Mediators of Weight Loss in a Family-Based
Intervention Presented over the Internet
Marney A. White, Pamela D. Martin, Robert L. Newton, Heather M. Walden, Emily E. York-Crowe,
Stewart T. Gordon, Donna H. Ryan, and Donald A. Williamson
WHITE, MARNEY A., PAMELA DAVIS MARTIN,
ROBERT L. NEWTON, HEATHER M. WALDEN,
EMILY E. YORK-CROWE, STEWART T. GORDON,
DONNA H. RYAN, AND DONALD A. WILLIAMSON.
Mediators of weight loss in a family-based intervention
presented over the Internet. Obes Res. 2004;12:1050–1059.
Objective: To assess the process variables involved in a
weight loss program for African-American adolescent girls.
Several process variables have been identified as affecting
success in in vivo weight loss programs for adults and
children, including program adherence, self-efficacy, and
social support. The current study sought to broaden the
understanding of these process variables as they pertain to
an intervention program that is presented using the Internet.
It was hypothesized that variables such as program adher-
ence, dietary self-efficacy, psychological factors, and fam-
ily environment factors would mediate the effect of the
experimental condition on weight loss.
Research Methods and Procedures: Participants were 57
adolescent African-American girls who joined the program
with one obese parent; family pairs were randomized to
either a behavioral or control condition in an Internet-based
weight loss program. Outcome data (weight loss) are re-
ported for the first 6 months of the intervention.
Results: Results partially supported the hypotheses. For
weight loss among adolescents, parent variables pertaining
to life and family satisfaction were the strongest mediating
variables. For parental weight loss, changes in dietary prac-
tices over the course of 6 months were the strongest medi-
Discussion: The identification of factors that enhance or
impede weight loss for adolescents is an important step in
improving weight loss programs for this group. The current
findings suggest that family/parental variables exert a strong
influence on weight loss efforts for adolescents and should
be considered in developing future programs.
Key words: adolescents, African American, Internet,
telehealth, family-based interventions
The rate of obesity among children and adolescents has
increased in recent years; recent estimates report a 15.3%
rate of overweight status among children 6 to 11 years old
and a 15.5% rate of overweight status among adolescents
between 12 and 19 (1). Although the prevalence of obesity
is increasing in general, particular subgroups of the popu-
lation are more likely to be obese than others and, as a
result, to be at greater risk of developing serious physical/
medical problems (2,3). African Americans are more likely
than whites to be obese (4–6) and to suffer cardiac and
metabolic disorders such as diabetes (7). Similarly, African-
American children are more likely to be obese than are
white children (7–9), and African-American girls, in partic-
ular, are at increased risk for developing obesity-related
chronic health problems (7). These population estimates
underscore the need for primary and secondary prevention
programs for obesity and associated chronic health prob-
Given the substantial need for interventions catered spe-
cifically to African-American children and adolescents, a
relatively small number of treatment studies have targeted
Health Enrichment Multi-site Studies (GEMS) employed an
after-school intervention for African-American girls and
concluded that parental involvement is a crucial aspect of
facilitating behavior change in adolescents (10). This sup-
Received for review August 4, 2003.
Accepted in final form May 3, 2004.
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.
Pennington Biomedical Research Center, Louisiana State University, Baton Rouge, Loui-
Address correspondence to Marney A. White, Yale Psychiatric Research, 301 Cedar Street,
Second Floor, New Haven, CT 06519.
Copyright © 2004 NAASO
1050OBESITY RESEARCH Vol. 12 No. 7 July 2004
ports and extends research conducted with white adoles-
cents and children, which determined that the involvement
of family members is an important mediator of successful
weight loss for children (11,12). Furthermore, a mother’s
psychological adjustment has been found to be strongly
associated with the presence of childhood obesity in her
children and with the child’s success at losing weight (13).
Overall, existing research suggests that the home environ-
ment of children may be a strong determining factor in the
development of obesity during adolescence and that the
behavior of the parents, especially the mother, is critical in
determining successful weight loss efforts.
In addition to family environment and involvement, the
literature on behavioral interventions for weight loss has
identified several factors that predict program success. Suc-
cessful weight loss has been shown to be related to program
adherence, defined as frequency of contact with an inter-
ventionist (14) or in terms of adherence to behavioral pre-
scriptions such as self-monitoring (15). Further, dietary
self-efficacy is considered to be an important factor in
weight loss efforts (16) and has been related to session
attendance and outcome in weight loss programs for adults
(17). Although investigated in intervention studies for Af-
rican-American children (10,18), the potential relationship
between dietary self-efficacy and successful weight loss has
not yet been explicitly reported for this group.
Although it is well-established that family-based treat-
ments for childhood obesity are more successful than those
treating the child alone (e.g., 11,12), an often-cited problem
in providing family-based treatments is parental adherence
due to low attendance at face-to-face treatment sessions
(19). Because barriers to treatment such as transportation
are common, a novel approach to providing weight coun-
seling services is needed. The use of computer technology
to communicate information to weight loss participants may
overcome these barriers by bringing the program compo-
nents “into the home” where participants may review infor-
mation and communicate at their convenience. Computer
technology also affords a unique advantage in terms of
tracking program adherence. Because participants are re-
quired to “log on” to access program information, the time
spent reviewing materials can be systematically assessed. In
addition, self-monitoring forms submitted in computerized
form afford an advantage in that they can be stored and
reviewed electronically. Although computer technology
may potentially serve to overcome some barriers to treat-
ment, it should be noted that factors such as computer
anxiety and lack of computer literacy may serve as addi-
tional barriers. Because computer anxiety may affect com-
puter usage, it logically follows that computer anxiety may
influence the extent to which participants would adhere to
telehealth program requirements and the extent to which
they would benefit from interventions.
