Two-Year Internet-Based Randomized Controlled Trial for Weight Loss in African-American Girls

Authors


Pennington Biomedical Research Center, 6400 Perkins Road, Baton Rouge, LA 70808. E-mail: WilliaDA@pbrc.edu

Abstract

Objective: A randomized controlled trial tested the efficacy of an internet-based lifestyle behavior modification program for African-American girls over a 2-year period of intervention.

Research Methods and Procedures: Fifty-seven overweight (mean BMI percentile, 98.3) African-American girls (mean age, 13.2 years) were randomly assigned to an interactive behavioral internet program or an internet health education program, the control condition. Overweight parents were also participants in the study. Forty adolescent-parent dyads (70%) completed the 2-year trial. Outcome data including BMI, body weight, body composition, and weight loss behaviors were collected at baseline and at 6-month intervals. A computer server tracked use of the web sites.

Results: An intention-to-treat statistical approach was used, with the last observation carried forward. In comparison with the control condition, adolescents in the behavioral program lost more mean body fat (BF) (−1.12 ± 0.47% vs. 0.43 ± 0.47% BF, p < 0.05), and parents in the behavioral program lost significantly more mean body weight (−2.43 ± 0.66 vs. −0.35 ± 0.64 kg, p < 0.05) during the first 6 months. This weight loss was regained over the next 18 months. After 2 years, differences in fat for adolescents (−0.08 ± 0.71% vs. 0.84 ± 0.72% BF) and weight for parents (−1.1 ± 0.91 vs. −0.60 ± 0.89 kg) did not differ between the behavioral and control programs.

Discussion: An internet-based weight management program for African-American adolescent girls and their parents resulted in weight loss during the first 6 months but did not yield long-term loss due to reduced use of the web site over time.

Introduction

Over the past 2 decades, the prevalence of obesity in adults has increased at an alarming rate in all segments of society (1, 2, 3, 4). This epidemic of obesity also applies to children and adolescents (5, 6). Given that overweight children and adolescents with at least one obese parent typically become obese adults (7), it is likely that prevention and early treatment efforts could be most effective if aimed at this target group (8, 9). African-American women are particularly vulnerable to obesity (10), have a higher mortality rate from cardiovascular diseases, and have a higher prevalence of diabetes and hypertension (11) in comparison with women from most other racial and ethnic groups. The increased prevalence of obesity in adult African-American women extends to African-American girls (5). Because puberty is associated with increased adiposity, it is desirable to introduce a treatment program during late childhood or early adolescence (8, 9). This reasoning led to our decision to target 11- to 15-year-old African-American girls who were at risk for chronic obesity. The study was planned as a secondary prevention program (12) whereby participants who were highly susceptible to developing chronic obesity as an adult received preventive weight loss therapy at a younger age (13).

Traditional face-to-face weight management counseling may pose several problems for African-American participants. For example, outpatient weight management studies have reported that African Americans attend fewer sessions, have higher attrition rates, and lose less weight compared with whites (14, 15, 16). Also, the university-based clinic is unlikely to be as familiar or convenient as a community-based site, which may affect participant's willingness to participate for extended periods of time. Furthermore, for low-income African Americans, transportation and childcare may pose considerable problems. Based on these considerations, we designed a randomized controlled trial to test the efficacy of an internet-based intervention that would overcome some of the difficulties associated with traditional face-to-face weight management counseling. To test the efficacy of this approach, an interactive internet-based behavioral weight management program was compared with a passive internet-based health education program. The research study targeted a population at great risk for chronic obesity in adulthood: African-American adolescent girls who were overweight, defined by a BMI > 85th percentile for girls ranging in age from 11 to 15 years, with at least one obese biological parent (defined by a BMI > 30). A main component to this weight loss trial was the use of the internet to disseminate information to study participants, reducing the number of face-to-face counselor sessions. The study design has been described in detail in earlier reports (17, 18). Recently, a series of studies have reported on the efficacy of internet-based weight management programs (19, 20, 21, 22, 23). These studies have yielded mixed results, and only one (19) has targeted pediatric obesity. Furthermore, no study has tested the efficacy of a family-oriented internet-based intervention that aimed to promote weight loss in at-risk adolescents with an overweight parent (24).

