Effects of Shift Work on the Eating Behavior of Police Officers on Patrol

Police officers, aged 20–67 years, working full-time on rotating shifts in the province of Quebec, Canada were contacted between September 2016 and December 2017 via official email channels following approval of both management and union representatives. Police officers who expressed interest were required to meet eligibility criteria during a phone-based screening questionnaire. Prospective participants were excluded if they regularly engaged in substantial commitments other than their police patrol duties in excess of 10 hours/week; had a medical condition or were prescribed medication that could disturb sleep or waking; or reported sleep disorders unrelated to their rotating work schedules. Those participants eligible based on these criteria underwent additional screening, including a physical assessment by a physician and assessment of blood and urine samples. Through this process, anthropometric data were collected, including the body mass index (BMI; kg/m²), waist-to-thigh ratio (WTR), waist-to-hip ratio (WHR), and estimated basal metabolic rate (BMR) [59]. Participants also completed the Munich ChronoType Questionnaire for Shift Workers (MCTQ^Shift) to derive a chronotype estimate from mid-sleep on rest days (MSF) [30].

A total of 81 out of 159 police officers who responded to the advertisement were eligible and expressed interest in participating in a larger field study. Of these, a subset of 53 officers was invited to participate in the present food study; 36 agreed to do so (Supplementary Figure S1A). From these, 31 officers contributed to the final dataset, as 5 had to be excluded due to dropout (n = 1), technical issues with equipment (n = 2) or withdrawal for medical reasons (n = 2).

Police officers participated in an observational field study in which they documented their sleep, work, and eating behavior whilst working their habitual rosters (Figure 1A). Participants provided their work rosters to the research team in advance of the study to facilitate planning. Of the 31 police officers contributing to the study, 27 habitually worked a 35-day cycle with a roster comprising series of 9- and 12-h morning/day shifts (0700–1600 h, 0700–1900 h), evening shifts (1500–2400 h), and night shifts (2230–0730 h, 2300–0800 h, or 1900–0700 h) separated by rest days (Supplementary Figure S2A). The remaining four police officers worked a 28-day cycle with a roster comprising 12-h day shifts (0700–1900 h) and 12-h night shifts (1900–0700 h) separated by rest days (Supplementary Figure S2B).

Participants were provided a smartphone (Nexus 5, LG Electronics, Seoul, Korea) and instructed to use it to photograph all food and beverage intakes on days pre-selected by the research team following examination of their work schedules. A staff member reminded participants to begin logging meals on the day prior to each nutrient collection period. For participants on the 35-day roster, these days encompassed the second rest day and the second morning, evening, and night shift in a sequence—i.e., four “day types” in total (Figure 1A, Supplementary Figure S2A). For participants on the 28-day roster, these days encompassed the second rest day and the second day shift and night shift in a sequence—i.e., three “day types” in total (Supplementary Figure S2B). For each selected day type, the nutrient analysis window began from the offset (i.e., wake time) of the main sleep period prior to the shift/rest day until the offset of the main sleep period that followed. After taking a photograph, participants were instructed to classify the meal type—e.g., “breakfast”, “lunch”, “dinner”, “snack”, or “beverage only”—using the movisensXS smartphone application (Movisens GmbH, Karlsruhe, Germany). If participants did not finish a meal, they were asked to take a second photograph of the remainder. Finally, at the end of each day, police officers were asked to complete a chart describing the type and quantity of any food supplements they had consumed (e.g., vitamins).

Throughout the work cycles, police officers wore an actigraph (Actiwatch Spectrum^®, Respironics, Philips, OR) that comprised an accelerometer, light sensor, and event marker on the wrist of their non-dominant hand to monitor sleep–wake and physical activity patterns; they were asked to press the event marker at each bedtime and rise time. The device was required to be worn at all times except when showering, swimming, or engaging in contact sports to avoid equipment damage. Multiple times each day, participants also completed timestamped questionnaires on the smartphone. These questionnaires were used to log their bedtimes, rise times, and work times (i.e., work start, work end), as well as to collect data on neurobehavioral performance and self-reported fatigue, sleepiness, and mood for another study. Participants were contacted weekly throughout the study to limit data loss and could contact a staff member at any time for study assistance. At the conclusion of the study, police officers returned their equipment with a copy of their final work schedules so the research team could cross-check and verify their actual work hours.

