Lifestyle-focused interventions at the workplace to reduce the risk of cardiovascular disease--a systematic review.

OBJECTIVE
The goal of this review was to summarize the evidence for an effect of lifestyle-targeted interventions at the workplace on the main biological risk factors for cardiovascular disease (CVD).


METHODS
We performed an extensive systematic literature search for randomized controlled trials (RCT) that met the following inclusion criteria: (i) targeted at workers; (ii) aimed at increasing physical activity and/or improving diet; and (iii) measured body weight, body fat, blood pressure, blood lipids and/or blood glucose. We used a nine-item methodological quality list to determine the quality of each study. A best-evidence system was applied, taking into account study quality and consistency of effects.


RESULTS
Our review included 31 RCT, describing a diversity of interventions (eg counseling, group education, or exercise). Of these studies, 18 were of high quality. Strong evidence was found for a positive effect on body fat, one of the strongest predictors of CVD risk. Among populations "at risk", there was strong evidence for a positive effect on body weight. Due to inconsistencies in results between studies, there was no evidence for the effectiveness of interventions on the remaining outcomes.


CONCLUSIONS
We found strong evidence for the effectiveness of workplace lifestyle-based interventions on body fat and, in populations at risk for CVD, body weight. Populations with an elevated risk of CVD seemed to benefit most from lifestyle interventions; supervised exercise interventions appeared the least effective intervention strategy. To gain better insight into the mechanisms that led to the intervention effects, the participants' compliance with the intervention and the lifestyle changes achieved should be reported in future studies.

In Western countries, the prevalence of cardiovascular diseases (CVD) and related disabilities remains high (1). CVD can be divided into three major categories: cerebrovascular disease, coronary heart disease, and peripheral vascular disease. All three disease categories are associated with excess body weight and fat, an elevated blood pressure, disturbed blood glucose, and an abnormal serum lipid profile (ie, low high-density lipoprotein (HDL) cholesterol, high low-density lipoprotein (LDL) cholesterol, and high triglyceride levels) (2). These abnormalities are mainly caused by unhealthy lifestyle behaviors, including smoking. Smoking leads to hypertension and low levels of HDL cholesterol (3,4). Diet is also strongly associated with several CVD risk factors. A diet rich in saturated fat negatively influences serum lipid profile (5,6), and excessive salt and alcohol intake contributes to hypertension (7,8). A diet rich in calories, combined with insufficient physical activity, leads to weight gain and obesity (9) and, more importantly, excess body fat (10). Not only the content of meals but also eating patterns are associated with being overweight and CVD risk (11). For example, skipping breakfast increases the likelihood of eating more energy-dense snacks throughout the day. Physical inactivity is another lifestyle behavior associated with an elevated CVD risk (12)(13)(14), not least due to its contribution to weight gain. Since lifestyle is a strong but modifiable risk factor for CVD, it has been the subject of research for many years.
For workers, an unhealthy lifestyle and being overweight not only affect CVD risk, but may also have major disadvantages related to work. Insufficient physical activity is negatively related to physical work capacity (15) and positively related to sick leave (16). Furthermore, in two recently published systematic reviews, it has been shown that obesity is a significant predictor of long-term sick leave (17) and disability pensions (18). Schmier et al also concluded that obesity is related to more injuries (19). Altogether, physical inactivity and obesity are important drivers of indirect costs (19,20).
Thus, changing smoking, dietary, and physical activity behavior has many benefits. Several studies have investigated the effects of lifestyle-focused interventions on CVD risk. A Cochrane systematic review concluded evidence for the effectiveness of lifestyle interventions (21). In the previous 20 years, some narrative and systematic reviews on health promotion aimed at workers have been performed, focusing on physical activity (22)(23)(24)(25)(26), diet (27), smoking (28), or health promotion in general (29)(30)(31)(32). However, most reviews have not reported the evidence for effects on biological risk factors, which are objectively measurable, reliable, and strong predictors of CVD risk. Also, not all of the reviews applied a systematic approach to their search or the determination of evidence. Finally, the majority did not make a distinction between interventions aimed at populations at risk for CVD and those aimed at populations including both healthy persons and persons at risk ("mixed populations").
Our goal was to summarize the evidence for the effectiveness of interventions aimed at improving physical activity and dietary behavior on body weight and fat, blood pressure and glucose, and serum lipids among workers. Interventions aimed at smoking cessation were not included since smoking cessation is not associated with body weight loss. First, we describe the evidence based on all the studies together. Second, we describe the evidence derived from studies aimed at populations at risk for CVD and those targeted at mixed populations. Third, we describe evidence for the effectiveness of the three most frequently used intervention methods.

