Practitioner points

• There is a broad spectrum of different therapeutic approaches to involve the family/caregiver in psychotherapy.
• Family involvement can achieve improved treatment outcomes for children and adolescents with depression.
• Practitioners should routinely involve family/caregivers in psychotherapy.

BACKGROUND

Interest has been growing over the last 30 years in research on the epidemiology, aetiology, and treatment of children and adolescents with depression (Weersing et al., 2017). Worldwide prevalence rates for any depressive disorder in children and adolescents vary from about 1.7% (Finsaas et al., 2018; Ghandour et al., 2019) to 2.8% in preschool-aged children (Polanczyk et al., 2015), and between 6.1% (Ghandour et al., 2019) and 17.3% (Ormel et al., 2021) in adolescents. Note that overall prevalence rates in adolescence resemble those in adulthood, while those for children are much lower. Even so, depressive symptoms or major depressive disorder (MDD) in childhood or adolescence tend to follow a chronic course or reappear in adulthood (Birmaher et al., 1996; Copeland et al., 2009; Rutter et al., 2006). Early onset depressive episodes are associated with more severe and chronic episodes of depression in later life phases (Johnson et al., 2018; Köhler et al., 2019), lower quality of life, and with higher levels of comorbidities associated with other medical and mental disorders (Zisook et al., 2007). Depression affects children and adolescents in transition stages especially during their lives, and has a deep impact on their general quality of life (Murray et al., 2013). We thus need to effectively treat depression in this age group via early and evidence-based interventions (Johnson et al., 2018).

Generally, effects of psychotherapy are positive for children and adolescents (Weisz et al., 2017), with an overall moderate effect size (g) of 0.46 that drops to 0.36 in follow-up assessments. In their analysis, they also investigated whether the target problem or type of therapy influences therapy impact. With respect to depressive disorders in childhood and adolescence, this results in an overall small effect (g) of 0.29 for post-treatment and 0.22 for follow-up assessments (Weisz et al., 2017). These small effects are in line with other reviews or meta-analyses focusing on psychotherapy's efficacy in children and adolescents with depression (Oud et al., 2019; Yang et al., 2017). A review and meta-analysis (Weisz et al., 2006) was recently updated (Eckshtain et al., 2019) and reports an overall effect size (g) of 0.36 at post-treatment and 0.21 at follow-up; the authors reported no relevant change in the effect sizes in the last 15 years. These studies replicate the effects of other investigations (Cuijpers et al., 2020), and confirm the lack of any significant changes in efficacy over the last decade. The evidence has emphasized the need for specific research into how the psychotherapy for depression in children and adolescents can be optimized. There have been even few reviews and meta-analyses that address their effectiveness. Wergeland et al. (2021) identified 8 studies with an overall effect size of 1.25; they included both uncontrolled and randomized, control trials.

One target to heighten effects might be involving the family of depressed youth, as there is evidence of an influence by the family environment on children's internalized symptoms (Armsden et al., 1990; Babore et al., 2016). Parental behaviour is assumed to influence children and adolescents in the development of depressive disorders (Feeny et al., 2009; Johnco & Rapee, 2018; McLaughlin et al., 2007), and some major risk factors for juvenile depression are associated with parenting, such as harsh parenting or parental over-involvement (Pinquart, 2017; Van Voorhees et al., 2008; Yap & Jorm, 2015). Studies also demonstrate positive aspects of family interaction when it is supportive and conducive (Roberts et al., 1996; Whittle et al., 2014). Positive relationships between caregivers and children have a long-lasting supportive effect on development (Sandler et al., 2011), and positive changes in the therapy of caregiver behaviour can also alleviate pre-existing symptoms (Schwartz et al., 2012).

