Patient Characteristics as Predictors of Weight Loss after an Obesity Treatment for Children


  • 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.

Department of Developmental, Personality, and Social Psychology, H. Dunantlaan 2, 9000 Ghent, Belgium. E-mail:


Objective: To examine pretreatment patient characteristics as predictors of treatment outcome 2 years after completion of an inpatient treatment for children with obesity.

Research Methods and Procedures: Children (N = 122) ranging in age from 7 to 17 years [mean, 12.7 ± 2.3 (SD) years] with a mean adjusted BMI of 179.5 ± 28.6% participated in an inpatient obesity treatment program. Children (90.2% response rate) participated in the 2-year follow-up. Eight predictors, administered at baseline, were entered in separate regression analyses, with weight loss and changes on three psychological health measures as the outcome measures.

Results: Analyses revealed that baseline degree of overweight, age, and initial weight loss were significant positive predictors of weight loss 2 years after treatment, whereas eating disorder characteristics were a negative predictor. Sex, socioeconomic status, global self-esteem, and symptoms of psychopathology did not predict weight loss. With regard to the psychological outcome measures, baseline symptomatology emerged as the most important predictor of treatment changes.

Discussion: Long-lasting weight loss is associated with severity of pretreatment characteristics. Identification of the clinical markers for long-term response to treatment is useful to set realistic weight loss goals for clients and to tailor treatment programs to patient characteristics.


Research on the effects of treatment of childhood obesity has shown substantial short-term benefits as well as some evidence for long-term maintenance of the treatment effects (1). The most successful programs adopt a multidimensional approach, characterized by the inclusion of diet and exercise and the application of behavior modification principles (2). Nevertheless, not all children with obesity benefit from this approach. Most pediatric obesity interventions are marked by small changes in relative weight and substantial relapse (1,2,3). In treatments showing superior outcome, children have succeeded in decreasing their overweight by 5% to 20%. Until now, only a few studies have identified those who stand to benefit most from a given treatment for childhood obesity (4,5). Better knowledge of treatment outcome predictors could facilitate effective tailoring of treatment to patient characteristics.

The primary aim of this study was to identify patient characteristics associated with successful long-term outcome after an inpatient treatment program for children with obesity. Despite encouraging results in the short-term, treatment for obesity in adults is typically followed by weight regain (6). In children, the long-term maintenance of weight is still an issue for research (1,2,3).

Only a few predictors of treatment outcomes for children with obesity have been identified. Initial relative weight predicts weight loss during the initial phase of treatment; the heaviest children are most successful in losing weight (7). Furthermore, initial treatment success in children has been shown to predict relative weight for children 1 year later (7).

Across studies, a number of factors have been hypothesized as being related to the development or maintenance of overweight problems: family history of overweight, age, sex, severity of the overweight, socioeconomic status (SES),1 binge status, psychopathology (defined as a global measure of psychological problems), and self-worth (defined as the confidence subjects felt in their own functioning). The relationship of these factors to treatment outcome is currently unknown (3). Epstein and Wing (8) suggested that children with slim parents have better chances for long-term weight loss than children with heavy parents, because of both genetic and environmental causes of obesity. The relevance of age and sex as predictors of weight loss is unclear (8). Interventions with younger children may be more effective than those with older children for at least two reasons. First, younger children have a shorter history of poor eating habits. Second, treatment programs for older children are hampered by a reduced amount of parental control on eating behavior. Different studies found that children of lower SES were more overweight (8,9,10), although this negative relationship is not very consistent (11). Whether or not SES is a predictor of treatment outcome and the specific mechanisms through which SES influences childhood obesity remain unspecified. Epstein and Wing (8) assumed that family size and income are usually inversely related and that they, in turn, determine the amount of time parents spend on managing children's (eating) behavior. Therefore, it is predicted that low SES will lead to less social support and hence serve as a negative predictor of long-term treatment outcome.

Research has suggested that eating disorder symptoms may play a crucial role in the etiology and maintenance of adult obesity (12) and that obesity in childhood may be a risk factor for the development of eating disorders in adulthood (13). Moreover, recent studies provide evidence for the prevalence of eating disorder symptoms in children who are overweight (14,15), but their predictive value for treatment outcome has not been studied yet.

Although it is well established that psychopathology predicts poor obesity treatment outcomes for adults, the predictive value of psychopathology for the effects of childhood obesity treatment has not been assessed (16). The results of studies on the impact of self-worth on the outcome of adult obesity treatment are mixed, probably because they are biased due to their retrospective nature (17). Higher self-worth is correlated with better social and interpersonal functioning and with higher levels of coping and achievement (18), whereas a high level of psychopathology usually correlates with a range of negative outcomes.