The primary aim of the current study was to investigate
the influence of process variables in a behavior change
program for weight loss, using an innovative form of infor-
mation transmission through the Internet. Research using
face-to-face interventions has shown the superiority of be-
havioral treatments to educational treatments for weight
loss. Further, research with adults has shown that Internet-
based behavioral treatments are superior to Internet-based
educational methods in promoting weight loss (20). Re-
search on process variables in behavior change programs
indicates that adherence to behavioral recommendations,
such as attendance at face-to-face treatment sessions, is
related to successful weight loss (21,22). One possible ex-
planation for this finding is that attendance at face-to-face
sessions increases the opportunity for learning about healthy
lifestyle choices. The current study provided a unique op-
portunity to assess the extent to which participants accessed
the program components and educational material by mea-
suring the number of times the participants accessed the
Web site. In addition, the use of weekly quizzes provided a
rough index of the extent to which participants read and
understood the educational materials.
It was hypothesized that the experimental condition (be-
havioral vs. control) would exert the strongest influence on
weight loss. However, the influence of the experimental
condition was hypothesized to be mediated by various pro-
cess variables (e.g., adherence) and psychological variables
(e.g., satisfaction with life) of the adolescents and parents.
Statistical mediation refers to a phenomenon in which some
third variable (the mediator variable) influences the rela-
tionship between the two primary variables of interest. A
mediator variable must be related to both variables; it is the
process or means through which an independent or predictor
variable influences a primary dependent variable. This pro-
cess is diagrammed in Figure 1. For example, it was hy-
pothesized that a significant mediator variable in the current
study would be change in eating habits. The experimental
group was hypothesized to influence both weight loss (path
a) and changes in eating habits (path b). In addition, changes
in eating habits were hypothesized to influence weight loss
(path c). A significant test of mediation would occur if, after
controlling for the influence of eating habits on weight loss,
the effect of the experimental group on weight loss (path a)
was eliminated or diminished.
Figure 1: Graphical depiction of statistical mediation.
Mediators of Weight Loss, White et al.
OBESITY RESEARCH Vol. 12 No. 7 July 20041051
The current study hypothesized that: 1) the adolescent
participants in the behavioral condition would lose more
weight at 6 months than participants in the control condi-
tion; 2) the effect of experimental condition on weight loss
would be mediated by program adherence, dietary self-
efficacy, parental adherence, and changes in eating and
exercise habits; and 3) depression would negatively influ-
ence adherence among adolescents. In addition, the follow-
ing hypotheses applied to the parent participants: 4) parental
weight loss would be influenced by experimental condition,
with parents in the behavioral condition losing more weight
than those in the control condition; and 5) the effect of
experimental condition on parental weight loss would be
mediated by adherence, changes in dietary and exercise
behaviors, and psychological variables.
Research Methods and Procedures
Participants were 57 overweight African-American girls
between the ages of 11 and 15 with one obese (BMI ? 30)
biological parent. Because BMI is highly variable during
adolescence, overweight or at-risk for overweight was de-
fined as a BMI at or exceeding the 85th percentile of body
mass according to age in months. In addition, each adoles-
cent participant had at least one obese biological parent,
indicating a genetic risk for developing adulthood obesity. It
should be noted that although a participant was required to
have a BMI in excess of the 85th percentile to qualify for
the study, the vast majority (?90%) of participants had a
BMI in excess of the 95th percentile for her age.
The recruitment of participants for this study involved
a media and advertising campaign that utilized talks in
the community, paid advertisements, printed stories in
newspapers and magazines, and radio and television ap-
pearances. The recruitment campaign targeted over-
weight African-American adolescent girls with at least
one biological parent who was obese. In response to this
recruitment effort, 230 interested participants contacted
the research center for further information, 210 met cri-
teria for inclusion in the study during a screening inter-
view on the telephone, and 116 expressed an interest in
the study and were scheduled for a screening interview in
the clinic. Of those 116 participants, 96 were interviewed
in the clinic, and 61 met the criteria for inclusion in the
study and were randomized to the two treatment arms of
the study. Of these 61 participants, 57 completed the
baseline assessment, and these 57 adolescent girls and
their parents formed the study sample. Random assign-
ment resulted in 28 participants in the behavioral group
and 29 participants in the control group. Six months later,
50 participants were available for assessment. Of the
seven participants who were not available for measure-
ment at 6 months, two were from the control group and
five were from the behavioral group.
