An Integrative Review of Interventions for Adolescent Weight Loss

Procedure

The initial intent of this study was to conduct a meta-analysis. A review of the literature was conducted by a computer-assisted search using CINAHL (Cumulative Index to Nursing & Allied Health Literature; 1982–2003) and Medline (1966–October 2003) search resources. Key words used in different combinations for the searches were overweight, adolescent, obesity, and intervention. Reference lists of articles identified for the critical review were reviewed for additional articles that may not have been identified in the computerized search. Journal articles published in nursing, psychology, nutrition, medicine, and exercise physiology literature between 1980 and 2003 were retrieved. A total of 152 studies were identified and reviewed. Most studies did not include comparison or control groups, included participants younger or older than the inclusion criteria, or were not interventions for weight loss; only 17 studies met the inclusion criteria.

A codebook was developed by the investigators to classify substantive and methodological variables (McCain, Smith, & Abraham, 1986). Substantive variables included age, gender, parental participation, ethnic group, Tanner stage, and socioeconomic status. Methodological variables included sampling method, recruitment methods, length of intervention, study design, and intervention specifics. Interventions were coded according to their behavioral focus: nutritional intake, exercise activity, or lifestyle change education. Interrater reliability was ≥90% when tested for 7 of the 17 coded studies by three researchers coding the same article and comparing the findings. The quality of each study was rated using a modified instrument reported by Beck (1995). The purpose of the quality assessment was to evaluate the rigor of methods used for each study. Areas in the quality rating instrument included the published experience of the first author, funding, sampling, measurement, design, and statistical reporting. If extensive variation in quality are found, low-quality studies should not be included in the analyses. Further statistical analysis of the studies was rejected because of inconsistencies in the types of interventions, of variables measured, and reporting of statistical findings. Such limitations prevented a meta-analysis to identify the effect size of the interventions used in the studies.

Descriptive Profile of the Domain

The 17 studies that met inclusion criteria are listed in Table 1. There was a diversity of sampling methods for the studies, with seven using a stratified random assignment method and the remaining samples divided between convenience samples and random samples. Recruitment methods included the use of schools, health care professionals, newspaper ads, a local newsletter, radio ads, community meetings, and a hired recruiter. A classic experimental design was used in 11 of the studies; four of the studies used a multigroup, repeated-measures quasi-experimental design. The remaining designs were a quasi-experimental nonequivalent control group and a one-group pretest–posttest design. The majority of the interventions lasted between 12 and 20 weeks. Quality scores ranged from 14 to 26, with a mean of 18.4 (SD 3.12) for the 17 studies (Table 1). The maximum possible quality score was 30.

The 17 studies that met inclusion criteria included 831 participants. In studies reporting demographic information, 68% of the participants were girls, and 32% were boys. Sample sizes ranged from 20 to 82. The studies that reported ethnic participation included 44% White/Caucasian, 38% African American/Black, 0.5% Asian American, 2.5% Latino/Hispanic participants, and 15% were described as “other.”

Types of Interventions

Interventions for adolescent weight loss included treatment with and without family members (Brownell, Kelman, & Stunkard, 1983; Coates, Killen, & Slinkard, 1982; Mellin, Slinkard, & Irwin, 1987; Thomas-Dobersen, Butler-Simon, & Flehner, 1993; Wadden, Stunkard, Rich, Rubin, Sweidel, & McKinney, 1990); interventions with and without exercise (Becque, Katch, Rocchini, Marks, & Moorehead, 1988; Emes, Velde, Moreau, Murdoch, & Trussell, 1990; Gutin, Barbeau, Lemmon, Bauman, Allison, Kang, & Litaker, 2002; Lansky & Vance, 1983; Resnicow, Yarock, Davis, Wang, Carter, Slaughter, Coleman, & Baranowski, 2000; Sasaki, Shindo, Tanaka, Ando, & Arakawa, 1987); interventions with dietary reductions (Ikeda, Fujii, Fong, & Hanson, 1982; Sondike, Copperman, & Jacobson, 2003); an intervention with Metformin to reduce glucose intolerance (Freemark & Bursey, 2001); the use of monetary reward structures (Coates, Jeffrey, Slinkard, Killen, & Danaher, 1982); a telephone and mail-based behavioral intervention (Saelens, Sallis, Wilfley, Patrick, Cella, & Buchta, 2002); and an intervention with sibutramine as an adjunct to behavioral therapy, diet, and exercise (Berkowitz, Wadden, Tershakovec, & Cronquist, 2003).