At the end of the study, a staff member contacted each participant to clarify and confirm the contents of their photographs. Descriptions of the food and beverage items and portion sizes were entered into a spreadsheet; these were then checked for ambiguities by at least one other staff member to ensure that the descriptions were clear. Photographs then underwent anonymization with respect to date, time, and day type. Nutrient analyses were conducted by a registered dietician of the Center of Clinical Investigations at Brigham and Women’s Hospital using the University of Minnesota Nutrition Data System for Research software [60]. The accuracy of entries into the software were verified by another staff member. Among the dietary information extracted from these assessments were total calories (kcal) and macronutrient calories obtained from carbohydrates, fats, and proteins. Consistent with a previously published methodology, meals of the same type (e.g., “snack”, “beverage only”, etc.) that were consumed within 15 min of each other were combined into a single intake event, using the timestamp of the last meal [52,53,54,55,56].

To verify the work times, self-reported work hours were cross verified with work schedules provided to the research staff and the timestamps of questionnaires completed at the start and end of the shifts. Any conflicting work-start and work-end times were investigated and corrected. After these had been confirmed, shifts were classified based on their timing. Work periods commencing between 0500 h and 0900 h were classified as morning shifts; those ending between 2200 h and 0200 h were classified as evening shifts; those that encompassed the period between 0100 h and 0500 h were classified as night shifts. To facilitate the comparability between shifts, atypical work periods longer than 13.5 h (i.e., overtime of more than half a 9-h shift) or shorter than 4.5 h (i.e., less than half a 9-h shift) were excluded from analyses. Figure 1B and Supplementary Figure S1B illustrate the number of participants contributing to each combination of valid shifts/rest days.

A standardized hierarchical algorithm adapted from a previously published methodology [61] was used to identify participants’ rest intervals (i.e., the period from bedtime to rise time) from sleep log information and timestamps, together with physical activity, event markers, and light exposure data retrieved from the actigraph in 15-s epochs. Once rest intervals had been identified, a sleep detection algorithm was used to define the sleep period onset and final offset (i.e., wake time) and apply a sleep/wake estimate for each 15-s interval between them based on the activity counts in surrounding epochs [62,63]. From these data, the primary measure was sleep duration, defined as the total amount of sleep (in hours) scored within the sleep period. When more than one sleep period occurred within a 24-h calendar day, the longest sleep period was defined as the main sleep and all others were defined as naps. For the purpose of this study, relevant sleep episodes, including naps, were those that enclosed nutrient collection days, between the start of the main sleep immediately prior to a shift/rest day and the wake time of the following main sleep period.

Data analyses and visualization were performed using R Foundation for Statistical Computing, Version 3.6.0 [64] and the packages “ggplot2” and “ggbeeswarm” [65,66]. Demographic and anthropometric measures were compared between sexes with Welch’s T-tests. Χ² tests were used to compare the proportion of meal intakes by size, time and location of consumption, and meal type. Analyses of circular data (i.e., clock times), including meal, sleep/wake, and work times, were conducted with Circular ANOVA using the R package “circular” [67]. Remaining analyses of meal frequency and composition, physical activity, and work/sleep durations were conducted with linear mixed-effects models using the R package “lme4” and “lmerTest” [68,69]. These models controlled for sex and age and included “subject ID” as a random effect. Post-hoc pairwise comparisons were made using the package “emmeans” [70]. Significance was determined at p < 0.05 for main effects and post-hoc tests, corrected for multiple comparisons using the Benjamini–Hochberg method [71].

The 31 (6 female) participants included in the analyses had a mean age of 32.1 ± 5.4 years (25–44 years; mean ± SD, range), a mean BMI of 25.0 ± 2.3 kg/m² (20.7–29.9 kg/m²), and mean experience in the police force of 8.6 ± 4.7 years (3–21 years) (Table 1). A correlation suggests overweight in this sample is associated with age (r(31) = 0.40, p = 0.026). Of the other characteristics, the only significant (p < 0.05) difference observed between sexes was for their estimated 24-h basal metabolic rate (BMR). As expected, the average BMR of male police officers (1782 ± 121 kcal, 1579–2032 kcal; mean ± SD, range) was higher than that of female police officers (1388 ± 109 kcal, 1238–1562 kcal, p < 0.001).

Of the 107 relevant data collection days obtained, 16 were removed due to atypical shifts (i.e., <4.5 h or >13.5 h) and one was removed due to insufficient data (i.e., only one meal documented), resulting in 90 valid days in the final analyses (Supplementary Figure S1B). No participant contributed more than once per day type. A Χ² square goodness-of-fit test indicated the distribution of days across rest day (n = 28, 31%), morning shift (n = 21, 23%), evening shift (n = 17, 19%), and night shift (n = 24, 27%) day type categories was comparable (X² = 2.89, p = 0.41).