Treatment Program: The Health Improvement Program
for Teens (HIP-Teens)1
The HIP-Teens program was a family-based Internet in-
tervention program for weight loss in African-American
girls in the age range 11 to 15 years old. The current study
involves outcome data from the first 6 months of the 2-year
program. The primary components of the study were com-
municated using a locked Web site created specifically for
the program. At the study outset, participants were provided
with a personal computer for the home and were given free
Internet access. On a weekly basis, participants accessed
new material, focused on weight loss, including information
regarding nutrition, physical activity, and healthy food
choices. Interested readers may reference a more detailed
description of the HIP-Teens Program, which is reported
elsewhere (23). Web site content was assessed for readabil-
ity on a sixth grade reading level. In addition, several study
consultants worked to generate Web site content that was
culturally and geographically specific, while interesting for
adolescents. As such, supplemental links and pages were
included that focused on recipes, hobbies and activities, and
“health and beauty tips,” as well as the inclusion of “chat
rooms” for study participants. Participants and their parents
also attended focus groups in which they provided feedback
as to the readability and interest level of the Web site
The project consisted of two primary conditions. The
behavioral condition incorporated behavior modification
techniques and heavy emphasis on e-mail communication
with a case manager who had at least graduate-level clinical
psychology training specializing in weight management.
Specific weekly topics in the behavioral condition included:
self-monitoring, goal setting for eating and physical activ-
ity, problem solving, behavioral contracting, and relapse
prevention. The control condition was primarily educational
in nature and provided basic information about nutrition and
physical activity. Participants in the control condition
logged in to a separate Web site and were managed by a
registered dietitian. In the control condition, topics included
lessons pertaining to serving sizes, the food guide pyramid,
hidden calories, and understanding food labels.
In the behavioral condition, participants were instructed
to complete daily food records and submit them using an
automated form housed on the Web site. Food records were
submitted to the project dietitian, who reviewed them for
accuracy and compliance with the recommended calorie
levels. In addition, automated feedback was provided such
that after submission of the forms, a computer program
embedded in the form generated an image of the Food
Guide Pyramid and indicated the extent to which the food
1Nonstandard abbreviations: HIP-Teens, Health Improvement Program for Teens; CDSS,
Child Dietary Self-Efficacy Scale; FFQ, Food Frequency Questionnaire; CDI, Childhood
Depression Inventory; SCL-90-R, Symptom Checklist-90-Revised.
Mediators of Weight Loss, White et al.
1052OBESITY RESEARCH Vol. 12 No. 7 July 2004
records complied with the recommended nutritional values.
In addition, participants were encouraged to engage in reg-
ular physical activity. In the behavioral group, participants
established physical activity goals that were incorporated
into behavioral contracts.
Assessment Measures: Adolescent Participants
Changes in Body Fat. Proportion of body fat was mea-
sured using the DXA procedure. DXA provides a valid
measure of body composition (24). Given the instability of
BMI as a measure of adiposity during growth years (25),
DXA was used as a primary anthropometric index of body
fat for adolescents. The DXA procedure measures photon
absorption rates of tissues (bone, fat, and other soft tissue)
and provides an estimate of body fat. Change in body
weight was also measured by BMI (kilograms per meter
squared). For parents, DXA, BMI, and weight were used in
Dietary Self-Efficacy. For adolescent participants, dietary
self-efficacy was measured using the Child Dietary Self-
Efficacy Scale (CDSS) (26). The CDSS defines dietary
self-efficacy as it relates to fat and sodium intake. Partici-
pants were asked to rate the extent to which they were
“sure” that they could opt for a healthy food alternative,
such as a baked potato in lieu of French fries. Items are
geared toward school-aged children, and the efficacy for
behaviors appropriate to that age is assessed. Therefore, the
items assess the efficacy for food selection, simple prepa-
ration, substitution, and requesting. The scale possesses
good internal consistency (coefficient ? ? 0.84). The CDSS
has good concurrent validity indicated by its correlation
with a measure of usual food choices. For parents, the
Eating Habits Confidence Questionnaire (27) was used to
measure participants’ confidence in their ability to change
eating habits, food selections, and cooking methods. The
Eating Habits Questionnaire is a reliable and valid measure
of dietary self-efficacy and has been shown to be related to
changes in eating behaviors.
Adherence Measures. Program adherence was assessed
using the frequency of “hits” on the Internet site. Because
visiting the Web site to access program information is
analogous to session attendance in standard weight loss
programs, this measure tabulated the number of times each
participant visited the Web site in terms of Web site “hits.”
Because participants were required to log on to the Web
site, the software automatically tracked the number of times
a participant visited the Web site. Hits were accrued each
time a participant “moved” within the Web site or submitted
a form over the Web site. The primary adherence measure
was a frequency count of the number of Web site hits during
the first 6 months of the program.
In addition, the completion of weekly quizzes was used as
a secondary measure of adherence for participants in the
behavioral group. Quizzes were provided at the end of each
weekly lesson and required the participant to complete a
series of multiple choice questions about the program ma-
terial and submit their responses using the Internet. After
submission, participants received immediate feedback re-
garding the accuracy of their responses. Therefore, comple-
tion of the quizzes served the purpose of informing the case
managers that the participant had accessed the weekly in-
formational material and provided a rough measure of
whether the participant understood the material. Within the
first 6 months of the study, there were 25 weekly quizzes.
Frequency counts of the number of quizzes with adequate
scores (i.e., 60% correct or greater) were tabulated. In
addition, average quiz scores were computed for each par-
ticipant. For participants in the behavioral group, frequency
counts of the number of food diaries and exercise self-
monitoring forms submitted were also tabulated.
Multipass 24-Hour Recall and Food Frequency Ques-
tionnaire (FFQ). The multipass 24-hour recall of dietary
intake is a procedure in which a registered dietitian inter-
views participants about the amount and types of foods
consumed over the prior 24 hours (28). Based on informa-
tion gathered through the recall procedure, the daily caloric
intake and macro-nutrient content of foods consumed is
estimated based on U.S. Department of Agriculture nutrient
values. The 24-hour recall procedure has been found to be
reliable with adults and children (29). The FFQ (30) is a
self-report measure in which participants report the fre-
quency of eating various foods and typical serving sizes.
