Study reference

Study characteristics

Patient characteristics

Prognostic factor(s)

Follow-up

Estimates of prognostic effect

Comments

1st author,

year of publication

Type of study:

Setting and country:

Funding and conflicts of interest:

Inclusion criteria:

Exclusion criteria:

N=

Mean age ± SD:

Sex: % M / % F

Potential confounders or effect modifiers:

Describe prognostic factor(s) and method of measurement:

Duration or endpoint of follow-up:

For how many participants were no complete outcome data available?

N (%):

Reasons for incomplete outcome data described?

(Adjusted) Factor-outcome associations (include SEs or 95%CI and p-value if available):

Incremental predictive value¹:

Goldschmidt, 2014

Type of study:

Randomized controlled trial

Setting and country: USA

Funding and conflicts of interest:

Work was supported by NIH grants R01 HD036904, K24 MH070446, and T32 HL007456, and NCRR grants KL2 RR025000, KL2 RR024994, and UL1 RR024992

Inclusion criteria:

Overweight children with at least one overweight parent (BMI ≥25), who were involved in RCT of MT interventions following FBT (Wilfley, Stein, et al., 2007)

Exclusion criteria for child and parents:

- medical or psychiatric disturbances that would preclude treatment participation

- use of appetite- and/or weight-affecting medications

- concurrent involvement in weight loss or psychological treatment.

N= 150 overweight children (20-100% above their age- and sex-specific median BMI)

Mean age ± SD: 7-12 year

Sex: 30.7% M/69.3% F

BMI z-score = 2.21 ±0.30

Group was randomized over two different maintenance treatment approaches:

- Social facilitation maintenance treatment (SFM);

- Behavioural skills maintenance treatment (BSM)

- Demographic measures; Child z-BMI, calculated using CDC normative data; socioeconomic status (Hollingshead four-factor index)

- Self-Efficacy for children: Child Dietary Self-Efficacy Scale (CDSS); child-reported Self-Efficacy Scale for Children’s Physical Activity (SESCPA)

- Self-Efficacy for parents: The Eating/Exercise Habits Confidence Survey (EHCS)

- Social support; self-reported using the Social Support for Eating/Exercise Habits Survey (SSEHS)

- Psychological Functioning for children: The Child Eating Disorder Examination (ChEDE); the Child Behavior Checklist (CBCL)
- Psychological Functioning for parents: Brief Symptom Inventory (BSI), Eating Disorder examination-Questionnaire (EDE-Q)

FBT consisted of 20 weekly sessions delivered over 5 months.

The two MT approaches (SFM and BSM) consisted of 16 weekly sessions delivered over 4 months.

Psychosocial and anthropometric data were obtained at:

- Baseline, prior to initiating FBT;

- Randomization (month 5 post-FBT)

- Post-weight maintenance (month 9)

- 1-year follow-up (month 17; 1 year post-FBT)

- 2-year follow-up (month 29, 2 years post-FBT)

For how many participants were no complete outcome data available?

N (%):

Due to attrition weight data were available for

- 145 children at post-weight maintenance (97.6% of those randomized

- 137 children at 1 year follow-up (91.3%)

- 124 children at 2-year follow-up (82.7%)

A subset of families were lost to follow-up (most of whom were assessment non-completers or FBT dropouts and hence not randomized to a maintenance condition).

No demographic variables were significant predictors or moderators

Prediction of child z-BMI changes from baseline to randomization [post-family-based weight loss treatment (FBT)]

ΔR² = 0.17, which means that the predictors in the model accounted for 17% to the change of z-BMI scores in the model. Significant predictors in Table 2:

- Child z-BMI, B:-0.57, SE=0.26, p=0.029

- Child age, B: -0.16, SE=0.06, p=0.008

- Parent EHCS Efficacy to Reduce Calories, B: .21, SE=0.08, p=0.006

Positive coefficients indicate that higher levels of the predictor are associated with greater z-BMI reduction.

A low baseline child z-BMI is associated with greater z-BMI reductions

A low child age is associated with greater z-BMI reductions, and

And a higher score of parents to reduce calories is associated with greater z-BMI reductions

Higher baseline z-BMI and older child age predicted smaller z-BMI changes from baseline to randomization, whereas greater parent self-reported efficacy to reduce calories predicted greater z-BMI change in their children from baseline to randomization. Parents' reported family encouragement and friend encouragement were not significant predictors in this model (ps>.09). Parents' weight loss remained a significant predictor of their children's z-BMI change from baseline to randomization, accounting for 8% of the variance in children's z-BMI change during FBT after accounting for the baseline predictors.

