The findings of this systematic review suggest that worksite interventions are effective in improving some measures of dietary behaviour. Effect sizes are variable but are generally small, although decreases of up to 9% in total dietary fat and increases up 16% in daily fruit and vegetable intakes have been reported. However, worksite intervention research has typically been methodologically weak and many studies have not included appropriately matched control groups, meaning reported effects may be due to trial participation rather than the actual worksite intervention programme. The use of self-reported dietary outcomes in most studies is a particular cause for concern because reporting bias due to dietary education makes it probable that effects on diet are over-estimated.
Our findings are fairly consistent with two recent systematic reviews of worksite weight loss interventions on body weight [11
]. Benedict and Arterburn reviewed 11 intervention studies published between 1995 and 2006 and reported that intervention groups lost -0.2 to -6.4 kg more body weight than controls over follow-up periods ranging from two to 18 months [12
]. Anderson et al reviewed 47 intervention studies published between 1966 and 2005, and a meta-analysis of a sub-set of nine RCTs produced a pooled effect estimate of -2.8 pounds of weight loss (95% confidence interval -4.6, -1.0) over 6-12 months of follow-up [11
]. Thus, it appears that worksite health promotion interventions also have positive effects on employee body weight but effect sizes are small.
A 2005 review of 13 worksite programmes with environmental changes concluded there was strong evidence for an effect on dietary intake but inconclusive evidence for an effect on physical activity and no evidence for an effect on health risk factors [34
]. The purported strongest evidence was for diet despite the fact that all dietary outcomes were self-reported rather than objectively measured. Similarly dietary outcomes in our review were predominantly self-reported, except in three of the 16 studies where sales data were used to supplement self-reported dietary changes [28
]. There is an urgent need for future worksite dietary intervention studies to include objective measures of dietary behaviour and environments. Examples of such objective measures include body weight, biological risk factor levels such as blood cholesterol, canteen and/or vending machine sales data, and nutritional analysis of foods available at worksites. Similar recommendations have been made with respect to evaluation of worksite physical activity interventions [14
]. Future studies should also consider assessing dietary intake outside the workplace because of the potential for compensatory behaviours elsewhere.
Previous reviews have highlighted the lack of long-term data on the effect of worksite health promotion programmes on health and economic outcomes [12
]. Although some studies in our review had a relatively long duration of follow-up (up to 2.5 years), none reported effects on economic outcomes. Assessment of health and economic outcomes in worksite health promotion interventions should be a priority for future research, particularly given the advent of statistical methods that facilitate estimation of effects of changes in nutrition-related risk factors on burden of disease [10
], and cost-effectiveness of interventions [37
]. A recent review of the effects of worksite interventions on body weight reported that such programmes appear cost-effective and have the potential to boost profits of employers by increasing employee productivity and reducing medical care and costs [11
]. However, robust evidence is still lacking.
The conduct of worksite-based research studies is clearly challenging. It frequently proves difficult to combine the need for academic rigour with the practicalities of delivering a community-based intervention that must meet employer and employee needs, often within short timeframes and constrained budgets. Nevertheless it is important that rigorous, independent, long-term evaluation of worksite health promotion initiatives occurs if we are to reach definitive conclusions about how effects on employee behaviour change translate into hard outcomes such as changes in body weight, health risks, healthcare utilisation, absenteeism, and productivity.
This review provides a comprehensive assessment of the impact of worksite interventions published over the past 15 years on dietary outcomes. It complements previous reviews that examined the impact of worksite interventions on physical activity [13
] and weight loss [9
] outcomes. Strengths include the systematic approach to searching the literature and inclusion of a broad range of study designs. Inclusion of study designs other than RCTs is important when evaluating complex interventions such as worksite programmes because application of an RCT design may be difficult and/or ethically inappropriate in practice. Limitations of the review include restriction of the search to studies published in English and use of a limited number of electronic databases. These search restrictions may account for the predominance of North American studies retrieved. However this may also be due to the fact that employer health insurance contributions are common in the United States, providing a greater incentive for US employers to implement and evaluate the effectiveness of worksite health promotion programmes. Publication bias may also mean some relevant worksite health promotion programmes were not included. This is a particular possibility with community health promotion initiatives where many non-academic schemes are not evaluated and/or published.
Public health strategies are placing increasing emphasis on the key role worksites can play in preventing illness and promoting health and well-being [6
]. However, this review highlights a critical lack of evidence regarding the most acceptable and cost-effective worksite health programmes. Strategies employed to promote healthy eating to date have largely focussed on individual responsibility (education and behaviour change). Some programmes have implemented changes to worksite environments in order to make healthy choices easier but these have largely focussed on changing the physical environment, i.e. food availability, and have mostly failed to tackle the economic, political, and socio-cultural aspects of the worksite. Greater use of frameworks for interventions that acknowledge the complexity of the environment and the need to intervene at many levels may help to achieve more meaningful changes [38
]. In particular, workplace canteens which frequently include a degree of food subsidisation provide an ideal environment in which to test the potential of economic incentives to change food purchasing behaviour [39
]. Evidence suggests that economic incentives impact positively on dietary behaviour [40
]; and favorable effects have been seen for weight loss [41
], purchase of low-fat snacks [43
], and self-reported fruit and vegetable consumption [44
]. Changes to political (the rules) and socio-cultural (social norms) aspects of the worksite also merit more consideration in future interventions.
Before worksite programmes can be implemented with confidence and rolled out on a large scale, more social and behavioural research is needed to help identify determinants of eating habits and predictors of uptake of worksite health promotion programmes. Some worksite programmes have been based on solid groundwork exploring factors influencing potential programme adoption and implementation [31
], but there remains a clear need to integrate qualitative and quantitative research methods in order to better evaluate reasons for success or failure of such complex interventions [45
There is also a need to radically improve the quality and reporting of worksite intervention studies. Many published studies suffer from design flaws including the absence of a comparison group, reporting of multiple outcomes in the absence of a pre-specified study hypothesis and primary outcome, lack of objective outcome measures, and inappropriate statistical analyses.