Psychosocial,pharmacological,and legal interventions for improving the psychosocial outcomes of children with substance misusing parents: A systematic review

Child abuse, maltreatment, or neglect

Thirteen studies measured the number of substantiated child abuse incidents, recurrence of child abuse incidents, intervention by child welfare services, open or active child welfare cases, and/or ‘involvement’ with child welfare services without further specification. Five of these studies did not report sufficient data to permit calculation of effect sizes, either due to missing data (e.g., standard deviations, exact n), unclear reporting of the timing of outcome measurements after intervention completion, or because not all participants had completed the intervention at the time official outcome data was collected (Boles et al., 2007; Carey et al., 2010b; Hall et al., 2015a, 2021; Kissick et al., 2015; Pollock & Green, 2015).

Of the studies with sufficient data for effect size calculation (Flannery et al., 2020; Green et al., 2007; Jansson et al., 2007; Lam et al., 2008; Penny & Pratt, 2011; Smirk et al., 2014; Summey et al., 2018), five studies measured active child welfare investigation or involvement, or open child welfare cases after intervention completion compared to TAU (Flannery et al., 2020; Lam et al., 2008; Penny & Pratt, 2011; Smirk et al., 2014; Summey et al., 2018). In terms of intervention models, one study examined the impact of a parent-focused intervention comprised of individual therapy for fathers with children aged 8–12 years and behavioural couples therapy for their female partner that integrated parent-skills training (Lam et al., 2008). The other three studies were predominantly focused on mothers (prenatally and/or postnatally) and children during the neonatal and/or infancy period. Penny and Pratt (2011) examined the impact of a formalised plan of clinical care approach for mothers identified as having substance misuse issues at the time of birth. The intervention was conducted during the first 12-months of the child's life and aimed to improve the coordination and delivery of services across multiple areas, including parent mental health and parenting. Smirk et al. (2014) examined the impact of an outpatient detoxification programme for mothers and babies, which also included alternate day home-visits, psychoeducation, and weekly outpatient assessments by a neonatologist (Smirk et al., 2014). Similarly, Summey et al. (2018) examined the impact of a multidisciplinary comprehensive programme for managing NAS in newborns (‘rooming in’, in and outpatient pharmacology, phone-based parenting support, developmental screening). A meta-analysis of these conceptually similar studies (Figure 3) indicated that the odds of having active child welfare involvement did not differ between families engaged versus not engaged with treatment (n = 611; OR = 0.54, 95% CI 0.10 to 3.01; Figure 3) and there was statistically significant heterogeneity among the effect sizes (τ ² = 2.02; χ ² = 21.89, df = 2, p < 0.0001; I ² = 91%). There was no difference in the odds of child protective services involvement following engagement with a parent-focused intervention comprised of individual therapy for fathers with children aged 8–12 years and behavioural couples therapy for their female partner that integrated parent-skills training versus TAU (OR = 0.58, 95% CI 0.07 to 4.56).

OPEN IN VIEWER

Figure 3

Forest plot for child welfare involvement meta-analysis (post-intervention, infancy focused interventions).

Another single study (Flannery et al., 2020) compared whether the rate of child welfare involvement differed for infants with NAS who were treated with methadone versus morphine more than 12-months after hospital discharge. The odds ratio (OR) for this study suggests there were no differences in likelihood of child welfare involvement between infants treated with methadone or morphine (n = 83; OR = 0.69, 95% CI 0.70 to 4.05).

Lam et al. (2008) also measured active child welfare involvement at 6- and 12-months after treatment completion and single odds ratios similarly indicated that the odds of having active child welfare involvement did not differ between families engaged versus not engaged with treatment (n = 20; 6-months: OR = 0.44, 95% CI 0.03 to 5.88; 12-months: OR = 0.58, 95% CI 0.07 to 4.56). Further, a single study by Green et al. (2007) measured the number of substantiated cases of abuse, maltreatment or neglect approximately 4-years after families completed FTDC programmes across 4 locations in the United States (n = 450). There were greater odds of a substantiated case of abuse, maltreatment, or neglect for families who engaged with a FTDC programme compared to TAU (OR = 1.74, 95% CI 1.06 to 2.85).

Child abuse potential

Twelve studies used standardised measures to assess the potential for child abuse either directly after intervention completion and/or at one or more follow-up time-points. Three of these studies (Luthar et al., 2000, 2007; Stover et al., 2019) did not report sufficient data to permit calculation of effect sizes, predominantly due to lack of standard deviations or incomplete reporting of statistical tests to allow alternate effect size calculation. Of the studies with sufficient data for effect size calculation (k = 8), 3 studies examined the examined the effectiveness of interventions without a parental substance misuse component, but at least one intervention component focused on parenting or family-functioning (Barlow et al., 2019; Black et al., 1994; Kim et al., 2019), and the remaining 5 studies integrated a parental substance misuse treatment component with at least one or more treatment components focused on the parent, child, and/or family unit (Brown et al., 2021; Dakof et al., 2010; Dawe et al., 2007; Donohue et al., 2014; Leakey, 2005). Studies could be further differentiated by the type of measure used and child population. Six studies used the validated Child Abuse Potential Inventory (brief or full) and focused on younger children (Barlow et al., 2019; Black et al., 1994; Brown et al., 2021; Dakof et al., 2010; Dawe et al., 2007; Donohue et al., 2014; Saldana et al., 2015), whereas two studies measured abuse potential using the Adult-Adolescent Parenting Inventory (AAPI-2; Kim et al., 2019; Leakey, 2005).

At the post-intervention time-point, child abuse potential, as measured by the Child Abuse Potential Inventory (brief or full), was lower for families with young children engaged with treatment compared to those who were engaged with TAU (n = 259; SMD = 0.36, 95% CI 0.12, 0.61). There was no heterogeneity detected (τ ² = 0.00; χ ² = 3.36, df = 4, p = 0.50; I² = 0%), and we did not conduct subgroup analysis given that there were too few studies and because the distribution of possible effect modifiers was equivalent across the studies included in the meta-analysis. Figure 4 provides a forest plot of the distribution of effect sizes.

OPEN IN VIEWER

Figure 4

Forest plot for child abuse potential meta-analysis (post-intervention, young children).

At follow-up within 6-months of treatment completion, child abuse potential as measured by the Child Abuse Potential Inventory (brief or full) did not differ between families with young children engaged with treatment compared to those who were engaged with TAU (n = 272; SMD = 0.25, 95% CI −0.07 to 0.57). There was moderate but nonsignificant heterogeneity (τ ² = 0.06; χ ² = 6.83, df = 4, p = 0.15; I² = 41%), yet too few studies to permit subgroup analyses (see Figure 5).

OPEN IN VIEWER

Figure 5

Forest plot for child abuse potential meta-analysis (follow-up, young children).

Examination of the a priori effect modifiers did not uncover any clear categories or outliers among the studies according to: child age, outcome measure, type of substance misuse, intervention setting, or practitioners implementing the intervention. Upon examining other features of the studies that may explain heterogeneity, one study differed from the others due to having a treatment duration of <3-months, compared to the other 4 studies with treatment duration of approximately ≥6-months (Dawe et al., 2007). A sensitivity analysis that removed this study increased the precision of the meta-analysis (SMD = 0.14, 95% CI −0.12 to 0.49) and reduced the heterogeneity (τ ² = 0.00; χ ² = 2.43, df = 3, p = 0.49; I² = 0%). On its own, this study indicated that child abuse potential was lower 6-months following engagement with both methadone outpatient services and a family-focused multicomponent intervention in the home compared to TAU methadone outpatient services (SMD = 0.88, 95% CI 0.24, 1.53). The other 4 studies (Barlow et al., 2019; Brown et al., 2021; Donohue et al., 2014; Leakey, 2005) were otherwise similar to Dawe et al. (2007) in terms of population, point of intervention (post-partum), and an intervention approach that integrated parental substance misuse treatment with one or more treatment components that focused parent wellbeing, child health and development, parenting, and/or overall family-functioning. This result should be interpreted with caution and does not indicate that the particular intervention approach for Dawe et al. (2007) is more effective than the other studies measuring child abuse potential at longer-term follow-up.

Two studies included families with adolescents and measured child abuse potential using a measured focused on risk of adolescent child abuse (AAPI-2), which is distinctly different measure than for younger children (Kim et al., 2019; Leakey, 2005). One study evaluated the effectiveness of a Solutions Focused Behavioural Therapy model on adolescent abuse potential using a sample of substance misusing parents. The intervention was comprised of weekly individual sessions over the course of approximately 3-months. The other study examined the effectiveness of The Nurturing Program, which formed part of a therapeutic community where mothers with substance misuse issues reside with their children. The study specifically compared the effectiveness of the relationally focused programme with or without the integration of ‘nurturing time’, which was an experiential component that provided mother-child dyads the opportunity to practice the relationship building skills emphasised in the programme. In light of the qualitative differences between these two intervention approaches, they were not combined with meta-analysis. Upon completion of the intervention, abuse potential did not differ between parents who participated in the Solutions Focused Behavioural Therapy model compared to TAU on any of the 5 AAPI-2 subscales (see Table 7 above). Likewise, abuse potential did not differ between those who participated in the Nurturing Parenting Program with or without the nurturing time experiential component (see Table 7 above).

OOHC

Twenty-four studies reported a measure related to child OOHC, yet there was a substantial amount of missing data across the studies, with only 13 studies reporting sufficient data to calculate effect sizes (Abrahams et al., 2010; Butz et al., 2001; Donohue et al., 2014; Flannery et al., 2020; Forrester et al., 2008; Hall et al., 2014, 2015b, 2016; Harwin et al., 2013; Hünsler et al., 2013; Lawlor et al., 2020; Schaeffer et al., 2011; Smirk et al., 2014). For most studies with missing data, the main issue was missing standard deviations or n by groups, unclear reporting of the timing of outcome measurements after intervention completion, difficulty verifying if all participants had completed the intervention, or variation in follow-up periods across participants without appropriate time-to-event analyses. In the paragraphs that follow, the studies with missing data are omitted, however, Supporting Information: Appendix F provides a summary of studies with and without missing data by each outcome category.

Two studies (Donohue et al., 2014; Schaeffer et al., 2011) measured the number of days in child protective custody or OOHC. Donohue et al. (2014) examined the effectiveness of a home-based enhanced Family Behaviour Therapy intervention comprised of weekly home-based sessions over the course of 6-months. The mean number of days in child protective custody did not differ between families who participated in the enhanced Family Behaviour Therapy versus TAU at either post-intervention (n = 72; SMD = 0.21, 95% CI −0.21 to 0.67) or at follow-up <6-months later (n = 72; SMD = 0.25, 95% CI −0.22 to 0.71). The second study (Schaeffer et al., 2013) examined the effectiveness of the Building Strong Families intervention, which integrated Multisystemic Therapy for Child Abuse and Neglect (MST-CAN) and Reinforcement Based Therapy (RBT). At follow-up more than 12-months after treatment completion, there were no differences between the intervention and TAU participants in relation to the number of days children spent in OOHC (n = 43; SMD = 0.27, 95% CI −0.34 to 0.88).

Six studies assessed whether receiving treatment impacted the odds of children being in an OOHC placement. Three studies evaluated pharmacological interventions combined with either a rooming-in approach to treat infants with NAS or TAU in neonatal intensive care (Abrahams et al., 2010; Hünsler et al., 2013; Lawlor et al., 2020), one evaluated a FTDC model (Harwin et al., 2013), one evaluated a home-visiting intervention focused on mother and child health and wellbeing from birth until the child reached 18-months of age (Butz et al., 2001), and another examined the impact of an outpatient detoxification programme for mothers and babies, which also included alternate day home-visits, psychoeducation, and weekly outpatient assessments by a neonatologist (Smirk et al., 2014). A meta-analysis of the 6 studies indicated that the odds of a child being in OOHC with a nonrelative was lower for families engaged in treatment versus those engaged with TAU (Figure 6; n = 1607; OR = 0.62, 95% CI 0.41 to 0.93). Non-signficant heterogeneity among the effect sizes was detected (τ ² = 0.08; χ ² = 7.90, df = 5, p = 0.16; I² = 37%).

OPEN IN VIEWER

Figure 6

Forest plot for out-of-home care meta-analysis (post-intervention, infants and toddlers).