With respect to psychotherapy in children and adolescents in general, Weisz et al. (2017) demonstrated in a meta-analysis that involving families and parents in psychotherapy increases the efficacy of the intervention, resulting in a slightly higher, though still moderate effect size (g) of 0.42. In line with this, Oud et al. (2019) noted a correlating association between favourable depression treatment outcomes in youths and parent involvement in psychotherapy, as did Carr (2019) in a systematic review on systemic family interventions. Even low threshold interventions with family involvement resulted in positive effects on child development and depressive symptoms (McLaughlin et al., 2007). Further, parental involvement in children's psychotherapy proved to be a predictor for positive follow-up effects on symptom improvement (Sun et al., 2019). Overall, there is correlational evidence that family and parent involvement in youths’ psychotherapy may positively impact symptoms overall (Weisz et al., 2017) as well as depression symptoms (Oud et al., 2019), though the existing studies vary in their results (Weersing et al., 2017; Weisz et al., 2013) and overall effects are at best small to moderate (Schwartz et al., 2012; Schwartz et al., 2012).

In sum, there are studies investigating effective treatments for children and adolescents with depression, but compared to similar common disorders of childhood and adolescence such as anxiety disorders or attention deficit/hyperactivity disorder, the number thereof is quite limited (Weisz et al., 2019). The few studies and meta-analyses available provide general evidence that psychotherapeutic interventions result in small to moderate effects that can be reinforced by involving the family and caregivers. The aim of the current study is thus to focus on the effects of family or caregiver involvement on treating depression, as there is no systematic evidence yet on this topic. In a systematic review and meta-analyses, we thus posed the question: “What are the effects of family/caregiver involvement on the efficacy of psychotherapeutic treatment of children and adolescents with depression?”. By using the term caregivers, we refer to biological parents or other custodians, and by family, we refer to caregivers with their children and other immediate family members who are involved in therapy sessions. This terminology aims to clearly differentiate these two groups.

METHODS

Study selection

To identify relevant studies, we employed the Participants, Interventions, Comparisons, Outcomes, and Study type (PICOS) scheme (Moher et al., 2010). Studies were included if participants (P) were 21 years of age or younger and been diagnosed with an MDD, dysthymia, or subthreshold depressive symptoms (SD). Diagnoses had to have been made by applying evaluated self-rating instruments, clinical or subclinical interviews, or external assessment instruments targeting depressive symptoms. Participants had to have undergone psychotherapeutic treatment aiming to treat depression and including their family or caregivers. The intervention group needed to have a specific, significant focus on integrating the family or caregiver within some or all the psychotherapy sessions (I), and they were compared against a comparison group without family or caregiver involvement (which could be a waitlist control or a treatment that did not involve family or caregivers or not to the same extent as the intervention group) (C). Outcome measures of studies needed to be operationalized as severity of depression, level of functioning, or depressive symptoms. Studies needed to compare their intervention group to their control group at post-measurement after the intervention (O). We only included randomized controlled trials. Case-control, pilot, or cohort studies with no randomized allocation were excluded (S). Trials with caregivers who had a history of depression were included in the analysis, although that was not an inclusion criterion. Studies had to be in English or German. There were no restrictions related to the publication year or any other filters.

Search strategy

To identify potential studies for this review, our search incorporated five different databases (PubMed, PsycINFO, ERIC, COCHRANE, and PSYNDEX). We additionally searched the reference sections of reviews and meta-analyses as well as treatment guidelines (see Appendix 1). To find unpublished material, we screened dissertations and theses and contacted authors in the given field directly. In case of systematic reviews or meta-analysis, those were screened for includable studies as well. Our search strategy relied on the topic-specific terms below (see Figure 1).

The data selection process was based on the PRISMA flow chart (Moher et al., 2010). Titles and abstracts of the studies were selected after searching databases or other additional sources based on the inclusion criteria. After this initial screening (KS, ND), a full text screening based on the inclusion criteria was performed by two independent researchers (KS, ND). Discrepancies about the eligibility of studies were resolved with the assistance of other independent team members (HC, ELB).

