Health promotion trials at worksites and risk factors for cancer

Health trials at and Studies of worksite health promotion have frequently reported larger effects than those at the community level. Many of these studies have serious methodological problems. Forty-five worksite health promotion trials following specific quality criteria were selected and estimated for behavioral changes in cancer risk factors and the effectiveness of different intervention components. Tobacco control programs found quit rates of about 5% with relapse rates of 40% to 80% at 6 months after the intervention. Effectiveness increased with the duration of the intervention for at least 6 months, repeated contacts with the participants, continuous support, and tailored messages. There was less evidence for the long-term effectiveness of incentives. Trials on diet, alcohol, physical activity, overweight, and solar radiation showed the same positive trends. The overall evidence indicates a modest but positive effect of health promotion trials at worksites and the effect, for smoking cessation trials, is slightly larger than that of community-based trials. Many of the recommendations made to increase participation and effectiveness were not based on empirical data.

Health promotion programs encourage people to adopt healthy life-style behavior on the assumption that it will result in improved health. Big community-based health promotion trials have, however, reported only modest effects [eg, COMMIT and other trials (1,2)]. In recent years, numerous health promotion studies have been conducted at the worksite level. Risk behavior has been addressed in health promotion trials concerning tobacco cessation, diet changes such as increases in fruit and vegetable consumption, control of cholesterol levels, physical activity and loss of weight, exposure to ultraviolet light, stress, alcohol and drug abuse, and other such issues.
These studies, and reviews of these studies, have frequently reported more significant changes than those of community-based trials, such as tobacco quit rates of the order of 10-15% (3,4). It has frequently been assumed that worksite health promotion must have a positive ef-fect, and many of these studies were not adequately designed to evaluate whether such an effect was present. Many studies did not have a reference group, they were based on a small number of subjects, and they did not give quantitative estimates of the changes achieved by the interventions. Properly controlled large intervention trials at worksites tend to report a positive, but only modest effect (4)(5)(6)(7).
In this review we focus on worksite health promotion trials on major cancer risk factors, including tobacco, diet, physical activity, obesity, ultraviolet light, and alcohol consumption. In contrast to other existing reviews, we selected studies on the basis of specific quality criteria and have reported quantitative estimates of the changes achieved and have drawn conclusions on the basis of the quantitative estimates. The application of meta-analytic techniques was possible only to a limited extent due to the heterogeneous design of the Scand J Work Environ Health 2002; 28(3):141-157 Reviews trials. This review was prepared within the framework of the European project PREWORK (prevention of risk factors for cancer at the worksite).

Search methodology
Bibliographic searches were done through the online database Medline, searches in our own archives, backward searches of articles cited in earlier literature reviews, and backward searches of reports on health promotion trials at worksites. The keywords used in the Medline search were "employee OR worksite OR workplace OR occupational health services" AND "wellness program OR health promotion OR health education OR health behavior" in combination with keywords for specific risk factors such as: "tobacco control OR tobacco", "diet OR nutrition", etc. We initially identified more than 1000 papers and reviewed approximately 200 articles with some reference on health promotion programs at worksites.

Selection criteria
For this review, we selected 45 trials on the basis of the following criteria: (i) trials referring to tobacco control, diet control, physical activity, obesity, ultraviolet light, and alcohol consumption; (ii) health promotion trials at worksites with a control no-intervention group or with an evaluation of different types of interventions in the same study; (iii) studies examining outcomes directly related to health (ie, studies evaluating only absenteeism or work productivity were excluded); (iv) studies including more than 100 subjects and a minimum of 50 subjects in each study group; (v) articles in English, French, or Spanish; (vi) studies in peer-reviewed journals. A high participation rate was also included a priori in the inclusion criteria but was dropped due to the difficulties in calculating the comparable rates of the trials.

Target population
Most trials reported interventions in the United States (N=29, 64%) (table 1). The criteria used to select workplaces for the trials were clearly stated in 12 (12,13,17,22,26,28,29,31,35,36,38,51) of the 45 trials. The criteria used included number of workers (13,22,26,29,36), low turnover rates (31,36), high response rate at baseline (22), or acceptability of the program before the randomization process (11). The individual inclusion criteria also varied among the studies in that most of the studies included all or a sample of workers in a worksite, two trials included only at-risk populations (32,42), and six trials used only volunteers (15-18, 32, 44). The companies selected were mostly services and manufacturing plants. The proportion of bluecollar workers varied among the studies but, overall, half of the workers covered by health promotion programs were blue-collar. Although programs were addressed to workers independently of gender, out of the 45 trials     reviewed, 15 included more than 70% men (8, 9, 14, 17, 19, 28-31, 35, 38-40, 50, 51) and 3 had more than 70% women (37,42,49).

