Multidimensional Family Therapy (MDFT) for Young People in Treatment for Non-opioid Drug Abuse: A Systematic Review

1.1 DESCRIPTION OF THE CONDITION

Youth drug abuse ¹ of the kind that persists beyond the experimentation phase is a severe problem worldwide (United Nations Office on Drugs and Crime (UNODC), 2010). Abuse of drugs such as cannabis, amphetamine, and cocaine, referred to in this review as non-opioids, are strongly associated with a broad range of negative health implications such as traffic accidents, sexually transmitted diseases, mental problems and suicide as well as social problems including poor academic achievement, delinquency and violent behavior (Bonner & van den Bree, 2009; Deas & Thomas, 2001; Essau, 2006; Lynskey & Hall, 2000; Office of National Drug Control Policy (ONDCP), 2000; Rowe & Liddle, 2006; Shelton, Taylor, Nordstrom & Levin, 2007).

While cannabis, amphetamine, cocaine and other non-opioid drugs remain illegal in most countries, surveys indicate widespread prevalence. In the US, 25.5 percent of 12^th-grade students report having used an illicit drug (any kind) within the last month (Johnston, O'Malley, Miech, Bachman & Schulenberg, 2014). In Canada, 21 percent of 15-24 year olds report having used of some kind of illicit drug within the last year (Health Canada, 2011). In Australia, seven percent 12-17 year olds report using some kind of drug within the last month (White & Smith, 2009). The European Monitoring Centre for Drugs and Drug Addiction has found that within Europe prevalence differs significantly from country to country but that overall around a quarter of Europeans report having used some kind of illicit drug in their lifetime (European Monitoring Centre for Drugs and Drug Addiction (EMCDDA), 2013).

The prevalence of specific kinds of illicit drug abuse varies significantly, with cannabis generally being the most commonly used drug. In the US, 22.7 percent of 12^th-grade students report having used marijuana/hashish (types of cannabis), 4.1 percent amphetamine, and 1.1 percent cocaine during the last 30 days before the National Survey on Drug Use conducted in 2013 (Johnston et al., 2014). The European Drug Report of 2013 indicates that 11.7 percent of the 15 to 34 year-olds in Europe has used cannabis, 1.3 percent amphetamine, and 1.9 percent used cocaine during the last year (EMCDDA, 2013).

Although not all drug users progress to severe drug abuse and dependence, some do and therefore warrant treatment (see e.g. Crowley, Macdonald, Whitmore & Mikulich, 1998). Individuals that warrant drug treatment are described variously as abusers, misusers or as dependent. These specific categorizations are used in the Diagnostic and Statistical Manual of Mental Disorders (DSM) (American Psychiatric Association, 1994, 2000). While DSM-IV is widely used, the International Statistical Classification of Diseases and Related Health problems (ICD, now ICD-10) developed by the World Health Organisation (WHO) is also in wide use. Differences between these rubrics concern both terminology and categorization criteria. For example, DSM-IV includes the category ‘abuse’, while ICD-10 explicitly avoids this term on the grounds of its ambiguity; harmful use and hazardous use are the equivalent terms in WHO usage, but the categories are not identical; and while ICD-10 uses only physical and mental criteria, DSM-IV also includes social criteria (WHO, 2011; Nordegren, 2002).

Research draws attention to the significant gap between the number of young people classified as in need of treatment and the number of young people who actually receive such treatment (SAMHSA, 2010; National Survey on Drug Use and Health (NSDUH), 2007). In the US, for example, 7.2 million young people are classified as needing treatment for illicit drug abuse, but only 1.4 million of these actually receive treatment at a specialty facility for an illicit drug abuse problem (SAMHSA, 2011). The treatment usually provided to young people is delivered in outpatient settings. Accordingly, 90 percent of the 89,521 clients under age 18 registered in substance abuse treatment in 2012 by SAMHSA were in outpatient treatment, which is the same proportion as the total treatment population (SAMHSA, 2013). Equal proportions of the clients under age 18 were enrolled in facilities with a primary focus on substance abuse treatment and in facilities whose primary focus were provision of a mix of mental health and substance abuse treatment services; this differs from the total treatment population as youth tend to be treated in dual focus facilities more often than adults (SAMHSA, 2013). Cognitive-behavioral therapy and motivational interviewing are specific therapeutic approaches that are used to some extent by most (respectively 91 and 87 percent) treatment facilities (SAMHSA, 2013).

There is growing public concern about the effectiveness and high cost of available treatments for young people, and the high rates of treatment dropout and post-treatment relapse to drug abuse (Austin, Macgowan & Wagner, 2005; Najavits & Weiss, 1994; Stanton & Shadish, 1997). While relapse must be acknowledged as an expected part of any treatment process targeting individual drug abuse, efforts should be made to make treatment as attractive, accessible and relevant as possible for young people in order to minimize the risk of unwarranted dropout and continuous relapse (Simmons et al., 2008; National Institute on Drug Abuse (NIDA), 2009). Furthermore, the services provided should be empirically supported to increase the likelihood that (a) treatment will be successful, and (b) public spending supports the interventions that are the most effective.

