Workplace cafeteria and other multicomponent interventions to promote healthy eating among adults: A systematic review

The objective of this review is to evaluate evidence for the effectiveness of workplace cafeteria and other supporting multicomponent interventions to promote healthy eating and reductions in health risks among adults. We conducted an electronic search in EMBASE, CINAHL, EconLit, Ovid, Cochrane, Web of Science and PubMed for English-language articles published from 1985 to July 2019. Studies were original articles reporting the results of workplace cafeteria interventions to promote healthy eating and reduction in health risks. Outcomes were classified as changes in fruit and vegetable intake, health risk indicators, dietary intake, and food sales. Interventions were categorized as interventions targeting food quality or quantity, targeting price, targeting food choice at point of purchase, targeting improved supply, targeting client’s information, education or motivation and targeting organization policies. Behavioral change conditions used in interventions were identified using the COM-B system of behavioral change. Results were presented in a narrative summary. A total of 55 studies out of 6285 articles were identified for this review. Several studies used multicomponent interventions and the most featured interventions included interventions targeting food quality or quantity, targeting client’s information, education or motivation and targeting food choice at point of purchase. There is evidence that workplace cafeteria and other supporting multicomponent interventions resulted in higher intake of fruit and vegetables, improved dietary intake, improved health outcomes and healthy food sales. The findings of this review have the potential to inform future cafeteria-based and other supporting multicomponent workplace health interventions. The review protocol was not registered in a repository.


Introduction
More than 39% of the world's population is classified as overweight and 13% as obese (World Health Organisation, 2018). Obesity increases the risk of developing multiple diseases including cardiovascular disease, hypertension, dyslipidemia, type 2 diabetes, stroke, osteoarthritis, and some cancers World Health Organisation, 2018) Prevention of obesity is an international public health priority, given the critical influence of obesity on health and well-being. In 2015, excess body weight contributed to 4.0 million deaths and 120 million cases of disability-adjusted life-years among adults globally (Collaborators et al., 2017). In addition, the associated health care costs of obesity is on the rise. In 2014 in the US, the average spending attributed to obesity was $1901 per single obese individual, accounting for $149.4 billion nationally (Kim and Basu, 2016).
The food environment, incorporating the availability, accessibility, cost, quality and promotion of certain types of food, is a major determinant of dietary intake (Glanz et al., 2005). An unhealthy food environment contributes to unhealthy eating patterns (Elbel, 2011;Marteau et al., 2012;Schwartz et al., 2012). The modification of the food environment has the potential to promote and encourage healthy actions and can be used as a basis of workplace health interventions (Engbers et al., 2005). Workplaces are sedentary settings and places where energydense foods and beverages are commonly available (Anderson et al., 2009). From the economic lens, there is a growing concern about the economic burden of obesity in the workplace, induced by costs associated with absenteeism, sick leave, disability, injuries, and healthcare claims (Popkin et al., 2006). Nonetheless, on the positive side, the worksite provides a strategic setting for implementing programs to promote healthy eating, since employees spend up to 60% of their waking hours at the worksite (Engbers et al., 2005). The worksite can thus reach a large proportion of adults, including those unlikely to engage in a preventive health behavior program (Gorman et al., 2013). However, several factors could impede workplace health promotion initiatives, including worksite readiness, and intervention implementation (Wolfenden et al., 2018). Several systematic reviews have been conducted that evaluate the effectiveness of worksite health promotion trials (Engbers et al., 2005;Geaney et al., 2013;Ni Mhurchu et al., 2010). However, results from one review found that there are few studies that focused on the impact of food environmental modifications on dietary intakes and that the few studies containing an environmental component obtained inconclusive results (Anderson et al., 2009). Besides, it proves challenging to filter out successful intervention components in changing dietary behaviour (Schliemann and Woodside, 2019).
With a high proportion of adults around the world working in the formal workplace setting, it is of great interest to examine the food environment in the workplace to inform the development of health promotion initiatives. Hence, this systematic literature review aims to identify and assess the effectiveness of workplace cafeteria and other supporting multicomponent interventions to promote healthy eating. This review is positioned differently from other reviews as it reports outcome measures to improve healthy eating at worksites; changes in fruit and vegetable intake, health risk indicators, diet and food sales. Moreover, it catalogues interventions and sub-interventions as cafeteria and supporting non-cafeteria interventions and identifies behavioral components within interventions for translation into intervention success.