Literature search
We performed a literature search of several electronic databases (ie, Embase, PubMed, PsycINFO, SPORTDiscus, and the Cochrane Central Register of Controlled Trials). We then screened the reference list of a key systematic review on multiple risk factor interventions for primary prevention of coronary heart disease (21); personal databases of the first two authors of this paper were also searched for additional publications. We sought randomized controlled trials (RCT) and controlled trials, evaluating worksite lifestyle or healthpromotion interventions (such as individual counseling, group education, or self-help) aimed at the promotion of physical activity and/or a healthy diet. Due to their inferior design, controlled trials would only be included when the number of RCT would be too low to draw conclusions (33). As for the study population, we included interventions aimed at blue-and white-collar workers of all ages and both genders. Furthermore, interventions had to be implemented in an occupational setting (ie, at the workplace and/or during working hours and/or facilitated by the employer). Outcome measures were defined as biological risk factors for CVD. The search was limited to studies published in English, between 1 January 1987 and 31 December 2008.

Outcome measures
Several biological CVD risk factors were defined as outcome measures: body weight, body mass index (BMI), total, HDL and LDL cholesterol, triglycerides, systolic and diastolic blood pressure, blood glucose, and body fat. Because some measures were highly comparable or only measured sporadically, they were clustered. Body weight and BMI were clustered into "body weight/ BMI". The body fat-related measures were categorized into four categories: (i) overall body fat, as measured by dual energy X-ray absorptiometry (DXA) or bioelectrical impedance; (ii) "central" body fat, as measured by waist circumference, waist-hip ratio, or DXA; (iii) "peripheral" body fat, as measured by DXA or skin folds; and (iv) hip circumference. Peripheral and central body fat both have a positive relation with CVD, but the influence of the latter is largest.

Selection and data extraction
The first and second authors evaluated all titles and abstracts; both based their decision on the previously established inclusion criteria. In case an abstract contained insufficient information, or where both authors disagreed, the full paper was read. If disagreement remained, the third author made the final decision. In the situation where certain quality criteria on study design were not mentioned in the article, we checked if the authors referred to a previous publication that contained more detailed information on these topics. After having collected all relevant publications, the first author extracted the data. Due to the variety of outcome variables, measurement methods, and timing of measurements used in the studies, a meta-analysis was considered inappropriate.

Quality criteria list
The criteria list used for assessing the methodological quality of each study, was based on the Delphi list, developed by authors, epidemiologists, and statisticians (34). The list was adjusted to meet the specific purpose of this systematic review. Two items were added, referring to dropout and the length of follow up, as previously described by Proper et al (35). We pilot tested the adjusted version of the quality criteria list and independently scored two articles. Some items were described in more detail because of interpretation difficulties. Once the first three authors had agreed upon the modified list, as shown in table 1, the first two authors independently assessed the methodological quality of each study. Items were scored negative where they were neither mentioned nor properly explained. In case of disagreement, the third author also scored the article.

Best-evidence synthesis
Conclusions about the effectiveness of the interventions were based on a best-evidence synthesis. For each outcome, four levels of evidence for the effect of the intervention were discerned. The level of evidence depended on the quality of the studies showing this effect, and the consistency of the results. The levels of evidence, adapted from Van Poppel et al (36), were described as: level 1 (strong evidence = multiple high quality RCT with consistent outcomes); level 2 (moderate evidence = 1 high quality RCT and ≥1 low quality RCT, all with consistent outcomes); level 3 (limited evidence = only 1 high quality RCT or >1 low quality RCT, all with consistent outcomes); level 4 (no evidence = only 1 low quality RCT or contradictory outcomes of the studies).
A study was categorized as being of high quality in case >50% of the methodological quality items scored positively; otherwise a study was categorized as being of low quality. Consistency of results for a certain outcome measure was reached when at least 75% of relevant studies had results in the same direction (ie, significantly positive in the intervention group, no difference between groups, or significantly positive in the control group). Where there were ≥2 high quality RCT, the conclusion was based on these RCT only. If not, the results of the low quality RCT were also taken into account.
In addition to applying the best-evidence synthesis to all the studies together, we applied it to studies aimed at populations at risk for CVD only [ie, studies in which "having ≥1 CVD risk factors at or above a certain cutoff value" was one of the inclusion criteria (as defined by the authors of the studies themselves)]. We also applied the best-evidence synthesis to studies aimed at mixed populations only (ie, studies that had no inclusion criteria related to CVD risk status). Moreover, the evidence for the effectiveness of lifestyle interventions was determined for the three main intervention types separately (ie, individual counseling, group education, and supervised exercise). Figure 1 shows the flow diagram of the studies identified and subsequently included or rejected. The electronic database search resulted in 1193 studies. The personal database search identified four additional studies, and three were found in the reference lists. Of these 1200 studies, 1130 were excluded as a first step mostly due to a lack of a control group or because the study did not describe the outcome measures sought. After having read the whole text, another 31 studies were excluded, because they did not fulfill the eligibility criteria. This left us with 39 studies, of which 79.5% (N=31) were RCT.