Data extraction and analysis

The extracted information included intervention format, study population, sample size, participant demographics and baseline measurement characteristics, details on the intervention and control conditions, study methodology, recruitment and study completion rates, outcomes, and measurement times. If data were unavailable, we contacted the study's authors. The extracted outcome of the studies addressed the efficacy of treatments, that is, the operationalization of the change in severity of depression or depressive symptoms, and is basically oriented on DSM-III, -IV, −5 (American Psychiatric Association (APA), 2013) or ICD-10 (Dilling et al., 2015). Based on Hazell et al. (2002) and Loechner et al. (2018), we applied the hierarchical ranking according to Petti (1985) and Hazell et al. (2002) of self-report depression measurement instruments. If there was more than one outcome measurement of depression symptoms reported and none thereof was in the ranking, we applied the Hazell et al. (2002) criteria and used the most valid instrument to calculate effect sizes. To examine the efficacy of the investigated interventions with family or caregiver involvement, we focused on comparing the post-treatment measures with the control and intervention group. We calculated effect sizes based on Cohen's d. For this purpose, we extracted the sample size, mean value, and standard deviation from the included studies. If there was no way to extract these data, we calculated the effect sizes based on t-values, ANOVA, or z-scores. Studies not providing such information were excluded for meta-analysis.

Based on Borenstein et al. (2010), we conducted a meta-analysis with the random-effects model. Analyses were done in R version 3.6.1, using the “meta” (Schwarzer, 2007), “metafor” (Viechtbauer, 2010), and “dmetar” (Harrer et al., 2019) packages. We assessed the post-intervention assessments with 95% confidence intervals. We used the DerSimonian–Laird estimator for the in-between study variance τ² of the distribution of true effect sizes (DerSimonian & Laird, 1986) orientated on Jackson et al. (2017) and the Q-Profile method for the in-between study variance (Veroniki et al., 2016). We assessed heterogeneity using the Cochran´s Q to compute I² (Higgins et al., 2003). Relying on Higgins et al. (2003), we interpreted the heterogeneity index as follows: “low” (25%), “moderate” (50%), and “substantial” (75%). We excluded outliers if their upper or lower bound is lower or higher than the pooled effect confidence interval (Viechtbauer & Cheung, 2010). To visualize any small studies lacking small effect sizes (publication bias), we use funnel plots and testing for asymmetry via the Egger's test (Egger et al., 1997). If the result is significant, we follow Duval and Tweedies's Trim-and-fill procedure (Duval & Tweedie, 2000) to estimate the effect size if those studies had been published.

Study quality

To capture the risk of bias and quality of the randomized trials included, we used the Cochrane Collaboration's tool version 2 for assessing risk (Higgins et al., 2018), which is a domain-based evaluation of six different biases (selection, performance, detection, attrition, reporting, and performance bias). Two reviewers rated the included studies on the five different sections based on the manual of bias assessment (KS, ND). After the two ratings, one author (ND) resolved any conflicts. HC and ELB were consulted in case of difficult conflicts.

RESULTS

Results of search

Our initial search ended in August 2021 and yielded 10.064 results (8.167 after removing duplicates). See Figure 2 for the full screening process. While full text screening, we removed another 10 duplicates and 11 studies in this step because full text data were not accessible. Applying our inclusion criteria, we were left with a total of 195 abstracts for full text screening, after which we had to exclude 110 studies, as they failed to meet our inclusion criteria. Figure 2 contains more information on our specific exclusion reasons. Two researchers (KS, ND) reviewed the final selection of studies to extract data. Two researchers (HC, ELB) helped to settle disagreements. Some of the included studies relied on the same data sets, in which case we included the study reporting a stronger focus on our research question (family involvement, pre-post comparison). In the end, we were able to include 19 studies in our review and 17 in our meta-analysis. The reduced number of studies in the meta-analysis resulted to the missing possibility of differentiation in the data extraction for the intervention group (Bernal et al., 2019) and of depression-specific data (Fristad et al., 2002).

Study characteristics

All the included studies were published between 1990 and 2020 in England, the USA, Norway, and Australia. An overview thereof is in Table 1. In total, N = 1553 participants were examined, the mean number of participants per trial was M = 81.74. Overall, 56.73% of the participants were girls. In one trial, the authors examined children and adolescents suffering from affective disorders and included patients with bipolar disorder and depression (Fristad et al., 2002). In this case, we focused on the children with depressive disorders. Children and adolescents included in the trials were aged between 3 and 18 years. Overall, N = 1308 patients were included with post-treatment measures and 11 of the included studies conducted a follow-up assessment (N = 655).

