This study is designed as a prospective observational study of children and adolescents undergoing an intensive combined lifestyle intervention during one year for their severe obesity. The treatment program has either a 2 months or a 6 months inpatient period.
The Medical Ethics Committee of VU University Medical Center Amsterdam has approved the study design, protocols and informed consent procedure. This study is a collaboration with a study that aims to provide insights into the effectiveness and cost-effectiveness of the interventions .
This study is carried out in the childhood obesity clinic Heideheuvel (part of Merem Treatment Centers) in Hilversum, The Netherlands between August 2009 and July 2013.
The aim is to include 120 children and adolescents and their parents/caregivers in the study. 80 of the patients will undergo a 2 months inpatient period and 40 a 6 months inpatient period.
Patients (8–19 years of age) are admitted to Heideheuvel for their severe obesity. Inclusion criteria are a SDS-BMI ≥ 3.0 according to the growth curves based on the fourth Dutch National Growth Study of 1997 (this corresponds to the 99.9th percentile) or a SDS-BMI ≥ 2.3 (corresponding to the 99th percentile) with obesity related morbidity (e.g. obstructive sleep apnea syndrome, hyperinsulinemia, diabetes type 2, liver function disorders, dyslipidemia, musculoskeletal problems). Before admission, these children and adolescents have, unsuccessfully, received outpatient obesity care elsewhere. Some patients, who are diagnosed to not be able to profit from outpatient care, are referred directly. All patients are referred by their local pediatrician.
Criteria for exclusion from the study are: syndromal/chromosomal determined obesity, obesity caused by endocrine disorders (e.g. hypothyroidism, Cushing syndrome, primary hyperinsulinemia, pseudohypoparathyroidism, acquired (structural) hypothalamic damage) or use of medication (e.g. antiepileptic drugs, antidepressants), psychiatric disorders (e.g. severe depression, schizophrenia) that may obstruct adequate treatment, presence of eating disorders (binge eating disorder, bulimia nervosa) to such a degree that specific therapeutic attention is needed before starting the intervention, children/adolescents or parents that can or will not give ‘informed consent’, parents that can or will not participate in the treatment, children/adolescents with an IQ below 75 or attending a school for intellectually challenged children. Written informed consents are obtained from both the participants (from age 12) and their parents.
Participants do not pay for the treatment which is temporarily reimbursed by the Ministry of Health, Welfare and Sports. For their participation in the research they get reimbursed with a 20 Euros gift voucher to cover travel expenses.
The study period is four years. Since the capacity to treat patients is limited in the clinic, the patients are treated in sequential groups of 10 children (8–13 years) or 10 adolescents (13–19 years). At the end of the fourth year 120 patients will have completed the program and follow up period.
Measurements will be taken at three points in time: at baseline (start of treatment; T0), at the end of treatment (12 months after baseline; T12), at the end of follow-up (24 months after baseline; T24). Besides weight and height, measured at T0, T12 and T24, measurements include (also for the parents of the patients) several questionnaires on psychosocial characteristics: Dutch Eating Behavior Questionnaire for children (administered at T0, T12, T24), General self-efficacy scale (T0, T12, T24), Self-Perception Profile for Children/Adolescents (T0, T12, T24), Treatment Self-Regulation Questionnaire (T0), parent and child version of the Parent–child Interaction Questionnaire (T0), Child Behavior Checklist (T0, T12, T24), Parental Feeding Style Questionnaire (T0, T12, T24), Positive and Negative Affect Schedule (T0), line drawings of silhouettes to choose from (T0, T12, T24). In addition the children’s self-regulation abilities are evaluated by two behavioral computer tasks: The Stop Signal Task (T0, T12, T24) and the Balloon Analogue Risk Task (T0, T12, T24). All questionnaires and computer tasks are described in the measurements section below.
The treatment is an intensive one year lifestyle intervention program by a multidisciplinary team with emphasis on nutrition, exercise and behavior and implementation of the learned behavior in the home situation. Treatment consists of group treatment as well as individual sessions. For an extensive description of the treatment components, see Makkes et al. .
There is active and frequent participation of the parents during the whole treatment because parents have a major influence on their children’s weight and weight-related behaviors through modeling and encouraging of appropriate health behaviors, controlling the quality and quantity of the food available in the household, facilitating physical activity, and engaging in appropriate feeding practices. An additional advantage of family-based interventions is that they can improve health behaviors and weight not only in the treated child but also in siblings and adults living in the household [61, 62].
