The effect of shift work on eating habits: a systematic review

by This is the first systematic review examined the association between shift work and eating habits. The results suggest that dietary quality can be altered by shift work. Shift workers showed changes in meal pattern, skipping more meals and consuming more food at unconventional times. They also seem to consume foods enriched with saturated fats and lower amount of fibers. The effect of shift work on eating habits: a systematic review. Scand J Work Objective This systematic review aimed to evaluate the association between shift work and eating habits. Methods The protocol was registered in PROSPERO (number 42015024680). PubMed, EMBASE, Scopus, and Web of Science were searched for published reports. Of 2432 identified articles, 33 observational studies met the inclusion criteria. Their methodological approaches were assessed using the Newcastle-Ottawa Scale. Data were extracted using a standardized form. Studies were considered to have a low or a high risk of bias according to a percentage score of quality. Results The majority of the studies presented a quality score of <70% and a high risk of bias for comparability, sample selection and non-respondents. Shift workers show changes in meal patterns, skipping more meals and consuming more food at unconventional times. They also show higher consumption of unhealthy foods, such as saturated fats and soft drinks. Conclusions This review suggests that shift work can affect the quality of workers’ diets, but new studies, espe cially longitudinal studies, which examine the time of exposure to shift work, the duration of the workday and sleep patterns, are necessary to confirm this association.

Shift work is characterized by schedules that differ from conventional working hours (for example, work hours in Brazil are commonly between 08:00-09:00 and 17:00 -18:00 hours) (1,2). Night work and work that occurs continuously for 24 hours with rotating shifts (2,3) also fall within this definition of shift work. During the last few decades, the proportion of shift workers has been increasing (2,4). This work system has been described as an important risk factor in the etiology of metabolic disorders and chronic diseases (1,(5)(6)(7)(8)(9)(10).
The causal mechanisms of this association are not fully elucidated, but observational studies indicate that changes in work schedules result in physiological and behavioral changes in shift workers (11)(12)(13)(14). These workers suffer from a disruption of the circadian rhythm and therefore experience hormonal alterations due to being awake at the time biologically reserved for sleep (11,15,16). In addition, in shift work, the reorganization of nighttime and daytime activities involves changes in lifestyle, including eating habits (7,14,17,18).
Working in shifts can affect eating habits in a variety of aspects. Epidemiological studies have shown differences in relation to the consumption of calories and macronutrients and the quality of the food eaten by shift workers (7,11,17,19). Changes in meal patterns, ie, meal frequency, types of meals and times for meals (20), have also already been previously associated with changes in work schedules and/or sleep deprivation (17,21,22).
Systematization of the knowledge about the influence of shift work on eating habits is critical in understanding the relationship between work shifts and metabolic disorders and chronic diseases. In this sense, three published review studies on this topic were found. One narrative review described dietary characteristics among shift workers but did not compare dietary characteristics with those of non-shift worker controls (23). A systematic review investigated dietary habits in conjunction with the nutritional status of workers (5). In addition, a metaanalysis study focused only on energy intake assessment (13). Therefore, it is necessary to carry out review studies with a high-quality systematic approach that investigate the association between the timing of work shifts and dietary habits, considering eating habits in a broad way, including quantitative and qualitative assessments of diet, as well as the meal patterns of workers.
Thus, the aim of this study was to perform a systematic review of evidence of the association between shift work and eating habits. We hypothesized that shift workers have irregular eating times, skip main meals and have a higher consumption of snack meals and foods rich in sugars and fats than daytime workers.