Psychological Indices. For adolescents, depression was
measured with the Childhood Depression Inventory (CDI)
(31). The CDI is a well-validated measure with acceptable
internal consistency (0.70 to 0.86) and test-retest reliability
over a 1-month interval (0.82). For parents, depression was
measured with the depression subscale of the Symptom
Checklist-90-Revised (SCL-90-R) (32). The brief Satisfac-
tion with Life Scale (33) was used to measure global life
satisfaction in both adult and adolescent participants. The
Satisfaction with Life Scale has adequate internal consis-
tency and has been found to correlate highly (r ? 0.50) with
longer measures of subjective well-being. The Kansas Fam-
ily Life Satisfaction Scale (34) was used to assess parents’
satisfaction with their spouse, children, and quality of rela-
tionships within the family. The scale possesses good inter-
nal consistency (0.85) and has been shown to correlate
highly with other indices of quality of life.
An ? level of 0.05 was used for all analyses. To detect
group differences for adiposity and other variables, inde-
pendent Student’s t tests were conducted. Analyses were run
independently for adolescents and parents. Tests for medi-
ation were first explored using correlation analyses, fol-
lowed by a series of linear regressions after the methods of
Baron and Kenny (35). Tests of several adherence variables
Mediators of Weight Loss, White et al.
OBESITY RESEARCH Vol. 12 No. 7 July 20041053
were conducted on data available from participants in the
behavioral arm of the study. Missing data were imputed
using the mean of other participants within each experimen-
tal group. All statistical analyses were conducted using the
SPSS/PC statistical program (version 10.0.7 for Windows;
SPSS, Inc., Chicago, IL).
Of the original 57 families enrolled, 50 participants re-
mained in the study at the 6-month assessment. The baseline
sample consisted of 29 participants in the control condition
and 28 participants in the behavioral condition. Independent
Student’s t tests were conducted to confirm that the partic-
ipants in each group were equivalent in terms of age, BMI,
and adiposity measures at baseline. The descriptive statis-
tics for the adolescent participants and their participating
parents are in Table 1. A total of 7 participants dropped out
of the study before 6-month data could be collected; of
those remaining at 6 months, 27 participants were in the
control condition, and 23 participants were in the behavioral
Manipulation Check: Adherence Measures
As a function of the more interactive nature of the be-
havioral condition, it was hypothesized that participants in
the behavioral condition would visit the Web site more
frequently than participants in the control condition. This
hypothesis was supported for both adolescents [t(55) ?
5.07, p ? 0.001] and parents [t(55) ? 3.38, p ? 0.001, one
tailed]. Teens in the behavioral condition accrued an aver-
age of 624.9 hits (SD ? 444.1) over the 6-month period,
whereas those in the control condition accrued an average of
186.4 hits (SD ? 137.6). A parallel pattern was observed
among parents in the behavioral (mean ? 557.3, SD ?
500.4) and control (mean ? 226.8, SD ? 161.8) conditions.
Test of Treatment Effect
To determine whether the treatment groups differed in
weight loss at 6 months, independent Student’s t tests were
conducted on change scores for DXA and BMI for both
adolescents and parents. Results for the adolescents indi-
cated significant body fat and weight changes in the hy-
pothesized directions between the behavioral and control
groups. For parents, only BMI changes were significant at 6
months, with a nonsignificant trend in the anticipated direc-
tion for DXA. Descriptive statistics for change scores ap-
pear in Table 2. A more thorough presentation and discus-
sion of the treatment effect (behavioral vs. control) is
reported elsewhere (23).
The primary hypothesis was that for both adolescents and
parents, experimental condition would be associated with
weight loss at 6 months, such that the participants in the
behavioral group would lose more weight than those in the
control condition. It was further hypothesized that adher-
ence measures, psychological indices, and dietary change
measures would mediate this effect. The methods of Baron
and Kenny (34) were used to test possible mediating effects
on weight loss.
Adolescents. Regression analyses were performed to con-
firm the effect of experimental group on changes in adipos-
ity, after controlling for baseline adiposity (path a). Due to
the variability of body weight during growth periods, DXA
was used as the primary measure of adiposity for the ado-
lescent participants. The regression analysis indicated that
experimental condition influenced weight loss as measured
by DXA, such that adolescents in the behavioral group lost
more fat than those in the control group (F ? 3.44, p ?
0.05, b ? ?0.28, p ? 0.05).
The hypotheses that program adherence, dietary self-
efficacy, and parental adherence would be significant me-
diators of experimental group on weight loss were not
supported. Of the potential mediators, only those pertaining
to parent satisfaction emerged as significant mediator vari-
ables. To identify potential mediator variables, correlational
analyses were conducted between DXA change and adher-
ence (Web site hits), dietary changes as measured by the
Table 1. Participant characteristics: baseline
Control (n ? 29)Behavioral (n ? 28)Total (n ? 57)
DXA (% fat)
DXA (% fat)
Mediators of Weight Loss, White et al.