The behavioral part of the treatment program is carried out by psychologists, social workers and group coaches in individual as well as group sessions with children and parents together and separately. The program uses behavior change techniques, e.g. from cognitive behavioral therapy, to improve the general ability to self-regulate. This improvement in general self-regulation abilities is assumed to facilitate the self-regulation of the newly learned healthier behavior. For example, the program teaches patients to delay gratification (I want a cookie now, but I will wait until it is tea time), to plan behavior (e.g. forming if-then plans: if it rains, I will still take my planned thirty minute walk) or to self monitor behavior (keeping a diary of the cookies eaten and the walks taken). The treatment further addresses: disordered eating behaviors, self-worth, self-efficacy, behavioral and emotional problems, autonomous motivation, body image, outcome expectations, mood disorders, eating and exercise behavior, interaction between parent and child, parental feeding styles, relationships with peers, body acceptance and coping. To achieve its aims, the program applies a number of behavior change techniques. These techniques can be specified with a 40-item taxonomy, called CALO-RE, that is developed by Michie et al. . The chief psychologist of the treatment program scored the intervention with this CALO-RE taxonomy, which gives an overview of relevant behavior change techniques for interventions aimed at increasing physical activity and healthy eating . The treatment program uses 27 of the 40 defined behavior change techniques to some extent (number 1–13, 15–16, 19, 21–22, 28–29, 33, 35–40). For example, the program uses goal setting to define a desired behavior that can be related to changes in diet or physical activity. Goal setting is also used to define health outcomes. Another used behavior change technique is prompting generalization of a target behavior from the clinic to the home situation. To learn how to cope with barriers to change the behavior, like certain feelings and desires, the treatment uses barrier identification and problem solving techniques. The treatment also uses modeling of behavior, which involves showing how to perform a certain behavior in for example a role play, by the staff, the peers and the parents. Another applied technique is motivational interviewing in which the staff is trained and that is used to assess and if necessary help change the phase of behavior change the patient is in.
The primary outcome measurement is the gender and age-specific ΔSDS-BMI between baseline (T0) and follow up (T24) indicating total weight loss from baseline to follow up and ΔSDS-BMI between end of treatment (T12) and follow up (T24) indicating weight loss maintenance after treatment.
Determinants of long-term weight loss and weight loss maintenance
Behavioral measures of general self-regulation ability
Our hypothesis is that a primary determinant of (sustained) weight loss is the general ability to self-regulate. We assess two separate aspects of general self-regulation: sensitivity to reward and inhibitory control. Both are measured with two computerized behavioral tasks that are not food related and an objective measure of the general ability to self-regulate: the Stop Signal Task and the Balloon Analogue Risk Task [64, 65].
The Stop Signal Task is a measure of inhibitory control  also referred to as executive response inhibition . The task correlates significantly with related measures for self-control . A Dutch adaption of the task by C. Nederkoorn was used . The Stop Signal Task is a computer task consisting of a series of 4x64 trials, preceded by two practice blocks of 8 and 16 trials, in which individuals are exposed to a colored quadrangle on the right or the left side of the screen (the go signal). Individuals have to press the shift button on the side where the quadrangle appears as quick as possible. However, in 25% of the trials respondents hear a tone, after which individuals should inhibit their response to press the shift button. Failing to inhibit the response when the stop signal sounds, indicates lack of self-regulation. At the start of the task the delay between the go signal and the stop signal is set to 250 milliseconds. The computer changes the stop signal delay after every stop signal trial. Following an unsuccessful inhibition the delay is decreased by 50 milliseconds, making the task easier. After a successful inhibition the task is made more difficult by increasing the stop signal delay by 50 milliseconds. This dynamic adjustment results in a stop signal delay that enables the participant to stop on approximately 50% of the stop trials. The main outcome is the stop signal reaction time (SSRT), which is calculated by subtracting the stop signal delay from the mean go signal reaction time, measured in milliseconds. A longer SSRT means that the participant takes longer to inhibit his/her response and indicates less inhibitory control . The Stop Signal Task demonstrated less inhibitory control in children with impulse control problems like ADHD [64, 69] and in obese children [28, 70], making it a valid test to study self-regulation in the present sample as well.