Baxter, 2013

Type of study:

Prospective cohort study

Setting and country:

Queensland, Australia

Funding and conflicts of interest:

No conflict of interest. Study was funded in part by a National Heart Foundation Grant-in-Aid, the Children’s Hospital Foundation, ANZ Trustees and the University of Queensland. First author was supported by an Australian Post Graduate Award Scholarships. Template systems were donated by Pharmacy Health Solutions Pty Ltd.

Inclusion criteria:

Cohort members were sourced from the Eat Smart Studies, which included a pilot (2007) and a RCT (2008-2010)

- Age 10-17 years

- Simple obesity, measured by BMI > 90^th percentile for sex and age based on CDC

-referral from health professional

- motivated to lose weight (self-reported)

Exclusion criteria:

- prescribed medication known to alter body composition, insulin sensitivity or metabolism

N= 88 individuals

Mean age ± SD:

13.1 years (±1.9)

Sex: 30.7% M/69.3% F

BMI: 32.7 (5.7) kg/m²

BMI z-score: 2.2 (0.4)

Caucasian: 91%

Asian: 2%

Indigenous or Pacific Islander: 7%

Change in BMI z-score from baseline to week 12 of the intervention

- Demographic characteristics (age, sex and ethnicity)

- Social disadvantage was determined using the Socio-economic index for Areas (SEIFA)

- Insulin resistance was determined using the homeostatic model assessment for insulin resistance (HOMA-IR)

- Physical Activity level (PAL) was calculated via a 4 day physical activity diary

- Dietary restraint was measured with the Dutch Eating Behaviour Questionnaire – restraint subscale, child version (DEBQ-R)

Primary dataset contained 116 individuals. Those who did not complete the intervention in its entirety to the week 12 assessment (n=14) and those in the untreated control group (n=14) were removed.

This resulted in a dataset of 88 individuals, 45 in the modified carbohydrate group and 43 in the low fat dietary group.

This model accounted for 36% of explained variance for change in BMI z-score.

Table 2:

Variable, [Change (95% CI), p-value]

- Baseline BMI z-score, [0.112 (0.059-0.165, 0.000]

- Baseline insulin resistance (HOMA-IR), [0.013 (0.002-0.025), 0.023]

- Social advantage (SEIFA), [-0.041 (0-0.078, -0.004), 0.030]

- Referral source, [0.047 (0.010-0.085), 0.015]

A positive coefficient indicates a decrease in the explanatory variable is associated with greater decrease in outcome.

A lower baseline BMI z-score, a lower baseline insulin resistance, a higher social advantage and lower referrals from a specialist doctor is associated with a higher decrease in z-BMI scores.

Frohlich, 2011

Type of study:

Longitudinal study. Multivariate regression analysis

Setting and country:

Germany

Funding and conflicts of interest:

No financial relationships relevant to this article to disclose

Inclusion criteria:

1. BMI > 97^th percentile for age and gender or BMI > 90^th percentile with the presence of additional risk factors or diseases (eg. Hypertension, dyslipidemia, diabetes, orthopaedic problems)

2. referral for weight-reduction treatment by a local paediatrician

3. attendance at a regular school

Exclusion criteria:

Not specified

N= 111 parent-child/adolescent dyads with children aged 7 to 15 years (mean: 11.5 [SD: 1.84] years)

BMI SD score mean: 2.43 [SD: 0.44] [range: 1.31–3.54])

Longitudinal analysis with 3 assessment waves: at baseline (T0: within 3 weeks before the start of the intervention), body weight and height of participants and family member and psychosocial family characteristics were assessed at the conclusion of the program (T1: 1 year after T0). Body weights and heights were reassessed 1 year after conclusion (T2: 2 years after T0)

Parent/child dyads took part in the outpatient lifestyle intervention Fit Kids for overweight and obese children and adolescents. The program combines behavioural therapy for eating and physical activity, physical exercise and dietary courses. Families are continuously involved.

Duration or endpoint of follow-up:

2 years

For how many participants were no complete outcome data available?

