Systematic prevention of overweight and obesity in adults: a qualitative and quantitative literature analysis


S Kremers, Department of Health Promotion, Nutrition and Toxicology Research Institute Maastricht (NUTRIM), Maastricht University, PO Box 616, 6200 MD Maastricht, the Netherlands. E-mail:


To date, most interventions aimed at preventing obesity have underemphasized the application of systematic intervention development, implementation and evaluation. The present review provides a thorough insight in factors promoting implementation and/or effectiveness in interventions aimed at preventing overweight/obesity among adults. A total of 46 studies evaluating interventions aimed at preventing obesity were reviewed, followed by both qualitative and quantitative analyses. The Intervention Mapping protocol and the Environmental Research framework for weight Gain prevention (EnRG) were applied to analyse and classify the included studies. The interventions were categorized by setting (workplace, community, health care) and target group (ethnic minorities, pregnant women, [pre]menopausal women, smokers, people with intellectual disabilities). Generally, interventions were found to have potential in changing energy balance-related behaviours and anthropometric outcomes. Effect sizes for changes in body mass index ranged between −0.09 and 0.45. When the programme goal specifically aimed at weight management, the intervention was found to be more successful than interventions with programme goals that were aimed at preventing cardiovascular disease or improving general health status. Although a considerable part of studies included motivational interventions, only some actually assessed the effects on potential cognitive mediators. A general lack of reporting underlying theoretical models for behaviour change was observed as well as the inclusion of linkage groups and strategies to promote empowerment.


The prevalence rates of obesity and overweight are rapidly increasing, and the ‘obesity epidemic’ is globally recognized (1–3). Obesity has become one of the major risks to health as it is associated with a wide spectrum of chronic diseases. Lifestyle changes in dietary intake and physical activity contribute to the present situation (3). Although to date interventions have mostly focused on weight loss, prevention of overweight and obesity has increasingly received scientific attention (1,4–11). The goal of these programmes usually is to prevent weight gain across population segments (12). The need for a systematic and theory-based integral approach in interventions has been emphasized (13), but both systematic reviews and meta-analyses addressing this issue are lacking.

The purpose of the present paper was to gain insight in theory-based factors promoting implementation and/or effectiveness in interventions aimed at preventing overweight and obesity in the adult population (≥18 years). The review presents an overview of the evidence that studies have provided on promising and/or effective strategies to change (determinants of) excessive weight gain, and recommendations are given for future effectiveness and implementation studies.


Analytic framework

Several studies revealed that interventions that were developed and implemented in a systematic and planned way were more likely to be effective (14). The Intervention Mapping (IM) protocol is a useful planning framework in the field of overweight and obesity prevention (15,16) and can also be used to evaluate interventions on their merit and to establish gaps in knowledge. Therefore, the IM protocol (14) was applied to analyse the included studies. Although the empirical literature rarely describes all steps and tasks of the IM protocol, the following aspects were studied in the present study:

  • • Description of the target group.
  • • Programme goals and behavioural goals.
  • • Description of the intervention (components, strategies).
  • • Pilot work.
  • • Inclusion and use of strategies to increase sustainability.
  • • Evaluation of the programme.

Classification of interventions

The interventions will be categorized by target group (age, gender, nationality, ethnicity) and setting (workplace, community, health care). Additionally, the Environmental Research framework for weight Gain prevention (EnRG) is applied to classify interventions (17). This framework is proposed as a dual-process view on the environment–behaviour relationship in which energy balance-related behaviours can be the result of both indirect influences (i.e. the mediating role of behaviour-specific cognitions) and direct mechanisms (automatic, unconscious influences). EnRG integrates the ecological perspective on energy balance-related behaviour with the socio-cognitive view. The ecological approach incorporates the Analysis Grid for Environments Linked to Obesity Framework (18). This framework enables the identification of potential intervention settings and strategies by distinguishing four types of environments: physical, economic, political and socio-cultural. The other main ingredient of the EnRG model is derived from one of the most popular socio-cognitive behavioural determinants models, the Theory of Planned Behaviour (19). Using the EnRG framework to analyse and classify interventions may provide a better insight into specific theoretical contents of interventions.