Study quality

Figure 3 provides an overview of the quality ratings. Overall, study quality was mixed. Except for blinding of participants, risk of bias across categories was mostly low or there were some concerns. The first area of risk assessment is the selection bias. Every included study conducted a randomized allocation of participants to various groups, but general information about the method of the sequence generation was provided in just 11 studies, and allocation was concealed in 5 thereof. In most trials, only the outcome assessors were masked for all outcomes. Eight trials revealed a low risk for detection bias. To assess the risk of reporting bias, we sought to identify published study protocols or registrations, and succeeded in six cases (Diamond et al., 2010; Dietz et al., 2015; Esposito-Smythers et al., 2019; Luby et al., 2018; Tompson et al., 2017; Waraan et al., 2020). With regard to outcome reporting, we searched for intent-to-treat analysis or comparable methods and succeeded in 15 studies. A total of 16 trials reported on dropouts and exclusions in full. The three other trials failed to provide a thorough description of missing outcomes. No study revealed a high risk for attrition bias.

Intervention participants

The majority of studies included adolescents; only a few included children. Just one trial focused on preschool children exclusively: Luby et al. (2018) examined a therapy programme for children aged 3 to 7 years. The studies published before the year 2000 were most likely to have samples of adolescents aged 13 to 18 years (Brent et al., 1997; Clarke et al., 1999; Lewinsohn et al., 1990). Studies published after 2000 focused more on younger adolescents and children, of which five studies included patients 8 years of age and older (Asarnow et al., 2002; Dietz et al., 2015; Fristad et al., 2002; Tompson et al., 2017; Trowell et al., 2007). The oldest patients included were 18 years old. We identified no study investigating young adults up to 21 years of age.

Caregiver/Family involvement

The majority of the included studies reported significant effects by comparing the pre- and post-assessments of the intervention group with family/caregiver included to a control group. Just seven studies reported no significant differences related to the study's primary outcome (Asarnow et al., 2002; Bernal et al., 2019; Brent et al., 1997; Clarke et al., 2002; Esposito-Smythers et al., 2019; Fristad et al., 2002; Waraan et al., 2020). To complement the efficacy of the various programmes, descriptive data on the extent and format of family and caregiver involvement are presented in the section below. Further information is in Table 2.

Overall, between 50% and 25% of the conducted sessions took place together with caregivers or the families. Most of the studies implemented sessions with the children/adolescents, and meet afterwards or separately with either the caregivers or family. A total of eight studies incorporated, when indicated, the entire family (caregiver, patients, and siblings) in family sessions (Brent et al., 1997; Diamond et al., 2002, 2010; Dietz et al., 2015; Israel & Diamond, 2013; Tompson et al., 2017; Waraan et al., 2020). The general proportion of caregiver/family involvement was the highest in these eight studies; however, no individual session was extended to children/adolescents or caregivers. In contrast, we found seven studies that explicitly integrated the caregivers in all sessions, but no siblings or other family members (Asarnow et al., 2002; Bernal et al., 2019; Clarke et al., 1999, 2002; Fristad et al., 2002; Lewinsohn et al., 1990; Luby et al., 2018). When examining these data, it appears that the earlier studies in particular tended to explicitly include only the caregivers, implementing caregiver involvement mostly through parallel group programmes. The remaining four studies did not specify whether they had included the family or just caregivers.

To compare the frequency of inclusion to some extent, we examined the total number of (maximum) feasible sessions in relation to the (maximum) number of potential family sessions in the different programmes. This extent varied from 10% (Asarnow et al., 2002) to 100% (for all studies implementing family concepts without individual sessions, see Table 2). The studies that conducted individual therapy with the children/adolescents did not have an overall proportion of caregiver integration above 50%. The total number of sessions was not mentioned in one study (Esposito-Smythers et al., 2019). For an overview of the number of conducted sessions in the different interventions, see Table 2.