The Balloon Analogue Risk Task (BART) is a measure of sensitivity to reward , which has been shown to correlate significantly with related measures of self-control  and impulsivity . The BART assesses the willingness to consciously risk long term loss in order to get higher rewards on the short term . A Dutch adaption of the task by C. Nederkoorn was used. This version of the BART is a computer task consisting of a series of 29 trials. In every trial a small red balloon is shown on the computer screen. The participant is instructed to inflate the balloon by pressing a pump button. He or she is warned that the balloon will explode at some point, but that this can happen any time from after the first pump until after the balloon has expanded to fill the whole screen. Every pump produces a gain of 5 points in a temporary bank. As long as the pumping continues, the points go up. However, the balloons all explode at some point, according to a predetermined sequence with the weakest balloon exploding on the first pump and the strongest balloon exploding after 128 pumps. If that happens before the gained points are transferred to the permanent bank, all points for that balloon are lost. Transfer to the permanent bank can be done at any chosen moment by pressing the collect button and results in the next trial with a new uninflated balloon. The total amount of points earned is shown on the computer screen . In deciding whether to make each pump, the participant must balance the potential gain of accruing more money against the potential risk of losing all money accrued for that balloon . Individuals who are more sensitive to rewards will demonstrate a higher number of pumps on each balloon prior to money collection . For analysis of the results the adjusted value is used, defined as the average number of pumps excluding balloons that exploded [65, 72, 73].
Self-report measure of eating specific self-regulation ability
In addition to the computer tasks, self-reported indicators of eating-specific self-regulation are assessed. Hereto, a version specifically adapted to children  of the Dutch Eating Behavior Questionnaire (DEBQ), originally developed by Van Strien and colleagues , is used. This 33-item questionnaire assesses three eating styles that reflect dysfunctional regulation of food intake: 1) external eating (i.e. the tendency to eat in response to external cues), 2) emotional eating (i.e. the tendency to overeat in response to emotions) and 3) restrained eating (i.e. the tendency to attempt to refrain from eating) .
Moderators of the relation between self-regulation ability and weight loss and weight loss maintenance
As outlined earlier, factors related to competence, motivation, relatedness and outcome expectations will be examined as moderating factors. Hereto, all moderating factors are assessed at baseline and at follow-up examinations. Except for a few moderators that are only measured at T0.
Self-efficacy is measured with the Dutch adaptation of the 10-item General self-efficacy (GSE) scale, which assesses one’s sense of personal competence to cope across a variety of demanding or novel situations .
Self-worth is measured with Dutch adaptations [77, 78] of the Self-Perception Profile for Children (SPPC) and Adolescents (SPPA) developed by Harter [79, 80], which are questionnaires with 36 items (SPPC) or 35 items (SPPA) in six (SPPC) or seven (SPPA) domains: 1) global self-worth, 2) scholastic competence, 3) social acceptance, 4) athletic competence, 5) physical appearance, 6) behavioral conduct, and 7) close friendship (only SSPA).
Motivation to enter a weight loss program is measured with a 18-item version of the Treatment Self-Regulation Questionnaire (TSRQ) that assesses reasons to enter a weight-loss program . The questionnaire was translated into Dutch by two of the authors of this article (JH and EdV) with help from a native English speaker, in which process it was slightly adapted for use with children and adolescents. It assesses to what extent children and adolescents enter a weight-loss program for autonomous (e.g. “I really want to make some changes in my life”) and controlled reasons (e.g. “People will like me better when I’m thin”) on a 7-point Likert scale ranging from 1 (not at all for this reason) to 7 (totally for this reason).
Interaction between child and parent is measured with the Dutch version of the Parent–child Interaction Questionnaire – Revised (PACHIQ-R)  that assesses how the parent evaluates the relationship with the child (PACHIQ-R Parent version, 21 items) and how the child assesses the relationship with the parent (PACHIQ-R Child versions for mother and father, 25 items). The questionnaire concerns both behavioral interaction and attitudes. It has two subscales: conflict resolution (quality of preventing and solving conflicts) and acceptance (warmth, comfort, protection).
Social competences and social problems as assessed by the parent are measured with the Child Behavior Checklist (CBCL/6-18)  that rates children’s functioning. We use the Dutch translation  of the questionnaire that assesses competences (16 items) and behavioral and emotional problems (120 items) in the past six months. In the present study, only two scales are included: 1) The “Social scale” (this is one of the three competence scales) that comprises 6 questions, which concern group activities and social relationships (e.g.: “Please list any organizations, clubs, teams, or groups your child belongs to”; “About how many close friends does your child have? (Do not include brothers & sisters)”; “Compared to others of his/her age, how well does your child play and work alone?”) . 2) The “Social problems scale” (this is one of the five emotional and behavioral problem scales) that consists of 11 items, which concern topics like loneliness, jealousy, getting teased and clumsiness (e.g.: “Clings to adults or too dependent”; “Complains of loneliness”; “Easily jealous”; “Feels others are out to get him/her”) .
Parental feeding style is measured with a Dutch translation  of the 27-item Parental Feeding Style Questionnaire (PFSQ) which asks parents to report the frequency with which they use a number of feeding strategies that are grouped into 1) emotional feeding (i.e. “I give my child something to eat to make him/her feel better when s/he is feeling upset), 2) instrumental feeding (i.e. “I reward my child with something to eat when s/he is well behaved”), 3) prompting/encouragement to eat (i.e. “I praise my child if s/he eats what I give him/her”) and 4) overt control over eating (i.e. “I decide how many snacks my child should have”) .