Search strategy

For the present study, all individual papers in three recent reviews on interventions aimed at preventing obesity (1,3,12) were reviewed. Additionally, we updated these reviews by performing a literature search to identify relevant studies in 2006 and 2007. In addition to a review of personal archives, studies were searched in the computerized database of PubMed including the following keyword combinations: prevention, overweight, obesity, weight gain, community and workplace. Publications were screened on the title (one reviewer; A. R.) and second on the abstract (two reviewers independently of each other; A. R. and S. K.). In order to be included in the present study, studies had to (i) be published between January 1990 and the onset of the review; (ii) be written in English or Dutch; (iii) be primary studies and (iv) include one or more anthropometrical outcome measures (e.g. body weight, body mass index or skin-fold thickness).

Qualitative evaluation

The present study focused on lifestyle interventions aimed at prevention of overweight and obesity. Typically, primary programme objective of such interventions is weight management, prevention of weight gain or moderate weight loss. Included studies were examined using a standard data abstraction form. The studies were systematically evaluated on the quality of the intervention according to the IM protocol, the type of intervention according to the EnRG framework, the quality of the research design using the Maryland Scientific Method Scale (SMS) (Box 1) and effectiveness.

Box 1 The Maryland Scientific Method Scale, adapted for anthropometric outcome measures.

1Correlation between the prevention programme and an anthropometric measure at one point in time.
2Anthropometric measures before and after the programme but with no comparable control group.
3A comparison between two matched groups, one with and one without the intervention.
4Anthropometric measures before and after the programme with multiple experimental and comparison condition, controlling for other variables that influence outcome.
5Random assignment of individuals to programme and control conditions plus anthropometric measures before and after the programme.

Quantitative evaluation

Multiple potential moderators of intervention effectiveness were distracted from the qualitative analyses (Tables S1–S16), and they were quantified. These potential moderators represented each of the steps of systematic health promotion. Each of the steps and tasks of the IM protocol, as described above, was operationalized in the following manners:

  • 1Description of the target group.
    • • Sample characteristics: mean age, proportion of females in the sample.
    • • Setting: community vs. other settings.
  • 2Programme goals and behavioural goals.
    • • Programme goal: weight management vs. other goals.
    • • Behavioural goal: diet and physical activity vs. single behaviours.
  • 3Description of the intervention.
    • • Cognitive and environmental goals: both types of goals included vs. one of both.
    • • The duration of the intervention period (number of months).
  • 4Pilot work.
    • • Use of pre-test or pilot test (yes/no).
  • 5Inclusion and use of strategies to increase sustainability.
    • • Inclusion of strategies to increase empowerment (yes/no).
  • 6Evaluation of the programme.
    • • The quality of the research design (SMS; Box 1).
    • • The use of formative evaluation (yes/no).

The operationalization of the research model does not aim at incorporating the full set of potential determinants of intervention effectiveness. However, it can be interpreted as a structured set of indicators of systematic prevention of overweight and obesity. We hypothesized that each of the indicators would be positively associated with intervention effect size.

We calculated effect sizes – Cohen's d; (20)– for tests of differential change in weight across the intervention and control conditions, by dividing weight changes by standard deviation (SD) of baseline weight. As Cohen's effect size (d) has an upward bias, we calculated an adjusted effect size obtained by multiplication with [1 − 3/(4n − 9)](21) (n being the sample size). To calculate the average adjusted effect size and its confidence interval (CI) as well as to perform a fixed-effects meta-regression examining potential moderators of the effect size, spss macros developed by Lipsey and Wilson (22) were used. Although other indicators of change in body composition were used in several trials such as skin-fold thickness and waist circumference, these latter outcomes were operationalized inconsistently and were collected only in a subset of trials. Effect size was calculated when the research design included a control group (SMS score ≥ 3), when the sample included more than 20 participants and when the necessary data (baseline body mass index/weight and SD as well as change in body mass index/weight of both intervention and control group) were available in the selected papers. In case of more than one intervention group, we assessed the effect size for the most intensive intervention.

Next to descriptive analyses, both univariate and multivariate fixed-effects meta-regression analyses were executed. We performed the Q-test to examine (residual) heterogeneity in order to determine the necessity of executing random-effects analyses. The potential moderators were first separately entered into the analysis model (univariate), followed by an analysis of the full model by simultaneously entering all variables. As a result of the relatively small number of studies in the analyses, the critical P-value was set at ≤0.15.


Qualitative analyses

Worksites provide many opportunities to reinforce the adoption and maintenance of healthy lifestyle behaviours and to reach large numbers of individuals of various socioeconomic levels and ethnic backgrounds. Five studies in which workplace interventions were evaluated are described in more detail in Tables S1 and S9 (23–27). The number of worksites in the selected intervention studies ranged 3–32 worksites, including a range of 299–2041 respondents. None of the studies reported specified data on ethnicity.

