Figure 1 provides an overview of the participant flow within the trial. In total, 38 out of 50 (76%) randomized families completed the intervention. Twelve EG families withdrew after randomization, mostly for reasons of time (8) or because the child was affected by a psychiatric disorder (1). Some did not give reasons (2) or could not be contacted (1). One family (EG) discontinued the study after the intervention was completed because they had moved. Four families in the CG discontinued the study after randomization for unknown reasons (3), or because of marital disagreement about the risks of including their children in the study (1). Two families (CG) dropped out during the intervention period because the child no longer wanted to complete the questionnaires (1) or for unknown reasons (1). To our knowledge no severe adverse event and withdrawal of patients due to an adverse event occurred[2].
Missing data
Not all children completed all measures at all time points (see Figure 1). 118/135 (87.41%) provided data on at least one of the symptom measures at least once across the four time points. 66/135 (48.89%), 30 EG, provided data at all four time points. 16/135 (11.85%), 9 EG, provided data at three time points. 8/135 (5.93%), 4 EG, provided data at two time points. 27/135 (20.0%), 15 EG, provided data at one time point. The amount of missing data did not vary significantly between groups at any of the time points[3]. Only two outcome variables showed a correlation with the amount of missing values at T1: the FEEL-KJ subscales acceptance (r = 0.27) and reappraisal (r = 0.24). Since the outcome of MLMs for these variables was unchanged whether or not the number of completed assessments was controlled for we report estimates without the control.
Hypothesis 1: Treatment effects on depression onset
Data were available for 31 children (44.93%) in the CG and 32 (48.48%) in the EG. The frequencies of any psychiatric disorder were so low in both the EG (n = 2; 6% of those who provided data) and CG (n = 5; 16% of those who provided data) that the planned statistical analyses of these data were not possible[4]. Two children in the CG, versus no children in the EG, met criteria for depression at T4.
Hypothesis 2: Intervention effects on symptom severity
Table 2 reports descriptive statistics for the secondary outcome variables internalizing and externalizing symptoms (self- and parent-report) and symptoms of depression (self-report) across the two groups and all four time points.
Child-reported internalizing and externalizing symptoms (YSR).
As depicted in Figure 2 there were significant interactions between group and time for self-reported internalizing symptoms from T1 to T3 (estimate = -0.42, SE = 0.19, t = -2.22, p = .027, d = -0.45, 95% CI [-0.84, -0.06]) as well as from T1 to T4 (estimate = -0.41, SE = 0.19, t = -2.12, p = .035, d = -0.44, 95% CI [-0.84, -0.05]) but not from T1 to T2 (p > .05). Simple slope (post-hoc) analyses revealed a significant increase from T1 to T3 in the CG (estimate = 0.36, SE = 0.14, t = 2.55, p = .012, d = 0.38, 95% CI [0.09, 0.67]) but no change in the EG (estimate = -0.06, SE = 0.13, t = -0.46, p = .647, d = -0.06, 95% CI [-0.32, 0.20]). From T1 to T4 the CG showed a non-significant increase (estimate = 0.18, SE = 0.14, t = 1.27, p = .208, d = 0.19, 95% CI [-0.09, 0.48]) whereas the EG showed a non-significant decrease over time (estimate = -0.22, SE = 0.13, t = -1.77, p = .080, d = -0.25, 95% CI [-0.51, 0.01]). No significant effects were found for self-reported externalizing symptoms (all ps > .05).
Self-reported depressive symptoms (DIKJ).