Analysis
A characteristic feature of cluster randomization studies is that the natural variability in response among clusters exceeds the variability in response within clusters (62). Eight studies (11,18,23,26,27,31,39,48) matched subjects or clusters to factors such as worksite size, gender distribution, blue-and white-collar distribution, and other factors. Four studies stratified worksites in the design phase before the randomization (11,13,36,37). Most of the trials used worksite, or whatever unit of cluster randomization, as the unit of analysis (8,11,13,22,26,27,39). Those using workers as the unit of analysis took into account the effect of the cluster randomization procedure by incorporating a nested design structure (29,36,37,51) or by using other procedures (18,30,31,35,47). Seven trials did not take cluster randomization into account when analyzing their data (12,16,24,28,38,47,49).

Intervention characteristics
The interventions included in the health promotion programs reviewed are described in table 2. Methods for recruitment and the intensity of the recruitment process differed among the studies and included, for example, annual medical examination (8,14), letters or phone calls (12,13,15,17,22,38,39), and recruitment through group leaders or union representatives (11,29,50).
The process (contents) of the interventions can be classified into those aiming at improving workers' knowledge and awareness of certain risk factors, helping workers to make behavioral changes, and creating environments facilitating healthy life-styles. The first strategy offers information, such as an explanation of the quit process or of the relation between physical activity and health. The second strategy offers persuasion and techniques to develop skills, such as learning to avoid tempting situations or identifying personal barriers to the adherence to healthy behaviors. Environmental or structural interventions start changing the environment without requiring individual participation in educational activities, such as banning smoking in certain areas or changing available food in the canteen. Most programs included both awareness and behavioral change factors. Occasionally, environmental and structural interventions were added.
Most health promotion programs offered an initial medical screening, questionnaires, or interviews to obtain worker health profiles. Thereafter, the treatment group received some kind of intervention. Occasionally intervention messages were adapted to the results obtained in this baseline assessment (tailored interventions, see table 1), following in some studies the stages-ofchange theory (20,26,27,29,35,46). For example, in a smoking cessation trial, workers would be classified as at the precontemplation stage if they did not intend to quit smoking in the next 6 months, as at the contemplation stage if they intended to quit smoking within the next 6 months, and at the preparatory stage if they wanted to quit smoking in the next month or if they had seriously attempted to quit smoking during the last year. Workers would then receive different types of intervention depending on the stage. For example, workers in an "early" stage would be asked to consider the health risks of being overweight, having high cholesterol, or smoking, while those in a "later" stage would be asked to change their diet or tobacco use or to maintain a healthy life-style.

Outcomes
The outcomes measured in the trials were participation rates in health promotion activities, knowledge improvement on risk factors, alterations of states of change, behavioral changes, changes in biological parameters, and morbidity and mortality incidence reduction.