Of 111 children and adolescents who started, 95 (85.6%) completed the 12-month program and took part in the T1 assessment. Of these children, 78 (82.1%) attended the 1-year follow-up (T2)

41 children/adolescents showed long-term success, 53 children failed to reach the criterion of < 5% BMI-SDS reduction (n=37) or dropped out of the intervention (n=16).

Reasons for incomplete outcome data are not described

Success versus failure in weight reduction was not significantly associated with the presence versus absence of somatic risks (e.g. hypertension, dyslipidemia, diabetes, orthopaedic problems).

Weight reduction up to 12 month follow-up:

The following predictors explained for 27% the variance in the model: Age of child, maternal obesity, obese sibling, family adversity index, maternal depression, maternal attachment style, AAS close, anxiety.

Only the presence of obese siblings was statistically significant, with an Odds Ratio of 5.57. More details are listed in Table 2 in the paper.

A second stepwise regression revealed that the predictors maternal depression and obese siblings accounted for 18% the variance in the model, which was statistically significant. Meaning that children with obese siblings and children whose mothers report pronounced depression are at an increased risk of not reducing weight considerably.

Factors predictive for weight change during the period after completion of the program (T1 tot T2):

The β coefficients of the variable in the multiple regression analysis indicates that the anxious attachment attitude of the mother exclusively explains significant unique variance (β=0.27), 14%.

Wrotniak, 2004

Type of study:

Secondary data analysis from 3 family-based randomized controlled weight studies. Hierarchical regression models

Setting and country:

USA, New York state university at Buffalo.

Funding and conflicts of interest:

Study was supported by grants HD 25997 and HD 20829 from the National Institute of Child Health and Human Development, Bethesda.

Inclusion criteria:

Children with BMI > the 85^th percentile and parents’ BMI > than the 70^th percentile.

Only families with complete parent and child data at all time points were included for analyses.

Exclusion criteria: not specified

N= Obese 8- to 12-year-old children and their parents from 142 families

Mean age ± SD: 10.2 (1.2) years

Sex: 41% M / 59% F

58 boys, 84 girls

Potential confounders or effect modifiers:

All families were provided with an educational program based on the Traffic Light diet and a physical activity program.

Hierarchical regression models were used to establish if parental change in z-BMI accounts for a significant amount of incremental variance in child z-BMI change when controlling for variables forced into step 1 of the model. Variables forced into step 1 included: child’s sex, parent’s sex, SES, child’s age, parent’s age, baseline child z-BMI, baseline parent’s z-BMI. To control for the different studies dummy variables were created. Hierarchical regression models also examined the influence of parent weight change on child weight change.

No information is given about incomplete or missing data.

The participants included in the report are from 3 studies that differed in goal and hypotheses. Each study observed participants during a 2-year period, with assessments at baseline, 6, 12 and 24 months.

Regression analysis revealed that parent z-BMI change significantly predicted child z-BMI change for the 0-6 month time point (β=0.46 [0.31-0.61], p<0.001, R²=0.21) and also for the 0-24 month time point (β=0.29 [0.07-0.51], p=0.009, R²=0.05).

Regression analysis also determined that parent weight change was a significant predictor of child weight change for both time points:

0-6 months: β=0.24 [0.15-0.32], p<0.01, R²=0.18)

0-24 months: β=0.19 [0.04-0.34], p=0.02, R²=0.04)

(Information from text, results section, las paragraph on page 344)

Numbers in text (last paragraph results section on page 344) and Table 2 do not correspond.

Epstein, 2001

Type of study:

Randomized controlled trial

Setting and country:

Department of Pediatrics, University at Buffalo School of Medicine and Biomedical Sciences,

Buffalo, New York.

Funding and conflicts of interest:

This research was supported in part by Grant HD34284

Inclusion criteria:

1) child at or >85th

body mass index (BMI) percentile for age and sex and <100% over the average BMI for age and sex; 2) child

between 8 and 12 years of age; 3) neither parent over 100%

overweight; 4) one parent willing to attend weekly treatment

meetings; 5) neither parent or child on an alternative

weight-control program; and 6) no medical restrictions on

the participating parent or child that would prevent exercise,

no current psychiatric disorders in parents or children, and

no history of eating disorders in parents.

N= 67 families with obese children and 89 siblings.