In a community-based approach, the community is defined as the context in which health, health behaviour and health promotion attempt to reduce the population burden of, for example, obesity and overweight. Usually, such interventions involve programmes in a neighbourhood, village or suburb. Fifteen community-based intervention studies were selected (see Tables S2 and S10) (28–42). The number of respondents included in the selected studies varied from 19 to 2829, and participants were primarily White Americans. The interventions were mainly developed and evaluated in the USA. One intervention study was executed in the Netherlands (40).

Five intervention studies were identified that targeted a general patient population or high-risk patients for primary prevention (Tables S3 and S11) (43–47). All studies were conducted in the UK or the USA; the number of participants ranged from 40 to 12 472. Three studies included high-risk patients (44,45,47), and two studies were led by healthcare providers (43,46).

The fact that health disparities exist between racial and ethnic groups emphasizes the need to develop culturally appropriate programmes for diverse ethnic groups (14). We identified four studies that met the inclusion criteria (see Tables S4 and S12) (48–51). The participating individuals in the identified intervention studies were African–American women (49), Hispanics (48), women derived from diverse ethnic and socioeconomic backgrounds (50) as well as African–Americans (51).

Prevention of excessive gestational weight gain can be realized by effective interventions for pregnant women. Four studies were identified (see Tables S5 and S13) (52–55). The number of participating pregnant women varied from 79 to 560, and the majority was White (52–54).

For women, the years surrounding the menopause are associated with weight gain and increases in low-density lipoprotein-cholesterol and total cholesterol (56,57). Therefore, prevention of weight gain during this time in a women's life is an important health goal (Tables S6 and S14) (56–58). The Women's Health Initiative Dietary Modification Trial (58) included almost 49 000 women, whereas the Women's Healthy Lifestyle Project (56,57) included 535 women. Both interventions were long-term projects and conducted in the USA. The participating women were primarily White (>80%).

Individuals that quit smoking, especially women, report concerns about post-cessation weight gain. Eight studies were identified that investigated the effect of smoking cessation projects including weight-control interventions (Tables S7 and S15) (59–66). A total of six intervention studies included only female smokers (60–62,64–66), and two were targeted at both male and female smokers (59,63). The number of participants ranged from 67 to 417 smokers, and most studies were conducted in the USA.

Prevalence of overweight and obesity in persons with intellectual disability is at least as high as in the general population (67). A number of risk factors have been suggested to explain the increased prevalence of obesity in this group such as independent mobility, barriers to leisure and exercise facilities as well as specific syndromes. Two intervention studies were identified that targeted at adults with learning disabilities (68) and adults with Down's syndrome (67) (see Tables S8 and S16).

Quantitative analyses

Of the 46 identified studies, 28 studies (61%) were executed in the USA, 14 in Europe (five in UK), two in Australia and two in Canada. The participants in the studies ranged between 19 and 48 835 (mean 1892; SD 7328). Mean body mass index at baseline was 27.3 (SD 3.4). The mean age of the population across all included studies was 42.1 years (SD 9.8; Table 1). Excluding 18 studies that consisted exclusively of women (e.g. pregnant or [pre]menopausal women), the samples consisted of, on average, 58% (SD 21.0) of women. Mean duration of the interventions was 18.9 months (SD 22.4). Slightly less than half of the studies (45.7%) revealed significant intervention effects on body mass index. In total, 34 studies (74%) enabled the calculation of effect sizes (see Tables S9–S16 for effect sizes of individual studies).

