Predictors of Child Psychological Changes During Family-Based Treatment for Obesity

Objective:  To explore the influence of 1-year changes in child obesity and maternal psychopathology on changes in child psychological problems.

Design  Hierarchical regression models were used to predict child psychological change, with demographic variables, maternal psychological change, and child percentage overweight change as predictors.

Setting  Pediatric obesity research clinic.

Participants  Clinic sample of 116 obese 8- to 12-year-old children and their mothers.

Interventions  Family-based behavioral weight-control program.

Main Outcome Measures  Child psychopathology was assessed via mother-reported Child Behavior Checklists and maternal psychopathology was determined by standardizing scores on the Cornell Medical Index and the Symptoms Checklist-90–Revised.

Results  Significant improvements were observed in child percentage overweight (−20.1% overweight), and child and maternal psychopathology. Improved maternal psychopathology accounted for a significant amount of variance in improvements in the Child Behavior Checklist total Problems Scale and internalizing and externalizing problems subscales. Decreased obesity accounted for a significant amount of variance in improvements in the Total Competence scale and, somatic complaints, social problems and social competence subscales of the Child Behavior Checklist. Significant interactions of child obesity change by sex were found for Total Problems and externalizing scores. The interactions were due to girls with greater obesity reduction showing greater improvement in Total Problems, whereas boys with greater obesity reduction showed less improvement in externalizing problems.

Conclusions  These results highlight the multidimensional nature of psychosocial functioning in obese children and call attention to multiple avenues for intervention to improve their psychosocial functioning.

OBESITY IS a chronic and prevalent pediatric disease.¹ Obese children seeking treatment have an increase in social and psychological problems compared with population normative data.^2-4 These psychological problems are generally thought to be related to the negative psychosocial consequences of obesity.⁵ However, parental psychopathology is also a reliable predictor of child psychological problems.⁶ We have explored the independent associations of these 2 types of predictors in cross-sectional studies and consistently have found maternal psychopathology to be a strong correlate of child psychopathology.^7,8 In addition, when maternal psychopathology is controlled for, child obesity does not have a strong relationship with child psychosocial problems.^7,8 Finally, both parent and child psychological problems predict the degree of the child's success in treatment.⁹

The present study sought to extend our cross-sectional findings on the relationship between child psychopathology, maternal psychopathology, and child obesity^7,8 using a prospective design over the 1-year course of a family-based treatment for pediatric obesity. Although cross-sectional replication of the relationship between child and maternal psychopathology in obese children increases confidence in the findings, the demonstration that changes in child psychopathology are related to maternal psychopathology change provides a stronger prospective test of this relationship than the previous analyses.^7,8 The current study also sought to determine whether child obesity change accounts for incremental variance in child psychopathology change beyond that accounted for by change in maternal psychopathology. While cross-sectional studies have not found a relationship between child obesity status and child psychosocial functioning when maternal psychopathology was controlled,^7,8 demonstration that child obesity change is related to change in child psychosocial distress independent of changes in maternal psychopathology would provide evidence that obese children's psychosocial status is, in part, a function of their obesity. Finally, since the stigmatizing nature of obesity appears to be greater for females than males,^10,11 the interaction of child obesity change by child sex with changes in child psychopathology was explored.

Participants were 116 children (8 to 12 years old) and their mothers who had participated in 1 of 2 16-week family-based weight-control programs. Inclusionary criteria for acceptance in the program were as follows: child between 20% and 100% overweight at prescreening, neither parent greater than 100% overweight, one parent willing to attend treatment meetings, and no family members on an alternative weight-control program or receiving psychotherapy. Because of the small number of fathers assessed at baseline and 1 year (n=40) only mother-reported Child Behavior Checklist (CBCL) data were included in the analyses reported here. Of the 150 families who began treatment, 92% (138) had psychopathology data completed by the mother at baseline and 84% (116) of these children were assessed at 1-year follow-up. Program 1 families (n=66) participated in a study exploring various ways to increase children's activity, whereas program 2 families (n=50) participated in a study that trained various members of the family in problem-solving skills.