Search strategy and procedures
We used the preferred reporting items for systematic reviews and meta-analyses (PRISMA Checklist) to guide this systematic review (S1 File) (Moher et al., 2015). We searched multiple databases including EMBASE (general medicine), CINAHL (nursing & allied health), EconLit, Ovid, Cochrane, Web of Science and PubMed from 1985 to July 2019. MeSH search terms included: (1) Setting-based: cafeteria, canteen, school, workplace, worksite, campus, industry; (2) Intervention-based: nutrition, diet, dietary intervention, health promotion, primary prevention, health behaviour, health education, food, program evaluation (S2 File). We searched the citations of sentinel papers for additional sources. We included peer reviewed intervention studies published in English. Inclusion criteria were: (a) targeting adult employees aged ≥18 years; (b) non-drug and non-surgical interventions aimed at modifying the food environment and (c) intervention delivered at a workplace cafeteria (front and/or back of house), including studies with noncafeteria interventions implemented out of the cafeteria space. We excluded interventions involving (a) vending machines, kitchenettes or food trucks, (b) studies that reported results of hospital staff, patients and visitors, and university staff and students collectively, (c) interventions focused on eating disorders, (d) intervention studies that evaluated commercial weight-loss programs or products, (e) studies only involving the delivery of nutritional advice/education to employees, (f) pharmacological (drug-based studies) and clinically based interventions, observational and modelling (analytical methodology) studies aimed at improving health outcomes of participants. Studies had to report the effect of workplace cafeteria interventions and other supporting multicomponent interventions on changes in (a) fruit and vegetable intake, (b) health risk indicators (body mass index (BMI), blood pressure, serum cholesterol levels, blood glucose levels), (c) dietary intake (macro or micronutrient) or (d) food sales, such as the sales of healthy food. Interventions were catalogued as interventions targeting food quality or quantity, targeting price, targeting food choice at point of purchase, targeting improved supply, targeting client's information, education or motivation and targeting organization policies (S3 File). Using the COM-B system of behavior change (Michie et al., 2011), essential conditions for behavioral change: capability, opportunity and motivation were identified in interventions to translate intervention success (Table 3 S5 File). The methodological heterogeneity of the studies precluded meta-analysis and subsequently, a narrative summary of each study's characteristics and findings is presented. We imported all papers (title and abstracts) into an endnote database and removed duplicates. Two researchers (AN, CJ) screened the titles and abstracts, and full paper if necessary, separately and independently using a screening verification checklist. Any disagreements and unsure studies regarding inclusion were resolved by discussion with the third researcher (AS) until consensus was reached.

Data extraction
A copy of the full text of papers were obtained for each of the included studies. The screening checklist was re-applied in assessing the content of the paper. Studies not meeting the review inclusion criteria were excluded; however, studies meeting the inclusion criteria and belonging to the same trial was included. Two reviewers in parallel, (AN and CJ) independently extracted information from all 55 studies using the Data Abstraction Form published by the Guide to Community Preventive Services  to classify and describe key characteristics of the intervention. The Guide to Community Preventive Services data collection instrument and procedure for systematic reviews balances the flexibility for evaluating papers with different study designs and intervention types with the need to ask specific questions to maximize validity and reliability, providing a structured format for reviewing paper content and quality . A third reviewer (AS) double-checked 20% of the extracted studies for accuracy of data extraction. Thereafter, extraction results were compared for agreement and differences regarding data extraction were resolved by discussion until consensus was reached by all reviewers.