Both groups showed significant reductions from T1 to T3 (estimate = -0.30, SE = 0.14, t = -2.19, p = .029, d = -0.44, 95% CI [-0.64, -0.24]) and from T1 to T4 (estimate = -0.39, SE
Table 2 Descriptive Statistics of Internalising, Externalising and Depressive Symptoms for Measurement Points T1, T2, T3 and T4
Variables
|
Experimental Group (n, M, SD)
|
Control Group (n, M, SD)
|
T1
|
T2
|
T3
|
T4
|
T1
|
T2
|
T3
|
T4
|
Self-report symptoms
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Internalising (YSR)
|
54
|
9.00
(7.69)
|
42
|
7.38
(7.82)
|
40
|
7.95
(8.03)
|
40
|
5.95
(5.97)
|
40
|
6.75
(6.20)
|
31
|
8.29
(7.4)
|
35
|
10.38
(9.17)
|
32
|
8.82
(9.12)
|
Externalising (YSR)
|
54
|
8.85
(5.51)
|
42
|
8.52
(6.25)
|
40
|
7.63
(5.57)
|
40
|
7.65
(5.97)
|
40
|
8.82
(6.86)
|
31
|
9.84
(6.36)
|
35
|
8.91
(6.91)
|
32
|
8.97
(6.86)
|
Depression (DIKJ)
|
52
|
7.83
(5.85)
|
34
|
8.59
(8.19)
|
35
|
6.86
(7.67)
|
39
|
5.64
(5.27)
|
41
|
7.56
(4.91)
|
28
|
6.46
(4.88)
|
29
|
5.38
(4.19)
|
35
|
5.60
(5.98)
|
Parent-report symptoms
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Internalising (YSR)
|
56
|
7.96
(6.46)
|
38
|
7.34
(8.35)
|
41
|
6.00
(6.48)
|
39
|
5.05
(5.63)
|
43
|
7.05
(6.09)
|
33
|
6.55
(7.06)
|
32
|
6.03
(4.80)
|
37
|
6.30
(5.52)
|
Externalising (CBCL)
|
56
|
6.93
(6.47)
|
38
|
6.11
(5.94)
|
41
|
5.15
(4.98)
|
39
|
4.69
(5.42)
|
43
|
4.60
(5.70)
|
33
|
3.52
(3.86)
|
32
|
4.00
(4.12)
|
37
|
3.30
(3.19)
|
Note CBCL = Child Behavior Checklist; DIKJ = Depressions-Inventar für Kinder und Jugendliche; YSR = Youth Self-Report
= 0.14, t = -2.75, p = .006, d = -0.57, 95% CI [-0.77, -0.37]), but not from T1-T2. The groups did not differ between each other in changes across time (p > .05).
Parent-reported internalizing and externalizing symptoms (CBCL).
There was no evidence of group differences in changes in internalizing or externalizing symptoms across time (all ps > .05).
Hypothesis 3: Treatment effects on the potential mediators
3a. Group-dependent change in potential mediators.
Descriptive data for the variables showing group-dependent change (Parenting Style and ER) are shown in Table 3. Descriptive data for the other variables (attributional style, knowledge of depression) are reported in Supplementary Table S2.
MLM revealed a significant group by time interaction for negative (but not positive) parenting style (ESI) between T1 and T3 (estimate: -6.36, SE = 2.15, t = -2.96, p = .004, d = -0.44, 95% CI [-0.82, -0.05]). This reflected a decrease in negative parenting in the EG (estimate = -4.43, SE = 1.48, t = -3.00, p = .003, d = -0.28, 95% CI [-0.55, -0.02]) and no change in the CG (estimate = 1.93, SE = 1.57, t = 1.23, p = .22, d = 0.15, 95% CI[-0.13, 0.42]). There were no group differences at any other time points for negative or positive parenting style.
Significant group by time interactions were found for one adaptive ER strategy (cognitive problem solving: Supplemental Figure S1)[5] and two maladaptive strategies (self-devaluation: Supplemental Figure S2 and aggressive actions: Supplemental Figure S3).
The group by time interaction for cognitive problem-solving occurred from T1 to T2 (estimate = -2.63, SE = 0.99, t = -2.66, p = .010, d = -0.46, 95% CI [-0.84, 0.08]-0.07;-0.84) and reflected a non-significant reduction in the EG (estimate = -0.88, SE = 0.66, t = -1.33, p = .187, d = -0.16; 95% CI [-0.42, 0.09]) and a significant increase in the CG (estimate = 1.75, SE = 0.74, t = 2.38, p = .019, d = 0.29; 95% CI [0.01, 0.56]). The group by time interaction in self-devaluation occurred from T1-T3 (estimate = -2.71, SE = 1.05, t = -2.58, p = .010, d = -0.53, 95% CI [-0.91, -0.15]) and reflected an increase in the CG (T1 to T3; estimate = 2.89, SE = 0.77, t = 3.75, p < .001, d = -0.61, 95% CI [0.31, 0.91]) and no change in the EG (p > .05). The significant interaction in aggressive behaviour occurred from T1 to T2 (estimate = 2.59, SE = 0.88, t = 2.93, p < .001, d = 0.66, 95% CI [0.27, 1.04]) and reflected an increase in the EG (estimate = 1.81, SE = 0.59, t = 3.07, p < .001, d = 0.50, 95% CI [0.23, 0.77]) and no change in the CG (p > .05).