Describe prognostic factor(s) and method of measurement:

Participants were all children and parents living at home

from families who participated in two cohorts of the Childhood

Weight Control Program at the University at Buffalo

Duration of follow-up was 1 year

One family was excluded from the analyses after a severe

accident unrelated to the program prevented participation.

One father was dropped from analysis because he entered

another weight-control program. One mother was

dropped from analysis after she became pregnant during

follow-up, and eight families did not provide complete 6-

and 12-month follow-up data.

The final sample with complete

data for targeted children at baseline, 6-month, and

12-month measurements was based on 56 of the 67 families

that were randomized (84%), which included 245 family

members. The data that were excluded included five children

in the increase activity group (one boy and four girls)

and six children in the combined group (three boys

and three girls).

The best fitting multiple regression

model to predict changes in percentage of overweight

over the year is shown in the top of Table 3, including

predictors of weight change of adherence (better adherence,

better weight change) and predilection (higher predilection,

better weight change) for physical activity. Although child

sex was not an independent predictor when adherence and

predilection were considered in the model, this may be

because child sex was related to predilection for physical

activity (r = 0.31, p = 0.021), with boys having greater

predilection scores (34.2 ± 3.6) than girls (31.0 ± 6.2), and

to adherence to treatment as previously presented.

Table 3

At 12 months, boys showed significantly better

percentages of overweight changes (-15.8%) for the combined

treatment than girls (-1.0%), with no significant

differences for the increase intervention for boys (-9.3%)

or girls (-7.6%). Boys adhered to treatment better than girls

(p < 0.01).

There seems a discrepancy between this text from the abstract and the data from the multivariate regression model.

Shalitin, 2016

Type of study:

Retrospective cohort study

Setting and country:

The study was conducted in the obesity clinic of the

Institute of Endocrinology and Diabetes at Schneider

Children’s Medical Center, Israel, a tertiary, universityaffiliated

hospital

Funding and conflicts of interest:

Funding not mentioned/ no conflict of interest

Inclusion criteria:

Patients who had made at least three clinic visits with at least one year of follow-up were included to evaluate the short- and long-term impact of standard interventions on weight reduction.

Exclusion criteria:

The exclusion criteria were the presence of any syndrome or disease known to affect body composition or

fat distribution such as Cushing’s syndrome, hypothyroidism

or craniopharyngioma, the use of anti-obesity

medications or other medications that can affect body weight, current or prolonged steroid therapy and a history of bariatric surgery.

286 participants had a BMI above the 95th

percentile for age and gender and were eligible for the

study.

They comprised 175 girls (61.2%) and 111 boys (38.8%).

Their median age was 9.3 years (range 2.7–18.7 years) at

admission, and 61.3% were prepubertal. Mean BMI at

admission was 25.57 ± 5.67, and 32.2% had obesity-related

comorbidities.

Patients were scheduled for clinic visits every 4– 6 months.

Emphasis was placed

on the entire family’s lifestyle, and the patients received a

standard intervention in which they and one or both of

their parents were referred for consultation with a paediatric dietician who recommended age-adjusted dietary modifications.

To compare height, weight and BMI values by age and

gender, height-standard deviation score (SDS) percentile, weight-SDS percentile and BMI-SDS

percentile according to the Centers for Disease Control and

Prevention growth charts were calculated.

A stepwise logistic regression model was formulated to identify baseline predictors of successful weight reduction and maintenance. Significance was set at p ≤ 0.05.

The 110 (38.5%) of the patients with successful weight

reduction and maintenance, defined as a reduction in BMISDS

of ≥0.2 between the first and last visits, were classified

as responders and the rest of the cohort as nonresponders.

Mean BMI-SDS was significantly lower in the responders

than the nonresponders at the last visit (1.7 ± 0.63 vs.

2.09 ± 0.4, p < 0.001).

On multiple logistic regression, the stepwise results included

the following baseline characteristics: age, BMI-SDS, presence of acanthosis nigricans and previous participation in a

dietary weight-reduction programme.

During the study period, 834 patients were referred to the

clinic for obesity evaluation and 474 were followed for at

least one year. Of these, 286 had a BMI above the 95th

percentile for age and gender and were eligible for the

study.

The 110 (38.5%) of the patients with successful weight

reduction and maintenance, defined as a reduction in BMISDS

of ≥0.2 between the first and last visits, were classified

as responders and the rest of the cohort as nonresponders.