Quality assessment
Quality of study execution included an evaluation of five categories of threats to validity; study population and intervention descriptions, sampling, exposure and outcome measurement, data analysis, interpretation of results and other biases (S4 File), based on the Guide to Community Preventive Services guide which allows for the evaluation of different study designs with questions to evaluate a general concept. . All studies that met the inclusion criteria were assessed by the two reviewers independently for their methodologic quality (S6 File). The reviewers scored the item as positive (+) if the item was met, negative (− ) if the item was not met, and unclear (?) if insufficient information was provided. The total quality score was calculated by counting the number of items scored positively. Studies with none or one limitation were classified as good, 2-4 limitations as fair and 5 or more limitations as limited . Results were compared for agreement and differences regarding the quality score were resolved by discussion until consensus was reached by the reviewers.

Study selection
The PRIMSA diagram showing the literature search and selection process is presented in Fig. 1. Electronic database searches generated 6285 potentially relevant references. After screening the title and abstract, 6112 articles were excluded; 1486 duplicates and 4626 did not meet the inclusion criteria. Following the screening of the full text, 120 articles further did not meet the inclusion criteria. We added two hand A. Naicker et al. Preventive Medicine Reports 22 (2021) 101333 searched articles (Iriyama and Murayama, 2014;Uglem et al., 2013). A total of 55 articles conducted from 1994 to July 2019 were retrieved for detailed evaluation. Table 1 summarizes the study characteristics of the 55 papers reviewed. The papers included a range of different study designs; 23 randomized controlled trials, 13 non-randomized trails 14 pretest--posttest design and 5 time series. Many papers reported on more than one outcome. Papers classified by outcome yielded 17 papers on changes in fruit and vegetable intake, 16 papers on changes in health risk indicators, 21 papers on changes in dietary intake and 24 papers on changes in food sales. The duration of the intervention delivery ranged from 3 weeks to 5 years. Studies were conducted across multiple countries; 26 studies were conducted in the USA, 19 in Europe, 3 Japan, 2 in Brazil; and one each in Chile, Mexico, New Zealand, Australia and Taiwan. Studies were conducted in the private and public workplace settings ranging from government, factory, manufacturing, research, military, shipping, hospital, finance, farm, sports club, university and education. Eight studies took place in multiple worksites, and 15 studies did not indicate the worksite type. The study sample sizes ranged from 26 to 5695 employees and education level, ethnicity and male and female ratio varied among studies. Twenty-four studies used a single component intervention strategy, while 31 studies used multicomponent intervention strategies. Thirty-three studies used interventions targeting food quality or quantity, 12 studies used interventions targeting price, 24 studies used interventions targeting food choice at point of purchase, 5 studies used interventions targeting improved supply, 31 studies used interventions targeting client's information, education or motivation, and 5 studies used interventions targeting organizational policies. Table 2 and Table 3 S5 File provides a summary of studies reporting changes in fruit and vegetable intake, health risk indicators, dietary A. Naicker et al. Preventive Medicine Reports 22 (2021) 101333 intake and food sales as outcomes.

Changes in fruit and vegetable intake
There is evidence that workplace cafeteria and other supporting multicomponent interventions resulted in a higher intake of fruit and vegetables at the workplace. While five studies used single component interventions, 13 studies used multicomponent interventions in which 13 studies featured cafeteria-based interventions and five studies used both cafeteria and non-cafeteria interventions.
Four studies used multicomponent interventions. One study featured the interventions targeting food quality and quantity, targeting price and targeting client's information, education or motivation by offering healthy canteen choices, a free fruit program and information resources respectively (Lassen et al., 2011), one study featured the interventions targeting food quality and targeting price by offering free healthy takeaway meals (Lassen et al., 2012), and one study featured the intervention targeting food quality and targeting client's information, education or motivation by offering a salad bar, increasing vegetable dishes and providing information about the health benefits of a diet rich in fruit and vegetables through posters (Uglem et al., 2013) and one study featured the interventions targeting food quality and quantity, targeting food choice at point of purchase, targeting price and targeting client's information, education or motivation by introducing low energy dense foods and education thereof (Lowe et al., 2010).