Thirteen studies used group interventions for behavioral change (Becque et al., 1988; Berkowitz et al., 2003; Brownell et al., 1983; Coates, Killen et al., 1982; Coates, Jeffery, et al., 1982; Emes et al., 1990; Gutin et al., 2002; Ikeda et al., 1982; Lansky & Vance, 1983; Mellin et al., 1987; Resnicow et al., 2000; Thomas-Dobersen et al., 1993; Wadden et al., 1990), and two studies used an individual, nongroup intervention for behavioral change associated with food behavior and physical activity behavior (Saelens et al., 2002; Sondike et al., 2003). The remaining two studies used physiologic interventions (Freemark & Bursey, 2001; Sasaki, Tanaka, Ando, & Arakawa, 1987).

Theoretical Base

Only one study identified the theoretical framework used to guide the intervention as social cognitive theory (Resnicow et al., 2000). Thirteen of the remaining studies focused on behavioral changes but did not clearly describe any theoretical framework. Many of the studies addressed concepts that are part of social cognitive theory such as the knowledge and skill needed for the behavior, expectations, self-control, observational learning, self-efficacy, and reinforcements (Baranowski, Perry, & Parcel, 2002).

Parental Participation

Coates, Killen, and Slinkard (1982) examined the role of the family in adolescent obesity interventions by comparing outcomes produced with and without parental involvement. Both intervention groups attended weekly lifestyle change classes, and at 9-month follow-up, there was no significant difference in the amount of weight loss between the two groups. In 1983, Brownell and colleagues reported a study of the benefit of having parents attend a group separate from the adolescent. Adolescents were assigned to one of three groups: mother–child separately, mother–child together, and child alone without parent involvement in the treatment program. This study found differences in adolescent weight loss between the parental involvement groups; a 1-year follow-up found that adolescents in the mother–child separately treatment group had a 20.5% decrease in the percentage of overweight (Brownell et al., 1983). The effects of frequency of parental involvement on obesity intervention outcomes were examined in the SHAPEDOWN study (Mellin et al., 1987). In this study, parents were asked to attend 2 of 14 sessions; 19.1% of the parents attended both sessions, 68.1% attended one session, and 12.8% did not attend any session (Mellin et al., 1987). A 15-month follow-up found the SHAPEDOWN study group had achieved a significant weight loss (Mellin et al., 1987), but less than the losses found in the Brownell study (1983). The investigators proposed that the difference between the Brownell study and the SHAPEDOWN study was a result of the number of sessions attended by the parents, and possibly that the Brownell study, because of requirements for parental participation, may have self-selected adolescents with more supportive parents (Mellin et al., 1987). In a study of female African American adolescents, there was no evidence of greater weight loss when separate parent and adolescent groups were used (Wadden et al., 1990). Although positive interactions in the joint parent and adolescent group were reported, no significant difference in weight loss was found between the treatment groups (Wadden et al., 1990).

Exercise Interventions

School-based interventions have achieved effective weight loss with combined lifestyle change education and exercise (Lansky & Vance, 1983), as well as an exercise-only intervention (Sasaki et al., 1987). Lansky and Vance (1983) conducted 12 weekly meetings with 30 adolescents aged 12–14. The 45-minute meetings were used to encourage changes in eating and exercise behaviors. At the end of 12 weeks, there was a significant difference in the percentage of overweight between the control group and the intervention group (p <.01). No further follow-up was reported.

Sasaki et al. (1987) conducted a 2-year study of 41 overweight children with a 7-day-a-week aerobic exercise program. At the end of the study, the participants demonstrated a significant reduction in their percent of body weight that was in excess of the average body weight for age in Japan (p <.001). Gutin and colleagues (2002) reported cardiovascular fitness improvements and a decline in total body and visceral adiposity after a lifestyle education and physical training program conducted after school. No benefits from a faster increase for level of exercise were found for reduction of BMI (Emes et al., 1990) or percentage of body fat (Gutin et al., 2002) compared with a progressively slower increase to the same overall level of exercise.