Weight was recorded on a calibrated balance-beam scale and height was recorded on a stadiometer at baseline and 1-year follow-up. Child percentage overweight was calculated in reference to the 50th percentile body mass index (BMI) (which is calculated as weight in kilograms divided by square height in meters) based on child age; mother percentage overweight was calculated in reference to the BMI 50th percentile for women aged 25 to 29 years.¹² Children and mothers were considered to be clinically obese if they were at least 20% overweight.

Psychological measures were completed at baseline and at 1-year follow-up. The CBCL¹³ assesses behavioral problems in children via parental reports, and its validity and psychometric strengths have been well documented. The CBCL was scored using the IBM-PC version of the CBCL computer scoring program (Burlington, Department of Psychiatry, University of Vermont) that generates standardized T scores for a Total Problems score, 2 broad-band scores (internalizing and externalizing problems) and 8 behavior problem scores. Cutoffs used to establish clinical criteria were T ≥ 67 for the behavior problem scale, and T ≥ 60 for the internalizing and externalizing subscales and Total Problem subscales.¹³ In addition, competence scales assessing involvement in activities, social interaction, and school performance, as well as a Total Competence scale are determined, with a clinical cutoff of T ≤ 33 for these scales. Improvement is demonstrated by decreases in scores on the CBCL Total Problems scale, and on the broad-band and behavior problem subscales, and increases in scores on the competence scales. All CBCL results were based on maternal reports.

Maternal psychosocial problems were assessed differently in the 2 programs. In program 1, maternal psychiatric distress was measured using the Cornell Medical Index (CMI),¹⁴ a self-report questionnaire assessing adult physical and psychological complaints. The total CMI score has been established as a screening device to detect psychological disturbance¹⁵ and has been shown to discriminate emotionally maladjusted from normal individuals.¹⁶ Clinical criteria for psychiatric distress was 8 or higher for the mothers.¹⁷ In program 2, maternal psychopathology was assessed with the Symptoms Checklist-90–Revised (SCL-90–R), a reliable and valid self-report inventory¹⁸ designed to be used with community, medical, and psychiatric samples. The SCL-90–R provides a global distress index (Global Severity Index) as well as 9 symptom dimensions (somatization, obsessive-compulsive, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, psychoticism). The SCL-90–R raw scores were converted to T scores based on norms provided for nonpatient women.¹⁸ Clinical criteria for the SCL-90–R was met if T ≥ 63 was observed for the Global Severity Index or 2 or more symptom dimensions. The total CMI score and SCL-90–R Global Severity Index were standardized within mothers for programs 1 and 2, respectively, to create a standardized maternal psychopathology variable.

Family socioeconomic status was assessed at baseline, using the Hollingshead Four Factor Index of Social Status.¹⁹ Scores are derived from information about education level, occupation, sex, and marital status of the head of household. If there are 2 heads of households, their mean scores are averaged.

Participating children and parents across both programs attended 16 weekly meetings followed by 2 bimonthly booster sessions, a 6-month assessment, and a 1-year follow-up. All groups received the Traffic Light diet²⁰ and intensive behavioral intervention.²¹ Treatment outcome results from these 2 studies will be presented in other future articles. This research was approved by the University at Buffalo Human Subjects Review Board, Buffalo, NY, and informed consent was obtained from the parent and the child.

The CBCL provides global (Total Problems and Total Competence scales), broad-band (internalizing and externalizing subscales), and specific (8 behavioral problem subscales and 3 competence subscales) measures of child psychopathology. All analyses were first conducted on the global measures of child psychopathology (ie, Total Problems) or child competence (ie, Total Competence). Bonferroni corrections were done on subsequent analyses to control for type I error, with P <.025 (.05/2) for the analyses on the broad-band scales, P <.006 (.05/8) for the behavioral problem scales, and P <.017 (.05/3) for the competence scales. Repeated measures analyses of variance (ANOVAs) were conducted on the 9 SCL-90–R symptom dimensions, with a Bonferroni correction of P <.006 (.05/9).

Preliminary 1-way between-group ANOVAs explored demographic, obesity status, and psychopathology differences between families who completed the 1-year follow-up and those unavailable for assessment. Then, the weight and psychopathology change data were explored in 2 ways. First, changes across time in child and mother percentage overweight and psychopathology were assessed by repeated measures ANOVAs with baseline and 1-year values as the repeated variable. χ²Analyses were used to determine changes across time in the percentage of children or mothers scoring in the clinical range for obesity or psychopathology.