Affect of the parents is measured with the Dutch translation of the 20-item Positive and Negative Affect Schedule (PANAS) [87, 88], which asks parents to indicate what they currently feel on two scales: 1) positive affect (the extent to which a person feels enthusiastic, active and alert) and 2) negative affect (subjective distress and unpleasurable engagement that subsumes a variety of aversive mood states, including anger, contempt, disgust, guilt, fear and nervousness).
Peer body size is measured by asking children and adolescents to select from a range of line drawings (separate for boys and girls) the silhouette that most closely resembles how their five best same sex friends look. The used drawings are taken from Colleen S.W. Rand , who composed the drawings based on drawings by M.E. Collins  and T.I.A. Sorensen et al. .
The same silhouette drawings are also used to assess outcome expectations. Children and adolescents are asked to select 1) the silhouette that most closely resembles their current body size and 2) the silhouette that most closely resemble the body size they expect to have after treatment. Outcome expectations are calculated by measuring the difference between these current and expected own body sizes.
Several covariates will be assessed. It is likely that the intensity and duration of treatment are related to both the ability to self-regulate and ΔSDS-BMI. We will investigate the intensity of the treatment, expressed by the duration of the residential period (2 versus 6 months) as it might explain part of the variation between individual children and adolescents in weight loss and in the ability to self-regulate at the end of treatment (T12). We will also investigate the health care the children and adolescents receive in the period between the end of the treatment at Heideheuvel and the follow up one year later as the prolongation of treatment might influence weight loss maintenance after the first year of treatment. Furthermore we will assess the gender and age-specific ΔSDS-BMI between baseline (T0) and end of treatment (T12) because the amount of weight lost during treatment might be a predictor of weight loss maintenance after treatment. Finally the influence of SDS-BMI at the start of treatment, gender, age and educational level of the children and adolescents on the long-term weight outcome will be assessed.
First we will provide information on the baseline characteristics of the patient group. Secondly, SDS-BMI levels and changes will be described at the three time points, i.e. at baseline (T0), after one-year (T12) and at two-year follow-up (T24). This allows for the description of total weight loss (computed as ΔSDS-BMI between T0 and T24) and weight loss maintenance (computed as ΔSDS-BMI between T12 and T24). Next, determinants (general and eating-specific self-regulation) and moderators (factors related to competence, motivation, relatedness and outcome expectations) will be described and compared at baseline (T0) and one year later (T12). This enables detailed scrutiny of the variability of the determinants and moderators in this specific target group of severely obese children and adolescents. Furthermore, it will provide insight into the change in determinants and moderators after following a treatment of one year.
Third, the relation between self-regulation and initial weight loss at one-year follow-up and potential moderators of this relation will be tested using hierarchical multiple regression analyses. Linear regression analyses (with ΔSDS-BMI between T0 and T12 as dependent variable) and logistic regression analyses (with a dichotomous variable high versus low weight loss as dependent variable) will be conducted. To discriminate participants with a high and low weight loss a median split in ΔSDS-BMI between T0 and T12 will be conducted. In the first step, other factors will be added (e.g., age, gender, SDS-BMI at T0, treatment intensity). In the second step, the self-regulation indicators (at T0) will be entered simultaneously as independent variables. In the third step, interaction terms between the moderators (at T0) and (the most important) self-regulation indicators will be computed and added to the model. Significant interaction terms will subsequently be decomposed to gain insight into the relation between self-regulation and initial weight loss at high and low levels of the moderator.
Finally, the relation between self-regulation and weight loss maintenance at two year follow-up and potential moderators of this relation will again be examined using hierarchical multiple regression analyses. Linear regression analyses (with ΔSDS-BMI between T12 and T24 as dependent variable) and logistic regression analyses (with a dichotomous variable high versus low maintenance of weight loss as dependent variable) will be conducted. To calculate high versus low maintenance of weight loss a median split in ΔSDS-BMI between T12 and T24 is conducted. The same steps as in the one-year analyses will be conducted, except that for the determinants and most moderators the data collected at T12 will be entered. A few moderators that are only measured at T0, will be entered as assessed at T0.
We will explore the possibility of multiple imputations to deal with missing data. Missing data will be imputed multiple times, using the ‘Predictive Mean Matching’ (PMM) procedure that runs under the SPSS 20.0 program. All analyses will be performed on the complete case data and on the imputed data. In case of missing item scores, we will explore the possibility of a single imputation technique.