Dietary Interventions

Ikeda and colleagues (1982) compared the use of behavioral change education in groups with and without diet restrictions. Less focus was placed on behavioral changes in the group with the dietary restriction, but they were weighed weekly. The group exposed to a behavioral-focused intervention was weighed three times during the 14-week program. There was no significant difference in weight loss between the two treatment approaches.

Sondike and colleagues (2003) compared a low-carbohydrate diet with a low-fat diet and found that the low-carbohydrate group achieved the greater weight loss (p <.05) at the end of 12 weeks. Informal follow-up at 1 year identified sustained weight loss in 8 participants of the low-carbohydrate group, leading investigators to theorize that a low-carbohydrate diet may achieve better compliance with adolescents than low-fat diets (Sondike et al., 2003).

Monetary Rewards

Coates, Jeffery, and colleagues (1982) evaluated the effect of monetary incentives combined with a therapeutic intervention to evaluate motivational strategies for adolescent weight loss. Participants and their parents established cash deposits at the beginning of the program that were paid out contingent on weight loss or caloric reduction either weekly or five times a week for a 15-week period. Deposits that were not earned back as incentives were forfeited. Greater weight loss was achieved when cash incentives were provided more frequently. In a later study by the same group, however, similar loss of weight was achieved without the financial incentive (Coates, Killen, et al., 1982) .

Attrition and Attendance Rates

Of the 17 studies examined, 13 studies reported attrition rates, which ranged as low as 10% (Brownell et al., 1983; Freemark & Bursey, 2001) to as high as 44–45% (Ikeda et al., 1982; Resnicow et al., 2000), with a mean of 23.23% (SD = 11.6; Table 1). One behavioral intervention study with female adolescents reported that participants missed 43% of the sessions, and the overall attrition rate was 45% (Resnicow et al., 2000). Increased attendance at group sessions was found to be related to increased nutrition knowledge, perceived diet changes, and social support but did not have a statistically significant relationship to physiologic changes, including weight loss (Resnicow et al., 2000).

Telephone and Mail-Based Interventions

Saelens et al. (2002) developed an intervention using a computerized program to enable intervention-group adolescents to enter assessment information and create individualized action plans. The assessment and plan was subsequently reviewed with a pediatrician and the participant. The adolescent and a parent met in person with the investigator to learn food self-monitoring. The intervention staff followed up on the action plan with weekly phone calls to encourage behavioral changes, self-recording of weekly calories and physical activity, and entry into a prize lottery for successful participants. Although weight loss trends were noted in the intervention group, and the comparison group demonstrated weight gain trends, a statistically significant difference between the groups was not identified at the follow-up. The phone intervention period lasted between 14 and 16 weeks. The investigators theorize that a longer intervention might have had more favorable outcomes, based on observed trends for BMI in both groups (Saelens et al., 2002).

Medication Interventions

In a 6-month pilot study of 29 overweight adolescents, Freemark and Bursey (2001) reported that BMI was reduced by 1.3% in an adolescent group taking metformin, whereas the control group BMI increased 2.3% (p < .02). Metformin, an antidiabetic, has been used with obese nondiabetic adults to reduce blood glucose, insulin, and hemoglobin A1c (Freemark & Bursey, 2001). Berkowitz et al. (2003) reported that adolescents who took sibutramine, an antiobesity agent and appetite suppressant, combined with a behavioral, diet, and exercise intervention, had a 4.5% greater BMI loss (p = .001) than adolescents participating in the same treatment with a placebo. Parents of participants in the control group and the intervention group in the sibutramine study participated in separate groups but received the same information provided to the adolescents (Berkowitz et al., 2003).

Lessons From the Research for Evidence-Based Practice

The first lesson learned from this review was that there is a marked absence of studies testing similar interventions and outcomes, which makes it difficult to compare findings. Lifestyle change education was included in 14 of the 17 studies; however, there was a lack of coherence across the types of interventions, with few studies using the same instruments across samples. Other variables associated with lifestyle change included manipulating of levels of parental involvement, reward structures, dietary reductions, frequency of contact, and exercise programs. The variety of interventions created treatment diversity that provided unique evaluations but reduced opportunities for comparison and validation of findings to identify effect sizes reflective of “best practice.”