A single study (Flannery et al., 2020) compared whether the rate of OOHC differed for infants with NAS who were treated with methadone versus morphine more than 12-months after hospital discharge. There were no differences in the odds of OOHC between infants treated with methadone or morphine (n = 83; OR = 0.72, 95% CI 0.24 to 2.14). Similarly, Kelly et al. (2015) found that when mothers and infants participated in supported at-home weaning for treatment of NAS, there was no difference in odds of the infant being apprehended by child protection services (with no further specification of where the child was placed) compared to in-hospital weaning with additional support from a caseworker and public health nurse (n = 80; OR = 0.69, 95% CI 0.26 to 1.79).

Five studies examined the effect of interventions on the likelihood of a child being in the care of family members other than parents. Butz et al. (2001) examined whether a home-visiting intervention with mothers and their infants influenced the likelihood of placement in kinship care at 36-months of age (∼6 to 18-months after end the intervention) compared to TAU. There were no differences in the odds of kinship care for families who engaged with treatment versus TAU (n = 117; OR = 0.76, 95% CI 0.34 to 1.70).

Four studies compared alternate treatments for NAS (Hall et al., 2014, 2015b, 2016; Lawlor et al., 2020). Lawlor and collegues found that the odds of discharge to a family member did not different according to whether infant NAS was treated with a structured care-by-parent protocol involving rooming-in with the infant or TAU (OR = 1.00, 95% CI 0.26 to 3.86). Hall and colleagues found that the odds of discharge to a family member did not differ according to whether NAS was treated with (1) a stringent versus non-stringent weaning protocol (Hall et al., 2014; OR = 0.74, 95% CI 0.45 to 1.22); (b) a structured versus unstructured methadone weaning protocol (Hall et al., 2015; OR = 0.71, 95% CI 0.41 to 1.23); or (3) buprenorphine versus methadone (Hall et al., 2016; OR = 0.64, 95% CI 0.25 to 1.64). These same three studies also found that the odds of discharge to adoption or foster care did not differ according to whether NAS was treated with (1) a stringent versus non-stringent weaning protocol (Hall et al., 2014; OR = 1.13, 95% CI 0.67 to 1.90); (b) a structured versus unstructured methadone weaning protocol (Hall et al., 2015; OR = 0.96, 95% CI 0.55 to 1.69); or (3) buprenorphine versus methadone (Hall et al., 2016; OR = 0.64, 95% CI 0.25 to 1.64).

Two studies examined longer follow-up periods after treatment completion. Schaeffer et al. (2011) examined whether a multisystemic therapy approach (Building Stronger Families) impacted the likelihood of entering OOHC or the number of OOHC placements compared to TAU. Approximately 12–15 months after treatment completion, there were no differences in the likelihood of entering OOHC (n = 43; OR = 0.35, 95% CI 0.07 to 1.73) or the number of OOHC placements between the groups (n = 43; SMD = 0.29, 95% CI −0.32 to 0.90). Similarly, Forrester et al. (2008) examined if an intensive family preservation service (‘Option 2’) impacted whether children entered OOHC during a follow-up period of 3.5 years (range: 8 months–6.5 years, no statistically significant differences between groups on length of follow-up). There were no differences in the likelihood of entering OOHC (n = 368; OR = 0.95, 95% CI 0.59 to 1.54) between the groups.

Parental care

Thirteen studies examined if engagement with treatment impacted whether parents retained care or custody of their children (without specific reference to reunification or permanency). One study combined randomised and non-randomised treatment participants into reported counts for retention of parental care and so no effect size could be calculated (Ernst et al., 1999). Ten studies compared a psychosocial, pharmacological and/or legal intervention to a TAU condition at the post-intervention time-point (Abrahams et al., 2007, 2010; Butz et al., 2001; Forrester et al., 2008; Dakof et al., 2010; Hünseler et al., 2013; Lawlor et al., 2020; Singh et al., 2020; Stevens & Patton, 1998) and three studies compared alternate treatments (Hall et al., 2014, 2015b, 2016). One study compared a long-term therapeutic community with or without child(ren) residing with parents at 6-month post-discharge (Stevens et al., 1998) and another single study (Forrester et al., 2008) compared intensive family preservation service (‘Option 2’) to a TAU condition with a follow-up period of 3.5 years (range: 8 months–6.5 years, no statistically significant differences between groups on length of follow-up).

At post-intervention completion, the odds of parents retaining care of their children was significantly higher for parents engaged in treatment compared to TAU (n = 1702; OR = 1.54, 95% CI 1.14 to 2.09, see Figure 7). There was moderate but nonsignificant heterogeneity (τ ² = 0.05; χ ² = 8.72, df = 6, p = 0.19; I² = 31%), yet too few studies to permit formal subgroup analyses to explore this heterogeneity.

OPEN IN VIEWER

Figure 7

Forest plot for retaining parental care meta-analysis (post-intervention, all ages).

Upon examination of the seven studies, the nature of the interventions, and the a priori moderators specified for subgroup analysis, three main differences were identified across the studies. First, the studies could be divided by whether the intervention was focused in the prenatal/immediate birth or the postpartum period. Second, those focused on the prenatal/immediate birth period were also all conducted in health settings versus the post-partum studies that were set in community, home, and judicial settings. Third, all of the prenatal/immediate birth studies were also quasi-experiments versus the post-partum studies which were all RCTs. In light of this, and the inappropriateness of subgroup analysis, we chose to conduct two separate meta-analyses. The first included studies with interventions focused in the prenatal and immediate birth period and that were implemented in health settings (Abrahams et al., 2007, 2010; Hünsler et al., 2013; Lawlor et al., 2020; Singh et al., 2020). The second included studies where the intervention was focused on the post-partum period and not implemented in health settings (Butz et al., 2001; Dakof et al., 2010).

The odds of parents retaining care of their children were higher for parents who engaged in rooming-in with or without comprehensive prenatal care and/or pharmacological treatment for NAS (if required) compared to parents who engaged with TAU (n = 1523; OR = 1.66, 95% CI 1.14 to 2.43; see Figure 8). However, nonsignificant and moderate heterogeneity remained (τ ² = 0. 07; χ ² = 7.13, df = 4, p = 0.008; I² = 44%) and no clear factors could be identified that would permit sensitivity analysis.

OPEN IN VIEWER

Figure 8

Forest plot for retaining parental care (prenatal/neonatal focus, post-intervention).

The odds of parents retaining care of their children was no different for parents who engaged in post-partum and early childhood treatments than for parents who engaged with TAU (Butz et al., 2001; Dakof et al., 2010; n = 179; OR = 1.18, 95% CI 0.63 to 2.18). In addition, there was no statistically significant heterogeneity in the distribution of effect sizes (τ ² = 0.00; χ ² = 0.78, df = 1, p = 0.38; I² = 0%).

Stevens et al. (1998) examined whether residing in a long-term therapeutic community with or without children impacted whether mothers retained care of their children 6-months post-discharge. Findings from this single study suggest that the odds of mothers retaining care does not differ between mothers who do and do not have their children residing with them during treatment (n = 38; OR = 2.32, 95% CI 0.63 to 8.58). In comparison, Forrester et al. (2008) examined if an intensive family preservation service (‘Option 2’) impacted whether children remained with their parents at the end of a follow-up period of 3.5 years (range: 8 months–6.5 years, no statistically significant differences between groups on length of follow-up). Families engaged with Option 2 were more likely to be residing together than families engaged in TAU (n = 368; OR = 1.67, 95% CI 1.02 to 2.71).

Finally, in a series of studies examining different treatments for NAS, Hall and colleagues found that the odds of discharge to an infants' mother did not differ according to whether NAS was treated with (1) a stringent versus non-stringent weaning protocol (Hall et al., 2014; OR = 1.12, 95% CI 0.75 to 1.69); (b) a structured versus unstructured methadone weaning protocol (Hall et al., 2015; OR = 1.38, 95% CI 0.86 to 2.22); or (3) buprenorphine versus methadone (Hall et al., 2016; OR = 1.43, 95% CI 0.71 to 2.91).

Permanency

Thirteen studies reported permanency-related outcomes, yet only 7 of these studies reported sufficient data to calculate effect sizes (Boles et al., 2007; Chuang et al., 2012; Dakof et al., 2010; Harwin et al., 2018; Lloyd et al., 2021; Moore et al., 2020; Sagatun-Edwards, 2000; see Supporting Information: Appendix F for studies with missing data). Similar to other studies that use child welfare official data, the outcome time-point in relation to the intervention was often unclear or studies did not report appropriate time-to-event analyses to permit synthesis of studies examining time to permanency. Further, other studies only presented permanency data for children who had been reunified or who had experienced an OOHC rather than the entire randomised or assigned sample. These issues are likely to create biased effect estimates and their inclusion to estimate the effects of interventions is only recommended when appropriate data is available (e.g., Cox proportional hazards models, complete individual participant data; Deeks et al., 2019).

Of the 7 studies with sufficient data, 2 studies (Dakof et al., 2010; Harwin et al., 2018) assessed if engagement with an FTDC programme impacted whether children were permanently placed with a relative. At the post-intervention time-point, engagement with an FTDC did not increase the odds of children being placed permanently with relatives (n = 412; OR = 1.02, 95% CI 0.37 to 2.79; τ ² = 0.31; χ ² = 1.96, df = 1, p = 0.16; I² = 49%). Harwin et al. (2018) also examined if engagement with an FTDC impacted whether or not children were in a permanent placement at all and found that at the closure of the case, engaging with the FTDC model did not increase the odds of children being in a permanent placement (n = 350; OR = 1.16, 95% CI 0.71 to 1.90).

Two studies examined if engagement with an FTDC model increased the likelihood of children being permanently placed with one or both parents (Harwin et al., 2018; Sagatun-Edwards, 2000). At treatment completion, engagement with an FTDC did not increase the odds of children being placed permanently with a parent (n = 439; OR = 1.24, 95% CI 0.84 to 1.81; τ ² = 0.00; χ ² = 0.42, df = 1, p = 0.52; I² = 0%). Likewise, a meta-analysis of two studies (Boles et al., 2007; Sagatun-Edwards, 2000) indicated that engagement with a FTDC model did not increase the likelihood of permanency for children via a long-term foster parent/guardian or long-term placement (n = 4172; OR = 0.25, 95% CI 0.04 to 3.28; τ ² = 2.52; χ ² = 22.19, df = 1, p < 0.00001; I² = 95%). Based on the Boles et al. (2007) study, comparison children were more likely than FTDC children to be placed with a guardian or in a long-term placement at 6-months follow-up (18-months after beginning the intervention; n = 3595; OR = 0.27; 95% CI 0.19 to 0.38), 12-months follow-up (n = 3595; OR = 0.24; 95% CI 0.17 to 0.34), or 24-months follow-up (n = 2990; OR = 0.32, 95% CI 0.22 to 0.45).

A meta-analysis of 3 studies (Boles et al., 2007; Harwin et al., 2018; Sagatun-Edwards, 2000) indicated that engagement with an FTDC model did not increase the likelihood of permanency via adoption (n = 4523; OR = 0.79, 95% CI 0.31 to 1.99; Figure 9) and there was no statistically significant heterogeneity in the distribution of effect sizes (τ ² = 0.35; χ ² = 4.37, df = 2, p = 0.11; I² = 54%). Based on the Boles et al. (2007) study, however, comparison children were more likely than FTDC children to achieve permanency via adoption at 6-months follow-up (18-months after beginning the intervention; n = 3595; OR = 0.67; 95% CI 0.48 to 0.93), 12-months follow-up (n = 3595; OR = 0.60; 95% CI 0.43 to 0.83), or 24-months follow-up (n = 2990; OR = 0.64, 95% CI 0.46 to 0.89).

OPEN IN VIEWER

Figure 9

Forest plot for permanency via adoption meta-analysis (post-intervention, FTDC model).

Two other studies (Lloyd et al., 2021; Moore et al., 2020) used time-to-event data to examine whether engagement with an FTDC programme increased the likelihood of children exiting the child welfare system to a permanent placement. It was not appropriate to combine these studies with Harwin et al. (2013) as there was no fixed time-point in relation to the treatment for all participants (Deeks et al., 2019). Lloyd et al. (2021) appropriately analysed the data using a Cox proportional hazards model and found a medium to large effect, suggesting that the likelihood of a permanent child placement was higher for families engaged with a FTDC than for comparison cases who received TAU (unadjusted HR = 2.18, 95% CI 1.51 to 3.15). Similarly, Moore et al. (2020) analysed their data using a Cox regression model and found that the likelihood of a permanent child placement was higher for families engaged with a FTDC than for comparison cases who received TAU (adjusted OR = 2.76, p < 0.001 (CI not reported by authors)).