The median duration of follow-up at the clinic was 2.2 years

(mean: 2.4 ± 1.2 years).

Nothing is mentioned about incomplete data.

Those that significantly predicted successful weight reduction were

higher BMI-SDS [OR =2.4 (1.28-4.51), p = 0.004] and absence of acanthosis nigricans [OR = 2.9 (1.35 – 6.21), p = 0.007] at baseline. The area under the receiver operating characteristics curve for these parameters

was 0.667.

Reinher, 2010

Type of study:

Retrospective, multicentre study

Setting and country:

Munich, Germany

Obese children motivated to participate in the

outpatient intervention “Obeldicks” in 4 different outpatient

centers during the years 1999–2006.

Funding and conflicts of interest:

Grant support was received from the German “Competence Net

Obesity,” which is supported by the German Federal Ministry of Education

and Research (grant number 01 GI0839). There was partially support from the Munich Center of Health Sciences (LMUinnovativ) subproject II “Evidence

Based Prevention and Modelling of Chronic Diseases.” Furthermore, a grant support was received from the University of Witten/Herdecke.

None of the authors

declared a conflict of interest.

Inclusion criteria for lifestyle interventions were an age of 4–16 y, sufficient motivation

for a lifestyle change, and no BMI reduction in the past 6 mo.

A total of 663 obese children were included in the study (mean

age at baseline: 10.6 y; minimum: 4; maximum: 16 y; 55%

female). No age differences between female and male individuals

were observed (t test, P = 0.212).

Mean BMI-SDS at baseline was 2.46 (95% CI: 2.43, 2.50;

minimum: 1.90; maximum: 4.16).

A reduction in BMI-SDS was observed in 470 (71%)

children at the end of intervention (intention-to-treat analysis).

Briefly description of the 1-y intervention “Obeldicks”: The intervention was based on

physical exercise, nutrition education, and behavior therapy,

including the individual psychological care of the child and his or her family.

Multilevel growth curve models were used to estimate

the association of age and sex with individual patient BMI-SDS

measurements and how these associations vary over time

Duration or endpoint of follow-up:

5 year follow-up

A total of 116 (18%) children dropped-out during the intervention, of whom 3 children were again seen at later time

points.

Reasons for incomplete outcome data are not described.

The dropout in the intervention period did not differ significantly

with respect to age, sex, or BMI-SDS at baseline as

compared with children finishing the intervention. The children

lost to follow-up after the end of the intervention did not differ

significantly with respect to age, sex, or BMI-SDS at baseline as

well as a change in BMI-SDS in the intervention as compared

with the children with complete follow-up.

Age was significantly associated with BMI-SDS at baseline with the lowest values observed at the age of 11–12 y

and the highest BMI-SDS in children aged 4–7 y (regression with

age as dummy variable; P , 0.001; Figure 2, Table 2).

Age was significantly associated with both height of the BMI-SDS curve and the course of the BMI-SDS curve, even after adjustment for sex and baseline BMI-SDS. The older the children the lesser the BMI-SDS reduction within 2 y after intervention with a reduction of 0.58 BMI-SDS units for young children aged 4-6 y decreasing to 0.25 BMI-SDS units voor adolescents aged 13-16 y.

The efficacy of the lifestyle intervention was not assessed in the current study and similar changes in BMI-SDS in motivated children not receiving an intervention could not be ruled out.

Moens, 2010

Type of study:

8-year follow-up study of treated obese children.

Setting and country:

Belgium

Funding and conflicts of interest:

unknown

Inclusion criteria:

-between 4-17

years

-between 20% and

80% overweight

-medical clearance

from physician

-living within a radius

of 50 km from the

Clinic.

participating children (n = 90) (34 boys and 56 girls). Mean age: 10.12 years (SD = 2.56). Mean adjusted BMI of 153.12% (SD = 20.65). Mean

BMI z-score was 2.17 for boys (0.97 to 2.47) and 2.09 for girls (0.97 to 2.64). Their mothers had a mean BMI of

26.26 (SD = 4.88), fathers had a mean BMI of 27.49 (SD = 3.54). At

baseline, 87.5% of the participating families were intact. All

economic classes were represented: 20% of the participating

families were lower social class, 65% were middle class and 15%

were classified as upper class (categorized according to Hollingshead

index, 1975).