Recently, we reported preliminary results (18) at 6 months for the study named Health Improvement Program for Teens (HIPTeens).1 The preliminary results indicated that in comparison with a control condition, adolescents in behavioral internet-based treatment lost more body fat (BF), and parents lost significantly more body weight during the first 6 months of a 2-year randomized clinical trial. The present paper provides the long-term (24-month) outcomes for the HIPTeens study. Another paper (17) reported psychosocial correlates of changes in adiposity after 6 months of the intervention. This paper also extends this discussion to outcomes after 2 years of treatment.

Research Methods and Procedures

Research Design

The study compared two treatment arms in a randomized clinical trial: interactive behavior therapy vs. a passive health education control group. The treatment programs were continuously available for use by participants over a 24-month period. Outcomes were assessed at baseline, before initiating the trial, and at 6-month intervals over the course of the 24-month trial. The primary hypothesis of the study was that an interactive behavioral intervention would yield greater reductions in adolescent and parental adiposity relative to a passive health education intervention. The number of face-to-face counseling sessions (four) received by each group was equivalent, with both of the interventions being delivered by the internet. To overcome the obstacles related to disparities in access to the internet, the project provided low-cost computers and free internet access to participants in both treatment groups. The primary participants were the adolescents, and they were randomly assigned to the treatment arms using a stratified randomization strategy based on BMI percentile and age.

The two treatment arms had a similar structure to control for non-specific factors such as amount of therapeutic contact. The internet-based behavioral intervention included the provision of nutrition education plus an internet counseling behavior modification program that targeted the lifestyle eating and physical activity habits of the overweight adolescent and obese parent. The control condition included education on healthy nutrition and exercise, but behavioral changes for the adolescent or parent were not prescribed, and internet counseling was not provided. Thus, the behavioral program was highly interactive, and the control program was very passive, in terms of the structure and content of the two programs. In families with two obese parents, one parent was required to attend all measurement sessions and all face-to-face counseling sessions so there was only one designated obese/overweight parent for each family. This designated parent (56 mothers and 1 father) was required to have a BMI > 27. BMI > 27 (as opposed to BMI > 25) was selected to increase the likelihood that one parent would be motivated to lose weight and, thereby, be an active participant with the daughter.

Participants

The study enrolled 57 African-American adolescent girls (ages 11 to 15 years) who were overweight or obese and had at least one biological parent who was obese (BMI > 30). Inclusion criteria for the study included the following parameters: African-American girl between the ages of 11 and 15 years; BMI above the 85th percentile for age and gender based on 1999 National Health and Nutrition Examination Study normative data; at least one obese biological parent, as defined by BMI > 30; one designated parent who was overweight (BMI > 27) and was willing to participate in the study; adolescent's family was willing to pay $300 out-of-pocket expenses toward the purchase of the computer worth ∼$1000; and the family home had electricity and at least one functional telephone line. Of the 57 participants who enrolled in the study, 40 (70%) completed the 2-year trial. Of the 17 participants who dropped out of the study, 7 were from the control group, and 10 were from the behavioral group.

Weight Management Programs

The two secured web sites were designed to present culturally specific information, including links to other web sites containing culture-specific information such as recipes for foods commonly eaten by African Americans. The counselors were educated on culturally relevant dietary and physical activity issues, and they incorporated this information into the face-to-face and internet counseling sessions.

An internet/e-mail account for each participant was established after the family was enrolled in the study. Both parents and adolescents had e-mail accounts and could log on the web site using separate log-in names and passwords. After randomization and before baseline assessment, adolescents and parents were trained to set up their new computers at home, access and use the web site, and use e-mail to communicate with their counselor by the internet. Each family was assigned a counselor who conducted face-to-face sessions and corresponded regularly using e-mail.

The adolescent participant and her overweight/obese parent (in both treatment conditions) were scheduled to attend four face-to-face counseling sessions during the first 12 weeks of the program, at Weeks 1, 3, 6, and 12. The primary purpose of these sessions for those in the behavioral intervention was to encourage adherence to behavioral principles, to provide additional training related to using their computers to participate in the internet-based programs, and to solve any computer problems. Participants in the control group received nutrition education from a registered dietitian and additional computer training but were not prescribed behavioral tasks to yield weight loss. At the randomization visit, families in both treatment groups were given digital scales to be used for body weight measurement during the 24-month trial.