Although outpatient childhood obesity treatment programs are sometimes effective, for those who are not successful in these programs, few alternative treatment options are available. This has led to the development of inpatient treatment programs for obese children who do not seem to benefit from outpatient treatment. The short-term results of inpatient treatment have indicated that treated children lose significantly more weight than equally motivated children on the waiting list (19). However, and not totally unexpected, when the children return to their everyday environment and leave the strictly regulated inpatient care setting, they are faced with much more freedom, resulting in a weight regain. It is unknown what type of child is able to maintain a significant weight reduction in the long-term. Therefore, this study focused on patient characteristics as prognostic factors for long-term treatment outcomes of children with obesity treated in an obesity clinic.

Over the past few years, it has become clear that moderate weight loss is associated with significant health benefits. This has prompted a move toward redefining what constitutes success in the treatment of obesity. Different organizations (20,21,22) suggest that losing 5% to 15% weight is reasonable for adults. There is no clear consensus on how successful weight loss should be defined for children, although a relative weight loss of 5% to 15% seems realistic (1). However, weight loss is only one of the many parameters of treatment success. Recently, it has been argued that weight control programs for children need to be evaluated both in terms of their desired effect on body weight and their impact on psychological outcome variables (1,3,23). These secondary outcome measures are just as important as the primary outcome of weight control, because obese children seeking treatment often experience concomitant psychological problems that may be alleviated by effective obesity treatment, resulting in improved psychological or mental health status (24,25). Improving children's self-esteem may prevent development of a wide range of emotional and social problems and contribute to the development of a healthy personality (18). However, the benefits of enhancing self-esteem in the context of child obesity treatment programs have yet to be established by research (3). Thus far, to our knowledge, only one study has evaluated the effects of an outpatient family-based treatment on secondary outcome measures such as disordered eating and symptoms of psychopathology (26). The results documented a decrease in child behavior problems but no significant reduction in symptoms of disordered eating. This finding may be attributed to the use of an inappropriate measure of disordered eating. Furthermore, the children in this study were quite young and below the age at which disordered eating symptoms tend to increase.

To conclude, the main objective of this study was to examine which patient characteristics predict weight loss and mental health 2 years after the completion of an inpatient treatment program for children with obesity. Apart from the primary outcome measure, defined as total weight loss 2 years after treatment, the study addresses long-term impact of the intervention on self-worth, symptoms of psychopathology, and disordered eating behavior.

Research Methods and Procedures


Participants were 122 patients (7 to 17 years of age) admitted to a 10-month inpatient treatment program for obesity. Medical doctors referred all children because of outpatient treatment failure. Only children of normal intelligence (IQ >70 on the Wechsler Intelligence Scale for Children-Revised), with primary obesity, and a BMI >95th percentile according to their age and sex were admitted. The mean age of the children at admission was 12.7 ± 2.3 (SD) years. Mean weight was 84.7 ± 19.7 kg; mean height was 160.7 ± 10.9 cm. Mean adjusted BMI was 179.5 ± 28.6%. The majority of the sample was white (93%), and 7% were of African-American (n = 4) or Asian origin (n = 4). According to the Hollingshead index (27), 40% of the parents were lower social class; 44% were middle class; and 16% were classified as upper class. SES correlated negatively with overweight (r = −0.19, p < 0.01). The sample consisted of 41 boys (34%) and 81 girls (66%). There were no sex or age differences in adjusted BMI.

At the 2-year follow-up, 110 participants recorded weight and height (response rate: 90.2%), enabling the primary treatment outcome measure to be determined. Furthermore, 86 participants completed all psychological measures (response rate: 70.5%) at both assessment points. Preliminary analyses showed no significant baseline differences between the subjects who provided psychological measurements at the 2-year follow-up and those who did not.

Most of the children (78.0%) sought additional help to maintain their weight after the treatment, as recommended by their medical doctor at discharge. In several cases, combined care programs were reported. A majority of the children (49.1%) planned exercises through clubs or with the help of a physiotherapist. Others were monitored by a dietitian (19.5%), a psychologist (17.1%), or a medical doctor (12.2%) or enrolled in commercial weight reduction programs (13.4%).


To compare the BMI of the children at different ages, the adjusted BMI was used in this study. The formula is as follows: (Actual BMI/Percentile 50 of BMI for age and sex) × 100. The 50th percentiles of BMI are based on normative data (28).