Table 1.  Descriptive information regarding potential moderators of intervention effectiveness
ModeratorNumber of studies (%)Mean (standard deviation)
Mean age (years) of sample (range 19.7–62.3) 42.1 (9.8)
 <305 (10.9) 
 30–4012 (26.1) 
 40–5018 (38.1) 
 >5011 (23.8) 
Female proportion in the sample (range 0–100)74.8 (26.3) 
 Workplace interventions5 (10.9) 
 Community-based interventions15 (32.6) 
 Interventions in healthcare setting5 (10.9) 
 Interventions for ethnic minorities4 (8.7) 
 Interventions for pregnant women4 (8.7) 
 Interventions for (pre)menopausal women3 (6.5) 
 Interventions for smoking cessation8 (17.4) 
 Interventions for people with intellectual disabilities2 (4.3) 
Programme goal  
 Weight management32 (69.6) 
 Cardiovascular disease prevention12 (26.1) 
 General health2 (4.3) 
Behavioural goal  
 Diet and physical activity26 (56.5) 
 Diet6 (13.0) 
 Physical activity7 (15.2) 
 No specific behavioural goal reported7 (15.2) 
Intrapersonal/cognitive goals  
 Awareness26 (56.5) 
 Knowledge34 (73.9) 
 Attitude30 (65.2) 
 Perceived social influence19 (41.3) 
 Self-efficacy21 (45.7) 
 Goal-setting8 (17.4) 
 Habit1 (2.2) 
 Taste preference1 (2.2) 
Environmental goals  
 Physical9 (19.6) 
 Political5 (10.9) 
 Economical3 (6.5) 
 Socio-cultural14 (30.4) 
Both cognitive and environmental goals21 (45.7) 
One of both types of goals25 (54.3) 
Use of pre-test/pilot test  
 Yes12 (26.1) 
 No34 (73.9) 
Duration (months) of the intervention period (range 1.5–84) 18.9 (22.4)
 Not reported1 (2.2) 
 <311 (23.9) 
 3–69 (19.6) 
 6–1212 (26.1) 
 >1213 (28.3) 
Strategies to increase empowerment  
 Yes18 (39.1) 
 No28 (60.9) 
Quality of the research design (Maryland Scientific Method Scale [SMS]; range 2–5) 4.1 (1.0)
 Anthropometric measures before and after the programme but with no comparable control group (SMS 2)3 (6.5) 
 A comparison between two matched groups, one with and one without the intervention (SMS 3)10 (21.7) 
 Anthropometric measures before and after the programme with multiple experimental and comparison condition, controlling for other variables that influence outcome (SMS 4)12 (26.1) 
 Random assignment of individuals to programme and control conditions plus anthropometric measures before and after the programme (SMS 5)21 (45.7) 
Use of formative evaluation  
 Yes7 (15.2) 
 No39 (84.8) 
Significant intervention effect on body mass index  
 Yes21 (45.7) 
 No25 (54.3) 

The adjusted effect sizes ranged from −0.09 to 0.45; although the average was small (0.06), it significantly differed from 0 for both a fixed effects (95% CI = 0.04–0.08, z = 7.20, P < 0.001) and a random-effects analysis (95% CI = 0.04–0.08, z = 6.75, P < 0.001). As can be seen from Table 2, the univariate regressions showed that when the programme goal was specifically aimed at weight management, the intervention was found to be more successful than programme goals that were aimed at preventing cardiovascular disease or improving general health status (standardized β = 0.37; P = 0.01). Other relatively large positive standardized β (P < 0.15) were found for the application of a formative evaluation and the mean age of the sample. When the intervention was accompanied by a formative evaluation, this increased the effectiveness as measured through the adjusted effect size (standardized β = 0.27, P = 0.12). In addition, age was found to benegatively associated with programme effectiveness (standardized β = −0.29, P = 0.11), indicating that the interventions are less likely to result in an impact on the participants' body mass index as the target group is older. Examining the results of the multivariate model, again, focusing the programme on weight management was found to increase the effectiveness (standardized β = 1.18, P = 0.01). Additionally, duration of the intervention was found to be positively associated with intervention effectiveness at a level of P ≤ 0.15, while the statistical significance of the explanatory value regarding both age and the use of a formative evaluation was reduced in comparison with the univariate models. No residual heterogeneity was detected (Q = 10.44, degree of freedom = 19, P = 0.94), indicating that no random-effects analysis was needed. This was supported by an additional random-effects analysis, showing that the variance of the random effects was virtually 0.

Table 2.  Results of the univariate and multivariate (enter procedure) fixed-effects meta-regression analyses, with adjusted effect size (Cohen's d) as dependent variable
  1. SMS, Maryland Scientific Method Scale.