Beyond looking at who was included and to what extent, we considered the quality of caregiver/family involvement. In the investigations that studied family interventions (no individual sessions), we noted that in each intervention described, the focus was on a change in interaction (Brent et al., 1997; Dietz et al., 2015; Tompson et al., 2017), attachment (Diamond et al., 2002, 2010; Israel & Diamond, 2013; Waraan et al., 2020), or relationship (Luby et al., 2018). Apart from those studies, we observed a focus on caregiver/family psychoeducation on depression via several interventions, particularly evident in those that conducted separate caregiver sessions (Bernal et al., 2019; Clarke et al., 1999, 2002; Fristad et al., 2002; Lewinsohn et al., 1990). Asarnow et al. (2002) revealed the most limited caregiver involvement, but they only addressed depression and family-related themes explicitly in their sessions with children and adolescents. Table 2 provides detailed descriptions of the programmes’ contents. In addition, the studies use different therapeutic approaches. Six studies involved cognitive-behavioural approaches (Bernal et al., 2019; Clarke et al., 1999, 2002; Esposito-Smythers et al., 2019; Lewinsohn et al., 1990; Spirito et al., 2015). One study examines a systemic approach (Brent et al., 1997) and another examines a psychodynamic approach (Trowell et al., 2007). In the other 11 studies, it is not clearly stated which psychotherapeutic approach was used.

Implementation

In our analysis, we distinguished therapies implemented in individual formats (either with children/adolescents or caregivers) and group therapies (with the respective) or family sessions. The majority of studies, however, combined different formats and, as mentioned earlier, family sessions exclusively in several therapies. The depression-specific therapy components for children/adolescents were implemented in group format in five studies (Asarnow et al., 2002; Clarke et al., 1999, 2002; Duong et al., 2016; Lewinsohn et al., 1990) and in an individual format in another four (Bernal et al., 2019; Esposito-Smythers et al., 2019; Spirito et al., 2015; Trowell et al., 2007). Two studies reported treatment as usual (Fristad et al., 2002; Sanford et al., 2006). The sessions with the caregivers, which were not family sessions, also took place in a group format in almost all the remaining studies (Bernal et al., 2019; Clarke et al., 1999, 2002; Fristad et al., 2002; Lewinsohn et al., 1990). Only one of the studies also carried out individual caregiver sessions (if needed) (Trowell et al., 2007). Two interventions took place during or after school (Asarnow et al., 2002; Duong et al., 2016). One study entailed sessions at the family's homes (Duong et al., 2016) and another included family meetings with the therapists that took place after each individual session with the adolescents (Spirito et al., 2015). Likewise, the families were included as a group together with other families in only one therapy (Asarnow et al., 2002).

Mean intervention duration was 14.53 weeks (SD = 9.57; range 5–48 weeks) and the frequency of sessions was usually one weekly session lasting between 60 and 120 minutes. Esposito-Smythers et al. (2019) scheduled therapy sessions in increasing frequency over a total period lasting 9 months. Tompson et al. (2017) implemented therapies over a 6-month period. Three studies reported booster sessions after finishing the intervention (Clarke et al., 1999; Sanford et al., 2006; Spirito et al., 2015).

Control group

As randomized allocation to the intervention or control group was an inclusion criterion, every study had to have a control group. In none of the control groups was there any involvement of caregivers/families. In two studies, the control group was exclusively a wait list condition (Asarnow et al., 2002; Luby et al., 2018). Ten studies examined a not specified usual care condition or a nondirective supportive therapy (Clarke et al., 2002; Diamond et al., 2002, 2010; Duong et al., 2016; Esposito-Smythers et al., 2019; Fristad et al., 2002; Israel & Diamond, 2013; Sanford et al., 2006; Tompson et al., 2017; Waraan et al., 2020). Seven studies examined manual-based psychotherapies as a control group (eg. Spirito et al., 2015) in addition to treatment as usual or exclusively. Examined manual-based active control conditions were coping with a depression course (CWD-A), child-centred therapy (CCT), adolescent-only cognitive-behaviour therapy (AO-CBT), individual support programme (ISP), and adolescent group cognitive-behaviour therapy (AG CBT).