Internet-Based Behavioral Intervention

This intervention was based on the family treatment methods developed by Epstein et al. (13). The web site provided nutrition education and behavior modification for adults and adolescents using a family-oriented format, i.e., a program that invited the parents, the child, and other members of the family to be involved using mutual problem-solving and behavioral contracting.

Counseling for behavior modification was accomplished primarily by asynchronous e-mail communications. Participants were encouraged to send e-mails weekly to their counselor regarding their progress in the program. Each participant had their own e-mail account that could be accessed through the web site, but the participants could also send e-mails from their own personal accounts. The e-mail communication also allowed counselors to provide feedback on program components (e.g., quizzes, lessons, weight graphs, goal-setting, clinic appointments).

Along with e-mail, the web site contained a variety of interactive components (see Table 1) that were available for the duration of the program. For example, the web site contained weight and activity graphs. Participants were able to self-monitor their weight and physical activity weekly by entering their body weight and minutes of physical activity, which would then be viewable as a graph. Participants could also self-monitor their food intake by entering foods consumed into the web site. They would receive feedback (modeled after the traffic light diet) on the number of servings and caloric content of the food (13). Figure 1 depicts an example of the output from this component of the program. Parents and adolescents were trained to use problem-solving to overcome barriers to success. Figure 2 depicts the screen shot of the page that guided this on-line process. Participants were also instructed to read 52 lesson plans during the 1st year of the program. An example of the first page of Week 1 lesson is shown in Figure 3. A quiz followed every lesson, and participants received instant feedback on their quiz performance. Thus, the web site contained many different elements that were interactive, and none of these elements (except the weight graph and e-mail contact) were available to the participants in the control condition.

Table 1.  Internet intervention components
ComponentsBehavioral siteControl site
E-mail to study staffXX
Weight graphXX
Links to African-American health web sitesXX
Links to women's health web sitesXX
Lessons to be read by adolescents and parentsXX
 Educational for healthy lifestyle X
 Promotion of dietary and exercise behavior changeX 
Physical activity graphX 
Food monitoring with feedback using Traffic Light DietX 
Rate your plate exercise for healthy eating habitsX 
Behavioral contracting with parent, adolescent, counselorX 
Problem-solving for overcome obstacles for successX 
Goal-setting for specific behavior changeX 
Television self-monitoring to reduce sedentary behaviorX 
Quizzes that tested knowledge related to weekly lessonsX 
Figure 1.

Screen shot of feedback from the Food Guide Pyramid. Participants would receive this feedback after entering in foods in the Food Guide Pyramid. Green lights (e.g., turkey sandwich) indicate that this is a low-calorie food. Yellow lights (e.g., potatoes, scalloped) indicate that this food is moderately high in calories. A red light (e.g., chocolate candy) indicates that the food is very high in calories. The number of lights indicates the number of servings reported.

Figure 2.

Screen shot of problem-solving worksheet. This page allows participants to engage in the problem-solving process. This information is sent to the participant's internet counselor.

Figure 3.

Screen shot of the lesson plan for Week 5. This screen captures one of 52 lessons created for the behavioral group. These lessons contained information on healthy eating and exercise.

Internet-Based Control Intervention

Participants in the control condition were provided health education in a coordinated program between face-to-face sessions and links to a variety of web sites promoting healthy lifestyle. The control web site did not provide explicit prescriptions for behavior change, behavioral contracts, or opportunities for self-monitoring. Instead, the control web site was designed to be a passive (non-interactive) program that provided useful health education for the parents and the adolescents by electronic links to other health-related web sites.

Data Collection Procedures

Data were collected in the clinic and over the internet at baseline and Months 6, 12, 18, and 24. Measurement personnel were not blind to treatment arm assignment of the families. For completing each assessment protocol, parents were compensated $30 for their time, and the adolescent received a gift worth approximately $10. The clinic visit included a medical examination, dietary assessment, and measurement of height, weight, and body composition. All other measures were administered over the internet.

Measures

Medical and Eating Disorder Assessment

Adolescent participants and their parents had a medical examination during the baseline period that included assessment of Tanner Stage to record pubertal status of the adolescents. This baseline assessment indicated that 80% of the participants had achieved pubertal status. During the baseline assessment, the adolescents and their parents were screened for eating disorder symptoms and body weight concerns using the Interview for Diagnosis of Eating Disorders IV (25). No cases of eating disorder were identified.