Methods
A systematic literature review was carried out to summarize the results of observational studies that evaluated the impact of different work shifts on the eating habits of individuals. The protocol of this review was done in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (24) and the Meta-Analysis Of Observational Studies in Epidemiology Check List (MOOSE) (25) and was registered in PROSPERO (number 42015024680). Because it was a literature review, no ethics committee approval was required. The PICO (Population, Intervention, Control and Outcome) strategy was used to construct the research question (table 1). The articles were selected according to the following inclusion criteria: (i) have as an outcome the evaluation of eating habits (consumption of calories, nutrients, food groups or meal patterns); (ii) compare individuals in different shifts (rotating or night vs. regular or day shift workers); (iii) be an original article; and (iv) have full text available for reading.
The searches were conducted in PubMed, Scopus, EMBASE, and Web of Science from February to June 2017. Articles published by September 2017 were included. The descriptors are defined according to Medi-cal Subject Headings (MeSH) for searches performed in PubMed and according to EMBASE subject headings (Emtree) for searches performed in EMBASE. The research strategy used in PubMed is presented in table 2. Additional studies were identified in the references of articles selected.
Two investigators independently performed the initial selection of articles by reading all of the titles and abstracts. Endnote X7 software was used to manage the selection of articles. In a second step, the two reviewers performed the complete reading of the articles according to the inclusion criteria. In case of divergence between the two researchers, a third researcher was consulted for a final decision. Afterwards, the data from the studies that met the criteria were extracted, and the following information was considered: sample characteristics, losses (%), exposure classification (shift work), assessment tool of eating habits, classification of food consumption and main results. Authors were contacted when the information was not available in its entirety (26)(27)(28).
The Newcastle-Ottawa Scale (NOS) instrument was used to assess the quality of studies (29). In crosssectional studies, we used an adaptation to specific criteria for sampling, non-response rate and statistical tests employed, as proposed by Herzog et al (30). The instruments include items divided in three areas: selection, comparability and outcome. All items received a maximum score of 1 when the evaluated criteria were met and 0 if they were not identified. In cross-sectional studies, a maximum of 2 points could be attributed when the best criterion was identified in the questions that assess the definition of exposure and control for confounding factors (comparability) and outcome measurement. Two researchers carried out all evaluations independently. In case of disagreement, the third researcher was consulted.
The percentage score of quality evaluation was set as follows: the total score of each study divided by the maximum score applied to the checklist (10 for crosssectional studies and 9 longitudinal studies) and finally multiplied by 100. In a second stage, a general evalua-  Exposure: "Shift systems" OR "shift system" OR "Shift work" OR "Shift-work" OR "Night shift work" OR "Night shift-work" OR "Night-work" OR "Night worker" OR "Night workers" OR "shift workers" OR "shift worker" OR "Sleep disorder" OR "Sleep disorders" OR "Circadian Rhythm" OR "Circadian Rhythms" OR "rotating shift work" OR "rotating shift-work" Outcome: "Meal frequency" OR "meal frequencies" OR "meals" OR "meal time" OR "mealtime" OR "mealtimes" OR "meal times" OR "eating frequency" OR "eating frequencies" OR "eating episodes" OR "meal pattern "OR" meal patterns "OR" eating pattern "OR" eating patterns "OR" behaviors eating "OR" eating behavior "OR" dietary pattern "OR" dietary patterns "OR" dietary habits "OR" diet habit "OR" diet habits "OR" food intake "OR" foods intake "OR" nutrient intake "OR" nutrient intake "OR" caloric intake "OR" caloric intake " Limit "Humans" tion of the quality of the articles was performed for each item evaluated in the instrument. Studies that received 1-2 points for each domain were considered to have a low risk of bias, and those that received 0 points were considered to have a high risk of bias.
Differences in macronutrient intake were found in 9 studies, when shift workers or rotating shift workers were compared to daytime/fixed workers. (11,19,38,39,42,43,45,46,52). If, on the one hand, 3 studies observed a lower protein intake among night workers (39) and workers with rotating shifts (19,42), on the other hand, 1 study found increased protein intake by night workers (11). In 3 studies, lower carbohydrate intake was found for night workers (11,39,42), and 2 found a greater consumption of carbohydrates (38,45). Lower consumption of fiber among night shift workers (42) and rotating shift workers (43) were found in 2 studies. Regarding lipids, shift workers had a higher Table 3. Summary of population, design characteristics and methodological approaches of the studies sorted (N= 33). Check list The New Castle-Ottawa Scale (27) and its version adapted to cross-sectional studies (28 intake of total lipids in 2 studies (11,38) and a lower consumption of total lipids in 2 others (39,42). Five studies evaluated the type of fat consumed (19,42,43,45,52), three studies (43, 46, 52) observed a higher consumption of saturated fat among rotating shift workers, and one study (42) observed lower consumption of saturated fat among night workers. The evaluation of micronutrient intake was performed in only 3 studies (39,42,43), of which 2 found a lower consumption of calcium in rotating shift (43) or overnight workers (39). Lower potassium intake was also observed among night shift workers, and a lower intake of vitamins A and B1 and iron (39) was observed in rotating shift workers aged 20-29 years (43).
Food habits were evaluated in 9 studies through diet quality scores or dietary patterns (6,17,19,22,32,36,37,54,55). In a study by Barbadoro et al (6), working in a rotating shift system was inversely associated with an increased cardiovascular risk score, whereas in 2 other studies, rotating shift work was positively associated with unhealthy eating patterns (22) and inflammatory potential (54), characterized by the consumption fatty and fried foods (tables 5 and 6). Finally, 12 studies investigated meal patterns, comparing shift workers or rotating shift workers to daytime/ fixed workers. The number of meals consumed was evaluated in 6 studies (17,21,33,35,45,49) and were higher among night shift workers in 2 studies (17,21). Of the 10 studies that compared the types of meals (17,21,33,35,41,45,46,49,52,53), 7 had the following associations: night and rotating shift workers skip more meals at lunch and breakfast in the 4 studies that evaluated this behavior (17,33,35,52). Of the 6 studies that evaluated the time of food intake, 5 (17,21,33,41,46) showed higher consumption of food at night among night shift and rotating shift workers. Nevertheless, at least 1 study showed higher consumption of breakfast at inappropriate times for night workers and dinner for daytime workers (17).