Describe prognostic factor(s) and method of measurement:

The treatment consists of a non-diet healthy lifestyle program

focusing on three major components: healthy eating habits, moderate exercise and cognitive-behavioural techniques.

Hierarchical regression analyses assessed independent

and incremental correlates of changes in weight.

Within a period of 4 years, 230 children and families sought advice, 88 children (38%) felt not within the inclusion criteria, mainly because of the distance criterium, 142 children (62%) met the inclusion criteria.

After 8 years, 101 children (71%) were retrieved, 41 children (29%) dropped out (moved or changed phone number). 90 children agreed to participate. Response rate was 89%.

The children obtained a mean reduction of 8% in adjusted BMI at the 8-year follow-up. A total of

59 children (66%) were successful in obtaining weight control (i.e. maintaining their original % adjusted

BMI); 40% even decreased their adjusted BMI by 10% or more.

Table 1

Hierarchical regression models of children’s, process and parental characteristics on

long-term weight loss:

Child’s age at baseline (β = .33), adjusted BMI at baseline (β = .38) and global self- worth (β = .30 were significant predictors.

Adding the parental characteristics to the equation produced a significant

increase of 8% in the explained variance over and above child and

process variables. This can be attributed to the significant negative contribution of maternal psychopathology

(β = -.30).

The entire model (including all predictor variables)

accounts for a significant cumulative variance of 35% showing

significant positive contributions of child’s age, adjusted BMI at

baseline, child’s global self-worth and a negative contribution of

maternal psychopathology.

Analyses revealed that the child’s age, the degree of overweight at baseline and the child’s global self-worth were positive predictors of long-term weight loss 8 years after treatment, whereas psychopathology in the mother was a negative predictor. The total explained variance was R² = 35%.

Pinhas- Hamiel, 2008

Type of study:

Retrospective chart review

Setting and country:

Tel-Aviv, Israel

Funding and conflicts of interest:

This study was funded in part by the NASH foundation clinical scholars

grant

134 adolescents between 10- 18 year above the 95^th percentile for body mass index (BMI)

Patients with secondary causes of obesity like

hypothyroidism, or who were on medications or with known eating disorders, were excluded.

The study population comprised 134 obese patients. Patients

were categorized into two groups according to the

delta BMI Z-score, 99 (73.9%) belonged to the S group (delta BMI Z ≤0) and 35 (26.1%) to the NotS group (delta BMI Z > 0).

There were no differences in severity of obesity, male-to-female ratio, mean age, or Tanner stages between the groups.

Adjusted odds ratio and 95% confidence intervals (CI) were obtained from multivariate logistic regression models with success or no success as

dependent variables, and age, gender, elevated blood pressure,

abnormal lipid profile, parental morbidity duration of

follow-up, baseline BMI Z-score, and insulin as predictors.

BMI, fasting insulin, homeostasis model assessment insulin resistance (HOMAIR).

Weight management success was defined as BMI Z-score at final exam minus BMI Z-score at

initial exam ≤0. 99 were successful (group S) (reduced or maintained BMI Z-score) and

35 were not successful (NotS) (increased BMI Z-score)

Patients must had an initial visit plus a minimum of two

follow-up visits. The prevention program included a preliminary

visit with the pediatrician where patients had medical

and family history evaluation and a complete physical

examination. In a second visit 1 month later the patient met

with a registered dietitian. He or she received a 60-minute

counseling session, during which the necessary individual

nutrition and physical activity recommendations were discussed

with the subject and family in detail. They were educated regarding the effects of dietary and lifestyle factors that either promote or reduce pediatric obesity.

At subsequent follow-up visits, the patient’s height

and weight were obtained

Table 3

Multivariate logistic regression model with family and baseline

parameters including insulin as predictors of failure

Failing to respond to standard therapeutic lifestyle change intervention was dependent

on baseline insulin resistance (OR: 3.13 (1.79-6.01, p= .0002)), and parental obesity-related comorbidity (OR: 12.6 (1.93-82.6), p=0.008)).

A higher baseline BMI Z-score was

associated with better improvement in weight status, 0.3

(95% CI 0.11– 0.72, p = .01).

In a second model HOMA-IR was studied instead of insulin.

An increase in 1 unit of HOMA-IR was associated with a 1.64-fold increased odds (95% CI 1.27–2.21, p = .0004).