Height and Body Weight

Height and weight of the child and parents were measured (at baseline and at Months 6, 12, 18, and 24) and were converted to BMI (kilograms per meter squared). BMI for adolescent participants was also expressed in terms of percentile for age using the 1999 National Health and Nutrition Examination Study database to correct for instability of BMI values for ages before 18 years.

DXA

DXA was used to estimate percentage BF in adolescents and obese parents at each of the assessment points, i.e., baseline and Months 6, 12, 18, and 24. This procedure, using DXA model QDR 2000 (Hologic, Waltham, MA) in the fan beam mode, involves the measurement of differential absorption rates by tissues of photons or x-rays.

Weight Loss Behavior Scale (WLBS)

A questionnaire developed by Smith et al. (26) measured changes in weight loss behaviors. The reliability and validity of the WLBS have been established (26). Four of the WLBS scales were used in the study: dieting/weight concerns, exercise, overeating, and avoidance of fattening foods. The WLBS was administered to adolescents and parents over the web site at baseline and at each of the four assessment periods.

Web Site Use

Participants were required to log on to a secured web site (password protected). A computer server tracked use of the web sites. Participation in the program was measured by the number of hits on the web site. A hit was defined as accessing a page on the web site. Three indices of web site use were computed: total number of hits (applicable to both web sites), number of hits for accessing the weight graph (applicable to the behavioral program), and number of hits for accessing the physical activity graph (applicable to the behavioral program).

Computer Opinion Survey

This measure was only completed at baseline. The 26-item Computer Opinion Survey (27) assesses attitudes toward computers through questions related to their utility and one's enjoyment, satisfaction, and comfort with computers. Each question is answered using a six-point Likert rating scale ranging from 1 (strongly agree) to 6 (strongly disagree). Scores are summed to form the Computer Anxiety Index, a single-factor test (28), which measures one's interest in and comfort with using computers.

Statistical Analyses

An intention-to-treat analysis was used for all data analysis, using the last value carried forward procedure. (An intention-to-treat analysis was used for all data analysis, although it should be noted that the same results were found with a completers-only analysis.) To test the hypothesis of a treatment effect on outcome measurements, analysis of covariance was performed on the change from baseline values for data collected at each of the 6-month measurement points. For each variable of interest, the baseline score was included in the statistical model as a covariate. To test for the effect of time, the change from baseline was tested against zero. Pearson product moment correlations were calculated between the variables to test for linear relationships. All tests were performed with an α of 0.05. SAS version 8.2 software (SAS Institute, Cary, NC) was used for all analyses performed. The analysis of change from baseline is equivalent to a prepost analysis if the baseline value is included as a covariate. In SAS, the test of least square means for each variable translates into a test of the effect of time. Both treatment and time effects were of interest to us, so by analyzing change from baseline, we were able to test these effects. It should be noted that the results differ slightly from those reported for the first 6 months of treatment (16, 18) because of the higher number of drop-outs in this report of 24-month outcomes.

A sample size/power analysis was performed for the proposed 24-month study. A Student's t test analysis for each variable was used to calculate power for end of the study measurements at 24 months. The response variables of interest were the two primary endpoints BMI and percentage BF. The basic assumptions of the power analysis are those of a basic Student's t test for means, including normality, independence, and common variance between the two groups. Based on the literature, a difference of 15% for BF and 2.5 units for BMI was hypothesized between the group means. The sample size analysis for detecting this difference was powered for at least 80% for both endpoints. The power analysis indicated that at an end of the study sample size of 20 subjects per group was sufficient to detect these differences between groups.

Results

Baseline

Table 2 summarizes the baseline characteristics of the sample. Adolescents in the program were very overweight, averaging in the 98th percentile for age and height. The parents were middle-aged and moderately obese.

Table 2.  Baseline characteristics of the sample
 AdolescentsParents
  1. Means (and standard deviations) are shown for each variable (n = 57 adolescents and 57 parents).