Changes in health risk indicators
While two studies used single component interventions, 14 studies used multicomponent interventions to affect changes in health risk indicators. Of these interventions, three studies used cafeteria-based interventions, while 13 studies used a combination of cafeteria and noncafeteria interventions. In general, at least half of the studies had the expected benefits on health outcomes.
Blood pressure: Seven studies reported the effect on blood pressure. Out of these, four studies reported a significant reduction in systolic and diastolic blood pressure (Cook et al., 2001;Goetzel et al., 2010;Inoue et al., 2014;Leighton et al., 2009); two showed no significant difference Geaney et al., 2016); whereas one showed a significant increase in blood pressure (Engbers et al., 2007). Using a single component intervention targeting food quality, a large reduction in blood pressure was observed after one year of the Mediterranean diet; SBP decreased by 13 mmHg and DBP decreased by 15 mmHg (Leighton et al., 2009) and in another study SBP decreased by 5.6 mmHg and DBP decreased by 7.6 mmHg through a Japanese style healthy lunch (Inoue et al., 2014). Two studies used multicomponent interventions with a significant reduction in systolic and diastolic blood pressure. Goetzel              n/a n/a n/a -Energy intake decreased by 310 kJ, fat intake lowered by 4.3 g, cholesterol lowered by 19 mg/dL, and sodium by 106 mg when both full and reduced sized entrees were offered (P < 0.0001).