The mean work start and end times (circular mean) were: 0658–1627 h for morning shifts, 1458–0002 h for evening shifts, and 2153–0751 h for night shifts (Supplementary Table S1). Start and end times significantly differed between shift types (p < 0.001), but work durations did not (p = 0.098). Activity counts from the actigraphs indicate that police officers were more active on rest days than night-shift days (292 ± 79 vs. 251 ± 64 counts/min, p = 0.017) (Supplementary Table S1). Police officers were also more active at work during morning shifts than they were at work during night shifts (280 ± 52 vs. 230 ± 52 counts/min, p = 0.003).

Sleep onset and wake times were significantly affected by day type (). Wake times prior to morning shifts (0532 h ± 7 min) were significantly earlier than on rest days (0800 ± 24 min), or prior to evening shifts (0834 h ± 26 min) and night shifts (1558 h ± 39 min). Sleep onset times at the end of the wake period also reflected the work schedules; sleep onsets following night shifts (0914 h ± 17 min) were significantly different from sleep onsets on rest days (2256 h ± 18 min), or following morning shifts (2215 h ± 15 min), and evening shifts (0111 h ± 9 min). Supplementary Table S1

Day types did not significantly differ in “day length” from wake time of the main sleep period before the shift/rest day until the wake time of the main sleep period that followed (mean range: 23.4–24.6 h, p = 0.165). Police officers spent a greater proportion of time asleep on rest days (34%) than work days (27–30%, p < 0.05), and more on morning-shift than night-shift days (30% vs. 27%, p < 0.05; Supplementary Table S1).

A total of 569 meal intakes were analyzed, distributed between rest days (n = 210, 37%), morning-shift days (n = 133, 23%), evening-shift days (n = 85, 15%), and night-shift days (n = 141, 25%). Since different numbers of participants were included for each day type (Supplementary Figure S1B), a Χ² goodness-of-fit test was conducted to test whether the total count of meals documented for each day type was proportionate to the number of contributing participants for each day type. The test indicated that the number of meals significantly differed from expected values (Χ² = 11.61, p = 0.009). Pairwise comparisons showed more meals than expected were logged on rest days relative to evening-shift days (Χ² = 10.09, p = 0.004) and night-shift days (Χ² = 5.05, p = 0.037).

On average, police officers obtained significantly more meals on rest days (7.5 ± 2.2), and significantly fewer meals on evening-shift days (5.0 ± 1.4, p < 0.001), than all other day types (Table 2). In further analyses, Χ² tests of independence showed that participants consumed a greater proportion of their meals at night on night-shift days (33%) than they did on other days (1–14%) (Supplementary Table S2). However, there were no differences between day types in terms of: the proportion of meals within different caloric ranges; the proportion of intakes classified as “main meals” (48–53%), “snacks” (27–32%), or “beverage only” (16–26%); and the proportion of meals at work (40–47%). The average size of officers’ main meals (39–41% BMR, p = 0.84) and snacks (14–18% BMR, p = 0.84) did not differ between day types.

The average estimated 24-h basal metabolic rates of participants contributing to each day type did not differ (Table 2). Participants consumed the greatest percentage of their basal daily energy requirements on rest days and the smallest percentage on evening-shift days (Figure 2). Consumption on night-shift and evening-shift days was significantly lower than on rest days (Table 2). Consumption on evening-shift days was also lower than on morning-shift days. When assessed at a macronutrient level, participants consumed more calories as a percentage of their BMR from fat and saturated fat on rest days than on evening-shift days and night-shift days. Participants also consumed more fat on morning-shift days than evening-shift days. Intake of other macronutrients did not differ between day types, although a trend was observed for carbohydrate intake (p = 0.064; Table 2).

On average, the first meal (>5 kcal) occurred between ~1 h to 2 h of elapsed time awake, but there was no significant effect of day type (p = 0.203) (Supplementary Table S3). Similarly, final meals occurred, on average, between ~2 h to 3 h before sleep onset but did not differ between day types (p = 0.303). A significant effect of day type was found for the eating window (F(3,63) = 5.65, p = 0.005) (Supplementary Table S3). It showed that the duration between first and final meals was significantly longer for night-shift days than rest days (13.86 h ± 3.14 h vs. 11.26 h ± 1.81).