Participation
Almost all the trials provided some information on participation rates ( figure 1), but only a few reported comprehensive data for different stages of the study (10,12,22,28,50). Worksite or physician participation rates were reported in 11 trials, and they ranged from 17% (8) to 100% (19). Reasons for no participation were lack of interest in health promotion programs, strikes, or participation in other health promotion programs. Response rates for individuals were reported in most trials for baseline and follow-up assessments. The response rates of selected workers in the baseline assessment ranged from 20% (15) to 90% (43) with an average of 71%, while those at follow-up ranged from 43% (27) to 98% (11) of the subjects who completed the baseline assessments, the average being 69%. Finally, participation in the health promotion activities was reported in seven trials, and it ranged from 12% (12) to 83% (32). The median participation of the target population at follow-up (including also those not participating in the baseline survey) was 49% ( figure 1).
There are limited data on whether (and which) baseline characteristics of the target population are associated with participation rates. Factors that have led to higher response rates in the baseline or follow-up surveys in some studies have been gender (male), higher educational attainment, white-collar worker, older age, marriage, white non-Hispanic worker, and other factors related to specific risk factors, such as lower smoking prevalence (24). Factors influencing participation in program activities have been smaller worksites (12), gender (female), professional and clerical or sales job in a tobacco control program (22), and younger, better educated person, and active employment in a nutrition program (35).
The strategies used to recruit a high number of participants and to maintain low attrition rates differed between the studies. The involvement of workers in the planning and implementation of the intervention was the most common policy used to increase participation. Participation rates in the intervention activities with workers' participation (16, 17, 23, 26-28, 31, 33, 39) were, however, the same if not lower than those without it. A direct comparison of participation rates was, however, complicated since the studies differed in many aspects apart from worker participation. Only one trial offered the same intervention first without and then with the involvement of workers in the planning and implementation and increased the participation rate from 15% to 65% (24).
Some health promotion programs offered economic incentives or within or between worksite competitions intending to improve both participation rate and behavioral change. It was difficult to evaluate the influence of incentives on participation by comparing studies with or without incentives because there were many other differences between the studies and only a few of them reported data on participation. One trial offered the same intervention, however, with and without incentives and did not find any improvement in the participation rates (29).
The statistical treatment of nonresponse varied among the studies. In the analysis, one trial included the companies that did not agree to participate after having been assigned to the treatment group (36), while all the other trials excluded these companies from the analysis. Workers who were assigned to the intervention group but did not participate in the health promotion activities were usually included in the analysis, while those who did not complete the baseline assessments could not be included in the analysis. Occasionally, in smoking control trials, those lost to follow-up were included in the analysis and were treated as if they were still smokers (8,13,15,17,19,42).