Intention-to-treat analysis
Positive if an intention-to-treat analysis was performed.

Control for confounders
Positive if the analysis controlled for potential confounders.
This number was considered sufficient to draw conclusions. Finally, 32 publications describing 31 RCT were included . Concerning the methodological quality, the first two researchers disagreed on 11.8% of the items. Despite a consensus meeting, disagreement remained for one item. After consulting the third reviewer, the scoring process was completed.

Description of studies
Of 31 studies, 18 were high quality (37, 39, 41-44, 46, 47, 49, 55, 56, 58-64, 66). Study populations varied between 37-2791 workers. Most studies (N=21) were designed as a two-arm RCT that evaluated one or more intervention strategies in the intervention group. In ten studies, more than one intervention group was involved. The three intervention strategies most frequently used were individual counseling (N=18), group education (N=15), and supervised exercise (N=11). Between studies, these strategies showed large differences in frequency, intensity, and duration. Other methods, such as general written advice, a prescribed diet, self-help materials, environmental changes, or monetary incentives, were investigated only sporadically. Of the total, 12 studies aimed at populations at risk for CVD, and 19 targeted mixed populations. In table 2, the characteristics of all studies and the intervention methods are presented in detail. In the last column, all significant effects are indicated.

Body weight/body mass index
Body weight/BMI were reported in 20 studies, 14 of which were high quality. Seven high quality RCT showed a significant difference between groups in favor of the intervention group, whereas six high quality RCT showed no effect, and one showed a significantly positive effect in favor of the control group. Thus, there was no evidence for an effect on body weight/BMI. When considering studies aimed only at populations at risk (N=12), there was strong evidence for an intervention effect on body weight/BMI; one high quality RCT showed no effect on body weight and six high quality RCT showed a significantly positive intervention effect. Among mixed populations (N=19), there was no evidence for an effect on body weight/BMI. There was no evidence for an effect on body weight for any of the three main intervention strategies.

Body fat
Even though three low quality RCT showed no intervention effect, all three high quality RCT in which overall body fat was measured showed a significantly positive effect in favor of the intervention group. Thus, strong evidence was concluded for a beneficial effect on overall body fat. As for central body fat, the results were mixed, and there was no evidence for an effect on this outcome measure. Peripheral body fat was measured in three high quality RCT, two of which showed a significant favorable intervention effect, resulting in no evidence. Hip circumference was measured in two high quality studies, both showing no effect. Therefore, strong evidence for no intervention effect on hip circumference was concluded. Among populations at risk for CVD, the evidence for an effect on overall body fat and peripheral body fat was limited, since it was measured in only one study. When considering studies aimed at mixed populations, there was strong evidence for no effect on central body fat and hip circumference, based on four and two high quality RCT respectively. Among studies using counseling as part of the intervention (N=18), there was strong evidence for a positive effect on peripheral body fat and limited evidence for an effect on overall body fat. The latter was also true for studies using group education (N=15). In studies evaluating exercise interventions (N=11), there was strong evidence for (i) a positive intervention effect on overall body fat and (ii) no effect on hip circumference, the latter conclusion based on two high quality RCT.