Limited
(continued on next page) A. Naicker et al. ✓Targeting food quantity: IG 1: smaller portion (2/3 the size of the existing portion) was offered in addition to the existing portion and proportional pricing. IG 2: smaller portion was added to the assortment and value size pricing (that is, a lower price per unit for large portions than for small portions). Duration: 3 months.
Fair   ✓Targeting food choice at point of purchase: Phase 1 was a 3-month color-coded labeling intervention (red = unhealthy, yellow = less healthy, green = healthy). Phase 2 added a 3-month choice architecture intervention that increased the visibility and convenience of some green items. Duration: 9 months.
Limited (Vyth et al., 2011) ✓Targeting price: Reduced price of salad bar purchases by 50%. The subsidy was publicized through an email to all employees and by a large poster in the cafeteria. Duration: 1 month.
n/a n/a n/a n/a -Daily salad bar sales in March averaged 83% higher than sales averaged for other months (P = 0.008 ✓Targeting food choice at point of purchase: The first phase was a traffic light color-coded labeling system: healthy items (labeled green) and unhealthy items (labeled n/a n/a n/a n/a -Labeling decreased all employees red item purchases by − 11.2% (95% CI: − 13.6%, − 8.9%; P < 0.001) and increased green purchases by 6.6%  ✓Targeting food quality: 7 entrees from the cafeteria were modified to low total fat to < 30% of energy and sodium to < 1000 mg per serving and with nutrient information available. Modified standardized recipes and marketing of modified entrees was developed. ✓Targeting client's information, education or motivation: Nutrient information was displayed on a large sign for all modified entrees being served that week. Duration: 7 months.
n/a n/a n/a n/a -No significant differences were observed in sales data.-No significant changes in overall acceptability were found for any entrée.
✓Targeting food choice at point of purchase: Dissemination of information on traffic light labelling. Phase 2: implementation of the traffic light labelling in the buffet. The labeling included red (unhealthy/ stop), yellow (moderately healthy/wait) and green traffic light labels (healthy/ go). Duration: 11 months.
n/a n/a n/a n/a -Proportion of customers who reported positive attitudes towards traffic light labelling increased by 12% (P < 0.01).-Proportion of buffet customers whose chose green light entrées increased by 23% (P < 0.001)-Red-light entrees choice decreased by 42% (P < 0.001).
Limited (Levin, 1996) ✓Targeting food choice at point of purchase: Poster on low-fat entrée selection and heart shaped labels were placed next to 3 targeted entrees (bean burritos, potato and chili burritos and a turkey, lettuce and tomato sandwich) on the menu board. Duration: 7 months.
No intervention in CG. n/a n/a n/a -The sales of the targeted low-fat entrees increased significantly at 6 weeks X 2 = 50.24; p < 0.001) at the IG cafeteria.
Limited n/a n/a n/a Limited (continued on next page) A. Naicker et al. n/a n/a n/a n/a -The proportion of respondents that identifiedhealth and nutrition as being an important factor in making their food or beverage choice increased by 20% (p = 0.004).-The proportion of respondents that reported looking at nutrition information increased by 18% (p < 0.0001).-Respondents who reported noticing the labels at the time of their purchase bought a higher proportion of green and lower proportion of red items compared to respondents who did not notice the labels (p < 0.001).
Limited (Stites et al., 2015), Location: Pennsylvania, USA Pre/Post. Hospital participants with a BMI of at least 25.0 kg/ m 2 , ate 3 lunches a week at the cafeteria and had access to a computer at work. Participants diagnosed of unstable hypertension, dyslipidemia or coronary heart disease, whose medical therapy changed in the last 3 ✓Targeting point of purchase: The online pre-ordering system was designed to allow employees to order their lunches hours in advance of mealtime while viewing the nutrient content of the food choices. Daily and weekly specials rotated on a 4-week cyclic menu. ✓Targeting price: Participants were provided Delayed treatment group CG.
n/a n/a n/a -The treatment group purchased lunches with   ✓Targeting food quality or quantity: Provided a total of 6 FV (such as fresh fruit, salads or salad sandwiches) and non-sugar-sweetened drink products for sale at their club canteen. Substitution of high fat/ energy products with low fat/energy products and introduce other 'healthier' products for sale. ✓Targeting food choice at point of purchase: Clubs were required to ensure at least 75% of non-alcoholic drinks in the canteen fridge were non-sugar-sweetened beverages and were positioned in the upper half of the fridge. Clubs were to ensure FV and non-sugarsweetened drink products were displayed within view of consumers at all times. ✓Targeting price: Pricing strategies were encouraged to ensure that FV and nonsugar sweetened drink products were priced competitively compared to similar less healthy products.✓Targeting client's information, education or motivation: Promotional strategies to improve the physical environment No intervention in CG. CG received printed material on topics unrelated to trial outcomes.
n/a n/a n/a -Post-intervention, clubs receiving the intervention reported a significant increase in the availability of FV products (OR = 5.13; 95% CI:  n/a n/a n/a n/a -Fruit and salad bar purchases increased markedly, p = 0.0001. Limited (Mazza et al., 2018)  Calorie-only labels. n/a n/a n/a Participants exposed to PACE labels purchased 40.4 fewer calories (p = 0.002), and participants exposed to calorie-only labels purchased 38.2  All 6 sites received intervention at different periods: Usual product availability (no intervention in CG) in site 2 for period 1, in site 3 for period 1 and 2, in site 4 for period 1, 2 and 3, in site 5 for period 1, 2, 3 and 4 and in site 6 for period 1, 2, 3, 4 and 5. Each period = 2 weeks.
n/a n/a n/a -A reduction of 6.9% total daily energy purchased from targeted food categories for all sites (95% CI: − 11.7, − 1.7; p = 0.044  ✓Targeting food choice at point of purchase: Labelling all cafeteria products for which such information was available with their calorie content (e.g., "250 Calories") displayed in the same font style and size as for price. Duration: 17 weeks.
Good (Vasiljevic et al., 2019)  FV fruit and vegetable; BL Baseline; IG intervention group; CG control group; F/U follow up; n/a not assessed; WC waist circumference; SBP systolic blood pressure; Diastolic blood pressure; METs metabolic syndrome; BMI body mass index; EAP employment advisory board used the interventions targeting food quality, targeting food choice at point of purchase and targeting client's information, education or motivation by changing the menu to promote healthy eating with point of choice prompts supplemented with staff counselling (Goetzel et al., 2010); and Cook used the interventions targeting food quality, targeting food choice at point of purchase and targeting client's information, education or motivation by including low-fat meal options, offered water as a beverage, introduced point of choice messages promoting fruit and vegetables and installed nutrition displays in the cafeteria (Cook et al., 2001). Body Mass Index: Out of ten studies that reported BMI (Brehm et al., 2011;Cook et al., 2001;Ferdowsian et al., 2010;Fernandez et al., 2015;Geaney et al., 2016;Goetzel et al., 2010;Iriyama and Murayama, 2014;LaCaille et al., 2016;Mishra et al., 2013b;Thorsteinsson et al., 1994), five studies showed a significant reduction in BMI Fernandez et al., 2015;Geaney et al., 2016;Iriyama and Murayama, 2014;Mishra et al., 2013b). Three multicomponent intervention studies showed a significant reduction of greater than 1 kg/m 2 Geaney et al., 2016;Mishra et al., 2013b). Using the intervention targeting food quality, targeting price, targeting food choice at point of purchase and targeting client's information, education or motivation Geaney reduced saturated fat, sugar and salt in meals, increased fruit and vegetables, discounted fruits, strategically positioned healthy alternatives and had monthly group nutrition presentations (Geaney et al., 2016), Mishra using the intervention targeting food quality and targeting client's information, education or motivation implemented a low-fat plant-based diet in addition to weekly nutrition classes in the cafeteria (Mishra et al., 2013b) and Ferdowsian using the intervention targeting food quality and targeting client's information, education or motivation offered low-fat vegan options and group presentations . Two multicomponent intervention studies showed a small significant reduction (less than 1 kg/m 2 ) (Fernandez et al., 2015;Iriyama and Murayama, 2014). Fernandez included reduced sodium and calorie meals, fruit and vegetable subsidies, chef training workshop and brochures on nutrition (Fernandez et al., 2015). Iriyama included healthy meals in the menu and offered nutrition counselling (Iriyama and Murayama, 2014).
Waist circumference: Waist circumference was reported in six studies (Cook et al., 2001;Ferdowsian et al., 2010;Geaney et al., 2016;LaCaille et al., 2016;Leighton et al., 2009;Levin et al., 2010); three of them showed a significant reduction Leighton et al., 2009;Levin et al., 2010). A common intervention among these studies was the intervention targeting food quality or quantity were healthy meals were offered. Two of the studies also featured the intervention targeting client's information, education by raising awareness on healthy eating among workers.
Glycated haemoglobin: One multicomponent intervention study reported change in HbA1c (%) with a 0.7% reduction through a low-plant based diet in combination with weekly classes (Mishra et al., 2013b) whereas another multicomponent intervention study showed a significant reduction in fasting blood glucose through the improvement of the nutritional value of foods served in the cafeteria (Brehm et al., 2011).
Metabolic syndrome: One multicomponent study targeting improved supply and physical activity reported a decrease in the prevalence of metabolic syndrome by 9% among participants after implementing a cooking course to chefs of the cafeteria (Hjarnoe and Leppin, 2013).