1054OBESITY RESEARCH Vol. 12 No. 7 July 2004
FFQ and free recall assessments, and baseline psychological
variables for both parents and adolescents. To be considered
a potential mediator variable, the variable had to differ
across experimental groups and correlate with DXA change.
The variables meeting this criterion were parent family
satisfaction (r ? ?0.23, p ? 0.05) and parent satisfaction
with life (r ? ?0.29, p ? 0.05).
A second series of regression analyses was conducted to
confirm the influence of the potential mediating variables
on changes in DXA after controlling for baseline DXA.
Parent satisfaction with life was significantly related to
change in DXA (F ? 3.45, p ? 0.05, R ? 0.34, b ?
?0.285, p ? 0.05). After controlling for baseline DXA, the
relationship between parent family satisfaction at baseline
and change in DXA approached significance (F ? 2.43, p ?
0.09, R ? 0.29, b ? ?0.22, p ? 0.09).
A final series of regression analyses was conducted to
determine whether the hypothesized mediator variables
(i.e., those that emerged as significantly related to changes
in DXA) significantly attenuated the effect of the experi-
mental group on changes in DXA. For change in DXA, the
effect of experimental group alone was significant (b ?
?0.28, p ? 0.05). After controlling for parent family sat-
isfaction, the effect from group treatment was no longer
significant (b ? ?0.24, p ? 0.079). The same effect was
observed with parent satisfaction with life, such that after
entering this variable into the equation, the effect of assign-
ment to the experimental group was diminished (b ?
?0.20, p ? 0.162).
Parents. Regression analyses were conducted to deter-
mine the effect of experimental group on adiposity change
measures for parents. After controlling for baseline adipos-
ity, only change in BMI was significantly influenced by
experimental condition (F ? 6.37, p ? 0.01, R ? 0.44, b ?
?0.28, p ? 0.05).
The hypothesis that program adherence (as measured by
Web site hits) would mediate weight loss among parents
was not supported. To begin the tests for mediation, a series
of correlational analyses was conducted to investigate sig-
nificant relationships between change in BMI and potential
mediator variables (i.e., those that differed across experi-
mental groups). Due to the theorized directional nature of
familial influence (parent to adolescent), adolescent base-
line measures were not considered as possible mediator
variables in these analyses. As such, potential mediators
were limited to parent baseline motivational/psychological
variables and parent dietary change measures. Of these,
only change in percentage fat intake (as measured by the
FFQ) emerged as a significant correlate of change in BMI.
After controlling for baseline BMI, the FFQ percentage fat
decrease was significantly related to changes in BMI (F ?
9.66, p ? 0.01, R ? 0.51, b ? 0.39, p ? 0.01). Finally, to
test the effect for mediation, FFQ percentage fat change was
entered into regression analyses to test whether the effect of
experimental group would diminish. The results indicated
that before controlling for percentage fat dietary change, the
effect of experimental group on BMI was significant (b ?
?0.28, p ? 0.05). This effect was not observed after con-
trolling for FFQ percentage fat reduction (b ? ?0.21, p ?
0.087). As an additional test of mediation, data were again
analyzed using a repeated measures analysis of the BMI
data from the behavioral group only. Within the behavioral
group, the percentage fat reduction was correlated with BMI
change [r(26) ? 0.46, p ? 0.01]. The repeated measures
ANOVA indicated that the effect for time was significant
[F(1,25) ? 5.61, p ? 0.05]. However, after controlling for
percentage fat intake, this effect was not significant
[F(1,24) ? 0.11, p ? 0.74].
Adherence within the Behavioral Group. It was hypoth-
esized that the adherence measures would be correlated,
Table 2. Group differences for changes in adiposity
BMI (kg/m2) change
Weight (kg) change
BMI (kg/m2) change
Weight (kg) change
Negative numbers reflect difference scores from baseline to 6 months and are indicative of weight loss.
* Positive t values indicate greater improvements in the behavioral group.
Mediators of Weight Loss, White et al.
OBESITY RESEARCH Vol. 12 No. 7 July 2004 1055
such that within the behavioral group, the number of Web
site hits would be related to the number of quiz submissions
and food and physical activity records submitted by both
adolescents and parents. Descriptive statistics of Quiz Sub-
missions appear in Table 3. There was a significant decrease
in quiz completion from the first 3 months of the study to
the second 3 months for both adolescents [t(27) ? 5.37, p ?
0.001] and parents [t(27) ? 4.26, p ? 0.001].
Correlation analyses were conducted to determine the
interrelationships among overall Web site hits; e-mails;
weight graph, food record, and activity graph submissions;
quiz completions; average quiz scores; and computer anxi-
ety. The correlation matrix for both adolescents and parents
appears in Table 4. For adolescents, there was a significant
relationship between the adiposity measures and weight
graph submissions. Weight graph submissions also related
to the number of quizzes completed. Of the adherence
measures, the average quiz score was most consistently
correlated with other adherence measures and may indicate
that those participants who correctly understood the mate-
rial may have been more adherent with the other study tasks.
For parents, several more significant relationships
emerged, with quiz submissions most consistently related to
the other compliance variables. However, none of the ad-
herence measures significantly correlated with weight loss.
Parent data are represented in bold type.