Smoking cessation programs
A higher quit rate was observed in the group receiving the intervention in all 16 of the smoking cessation programs that reported this outcome (8-15, 17-19, 25-29). The magnitude of the effect of the intervention varied considerably between the studies (figure 2) and was statistically significant in 9 (9-12, 14, 15, 18, 19, 28) of the 13 studies reporting statistical significance (8,26,27,29). The average smoking cessation rate was 6.1%, ranging from 0.9% to 12%, when quit rates in the intervention groups were compared with those of the control groups (figure 2). For the 13 studies providing complete information allowing the calculation of means and standard errors (8-10, 12-15, 18, 19, 26-29), this average was 5.3%. The weighted mean of the quit rates (weighted by the inverse of the standard error) was very similar (5.4%). For the calculation of this weighted sample, the size of some studies had to be estimated because exact data were not provided.
The quit rates of the intervention groups were of the order of 10-15%, if the changes observed in the control group were not taken into account. In some of these studies (N=6), the control groups were not re-contacted after the baseline surveys examining the prevalence of risk factors in the study population (10,12,19,(26)(27)(28), while in others (N=8) the control groups received a variable amount of information on tobacco control (8, 9, 13-15, 18, 25, 29). On most occasions this information was limited to minimal informative material distributed at the beginning of the intervention. The results varied considerably between these studies. Overall, however, the quit rates were similar in the studies using control groups without intervention (6%) as compared with those using control groups with minimal intervention (6%).
Changes in smoking prevalence at the end of the intervention were sometimes reported (8,22,27,(31)(32)(33)(34). These changes accounted for quit and relapse rates and ranged from 0.5% to 6%. Quit rates could not be calculated, therefore, for these studies. Similarly, changes in smoking prevalence could not be calculated for the remaining studies since they did not provide relapse rates for the whole study population. Continuous abstinence rates were reported in six trials, and they ranged from 1% to 6% (8,14,16,17,21,23). An increase in the number of attempts to quit smoking (17% and 13%) was found in both studies that examined this outcome (12,27). Progression through stages of change was evaluated in one study, which found a progressive tendency towards higher stages of the quitting process (27). A reduction in the number of cigarettes smoked per day (2.5 and 3.6 cigarettes/day) was found in two (11,31) out of four studies evaluating this outcome (10,14). Another trial reported a 20% reduction in heavy smokers (more than 30 cigarettes/day) (34).
The effect of the interventions on smoking behavior over time was evaluated in 11 studies (9-12, 13, 15, 17-19, 21, 29). In nearly all of them, the effects observed after the intervention attenuated over time ( figure 3). At about 6 months after the end of the intervention, the effect of the intervention was reduced by approximately 40% to 80%. It seems that the reduction obtained at about 6 months is maintained over time, although only few studies followed the study population for long periods of follow-up. The reduction of the effect after   Effect of the length of the intervention and the number of contacts on quit rates. Only smoking cessation trials reporting quit rates for intervention groups were included. For the trials reporting more than one follow-up, we included the first data reported above 3 months of follow-up. For those offering different types of intervention, we included the mean of the intervention effects, unless the length of the intervention or the number of contacts differed between the groups. the end of the intervention may, in part, be due to a late effect of self-help material in the control groups. Razavi et al (21) evaluated the effectiveness of two relapse prevention programs and found that relapse was higher in the control group (50.4%) versus the two intervention groups (42.2% and 46.6%) although the differences were not significant.
Efficiency increased with the length of the intervention up to 6 months ( figure 4). A few studies offered intervention for more than 6 months (10,17,(26)(27)(28), and they did not indicate any additional beneficial effects when the intervention was prolonged. The intensity of the intervention, measured by the number of contacts, was weakly associated with higher smoking quit rates ( figure 4). Although the highest quit rates were obtained in intense intervention programs (22,29), high intensity did not always lead to a more successful intervention (13,28,33). The quality parameters of these studies, such as participation rates, size of study, or matching, did not account for the differences in the quit rates between the high-intensity intervention programs.
Trials including incentives or competitions reported higher quit rates (8%) than those not offering incentives (5%). Most of the former type of trials did not offer any kind of intervention to the control group, while most studies without incentives included a minimal intervention in the control group. Only five studies compared the effectiveness of providing incentives within the same trial (13,15,18,20,29). Four found a higher smoking quit rate in the incentive group (10.4% versus 7.5%, 20% versus 16.5%, and 12.9% versus 12%) (13,18,29) or a higher abstinence smoking rate (40% versus 23%) (20). In one of these studies the effect of the incentives was substantially reduced in the long-term (20), while in another (29) the long-term quit rates showed a negative effect of incentives. In one trial (13) effects were statistically significant only in the subgroup of less-educated workers (quit rates of 18.6% versus 8.8%).
One trial examined the effect of social support among workers on the effectiveness of the health promotion programs. Actions in favor of nonsmoking, such as being asked by a fellow worker not to smoke, was correlated with greater changes in quit rates (standardized regression coefficient 0.195, P< 0.005) (12).
Intervention trials on the reduction of cardiovascular disease risk, although frequently focusing on tobacco control, have not always measured outcomes related to smoking behavior. Significant reductions in mortality and morbidity have been reported (30,33), but the extent to which these changes were due to smoking behavior modification was difficult to evaluate since these programs also included intervention focusing on diet, physical activity, and alcohol.
The effectiveness of tobacco control programs has been associated with some baseline characteristics of the study populations. It has been found to be higher among managers, moderate smokers (<1 daily packet of cigarettes or <10 cigarettes/day depending on the trial), smokers of light tobacco, smokers having had previous attempts at quitting smoking, smokers that found quitting significantly less difficult, and smokers who had smoked for less time (8,11). Including incentives was effective for workers with less education (13).

Length of intervention
Number of contacts 40). One study did not find any differences between the follow-up at 3 and 6 months (41).

Physical activity
Two types of physical activity intervention could be distinguished, that based on counseling and educational sessions and that offering facilities, space, or time to workers. The outcomes measured in the worksite health promotion programs on physical activity varied, including body fat, aerobic capacity, and the proportion of workers engaged in regular exercise (table 3).