Finally, a single study analysed time-to-event data to examine whether engagement with an FTDC impacted the amount of time taken for children to reach a permanent placement (Chuang et al., 2012). Results of the authors' Cox proportional hazards model indicated that the time taken to achieve permanency was longer for families who had engaged with an FTDC compared to families who received TAU (HR = 0.52, robust SE = 0.10).

Reunification

Twenty-two studies reported reunification outcomes, however, only 6 reported sufficient data to calculate effect sizes (Boles et al., 2007; Brook et al., 2007; Lloyd et al., 2021; Quittan, 2004; Ryan et al., 2016; Sanmartin et al., 2020; see Supporting Information: Appendix F for studies with missing data). The studies with missing outcome data had issues similar to those reported for the other child welfare outcomes (see above). A particular issue with this outcome category was that not all children were removed from their parents at the beginning of the intervention, meaning that it was not possible for all families randomised or assigned to achieve the outcome. Further, studies did not consistently report if the reunification related to a removal prior or during the intervention, or whether the reunification was related to an initial or subsequent removals. These issues combined, rendered many of the studies unsuitable for estimating the effect of the intervention for this particular outcome.

Of the 6 studies with sufficient data to calculate effect sizes, 5 examined the impact of treatment on the rate of reunification. One study (Sanmartin et al., 2020) examined the impact of legislative responses to parental substance misuse whereby children are removed and placed into foster care. Specifically, Sanmartin and colleagues (2020) compared the rate of reunification for children under age 1 who had been removed due to parental substance misuse for 24 American states with such legislation to 26 American states without this type of legislation. Based on discrete time-hazard models, the likelihood of reunification with parent(s) after removal was lower in states with legislation that criminalises or classes parental substance misuse as child abuse compared to those states without the legislation (n = 350,604; HR = 0.95, 95% CI 0.94 to 0.96). The four remaining studies focused on rate of reunification included treatment and comparison groups where removal had occurred for all children due to parental substance misuse, but used different analytic methods and time-points, which precluded combining all three into one meta-analysis. Lloyd et al. (2021) examined whether engagement with an FTDC programme increased the likelihood of reunification, tracking families from the point of referral to the FTDC until an end date of 2019 (study period: October 2015–March 2019). The authors found that the likelihood of reunification during the study period was higher for families engaged with a FTDC than families receiving TAU (unadjusted HR = 3.92, 95% CI 2.42 to 6.33).

Boles et al. (2007), Ryan et al. (2016), and Quittan (2004) reported counts for reunification. Boles et al. (2007) and Quittan (2004) reported both post-intervention and follow-up data capturing multiple time-points between 6 and 18 months after intervention completion, whereas Ryan et al. (2016) began tracking families from the point of petition (initiated randomisation of cases) to 36-months later (i.e., approximately 24-month follow-up point). Boles et al. (2007) compared an FTDC model with parallel specialised treatment and recovery services to TAU (usual child welfare services), Quittan (2004) compared a standard FTDC model with a FTDC model enhanced with the Celebrating Families! intervention (cultural adaption of the Strengthening Families programme), and Ryan et al. (2016) compared TAU child welfare services to a Recovery Coaches model. There were no differences in the odds of reunification between a standard FTDC model and FTDC + Celebrating Families! at post-intervention (n = 51; OR = 1.53, 95% CI 0.46 to 5.02) or at the 6–12 month follow-up mark (OR = 1.28, 95% CI = 0.38 to 4.27). However, Ryan et al. (2016) found that the odds of reunification was higher for families assigned to Recovery Coaches compared to TAU (n = 3440; OR = 1.34, 95% CI 1.11 to 1.62) approximately 24-months after treatment completion (or 3-years after initial petition). Likewise, Boles et al. (2007) found that the odds of reunification was higher for families participating in an FTDC model with parallel specialised treatment and recovery services compared to TAU (usual child welfare services) at treatment completion (12-months after beginning treatment; n = 4084; OR = 2.85, 95% CI 1.94 to 4.18), at 6-months follow-up (n = 3595; OR = 2.64; 95% CI 1.87 to 3.71), 12-months follow-up (n = 3595; OR = 2.96; 95% CI 2.08 to 4.20), and 24-months follow-up (n = 2990; OR = 2.60, 95% CI 1.83 to 3.67).

A final study (Brook et al., 2007) compared the time taken to reunify for families engaged with an intensive service coordination treatment to families receiving TAU. Although reunification occurred at a faster rate for families engaged with treatment, the effect size representing the differences in trajectories was small and not statistically significant (Wilcoxon [Gehan] = 1.763, p = 0.1842).

Termination of parental rights

Six studies examined the effects of treatment using outcomes relating to termination of parental rights (Carey et al., 2010b; Dakof et al., 2009, 2010; Forrester et al., 2008; Green et al., 2007; van Wormer et al., 2016), however, only 3 studies reported sufficient data to calculate effect sizes. Studies with missing data generally did not clearly report the timing of outcome measurement in relation to treatment and so it was unclear whether all participants had completed the intervention and whether there was variability in when the outcome was measured after completion of the intervention. Two studies by Dakof and colleagues (2009, 2010) compared the FTDC model with either comprehensive case-management or a family-focused programme set in the home (Engaging Moms; Dakof et al., 2009, 2010). The initial quasi-experimental evaluation (Dakof et al., 2009) examined if the two approaches differed in relation to whether an application for termination of parental rights (TPR) had been filed and whether parents voluntarily surrendered their parental rights permanently. At treatment completion, the odds of a TPR application being filed were lower for families who participated in the FTDC+Engaging Moms approach than for families engaged with FTDC+case-management approach (OR = 0.28, 95% CI 0.11 to 0.74). There were no differences between the groups for voluntary and permanent surrender of parental rights (OR = 0.85, 95% CI 0.20 to 3.65). In the randomised evaluation of the FTDC+Engaging Moms model found that the odds of a child being placed in foster care or with a relative, along with TPR, did not differ between the treatment and comparison condition (foster: OR = 0.36, 95% CI 0.10, 1.34; relative: OR = 0.56, 95% CI 0.12, 2.57). The third study with sufficient data for effect size calculation also examined an FTDC model (Green et al., 2007) and tracked families for 5-years from enrolment (∼4-years after treatment completion). There were no differences between FTDC and comparison children in terms of whether the permanency type was termination of parental rights (n = 446; OR = 0.79, 95% CI 0.51 to 1.21).

Cognitive development

Eleven studies examined whether treatment impacted young children's cognition, yet three of these studies (Field et al., 1998; Jimenez, 2018; Kaltenbach et al., 1986) reported insufficient data to calculate effect sizes (e.g., no standard deviation or other statistical data to permit hand calculations). Five studies assessed intervention effectiveness using TAU comparison conditions and measured child cognition post-intervention when children were between 18 and 46 months (Black et al., 1994; Conerly, 1999; Ersnt et al., 1999; Kilbride et al., 2000; Porges et al., 2018). These interventions began at birth or early infancy and all focused on child health and development. None of the interventions included an intervention component focused on treating parental substance misuse, two included case-management (Ersnt et al., 1999; Kilbride et al., 2000), and two integrated a parent-focused therapeutic component (Black et al., 1994; Ersnt et al., 1999). A meta-analysis of these 5 studies indicated that early child cognition did not differ for those who had received treatment versus TAU (Figure 10; n = 245; SMD = 0.17, 95% CI −0.09 to 0.43) and there was little heterogeneity among the effect sizes (τ ² = 0.00; χ ² = 1.82, df = 3, p = 0.77; I² = 0%).

OPEN IN VIEWER

Figure 10

Forest plot for child cognition meta-analysis (post-intervention).

Three studies examined the impact of alternate treatments on child cognition and three of these were pharmacological interventions. The first study compared cognitive outcomes at age 12-months after treatment for NAS with either methadone or clonidine (Bada et al., 2015), and again there was no difference between treatments (n = 24; SMD = −0.05, 95% CI −0.85 to 0.75). The second study compared cognitive outcomes at age 18-months after treatment for NAS with either methadone or morphine (Flannery et al., 2020), and again there was no difference between treatments (n = 83; SMD = −0.10, 95% CI −0.53 to 0.33). The third study comparing alternative treatments examined the effectiveness of a child-focused developmental curriculum with young children with prenatal cocaine exposure delivered either one-on-one in the home twice a week or at in childcare group setting 5 days per week (Claussen et al., 2004). Mothers in both groups, who continued to experience significant psychosocial stressors (including substance use) also received other support services (e.g., medical care, social work services, transportation). Upon intervention completion, when children were approximately 36-months of age, there were no differences in cognition between the two treatment approaches (n = 96; SMD = −0.29, 95% CI −0.69, 0.11).

Executive functioning and hyperactivity

Five studies reported measuring child executive functioning, yet 2 did not report complete data for the executive function measure after intervention completion (Margolis et al., 2017; Sullivan, 1984). Kim et al. (2019) examined the impact of Solution Focused Behavioural Therapy for substance misusing parents – without inclusion of children – on child executive functioning measured via parent self-report on a standardised tool. At the post-intervention time-point, children of parents engaging with Brief Solution-Focused Therapy did not differ from parents engaged with TAU on: (1) a composite measuring global executive functioning (n = 75; SMD = −0.16, 95% CI −0.62 to 0.30); (2) a composite measuring ability to problem-solve, plan and organise, sustain working memory, and/or monitor performance (n = 75; SMD = −0.19, 95% CI −0.64 to 0.27); or (3) a composite measuring the ability to regulate behaviour (n = 75; SMD = −0.11, 95% CI −0.56 to 0.35). There was also no difference between groups on the 8 individual subscales that made up the executive functioning composites (see Table 8) or on another measure of attention (CBCL) completed by parents (see Table 8). Davis-Susser (1990) examined the impact of a child-focused group programme for children of alcoholics on two measures of child executive functioning measured via parent self-report on a standardised tool. At the post-intervention time-point, children engaging with the group programme did not differ from children who did not participate in the programme on: (1) a measure of attention (n = 40; SMD = −0.12, 95% CI −0.83 to 0.60); or (2) a measure of hyperactivity (n = 40; SMD = −0.28, 95% CI −1.00 to 0.44). The third study (Adelman, 1990) found that children engaging in a child-focused group therapy programme for children of alcohol-misusing parents did not differ on self-rated levels of self-directiveness than children of alcohol misusing parents who did not engage in treatment (n = 44; SMD = −0.16, 95% CI −0.76 to 0.43). While Sullivan (1984) did not report complete executive function outcome data, there was sufficient data for effect size calculation for a measure of hyperactivity. Sullivan (1984) examined the effectiveness of a therapeutic intervention for alcoholic parents compared to no treatment at 8-months follow-up. The intervention comprised 6 parent group therapy sessions, 6 child group therapy sessions, and 6 family therapy sessions focused on psychoeducation and supportive psychotherapy. There were no differences in the level of parent-reported levels of child hyperactivity after participating in the intervention compared to the no treatment control group (n = 33; SMD = −0.27, 95% CI −0.64 to 1.14).

Language development

Six studies examined whether treatment impacted young children's language development, yet one of these studies (Jimenez, 2018) reported insufficient data to calculate effect sizes (e.g., no standard deviation or other statistical data to permit hand calculations). Two studies assessed intervention effectiveness using TAU comparison conditions and measured child language post-intervention when children were between 18 and 46 months (Conerly, 1999; Porges et al., 2018). These interventions began at birth or early infancy and all focused on child health and development. None of the interventions included an intervention component focused on treating parental substance misuse or any other areas of family-functioning (e.g., parent wellbeing). A meta-analysis of these 2 studies indicated that receptive language development did not differ for those who had received treatment versus TAU (n = 84; SMD = 0.44, 95% CI −0.13 to 1.01) and there was modest but nonsignificant heterogeneity among the effect sizes (τ ² = 0.06; χ ² = 1.50, df = 1, p = 0.22; I ² = 33%). However, children receiving early developmental intervention had better expressive language development upon treatment completed compared to children receiving TAU (n = 85; SMD = 0.55, 95% CI 0.08 to 1.01) and there was no heterogeneity among the effect sizes (τ ² = 0.00; χ ² = 0.70, df = 1, p = 0.40; I² = 0%).