Blood pressure
Systolic blood pressure was measured in 18 studies, 12 of which were of high quality. Of those 12 studies, 25% (N=3) showed a positive effect and 75% (N=9) showed             The quality score is reported, defined as the number of quality items scored positively as opposed to the total amount of quality items (eg 5/9). b Only outcomes measures with statistically significant (P<0.05) intervention effects, as determined by between-group differences at follow up or linear regression analyses.
no effect. Thus, strong evidence for no intervention effect on systolic blood pressure was concluded. Diastolic blood pressure was reported in 17 studies, 11 of which were high quality RCT. Of these, two showed a significantly positive effect in favor of the intervention group, one showed a significantly positive effect in favor of the control group, and 8 showed no effect. These data led to the conclusion of no evidence for an intervention effect on diastolic blood pressure. There was no evi-dence for an effect on either systolic or diastolic blood pressure in populations at risk. In studies aimed at mixed populations, there was strong evidence for no effect on systolic or diastolic blood pressure, based on six and five high quality RCT respectively. From counseling interventions studies, no evidence for an effect was concluded. In studies using group education, there was strong evidence for no effect on systolic blood pressure, based on five high quality RCT. The same was true for studies using exercise. Moreover, in exercise-based studies, there was no evidence for an effect on diastolic blood pressure.

Serum lipids
Of the 21 studies reporting ≥1 total, HDL, or LDL cholesterol measures, 15 were high quality, and no evidence was 15 were high quality, and no evidence was no evidence was concluded for an intervention effect. Triglycerides were significantly positively influenced by the intervention in two high quality studies, but no effect was found in five in five high quality studies. Thus, no evidence for an intervention . Thus, no evidence for an intervention effect on triglycerides was concluded. In populations at risk for CVD, there was no evidence for an effect on total or LDL cholesterol and triglycerides. There was strong evidence for no effect on HDL cholesterol, as concluded from five high quality RCT. In mixed populations, there was no evidence for an effect on any of the serum lipids, except for triglycerides, for which we found strong evidence for no effect. Among studies using counseling as (part of the) intervention, there was strong evidence for no effect on LDL cholesterol and no evidence for effects on the other serum lipids. We concluded strong evidence for no effect on triglycerides from studies using group education. Exercise-based studies provided strong evidence for no effect on total, LDL, and HDL cholesterol.

Blood glucose
Of three studies that measured blood glucose, one high quality RCT showed no effect and another showed a significantly positive effect. That said, there was no evidence for an effect on blood glucose. There was no evidence in populations at risk or mixed populations, limited evidence for an effect among studies that used counseling, and no evidence among group education or exercise-based studies.

Main findings
Based on the 31 studies examined, we found there was no evidence for a positive effect of workplace lifestylefocused interventions on body weight, blood pressure, serum lipid profile, blood glucose, and triglycerides. However, there was strong evidence for a favorable intervention effect on overall body fat, which is a better predictor for CVD than body weight; when fat mass is lost and muscle mass is gained, body weight remains unchanged.
The effectiveness of a lifestyle intervention often depends on whether the participants enrolled in the study have an elevated disease risk or not. Studies aimed at high-risk populations may yield different results, have a larger health impact, and be more cost-effective (69) than those targeting non-risk populations. To provide insight into this issue, we separately evaluated the studies aimed at populations (i) with an elevated CVD risk and (ii) for whom no CVD risk-related inclusion criteria were defined (ie, mixed populations). We found that among the latter, there was strong evidence for no effect on most outcome measures. For high risk populations, however, even though there was strong evidence for no effect on HDL cholesterol, there was strong evidence for an effect on body weight. For the other outcome measures, there was limited or no evidence -due to heterogeneous results or small sample sizes. We agree with Fleming et al (70) that lifestyle interventions aiming at low risk populations may be of marginal benefit and resources are better spent on those with an elevated risk of CVD. With respect to intervention strategies, we found that counseling, group education, and exercise were most frequently used. Studies focused on individual counseling and group education were more likely to find positive intervention effects than those examining supervised exercise. In fact, among studies looking at supervised exercise, for half of the outcome measures, there was strong evidence for no effect. These inconsistencies are probably related to differences in study populations.
The lack of evidence for most outcome variables resulted from inconsistencies between the studies' results. These inconsistencies are probably related to study populations, intervention strategies, and measurement methods. Other factors that may have contributed to the inconsistencies in results could be differences in the participants' compliance with the intervention, and the lifestyle changes that they actually achieved. Unfortunately, most articles lacked information in this respect (eg, the frequency and duration of sessions attended and the number of self-help assignments completed). The exact contents of the counseling or group education sessions were also usually not mentioned. Finally, from most studies it was unknown to what extent the interventions led to the intended dietary or physical activity change. Therefore, it was difficult to conclude what exactly happened to the participants during the study, and what was the mechanism that led to the effects.
When considering the results, not only significance, but also clinical relevance should be considered. The clinical relevance of a change in a certain CVD risk factor depends on its initial value and the presence of other risk factors, as illustrated by various (coronary heart disease) risk assessment instruments such as the Framingham risk score (71). Considering body weight, every kilogram of body weight loss was proven to correspond to a 16% reduction in diabetes risk (72). Thus, small changes in body weight and body fat may already be clinically relevant. Consequently, the findings of strong evidence for intervention effectiveness on body fat, and body weight in populations at risk, are certainly interesting.