Mean age of sample−−0.160.740.65
Female proportion in the sample−0.060.720.00−0.040.94
Intervention in community setting (no/yes)0.070.670.010.240.60
Programme goal weight management (no/yes)0.370.
Behavioural goal diet and physical activity (no/yes)0.100.560.01−0.430.33
Both cognitive and environmental goals (no/yes)0.140.430.02−0.280.62
Duration of the intervention period (number of months)−
Use of pre-test/pilot test (no/yes)−−0.300.53
Strategies to increase empowerment (no/yes)0.130.450.020.000.99
Quality of the research design (SMS)−0.040.800.00−0.020.97
Use of formative evaluation (no/yes)


To date, review studies have focused primarily on evidence regarding the effectiveness of overweight prevention interventions (1,3,4,10–12). To our knowledge, the present study is the first to apply a theory-based approach, within the framework of systematic health promotion, in reviewing factors promoting effectiveness of interventions aimed at changing determinants of weight gain in adults. A total of 46 studies evaluating interventions aimed at preventing obesity were reviewed, followed by both qualitative and quantitative analyses.

Interventions were found to have potential in changing energy balance-related behaviours and anthropometric outcomes. Effect sizes for changes in body mass index ranged between −0.09 and 0.45, with a mean effect size of 0.06 (95% CI = 0.04–0.08). Although significantly different from 0, the mean effect size is small, and it is important to recognize that publication bias is likely to favour interventions that show positive impacts. Less-successful studies may not have been reported in the peer-reviewed literature.

As intervention components were generally not evaluated in isolation, it is unclear which separate intervention components contribute most strongly to the impact. Our quantitative analyses did show the importance of formulating the programme goal specifically towards weight management. Those interventions were found to be more successful than interventions that had more broadly defined programme goals (e.g. prevention of cardiovascular disease or improving general health status). The results of the meta-analysis also indicate the importance of executing a formative or process evaluation in addition to assessing effects of the intervention on primary outcomes. Formative evaluation enables the intervention designers to guide programme development or to adapt the intervention as a consequence of identified barriers of successful implementation (14). The positive association between the uses of process-evaluative research methods may also be an indicator of the level of systematic development and implementation of the intervention. In addition, our results indicated that corrected for factors in our model longer interventions appeared to result in a larger effect size. The major problem with weight management programmes has been the difficulty of maintaining health behaviour changes and multiple authors and institutes have stressed the need for duration of programmes of longer than 5 years in order to assess the sustainability of behaviour change and the effects on weight (1,3,12). However, a majority of the identified intervention studies was not designed to show long-term effects on weight-related outcomes. This need for sustained interventions to prevent weight gain applies to all age groups. But it may especially be true for older target groups. Habits become more and more routinized as the behaviours are more frequently performed, and physical abilities and the corresponding variety of opportunities to be physically active decrease as people grow older. These characteristics require interventions to apply a continuous lifestyle approach that incorporates knowledge and promotion of personal opportunities to engage in behaviours that enable the maintenance of an energy balance.

A considerable part of interventions included motivational interventions, but only some studies actually assessed the intervention effects on potential cognitive mediators. In order to gain insight into the effectiveness of the intervention strategies in positively influencing specific cognitive mediators, it is recommended to investigate and report these effects. Although we feel that research aimed at investigating causal chains should primarily be subject of research in smaller scale, relatively controlled experimental research settings, the operationalization of cognitive changes as intermediary indicators is likely to provide important information about the necessary elements of programmes. However, accurate assessment of nutrition, physical activity and psychosocial factors without annoying respondents with very long questionnaires remains challenging.

Although the inclusion criteria for the present review were relatively broad and led to a large number of included studies, the selected studies may still lack several interesting important approaches in the area. For example, faith or church-based interventions may represent a valuable development in the attempt to more efficiently reach ethnic minority groups (69). Although a general lack of this type of interventions has been observed recently (69,70), we did encounter some interesting manuscripts on this issue (71–73). For various reasons, e.g. not incorporating anthropometric outcomes (74), many of these approaches were not reviewed in our study.

A general lack of reporting underlying theoretical models and applied methods for behaviour change was observed. The use of theoretical frameworks may increase effective intervention development and potentially generates data that can be analysed more efficiently and effectively (75). Consequently, the processes underlying successful or unsuccessful interventions will be revealed. An interesting development in this respect is the formation of the Workgroup for Intervention Development and Evaluation Research Group (76), dedicated to improving standards in conduct and reporting of behaviour change interventions. We recommend intervention designers and researchers to increasingly apply systematic approaches in the design, implementation and long-term evaluation of interventions.

Conflict of Interest Statement

No conflict of interest was declared.


The authors would like to thank Babs Franssen for her cooperation in the data-gathering process. The present study was funded by the Netherlands Organisation for Health Research and Development (ZonMw).