Three studies examined the impact of alternate treatments on child language, two of which were pharmacological interventions. These were the same treatment-treatment studies described above in the ‘Child cognition’ section (Bada et al., 2015; Flannery et al., 2020; Claussen et al., 2004). Flannery et al. (2020) found no differences in language development for children 18-months after treatment for NAS with either methadone or morphine (n = 83; SMD = −0.10, 95% CI −0.54 to 0.33). Similarly, Bada et al. (2015) found no difference in language development 12-months after treatment for NAS with either methadone or clonidine, for either total language development (n = 24; SMD = 0.17, 95% CI −0.63 to 0.98); expressive language development; (n = 24; SMD = 0.06, 95% CI −0.74 to 0.86) or receptive language development (n = 24; SMD = 0.44, 95% CI −0.37 to 1.25). The third study comparing the centre- versus home-based developmental curriculum (Claussen et al., 2004) found that both expressive and receptive language were more developed in children who participated in the centre-based versus home-based intervention (Expressive: n = 96, SMD = −0.75, 95% CI −1.17 to −0.34; Receptive: n = 96, SMD = −0.48, 95% CI −0.89, −0.07).

Motor development

Seven studies examined whether treatment impacted young children's motor development, yet two of these studies (Field et al., 1998; Jimenez, 2018) reported insufficient data to calculate effect sizes (e.g., no standard deviation or other statistical data to permit hand calculations). Two studies assessed intervention effectiveness using TAU comparison conditions and measured child motor skills post-intervention when children were between 18 and 36 months (Black et al., 1994; Ersnt et al., 1999). Both interventions began at birth or early infancy and focused on child health and development. Neither study included an intervention component focused on treating parental substance misuse, one included case-management (Ersnt et al., 1999), and both integrated a parent-focused therapeutic component. A meta-analysis of these studies indicated that early child motor skills did not differ for those who had received the intervention versus TAU (n = 91; SMD = −0.16, 95% CI −0.57 to 0.26), and no heterogeneity was detected among the effect sizes (τ ² = 0.00; χ ² = 0.65, df = 1, p = 0.42; I² = 0%).

Three studies examined the impact of alternate treatments on child motor skills and three of these were pharmacological interventions. These were the same treatment-treatment studies described above in the ‘Child cognition’ section (Bada et al., 2015; Flannery et al., 2020; Claussen et al., 2004). In contrast, Flannery et al. (2020) found no differences in motor skills for children 18-months after treatment for NAS with either methadone or morphine (n = 83; SMD = −0.25, 95% CI −0.68 to 0.19). Bada et al. (2015) also found no difference in motor development 12-months after treatment for NAS with either methadone or clonidine (n = 24; SMD = 0.19, 95% CI −0.61 to 1.00). The fourth study, which compared a centre- versus home-based developmental curriculum (Claussen et al., 2004) found that both fine and gross motor skills did not differ for children who participated in either the centre- or home-based intervention (Fine: n = 96, SMD = −0.12, 95% CI −0.52, 0.29; Gross: n = 96, SMD = −0.25, 95% CI −0.65, 0.16).

Sleep

Two studies measured child sleep problems. One was the study described above which compared a centre- versus home-based developmental curriculum (Claussen et al., 2004) and the other compared a parent and child-focused nurse home-visiting intervention to usual community care approximately 6 to 18-months after treatment completion (Butz et al., 2001). Claussen et al. (2004) found that the level of sleep problems did not differ for children who participated in either the centre- or home-based intervention (n = 96, SMD = −0.30, 95% CI −0.70, 0.10). Likewise, Butz et al. (2001) found that the level of sleep problems did not differ for children who received the home-visiting or usual community care at ∼18-months after treatment (n = 100, SMD = 0.21, 95% CI −0.18, 0.61).

Social skills

Eight studies reported child social skills as an outcome measure, however, three of these studies (Field et al., 1998; Margolis et al., 2017; Orte et al., 2015) reported insufficient data to calculate effect sizes (e.g., no standard deviation or other statistical data to permit hand calculations). One study utilised a child-report measure to examine the effectiveness of child-focused group therapy programme compared to a waitlist control group (Scheer, 1996). At the post-intervention time-point, children's self-reported social skills did not differ between those who were assigned to the intervention versus waitlist control condition (n = 30; SMD = 0.33, 95% CI −0.39 to 1.05).

Two studies compared psychosocial interventions to TAU conditions using parent-report measures of child social skills (Conerly, 1999; Kim et al., 2019). One study was a solution-focused individual therapy intervention for parents (Kim et al., 2019) and the other was a developmentally focused child and family early intervention programme (Conerly, 1999). Upon treatment completion, child social skills did not differ between the intervention and TAU conditions (n = 100; SMD = 0.11, 95% CI −0.29 to 0.52) and there was no heterogeneity detected in the distribution of effect sizes (τ ² = 0.00; χ ² = 0.64, df = 1, p = 0.42; I² = 0%). A third study examined the effectiveness of a therapeutic intervention (Sullivan, 1984) for alcoholic parents compared to no treatment at 8-months follow-up. The intervention comprised 6 parent group therapy sessions, 6 child group therapy sessions, and 6 family therapy sessions focused on psychoeducation and supportive psychotherapy. For this study, parents rated their children's social skills lower in after participating in the intervention compared to the no treatment control group (n = 33; SMD = −0.73, 95% CI −1.46 to −0.01). The final study compared two alternative treatments and is detailed in the ‘Child cognition’ and subsequent sections (Claussen et al., 2004). At post-intervention, child social skills, as rated by parents, did not differ for children who participated in either the centre- or home-based intervention (n = 96, SMD = 0.19, 95% CI −0.22, 0.59).

Other child development measures

Four studies measured developmentally related outcomes that did not fit clearly under the abovementioned outcome categories. Flannery et al. (2020) found no differences in the odds that a family was engaged with early intervention services to support child development 18-months after treatment for NAS with either methadone or morphine (n = 83; OR = 0.95, 95% CI 0.39 to 2.30). The second study (Sullivan, 1984), described under ‘Child social skills’ above, utilised a comprehensive standardised and normed parent-report measure of numerous aspects of child development and functioning. In addition to social skills and other areas examined under the behavioural/emotional outcome domains, three other scales were developmentally related: Cerebral Dysfunction, Intellectual Screening, and Development. At 8-months follow-up, parents in the treatment and control groups did not differ in reports of their children's difficulties in these three areas (n = 33; Cerebral Dysfunction: SMD = −0.39, 95% CI −1.09 to 0.32; Intellectual Screening: SMD = −0.58, 95% CI −1.29 to 0.14; Development: SMD = −0.30, 95% CI −1.00 to 0.40). The third study in this outcome category assessed whether children's development was ‘within normal limits’ with no further specification (Penny & Pratt, 2011). This study compared usual early childhood health care to a multidisciplinary coordinated clinical pathway to care model for families affected by parental substance misuse with children from birth to 12-months. At the post-intervention time-point, children in the coordinated care and TAU groups did not differ in terms of whether they were developing within normal limits (n = 24; OR = 1.43, 95% CI 0.11 to 18.00). The final study (Adelman, 1990) found that children engaging in a child-focused group therapy programme for children of alcohol-misusing parents did not differ from children who did not engage in treatment based on a teacher-rated measure of extroversion which included concepts such as self-centredness, energy levels, and social skills (n = 44; SMD = −0.26, 95% CI −0.86 to 0.33).

Total problems

Twenty-three studies examined the effectiveness of treatment using parent-report standardised measures that produce a composite for overall emotional and/or behavioural problems for children at their developmental age. Eight of these studies either reported no outcome data at all or insufficient data to calculate effect sizes (e.g., no standard deviation or other statistical data to permit hand calculations) (Barlow et al., 2019; Dore et al., 1999; Logan et al., 2001; Luthar et al., 2000, 2007; Margolis et al., 2017; Roosa et al., 1990; Sullivan, 1984). Of the studies with complete data (k = 15), 9 studies compared an eligible intervention to a TAU comparison condition at post-intervention, with 6 studies using both a post-intervention and short-term follow-up (≥3 months ≤6-months), and 3 studies also including a longer follow-up (>6-months ≤12-months). An additional two studies compared alternative treatments (Claussen et al., 2004 and Flannery et al., 2020 described above). Three studies included both a parent and child-report measure for total emotional and behavioural problems (Luthar & Suchman, 2000; Luthar et al., 2007; Margolis et al., 2017), but reported insufficient data to permit calculation of effect sizes (e.g., no standard deviations or other statistical data to permit hand calculations).

The meta-analysis of the 9 studies using a post-intervention outcome measure produced a small effect size, indicating that treatment was more effective than TAU for reducing emotional and behavioural problems (Figure 11; n = 567; SMD = 0.39, 95% CI 0.06 to 0.72). There was moderate and significant heterogeneity among the effect sizes (τ ² = 0.17; χ ² = 26.90, df = 8, p = 0.0007; I² = 70%), yet too few studies to conduct subgroup analyses. To explore this heterogeneity, we first examined the included studies according to the a priori defined moderators: child age, type of outcome measure, type of parental substance use, intervention setting, and practitioners implementing the intervention. In terms of child age, 3 studies included children with a mean age of ≥3 but ≤5 years (Brown et al., 2021; Conerly, 1999; Dawe et al., 2007), whereas 5 studies included children with a mean age of >5 to ≤10 years (Catania, 1985; Dyba et al., 2019; Kelley et al., 2002; Kim et al., 2019). However, dividing the studies by age was considered inappropriate as proportions of study samples fell into both age categories. All studies used either well-validated standardised measure or a measure with associated psychometric data to establish its reliability and validity. All interventions were implemented by clinicians, mostly with postgraduate training. Four studies did not specify or report the specific type of parental substance misuse (Catania, 1985; Dawe et al., 2007; Dyba et al., 2019; Kim et al., 2019), 2 studies included alcohol misusing parents (Kelley et al., 2002; Satyanarayana et al., 2016), 2 studies included drug misusing parents (e.g., cocaine; Conerly et al., 1999; Kelley et al., 2002), and 1 study included polysubstance misusing parents (Brown et al., 2021). This meant we could not divide the studies easily by type of substance misused by parents. The intervention was predominantly implemented in the home for 2 studies (Brown et al., 2021; Dawe et al., 2007) but in clinics or community settings for 7 studies (Catania, 1985; Conerly, 1999; Dyba et al., 2019; Kelley et al., 2002; Kim et al., 2019; Satyanarayana et al., 2016). A sensitivity analysis removing the two studies predominantly set in the home reduced the precision of the meta-analysis (SMD = 0.34, 95% CI −0.04. to 0.73) and slightly increased heterogeneity (τ ² = 0.20; χ ² = 24.54, df = 6, p = 0.08; I² = 76%). This suggested that the setting of the intervention may not have been the source of heterogeneity.

OPEN IN VIEWER

Figure 11

Forest plot for total problems meta-analysis (post-intervention, parent-report).

Because there were no clear outliers or cohesive categories in relation to a priori moderators, aside from intervention setting, we also examined the nature of the research designs and interventions to identify potential sources of heterogeneity. A range of possible sources of heterogeneity were identified, with the results of these sensitivity analyses presented in Table 10 below. Use of a non-randomised evaluation design, absence of a parental substance misuse treatment, an intervention duration of >6-months, or an intervention spanning prenatal and postnatal periods did not reduce heterogeneity in the results and reduced the precision of the meta-analysis. However, removal of the two studies that used a group delivery format (Cantania, 1985; Dyba et al., 2019) – compared to the others that were individual or a mix of delivery formats – reduced detected heterogeneity to non-significance and 28%, but the meta-analysis no longer reached statistical significance. Both of these studies evaluated an intervention with a parent-focus in terms of parental substance misuse and/or wellbeing, in addition to parent-skills training without an explicit focus on the parent-child relationship or attachment.