Nutrition
Worksite health promotion trials on diet have focused on behavioral changes, particularly an increase or reduction in the consumption of specific food groups. Fruits, vegetables, and foods with a high proportion of fiber or fat have been the most common targets. Physiological outcomes, including cholesterol levels, body fat, and blood pressure have also been reported. Only body fat is included in this review (see the section on weight control) because the other factors are not directly related to cancer.
Programs targeting environmental changes, such as modification of the food offered in the company canteen and vending machines were evaluated in five trials (23,26,34,36,37). They all included some other kind of intervention. The results were similar to those obtained in trials not including environmental changes. Some of the intervention programs (7 of 13) included workers in the planning and implementing of the health promotion, but a larger effect could not be associated with this procedure (23,26,27,31,34,39,42).
Five studies included more than one period of follow-up (31,35,(39)(40)(41). The percentage of change maintained at 6 or 12 months with respect to the intake increase or reduction ranged from 30% to 65% (35, 39, and informative materials (25, 29, 31-34, 44, 46, 48) and evaluating outcomes directly related to physical activity, four reported significant changes (25,34,46,48). Significant changes were reported in three (44,46,48) of four trials (44) offering facilities, time, and space. One trial comparing both types of intervention found better results with educational activities focusing on behavioral changes, with 65% of at-risk employees exercising weekly, compared with 50% when facilities were offered (44). One study examined the effectiveness of self-help intervention tailored to the individual's stage of motivational readiness for exercise adoption (48). Motivationally tailored intervention appeared to be more effective in nearly all the stages of readiness, with differences from 4% to 17% in the proportion of workers progressing to a higher stage, for example, from contemplation of starting exercise to actually starting it (48).
Some studies included at least two follow-up periods to evaluate the maintenance of the intervention effect over time (29,30,45). The longest follow-up period was 7 years, and no reduction in the intervention effect was found. Participation in the activities was 13% at 7 years, the same as at the end of the intervention (45). Studies with repeated follow-up periods initially showed a change in the direction expected, but later a reduction occurred in this change (29,30). Among all the remaining studies that reported data only from one follow-up, four evaluated long-term effects at 2, 3, or 10 years (25,31,32,44). The study evaluating a 10year period did not find any difference between the control and intervention groups (31).

Weight control
Weight control programs were offered in three studies (22,43,49). Other studies aiming, for example, to reduce cardiovascular risks, offered intervention in nutrition or physical activity and also reported body fat changes (28,29,32,38,45). The results varied among the studies, the changes in body weight ranging from a 0.25-kilogram increase to a 3.5-kilogram decrease (22,28,32,43,49). The changes achieved in programs including weight control evaluation after a short-term period (43,49) were greater than those of studies reporting longer term data, at 2 to 3 years (22,32). Body fat reduction ranged from 0.45% to 1.19% (29,45), and the body mass index reduction ranged from 0.3 (29) to -0.3 (38). Maintenance of the intervention effect over Table 3. Intervention effects of some of the health promotion programs at worksites using physical activity as a form of intervention. a Statistically significant change in score only in one study (43). *P< 0.05.
1 year was assessed in one study with almost no reduction in the effect being found (29).

Other risk factors
Three alcohol intervention programs were identified. Kishchuk (50) reported significant improvement in alcohol knowledge, socially responsible attitudes, and feelings of self-efficacy and a statistically significant reduction in weekly consumption in the treatment group compared with the control group (decrease of 1.5 drinks/ week). In the study by Richmond et al (51) a decrease of 4.8 and 0.4 drinks/week was found for the men and women, respectively, but the results were statistically significant only for the women. A reduction in regular excessive drinking and binge drinking was seen among the men, but it was not significant. The third study (34) included alcohol among other types of life-style behavior and found a statistically significant higher increase for moderate drinkers (among all drinkers) in the intervention group (from 34% to 50%) than in the control group (from 38% to 44%).
Reducing occupational exposure to solar radiation was the objective of one program (52). The intervention, based on awareness and change of behavior, led to a significant improvement in solar protection knowledge and behavior among the participating workers, such as a 16% increase in the number of workers using sunscreen.
In neither the alcohol nor the solar radiation programs was there a follow-up of the effects obtained at posttests.

Discussion
The overall evidence from the 45 controlled trials evaluating worksite health promotion for cancer risk factors indicates a modest but positive effect. For smoking cessation, this effect was slightly larger than that reported in several large community-based trials. The bulk of the evidence comes from smoking cessation trials, but the effect was found for interventions focusing on all other risk factors also. Previous reviews on health promotion trials at work have also supported a positive effect (3)(4)(5)(6)(63)(64)(65). A few of these reviews, for example, that of Glanz et al (5), applied quality criteria to classify the trials evaluated but did not seem to have based their conclusions on trials considered to be of higher quality with respect to their design. Furthermore none of the existing reviews provided a comprehensive quantitative analysis. When only the big and properly controlled trials included in this review are considered, the results of the health promotion trials at work appear to be less impressive than those frequently quoted.
A quantitative approach to intervention effects and intervention component effectiveness necessarily implies simplifications. There are many differences between the trials, for example, target population, design, or intervention strategy, which may affect the results obtained. Control for these factors or a comparison of different options presents difficulties due to the insufficient information provided by many trials on the intervention delivered.