It was hypothesized that adherence would be negatively
correlated with depression as measured at baseline. This
hypothesis was supported for adolescents; CDI scores were
modestly related to Web site hits [r(55) ? ?0.26, p ? 0.05,
one tailed] such that more depressed adolescents were less
likely to log on to the study Web site. For parents, the
hypothesis was partially supported; Web site hits were not
Table 3. Descriptive statistics of quiz submissions for
adolescents and parents
Minimum Maximum Mean SD
weeks 1 to 13
weeks 14 to 26
weeks 1 to 13
weeks 14 to 26
0 25 9.75 8.17
0 24 7.96 7.10
0 13 5.93 4.15
0 11 2.04 3.91
0 100 64.46 0.33
0 23 8.89 8.11
0 228.00 7.47
0 125.57 4.34
0 122.43 4.09
0 100 66.48 0.37
Table 4. Correlation matrix of compliance measures with changes in adiposity
1) 2) 3)4)5)6)7) 8) 9)10)
1) BMI change
2) DXA change
4) Weight graph
5) Activity graph
6) Food records
9) Average quiz score
10) Computer anxiety
Parent data appear in bold type. Degrees of freedom range from 16 to 28.
* Correlation is significant at the 0.05 level (one-tailed).
† Correlation is significant at the 0.01 level (one-tailed).
Mediators of Weight Loss, White et al.
1056OBESITY RESEARCH Vol. 12 No. 7 July 2004
related to SCL-90-R Depression subscale scores [r(55) ?
?0.03, p ? 0.40, one tailed]. However, when including the
potentially more valid measure of adherence (average quiz
score) among parents in the behavioral group, average quiz
score was negatively correlated with SCL-90-R Depression
[r(26) ? ?0.34, p ? 0.05, one tailed].
The primary finding of this study was consistent with
previous weight loss studies showing the superiority of
behavioral interventions to purely educational interventions
(36). This study was the first to use an Internet-based
approach to promote weight loss among adolescents and
their parents. Because the primary finding parallels the
findings of in vivo weight loss studies, results indicate that
the Internet may be an effective means to transmit informa-
tion and facilitate behavior change among adolescents and
parents. The current study extends the accumulating litera-
ture on the effectiveness of computers as either a means or
supplement to behavioral health interventions.
Overall, the results confirmed that the Internet may be an
effective means to transmit nutritional information to pro-
mote healthier dietary outcomes. Both treatment groups
showed improvements in self-reported eating behaviors and
attitudes toward healthy eating from baseline to 6 months.
Further, the behavioral intervention was more successful in
changing these behaviors among both adolescents and par-
ents, as evidenced by significantly greater reductions in fat
intake over the course of the study. However, given the
rather modest changes in adiposity compared with face-to-
face interventions, the Internet may be best utilized in the
context of preventing weight gain. For example, the Internet
may be an effective supplement to school- or community-
based programs for the prevention of weight gain and/or the
promotion of other health behaviors.
This study also investigated potential mediating variables
in the relationship between experimental group and weight
loss. For adolescent weight loss, variables relating to family
climate (i.e., parents’ satisfaction with life and family sat-
isfaction) emerged as significant mediators. This finding
may be interpreted to mean that the family context had a
powerful influence on treatment effectiveness. Indeed, pre-
vious research has indicated that parental family life satis-
faction is related to weight loss for children, with higher
family satisfaction ratings associated with successful weight
loss (37). These authors concluded that family climate vari-
ables have the potential to impede or accent child weight
loss efforts. In addition, mothers’ psychological well-being
has been shown to be related to health behaviors among
children (13,38). From a global viewpoint, family climate
(like most environments) would influence the physical and
psychological health of children. Previous research has
shown that parental attitude and expectations of success
significantly impact weight loss efforts for children (39–41).
The influence of family setting is also observed in the
interrelationships of adherence measures within the behav-
ioral group. Social learning theory predicts that parents
would model healthy eating and exercise habits and pro-
gram adherence, leading to similar behavior patterns among
adolescents. Indeed, results showed that adolescent adher-
ence measures were associated with parent adherence mea-
sures, implying that parents may have been powerful agents
of change within the family. Furthermore, adolescent
changes in dietary habits as measured by calorie and fat
intake were strongly associated with parent changes in
dietary intake. These findings support previous research
showing that parental modeling of dietary practices is a
powerful determinant of children’s eating behaviors (42). In
addition, a parent’s willingness to change eating habits has
been shown to influence effective weight loss for children
(39). On a more global level, the relationships among ad-
herence measures, dietary intake, and weight loss may in-
dicate that the parents’ behaviors and investment in the
program significantly influenced outcome for adolescents.
It was further hypothesized that for adolescents, the effect
of experimental condition on weight loss would be mediated
by program adherence (as measured by Web site hits). This
hypothesis was partially supported. Within the behavioral
group, several adherence measures indicated that the com-
pletion of behavioral prescriptions was significantly related
to weight loss. For example, submission of weight graphs
and scores on weekly quizzes were associated with weight
loss among adolescents.
Dietary self-efficacy was also hypothesized to mediate
the effect of experimental group on weight loss at 6 months.
Previous research has shown that among adults, dietary
self-efficacy is related to successful weight loss (17). It was
hypothesized that among adolescents, dietary self-efficacy
would influence treatment outcome, such that those partic-
ipants with greater confidence in their ability to change
dietary habits would be more successful in the program.
This hypothesis was not confirmed and parallels the find-
ings of Martin et al. (43), who found that self-efficacy as
measured by a brief questionnaire was unrelated to weight
loss. These researchers suggested that when initiating a
weight loss program, participants may overestimate their
ability to practice healthful behaviors.