The second lesson learned is that findings from previous research with overweight adolescents may not be generalizable to today’s overweight adolescent. Twenty-seven percent of male Latinos between the ages of 12 and 19 are reported to be overweight, and an additional 44% are at risk of being overweight with a BMI for age equal to or greater than the 85th percentile (Ogden et al., 2002). However, although male Latino adolescents are the fastest increasing overweight adolescent population, only 2% of participants in published intervention studies were Latino. African American adolescents follow closely with 26.6% of female adolescents overweight and 25% at risk of becoming overweight; 20% of African American male adolescents are overweight, with 35% at risk of becoming overweight. Nine studies reported participation by race or ethnic group and included 44% White participants who represent less than 13% of overweight adolescents. Gender was also disproportionately represented, with studies including only 32% male participants although rates of overweight and at-risk adolescents continue to increase in the male Latino and African American adolescent populations.

The third lesson learned is that parental participation is an important variable that needs to be addressed in future studies. The effect of parental involvement was examined in six of the studies (Brownell et al., 1983; Coates, Killen, et al., 1982; Lansky & Vance, 1983; Mellin et al., 1987; Resnicow et al., 2000; Wadden et al., 1990). Initial findings suggest that separate adolescent and parental treatment groups are more effective for most adolescents (Brownell et al., 1983). African American adolescents responded in a positive manner in treatment groups that included their mothers (Wadden et al., 1990).

A fourth lesson learned is that planned interventions with overweight adolescents must consider and develop methods to reduce attrition rates. Attrition rates and attendance rates in adolescent weight loss intervention studies ranged as high as 45% (Resnicow et al., 2000). High attrition rates place studies at risk of a loss of statistical power, may compromise internal and external validity, and contribute to concerns relating to utilization of research time and funds (Carroll, 1997; Hunt & White, 1998). The impact of attrition lengthens the time required to identify effective outcomes (Davis, Broome, & Cox, 2002). Specific reasons for high attrition rates and poor attendance rates have not been explored in depth. These reasons need to be identified to develop effective methods to reduce their effect within adolescent weight loss studies. Although it must be acknowledged that the duration of the studies, the complex and demanding interventions used, and adolescent academic obligations likely influence dropout rates, future researchers need to develop clear retention strategies that address issues known to influence participation in research (Clark-Jones & Broome, 2001).

The fifth lesson learned is that there are excellent opportunities for future research with overweight adolescents to expand or test more comprehensive theories that address the complexity of adolescent behavior, as well as physiologic and environmental variables. Historically, studies have been limited to theoretical frameworks grounded in social cognitive and behavioral theories. Adolescent obesity interventions may benefit from the inclusion of developing theories of environmental factors as contributors to the obesity epidemic (Egger & Swinburn, 1997; Sallis & Owen, 2002). Ecological models are evolving to address global societal obesity trends with a focus in the areas of the biophysical environment (Egger & Swinburn, 1997; Sallis & Owen, 2002). Ecological models have roots in public health and psychology. Ecological models for health behavior propose that behaviors are influenced by intrapersonal, social, cultural, policy, and physical–environmental factors (Sallis & Owen, 2002). Examples from the ecological model of influences on adolescent weight management include individual food preferences, family and cultural influences, peer influence, availability of vending machines in a school setting, accessibility to fast food, mass-media influences, and policies relating to food availability and physical activity in the school setting. When using the ecological model, planning for weight management would extend beyond the individual to include environmental influences.

A newer contrasting area of emerging theory related to obesity research explores the role of addiction in obesity (Del Parigi, Chen, Salbe, Reiman, & Tataranni, 2003; Pelchat, 2002). However, the evaluation of the possible role of addiction in obesity is still in early stages. Rogers and Smit (2000) described similarities between the positive effects of drugs and foods but concluded that there was a lack of support for physiologic addiction because of the absence of clinical responses to food withdrawal that are present with drug withdrawal. Technology advances, such as positron emission tomography studies, are allowing researchers to evaluate brain response to food intake and are expected to provide a better foundation for understanding physical responses to food intake in the future (Del Parigi et al., 2003).