The Self-Perception Profile for Children (SPPC) (29,30) assesses the child's self-perception in five different areas. In this study, only the global self-worth score was used. Higher scores reflect a better confidence in their own overall functioning. Reliability and validity coefficients vary from 0.68 to 0.79 (30).

The Child Behavior Checklist (CBCL) (31,32) measures more than 100 different emotional or behavior problem items, based on the observation of children through epidemiological studies. The checklist provides normalized T-scores on several problem areas and a total problem score, which was used in this study. The CBCL is generally accepted as an objective tool for screening symptoms of psychopathology. The scale has shown good reliability and validity. Test—retest reliability correlations are between 0.82 and 0.95. The CBCL was sent to the parents at home. In addition, all children of ≥11 years of age completed the Youth Self-Report (YSR), which consists of the same items as the CBCL (33,34,35). A composite score was calculated as the mean of the total problem scores of the CBCL and the YSR.

The Eating Disorder Examination (EDE) (36) was administered to assess eating disorder symptoms through an interview with a trained psychologist. The EDE contains four subscales and additional questions (e.g., total number of binges per month) as well as a total score, which was used in this study. The EDE has shown good internal consistency, validity, and sensitivity to change (37). In this study, its reliability ranged from 0.51 to 0.78 (38).


A total of 150 patients who entered the clinic between September 1996 and September 1999 were selected at random, and informed consent was obtained. During the study, 28 children left the study; 21 of them reached an acceptable weight in a shorter time (mean = 5 months). The weight criterion for discharge was defined as an adjusted BMI <130%. The mean age of the early-leavers (N = 28) was 12.4 years. The mean adjusted BMI of the early-leavers was 170% at baseline and 137% at the end of their stay. Although this successful subgroup is of interest for further analysis, this study focused primarily on those who completed the treatment.

The obesity treatment center is a recognized institute for treatment of children with chronic diseases. All children were treated as inpatients. Based on pilot studies, a 10-month non-diet healthy lifestyle program was implemented (39,40,41). Children learned to make healthy food choices at fixed times during the day in combination with daily physical activities, necessary for a healthy body weight. Cognitive-behavioral techniques are considered as important tools in changing the behavior and were integrated as part of the program. The program is described in detail in Braet et al. (25).


Weight loss during the treatment (initial weight loss; baseline adjusted BMI minus adjusted BMI at the end of treatment) was expressed as a change score. Furthermore, four other change scores were calculated using the 2-year follow-up data (pretest score minus follow-up score), reflecting change in psychopathology, eating disorder symptoms, global self-worth, and total weight loss (baseline adjusted BMI minus adjusted BMI at follow-up).

Regression analyses were conducted to examine the relationship between predictors and outcome measures. Total weight loss was the dependent variable and the eight independent variables were as follows: baseline adjusted BMI, SES, sex, age, initial weight loss, and three baseline psychological measures (global self-worth, psychopathology, eating disorder symptoms). Baseline adjusted BMI was entered first, and the variables SES, sex, and age were entered in a second block. Initial weight loss was entered third, and, finally, the three psychological characteristics were entered in the fourth block. Next, predictors of change in psychopathology, self-worth, and eating disorder symptoms were determined by three separate regression analyses, with the same predictors as described above.


At the 2-year follow-up, the children had grown 8.0 ± 10 cm and their mean weight was 84.5 ± 17.3 kg. The mean adjusted BMI was 150.6 ± 30.2%. Total weight loss was 27.1 ± 26.3% (Table 1 and Figure 1). Moreover, 77.3% of the sample had succeeded in reducing their adjusted BMI by 10%. Effect size (Cohen's d for repeated measures) (42) for total weight loss was d = 0.98. A Student's t test revealed no sex differences in weight loss. At follow-up, the adjusted BMI was 151.6% for girls and 148.7% for boys. Total weight loss was correlated with baseline overweight (r = 0.40, p < 0.01) and with initial weight loss (r = 0.43, p < 0.01). Only initial weight loss was correlated with age (r = −0.32, p < 0.01).

Table 1. . Mean scores on weight parameters and psychological measures for obese children
 PretestPost-testTwo-year follow-up
  1. SD, standard deviation; SPPC, Self Perception Profile for Children; EDE, Eating Disorder Examination; CBCL, Child Behavior Checklist; YSR, Youth Self Report Form.