At ≥3-months to <6-months follow-up, a meta-analysis of 7 studies (see Figure 12) indicated that parents engaged in treatment did not report lower levels of total child behavioural and emotional problems compared to those receiving TAU (n = 463; SMD = 0.28, 95% CI −0.03 to 0.60), and there was statistically significant heterogeneity among effect sizes (τ ² = 0.10; χ ² = 14.97, df = 6, p = 0.02; I² = 60%). Similar to the post-intervention meta-analyses for total problems, we explored this heterogeneity by first examining the included studies according to the a priori defined moderators: child age, type of outcome measure, type of parental substance use, intervention setting, and practitioners implementing the intervention. In terms of child age, 1 study included children with a mean age of ≥3 but ≤5 years (Dawe et al., 2007), whereas 6 studies included children with a mean age of >5 to ≤10 years (Dyba et al., 2019; Kelley et al., 2002; Saldana et al., 2015; Satyanarayana et al., 2016). However, dividing the studies by age was considered inappropriate as proportions of study samples fell into both age categories. All studies used either well-validated standardised measure or a measure with associated psychometric data to establish its reliability and validity. All interventions were implemented by clinicians, mostly with postgraduate training. Four studies did not specify or report the type of parental substance misuse, 2 studies included alcohol misusing parents (Kelley et al., 2002; Satyanarayana et al., 2016), and 1 study included drug misusing parents (e.g., cocaine; Conerly, 1999). This meant we could not divide the studies easily by type of substance misused by parents. The intervention was implemented primarily in the home for 2 studies (Dawe et al., 2007; Saldana, 2015), in outpatient clinics or community settings for 4 studies (Cantania, 1985; Dyba et al., 2019; Kelley et al., 2002), and an inpatient setting for 1 study (Satyanarayana et al., 2016). A sensitivity analysis removing the two studies set in the home reduced the precision of the meta-analysis (n = 400; SMD = 0.24, 95% CI −0.11 to 0.59) and slightly increased heterogeneity (τ ² = 0.10; χ ² = 10.94, df = 5, p = 0.03; I² = 63%). A sensitivity analysis removing the single study using an inpatient setting indicated that parents engaged in treatment reported lower levels of total child behavioural and emotional problems compared to those receiving TAU (n = 286; SMD = 0.40, 95% CI 0.10 to 0.69), with nonsignificant heterogeneity (τ ² = 0.04; χ ² = 7.46, df = 5, p = 0.19; I² = 33%). This suggested that the divergent inpatient setting of the intervention may have been the source of heterogeneity.

OPEN IN VIEWER

Figure 12

Forest plot for total problems meta-analysis (short follow-up, parent-report).

However, other potential sources of heterogeneity were tested with sensitivity analyses, with the results presented in Table 10. These indicated that (1) use of a non-randomised evaluation design or an an intervention spanning prenatal and postnatal periods (Dyba et al., 2019); (2) absence of a parental or couples focused treatment component (Cantania, 1985); (3) use of an explicit child health and/or development intervention component (Dawe et al., 2007); (4) or use of a group-only delivery format (Cantania, 1985; Dyba et al., 2019) did not reduce heterogeneity in the results or reduced the precision of the meta-analysis.

At ≥6-months follow-up, a meta-analysis of 3 studies (Kelley et al., 2002; Butz et al., 2001; see Figure 13) indicated that parents engaged in treatment continued to report lower levels of total child behavioural and emotional problems compared to parents receiving TAU (n = 192; SMD = 0.31, 95% CI 0.03 to 0.60), and this overall effect had nonsignificant and minimal heterogeneity (τ ² = 0.00; χ ² = 0.09, df = 2, p = 0.96; I² = 0%). Two of these studies examined the effectiveness of 32 sessions of treatment for either paternal drug or alcohol misuse, which was comprised of 12 sessions of Behavioural Couples Therapy and 20 sessions of individual cognitive behavioural therapy (CBT) for fathers (Kelley & Fals-Stewart, 2002 [alcohol]; Kelley & Fals-Stewart, 2002 [drug]). The third study examined the effectiveness of a home-visiting intervention focused on parent and child wellbeing from birth until the child was aged 18-months (Butz et al., 2001).

OPEN IN VIEWER

Figure 13

Forest plot for total problems meta-analysis (long follow-up, parent-report).

Two studies (Forrester et al., 2008; dos Reis Viela et al., 2016) reported the number of participants who scored in the clinical range of the instrument used to measure total child emotional and behavioural problems. The Forrester et al. (2008) study is described in previous sections and reported the number of children who scored in the clinical range for the Strengths and Difficulties Questionnaire. The authors found that the odds of having clinical levels of emotional/behavioural problems did not differ for children engaged versus not engaged with treatment (n = 27; OR = 2.00, 95% CI 0.38 to 10.58). Dos Reis Viela et al. (2016) examined the impact of a school-based intervention model that provided mental health support and a range of other activities aiming to prevent substance use in children being raised in families affected by substance misuse. The authors found that children engaged with treatment were less likely to obtain a score in the clinical range for total emotional/behavioural problems, as reported by caregivers (n = 101; OR = 0.24, 95% CI 0.10 to 0.57).

Two studies compared alternative treatments using a measure of total child emotional and behavioural problems, the intervention for both of which are described in sections above. Flannery et al. (2020) found no differences in parent reported total problem behaviours 18-months after treatment for NAS with either methadone or morphine (n = 77; SMD = 0.00, 95% CI −0.45 to 0.45). Claussen et al. (2004) found that, at post-intervention, child behavioural and emotional problems, as rated by parents, did not differ for children who participated in either the centre- or home-based child development intervention (n = 96, SMD = −0.16, 95% CI −0.56, 0.24).

One study examined whether participating in a therapeutic group in the school setting impacted overall teacher-reported behavioural problems for children with alcohol misusing parents (Adelman, 1990). At post-intervention children engaging with this intervention did not differ from children who did not (n = 44; SMD = 0.04, 95% CI −0.55 to 0.63).

Externalising

19 studies examined the effectiveness of treatment using parent-report standardised measures that produce a composite for child externalising problems. Three studies had missing data, by either reporting no between group data or only data for participants with scores over a particular threshold on the outcome measure (Bartle-Haring et al., 2018; Pirnia, 2019; Margolis et al., 2017). Of the studies without missing data (k = 15), 11 studies compared an eligible intervention to a TAU comparison condition at post-intervention, 3 studies used both a post-intervention and short-term follow-up (≥3 months ≤6-months), and 5 studies also included a longer follow-up (>6-months ≤12-months). An additional two studies compared alternative treatments (Claussen et al., 2004 and Flannery et al., 2020 described above). In addition, one study provided both a child- and parent-report measure of externalising problems (Mitrani et al., 2010) and another provided teacher-reported externalising behaviours (Adelman, 1990).

Of the 11 studies using a post-intervention outcome measure, 4 were direct therapeutic interventions for children of substance misusing parents aged >5 but ≤13 years (Davis-Susser, 1990; dos Reis Viela et al., 2016; Haight et al., 2010; Scheer, 1996), and were meta-analysed separately to the other 7 studies which did not integrate a child-focused treatment component. Instead, these 7 studies were focused on parent wellbeing, parental substance misuse, parenting skills or a combination of these foci (Brown et al., 2021; Guo et al., 2016; Kim et al., 2019; Lam et al., 2008; Maguin, 1991; Mitrani et al., 2010; Stanger et al., 2011).

One of the child-focused studies could not be combined with the other 3 child-focused studies (dos Reis Viela et al., 2016) because the authors reported the number of children scoring in the clinical range for externalising problems rather than mean scores like the other three studies. The remaining 3 studies (Davis-Susser, 1990; Haight et al., 2010; Scheer, 1996) were meta-analysed with results indicating that child externalising problems, as reported by parents, did not differ between children who participated in a psychosocial treatment versus those that did not (see Figure 14; n = 85; SMD = 0.22, 95% CI −0.24, 0.67). There was no detected heterogeneity among the effect sizes (τ ² = 0.00; χ ² = 0.51, df = 2, p = 0.78; I² = 0%). For the dos Reis Viela et al. (2016) study, children engaged with the school-based intervention model treatment aiming to prevent substance use in children of substance misusing parents were less likely to obtain a score in the clinical range for externalising problems compared to children not engaged in TAU (n = 101; OR = 0.29, 95% CI 0.12 to 0.68).

OPEN IN VIEWER

Figure 14

Forest plot for externalising problems meta-analysis (post-intervention, child-focused interventions, parent-report).

Likewise, the meta-analysis of the 7 parent-focused studies (see Figure 15) found that child externalising problems did not differ between parents who engaged with a psychosocial treatment compared to TAU or no treatment (n = 281; SMD = 0.17, 95% CI −0.07, 0.41). There was no detected heterogeneity among the effect sizes (τ ² = 0.00; χ ² = 4.53, df = 6, p = 0.61; I² = 0%). For the Lam et al. (2008) study, female partners of the alcohol misusing fathers also completed the same measure for child externalising problems. The results suggested no differences between the psychosocial treatment that integrated individual treatment for fathers with parent-focused intervention with female partners compared to TAU (n = 20; SMD = 0. 37, 95% CI −0.51 to 1.26).

OPEN IN VIEWER

Figure 15

Forest plot for externalising problems meta-analysis (post-intervention, parent-focused interventions, parent-report).

Three studies tracked parent-reported externalising behaviour at short-term follow-up, approximately >3 months but ≤6-months after treatment completion. Mitrani et al. (2010) examined the impact of a family-focused Structural Ecosystems Therapy for substance misusing mothers with HIV. Guo et al. (2016) examined the effectiveness of supplementing housing with Ecologically-Based Treatment for homeless substance-misusing mothers. Lam et al. (2008) examined the impact of a parent-focused intervention comprised of individual therapy for fathers with children aged 8–12 years and behavioural couples therapy for their female partner that integrated parent-skills training. A meta-analysis of these studies (see Figure 16) found no differences in externalising problems between the intervention and TAU children at short-term follow-up (n = 101; SMD = 0.18, 95% CI −0.23 to 0.58). There was no detected heterogeneity among the effect sizes (τ ² = 0.00; χ ² = 0.28, df = 2, p = 0.87; I² = 0%). For the Lam et al. (2008) study, female partner reports suggested no differences in child externalising problem at short-term follow-up between the treatment compared to TAU (n = 20; SMD = 0.44, 95% CI −0.45 to 1.33). Mitrani et al. (2010) also included a child-report measure of externalising behaviours and there were no differences between groups at post-intervention (n = 24; SMD = −0.81, 95% CI −1.68 to 0.05) or short-term follow-up (n = 21; SMD = −0.23, 95% CI −1.24 to 0.77).

OPEN IN VIEWER

Figure 16

Forest plot for externalising problems meta-analysis (short follow-up, parent-focused interventions, parent-report).

Five studies also included a longer-term follow-up ≥6-months after a psychosocial treatment (Butz et al., 2001; Lam et al., 2008; Maguin, 1991; Mitrani et al., 2010; Sullivan, 1984). All but one study (Lam et al., 2008) examined the effectiveness of interventions that included both child and parent involvement, and integrated one or more components focused on parent-wellbeing, parenting, child health and development, child psychological wellbeing, and/or overall family functioning. None included a case management, legalistic, or child welfare component. A meta-analysis of these 4 conceptually similar studies (see Figure 17) indicated that child externalising problems, as reported by parents, did not differ for participants who engaged in a psychosocial treatment versus those who received TAU (n = 208; SMD = −0.07, 95% CI −0.55, 0.41). However, there was statistically significant and moderate heterogeneity among the effect sizes (τ ² = 0.14; χ ² = 7.77, df = 3, p = 0.05; I² = 61%). A single effect size for the Lam et al. (2008) study (see above) suggested no differences in externalising problems – as reported by participating fathers – between the intervention and TAU children at the 12-month follow-up point (n = 20; SMD = 0.40, 95% CI −0.48 to 1.29). There was also no difference between intervention and TAU children when the mother was the respondent (n = 20; SMD = 0.34, 95% CI −0.55 to 1.22).

OPEN IN VIEWER

Figure 17

Forest plot for externalising behaviours meta-analysis (long follow-up, parent-report).

Mitrani et al. (2010) also included a child-report measure of externalising behaviours and there were no differences between groups at long-term follow-up (n = 25; SMD = 0.13, 95% CI −0.68 to 0.95). A single study compared child externalising difficulties 18-months after treatment for NAS with either methadone or morphine (Flannery et al., 2020). The effect size for this study indicated no differences between the alternate treatments (n = 77; SMD = −0.03, 95% CI −0.48 to 0.42).

As there were too few studies to conduct subgroup analyses, we first assessed the 4 studies on the a priori moderators to identify any clear outliers or categories among the studies. All studies used either well-validated standardised measure or a measure with associated psychometric data to establish its reliability and validity. All interventions were implemented by clinicians, mostly with postgraduate training. Two studies only reported information on parental alcohol misuse (Maguin, 1991; Sullivan, 1984) but did not appear to exclude those using drugs, whereas the other two reported polysubstance use for parents (Butz et al., 2001; Mitrani et al., 2010). The intervention was implemented in the home for 1 study (Butz et al., 2001), but in a mixture of what appeared to be community settings for the other 3 studies (Maguin et al., 1991; Mitrani et al., 2010; Sullivan, 1984). In terms of child age, 3 studies included children over the age of 3 years (Maguin et al., 1991; Mitrani et al., 2010; Sullivan, 1984), yet the same study that was set in the home, also differed from the other 3 studies by including children only from birth (Butz et al., 2001). Removal of this study from the meta-analysis slightly increased the precision of the meta-analysis (n = 109; SMD = −0.26, 95% CI −0.80 to 0.28), but only slightly reduced heterogeneity (τ ² = 0.10; χ ² = 3.64, df = 2, p = 0.16; I² = 45%).