Mixed-effects models showed a significant fixed effect of day type on the weighted rho-value used to measure the spread of caloric intake around the clock (F(3,84) = 8.08, p < 0.001). It showed that meals were significantly more dispersed on night-shift days than rest days or other work days (Figure 3A). Analyses of the macronutrients showed similar results (Supplementary Table S3; Figure S3).

A circular ANOVA used to assess the weighted mean times of meal intakes revealed a significant main effect of day type (F(3,86) = 29.27, p < 0.001). Mean caloric intake occurred significantly later on night-shift days (2308 h ± 74 min) than on rest (1525 h ± 29 min), morning-shift (1410 h ± 27 min), and evening-shift (1453 h ± 39 min) days (Figure 3B). Mean intake on morning-shift days was significantly earlier than on rest days. Macronutrient results were similar (Supplementary Table S3; Figure S3).

Mixed-effects analyses of the percentage of total caloric intake revealed significant interactions of day type with time-of-day for the percentage of total caloric intake (F(15,516) = 7.33, p < 0.001), and the percentages of total caloric intake from carbohydrate (F(15,516) = 6.72, p < 0.001), fat (F(15,516) = 5.29, p < 0.001), and protein (F(15,516) = 4.66, p < 0.001) (Figure 4; Supplementary Figure S4). For all day types, the distribution of caloric intake varied as a function of time-of-day. In each, the greatest proportion of calories (~30–35%) occurred between 1600–2000 h. For rest days and for morning-shift and evening-shift days, the trough of consumption occurred between 0000–0800 h. On morning-shift days, there was also a reduction of caloric intake from 1200–1600 h compared to 1600–2000 h. Night-shift days displayed a less robust diurnal rhythm of caloric intake, with the lowest proportion of total calories shifted to 0800–1600 h. There were similar rhythms at the macronutrient level (Supplementary Figure S4).

Mixed-effects analyses revealed significant interactions of day type with time awake for the percentage of total caloric intake (F(12,440) = 6.66, p < 0.001) and the percentages of total caloric intake from carbohydrate (F(12,440) = 5.60, p < 0.001), fat (F(12,440) = 4.95, p < 0.001), and protein (F(12,440) = 6.53, p < 0.001) (Figure 4; Supplementary Figure S5). The distribution of total caloric intake remained stable for the first 8 h after waking on rest days, morning-shift days, and evening-shift days (Figure 4). Caloric intake was significantly reduced between 8 to 12 h after waking on morning-shift days, before increasing again 12 to 16 h post-waking. On both evening-shift and night-shift days, the greatest percentage of calories were consumed within the first 4 h post waking. For night-shift days, this was significantly more than consumed at any other time awake (Figure 4). For evening-shift days, this was significantly more than consumed after 12 h of time awake. For all day types, the proportions of total caloric intake significantly tapered beyond 16 h of wakefulness. Similar distributions of caloric intake with time since waking were observed for the macronutrients (Supplementary Figure S5).

Mixed-effects model ANOVA revealed that the percentages of total calories consumed from carbohydrates, sugars, fats, saturated fats, and proteins did not significantly differ between the day types (Table 3). Overall, the average sources of daily calories were 44% from carbohydrate, 36% from fat, and 18% from proteins.

When comparing the percentages of total calories consumed at night (i.e., 2300–0600 h), a significantly greater percentage was consumed on night-shift days than on all other day types (Table 3). Participants also consumed a greater percentage of total calories at night on evening-shift days than they did on rest days and morning-shift days.

When comparing the percentages of total caloric intake from all meals or macronutrients at work, there were no significant main effects of day type (Table 3). Post-hoc analyses of the main effect of day type trends (p < 0.10) suggested that participants consumed more calories at work during morning shifts than at work during evening and night shifts (p < 0.05). Further, post-hoc analyses for the macronutrient trends indicate they consumed more carbohydrates at work during morning shifts than evening shifts (p = 0.029), and more fat at work during morning shifts than night shifts (p = 0.040) (Table 3).

The times at which police officers were able to eat on a given day were contingent on the sleep–wake times and work breaks permitted by their work schedule. Indeed, compared to rest days, officers had to wake up significantly earlier prior to morning shifts and began sleep later at night following evening shifts or in the morning following night shifts. As a consequence, police officers obtained significantly less sleep on work days than rest days as a proportion of the day. Associated with these altered sleep–wake times, police officers consumed their meals earlier on morning-shift days than rest days, and—likely of more importance for metabolic health—at considerably later times of day for night-shift days than other day types. As police officers’ circadian pacemakers were unlikely to have adapted to a nocturnal schedule by the second shift, these later meals would have occurred at later phases of their internal clock. The greatest proportion of daily calories (~30–35%) occurred between 1600–2000 h on all day types—i.e., after morning shifts, during evening shifts, and before night shifts. On morning-shift days, the reduction of caloric intake from 1200–1600 h coincided with the latter half of the shift and the post-work commute.