Changes in other dietary intake
While three studies used single component interventions, 17 studies used multicomponent interventions to affect changes in dietary intake. Among these interventions, 13 studies used cafeteria-based interventions and seven studies used a combination of cafeteria and noncafeteria interventions. There is evidence that changing the food environment resulted in improved dietary intake at the workplace.
Total fat: Four studies (Bandoni et al., 2011;Berkowitz et al., 2016;Geaney et al., 2010;Levin et al., 2010;Vanderlee and Hammond, 2014) reported a significant decrease in total fat intake out of which two studies had a greater than 15 g reduction in total fat intake (Geaney et al., 2010;Levin et al., 2010). Both studies used multicomponent interventions targeting food quality or quality and targeting client's information, education or motivation, one study included low fat vegan menu options and provided cooking demonstrations ; another restricted food high in fat, limited cooking methods with oil and provided nutrition information (Geaney et al., 2010).
Saturated fat: Five studies reported a decrease in saturated fat intake (Berkowitz et al., 2016;Brehm et al., 2011;Geaney et al., 2010Geaney et al., , 2016Levin et al., 2010;Vanderlee and Hammond, 2014). Common to these five studies was the intervention targeting food quantity or quality, one study introduced low fat vegan menu options , two reduced foods high in fat (Geaney et al., 2010(Geaney et al., , 2016, two conducted taste tests to modify healthy meals (Geaney et al., 2016) and one introduced reduced size entrees (Berkowitz et al., 2016).
Fiber: Eight studies reported a significant increase in fiber intake (Bandoni et al., 2011;Emmons et al., 1999;Ferdowsian et al., 2010;Inoue et al., 2014;Lassen et al., 2011;Levin et al., 2010;Mishra et al., 2013aMishra et al., , 2013b. Four studies had a common intervention targeting food quantity and quality to increase fruit and/or vegetables consumption through a low-fat vegan menu Levin et al., 2010;Mishra et al., 2013aMishra et al., , 2013b, one study labelled healthy food (Emmons et al., 1999), one study conducted culinary workshops for canteen operators (Bandoni et al., 2011), one study offered a Japanese style lunch with increased vegetables (Inoue et al., 2014) and one study had a free fruit program (Lassen et al., 2011).
Sugar products: One study provided cafeteria staff with a healthy cooking course, resulting in reduced intake of sugar products (Hjarnoe A. Naicker et al. and Leppin, 2013).
Whole grains: One study resulted in an increased consumption of whole grain bread by improving the whole grain content of bread by 50-100% and fiber content of 4-7 g/100 g at meals (Uglem et al., 2013).