We then proceeded to examine other features of the studies, including the research design, study populations, and characteristics of the interventions. We identified that one of the studies was a non-randomised study and also implemented an intervention of a much shorter duration than the other three studies (Sullivan, 1984). Removal of this study from the meta-analysis increased the precision of the meta-analysis (n = 176; SMD = 0.17, 95% CI −0.13 to 0.46) and substantially reduced heterogeneity (τ ² = 0.00; χ ² = 1.91, df = 2, p = 0.38; I² = 0%). These sensitivity analyses suggest the heterogeneity in effect sizes for child externalising problems may be due to the nature of the evaluation methodology, age of the children involved, and/or features of the specific intervention approach.

Four studies also reported results for individual subscales that comprised the externalising composite from the Child Behaviour Checklist (CBCL) (Butz et al., 2001; Claussen et al., 2004; Kim et al., 2019) or other measures (Personality Inventory for Children; Sullivan, 1984). Butz et al. (2001) used a version of the CBCL for younger children which uses an aggressive and a destructive subscale to make up the externalising composite. At the post-intervention time-point (6 to 18-months after treatment completion), 3-year-old children of parents who engaged with a home-visiting intervention did not differ from parents who received TAU on aggressive behaviour (n = 100; SMD = 0.36, 95% CI −0.03 to 0.76) or destructive behaviour (n = 100; SMD = 0.12, 95% CI −0.27 to 0.52). Kim et al. (2019) used a version of the CBCL for older children which uses an aggressive and a rule-breaking subscale to make up the externalising composite. The rule-breaking subscale results are reported in the ‘Antisocial behaviour’ subsection below. On the aggression subscale at post-intervention, there was no difference between children of parents who participated in a Solutions Focused Behavioural Therapy intervention comprised of weekly individual parent sessions over the course of approximately 3-months versus those who received TAU (n = 67; SMD = 0.02, 95% CI −0.46 to 0.51). Similarly, the single study that compared a centre- or home-based child development intervention (described above, Claussen et al., 2004; n = 96) found that, at post-intervention, children did not differ on parent-rated destructive behaviours (SMD = −0.10, 95% CI −0.50, 0.30); aggressive behaviours (SMD = −0.18, 95% CI −0.58, 0.22); or disruptive behaviours (SMD = −0.36, 95% CI −0.76, 0.05). Sullivan (1984) examined the effectiveness of a therapeutic intervention for alcoholic parents compared to no treatment at 8-months follow-up. The intervention comprised 6 parent group therapy sessions, 6 child group therapy sessions, and 6 family therapy sessions focused on psychoeducation and supportive psychotherapy. There were no differences in the level of parent-reported levels of child aggression after participating in the intervention compared to the no treatment control group (n = 33; SMD = 0.64, 95% CI −0.08 to 1.36).

The single study using a teacher-report measure (Adelman, 1990) examined whether participating in a therapeutic group in the school setting impacted overall teacher-reported externalising behaviour (aggression) for children with alcohol misusing parents. At post-intervention children engaging with this intervention did not differ from children who did not (n = 44; SMD = 0.13, 95% CI −0.46 to 0.73).

Internalising

Fourteen studies examined the effectiveness of treatment using parent-, child-, or teacher-report - standardised measures that produce a composite for child internalising problems. Two studies had missing data, by either reporting no between group data or only data for participants with scores over a particular threshold on the outcome measure (Bartle-Haring et al., 2018; Margolis et al., 2017). Nine studies compared an eligible intervention to a TAU comparison condition at post-intervention only, 3 studies included both a post-intervention and short-term follow-up (≥3 months ≤6-months), and 4 studies also included a longer follow-up (>6-months ≤12-months). An additional two studies compared alternative treatments (Claussen et al., 2004; Flannery et al., 2020; described above). Further, two studies provided both a child- and parent-report measure of internalising problems (Luthar et al., 2007; Mitrani et al., 2010). However, only one of these reported sufficient data to calculate effect sizes (Mitrani et al., 2010).

Similar to externalising behaviours, 3 of the 9 studies using a post-intervention outcome measure were direct therapeutic interventions for children of substance misusing parents aged >5 but ≤10 years (Haight et al., 2010; Scheer, 1996; dos Reis Viela et al., 2016). These were analysed separately to the other 6 studies which did not integrate a child-focused treatment component described above (Brown et al., 2021; Guo et al., 2016; Kim et al., 2019; Lam et al., 2008; Mitrani et al., 2010; Stanger et al., 2011). The meta-analysis of the 2 child-focused studies indicated that child internalising problems, as reported by parents, did not differ between children who participated in a psychosocial treatment versus those that did not (n = 45; SMD = 0.15, 95% CI −0.54, 0.84). There was minimal detected heterogeneity among the effect sizes (τ ² = 0.06; χ ² = 1.29, df = 1, p = 0.26; I² = 23%). For the dos Reis Viela et al. (2016) study, children engaged with the school-based intervention model treatment aiming to prevent substance use in children of substance misusing parents were less likely to obtain a score in the clinical range for internalising problems compared to children not engaged in TAU (n = 101; OR = 0.15, 95% CI 0.06 to 0.39).

The meta-analysis of the 6 parent-focused studies (see Figure 18) found that child externalising problems did not differ between parents who engaged with a psychosocial treatment compared to TAU or no treatment (n = 220; SMD = 0.16, 95% CI −0.10, 0.43). There was no detected heterogeneity among the effect sizes (τ ² = 0.00; χ ² = 3.75, df = 5, p = 0.59; I² = 0%). For the Lam et al. (2008) study, female partners of the alcohol misusing fathers also completed the same measure for child internalising problems. The results suggested no differences between the psychosocial treatment that integrated individual treatment for fathers with parent-focused intervention with female partners compared to TAU (n = 20; SMD = 0.39, 95% CI −0.50 to 1.26).

OPEN IN VIEWER

Figure 18

Forest plot for internalising behaviours meta-analysis (post-intervention, parent-report).

Three studies—described in the previous section—tracked parent-reported internalising behaviour at short-term follow-up (Guo et al., 2016; Lam et al., 2008; Mitrani et al., 2010). A meta-analysis of these studies (Figure 19) found no differences in internalising problems between the intervention and TAU participants at short-term follow-up (n = 101; SMD = 0.23, 95% CI −0.17 to 0.65). For the Lam et al. (2008) study, female partner reports suggested no differences in child internalising problem at short-term follow-up between the treatment compared to TAU (n = 20; SMD = 0.33, 95% CI −0.55 to 1.22). Mitrani et al. (2010) also included a child-report measure of internalising behaviours and there were no differences between groups at post-intervention (n = 24; SMD = −0.51, 95% CI −1.36 to 0.33) or short-term follow-up (n = 21; SMD = −0.22, 95% CI −1.23 to 0.79).

OPEN IN VIEWER

Figure 19

Forest plot for internalising behaviours meta-analysis (short follow-up, parent-report).

Four studies also included a longer-term follow-up ≥6-months after a psychosocial treatment (Butz et al., 2001; Lam et al., 2008; Mitrani et al., 2010; Sullivan, 1984). All but one study (Lam et al., 2008) examined the effectiveness of interventions that included both child and parent involvement, and integrated one or more components focused on parent-wellbeing, parenting, child health and development, child psychological wellbeing, and/or overall family functioning. None included a case management, legalistic, or child welfare component. A meta-analysis of the 3 conceptually similar studies (see Figure 20) indicated that child internalising problems, as reported by parents, did not differ for participants who engaged in a psychosocial treatment versus those who received TAU (n = 158; SMD = −0.10, 95% CI −0.93, 0.74). Similar to the externalising meta-analysis which also contained these studies, there was statistically significant and high heterogeneity among the effect sizes (τ ² = 0.43; χ ² = 10.02, df = 2, p = 0.0007; I² = 80%). As there were too few studies to conduct subgroup analyses, we assessed the 3 studies on the a priori moderators to identify any clear outliers or categories among the studies. All studies used either well-validated standardised measure or a measure with associated psychometric data to establish its reliability and validity. All interventions were implemented in similar settings in the United States by clinicians usually with postgraduate training. Removal of the one study that was a quasi-experiment (Sullivan, 1984) increased the effect size and changed the direction to favouring treatment, and also reduced the confidence intervals and heterogeneity (n = 125; SMD = 0.31, 95% CI −0.05, 0.66; τ ² = 0.00; χ ² = 0.07, df = 1, p = 0.79; I² = 0%). A single effect size for the Lam et al. (2008) study, which examined the impact of a parent-focused intervention comprised of individual therapy for fathers with children aged 8–12 years and behavioural couples therapy for their female partner that integrated parent-skills training suggested no differences in internalising problems between the intervention and TAU children at the 12-month follow-up point (n = 20; SMD = 0.34, 95% CI −0.54 to 1.23). There was also no difference between intervention and TAU children when the mother was the respondent (n = 20; SMD = 0.37, 95% CI −0.40 to 1.15).

OPEN IN VIEWER

Figure 20

Forest plot for internalising behaviours meta-analysis (long follow-up, parent-report).

Three studies also reported results for specific internalising behaviour subscales that contribute to the internalising composite from the CBCL (Butz et al., 2001; Claussen et al., 2004; Kim et al., 2019) or the Personality Inventory for Children (Sullivan, 1984). At the post-intervention time-point (6 to 18-months after treatment completion), 3-year-old children of parents who engaged with a home-visiting intervention (Butz et al., 2001) differed from parents who received TAU on withdrawn behaviour (n = 100; SMD = 0.51, 95% CI 0.11 to 0.91), but not on anxious-depressed behaviour (n = 100; SMD = 0.35, 95% CI −0.04 to 0.75) or somatic complaints (n = 100; SMD = 0.15, 95% CI −0.24 to 0.54). Similarly, the study that compared a centre- or home-based child development intervention (described above, Claussen et al., 2004; n = 96) found that, at post-intervention, children did not differ on parent-rated anxious-depressed behaviours (SMD = 0.06, 95% CI −0.34, 0.46) withdrawn behaviours (SMD = −0.08, 95% CI −0.48, 0.32), or somatisation (SMD = 0.04, 95% CI −0.36, 0.44). Likewise, Kim et al. (2019) found that older children of parents who engaged with a Solution Focused Behavioural Therapy for substance misusing parents did not differ to children of parents assigned to TAU at post-intervention on withdrawn-depressive behaviours (n = 67; SMD = 0.02, 95% CI −0.47 to 0.50), anxious-depressive behaviours (n = 67; SMD = −0.03, 95% CI −0.52 to 0.45), or somatic complaints (n = 67; SMD = −0.08, 95% CI −0.59 to 0.43). Sullivan (1984) examined the effectiveness of a therapeutic intervention for alcoholic parents compared to no treatment at 8-months follow-up. The intervention comprised 6 parent group therapy sessions, 6 child group therapy sessions, and 6 family therapy sessions focused on psychoeducation and supportive psychotherapy. There were no differences in the level of parent-reported somatic concerns in their children (n = 33; SMD = −0.12, 95% CI −0.82 to 0.58) after participating in the intervention compared to the no treatment control group. However, there were higher levels of somatisation overall in the treatment versus control group (n = 33; SMD = 0.99, 95% CI 0.24 to 1.73).

Mitrani et al. (2010) also included a child-report measure of internalising behaviours and there were no differences between groups at long-term follow-up (n = 25; SMD = 0.01, 95% CI −0.74 to 0.27). A single study compared child internalising difficulties 18-months after treatment for NAS with either methadone or morphine (Flannery et al., 2020). The effect size for this study indicated no differences between the alternate treatments (n = 77; SMD = 0.02, 95% CI −0.42 to 0.47).