Children in both groups showed increases in their knowledge of depression over time (T1 to T2: estimate = 1.51, SE = 0.62, t = 2.41, p = .002, d = 0.38, 95% CI [0.18, 0.58]; T1 to T3: estimate = 1.60, SE = 0.82, t = 1.95, p = .005, d = 0.40, 95% CI [0.20, 0.60]; T1 to T4: estimate = 1.96, SE = 0.62, t = 3.17, p < .001, d = 0.49, 95% CI [0.29, 0.70]), but groups did not differ from each other. There was no evidence of group by time interactions in any of the attributional style (ASF) subscales (all p > .05).
3b. Mediation models.
We estimated the dual simplex models with the candidate variables from phase 3a ER (cognitive problem solving, self-devaluation, aggressive action) and negative parenting (ESI) as mediators and the five symptom measures as outcomes[6]. There were no significant paths found for child- (YSR) or parent-report (CBCL) internalizing or externalizing symptoms. Although in the MLM analysis (hypothesis 2) there was no direct effect on the intervention of depressive symptoms (DIKJ), in the path models the ER strategy cognitive problem-solving (but not self-devaluation or aggressive actions) and the negative parenting subscale of the ESI revealed significant lagged effects on depressive symptoms (DIKJ). This suggests indirect effects of the intervention over time (see estimated path coefficients in Figures 3 and 4). The intervention had a significant effect on the mediator at T2, and the mediator has a further effect on the outcome at T3.
Table 3 Descriptive Data on Changes in Emotion Regulation and Parenting Style in Both Groups Across all Four Time Points
|
Experimental Group (n, M, SD)
|
Control Group (n, M, SD)
|
|
T1
|
|
T2
|
|
T3
|
|
T4
|
|
T1
|
|
T2
|
|
T3
|
|
T4
|
|
Adaptive strategies (FEEL-KJ)
|
57
|
129.33
(31.19)
|
43
|
134.37
(33.78)
|
40
|
141.65
(34.67)
|
40
|
138.40
(35.04)
|
40
|
132.96
(28.48)
|
35
|
142.63
(27.31)
|
38
|
139.27
(24.39)
|
38
|
141.39
(30.94)
|
Problem-solving
|
57
|
19.58
(5.00)
|
43
|
20.23
(5.24)
|
40
|
21.53
(5.26)
|
40
|
21.13
(5.71)
|
48
|
19.83
(5.35)
|
35
|
22.26
(5.41)
|
37
|
20.86
(5.30)
|
38
|
21.37
(5.6)
|
Distraction
|
57
|
19.35
(5.99)
|
43
|
20.56
(6.22)
|
40
|
21.00
(6.07)
|
40
|
20.88
(5.74)
|
48
|
19.43
(5.36)
|
35
|
20.86
(4.91)
|
37
|
19.16
(5.97)
|
38
|
20.05
(5.20)
|
Positive thinking
|
57
|
18.77
(5.95)
|
43
|
20.09
(5.69)
|
40
|
21.35
(5.82)
|
40
|
20.15
(6.17)
|
48
|
19.20
(5.76)
|
35
|
19.91
(5.83)
|
37
|
18.38
(6.12)
|
38
|
20.34
(5.21)
|
Acceptance
|
57
|
17.77
(6.18)
|
43
|
18.81
(6.34)
|
40
|
20.60
(5.94)
|
40
|
20.35
(6.15)
|
48
|
19.60
(5.57)
|
35
|
20.00
