Intervention(s) and comparator(s)

Sample size^a

Screened/eligible
[N]

Randomised
[N]

ITT
[N]

Analysed
[N]

Finishing trial
[N]

Randomised finishing trial
[%]

Follow‐up
(extended follow‐up)^a

(20) Resnicow 2015

I1: parent‐only PCP motivational interviewing

The study was powered to detect a 3‐point difference in BMI percentile between any pair of study groups at 2‐year follow‐up, with an assumed SD for BMI percentile between 4 and 6: power of 0.80 and 2‐tailed a of 0.05. Sample size was inflated to account for practice‐level clustering, assuming an intraclass correlation between 0.01 and 0.05. On this basis and a projected 25‐30% attrition at 2‐year follow‐up, 10‐12 practices per arm (30‐36 total) and a mean of 15‐20 children per practice at baseline were required

‐

16 practices

212 participants

145

145

145

68

2 years (2 years)

I2: parent‐only PCP + dietician motivational interviewing

15 practices

235 participants

154

154

154

66

C: usual care

11 practices

198 participants

158

158

158

80

total:

645

457

457

457

71

(19) Mazzeo 2014

I: parent NOURISH

‐

235

48

‐

46

10

21

12 weeks post 12‐week intervention (24 weeks)

C: parent control

45

‐

45

16

36

total:

93

91

26

28

(18) Van Grieken 2013

I: parent‐only

Sample size was calculated taking into account the intracluster correlation coefficient (ρ = 0.1), the number of clusters (44), the expected prevalence of overweight children in the study population, the SD, expected effect (a difference in mean), and the power of the study (80%). With a participation of 50%, an expected prevalence of overweight children of 9% and a loss‐to‐follow‐up of 30%, at least 11,301 children (and their parents) should be invited by the YHC teams to participate in the study to have a final sample of about 356 overweight children (178 in both the intervention and control group). Assuming a SD of BMI to be 1.0 kg/m², a difference in mean BMI of 0.35 kg/m² between the children in the intervention group and the children in the control group can be established under the assumptions mentioned above

22 clusters

7004 participants

349

‐

21 clusters

277 participants

277

79

2 years post up to 12‐month intervention

C: usual care

22 clusters

7004 participants

288

‐

21 clusters

230 participants

230

80

total

637

‐

42 (507)

507

80

(17) Small 2013

I: parent‐only

‐

‐

34

33

33

33

97

24 weeks post 16‐week intervention (41 weeks)

C: parent control

33

27

27

27

82

total:

67

60

60

60

90

(16) Esfarjani 2013

I: parent‐only

‐

550/156

70

‐

55

58

83

Intervention 6 months (not reported)

C: parent control

86

‐

52

59

69

total:

156

‐

107

117

75

(15) Moens 2012

I: parent‐only

‐

80/75

31

‐

‐

31

100

Immediately following 6‐month intervention

C: waiting list control

19

‐

‐

15

79

total:

50

‐

‐

46

92

(14) Raynor 2012a

I1: parent‐only

Sample size calculations presumed 2‐sided hypothesis testing at 6‐month assessment, with type 1 error rate = 0.05. To reject with 80% power the null hypothesis of no pre‐ to post‐treatment difference between intervention conditions vs. the alternative that the pre‐ to post‐treatment difference was 0.6 or greater (effect size), 24 participants per group were needed

549

33

33

33

29

88

24 weeks post 24‐week intervention (reported as '12 months')

I2: parent ‐ diet decrease

33

33

33

29

88

I3: parent ‐ diet increase

35

35

35

32

91

total:

101

101

101

90

89

(13) Raynor 2012b

I1: parent‐only

Sample size calculations presumed 2‐sided hypothesis testing at 6‐month assessment, with type 1 error rate = 0.05. To reject with 80% power the null hypothesis of no pre‐ to post‐treatment difference between intervention conditions vs. the alternative that the pre‐ to post‐treatment difference is 0.6 or greater (effect size), 24 participants per group were needed

549

29

29

29

26

90

24 weeks post 24‐week intervention (reported as '12 months')

I2: parent ‐ diet and activity traditional

26

26

26

24

92

I3: parent ‐ diet and activity substitute

26

26

26

24

92

total:

81

81

81

74

91

(12) Margarey 2011

I: parent healthy lifestyle

Sample size calculation was based on a reduction in BMI z score of 0.26 (SD 0.49) over 12 months (power 80%, alpha = 0.05, and drop‐out rate of 30%). This represents a 50% reduction in weight velocity over 12 months and no change in height velocity. We sought 42 children per group per site (168 children)

398

85

85

85

52

61

80 weeks post 24‐week intervention (104 weeks)

C: healthy lifestyle

84

84

84

54

64

total:

169

169

169

106

63

(11) Jansen 2011

I: parent CBT

‐

161

59

‐

54

54

92

12 weeks post 12‐week intervention (24 weeks)

C: waiting list control

39

‐

34

34

87

total:

98

88

88

90

(10) Collins 2011

I: parent‐only ‐ diet

Power: 80% chance of detecting significance (2‐sided 5% level), with a 0.26 BMI z score difference from baseline to 12 months as the initial end point, with an anticipated loss to follow‐up of 20%