Prosocial behaviour

Four studies evaluated the impact of psychosocial interventions using a parent-report measure of child prosocial behaviours compared to either a TAU comparison condition (Dawe et al., 2007; Maguin, 1991; Margolis et al., 2017) or an alternate treatment (Claussen et al., 2004, described above). Margolis et al. (2017) did not report sufficient data to calculate effect sizes and also combined child- and parent-report measures into their summary statistics for their measure of prosocial behaviours. Maguin (1991) recruited alcohol misusing fathers and their spouses to evaluate an intervention that combined parent-skills training with couples-based therapy focused on substance misuse, parent wellbeing, and relationships issues. Dawe et al. (2007) recruited parents from methadone clinics to evaluate a comprehensive home-based intervention that integrated components focused on parental substance misuse, parental wellbeing (e.g., emotion regulation and mindfulness), parenting skills, quality of the caregiving relationship, and overall family functioning. Both studies also integrated case management. A meta-analysis of these two studies found that children of parents receiving integrated psychosocial interventions were observed to display more prosocial behaviours than children whose parents received TAU at post-intervention (n = 92; SMD = 0.47, 95% CI 0.06 to 0.89). This effect was not maintained at 6-month follow-up (n = 92; SMD = 0.57, 95% CI −0.16 to 1.30) and there was nonsignificant and moderate heterogeneity among the effect sizes (τ ² = 0.18; χ ² = 2.91, df = 1, p = 0.09; I² = 66%).

Antisocial behaviour

Eight studies included a measure of antisocial behaviour to evaluate psychosocial interventions (Catalano et al., 1999; Davis-Susser, 1990; Kim et al., 2019; Logan et al., 2001; Maguin, 1991; Peitler, 1980; Ryan et al., 2016; Sullivan, 1984), with all but four (Catalano et al., 1999; Logan et al., 2001; Ryan et al., 2016; Peitler, 1980) using parent-report instruments. Three studies did not report sufficient data to calculate effect sizes (Logan et al., 2001; Peitler, 1980; Ryan et al., 2016).

The study using the child-report measure (Catalano et al., 1999) examined the effectiveness of the Focus on Families intervention, an intensive 9-month family-focused intervention integrating home-based case management with weekly parent groups focused on parental substance misuse, parenting skills, emotion regulation, and strategies for preventing child substance misuse. Children also attended 12 of the 32 group sessions to facilitate skills practice. Children participating in Focus on Families did not differ from children receiving TAU on a 9-item scale measuring self-reported delinquency at the 6-month follow-up (n = 104; SMD = 0.34, 95% CI −0.05 to 0.73) or the 12-month follow-up (n = 100; SMD = 0.15, 95% CI −0.25 to 0.55).

The four studies with parent-report antisocial behaviour measures used measures of delinquency or antisocial behaviours that varied in seriousness. One reported a rule-breaking subscale from a standardised measure of child behavioural problems (Kim et al., 2019), another constructed a composite of delinquent behaviours from a range of standardised measures (Maguin, 1991), another reported a delinquency subscale from a comprehensive standardised measure of childhood personality features (Sullivan, 1984), and the final study (Davis-Susser, 1990) used a Conduct Disorder subscale from a validated measure (Revised Behaviour Problem Checklist). The study using the rule-breaking subscale (Kim et al., 2019) found no difference between those who participated in a Solutions Focused Behavioural Therapy intervention comprised of weekly individual parent sessions over the course of approximately 3-months versus those who received TAU (n = 67; SMD = 0.13, 95% CI −0.35 to 0.61). Similarly, the study using the composite (Maguin, 1991) found no difference in parent-reported child delinquent behaviours between those who received no treatment and those who participated in combined parent-skills training with couples-based therapy focused on substance misuse, parent wellbeing, and relationships issues (n = 51; SMD = 0.09, 95% CI −0.46 to 0.64). This study also included a follow-up measure that aligned with the follow-up measure of Sullivan (1984) (the post-intervention data was not reported for this study). This third study, which used the subscale from the standardised personality measure (Sullivan, 1984) evaluated a psychosocial intervention comprised of 6 parent, 6 child, and 6 family therapy sessions. A meta-analysis of these 2 studies indicated that parent-reported childhood delinquency does not differ between participants who receive no intervention versus those who engage with psychosocial interventions comprised of parent and child components at 6–8 months follow-up (n = 83; SMD = −0.23, 95% CI −0.66 to 0.21; τ ² = 0.00; χ ² = 0.09, df = 1, p = 0.77; I² = 0%). Davis-Susser (1990) examined the effectiveness of a child-focused group therapy programme and found that participating children did not differ in terms of conduct disorder symptoms compared to a control group not receiving therapy (n = 40; SMD = −0.30, 95% CI −1.02 to 0.42).

Substance use and related measures

Seven studies included a measure related to either child substance use (Bartle-Haring et al., 2018; Catalano et al., 1999; Gance-Cleveland et al., 2008; Logan et al., 2001) or knowledge and attitudes related to substance misuse (Broning et al., 2019; Larson, 1989; Sullivan, 1984). Three of these studies (Bartle-Haring et al., 2018; Gance-Cleveland et al., 2008; Logan et al., 2001) did not report sufficient data to permit effect size calculation at all (e.g., no between group outcome data or results of statistical tests to allow effect size calculation). Another of the studies did not report sufficient data to permit effect size calculation at 6- and 12-month follow-up (Catalano et al., 1999), but sufficient data at a 5-year follow-up. This study evaluated the Focus on Families intervention (see above) and found that the odds of having any type of substance abuse or dependence diagnosis did not differ between treatment and comparison conditions (see Table 9).

In relation to the studies examining child knowledge and/or attitudes about substance misuse, one found that children who engaged with the parent-child-family intervention (see previous section, Sullivan, 1984) did not differ on their self-reported knowledge and attitudes about substance use upon treatment completion (n = 33; SMD = −0.58, 95% CI −1.30 to 0.13) or approximately 8-months later (n = 33; SMD = −0.37, 95% CI −1.07 to 0.34). The remaining studies evaluated child-focused interventions. Broning et al. (2019) evaluated a group resilience programme for substance misusing parents (alcohol and drugs) and Larson (1989) evaluated a group psychodrama intervention for children of alcohol misusing parents. A meta-analysis of these studies indicated that children self-reported higher levels of knowledge about substance misuse upon treatment completion (n = 197; SMD = 0.78, 95% CI 0.46 to 1.05; τ ² = 0.00; χ ² = 0.38, df = 1, p = 0.54; I² = 0%). Based on the Broning et al. (2019) study only, this effect was maintained approximately 6-months later (n = 176; SMD = 0.46, 95% CI 0.15 to 0.76).

Anxiety

Seven studies included a measure of child anxiety to evaluate psychosocial interventions with either a teacher-report measure (Adelman, 1990), parent-report measures (Bartle-Haring et al., 2018; Pirnia et al., 2017a, 2017b; Sullivan, 1984), child-report measures (Lam et al., 2008; Scheer, 1996), or a mixture of both (Davis-Susser, 1990). Two studies did not report between-groups effect size data or results of statistical tests that could be used to calculate an effect size (Bartle-Haring et al., 2018; Prinia et al., 2017b).

The single study using a teacher-report measure (Adelman, 1990) examined whether participating in a therapeutic group in the school setting impacted overall teacher-reported anxiety for children with alcohol misusing parents. At post-intervention children engaging with this intervention did not differ from children who did not (n = 44; SMD = −0.10, 95% CI −0.70 to 0.49). Two of the three studies utilising a child-report measure examined the effectiveness of child-focused group therapy programmes compared to control groups not receiving therapy (Davis-Susser, 1990; Scheer, 1996), and a meta-analysis of these studies found that there were no differences between groups (n = 70; SMD = −0.44, 95%CI −0.95 to 0.08), with no heterogeneity detected (τ ² = 0.00; χ ² = 0.02, df = 1, p = 0.90; I² = 0%). The third study using a child-report measure (Lam et al., 2008) examined the impact of a parent-focused intervention comprised of individual therapy for fathers with children aged 8–12 years and behavioural couples therapy for their female partner that integrated parent-skills training. Single effect sizes suggested no differences in child-reported anxiety between intervention and TAU children at the post-intervention (n = 20; SMD = 0.51, 95% CI −0.38 to 1.40), 6-month follow-up (n = 20; SMD = 0.47, 95% CI −0.42 to 1.36), or 12-month follow-up points (n = 20; SMD = 0.48, 95% CI −0.41 to 1.37).

The 3 studies using parent-report measures of child anxiety could not be combined with meta-analysis due to conceptually different interventions and comparison conditions. As such, single effect sizes for each study are presented here and in Table 9. One study compared alternate treatments (Pirnia et al., 2017a) and found that, at post-intervention, child anxiety (as rated by parents) did not differ for mother-child dyads who participated in either a CBT group programme versus a Parent-Child Interaction Therapy (PCIT) intervention targeting high-potency cannabis users (n = 50, SMD = −0.44, 95% CI −1.00, 0.12). Although a waitlist control group was included in this study, the data for these participants was not reported to allow for a comparison between the treatments and the waitlist control participants. Sullivan (1984) evaluated a psychosocial intervention comprised of 6 parent, 6 child, and 6 family therapy sessions and found that those engaged in treatment did not differ from those receiving no treatment on parent-rated anxiety approximately 8-months post treatment completion (n = 33; SMD = −0.54, 95% CI −1.25 to 0.17). Likewise, Davis-Susser (1990) found no difference in parent-rated anxiety for children who participated in a group therapy programme for children of alcoholics compared to children receiving no treatment (n = 40; SMD = −0.28, 95% CI −1.00 to 0.44).

Depression

Twelve studies included a measure of child depression behaviour to evaluate psychosocial interventions. One study used a parent-report instrument (Sullivan, 1984) and 11 used child-report instruments. Six of the studies using child-report instruments either did not report outcome data for the depression measure or did not report sufficient data to calculate effect sizes (Bartle-Haring et al., 2018; Gance-Cleveland et al., 2008; Logan et al., 2001; Luthar et al., 2007; Peitler, 1980; Roosa et al., 1990). Four of the remaining five studies with a child-report measure of depression were all child-focused group psychosocial interventions (Adelman, 1990; Davis-Susser, 1990; Hyun et al., 2010; Scheer, 1996). However, Adelman (1990) used a measure of hopelessness (symptom of depression), whereas the other 3 studies used validated measures of multiple depressive symptoms. A meta-analysis of these 3 studies indicated symptoms of depression did not differ for children participating in a child-focused group therapy compared to children not receiving treatment (Figure 21; n = 98; SMD = −0.34, 95% CI −1.27 to 0.59). Substantial heterogeneity was detected between effect sizes (τ ² = 0.53; χ ² = 9.16, df = 2, p = 0.001; I² = 78%) and the only difference between the studies based on a priori moderators was that the study by Hyun et al. (2010) was conducted in Korea with only males, and was slightly longer in duration (10-weeks vs. 8-weeks for the other 2 studies). A sensitivity analysis removing this study found that symptoms of depression were significantly lower for children participating in a child-focused group therapy compared to children not receiving treatment (n = 70; SMD = −0.77, 95% CI −1.48 to −0.05) and the heterogeneity between effect sizes reduced (τ ² = 0.12; χ ² = 1.81, df = 1, p = 186; I² = 45%).

OPEN IN VIEWER

Figure 21

Forest plot for depressive symptoms meta-analysis (post-intervention, child-report).

Adelman (1990) found that children engaging with a school-based group therapeutic intervention had higher levels of hopelessness compared to children who did not participate (n = 44; SMD = −0.61, 95% CI −1.22 to 0.00). The remaining study using a child-report measure (Lam et al., 2008) examined the impact of a parent-focused intervention comprised of individual therapy for fathers with children aged 8–12 years and behavioural couples therapy for their female partner that integrated parent-skills training. Single effect sizes suggested no differences in child-reported depressive symptoms between intervention and TAU children at the post-intervention (n = 20; SMD = 0.55, 95% CI −0.35 to 1.45), 6-month follow-up (n = 20; SMD = 0.54, 95% CI −0.36 to 1.43), or 12-month follow-up point (n = 20; SMD = 0.51, 95% CI −0.39 to 1.40).

In relation to the studies with a parent-report measures, one study (Sullivan, 1984) found that child depressive symptoms were lower for those receiving no treatment and those who participated in a psychosocial intervention comprised of 6 parent, 6 child, and 6 family therapy sessions (n = 33; SMD = −0.81, 95% CI −1.53 to −0.08).

Trauma

Only one study reported a child trauma outcome measure. With a small RCT, Haight et al. (2010) examined the effectiveness of individual narrative-informed therapy with children affected by parental substance misuse. At post-intervention, children receiving treatment did not differ from children in a waitlist control group on a parent-report measure of child trauma symptoms (n = 15; SMD = 0.00, 95% CI −1.01 to 1.01).

