 22:00 to 06:00 (overlap of work with such early chronotype's biological night=8 hours) than in someone who has a biological night of 03:00–11:00 (overlap of work with such late chronotype's biological night=5 hours).
Methodologically, chronotype is supposed to act as an effect modifier3 assuming that chronodisruption, and possibly associated relationships with cancer, should vary across different chronotypes given the same shifts. Several studies published after IARC2007 have dealt with chronotype by stratifying analyses of night work and cancer.11 ,14–16 However, analyses via stratification and interaction modelling cannot assume that all night shifts have the same—and that day shifts have no—potential for circadian disruption or chronodisruption irrespective of the individual chronotype.17 Taken together, when assessing chronodisruption, we can incorporate chronotype information in a way that takes account of the interaction structure at the design stage already.3
Looking back at research after IARC's 2007 classification, which is critically based on chronobiology, it comes as a surprise18 that merely a handful of studies included chronobiological information.11 ,14–16 ,19 Similar to Gyarmati and colleagues, other authors have mentioned that chronotype information was unavailable. Importantly, as long as we lack demonstration that chronotype information is relevant, traditional epidemiological studies such as the one by Gyarmati et al remain valuable. However, with such definition of night shift work, traditional studies will neither necessarily, nor exclusively, identify all ‘shiftwork that involves circadian disruption’. Disconcertingly, if we do not falsify the hypothesis that chronotypes are relevant, we cannot rule out that traditional epidemiology may miss direction and magnitude of risks associated with work at the biological night.
Finally, the rationale as to how we should conduct chronobiology-based shift work epidemiology could go way beyond cancer research. Most readers will expect that shift work is detrimental to human health. If so, why do not we see higher, let alone a doubling of, risks of some end points associated with shift work which is regularly required as one component or standard of causation and for compensating an occupational disease? A candidate answer to this riddle could be that epidemiology hitherto failed to focus on the biological night and thus failed to capture relevant doses of circadian disruption or chronodisruption.
Acknowledgments
The editorial was developed during LF's work as a visiting professor at the University of Cologne. TCE wishes to thank the Medical Faculty of the University of Cologne for supporting LF's visiting professorship. LF is supported by fellowships from the National Health and Medical Research Council and the Cancer Council Western Australia.
References
Footnotes
Contributors TCE conceived the work and wrote the first draft. TCE, JVG and LF developed, revised, approved and are accountable for the work.
Competing interests None declared.
Provenance and peer review Not commissioned; internally peer reviewed.