Age (years)13.2 (1.4)43.2 (6.2)
Height (cm)160.0 (8.1)162.3 (6.9)
Weight (kg)93.3 (22.5)101.2 (18.4)
BMI36.4 (7.9)38.4 (7.2)
Body fat DXA (%)45.9 (7.5)48.4 (6.3)
BMI percentile98.3 (2.5) 

Attrition/Attendance

Seven families dropped out of the control condition (24%), and 10 dropped out of the behavioral treatment (36%), but this difference was not statistically significant (χ2 = 0.91; p = 0.34). Completers and drop-outs were compared on baseline values for: BMI, percentage BF, and scores on the four WLBS scales. Parents and adolescents who dropped out of the program did not differ from those who remained, with one exception. Adolescents who dropped out had lower scores on the exercise subscale of the WLBS compared with those who remained. No differences were found for parents. Attendance at the four face-to-face counseling sessions was 100% for participants who completed the 24-month study.

Changes in Adiposity

Figure 4 depicts the findings related to changes in BMI, body weight loss (kilograms), and changes in estimates of percentage BF for adolescents and parents across the 24-month study. Table 3 summarizes changes in weight, BMI, and percentage BF at Month 24. The left side of Figure 4 (A C, and E) displays results for the adolescents, and the right side of the figure (Figure 4 B D, and F) displays results for the parents. As reported earlier (17, 18), in comparison with the control condition, adolescents in the behavioral treatment lost more BF and parents lost significantly more body weight during the first 6 months. During the next 18 months, parent and adolescent participants in both groups gained weight, and at 2 years, the weight/fat of the two treatment groups did not differ. Data analysis indicated that body weight at baseline was a significant covariate for body weight changes of parents, F(1,54) = 27.2, p < 0.0001, and that baseline BMI percentile was a significant covariate for changes in BMI percentile, F(1,54) = 10.89, p < 0.001. For both covariates, higher levels of body weight were associated with greater weight loss over the 24 months of treatment. The overall effect of treatment was significant for parental changes in BMI, F(1,54) = 4.78, p < 0.04, indicating that parents in the behavioral treatment arm had greater reductions in BMI in comparison with the control arm. Post hoc comparisons of groups (see Table 3) indicated that this difference was statistically significant (p < 0.05) at Months 6 and 12 but was not significant at Months 18 and 24 (p > 0.05). Significant changes over time were observed for adolescent changes in body weight, F(3,54) = 6.55, p < 0.001, BMI percentile changes, F(3,54) = 4.14, p < 0.02, and BMI, F(3,54) = 3.13, p < 0.04.

Figure 4.

Graphs of weight (kilograms) change [± standard error (SE)], DXA percentage BF change (±SE), and BMI change (±SE) from baseline for 6, 12, 18, and 24 months of treatment for adolescents (A, C, and E) and parents (B, D, and F) in each treatment group.

Table 3.  Change scores for body weight/adiposity outcome variables after 24 months
 ControlBehavioral
 AdolescentsParentsAdolescentsParents
  1. Adjusted means ± standard errors are shown for each variable.

Weight (kg)6.3 ± 1.6−0.60 ± 0.894.4 ± 1.7−1.1 ± 0.91
BMI1.2 ± 0.650.04 ± 0.340.73 ± 0.66−0.55 ± 0.34
Body fat DXA (%)0.84 ± 0.720.51 ± 0.46−0.08 ± 0.710.36 ± 0.46
BMI percentile−0.001 ± 0.003 −0.004 ± 0.003 

Web Site Use

Figure 5 summarizes use of the web sites during Years 1 and 2. Figure 5 A and B display the graphs for mean web hits for the behavioral and control adolescents and parents, respectively. These graphs display the significant differences in web site use between the behavioral and control groups, for both adolescents and parents, during the 1st year of treatment: adolescents, F(1,55) = 4.95, p < 0.05; and parents, F(1,55) = 4.15, p < 0.05. However, these differences diminished during the 2nd year of treatment, with both groups showing almost equal and very reduced web site usage.

Figure 5.

Mean (±SE) web hits (per participant) by treatment group for (A) adolescents and (B) parents at 12 and 24 months into treatment. (C) Mean (±SE) weight graph hits (per participant) for adolescents and parents in the behavioral group. (D) Mean (±SE) activity graph hits (per participant) for adolescents and parents in the behavioral group.