505/319

63

‐

42

22

35

80 weeks post 24‐week intervention (104 weeks)

C1: parent‐child (physical activity)

73

‐

63

35

48

C2: parent‐child (physical activity + diet)

70

‐

60

36

51

total:

206

‐

165

93

45

(9) Boutelle 2011

I: parent‐only

Sample size was determined by pragmatic factors, including budget and investigator time commitments. No interim analyses were done. The hypotheses tested related to non‐inferiority of the parent treatment to the parent‐child treatment on child and parent weight loss and child daily caloric intake and physical activity. The bound for non‐inferiority hypotheses related to BMI percentile was set to 1. This is the maximum value the parent‐child group could do better than parent‐only, below which non‐inferiority would be concluded. This bound could correspond to an mean‐aged child in this sample having a BMI of 26 in the parent‐child group and 28.5 in the parent‐only group at post‐treatment/follow‐up, assuming equivalence at baseline. For a non‐inferiority bound for child BMI, which was selected post hoc, we considered choosing a BMI that would correspond to the BMI percentile non‐inferiority bound (BMI = 2.5), but instead chose a more rigorous value of BMI = 1

157

40

‐

24

24

60

24 weeks post 20‐week intervention (week 44)

C: parent‐child

40

‐

28

28

70

total:

80

52

52

65

(8) West 2010

I: parent‐only

‐

205

52

52

52

34

65

40 weeks post 12‐week intervention (52 weeks)

C: waiting list control

49

49

49

46

94

total:

101

101

101

80

79

(7) Resnick 2009

I: educational material + personal encounters

‐

84/46

22

‐

18

18

82

Unclear (41 weeks between start and last mail out)

C: educational material

24

‐

24

24

100

total:

46

42

42

91

(6) Estabrooks 2009

I1: parent group + IVR

Sample size calculations were completed, varying the detectable effect sizes from small to medium with a power of 0.8. The result was a need for 42 participants per intervention to detect a medium effect and 64 participants to detect a small effect

1487/656

85

‐

63

63

74

28‐40 weeks post 12‐ to 24‐week intervention (52 weeks)

I2: parent group

85

‐

56

56

66

C: parent workbook

50

‐

36

36

72

total:

220

155

155

70

(5) Munsch 2008

I: mother‐only CBT

Trial authors did not reach the necessary sample size of 68 families with obese children within the given time span (the target sample size of 68 was based on a repeated‐measures analysis with alpha = 0.05, 1 ‐ beta = 0.8, and a medium effect size for the linear term of the interaction between treatment and time, assuming a drop‐out rate of 20%

181/60

25

‐

7

7

28

24 weeks post 10‐week intervention (34 weeks)

C: mother‐child CBT

31

‐

20

20

65

total:

56

27

27

48

(4) Janicke 2008

I: parent‐only

Post hoc power analyses were used to determine the detectable change in BMI z score from 0 to 10 months for the family based and parent‐only interventions relative to the waiting list control condition. Effect sizes (standardised BMI index) detectable with 80% power and 2‐sided level 0.05 tests were used. Standard deviations and sample sizes were set equal to their observed values. For comparing the family‐based and waiting list control conditions, trial authors reported 80% power to detect a shift from 0.022 to ‐0.145. For comparing the parent‐only and waiting list control conditions, trial authors reported 80% power to detect a shift from 0.022 to ‐0.135

111

34

‐

26

26

76

14 weeks post 16‐week intervention (40 weeks)

C1: parent‐child

33

‐

24

24

73

C2: waiting list control

26

‐

21

21

81

total:

93

71

71

76

(3) Golley 2007

I: parent intervention + lifestyle education

Sample size calculation was based on a fall in BMI z score reflecting a weight gain of only 50% of that expected over 12 months with normal growth. A sample size of 28 per group was estimated to have 80% power to detect a 12‐month fall in mean BMI z score from a baseline of 0.26 (SD 0.49), assuming no change in the control group, at a 2‐sided significance level of 0.05. To account for a drop‐out rate of up to one‐third (commonly 20‐50% in child weight‐management studies), 42 children per study group were sought (126 children)

262/115

38

‐

31

31

82

24 weeks post 24‐week intervention (48 weeks)

C1: parent intervention

37

‐

29

29

78

C2: waiting list control

36

‐

31

31

86

total:

111

101

101

‐

(2) Golan 2006

I: parent‐only

The study was designed to detect differences of 10% weight loss with a power of 90% and a significance level of 0.05, given a drop‐out rate of 10% with a sample of 12 in each group

102

14

‐

10

10

71

1 year post 26‐week intervention (18 months)

C: parent‐child

18

‐

17

17

94

total:

32

27

27

84

(1) Aragona 1975

I1: parent‐only with reinforcement

‐

‐

5

‐

4

4

80

12 weeks (51 weeks' follow‐up)

I2: parent‐only

5

‐

3

3

60

C: waiting list control

5

‐

5

2

40

total:

15

12

9

60

Grand total

All interventions

‐

1773^c

1276

All comparators

1284^c

942

All interventions and comparators

3057^c

2218