Second, the independent multivariate predictors of changes in child psychological status were established. Hierarchical regression analyses assessed independent and incremental correlates of changes in child psychopathology for each CBCL scale. Demographic control variables were entered on block I. Because previous cross-sectional research established maternal psychopathology as a strong correlate of child psychopathology, change in maternal psychopathology was entered next on block II. To determine whether differential change in maternal psychopathology across programs 1 and 2 accounted for variance in the dependent measures, the interaction term of maternal psychopathology change by program was entered on block III. Child obesity change was entered on block IV to determine the amount of incremental variance in child psychopathology accounted for by this variable when controlling for the effect of maternal psychopathologic change. Finally, the child obesity change×child sex interaction term was entered on block V.

Program and child sex were entered on block I in all models because of subsequent inclusion of the interaction terms (ie, maternal psychopathology change×program and child obesity change×child sex). To determine which other demographic and treatment control variables should be entered on block I, a preliminary regression model was done for each CBCL scale. Family socioeconomic status, child sex, child age and program (1 or 2) were entered simultaneously, and if socioeconomic status and child age contributed significantly (P<.05) to the model, these variables also were included in block I of the overall model for that scale.

Residualized change scores were used in all regression analyses to control for the variance in change attributable to baseline values. Residualized change scores were created by regressing baseline values on the baseline to 1-year change values for each variable. For example, residualized change scores from baseline to 1 year in percentage overweight for children and mothers were developed following regression of the baseline values of percentage overweight on the change score (baseline subtracted from 1-year percentage overweight values). Similarly, residualized change scores for changes in child and maternal psychopathology were created by regressing the baseline value for the measure (CBCL scales or standardized maternal psychopathology variable) on the change value for that measure.

Two-tailed tests of significance were used for all analyses. The ANOVA and regression analyses were done using both CBCL T scores and raw scores. As results were similar, only analyses involving T scores are presented. All analyses were conducted using SPSS.²²

Child and parent characteristics at baseline and 1-year are shown in Table 1 and Table 2, respectively. There were no Bonferroni-adjusted significant differences in overweight status or age of children or mothers, sex of children, socioeconomic status of family, or child or maternal psychopathology at baseline between families available for follow-up and those who had dropped out of the study. Seventy girls and 46 boys completed the 1-year assessment, with the average child 10.4 (1.2) years (mean [SD]) of age, and 61%±16% overweight (mean [±SD]). As a group, children scored near or above normative levels for all CBCL scales at baseline. However, there was considerable variability in this sample of obese children. For example, the range of scores for Total Problems was 26 to 74, and 29% of the children met criteria on at least 1 CBCL behavior problem subscale. The most prevalent problems were internalizing (24%) and social problems (21%).

Mothers were 39.9 (4.5) years (mean [SD])of age, and 36.8% (27.4%) overweight (mean [SD]). Forty-two (36%) of the mothers were nonobese. The percentage of mothers who met clinical psychological criteria in the 2 programs was very similar, with 17% of program 1 mothers and 18% of program 2 mothers meeting clinical criteria. The average Hollingshead Four Factor Index of Socioeconomic status was 47.8 (10.5) (mean [SD]), with a range of 20 to 66, indicating that the families were generally of middle class status (medium-sized business owners, minor professionals, technicians, etc).

Both children and mothers significantly reduced their percentage overweight and BMI from baseline to follow-up. At 1 year, 16% of the children were nonobese, and 22% of mothers who had been obese at baseline were nonobese.

As shown in Table 1, a significant improvement from baseline levels to 1 year was demonstrated for CBCL Total Problems, Total Competence, internalizing problems and 5 behavior problem subscales: withdrawn, somatic complaints, anxious/depressed, social problems, and attention problems. The social competence subscales significantly increased.

By 1-year follow-up, only 13% of children met criteria on at least 1 CBCL behavior problem subscale (vs 29% at baseline), as seen in Table 1. Social problems still were the most prevalent, but the number of children reaching clinical criteria had been reduced by half. The percentage of children meeting clinical criteria was reduced significantly for Total Problems and internalizing scores.

