Occupational and Environmental Medicine | 2021
Is a JEM an informative exposure assessment tool for night shift work?
Abstract
In this issue of Occupational and Environmental Medicine, Fernandez et al investigate the role of maternal night shift work in occurrence of urogenital anomalies in offspring. The authors inferred potential exposure to night shift work by applying a jobexposure matrix (JEM) to mothers’ recorded occupations in the Australian Perinatal Registry. This study’s assessment of night shift work across various occupations, with nurses reported separately from other types of workers, adds valuable knowledge on a rarely studied outcome. The authors acknowledge that the lack of individuallevel information on shift schedules precluded their ability to assess differences in duration and/or intensity of night shift work, which may variably interfere with reproductive function and recommend ‘investigation in a sample with more detailed exposure information’. This is an important recommendation given the widely recognised complexity of assessing exposures in epidemiological studies of night shift work. ‘Night shift work’ refers to work that occurs during the regular sleeping hours of the general population. As such, it is not an exposure in the traditional sense, but rather a proxy for a complex combination of exposures and circumstances leading to circadian disruption. In addition to a wide variety of schedule characteristics, these include light at night (LAN), phase shift, sleep disturbances, disrupted social behaviours and personal habits and other workplace hazards, many of which are strongly interrelated. Moreover, personal characteristics such as age, sex and chronotype or diurnal preference may affect individual responses to night shift work. In a recent effort to address this complexity, the International Agency for Research on Cancer (IARC) Monograph evaluation of the carcinogenicity of night shift work considered various domains of night shift work exposure (eg, exposure intensity/duration/temporality, potential for reference group contamination, shift start/end times) across the epidemiological studies reviewed. Greater weight was placed on the most informative studies, based on methodologic considerations that included night shift work assessment quality (most notably the potential for misclassification). Objective individuallevel data on night shift work exposure were considered most accurate, in part due to its suitability for multidimensional exposure assessment. Individuallevel data can be obtained from company records or collected with questionnaires, interviews or diaries wherein workers selfreport their work schedules along with other relevant exposures and characteristics. When using a JEM to translate job titles into exposure to night shift work, no individual characteristics are being considered. Fernandez et al acknowledge that ‘use of occupational title to impute night shift work involves a degree of misclassification of exposure...’. Such misclassification is an inevitable and wellrecognised limitation of JEMs since heterogeneity between workers within the same job is ignored by design. 7 The extent of this heterogeneity varies by the agent considered and may be larger for night shift work compared with other exposures (eg, dusts) for which JEMs have been frequently used. There are two major considerations in developing a JEM for night shift work: first, it has to be decided at the job level whether someone works at nighttime. Second, various exposures and circumstances associated with night shift work (LAN, other workplace hazards, etc) could be assigned at the job level. Both steps are prone to misclassification as these aspects of night shift work are strongly related to individual characteristics and heterogeneous within jobs. Exposure to night shift work in occupations with high proportions of night shift workers, such as nurses, might be relatively well estimated by a JEM that assigns probability by job, while occupations with less prevalent exposures will be less well characterised. However, even within nurses, night shift schedules and related exposures will vary between persons as well as within persons over time. Further, the variety of tasks, work environments and occupational hazards encountered by different types of nurses cannot be distinguished in many job coding systems that form the basis of JEMs. As such, a JEM is inherently limited in its ability to assign night shift work with sufficient exposure contrast and nuance to assess aetiologies of health effects in the general working population. With these considerations in mind, it is reasonable to ask: ‘Is a JEM an informative exposure assessment tool for night shift work?’ The JEM used by Fernandez et al was sufficient to detect a relationship between maternal night shift work and occurrence of urogenital anomalies in offspring. This JEM, however, was developed with data from Australian females, with limited applicability to other populations. It has been demonstrated that male workers may have very different patterns of night shift work, and circumstances of this exposure may also vary widely by region, time period and life stage. Lessstudied groups, including gig economy workers and nonCaucasian populations, are also not well captured in this and other night shift work JEMs focused on Western working populations. Opportunities to optimise night shift work JEM performance, such as incorporation of quantitative information on agents that may impact circadian disruption, should also be considered. For instance, noted variability in measured LAN exposures across occupations and work environments 11 could be leveraged to develop grouping variables with optimal exposure contrast or to test the validity of subjective measures such as selfreported light exposures or expert assessment. A recent Canadian study observed large betweengroup LAN variance relative to withingroup variance for various exposure metrics and groupings in healthcare, suggesting that highlevel exposure groupings informed by quantitative measurements could be an effective means to characterise individual exposures in epidemiological studies of night shift work. Although a JEM may be the only solution to assess night shift work in general population studies when no detailed job histories or exposure data are available, we advise caution in the application of JEMs. Due to Berkson error, a groupbased approach such as a JEM will in principle not lead to biased risk estimates, but to a considerable loss of power and consequently reduced precision of the risk estimate. A JEM cannot Institute for Risk Assessment Sciences, Utrecht University, Utrecht, The Netherlands Government of Canada, Charlottetown, Prince Edward Island, Canada