The hypothesis that parental adherence would mediate
the effect of experimental condition on weight loss for the
adolescent participants was not supported. Parental adher-
ence measures were not significantly correlated with weight
loss for adolescents. However, parent adherence measures
were correlated with adolescent adherence measures, indi-
cating that parent’s behavior likely influenced that of the
adolescent. Indeed, social learning theory would predict that
parental modeling of specific behaviors would increase the
likelihood of the adolescent performing the desired behav-
Mediators of Weight Loss, White et al.
OBESITY RESEARCH Vol. 12 No. 7 July 20041057
Depression was also hypothesized to predict poor adher-
ence, such that participants with higher levels of depressive
symptoms would be less likely to adhere to the program
requirements. This hypothesis was supported for both ado-
lescents and parents, indicating that psychological difficul-
ties may interfere with treatment for weight loss.
The current study has some limitations. The most valu-
able measures of adherence were collected only from par-
ticipants in the behavioral group. Although a tally of Web
site hits was gathered from the control participants, the
frequency of hits index was shown to be a misleading
variable when compared with other adherence measures
(e.g., average quiz scores). Because it was not possible to
track participants’ movement throughout the Web site, it is
not possible to distinguish whether participants were read-
ing educational material vs. spending time in the commu-
The primary contribution of this study is the identifi-
cation of variables that may either enhance or impede
weight loss programs presented over the Internet for
overweight adolescents. Previous research on weight loss
for children and adolescents has highlighted the impor-
tance of including family members to achieve optimal
treatment gains. Because parents can serve as both au-
thority figures and role models in changing health behav-
iors, inclusion of a parent in treatment will optimize
effects. In addition, the extent of inclusion will likely
influence treatment success. For example, including a
parent as a participant will likely yield better results than
merely eliciting emotional support. The current study
extends the literature on families and weight loss in that
it identified nonspecific family variables as important
mediators of treatment effectiveness. That parents’ levels
of family and life satisfaction emerged as the primary
mediators of experimental group on weight loss speaks to
the saliency of family climate in treatment interventions
for adolescents. The most effective treatments for chil-
dren and adults are those that view the individual as
operating within a system and address problematic as-
pects of the home. Future weight loss interventions
should adopt this holistic view and incorporate compo-
nents to address family climate.
This work was supported by NIH Grant 5 RO1
1. Ogden CL, Flegal KM, Carroll MD, Johnson CL. Preva-
lence and trends in overweight among US children and ado-
lescents, 1999-2000. JAMA. 2002;288:1728–32.
2. Bray GA. Contemporary Diagnosis and Management of Obe-
sity. Handbooks in Health Care Co.: Newton, PA.; 1998.
3. National Task Force on the Prevention, Treatment of Obe-
sity. Towards prevention of obesity. Obes Res. 1994;2:571–
4. Abrams KK, Allen LR, Gray JJ. (1993). Disordered eating
attitudes and behavior, psychological adjustment, and ethnic
identity: A comparison of black and white female college
students. Int J Eat Disord. 1993;14:49–57.
5. Klesges RC, DeBon M, Meyers AW. Obesity in African-
American women: epidemiology, determinants, and treatment
issues. In: Thompson JK, ed. Body Image, Eating Disorders,
and Obesity: An Integrative Guide for Assessment and Treat-
ment. Washington, DC: American Psychological Association;
1996, pp. 461–77.
6. Rand CS, Kuldau JM. The epidemiology of obesity and
self-defined weight problem in the general population: gender,
race, age, and social class. Int J Eat Disord. 1990;9:329–43.
7. Crawford PB, Story M, Wang MC, Ritchie LD, Sabry ZI.
Ethnic issues in the epidemiology of childhood obesity. Pe-
diatr Clin North Am 2001;48:855–78.
8. Dwyer JT, Stone EJ, Yang M, et al. Predictors of overweight
and overfatness in a multiethnic pediatric population. Am J
Clin Nutr. 1998;67:602–10.
9. Morrison JA, Barton B, Biro FM, Sprecher DL, Falkner F,
Obarzanek E. Sexual maturation and obesity in 9- and 10-
year-old black and white girls: the National Heart, Lung, and
Blood Institute Growth and Health Study. J Pediatr. 1995;
10. Story MS, Sherwood NE, Himes JH, et al. An after-school
obesity prevention program for African-American girls: the
Minnesota GEMS pilot study. Ethn Dis. 2003;13:S1–54-64.
11. Epstein LH, Wing RR, Koeske R, Valoski A. Long-term
effects of family-based treatment of childhood obesity. J Con-
sult Clin Psychol. 1987;55:91–5.
12. Epstein LH, Valoski A, Wing RR, McCurley J. Ten-year
follow-up of behavioral, family-based treatment for obese
children. JAMA. 1990;264:2519–23.
13. Favaro A, Santonastaso P. Effects of parents’ psychological
characteristics and eating behaviour on childhood obesity and
dietary compliance. J Pyschosom Res. 1995;39:145–51.
14. Davis K, Christoffel KK. Obesity in preschool and school-
age children: treatment early and often may be best. Arch Ped
Adolesc Med. 1994;148:1257–61.
15. Israel AC, Silverman WK, Solotar LC. The relationship
between adherence and weight loss in a behavioral treatment
program for overweight children. Behav Ther. 1988;19:25–33.