No significant differences across time were found on the CMI for program 1 mothers, nor had the percentage of mothers reaching clinical criteria been reduced significantly at 1 year (from 17% to 14%). A significant reduction in the SCL-90–R Global Severity Index score was found for program 2 mothers, with the interpersonal sensitivity symptom dimension showing significant improvement. The percentages of program 2 mothers scoring in the clinical range on the Global Severity Index and interpersonal sensitivity symptom dimension were reduced significantly at follow-up.

Table 3 presents the results of the hierarchical regression analyses in which changes in maternal psychopathology and/or changes in child obesity accounted for a significant amount of the variance in child psychological changes. Only models in which maternal psychopathology change, child obesity change, or the child obesity change×child sex interaction term accounted for a significant incremental amount of variance in the model are presented. Program (1 or 2) and child sex are entered on block I, with changes in maternal psychopathology on block II, the maternal psychopathologic change×program interaction term on block III, change in child obesity on block IV, and the child sex×child obesity change interaction term entered on block V. Family socioeconomic status and child age were not significant in any models, thus only child sex and program were entered on block I. The cumulative variance accounted for by the entire model (R²_cum), as well as the incremental variance accounted for (R²_Δ) and significance level for each block are presented.

Maternal psychopathologic change was significant for Total Problems and both broad-band subscales scores. Improved maternal psychopathology was associated positively with improved child psychopathology in these models. The interaction of maternal psychopathology change by program was not significant in any model. Thus, despite the different degree of change in maternal psychopathology seen in program 1 vs program 2, the effects of maternal psychopathology change on child psychopathology change were similar for both programs, regardless of assessment used.

Child's change in relative weight was the only significant correlate for the Total Competence scale and somatic complaints, social problems, and social competence subscales. A decrease in obesity was related to decreases in child psychological problems and increases in competence. Since being overweight is an item on the CBCL social problem scale, hierarchical analyses also were conducted using the raw score of the social problems scale excluding the overweight item as the dependent measure. This analysis revealed change in obesity status to account for 5% (P=.014) of incremental variance in the change in this revised social problems scale score.

The interaction term of child obesity change by child sex accounted for a significant, independent amount of variance above and beyond that of the main effect of child obesity change in Total Problems scale and externalizing subscale. To explore these interactions, a median split on residualized obesity change was done for both girls and boys, and the residualized change scores for these CBCL scales were compared across obesity change levels (high vs low). Girls with greater obesity reduction showed greater reduction in Total Problems scores than girls with less obesity reduction (−3.1 vs +1.7), with comparable changes in boys' scores regardless of obesity change. Boys with less obesity reduction showed better improvement in externalizing scores than boys with greater obesity change (−0.1 vs +2.3), with comparable changes in girls' scores regardless of obesity change.

Children who participated in the family-based behavioral childhood obesity treatments showed significant changes in percentage overweight, with an average decrease percentage overweight of −20%. In addition, children showed significant changes in psychopathology on a variety of behavior problem, competence, and total scale scores. Almost one third of the children met clinical criteria on at least 1 CBCL subscale at baseline, and this percentage was halved after 1 year of observation. Results of the hierarchical analyses showed that improvements in maternal psychopathology and child weight status each contributed to improvements in child psychological status. This study is unique in demonstrating the influence of participation in a pediatric weight-loss program on child psychological changes across a wide variety of syndromes, independent of changes in maternal psychosocial functioning.

Improved psychological status of mothers was related to improved psychosocial adjustment in their children across the Total Problems scale and both internalizing and externalizing broad-band subscales. Previous studies have demonstrated that parental distress and child psychosocial problems in obese children are related cross-sectionally,^7,8 and that baseline levels of parental and child psychological problems are related to success in obesity treatment.⁹ The current prospective study strengthens confidence in the relationship between parent and child psychological functioning in obese children by demonstration of the relationship between change in parent psychological functioning and change in child psychological functioning.