Meta-analysis

Effect of interventions versus control groups at post-assessment

We included 17 studies in our meta-analysis with a total of 1208 participants. The inclusion of the studies in the meta-analysis depended on whether it was possible to extract data from the studies with which the effect size could be determined at post-measurement. If studies investigated more than one control condition, we included the active control group to focus on the difference between family involvements; as such, comparisons currently considered the highest standard (Weisz et al., 2017). We were unfortunately unable to extract data from Fristad et al. (2002); as they did not report post-measurements separately for MDD and bipolar disorder, we could not extract their outcomes specifically for children versus their caregivers’ outcomes. One study failed to distinguish the intervention's data from the control group's (Bernal et al., 2019). Our meta-analysis of the remaining k = 17 studies revealed a significant effect (d = 0.34; [0.07; 0.60]; p = .01).

Cochrane's Q-Test revealed that heterogeneity was significant (Q = 77.89; df = 16; p < .0001); see Figure 4 for the forest plot of the analysis. The variance of the true effect was estimated to be τ² = 0.48 [−0.30; 0.84]. The amount of total variability between the observed effect size I² = 79.5% is substantial based on Higgins et al. (2003). We therefore see clear indications for great heterogeneity. One possible way to deal with such widely heterogeneous data is to identify outliers in the included studies. We classify a study as an outlier if its confidence interval does not overlap with the confidence interval of the pooled effect. We thus identified no outliners in our data set.

Exploratory subgroup analyses

To examine whether the heterogeneity was due to the extent of family involvement or study quality, we conducted an exploratory subgroup analysis and additional meta-analysis. To investigate the influence of family involvement, we divided the studies into two groups based on the extent of caregiver/family involvement criterion. One group included therapies involving the entire family. The other group included all studies with proportionally less or mixed caregiver/family involvement. Our analysis demonstrated no significant difference between these two subgroups (Q = 1.80; df = 1; p = .18), but our subgroup analysis did reveal differently estimated between-study heterogeneity variance and different pooled effect sizes between the subgroup with mixed (k = 9, d = 0.17 [−0.09; 0.45], τ² = 0.10, I² = 61.9%) and full (k = 8, d = 0.54 [0.09; 0.98], τ² = 0.33, I² = 84.5%) caregiver/family integration. The information on which study belongs to which group is also given in Figure 4.

We then looked to see whether the heterogeneity was attributable to study quality. As the sample sizes were considered too small for a subgroup analysis concerning the three risk of bias categories (low, unclear, high), we conducted an additional meta-analysis with a smaller subsample of studies showing an overall a low risk of bias. We therefore selected those exhibited a low risk of bias in at least four of seven risk of bias categories. The analysis of those remaining k = 8 studies (Clarke et al., 2002; Duong et al., 2016; Esposito-Smythers et al., 2019; Israel & Diamond, 2013; Luby et al., 2018; Sanford et al., 2006; Tompson et al., 2017; Waraan et al., 2020) revealed a significant effect (d = 0.44; [0.07; 0.80]; p = .02). We noted significant heterogeneity in this analysis as well: the Cochrane's Q-Test revealed that such heterogeneity was significant (Q = 43.34; df =7; p < .0001).

Publication Bias

Despite trying to identify unpublished literature, we were unable to access unpublished studies. This entails a risk for upward bias in our study sample. In terms of potential publication bias, a funnel plot inspection revealed a symmetric distribution of the observed effects around the average true effect (see Figure 5). Visual inspection is supported by Egger's test (Egger et al., 1997) indicating a non-significant result (t = −0.256; p = .80). Based on the non-significant result, we refrained from Duval & Tweedie's trim-and-fill procedure (Duval & Tweedie, 2000).

DISCUSSION

This review provides an overview of the latest pool of evidence of family/caregiver involvement in the psychotherapy for children and adolescents suffering from depressive disorders. The total number of studies is quite small, as there were only 19 we could include. Our purpose was to determine whether the involvement of families or caregivers during therapy is a beneficial and effective approach. Our meta-analysis showed a significant result in that we found a clear difference in efficacy between the intervention and control groups, but this should be interpreted taking into account the significant heterogeneity between studies. The results suggest that interventions involving the family/caregiver can be effective and are worthy of further investigation in the treatment of children and adolescents with depressive disorders, but the relatively small number of studies and the large variability in study quality reduce the confidence with which definite conclusions can be drawn. It is important to understand which factors make a relevant contribution to therapy and to incorporate the multifactorial aetiology and maintenance of the disorder into treatment.