Changes in food sales
From the 24 studies that assessed changes in sales of healthy food, 17 studies used single component interventions while seven studies used multicomponent interventions. Among these interventions, 20 studies used cafeteria-based interventions, while four studies used a combination of cafeteria and non-cafeteria interventions. Eighteen studies using interventions targeting food choice at point of purchase increased sales of healthy foods (Chen et al., 2017;Jeffery et al., 1994;Kottke et al., 2013;Levin, 1996;Levy et al., 2012;Mazza et al., 2018;Sonnenberg et al., 2013;Stites et al., 2015;Thorndike et al., 2019Thorndike et al., , 2014Thorndike et al., , 2012van Kleef et al., 2012;Vasiljevic et al., 2018Vasiljevic et al., , 2019Viera et al., 2019;Vyth et al., 2011;Wolfenden et al., 2015). Thirteen of the 24 studies reported significant increase in sales of healthy food and beverages. Seven studies used traffic light labelling (Chen et al., 2017;Levy et al., 2012;Mazza et al., 2018;Sonnenberg et al., 2013;Thorndike et al., 2019Thorndike et al., , 2014Thorndike et al., , 2012, and two studies used healthy symbol labels (Levin, 1996;Vyth et al., 2011) of which one significantly increased fruit sales but had no impact on the sale of sandwiches, soups and salads using the healthy symbol (Vyth et al., 2011). Four studies with significantly increased sales used interventions targeting price. One study increased sales of healthy food through the reduction of salad bar prices by 50% (Kottke et al., 2013), one study offered meal vouchers (Stites et al., 2015), one study offered competitive pricing of healthy drinks (Wolfenden et al., 2015) and one study reduced the price of fruits and salad by 50% (Jeffery et al., 1994).

Quality assessment
The assessment of the quality of included studies was impeded by incomplete reporting, and consequently, an unclear risk of bias judgement was reached for some domains. Six out of 55 studies were graded as good quality studies, 14 studies as fair quality and 35 studies as limited quality.