The other major finding of this study was the independent association between improvement in child obesity and improvements in Total Competence, somatic complaints, social problems, and social competence. Based on the hierarchical regression methods used, improvements in child obesity added to the prediction of child psychological problems beyond improvements in maternal psychological problems. The demonstration of obesity changes being related to psychological changes supports hypotheses regarding the causative role of obesity in the development of psychological problems.⁵

Based on the consistency of these findings across studies and the relevance of these findings to theoretical models of obesity and psychological processes, there are several profitable directions in which this research could lead. One important direction is specification of mechanisms that may account for the relationships between these variables. There are many ways in which maternal psychopathology may influence child psychopathology, or that child obesity could influence child psychopathology, especially in the social arenas. While the current study was not designed to test these mechanisms, brief mention of hypotheses is warranted. Research on maternal depression and child psychopathology⁶ has suggested that the association between parental psychopathology and child psychopathology may be mediated by such variables as parenting,⁶ parent-child interaction,²³ and marital relationship.⁶ For example, family-based behavioral weight-control treatment includes teaching parenting skills²⁰ to elicit behavioral change, and an improved parent-child relationship following more effective parenting may improve the psychosocial status of multiple family members.²³ In addition, children's psychosocial experiences may reciprocally influence the psychological adjustment of their parents.⁶ Improvements in the psychosocial functioning of their children may reduce the stress and/or worry experienced by their mothers, thereby improving maternal psychosocial status.²⁴

Children's social relationships have important psychosocial consequences,^25,26 because children who are rejected or have poor social skills are at greater risk for psychosocial problems both in childhood²⁷ and as adults.²⁵ Research has documented social stigmatization against obese children,^28,29 and changing a child's degree of obesity may result in a reduction in social stigmatization and improved social relationships for obese children. Alternatively, improved peer relationships may contribute to improved child weight loss. Spending time with peers (in nonfood-related activities) can compete with eating, and obese children who became increasingly integrated into their social network may have had less opportunity for eating or more opportunities and support for physical activity than their more isolated peers.

Child sex interacted with changes in obesity to influence general psychological functioning, with girls who had better obesity change showing improvement on the Total Problems scale score whereas boys who were more successful with obesity change showed an increase in externalizing scores. No interactions with sex were observed for any of the specific behavior problem scales. It is not surprising that there may be some differences in the influence of weight change by sex, since females have more weight and dieting concerns³⁰ and show stronger relationships between weight, body-esteem, and overall self-esteem than males.³¹ In addition, females may be more stigmatized by obesity than males,¹⁰ hold a more negative perception of the stigmatization associated with obesity,³² and be more accepting of the stigma.¹¹

It would be advantageous in future studies to increase father measurement and participation in treatment. Mother and father psychosocial variables may have different effects on child functioning,²⁴ and inclusion of fathers allows for a more complete investigation of important aspects of the family environment. In addition, it would be useful to obtain reports on changes in child psychological problems from multiple sources, particularly since a mother's perception of her child's behavior may be biased by her own psychological status.³³ While there is some disagreement on this point,^34,35 the possibility remains that the association between improved maternal psychosocial functioning and improved child psychosocial functioning may be explained partly by a more positive maternal bias in perception of child behavior resulting from better maternal functioning. In addition, although our analyses suggest that the differential change in maternal psychopathology by program did not influence the change in child psychopathology in these obese children, this study would have benefited from the use of the same measure of adult psychopathology in both samples. The significant change in maternal psychopathology found with the SCL-90–R used in program 2 may be a reflection of the increased sensitivity of this measure to assess maternal distress compared with the CMI used in program 1.

The importance of both maternal psychological status and child obesity status on the psychological status of obese children highlights the multi-dimensional nature of psychosocial problems in obese pediatric populations, as well as the multiple avenues for intervention. Behavioral regimens involve interventions that may improve family functioning, or exacerbate conflictual family relationships and increase familial stress. When both child and parent experience improved psychosocial functioning, better parent-child interactions may be more likely to occur, perhaps even facilitating adherence, and thus weight control. Furthermore, improvements in body weight have positive psychosocial consequences. Given the importance of the family environment for pediatric patients, family-based interventions that target weight control may not only create the best opportunity for maintenance of treatment gains but may also increase the spectrum of positive clinical gains to include psychosocial outcomes.

Accepted for publication April 6, 1998.

This research was supported in part by grants HD RO123713, HD RO125997, and HD RO120829 from the National Institute of Child Health, Bethesda, Md, awarded to Dr Epstein.

Corresponding author: Leonard H. Epstein, PhD, Department of Psychology, University at Buffalo, Buffalo, NY 14260-4110.

Editor's Note: Now we have more evidence that obesity is a very complex problem. Like diets, if there was a model answer, we'd all be models.—Catherine D. DeAngelis, MD