Socioemotional functioning

Eight studies included measures of child socioemotional functioning or competencies, with 4 studies including child-report measures (Adelman, 1990; Broning et al., 2019; Luthar & Suchman, 2000; Roosa et al., 1990), three including parent-report measures (Barlow et al., 2019; dos Reis Viela et al., 2016; Finkelstein et al., 2005), and one including a child-, parent-, and teacher-reported versions of the same measure (Orte et al., 2008). Only two studies reported sufficient data to permit effect size calculation (Adelman, 1990; dos Reis Viela et al., 2016). The studies with missing data had issues ranging from missing standard deviations or insufficient results of statistical tests to permit alternative avenues effect size calculation (e.g., Luthar & Suchman, 2007; Orte et al., 2008), or incomplete outcome data due to the measure not being appropriate to a large portion of the study sample due to child age (Barlow et al., 2019). The studies by Finkelstein et al. (2005) and Orte et al. (2008) are included under this outcome category, however, the measures used by the authors contain multiple composites meaning that if the complete data had been reported, these studies would likely be included under multiple outcome domains. Broning et al. (2019) provided data for other outcomes upon author contact but were unable to provide data for all requested outcomes.

The dos Reis Viela et al. (2016) study found no differences in the odds of scoring in the clinical range for caregiver-rated total competencies for children engaging with a school-based intervention model providing mental health support and a range of other preventative activities than children assigned to the control condition (n = 101; OR = 2.00, 95% CI 0.76 to 5.29). Adelman (1990) found that children engaging with a school-based group therapeutic intervention did not differ on overall self-reported adjustment compared to children who did not participate (n = 44; SMD = −0.43, 95% CI −1.03 to 0.17).

Mood, psychological wellbeing, emotion regulation, and symptoms of psychopathology

Nine studies reported miscellaneous measures of mood, child psychological wellbeing, emotion regulation and/or symptoms of psychopathology. Two of these studies reported insufficient data for calculating effect sizes (Gance-Cleveland et al., 2008; Pirnia, 2019).

Broning et al. (2019) compared the effectiveness of a child-focused group resilience programme (TRAMPOLINE) to a play-based educational attentional control condition and reported multiple outcomes using numerous outcomes, including general psychological wellbeing and physical stress symptoms. The authors were able to provide missing data to permit effect size calculation for physical stress symptoms. Children participating in the TRAMPOLINE intervention reported lower levels of physical stress symptoms compared to children who engaged the play-based attentional control activities at post-intervention (n = 181; SMD = 1.46, 95% CI 0.32, 2.60) and at 6-months follow-up (n = 172; SMD = 2.25, 95% CI 1.03 to 3.47). Pirnia et al. (2019) also provided a measure of stress (cortisol) to evaluate the impact of PCIT compared to no treatment in mother-child dyads, yet reported no outcome data to permit effect size calculation.

Two studies assessed the effectiveness of child-focused therapy groups using child-reported measures of happiness (Hojjat et al., 2015; Gance-Cleveland et al., 2008), however only one study reported sufficient data to permit effect size calculation. Hojjat et al. (2015) found that adolescents with substance misusing parents who participated in a group assertiveness training programme had higher overall levels of happiness (n = 57; SMD = 0.67, 95% CI 0.14 to 1.21).

Two studies assessed the effectiveness of child-focused therapy groups using child-reported measures of anger (Adelman, 1990; Hojjat et al., 2017). A meta-analysis of these studies indicated no difference in levels of state anger (intensity at point in time) between treated and untreated children (n = 104; SMD = 1.42, 95% CI −1.94 to 4.77; τ ² = 5.73; χ ² = 47.40, df = 1, p < 0.00001; I² = 98%). Similarly, there was no difference in the levels of trait anger (predisposition towards anger) between treated and untreated children (n = 104; SMD = 1.39, 95% CI −2.15 to 4.93; τ ² = 6.38; χ ² = 51.70, df = 1, p < 0.00001; I² = 98%). Both meta-analyses had significant heterogeneity. We examined this heterogeneity based on a priori moderators and found that the studies differed on age of the participants. Hojjat et al. (2017) finding a large effect size favouring the treatment for adolescent males, whereas Adelman (1990) found no differences between treated and untreated primary school children. Single effect sizes for the specific subscales of the validated anger scale reported in Adelman (1990) also indicated no differences between treated and untreated children (see Table 9).

Davis-Susser (1990) found that psychotic-related behaviours did not differ between children who participated in a child-focused group therapy intervention versus those who did not (n = 40; SMD = −0.09, 95% CI −0.80, 0.63). Likewise, Kim et al. (2019) found no differences between participants who received a parent-focused Solution Focused Behavioural Therapy (Kim et al., 2019) and TAU on parent-reported child thought problems (n = 67; SMD = −0.26, 95% CI −0.74 to 0.23). However, Sullivan (1984) found that children who engaged with a psychosocial intervention comprised of 6 child, 6 parent, and 6 family therapy sessions had lower psychosis symptomatology compared to children who received no treatment (n = 33; SMD = −1.08, 95% CI −1.83 to −0.33).

Isolation and social support

Three studies evaluated group psychosocial interventions for children of substance misusing parents using either a measure of child-report social support (Broning et al., 2019), child-reported loneliness (Dore et al., 1999), or teacher-reported problematic isolation behaviours. Broning et al. (2019) and Dore et al. (1999) did not report sufficient data to calculate effect sizes. Adelman (1990) found no differences on teacher-reported problematic isolation behaviours for children engaging with a school-based group therapeutic intervention versus children who did not participate (n = 44; SMD = −0.03, 95% CI −0.63 to 0.56).

Coping and resilience

Twelve studies evaluated predominantly child-focused psychosocial interventions using measures under the broad category of coping and resilience. Under this category of outcomes, we included assertiveness, self-esteem, self-worth, self-concept, self-efficacy, locus of control, resilience, coping behaviours, and measures explicitly labelled resilience by authors. Many of the studies in this category had missing data that precluded effect size calculation and/or meta-analysis.

Hojjat et al. (2015) found that adolescents with substance misusing parents who participated in a group assertiveness training programme had higher levels assertiveness (n = 57; SMD = 1.13, 95% CI 0.57, 1.69) compared to adolescents receiving no treatment. Four studies included measures of child-reported locus of control (Davis-Susser, 1990; Dore et al., 1999; Larson, 1989; Sullivan, 1984), yet one did not report data to permit effect size calculation (Dore et al., 1999). At 8-months follow-up Sullivan (1984) found that children who engaged with a psychosocial intervention comprised of 6 child, 6 parent, and 6 family therapy sessions did not differ from children who received no treatment on their self-reported locus of control (n = 33; SMD = −0.06, 95% CI −0.75 to 0.64). A meta-analysis of the two studies evaluating child-focused group therapy interventions (Davis-Susser, 1990; Larson, 1989) found that treated children who participated in a child-focused therapy group did not differ from children who received no treatment on their self-reported locus of control (n = 49; SMD = −0.27, 95% CI −0.91 to 0.37; τ ² = 0.00; χ ² = 0.18, df = 1, p = 0.67; I² = 0%).

Six studies included child-report measures of self-concept (Adelman, 1990; Broning et al., 2019; Dore et al., 1999; Hyun et al., 2010; Scheer, 1996; Sullivan, 1984), yet only three of these reported outcome data and/or sufficient data to calculate effect sizes (Adelman, 1990; Hyun et al., 2010; Scheer, 1996). A meta-analysis of these three studies suggests that children participating in a child-focused group therapy programme did not differ in how positively they viewed themselves than children who did not receive the programme (Figure 22; n = 103; SMD = −0.38, 95% CI −1.32 to 0.57). There was significant heterogeneity between the effect sizes (τ ² = 0.56; χ ² = 10.56, df = 2, p = 0.005; I² = 81%) and the only clear difference between the studies based on a priori moderators was that Hyun et al. (2010) was conducted in Korea and the other two studies were conducted in the United States. Removal of the Hyun et al. (2010) study from the meta-analysis suggested that children participating in a child-focused group therapy programme tended to report a less positive view of themselves than children who did not receive the programme (n = 75; SMD = −0.81, 95% CI −1.29 to −0.34) and heterogeneity between effect sizes reduced (τ ² = 0.00; χ ² = 0.85, df = 1, p = 0.36; I² = 0%).

OPEN IN VIEWER

Figure 22

Forest plot for self-concept meta-analysis (post-intervention, child-report).

Four studies used self-esteem and/or self-worth measures to evaluate child-focused therapy models. Sorensen (1993) evaluated the impact of a child-focused group with adolescent boys in group foster care homes using an unstandardised measure of self-esteem. Peitler (1980) evaluated the effectiveness of a child-focused group for boys held at a hospital where fathers were being treated for alcoholism using a measure self-worth. Neither of these studies reported sufficient data to enable calculation of effect sizes.

Haley (1992) culturally adapted an existing measure of self-esteem to evaluate the impact of a child-focused therapeutic group at school with primary-school aged children of alcohol misusing parents. The single effect size for Hayley (1992) indicates that children who participated in the treatment had higher levels of self-esteem than children who did not participate (n = 60; SMD = 0.96, 95% CI 0.43 to 1.50). This effect was consistent for all but one subscale (social self-esteem) of this measure that represent different aspects of self-esteem (see Table 9). A fourth study by Hyun and colleagues (2010) evaluated the effectives of a group CBT programme for adolescent males using a measure of self-worth. The authors found that adolescents participating in the CBT group programme did not differ in levels of self-worth compared to adolescents receiving a ‘brief education on mental health’ (p. 204) at post-intervention (n = 28; SMD = 0.33, 95% CI −0.42 to 1.08).

Two studies evaluated psychosocial interventions using child-reported measures of resilience. The first was a longitudinal follow-up of the Focus on Families evaluation study approximately 10-years after the intervention (Skinner et al., 2009). The authors created a metric of functional resilience comprised of three items: (1) enrolment in school or being employed; (2) absence of substance abuse or dependence; and (3) lack of criminal justice involvement within the previous 5-years. Unfortunately, the authors did not report sufficient data to enable computation of an effect size for this study. The second study by Hyun and colleagues (2010) evaluated the effectives of a group CBT programme for adolescent males and found that adolescents participating in the CBT group programme did not differ in levels of resilience compared to adolescents receiving a ‘brief education on mental health’ (p. 204) at post-intervention (n = 28; SMD = 0.09 95% CI −0.65 to 0.83). Three studies also examined the effectiveness of child-focused psychosocial group programmes on children's self-reported coping, but none of the studies reported sufficient data to enable calculation of effect sizes (Broning et al., 2019; Gance-Cleveland et al., 2008; Roosa et al., 1990).

Sexual risk behaviours

One study examined the long-term effectiveness of the Focus on Families model (see above) on sexual risk-taking behaviours approximately 10-years after the intervention (Skinner et al., 2014). The authors created a dichotomous measure of risk based on whether the participant had not used condoms consistently during sexual encounters, if they reported more than two sexual partners in the last 12-months, or if they reported having sex outside a committed relationship. In addition, the authors measured participants' number of sexual partners, their sexual activity outside of a committed relationship, whether they had had sex in exchange for money or drugs, had received diagnosis of a sexually transmitted infection, or had a partner who was an intravenous drug-user in the previous 12-months. The authors found that the odds of any of the above sexual risks were no different between those who participated in Focus on Families compared to those who did not (see Table 9).

Wellbeing and functioning specifically linked to parental substance misuse

Two studies (Broning et al., 2019; Davis-Susser, 1990) used measures pertaining to child wellbeing and functioning specifically in relation to parental substance misuse to evaluate child-focused group therapies. Broning et al. (2019) measured coping behaviours relating to parental substance misuse, as well as child-reported distress and isolation relating to parental substance misuse. Data was provided by the authors for the latter two measures. Children participating in the TRAMPOLINE intervention did not differ from children who engaged the play-based attentional control activities on the measure of distress at post-intervention (n = 202; SMD = 0.13, 95% CI −0.15, 0.42) or at 6-months follow-up (n = 173; SMD = 0.23, 95% CI −0.08, 0.54). In comparison, children participating in the TRAMPOLINE intervention reported lower levels of isolation compared to children who engaged the play-based attentional control activities at post-intervention (n = 166; SMD = 0.55, 95% CI 0.23, 0.87), but this effect was not sustained at the 6-months follow-up (n = 164; SMD = 0.22, 95% CI −0.10, 0.54).

Davis-Susser (1990) examined whether a child-focused therapy group for children of alcoholics impacted coping behaviours relating to the alcoholic parent(s). At post-intervention, parents rated whether children appeared to demonstrate a range of coping behaviours (e.g., express their feelings about misuse). Children participating in the group programme did not differ in their coping behaviours in comparison to children who received no treatment (n = 40; SMD = −0.08, 95% CI −0.80, 0.64).