Comparable studies
Ebrahim et al (21) published an extensive Cochrane systematic review on lifestyle interventions for lowering coronary heart disease risk in different settings, which was updated in 2006. They found insignificant changes in blood pressure but significant falls in blood cholesterol. It is important to note that Ebrahim and colleagues suggested that the changes in cholesterol levels may have been attributable to the use of cholesterol-lowering medication. Prescription of medication was not part of the intervention protocol in any of the studies included in our systematic review. Nevertheless, participants may have used medication before the study had started. Proper et al (23) published a systematic review similar to ours. They summarized the evidence for an effect on physical activity, fitness, and health among workers. Despite that they included only physical activity interventions, their conclusions were comparable to ours (ie, inconclusive evidence for an effect on body composition, and no evidence for an effect on blood pressure and serum lipids). More recently, Conn et al (26) showed that physical activity interventions based on supervised exercise and motivational and educational strategies led to significant improvements in lipids and anthropometrics. In contrast to our study, Conn and colleagues included RCT as well as non-controlled trials and unpublished reports -study designs that we considered less valid. In a systematic review on body weight loss among workers, Anderson et al (29) found a net body weight loss of 1.3 kg based on nine RCT. Since we did not pool our data, our findings cannot be compared to theirs. With respect to the quality of the studies, we differed strongly with the findings of Kjaergard et al (73) who reported that year of publication was not positively related to study quality. In our review, of all included studies published in or after 2000, 78.6% (N=11) were of high quality, whereas of the studies published between 1987-1999, only 41.2% (N=7) were of high quality. This may have been a result of the stricter quality criteria of scientific journals in recent years.

Limitations and strengths
One of the limitations of our study concerned the bestevidence synthesis. When determining the evidence for effectiveness on a certain outcome measure, adding one high quality study may change the conclusion from "no evidence" to "strong evidence". Besides, the cut-off point of 75% for consistency between results often leads to the conclusion that there is no evidence for an effect. However, there is no consensus about which levels of evidence criteria should best be used (74). Another drawback was related to the fact that our quality assessment was based on the data as reported in the articles. In reality, in some articles, relevant data (on, for example, randomization procedure, blinding, and type of analysis) were not presented. This may have led to an underestimation of the study's quality. One way to solve this problem would have been to ask all authors individually to provide missing information. However, in a study of Gibson et al (75), two thirds of the authors simply did not respond to their request for additional information. Lastly, measurements of waist circumference and skin folds are less accurate than DXA and bioelectrical impedance. Still, we decided to cluster all the studies in which body fat was measured. In our opinion, separating them according to the type of measurement would result in an inadequate numbers of studies to draw conclusions. Furthermore, when changes in body fat are determined over time, inaccuracy of measurements is less of a problem than when determining body fat cross-sectionally.
Several strengths of this systematic review can also be mentioned. All relevant publications on workplace lifestyle-focused interventions were systematically collected and evaluated. We described not only the intervention effects, but also the methods and population type used. The quality list was well adapted to the type of intervention studies. By independently scoring all articles, we maintained objectivity. Most importantly, in order to determine the population for whom lifestyle interventions seem most effective, we looked separately at studies aimed at populations at risk and mixed populations. Moreover, in order to define the most promising intervention strategy, we explored the evidence in three frequently used counseling strategies separately.

Concluding remarks
This systematic review fills a gap of knowledge on the effectiveness of lifestyle interventions on the main CVD risk factors among workers. Considering the cardiovascular health-and work-related risks of excessive weight and obesity, the findings of strong evidence for effectiveness on body fat and, among populations at risk, body weight are interesting for employers. For intervention planners and policy-makers it is worth knowing that populations at risk seemed to benefit more from lifestyle interventions than mixed populations, while supervised exercise interventions appeared the least effective intervention strategy. The lack of evidence for effects on most of the remaining CVD risk factors was mainly due to inconsistencies in results. In order to gain better insight into the mechanisms that led to the intervention effects, the participants' compliance with the intervention and their lifestyle change achieved should be reported in future studies.