All models fit well, with p-values of Hosmer and Leme-show

test from .44 to .50 and c-statistics from .802 to .804.

Reinher, 2003

Type of study:

Longitudinal, clinical intervention study

Setting and country:

Dortmunt, Germany

Funding and conflicts of interest:

unknown

A total of 75 obese children (age 7–15 y; standard duration scores of body mass index (SDS-BMI)+1.9– +3.8) taking part in the outpatient training ‘Obeldicks’ between between 1999 and 2001.

Inclusion criteria for the outpatient training ‘Obeldicks’

were a two-time presence in the obesity ambulance and to fill in a questionnaire according to exercise and dietary

habits.

The behaviour therapy is

composed of contingency contracting, self-monitoring,

praise, stimulus control techniques and problem-solving.

The exercise therapy consists of aerobic exercise, lifestyle

exercise and decreasing sedentary behaviour.

Participants were characterised as to their willingness to change their behaviour (changes in weight status, number of attempts at therapy, participation in exercise groups), somatic characteristics (BMI of children and family members, gender, age), socioeconomic status (level of education of the children and their parents, working mother), exercise and dietary habits, dietary intake as well as the quality of dietary records. These factors were related to the success of the treatment using a multiple logistic regression

A total of 27 children (37%) were

unsuccessful at the end of the training; 20 of these 27

children dropped out without finishing the training. All the

children who dropped out did not reduce their overweight

and dropped out in the first 6 months of the training.

At the end of the 1-y training, 48 children (63%) had

reduced their overweight in median by 0.4 SDS-BMI (range –

0.2 up to –1.1 SDS-BMI).

A total of 22 (96%) of the 23 children taking part in the

exercise groups before beginning of the training had success,

while one (2%) of the other 52 children who did not take

part in the exercise groups had success.

Multiple logistic regression showed that only the previous

participation in the exercise groups for the obese was an

independent factor that was associated with the success of

the training (see Table 2).

Multiple logistic regression of the dependent variable SDS-BMI

reduction (n = 75, r² = 0.45)

- Participation exercise group before training: Coefficient +0.41 (+0.29 - +0.53, p <0.0001)

- BMI, mother: Coefficient 0.00 (-0.01-+0.01, p=0.99)

- Single-parent family: Coefficient +0.05 (-0.12-+0.22, p=0.57)

- Number of therapies: Coefficient -0.04 (-0.10-+0.02, p+0.23)

A training for the obese children and their parents enables the majority of the participants to reduce weight.

Previous participation in exercise groups can be considered as a predictor to success of the treatment.

Mameli, 2017

Type of study:

Intervention study

Setting and country:

Milan, Italy

Funding and conflicts of interest:

The authors received no specific funding and have declared no conflict of interest.

Inclusion criteria:

From January 1st 2006

to December 31th 2016, patients with overweight and obesity aged between 2 and 18 years old

were invited by their family paediatrician to be evaluated in the Clinic

N= 864 patients entered the intervention. 453 patients (208 males), mean age 11.2 ±3.1 years,

392 with obesity (86%, z-BMI 2.90 ±0.80 SD) and 61 patients with overweight (z-BMI

1.73±0.21 SD) attended at least 1 follow-up visit.

Univariate and multivariate linear regression modelling to evaluate the association of demographic and anthropometric variables with weight loss on follow-up.

Predictand was taken to be weight change expressed as change in BMI z score from T0 visit to last

follow-up visit.

Predictor variables included demographics (age, sex) and anthropometrics

(BMI, normalized waist and hip circumferences).

Predictors found to be significantly associated (p < 0.05) with change in BMI z

score in the univariate regressions were included in a multivariate linear regression

model.

The proportion of subjects who made a recovery from obese to overweight was 19.8%

(78/392 subjects) and from obese to normal weight was 1.2% (5/392 subjects). The recovery

from overweight to normal weight was 13.1% (8/61 subjects)

325/864 (37%) refused to be included in the program. Reasons given by parents included lack

of interest in the program (102 patients, 31%), too much time required for follow-up visits

(87 patients, 27%) and absence of agreement of both parents for program participation (80 patients,

24%). 56 patients (17%) preferred to not give any explanation.