Figure 5 C and D, depict the average frequency of entry of weight and physical activity data by parents and adolescents in the behavioral group. Both figures show that adolescents and parents in the behavioral group utilized the weight and activity graph components more frequently during the 1st year of the study in comparison with the 2nd year. Correlations were conducted between the computer opinion survey and web site use during the 1st and 2nd years of the study. Although no significant effects were found, parents in the experimental condition who reported more interest and comfort with computers (at baseline) showed trends for more web site hits [r(13) = 0.52; p < 0.07] and more e-mail use [r(13) = 0.54; p < 0.06] during the 2nd year of the study.

Weight Loss Behaviors

Figure 6 displays changes in weight loss behaviors from baseline for adolescents and parents, as reported by the WLBS. For all four WLBS scales, the baseline score was a significant covariate (all p values < 0.0001), indicating that higher scores at baseline were associated with greater improvement for adolescents and parents. For dieting and weight concerns of adolescents, the interaction of treatment and time approached significance, F(3,48) = 2.44, p < 0.08 (Figure 6A). The adolescents in the control arm decreased dieting and weight concerns during the first 6 months, but their scores returned to baseline levels during the next 18 months. Figure 6 C and E illustrate that adolescents in both the behavioral and control groups reported improvement in exercise and overeating, in comparison with baseline (p < 0.05). Adolescents in the behavioral group reported eating less fattening foods (Figure 6G) in comparison with the control group, F(1,48) = 2.08, p < 0.05. Parents in both groups reported increased weight and dieting concerns (Figure 6B). Parents in the behavioral arm reported more exercise (Figure 6D) than did the parents in the control group, F(1,49) = 4.10, p < 0.05. Parents in the behavioral group reported greater reductions in overeating, especially during the first 6 months (Figure 6F). Parents in both groups reported improvements in avoidance of eating fattening foods (Figure 6H) during the 24-month trial, in comparison with baseline.

Figure 6.

Graphs of WLBS dieting/weight concerns, exercise, overeating, and avoidance of fattening foods change (±SE) from baseline for adolescents (A, C, E, and G) and parents (B, D, F, and H) in each treatment group at 6, 12, 18, and 24 months of treatment.

Behavioral Correlates of Changes in Adiposity

The four scales of the WLBS (at 24 months) were correlated with changes in body weight and changes in percentage BF at 24 months. Self-reported level of exercise was the only behavior significantly associated with changes in BF, r(18) = −0.65, p < 0.01, and body weight, r(18) = −0.42, p < 0.05, for adolescents in the behavioral arm. Similar correlations were found for parents in the behavioral arm: between exercise and BF, r(18) = −0.49, p < 0.05, and between exercise and body weight, r(18) = −0.55, p < 0.05. For adolescents and parents in the control arm, no significant behavioral correlates of BF and body weight changes were found.

Discussion

An internet-based behavioral intervention was found to be superior to internet-based health education because it yielded decreased BF for adolescent girls and decreased body weight for parents after 6 months of treatment (17, 18); however, these weight/fat losses were not maintained during the subsequent 18 months.

As shown in Figure 5, use of the web site decreased dramatically after the 1st year of the study. These findings suggest that four face-to-face sessions during the first 12 weeks combined with an ongoing interactive web site was insufficient to sustain web site use over a 2-year period.

Figure 6 illustrates that in general, both adolescents and parents reported improved weight loss behaviors, regardless of treatment condition. The only exception was the decrease in dieting and weight concerns seen by adolescents in the control group after the first 6 months. The exercise habits of parents in the behavioral arm improved to a greater degree than those in the control group. Furthermore, self-reported levels of exercise for adolescents and parents in the behavioral arm were associated with greater weight loss and fat loss at 24 months. This correlation was not observed in the control arm. Adolescents who dropped out of the program reported lower levels of exercise compared with those who completed the program. These findings suggest that increased exercise may have been important for successful weight and/or fat loss.