Another consequence of reduced sleep duration on work days, particularly following night shifts, was that police officers had inversely longer wake times during which they could eat [73]. Indeed, police officers had the shortest eating window (11.3 h ± 1.8 h) on rest days where the most sleep was obtained and had longest eating window (13.9 h ± 3.1 h) on night-shift days where the least sleep was obtained. Similar trends were present for the other work days. While fewer meals were eaten on night-shift days than rest days, this duration in itself is important because extending the eating window reduces the time during which bodies can enter an important fasting state that facilitates cellular repair and activates catabolic processes (e.g., the synthesis of glucose) [74,75]. Moreover, the variability in meal times and eating windows across rest days and work days is of significant concern for shift workers on rotating schedules because irregular eating patterns can also impair metabolic function [76,77].

A comparison of the number of meals documented for each day type revealed that more meals were consumed on rest days than evening-shift and night-shift days. However, none of the days significantly differed in terms of the proportion of meals that were snacks or only composed of beverages. As police officers did not contribute to all work days and rest days in the final dataset, we standardized caloric intake during these meals relative to their estimated daily BMR (see methods). This ensured we could evaluate differences in total intake between day types according to their individual requirements. The estimation of the BMR did not significantly differ between the police officers who contributed to the different day types.

Unlike much of the literature about dietary patterns in shift work, we observed that total caloric intake by police officers significantly varied between day types. Specifically, police officers consumed more calories relative to their metabolic requirements on rest days and morning-shift days on average (170%–184% BMR; Table 2) than on evening-shift and night-shift days (130–143% BMR). The increase in total caloric intake on rest days and morning-shift days appears to be due to an increase in calories from fats (p = 0.004) and carbohydrates (p = 0.064). Gill and Panda estimated that participants forget to log approximately 10% of their food/beverage or water events, meaning that caloric intake is likely higher than reported [52]. However, the false negative rate may be lower in the current study as participants only had to log meals for a single day at a time, rather than multiple weeks, and were reminded by staff the preceding day. Assuming that low to moderate physical activity each day would require a further ~40–60% in caloric intake [78] relative to BMR to account for additional energy expenditure, it is possible police officers overate relative to their metabolic requirements on rest days and morning-shift days and/or underate on evening-shift and night-shift days. However, without information about their actual total energy expenditure on each day, it is not possible to categorically state whether police officers were in positive (or negative) energy balance on any particular day of the study. Accelerometry data from the actigraphs suggest that police officers were most physically active on day-oriented schedules and least active on night-oriented schedules (Supplementary Table S1), so it is possible they were able to maintain energy balance. Nonetheless, the consumption differentials between each of the work days and rest days provide useful information about the police officers’ dietary choices and a potential area of focus for weight management.

In contrast to what we observed in our sample, Lennernas et al. found that shift workers consumed more on a 12-h day shift day than on rest days [32]. However, this was a consequence of these shifts always occurring on Sundays, and thus being accompanied by traditional large meals of meat and potatoes. Apart from the 12-h shift, consumption on rest days did not differ from the other morning, evening, or night shifts. It is not clear why there are differences in total caloric intake between day types in the current study, yet relatively small differences in other studies [32,39].

In terms of the overall macronutrient composition of meals, there were no significant differences between day types. On average, the sources of daily calories across all day types were proportionally 44% from carbohydrates, 36% from fats (including 12% saturated fat), and 18% from proteins. These quantities fell slightly short of the recommended levels of carbohydrates (45–65%) and slightly exceeded the recommended levels of both fat (20–35%) and saturated fat (<10%) [78], but are consistent with what has previously been reported in shift workers [39,79].

In other studies, the primary consequence of night work is the increase in meals and caloric intake consumed during the night relative to the daytime [32,39], particularly in the form of snacks [37]. Our results are consistent with these previous findings. In another study of the temporal eating patterns of shift workers, Shaw et al. found that 6% of meals occurred at night when working days, in contrast to 30% when working nights [39]. We found that the proportions of meals at night were similarly related to day type—4.5% for morning shifts, 14.1% for evening shifts, and 32.6% for night shifts (Supplementary Table S2). Caloric intake was distributed in a similar fashion, with 1.2%, 9.7%, and 30.3% of total intake consumed from 2300–0600 h at night on morning-shift, evening-shift, and night-shift days, respectively. For perspective, the proportion of calories between these hours on night-shift days, presumed to have been consumed in misalignment with the circadian pacemaker due to the low rate of adaptation, constituted a 14-fold increase relative to the amount consumed in the same period on rest days. Nonetheless, the overwhelming majority (~70%) of calories consumed on night-shift days were still consumed during the day.