Study quality helps us explain heterogeneity. We found that the overall quality of these studies resembled that of other psychotherapy-focused reviews and meta-analyses (e.g. Oud et al., 2019). A subgroup analysis was not possible because of the small sample size. Nevertheless, to examine whether study quality affected heterogeneity, we conducted an additional meta-analysis with all investigations exhibiting a low risk of bias, which also yielded a significant result, but also showed strong heterogeneity between studies in that group. In addition to the information on heterogeneity, our analysis revealed another strong indication (through the significant effect we detected) that involving the family in psychotherapy can make a genuine difference.

Another factor that can cause considerable heterogeneity when researching the psychotherapy of children and adolescents is the patients’ age. The studies we included were fairly homogeneous in this regard; most of them implemented programmes for adolescents within a similar age range. Thus we do not assume age to be a relevant factor in terms of the heterogeneity in our meta-analysis.

We distinguished between programmes that include the family or caregivers in part from programmes that include them entirely in psychotherapy. Our exploratory subgroup analysis demonstrated no significant result after making this distinction, but it did reveal a large difference in the examined pooled effect sizes of the groups. The programmes with full family involvement demonstrated larger effect sizes and a generally different focus in terms of content. They did not just focus on the psychoeducation of childhood and adolescent depression or CBT-based interventions – they also clearly focused on the relationship, attachment, or interaction, sometimes without a psychoeducational basis. This similarity may explain the different effect sizes, and highlights the proven need for positive, supportive caregiver–child relationships and interactions to reduce depressive symptoms.

Both the frequency of sessions and format are factors to consider when interpreting heterogeneity and the efficacy of psychotherapy independent of caregiver/family involvement. We observed many possible formats and different frequencies in the studies we included. This might explain the heterogeneity we detected. There is no descriptive evidence that certain programmes or certain therapy forms are more likely to be implemented in one certain format than others. Since the number of widely different implementations was too high, further subgroup analyses to examine heterogeneity were not possible.

Most importantly, the analysis of the different interventions shows that there are many different ways to implement caregiver/family involvement in practice. The included studies provide new ideas for more individualized solutions for involving caregivers/families in psychotherapy for children and adolescents suffering from depression. As we found no evidence of publication bias, we assume that our review indeed reflects the current paucity and heterogeneity of evidence in this area (Eckshtain et al., 2019; Weisz et al., 2006), which is also underscored by other reviews and meta-analyses focusing on general treatment effects in children and adolescents with depression (Forti-Buratti et al., 2016; Loechner et al., 2018; Oud et al., 2019). The literature provides solid evidence that the family environment plays a significant role in the development and maintenance of childhood and adolescent depression (Carr, 2019; Cole et al., 2016). Our analysis supports these conclusions with respect to psychotherapy. We see promising results for interventions that focus on adolescents with the entire family. In addition, we find that family-involved interventions are particularly efficacious when they focus on relationships or patterns of interaction.

Limitations

To the best of our knowledge, this study is the first systematic review and meta-analysis of family/caregiver involvement in the psychotherapy of children and adolescents with depressive disorders. We sought high-quality studies on caregiver/family involvement, and established strict inclusion criteria that are comparable to and oriented towards other reviews in this field.

First, the most important limitation in this study is the low number of included studies. As this research field seems to have inspired much too little research thus far, we were unable to include more studies. Furthermore, a substantial number of studies included in our full text screening failed to enable full access, even after we had contacted the author or tried to order the journal through the university library. That factor also effectively reduced the number of studies we possibly could include in our review and meta-analysis.