Researchers point to the fact that many research projects claim to have empirically validated different kinds of treatment approaches for young drug abusers (e.g. Austin et al., 2005; Rowe & Liddle, 2006; Waldron, Turner & Ozechowski, 2006; Williams, Chang & Addiction Centre Adolescent Research Group, 2000). The current dilemma in the field of youth substance use treatment is that it is not clear what works best and for whom as the research suggests that a number of interventions lead to reduced drug abuse (Waldron & Turner, 2008). Treatments identified as promising are individually based cognitive and motivational therapies including Cognitive Behavioral Therapy, Multisystemic Therapy, and Family therapies (Deas & Thomas, 2001; Galanter & Kleber, 2008; Kaminer, 2008; Waldron & Turner, 2008).

Family therapy covers a range of different interventions, and is based on different manuals and varying theoretical sources such as behavioral and cognitive behavioral theory, structural and strategic family theory, and family systems theory (Williams et al., 2000; Austin et al., 2005). Family-based interventions for the treatment of young drug abusers include Multidimensional Family Therapy, Brief Strategic Family Therapy, Functional Family Therapy and Family Behavior Therapy (Waldron & Turner, 2008; Austin et al., 2005; Rowe & Liddle, 2006; Alexander & Sexton, 2002; Waldron et al., 2006; Williams et al., 2000). Some reviews suggest that these family-based therapies are superior to individual-based programs in reducing youth drug abuse (Williams et al., 2000; Lipsey et al., 2010; Waldron, 1997).

Young people who abuse drugs persistently and to an extent that warrants treatment have unique needs due to their particular cognitive and psychosocial developmental stage. Recognizing that young people are particularly sensitive to social influences, families and peer groups being highly influential, authorities such as the U.S.

National Institute on Drug Abuse recommends that youth drug treatments facilitate positive parental and peer involvement (Institute on Drug Abuse (NIDA), 2009: 22). Moreover, they recommend that other systems in which the youth participates (such as schools and athletics) are also integrated into a comprehensive treatment approach to meet the unique needs of young drug abusers (ibid. 23). A number of studies and reviews show positive results for family therapies in general, but there is a need to synthesize individual study results for specific family therapies to determine whether and to what extent specific family therapy interventions work for young drug abusers (Williams et al., 2000; Austin et al., 2005; Waldron & Turner, 2008; Kaminer, 2008; Deas & Thomas, 2001).

This review has explored the specific family-based intervention of Multidimensional Family Therapy (MDFT) (Liddle, 2002; Liddle et al., 2001; Liddle, Rowe, Dakof, Henderson & Greenbaum, 2009) as aggregated evidence for MDFT's effects is needed. The review has attempted to clarify the effects of the MDFT program for relevant groups of young people aged 11-21 living with their families, and has focused on young people enrolled in treatment for drug abuse, independent of how their problem was labeled. Enrolment in drug treatment indicates that the severity of the young person's drug abuse has caused the young person or a significant adult close to the young person (such as teacher, parent, social services worker, school counselor) to seek treatment. The review focused on MDFT delivered as outpatient treatment ² and primarily on non-opioid drug abuse ³ ; it is one in a series of reviews on different manual-based family therapy interventions for young people in treatment for drug abuse ⁴ .

1.2 DESCRIPTION OF THE INTERVENTION

Multidimensional Family Therapy (MDFT) has evolved over the last twenty years and is a manual-based, family-oriented treatment designed to eliminate drug abuse and associated problems in young people's lives (Liddle, 1999; Liddle, 2002; Liddle et al., 2009). MDFT is one of several family therapy forms that meet the general characteristics of manual-based family therapies in that it deals with young people and their families as a system throughout treatment, and thereby recognizes the important role of the family in the development and treatment of young people's drug abuse problems (Liddle et al., 2001; Muck et al., 2001).

MDFT is designed to take into account a number of risk and protective factors, and it acknowledges that young people's drug abuse is linked to multiple dimensions: home life, friends, school and community (Liddle et al., 2004). As such it advocates that a multi-dimensional approach is needed to resolve the young person's problematic drug abuse, and therefore aims to modify multiple domains of functioning by intervening with the young person, family members, and other members of the young person's support network (Austin et al., 2005). This also means that MDFT is based on multiple therapeutic alliances; with the young drug-user, his or her parents and other family members, and perhaps school and juvenile justice officials. While some young people have only a single parent and few significant others relevant to therapy, others might have two sets of parents and many significant others relevant to therapy, and the therapist must organize the treatment accordingly.