Adjusted BMI (%)179.5228.59130.3118.33150.5830.15
Weight (kg)84.6819.7463.9615.0084.5417.30
Height (cm)160.6710.93162.8410.82168.6209.59
  SPPC-global self-worth2.430.722.680.682.780.70
Eating disorder symptoms      
  EDE-total score1.901.201.330.771.190.82
  Composite score56.7209.0155.1609.4354.4808.98
Figure 1.

Weight evolution of obese children before and after an inpatient treatment program.

Psychopathology composite score and eating disorder symptoms declined significantly (paired Student's t tests, p < 0.05), whereas global self-worth increased significantly (paired Student's t test, p < 0.001). Mean scores are presented in Table 1. Effect sizes for the change in eating disorder symptoms, psychopathology, and self-worth were 0.69, 0.25, and 0.49, respectively.

A first regression analysis was conducted with total weight loss as the dependent variable. As shown in Table 2, baseline adjusted BMI accounted for a significant portion of the variance [R2 = 17%; F (1, 106) = 21.30, p < 0.001]. Total explained variance was R2 = 34%. Additional significant predictors were age, initial weight loss, and eating disorder symptoms.

Table 2. . Multiple regression predictors of weight loss at 2-year follow-up
  1. SE, standard error; EDE, Eating Disorder Examination.

Block 1     
 Baseline adjusted BMI0.380.080.414.62<0.001
  Multiple R0.41    
Block 2     
  Multiple R0.47    
Block 3     
 Initial weight loss0.501.540.433.250.002
  Multiple R0.54    
Block 4     
 Eating disorder symptoms (EDE)−5.001.94−0.22−2.580.01
  Multiple R0.58    

Next, a second regression analysis was conducted, this time with change in psychopathology as the dependent variable (Table 3). Total explained variance was R2 = 37%. Only baseline psychopathology was found as a significant predictor. A separate regression analysis with change in self-worth as dependent variable revealed that change was predicted by baseline self-worth and by eating disorder symptoms. Finally, change in eating disorder symptoms was predicted by baseline eating disorder symptoms and sex. A closer inspection of this finding revealed that girls had more eating disorder symptoms both at baseline and at follow-up (p < 0.001); both boys and girls decreased in symptoms (p < 0.001).

Table 3. . Multiple regression predictors of changes in psychological variables at 2-year follow-up
  1. SE, standard error; EDE, Eating Disorder Examination; SPPC, Self Perception Profile for Children.

Predictors of change in psychopathology      
 Baseline behavior problem score (composite score)0.370.720.110.616.85<0.001
Predictors of change in eating disorder symptoms      
 Baseline eating disorder symptoms (EDE)0.620.880.070.8212.64<0.001
Predictors of change in self-worth      
 Baseline global self-worth (SPPC)0.390.810.110.667.72<0.001
 Baseline eating disorder symptoms (EDE)0.410.<0.05


Despite encouraging results in the short-term, treatment for obesity is typically followed by weight regain, and long-term weight control may, therefore, be the main challenge in the treatment for obesity (6). Especially for those children for whom outpatient treatment has already failed, new treatment options are now being researched. This study revealed that 77.3% of these severely overweight children were successful in reducing their adjusted BMI by at least 10% 2 years after completing an intensive inpatient treatment. Furthermore, the study showed significant improvements at follow-up in global self-worth and reduction of psychopathology and symptoms of eating disorders.

The primary aim of the study was to identify relevant predictors, based on pretreatment variables. Identification of clinical markers, associated with response to treatment, can help us to effectively tailor treatments to clients. The idea of tailoring is appealing because it acknowledges the heterogeneous nature of the population of overweight individuals and the likelihood that people with different characteristics will respond differently to various treatments.

A few important predictors of treatment success emerged in this study. First, as predicted, the baseline adjusted BMI explained a substantial part of the variance in weight loss. The higher the adjusted BMI, the greater the total weight loss. The knowledge that even the most severely obese children can move toward a reasonable weight is a hopeful prospect. For children with moderate obesity, it can be a frustrating experience that some children lose more weight than others, even when they adhered to the same program and displayed the same efforts. Therefore, personalized and realistic weight loss goals are to be formulated, taking into account initial weight.

Furthermore, age emerges as another important predictor. In this study, the older children were more successful, even after controlling for initial weight. Total weight loss of children older than 12 was 30%, whereas it was 23.4% for the younger children. It is now hypothesized that weight control requires self-control skills that older children are more able to master. For younger children, future programs will need to focus more on parental involvement to secure significant weight loss. After discharge, outpatient help focusing on supporting the child in maintaining weight control seems a worthwhile additional measure.