Second, there are limitations related to the inclusion criteria. We defined depression as the presence of depressive symptoms in self-ratings or DSM-based depression inventories. This inclusion criterion could disadvantage younger children, as self-reporting is hampered in children under 8 years of age (Luby et al., 2007). In addition, diagnostic interviews with caregivers about the internalizing of disorders by their children are often difficult to interpret, as internalized symptoms are much harder for caregivers to perceive and identify than are externalized disorders (De Reyes & Kazdin, 2005). In addition, the inclusion criteria, and thus the search for studies, focus on family-focused therapies rather than parent-focused interventions. This is done against the background of being as open as possible to studies that include a wide range of caregivers. But, of course, this may have led to limitations on parent–child therapies.

Third, the interpretation of our meta-analysis is somewhat limited by the fact that we were unable to conduct additional subgroup analyses of other relevant areas. Given the large study heterogeneity, it was not possible to form conclusive subgroups, so we decided against other subgroup analyses beyond the analysis we performed.

Conclusion

The family is the decisive formative environment for children and adolescents. As outlined above, there is a paucity of strong evidence, even though the family is known to be a key factor in the development of depression (Cole et al., 2016; Restifo & Bögels, 2009; Yap & Jorm, 2015). We therefore see a strong need for effective caregiver/family integration in psychotherapy, and we acknowledge evidence of the efficacy of family and caregiver involvement. This study provides the first overview of the current state of research on family/caregiver involvement in psychotherapeutic interventions in children and adolescents suffering from depression.

Our results demonstrate that it makes a difference whether family and caregivers are involved in the therapy or not. The integration of families or caregivers requires additional efforts and engagement in therapy. Without making significant changes, the current psychotherapy programmes for adults cannot be simply adapted to accommodate children and adolescents. Nevertheless, in light of our generally very promising results, we see this effort as being very worthwhile. There are many different ways to integrate family members within the therapy to fulfil the individual patient's needs, and there are effective ways of including caregivers or families. Studies involving younger children are especially lacking. Since developmental factors are likely to influence effects of family/caregiver involvement in the psychotherapy of youth with depression, studies should aim to include different age groups to identify when family/caregiver involvement might be most beneficial, and when it might cause detrimental effects. Also urgently needed is greater research focus on the format and extent of caregiver/family involvement. We identified a wide range of possibilities to integrate families/caregivers in the therapy, but a systematic evaluation is necessary of different approaches related to its formats and duration. Furthermore, the examination of different psychotherapeutic approaches such as cognitive-behavioural-therapy or systemic approaches is very relevant. Studies need to apply transparent outcome measurements describing the family relationship and patients’ depressive symptoms. More intensive and systematic research is essential to identify which children/adolescents with depression need psychotherapy, under what circumstances it should take place, and in what form the caregivers or family should be effectively involved in their psychotherapy.

APPENDIX 1

Meta-analyses and review that have been screened for potential relevant studies

(Arbeitsgemeinschaft der Wissenschaftlichen Medizinischen Fachgesellschaften (AWMF) - Ständige Kommission Leitlinien, 2013; Carr & Boyd, 2003; Cuijpers et al., 2020; Diamond et al., 2016; Elmquist et al., 2010; Forti-Buratti et al., 2016; Francis et al., 2012; Fristad et al., 1996; Goldberg-Arnold et al., 1999; Hetrick et al., 2015; Kaslow et al., 1994; Klein et al., 2007; Lewinsohn & Clarke, 1999; Luby, 2013; Michael & Crowley, 2002; Mufson et al., 1999; Nel, 2014; Northey et al., 2003; Oud et al., 2019; Puig-Antich & Weston, 1983; Ryan, 2005; Sander & McCarty, 2005; Sanderson et al., 2009; Schweitzer & Retzlaff, 2006; Sorensen et al., 2002; Sörheim Nilsen et al., 2013; Southam-Gerow et al., 2000; Spielmans et al., 2007; Stark et al., 1996; Stein et al., 2006; Straub et al., 2014; Tindall et al., 2017; von Sydow et al., 2008; Weersing et al., 2017; Weisz et al., 2006, 2013, 2017, 2019; Wozney et al., 2017; Yang et al., 2017)

All references are cited in the reference list.