Design of the trials
We reviewed only the trials that included a control group. The use of a control group is necessary to discriminate intervention effects from population trends and to take into account changes in the way a person behaves because of the mere awareness of being under observation (66). The interpretation of the results of the trials depends greatly on whether the control group was taken into account in the analysis or not. Without this comparison, estimates on smoking cessation would have been two to three times higher than those found when comparing with a control group. This phenomenon explains the wide discrepancy in the estimates of this review from the higher ones provided in reviews without a strict selection of trials (3). The use of control groups receiving some kind of intervention either intentionally or unintentionally, may, however, also bias the results negatively. Negative bias may occur through contamiation of the information among workers and transfer of the intervention effects from those participating to those not participating in the health promotion activities (12). Contamination of the intervention stresses the need for cluster assignment rather than individual assignment.
Participation rates for health promotion activities have been seldom reported for the original target population. Low participation rates in interventions (ie, presence in the health promotion activities) have probably helped to decrease the effectiveness of these programs (64). The response rates at both the baseline assessment and follow-up were around 50%. Self-selection bias is a generalized feature when participation rates and recruitment intensity are low. This bias would tend to inflate results unless close attention is given to appropriate evaluation methods and design procedures (67).

Intervention effects
Results from the tobacco control programs indicated a quit rate of about 6% that could be attributed to the workplace intervention. A similar, although somewhat lower quit rate of 1.8% was found in analyses of controlled community smoking cessation trials (68).
Trials focusing on diet also identified positive but modest intervention effects. A meta-analysis of community-based health promotion programs on diet (69) found similar fat intake changes when the number of contacts received by the participants was lower than 10. In community trials with more than 10 contacts, the reduction in fat intake was five times that observed in health promotion trials at work. Among trials at worksites, only 3 of 13 contacted the workers repeatedly, and only 1 referred to fat intake.
All the existing trials promoting physical activity reported positive effects, even though statistically significant results were obtained in only about half of the trials. Statistically significant results were mostly obtained in programs including fitness facilities or classes as compared with programs without these advantages.
The two available programs focusing on alcohol found some intervention effects, as reported in a previous review (63). It is not evident, however, whether the reduction in alcohol intake occurred among the at-risk (high consumption) population or among those who already had low and safe consumption.
A trial on solar radiation focused on primary and secondary preventive measures and reported encouraging effects on the increase in protective measures such as the use of sunscreens by workers. A similar program offered at the community level found lower cumulative mortality rates for melanoma in the intervention region than in neighboring regions (70). Existing trials were directed towards occupational exposures. We could not identify any prevention trials dealing with leisure-time sun exposure.

Persistence of the intervention effects
Behavioral changes are difficult to maintain over time, and relapse rates of 40% to 80% were observed in the tobacco control trials. Similar relapse rates (50-90%) are commonly found in community trials for smoking and other disorders or risk factors such as alcoholism, obesity, and drug addiction (67). Relapse seemed to occur mostly in the first 6 months after the intervention, a result that has frequently been reported in other trials (71). These results probably indicate that a 6-month followup period would be sufficient in health promotion trials to identify almost all the relapse occurrences. There is little evidence, however, suggesting that there is any specific point beyond which relapse should not be considered (67), since the reviewed health promotion programs at work rarely reported long-term effectiveness after the end of the intervention. Neither in alcohol programs nor in solar radiation programs has there been any follow-up of the effects obtained at posttests.
The decrease in the effects observed over time is not always due to relapses in the intervention group. Occasionally it may be due to a late effect in the control group, particularly among those receiving self-help materials. This minimal intervention effect may take place with a longer latency time than that of the intervention group (9).