These findings are best interpreted in the context of findings from other internet-based weight management trials for adults and children. It should be noted that these studies are primarily internet-based because all of them, including the current one, have included some face-to-face sessions. Five other studies have tested the use of the internet for the purpose of delivering a weight management program (19, 20, 21, 22, 23). Four were studies of adults and the duration of the internet-based interventions ranged from 6 to 12 months with study samples that were predominantly non-Hispanic white. The research designs of two of these studies (22, 23) were very similar to the design of this study in that an interactive internet-based intervention was compared with a health education web site, but the intervention was much shorter in duration than the current study. Both of these studies reported relatively small weight losses by adults assigned to the internet-based behavioral treatment arm. Harvey-Berino and colleagues (20, 21) have conducted two studies that used an internet-based approach for prevention of weight gain after weight loss induction. The first study (20) found that the internet-based intervention did not yield satisfactory weight maintenance. In contrast, a more recent study (21) reported no differences in weight maintenance between face-to-face contact and internet support. Therefore, the bulk of the evidence suggests that internet-based interventions can yield small weight losses at the initiation of treatment (22, 23), but satisfactory weight maintenance has been reported in only one study (21). In the only other study that has investigated the use of the internet with a pediatric population, Baranowski et al. (19) tested the efficacy of a 4-week summer camp, followed by an 8-week internet-based intervention, for prevention of weight gain in 8-year-old African-American girls. This study did not report significant weight changes in comparison with a control group, and the duration of the intervention was quite short. Overall, the findings of these studies of internet-based weight management interventions suggest that adults, children, and adolescents who are overweight will use the programs, and these programs generally yield small but statistically significant weight changes that are relatively short-lived. Furthermore, in the HIPTeens project, Figure 5 clearly shows that we were unable to sustain use of the web site over a period of 2 years with minimal supervision of this activity. Based on these observations, we have recommended (24) that internet-based weight management interventions should be integrated into other forums for supervised behavior change, e.g., health classes in schools, and that they might best fit with goals of relatively small weight losses or weight gain prevention (12).

The findings can also be interpreted within the context of clinic-based weight management programs for children and adolescents. Epstein et al. (13) have reported that family-based treatments result in significant weight loss in children that has been maintained for as long as 8 to 10 years after treatment (13). Other studies (29, 30, 31, 32, 33) have reported that clinic-based interventions, including family therapy, for children and adolescents yield initial weight loss, followed by weight regain over the course of several years of follow-up.

Some school-based interventions have also reported positive results (34, 35), whereas others have not found an effect on weight status (36, 37). Of special relevance, Fitzgibbon et al. (38) reported that a 14-week school-based program for young African-American children resulted in smaller increases in weight compared with weight changes of students in control schools.

In this context, our findings are similar to the findings of many other weight management programs with children and adolescents that followed the participants for 2 years or longer. With the notable exception of the studies reported by Epstein et al. (13), most other studies have reported initial weight loss followed by weight regain. Given current evidence, it appears that intensive family-based clinical interventions (13) may be required to yield sustained weight loss.

This study is not without limitations. One of the major limitations of the study was the small sample size. We did, however, retain sufficient sample size to test the primary aims of the study, i.e., comparison of the two treatment arms at 24 months. Although innovative, the intervention delivery may have also functioned as a limitation for this population. Some research related to African-American learning styles suggests that African Americans prefer a cooperative learning style (39, 40). Our program utilized more of an individual learning style, requiring children to learn information primarily on their own by reading information from a web site. Future studies should assess the ability of integrating more cooperative learning styles into a similar program, such as fostering more cooperation between mother and daughter for information or more interactive group-based sessions. Finally, African Americans have less access to the internet than other ethnic groups (41). Scores on the Computer Opinion Survey at baseline were positively correlated with greater web site use for parents in the behavioral group during the 2nd year of the program. It is possible that as the popularity of the internet increases in the African-American community, the approach tested in this study may become more viable.

In summary, the interactive internet-based program was not effective for long-term weight loss. Internet-based weight management interventions may be more appropriate for overweight people who need to lose small amounts of weight, and they may be a useful strategy to prevent weight gain in children, adolescents, or adults. Furthermore, it is our opinion that strategies for promotion of sustained use of internet-based interventions will be a necessary component of this approach, regardless of its use for weight loss or prevention of weight gain.

Acknowledgments

This work was supported by NIH Grant RO1 HD39104-03.

Footnotes

  • 1

    Nonstandard abbreviations: HIPTeens, Health Improvement Program for Teens; BF, body fat; WLBS, Weight Loss Behavior Scale; SE, standard error.

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Ancillary