Summary of main results
A summary of the associations in these studies is described in table 7. When the associations found in each study between work shift and eating habits are analyzed together, it is possible to conclude that shift workers have a higher consumption of saturated fats and soft drinks than daytime workers. Likewise, shift workers skip more meals and have a higher meal intake at night. With regard to energy intake and micronutrient consumption, the results of the studies are varied, and most studies found no associations. Most studies that assessed scores and dietary patterns also found no associations.

Discussion
This systematic review included results from 33 observational studies (30 cross-sectional, 2 longitudinal, and 1 case-control) that investigated associations between shift work and eating habits. Thus, the evidence found should be evaluated in consideration of the limitations of cross-sectional study designs. The evidence found suggests that shift work (night and rotating) can lead to changes in workers' eating habits when the pattern of meal consumption was evaluated. Aside from the skipped meals, the studies included in this review have shown that shift workers show differences in the distribution of food intake, with increased food consumption during the night among night shift workers than among rotating or daytime workers. In addition, there seems to be a trend toward greater consumption of foods rich in saturated fats and soft drinks among shift workers (22,42,43,54).
Studies of workers and the general population have shown that changes in the pattern of meals are an independent risk factor for weight gain (56), glucose intolerance, insulin resistance, dyslipidemia and obesity (57)(58)(59). Skipping breakfast is associated with excess weight and changes in metabolic markers, regardless of total caloric intake (60)(61)(62). In addition, the increased consumption of food and calories during the night has also been associated with metabolic changes and weight gain (62)(63)(64)(65). Experimental studies have shown that animals fed during the period considered inactive have greater weight gain and increased body fat, regardless of the type of diet provided (66)(67)(68). In humans, increased food consumption in the later hours of the day was positively associated with overweight among healthy men and women (65,69,70). Additionally, individuals undergoing a weight loss program who eat at later hours show less weight loss than those who have the same diet but consume food earlier in day (59).
The causal pathway that links changes in eating patterns to metabolic disturbances can be explained by circadian rhythm disruption. The production of hormones and metabolic function is synchronized with the circadian rhythmicity system (36,(68)(69)(70). Experimental (25,70) and epidemiological studies (60,61,64) show that healthy individuals with changes in biological patterns of sleep and wakefulness may have reduced production of leptin, increased ghrelin and increased insulin resistance. Thus, changes in the regulation of body temperature, digestion, energy metabolism and hormonal responses are experienced by individuals who eat during the rest period (63,68,70,71).
Changes in the consumption of food groups and nutrients that make up an unhealthy food pattern were also identified among shift workers. An increased intake of saturated fats and soft drinks was identified in shift workers' diets. Shift workers and permanent night workers -already exposed to chronobiological and hormonal changes as a result of their working hours -may present additional risk factors for the development of metabolic alterations due to the excessive consumption of these products, which contributes to excess weight, obesity and the development of non-transmittable chronic diseases (72). In addition, this finding suggests the recognition of eating habits as possible contributing risk factors for the development of chronic diseases. The occurrence of cardiovascular events has been positively associated with shift work. In a literature review and meta-analysis of observational studies, a higher risk of infarction and coronary events was found among workers performing shift work activities (73).
There is insufficient evidence of an association between the consumption of energy, protein, carbohydrates and micronutrients and shift work. These results reinforce the findings by Bonham et al (13) demonstrat- ing that there are no differences in the energy intake of rotating/night shift workers when compared to daytime workers. Moreover, only three studies evaluated the micronutrient intake of workers, demonstrating the need for more studies with this approach.
The results of this review should be interpreted in the light of the methodological characteristics of the included studies. One limitation of the cross-sectional study design is that the exposure and outcome are simultaneously assessed. In this sense, it is not possible to affirm if workers adopted these eating habits after or before beginning to work in the current shift pattern.