(4.58)
|
37
|
21.29
(3.76)
|
38
|
20.68
(5.56)
|
Forgetting
|
57
|
19.42
(4.98)
|
43
|
19.07
(5.09)
|
40
|
19.80
(5.1)
|
40
|
19.05
(5.32)
|
48
|
19.16
(4.66)
|
35
|
20.17
(4.46)
|
37
|
19.91
(3.90)
|
38
|
19.47
(4.86)
|
Reappraisal
|
57
|
20.07
(5.31)
|
43
|
19.53
(5.28)
|
40
|
20.45
(6.1)
|
40
|
20.10
(6.11)
|
48
|
19.75
(6.11)
|
35
|
22.00
(5.19)
|
37
|
121.18
(5.34)
|
38
|
20.76
(5.75)
|
Cognitive problem-solving
|
57
|
14.37
(5.33)
|
43
|
16.00
(5.31)
|
40
|
17.05
(5.86)
|
40
|
16.78
(5.84)
|
48
|
15.95
(4.67)
|
35
|
17.43
(5.20)
|
38
|
18.00
(3.89)
|
38
|
17.39
(5.04)
|
Maladaptive strategies (FEEL-KJ)
|
57
|
69.35
(13.65)
|
43
|
71.63
(17.15)
|
40
|
71.43
(17.85)
|
40
|
67.58
(15.41)
|
48
|
69.68
(15.78)
|
35
|
72.03
(15.04)
|
37
|
76.13
(17.97)
|
38
|
70.45
(17.14)
|
Giving up
|
57
|
13.89
(3.99)
|
43
|
13.93
(5.44)
|
40
|
14.53
(5.44)
|
40
|
12.85
(4.63)
|
48
|
13.06
(4.73)
|
35
|
13.11
(4.44)
|
37
|
14.45
(5.4)
|
38
|
13.61
(4.87)
|
Aggressive actions
|
57
|
10.02
(3.62)
|
43
|
11.93
(4.04)
|
40
|
10.80
(3.50)
|
40
|
10.33
(3.56)
|
48
|
12.89
(4.63)
|
35
|
10.77
(4.06)
|
37
|
14.48
(5.47)
|
38
|
10.55
(3.64)
|
Withdrawal
|
57
|
14.58
(4.39)
|
43
|
14.88
(5.36)
|
40
|
14.85
(5.81)
|
40
|
13.55
(4.67)
|
48
|
14.87
(4.80)
|
35
|
15.31
(4.8)
|
37
|
16.51
(4.63)
|
38
|
15.26
(4.64)
|
Self-devaluation
|
57
|
13.68
(5.39)
|
43
|
13.67
(4.89)
|
40
|
13.68
(5.01)
|
40
|
13.58
(5.49)
|
48
|
13.68
(4.71)
|
35
|
15.09
(5.2)
|
37
|
16.94
(6.07)
|
38
|
14.95
(5.65)
|
Rumination
|
57
|
17.18
(5.56)
|
43
|
17.21
(4.42)
|
40
|
17.53
(5.01)
|
40
|
17.28
(4.86)
|
48
|
16.95
(4.40)
|
35
|
17.74
(5.13)
|
37
|
16.81
(5.72)
|
38
|
16.08
(5.43)
|
Positive Parenting (ESI)
|
55
|
74.71 (11.69)
|
41
|
74.51 (14.70)
|
41
|
73.21 (17.40)
|
40
|
69.65 (19.30)
|
45
|
70.51 (12.51)
|
34
|
72.00 (16.66)
|
37
|
65.27 (16.70)
|
39
|
67.95 (17.17)
|
Negative Parenting (ESI)
|
54
|
67.50 (15.56)
|
42
|
64.62 (14.25)
|
39
|
60.82 (11.62)
|
40
|
65.35 (15.91)
|
45
|
67.11 (13.30)
|
33
|
67.70 (12.33)
|
37
|
69.35 (14.52)
|
39
|
69.85 (17.98)
|
Cognitive problem-solving (Figure 3). The negative effect on cognitive problem-solving at T2 reflects the significant increase in cognitive problem-solving from T1-T2 for the CG but not EG. The positive effect of the intervention on cognitive problem-solving at T3 reflects a greater increase in cognitive problem-solving in the EG (versus CG) from T2 to T3 (not directly tested in Hypothesis 3a). The negative “b” paths imply that increases in cognitive problem-solving were associated with decreases in depressive symptomology.