Discussion
The results of this systematic review demonstrate that cafeteria interventions and supporting non-cafeteria interventions at worksites promote healthy eating and influence health-related behaviors among adults. There is evidence that workplace cafeteria and other supporting multicomponent interventions resulted in a higher intake of fruit and vegetables, improved dietary intake, improved health outcomes and improved healthy food sales at the workplace. Several studies used multicomponent interventions, and the most featured interventions included interventions targeting food quality or quantity, targeting client's information, education or motivation and targeting food choice at point of purchase.
In this review, sixteen out of 18 studies demonstrated a mild to moderate effect in the increase of fruit and vegetable intake. Of those that were effective, most studies used interventions targeting food quality or quantity and client's information, education or motivation. Mechanisms for increasing fruit and vegetable intake included expanding fruit and vegetable availability and the provision of informational material on healthy eating. Likewise, two other reviews reported a positive impact on fruit and vegetable consumption through multiple component intervention strategies (Hendren and Logomarsino, 2017;Ni Mhurchu et al., 2010). In this review, at least half of the workplace cafeteria and other supporting multicomponent interventions had the expected benefits on health outcomes. Fourteen out of 16 studies that evaluated change in risk factors demonstrated a positive effect on either blood pressure, BMI, weight, WC, lipid, glycated haemoglobin or metabolic syndrome. Most studies used a combination of interventions, targeting food quality or quantity and targeting client's information, education or motivation. Mechanisms to affect changes in health risk indicators included introducing healthier cafeteria foods with reduced fat, vegan options, and lifestyle education. In this review, there is evidence that workplace cafeteria and other supporting multicomponent interventions result in improved dietary intake at the workplace. Eighteen out of 20 studies that evaluated changes in dietary intake, reported a significant effect on total fat, saturated fat, fiber and total energy intake. Most studies used a combination of interventions targeting client's information, education or motivation and food quality or quantity. Mechanisms for reducing fat with positive changes in dietary intake included low-fat vegan menu options, restriction of food high in fat, limiting cooking methods requiring oil and offering low-fat meal options. In this review, 13 out of 24 studies reported a significant increase in the sale of healthy food and beverages using environmental level changes; labelling and pricing. Most studies used cafeteria-based interventions targeting food choice at point of purchase. The review by Al-Khudairy, on choice architecture intervention to improve dietary behavior found that there was no strong evidence for the effect of pricing and on labelling alone on behavioral change (Al-Khudairy et al., 2019). However, interventions including the availability and proximity element were generally reported to be successful in changing behaviour (Al-Khudairy et al., 2019).
Given that most studies reported positive results using either all or one to two behaviour conditions, it is important that when selecting interventions, it has to be mapped to the behaviour target for intervention success. In general, the included studies were of fair to limited quality. Six studies were graded as good quality. The strength of studies could have been comprised due to the inherent limitations of a worksite setting and adherence to interventions. Moreover, our evaluation of the quality of studies was impeded by incomplete reporting.
This review has several strengths and limitations. We did a comprehensive search, covering more than 20 years of research including all types of worksites which improves the generalizability of the findings, however it is possible the search did not identify all studies published. Furthermore, the review study search was restricted to studies published in English and excluded unpublished studies. We assessed the quality of the studies using a standard quality assessment tool, with the built-in flexibility of assessing the quality of different study designs. The primary limitation of this review was the heterogeneity of the study designs, outcomes and outcome measures among studies which limited data pooling to perform a meta-analysis, hence limiting the direct comparison of studies to quantify the results to assess the effectiveness of specific interventions. Reporting the results by intervention type should be considered for future research to highlight the exact effect by intervention type to promote healthy eating and reductions in health risks. Several trails produced multiple papers; hence it is suggested that different papers that belong to same trail be reported together.

Conclusion
The review has the potential to inform future workplace health interventions in tackling workplace obesogenic environments and promoting positive dietary behavior changes. Understanding the components and processes included in such interventions has implications to inform employers and implementers about intervention options, components, format, duration and opportunities that exist to improve the health of a workforce. Future research should standardize the intervention assessment tool, outcome measures as well as evaluate the sustainability of the interventions in terms of cost and acceptability of interventions by employees. This will improve the quality of evidence available and allow for thorough assessment to identify the most effective interventions and implementation strategies. Multicomponent interventions, specifically interventions targeting food quality or quantity, interventions targeting client's information, education or motivation and interventions targeting food choice at point of purchase have the potential to produce positive health related behaviors at worksites.

Funding sources
This