The articles were found to show low methodological quality, mainly in relation to controlling for confounding factors. In most studies, the association between eating habits and shift work was not the main objective, so the analyses did not consider possible confounding factors. Sleep pattern, for example, is an important mediating factor of circadian and metabolic disorders (1,5,65,74). Thus, studies evaluating the relation between working shifts and eating habits should consider the aspects related to quality and duration of sleep. Another criterion evaluated with a high risk of bias was the selection of workers. Samples for convenience, without description of the sample population, as well as a lack of information on the number and characteristics of nonrespondents, allow for selection bias.
Nevertheless, methods of assessing the outcome varied among studies. The choice of instrument to assess food consumption should be considered a predefined objective, for example, quantitative assessment of nutrient intake, assessment of the consumption of food or food groups, or evaluation of dietary patterns (75). In this case, studies included in this review presented collection methods adequate for their objectives. The quantitative evaluation of feeding was performed from dietary records of two to seven days or 24HR. On the other hand, the evaluation of food group consumption and dietary patterns was carried out using FFQ validated for the population studied. Except for one study (54), all included studies repeated applications (two or more) of dietary questions about the previous 24HR and dietary records, thus reducing the limitations of these methods to assess the individuals' usual intake. The evaluation of the usual diet is important, since the effects of inadequate food intake can occur only after prolonged exposure to shift work.
However, it is known that the self-report of food     intake is susceptible to memory and information bias, particularly underreporting (75); additionally, there was no information about the validity of the meal pattern questionnaires and the simple questions used in several studies. Although there are limitations, food questionnaires (food recalls, food records and FFQ) are the most widely used methods for assessing food consumption in observational studies due their simplicity and low cost (75). In the quality assessment of the studies, the outcome measurements were classified as having a low risk of bias due to the characteristics of the quality checklist adopted. (30) Differences in exposure ratings should be considered. The routine of workers subjected to fixed night or day work may differ from that of workers who perform their activities on a rotating shift scale, implying different feeding, sleep and leisure times between groups and limiting comparability between them. In addition, some studies were not clear in the shift system rating (35,40,50,53,54), and none of the studies presented an instrument for the classification of shifts, for example, payroll, registration forms or access to information recorded at the workplace. It is known that the self-report of the working shift system has low sensitivity and moderate specificity, compared with objective data (76).
The investigation of eating habits by the studies included in this review was conducted under different contexts, such as food availability at the work place, place and time for worker meals, type of company and work (traveling/driving workers, blue or white collar workers, hospital or industrial workers), and cultural habits (food differences in Eastern and Western coun-tries). These differences can explain the heterogeneity in the methods and results of the studies.
Despite the division of outcomes into nutritional composition or food and food groups in most articles, the results summary and evaluation allow a broad understanding of dietary habits. Likewise, the collection of information that characterizes an eating pattern (both the quantity and quality of food and the temporal distribution of eating across the 24 hours in a day) allows the adequate identification of risk behaviors and interventions that may modify them.

Concluding remarks
This was the first systematic review of literature to investigate the association between shift work and eating habits. Although quantitative differences in calorie intake are not influenced by work shift, the timing of meals consumed and the pattern of eating seem to be different in shift workers. Shift workers also have a higher consumption of foods rich in saturated fats and sugar. The alterations in sleep-wake cycles in addition to unhealthy food habits have a possible mediating role in the relation between shift work and chronic diseases.
The results of this review highlight the need for attention to the quality of food these workers eat and the need for future specific studies regarding the association between shift work and eating habits. The methodology of these studies should consider the use of longitudinal designs, the statistical power and representativeness of the sample, objective methods for work shift measurement, the assessment of food habits based on different indicators (calories, nutrients, food and meal patterns) and the control of possible confounding factors, such as sleep characteristics, time of exposure to shift work and duration of the work shift. These studies are fundamental in proposing nutritional guidelines specific to the population of shift workers.

Ethics
The authors declare no conflicts of interest.