Negative parenting style (Figure 4). Although the individual effects of the intervention on negative parenting at T2 and T3 were not significant, the sum of their effects was (estimate: -.20, SE = .07, t = -2.75, p = .006, 95% CI [0.345, -0.058]). The negative correlation is consistent with the MLM findings (Hypothesis 3a), suggesting the intervention lead to reduced levels of negative parenting. The positive “b” paths indicate that reductions in negative parenting were correlated with reductions in depressive symptoms.
Aim 4: Moderating role of parental depression and stressful life events
Stressful life events
Data on the number and impact of stressful life events (CASE; T4) were available for 43 participants. There was no evidence that number or impact of negative life events (CASE; T4) moderated group-related changes (group x time interaction) in children’s depressive symptoms (DIKJ), or self-reported internalizing or externalizing symptoms (YSR) over time (all ps > 0.05). There was some evidence that the interaction between group (EG, CG) and time (T1 to T3) on parent-reported internalizing symptoms (CBCL) was moderated by the number of negative life events (estimate = 4.12, SE = 1.46, t = 2.83, p = .006). The positive values reflect the fact that EG children who experienced fewer negative life events showed a significant reduction in symptoms (simple slope = -4.878, SE = 1.46, t = -3.34, p = 0.001) whereas those with more life events showed no change (p > .05). The CG was unaffected by the number of negative life events. There was no evidence that the impact of negative life events moderated the intervention effects (the model did not converge).
Parental depression
Data on the severity of parental depression (BDI-II; T4) were available for 90 parents. There was no evidence that the severity of parental depression or diagnostic status (current versus past depression) at T1 moderated group-related changes (group x time interaction) on any of the five outcome variables (all ps > 0.05).
Aim 5: Effects of the intervention on parental depression
Table 4 shows changes in parental depression (BDI-II) from T1-T4 in both groups (EG, CG). MLM with parental depression (BDI-II) as the outcome revealed that both groups show decreased values from T1 to T4 (e = -1.38, SE = 0.63, df = 72.94, t = -2.19, p = .031), but there was no significance difference between EG and CG (e = -0.97, SE = 0.87, df = 72.43, t = -1,11, p = .269). There was no evidence that parental depression changed from T1 to T2 or T3.
Table 4 Changes in Parental Depression from T1 to T4 in Both Groups
Experimental Group (n, M, SD)
|
Control group
|
T1
|
T2
|
T3
|
T4
|
T1
|
T2
|
T3
|
T4
|
22
16.64
(9.09)
|
22
12.5
(11.77)
|
22
10.91
(9.22)
|
22
10.05
(9.39)
|
18
21.83
(12.26)
|
18
14.34
(12.11)
|
18
13.89
(11.27)
|
18
14.78
(12.85)
|
[2] Some families reported to feel burdened by the numerous questionnaires they were asked to fill in.
[3] Between-groups T-tests on data from the oldest child in each family revealed no significant differences at T1 (t1,99=0.06; p=0.415), T2 (t1,99=0.17; p=0.730), T3 (t1,99=-1.95; p=0.071) or T4 (t1,99=0.27; p=0.786).
[4] We exploratorily investigated whether the frequency of sub-clinical symptoms (mild or elevated) of psychiatric disorder varied between the groups, but again numbers were relatively low (EG = 7, CG = 7). Descriptively we note that anxiety disorders (n =5) were more common than depression (n = 2).
[5] There was some evidence that positive thinking increased in the EG from T1 to T3 (estimate = 2.26, SE = 0.88, t = 2.56, p = .012, d = 0.38) and remained the same in the CG (estimate = -0.13, SE = 0.95, t = -0.14, p = .891, d = - 0.29). However, this difference was not statistically significant (estimate: 2.39, SE = 1.30, t = 1.84, p = .067, d = 0.41). There were no significant effects on the other emotion regulation strategies (all ps >.05).
[6] In some models, variances of latent variables did not converge to positive values. This seems to be due to imbalance of the dataset (i.e., missing values) – therefore, we excluded individuals who did not complete the baseline questionnaires; leaving n = 87 for these mediation analyses.