16. Perry CL, Stone EJ, Parcel GS, et al. School-based cardio-
vascular health promotion: the Child and Adolescent Trial for
Cardiovascular Health (CATCH). J Sch Health. 1990;60:406–
17. Clark MM, Abrams DB, Niaura RS, Eaton CA, Rossi JS.
Self-efficacy in weight management. J Consult Clin Psychol.
18. Resnicow K, Yaroch AL, Davis A, et al. GO GIRLS!: results
from a nutrition and physical activity program for low-in-
come, overweight African American adolescent females.
Health Educ Behav. 2000;27:616–31.
19. Dunn PC, Lackey C, Kolasa K, Mustian D. At-home nutri-
tion education for parents and 5- to 8-year-old children: the
HomePlate pilot study. J Am Diet Assoc. 1998;98:807–9.
Mediators of Weight Loss, White et al.
1058OBESITY RESEARCH Vol. 12 No. 7 July 2004
20. Tate DF, Wing RR, Winett RA. Using Internet technology to Download full-text
deliver a behavioral weight loss program. JAMA. 2001;285:
21. Stolley MR, Fitzgibbon ML. Effects of an obesity prevention
program on the eating behavior of African American mothers
and daughters. Health Educ Behav. 1997;24:152–64.
22. Wadden TA, Stunkard AJ, Rich L, Rubin CJ, Sweidel G,
McKinney S. Obesity in black adolescent girls: a controlled
clinical trial of treatment by diet, behavior modification, and
parental support. Pediatrics. 1990;85:345–52.
23. Williamson D, Martin P, White M, Newton R, Walden H.
HIPTeens: Randomized controlled trial of the efficacy of an
Internet-based weight management program for overweight
African-American girls. Obes Res. 2003;11:A29.
24. Harsha DW, Bray GA. Body composition and childhood
obesity. Endocrinol Metab Clin North Am. 1996;25:871–85.
25. Himes JH, Dietz WH. Guidelines for overweight in adoles-
cent preventive services: recommendations from an expert
committee. Am J Clin Nutr. 1994;59:307–16.
26. Parcel GS, Edmundson E, Perry CL, et al. Measurement of
self-efficacy for diet-related behaviors among elementary
school children. J Sch Health. 1995;65:23–27.
27. Sallis JF, Pinski RB, Grossman RM, Patterson TL, Nader
PR. The development of self-efficacy scales for health-related
diet and exercise behaviors. Health Educ Res. 1988;3:283–92.
28. Johnson RK, Driscoll P, Goran MI. Comparison of multi-
ple-pass 24-hour recall estimates of energy intake with total
energy expenditure determined by the doubly labeled water
method in young children. J Am Diet Assoc. 1996;96:1140–4.
29. Freund A, Johnson SB, Silverstein J, Thomas J. Assessing
daily management of childhood diabetes using 24-hour recall
interviews: reliability and stability. Health Psychol. 1991;10:
30. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon
J, Gardner L. A data-based approach to diet questionnaire
design and testing. Am J Epidemiol. 1986;124:453–69.
31. Kovacs M. Rating scales to assess depression in school-aged
children. Acta Paedopsychiatry. 1980;46:305–15.
32. Derogatis LR. SCL-90-R: Symptom Checklist-90-R: Admin-
istration, scoring, and procedures manual. Minneapolis, MN:
National Computer Systems; 1994.
33. Diener E, Emmons RA, Larsen RJ, Griffin S. The Satis-
faction with Life Scale. J Pers Assess. 1985;49:71–5.
34. Schumm WR, McCollum EE, Bugaighis MA, Jurich AP,
Bollman SR. Characteristics of the Kansas Family Life Sat-
isfaction Scale in a regional sample. Psychol Rep. 1986;58:
35. Baron RM, Kenny DA. The moderator-mediator variable
distinction in social psychological research: conceptual, stra-
tegic, and statistical considerations. J Person Soc Psychol.
36. Jelalian E, Saelens BE. Empirically supported treatments in
pediatric psychology: pediatric obesity. J Pediatr Psychol.
37. Barbarin OA, Tirado M. Enmeshment, family processes,
and successful treatment of obesity. Fam Relat: J Appl Fam
Child Stud. 1985;34:115–21.
38. Florian V, Elad D. The impact of mothers’ sense of empow-
erment on the metabolic control of their children with juvenile
diabetes. J Pediatr Psychol. 1998;23:239–47.
39. Uzark KC, Becker MH, Dielman TE, Rocchini AP. Pycho-
social predictors of compliance with a weight control inter-
vention for obese children and adolescents. J Compliance
Health Care. 1987;2:167–78.
40. Uzark KC, Becker MH, Dielman TE, Rocchini AP, Katch
V. Perceptions held by obese children and their parents: im-
plications for weight control intervention. Health Educ Quart.
41. Wilson DK, Ampey-Thornhill G. The role of gender and
family support on dietary compliance in an African American
adolescent hypertension prevention study. Ann Behav Med.
42. Golan M, Weizman A. Familial approach to the treatment of
childhood obesity: conceptual model. J Nutr Educ. 2001;33:
43. Martin CK, O’Neil PM, Binks M. An attempt to identify
predictors of treatment outcome in two comprehensive weight
loss programs. Eat Behav. 2002;3:239–48.
Mediators of Weight Loss, White et al.
OBESITY RESEARCH Vol. 12 No. 7 July 2004 1059