Out of the 539 patients who attended at least 1 follow-up visit, 453 subjects (208 males, 245

female) with all predictor values recorded at the baseline visit and height and weight recorded

at the last visit were considered for the regression analysis.

In a multivariate model, initial BMI z score and number of follow-up visits were significant predictors

of weight change, while age, sex, ABSI, and HI were not significant predictors.

Results multivariate model, predictors of change in BMI z score on follow-up

Predictor: [Regression coefficients (95% confidence intervals), p-value]

Age (y): [-0.013 (-0.029 - +0.002), p=0.098]

z-BMI: [-0.220 (-0.276 - -0.164), p=8×10⁻¹⁴]

Follow-up duration (y): [-0.051 (-0.113 -+0.010), p=0.103]

Number of visits: [-0.040 (-0.069 - -0.011), p= 0.008]

Patients affected by overweight and obesity involved in a multidisciplinary weight loss intervention

reduced their mean BMI z score, while ABSI and HI were stable. Weight loss was

not predicted by initial ABSI or HI. More visits predict more weight loss, but dropout rates are

high. The great majority of patients leave the weight management program before having normalized their BMI.

It is not clear whether the results of this study could be linked to a successful life style intervention.

Goossens, 2009

Intervention study

Gent, Belgium

The work was supported by the Special Research Funds of Ghent University.

Nothing is mentioned about conflict of interest.

Participants were 132 overweight youngsters (8–18 years;

M=13.58; SD=2.15) who were admitted to a 10-month inpatient

treatment program for obesity. Exclusion criteria were mental

retardation and autism.

At baseline, all participants were diagnosed

as being primary overweight (BMI>85th percentile or adjusted

BMI>120%).

Hierarchical linear and binary logistic regression analyses were

conducted to examine the relationship between predictors and outcome

measures. Following predictors were considered:

weight parameters

[baseline adjBMI, 1 month weight loss and 4 months weight loss],

patient characteristics, [gender, age and SES] and

eating pathology variables [SBE (subjective binge episode), OBE (objective binge episode), cognitive restraint, eating, weight and shape concern;

emotional, external eating and restrained eating style].

10-month inpatient treatment program

Eleven youngsters (8.3%) left the treatment earlier

against medical/psychological advice and are considered as drop outs.

When considering total percentage of adjusted BMI change as outcome, the explained variance of the whole model was 69%, whereby 24% was

explained by baseline adjusted BMI, 3% by the patient characteristics, 6% by the eating pathology variables and 36% by the two parameters of early

weight loss (weight loss after 1 and 4 months of treatment).

Baseline adjusted BMI, restrained eating style and weight loss after 4 months of treatment positively predicted weight loss at the end, whereas larger weight loss after 1 month predicted less total weight loss.

When considering drop-out as outcome, the total explained variance of this model was 64%. Lower baseline adjusted BMI was associated with more drop

out.

SBE had a positive effect on program completion

See Table 2 in paper

Study reference

(first author, year of publication)

Study participation¹

Study sample represents the population of interest on key characteristics?

(high/moderate/low risk of selection bias)

Study Attrition²

Loss to follow-up not associated with key characteristics (i.e., the study data adequately represent the sample)?

(high/moderate/low risk of attrition bias)

Prognostic factor measurement³

Was the PF of interest defined and adequately measured?

(high/moderate/low risk of measurement bias related to PF)

Outcome measurement³

Was the outcome of interest defined and adequately measured?

(high/moderate/low risk of measurement bias related to outcome)

Study confounding⁴

Important potential confounders are appropriately accounted for?

(high/moderate/low risk of bias due to confounding)

Statistical Analysis and Reporting⁵

Statistical analysis appropriate for the design of the study?

(high/moderate/low risk of bias due to statistical analysis)

Study reference

Goldschmidt, 2014

low

moderate

low

low

moderate

low

Baxter, 2013

moderate

high

low

low

moderate

low

Frohlich, 2011

low

high

low

moderate

moderate

low

Wrotniak, 2004

low

high

moderate

moderate

moderate

low

Epstein, 2001

high

low

moderate

moderate

moderate

moderate

Shalitin, 2016

low

moderate

moderate

low

moderate

moderate

Reinher, 2010

high

high

moderate

low

moderate

low

Mameli, 2017

high

high

moderate

low

moderate

low

Goossens, 2009

high

moderate

moderate

high

moderate

moderate