Although the duration of work periods did not differ between shift types (Supplementary Table S1), there was a statistical trend (p = 0.083) in which police officers tended to have a lower caloric intake during night shifts (43.4%) than during morning shifts (52.5%). This pattern of greater caloric intake during morning shifts than night shifts is consistent with the observations of Lennernas et al. [32]; furthermore, it comports with findings that circadian rhythms of hunger—and its associated hormone, ghrelin—peak in the biological evening and reach a trough during the biological night [24,80,81]. The circadian rhythms of hunger may also explain the marked peaks in caloric intake between 1600–2000 h for all day types, including night-shift days, despite an overall dampening of intake across the wake period on night-shift days.

The strengths of the current study are that we were able to obtain information about the diets of police officers across the variety of representative days experienced in their work schedules and did so using a novel smartphone-based approach with timestamped photographs. The advantages of monitoring nutrition in this way are that it provides precise and reliable information about food timing and allows for the objective quantification of the various macronutrients depicted in the photographs by dieticians. Moreover, it is less susceptible than other measures to the limitations of human memory (food content and timing) and the lack of feedback or human interaction make it less prone to social desirability bias (over-reporting positive behaviors and under-reporting negative behaviors), and the observer effect (altering behavior in response to being monitored). The successful implementation of the method in this and other studies [52,53,54,55,56] demonstrates a viable path for the future assessment of long-term dietary patterns in different demographic populations and shift working occupations (e.g., nursing).

While acknowledging these strengths, it is important to consider that this study only captured a snapshot of eating behavior for each shift. As participants did not contribute more than one day of data for each day type, it is not possible to know the extent to which individual eating patterns may vary from day-to-day per shift type. Thus, our results assume data were collected on ‘typical’ days. We tried to ensure this was the case in a couple of ways: first, we selected the second day in sequences of shifts for data collection to avoid transitions between day types. Additionally, we excluded days from the analyses that were unusual, such as those with work periods that were unusually long or short. It is worth emphasizing that the police officers were all shift workers on rotating schedules; we did not compare shift workers with non-shift workers. Thus, it is feasible that the size and composition of meals consumed in our study differ from the habitual dietary patterns of permanent day workers and permanent night workers.

Furthermore, the police officers in this study are a select group of shift workers and may not represent the majority of shift workers in other occupations (e.g., those that are more sedentary or involve working indoors) whose daily routines may differentially affect their dietary choices/options and physical activity [31]. Indeed, the tendency for greater body mass and cardiovascular symptoms in nurses who work shifts [82,83] may be related to their reports of (i) reduced time and energy to prepare healthy meals, (ii) few options available at night, and (iii) limited opportunities to eat while at work [84,85]. These factors may increase the likelihood of choosing convenient and unhealthy “fast food” options [31]. Compared to shift workers primarily based at their desk or indoors, police officers out on patrol have different work conditions that can affect both accessibility to healthy food and levels of physical activity during the course of their duties.

It should be noted that, as a group, the police officers in our study were healthy, with no medical or psychiatric conditions based on their questionnaire and physical exam during screening. The officers underwent yearly fitness examinations as part of their professional responsibilities. While half the participants had a BMI that fell within the “overweight” range, their waist-to-thigh ratios — a reliable anthropometric measure of abdominal obesity—were within ranges for low diabetes and cardiovascular disease risk [86,87]. This is in contrast to previous reports of police officers having high rates of overweight and obesity, elevated blood pressure, and type 2 diabetes [57,58]. One factor that may explain this discrepancy is that we excluded participants who reported sleep disorders unrelated to their work schedule. In their study of North American police officers, Rajaratnam et al. [57] found that sleep disorders such as obstructive sleep apnea are highly comorbid with these health issues. Consequently, the dietary patterns and energy expenditure of this cohort may not reflect the behaviors of all shift-working police officers, who potentially eat different quantities of food at different times. Future research evaluating the efficacy of meal-timing interventions on the health of shift workers will require the inclusion of a broader cross-section of participants.