Initial weight loss during the treatment can be identified as another important predictor of treatment success. This predictor was already identified in previous studies in adults (43) as well as in children (7). Although as yet this has not been the topic of purposive research in children, it is possible that children who lose a significant amount of weight are more motivated to continue their weight control program and experience greater self-efficacy feelings contributing to sustained efforts to control weight. It is also worthwhile to consider that mechanisms responsible for reversing obesity in these obese children do not depend on subject characteristics such as SES and sex.

This study revealed that eating disorder symptoms adversely affect treatment outcome. Weight loss and change in global self-worth were affected by this parameter. This highlights the importance of considering eating disorder symptoms when devising treatment programs for children and adolescents suffering from obesity. Moreover, baseline symptoms were important predictors of long-term treatment changes in psychological measures of mental health. In this study, symptoms of psychopathology were the only important predictor for change in psychopathology. However, this may be because of omission of important predictors, lack of sensitivity of the included predictors, interactions among the included predictors that are not fully understood, or changing relationships among prognostic factors when children grow and mature. Future research needs to include new prognostic factors, based on theoretical models or ad hoc findings. In a study of Myers et al. (44), reduced maternal psychopathology accounted for a significant part of the decrease in the psychopathology of the child. Again, this strengthens the case for taking into account family characteristics as predictors of improvement in children. Next, child studies may profit from research in adults, where at least four psychological factors have been identified as potential predictors: self-efficacy, dietary restraint, coping skills, and weight goals (17). However, discrepant findings seem to reflect heterogeneity of factors contributing to obesity (45).

Unsuccessful treatment of obesity may have adverse psychological consequences. When children return to their everyday environment, it is not totally unexpected for them to regain weight (1). In this study, 24% of the children continued to lose relative weight after treatment, whereas all of the others showed weight increase. We wonder how far-reaching the experience of weight regain is and whether feelings of frustration adversely impact children's self-worth. In this study, global self-worth of the children increased. Furthermore, the children were found to experience less psychopathology at follow-up. Thus, the data in this study show no psychological side effects associated with the weight evolution. In line with others (26,46), this study shows that pediatric obesity treatment has no long-term detrimental impact on the child's eating behavior.

One of the strengths of this study is that pretreatment characteristics were not collected retrospectively, but prospectively. In addition, adopting a multi-method, multi-informant procedure by including self-reports, parent reports, and interviews avoids idiosyncratic or biased single informant perspectives and improves the reliability of measurements (47). Statistical analyses were limited to eight predictors to safeguard adequate predictive power. The program was conducted in a local clinic, and the children were not selected for participating in a university-based program. This design enhances the ecological validity. Enthusiasm on the grounds of this study must, however, be tempered with a cautionary note. It could be argued that regression to the mean may exaggerate the improvements in the psychological measures. Furthermore, this study did not include a control group; therefore, the interpretation of the changes must be seen as tentative. Normally, it is expected that, as girls progress through sexual maturation, they report increased body image dissatisfaction and disordered eating. Research also suggests that overweight children tend to gain weight over time (48). Furthermore, naturalistic weight reduction efforts by adolescents are more likely to result in weight gain than in weight loss probably because of deregulation of the normal appetite (49). In this light, the 2-year outcome data that show an improvement rather than worsening of the eating and weight problems are particularly encouraging.

At the 2-year follow-up, overweight was still prevalent in most children. Moreover, a trend toward weight regain was found in 73% of the children after discharge. More research is planned to study whether weight regain can be diminished if the treatment is followed by booster sessions in which the children learn maintenance strategies. Recently it was argued that continuous care of indefinite duration might be necessary to achieve long-term treatment effects for the majority of obese people (8). The challenge is to convince health care professionals, obese persons, and the general public that obesity is a complex, chronic condition that can be managed, although not cured, through intensive programs of ongoing care.

To conclude, pediatric obesity treatment is moderately successful, but long-lasting weight loss is associated with several patient characteristics. It is important to identify in a pretreatment assessment phase initial weight, age, and psychological variables, because they predict treatment success. Identification of these clinical markers associated with response to treatment can help us advise our clients about realistic weight loss goals. Moreover, information on pretreatment patient characteristics can be used to identify at- risk groups and to tailor obesity treatments to specific groups of patients.


The author thanks the Zeepreventorium De Haan staff and all of the children for participating in this study. There was no funding/outside support for this study.


  1. Nonstandard abbreviations: SES, socioeconomic status; SPPC, Self-Perception Profile for Children; CBCL, Child Behavior Checklist; YSR, Youth Self-Report; EDE, Eating Disorder Examination.