Factors that affect effectiveness
Various policies have been proposed to improve the effectiveness of trials by increasing participation rates and facilitating and encouraging participants to adopt healthier life-styles and maintain them over time. Many of these policies use a theoretical basis including concepts of cognitive theory, social support principles, and models of behavioral change (72,73,74). Factors thought to improve effectiveness include incentives, acting as a direct positive reinforcement by rewarding people who change to a healthier option; environmental changes, creating opportunities for action and removing barriers at the worksite level; tailored intervention, offering different messages according to the individual characteristics obtained in baseline surveys, generally following the stage-of-change model; involvement of workers in planning and implementing the intervention, suiting better workers' health priorities and identifying themselves with the program; intensity of the intervention, regarding the length or number of contacts with workers; and, finally, involvement of family or co-workers, acting as support to adopt and maintain the healthy options. However, the influence of many of these features has not been clearly associated with larger intervention effects.
In tobacco control programs a weak positive relation was found with a longer duration of the intervention and with the number of contacts. Health promotion trials focusing on factors other than smoking have too few and too heterogeneous data on intensity and duration of the worksite for these associations to be evaluated. In community-based trials on diet, a larger effect was observed in the trials with a higher number of contacts (69).
Comparisons between worksite trials on tobacco control regarding other components, such as incentives, company participation in intervention planning or implementation, environmental changes, focus on maintenance, or tailored message, were not feasible since they may be influenced by other design or intervention differences apart from the specific features examined. Within-trial comparisons to evaluate specific features have been done less often, but they found that the use of financial incentives, continuous support, and tailored messages were associated with improvement in the intervention effect. The impact, however, of incentives on quit rates was not only not maintained in the long-term but, in one study, seemed to be counterproductive.
Worksite trials on diet involving environmental intervention, such as changes in the worksite canteen or vending machines, did not prove to be more efficient than those without environmental components. A similar conclusion was reached in a previous review of worksite multicomponent health promotion programs (4).
Few trials evaluated the effect of providing tailored, as opposed to generic, strategies. The results indicated consistently larger intervention effects when a tailored approach was used. A similar conclusion has been reached in another review on worksite health promotion (4) and in community-based trials regarding physical activity and tobacco control (68,75).
It has been stated that, to optimize resources and results, it is necessary to integrate different risk factors into health promotion programs (36,52). Intuitively this assumtion seems apparent since life-style risk factors are often related and focusing on one specific factor may help modify another. There is only little direct evidence on this possibility (76). Existing studies do not evaluate the relative effectiveness of programs focusing on single or, alternatively, multiple risk factors.

Recommendations for future health promotion trials at worksites
Trials in North America have consistently shown a modest but positive effect of health promotion trials at worksites. Although much of the research has concentrated on tobacco control, findings for other cancer risk factors have indicated similar patterns. Low participation seems to be one of the main weak points of worksite health promotion. The involvement of workers in the preparation and conduct of the trials appears to be the only clue provided by existing studies on how participation could be increased. Ways of increasing participation and effectiveness are among the main issues to be resolved in future research. For participation to be evaluated, it is essential that studies report participation rates for different study phases. Increased duration, multiple contacts, tailored interventions, incentives, and continuous support are among the factors that seem to affect effectiveness, but data are not entirely consistent. These and other factors will have to be evaluated further. In addition, the information provided on the intervention delivered should be more complete.
There exists less experience on health promotion at work in Europe. Trials in North America may not be directly applicable to those in other industrialized countries with different sociocultural characteristics. Studies to evaluate the effectiveness of worksite health promotion should be conducted in those countries.
Existing studies indicate that a minimum of 6 months is necessary to evaluate the maintenance of interventions. There exist, however, too few studies examining effects over a longer term. Existing large studies could provide this evidence.
Finally, most of the worksite health promotion programs were conducted in large or medium-sized workplaces. The effectiveness of health promotion programs should be assessed in other kinds of worksites, including smaller worksites, those of workers in the primary sector, and those employing minorities. Although major intervention programs would not be feasible at small worksites, evidence from existing trials indicates that relatively modest interventions may have an effect.

Concluding remarks
Substantial evidence from health promotion programs at worksites, mostly from North America, currently shows that these trials have a modest but positive effect. For smoking cessation, this effect seems larger than that obtained in community intervention trials. Most trials have focused on tobacco control, but trials on diet (increase in fruit and vegetable consumption, decrease in fat consumption), alcohol, physical activity, overweight, and solar radiation have all shown the same positive trends. The reviewed health promotion programs at work included different intervention components and presented wide differences in issues such as recruitment intensity and worksite and workforce baseline characteristics. Low participation rates appear to be one of the main problems of worksite health promotion, and there is little empirical evidence regarding factors related to higher participation and lower attrition rates. Specific features that increase the effectiveness of the trials seem to be at least a 6-month duration of the intervention, repeated contacts with the participants, continuous support, and tailored messages. There is less evidence for the long-term effectiveness of incentives. Many of the other recommendations that have been made to increase effectiveness, although intuitively appealing, are not based on empirical data.