Treatment focuses on individual characteristics of the young person, their parents, and other key individuals in the young person's life, as well as on the relational patterns contributing to the drug abuse and other problem behaviors. A variety of therapeutic techniques are used to accomplish this and to improve the young person and the family's behaviors, attitudes, and functioning across the variety of domains (Liddle, 1999). MDFT aims to reorient the young person and family toward a more functional developmental trajectory on the basis a variety of key principles, including: 1) Individual biological, social, cognitive, personality, interpersonal, familial, developmental, and social ecological aspects can all contribute to the development, continuation, worsening and chronicity of drug problems; 2) The relationships with parent(s), siblings and other family members are fundamental domains of assessment and change; 3) Change is multifaceted, multi-determined and relates to the youths' cognitive and psychosocial developmental stages; 4) Motivation is not assumed, but is malleable and motivating the young person and family members about treatment participation and change is a fundamental therapeutic task; 5) Multiple therapeutic alliances are required to create a foundation for change; and 6) Therapist responsibility and attitude is fundamental to success (Liddle, 2010).

MDFT's theory of change hypothesizes that changing the family system constructively will produce changes in youths' drug abuse (reduction or elimination) as well as improvements in relation to other emotional and behavioral problems (Hogue, Liddle, Dauber, & Samoulis, 2004, Liddle et al, 2005). Specifically, MDFT process studies have emphasized the importance of addressing cultural themes and increasing youths' participation in treatment (Jackson-Gilford, Liddle, Tejeda & Dakof, 2001) and improving parents' skills (Schmidt, Liddle & Dakof, 1996). Moreover, the quality of therapeutic alliances between both the therapist and the youth, and the therapist and parent(s) are emphasized as decisive for the production of change (Robbins et al. 2006).

1.3 HOW THE INTERVENTION MIGHT WORK

Overall, MDFT proposes to produce positive changes through working simultaneously with different systems -- inside and outside the family -- to end drug abuse and related problems (Liddle, 2002; Rowe, 2010). MDFT outlines a three-phase model that explicates how the intervention is to be administered. Stage 2 is highlighted as the working phase of treatment, where significant change attempts are made within and across systems (Liddle et al., 2005).

The working phase of MDFT includes an adolescent module, a parent module and a family interaction module. In the adolescent module (Liddle et al., 2005, p. 140-141), the therapy seeks to produce change through a range of techniques such as helping the youth examine positives and negatives about their drug abuse as well as helping the youth articulate hopes and dreams for the immediate and long term. Overall, the module aims to help the youth see that they will have difficulties in achieving the things they say they want as long as current problem behavior continues, and the therapist helps the youth create concrete pathways toward a change of lifestyle.

In the parent module (Liddle et al., 2005, p. 141-142), the therapy seeks to produce change by, for example, clarifying and responding to parents' needs (e.g. for extra psychiatric service), instilling hope that change is possible (e.g. through bringing small signs of change to their attention), and advising parents -- respectfully, but in direct terms - on how to handle challenging situations. Overall, the module aims to improve parents' skills, such as their abilities to practice age-proportionate limit-setting and to enforce house rules, including defining both sanctions and rewards for adherence.

In the family interaction module (Liddle et al., 2005, p. 143), to the therapy seeks to produce change through helping families to establish more positive patterns of interaction, e.g. improving their communication and conflict resolution skills and their understanding of the importance of establishing positive, supportive relationships.

1.4 WHY IT IS IMPORTANT TO DO THIS REVIEW

Persistent drug abuse among young people is a problem worldwide as it causes a range of health problems and social problems. Drug treatment targeting young drug abusers is challenging and costly as interventions are often plagued by high dropout rates and post-treatment relapse into drug abuse. Research suggests that almost half of the young drug abusers who receive drug treatment do not complete that treatment (Substance Abuse and Mental Health Services Administration, 2008). While relapse must be acknowledged as an expected part of any treatment process targeting individual drug abuse, there is a need to identify effective treatments for addressing young people's drug abuse problems in order to minimize unwarranted treatment dropout and continuous post-treatment relapse (Simmons et al. 2008, NIDA 2009). Furthermore, the growing interest among policy makers in increasing funding for evidence-based interventions is a strong motivation to add to the evidence base with a systematic review on a promising treatment for young drug abusers.

Previous reviews (Vaughn & Howard, 2004; Waldron & Turner, 2008; Becker & Curry, 2008) indicate that MDFT is a promising treatment for young drug abusers. However, the only meta-analysis thus far conducted (Waldron & Turner, 2008) included MDFT as part of a broad category of family therapy rather than including MDFT as a distinct treatment model. In contrast, this review examines the effect of MDFT and by aggregating individual studies' results on MDFT, and so contributes to the knowledge about treatment of young drug abusers and their families.

The review informs practice by exploring whether results indicate that MDFT works better for some client groups than others based on characteristics such as age, gender, minority background, family composition (e.g., single parents), and co-occurring conditions. As previous reviews (e.g. Waldron & Turner 2008) indicate that individual treatment outcomes vary widely within intervention models, it is important to investigate who might benefit the most from MDFT. The hypothesis is that MDFT is not similarly efficacious for all client groups and the review investigates whether it is possible to identify subgroups that benefit more than others.

3.1 TITLE REGISTRATION AND REVIEW PROTOCOL

The title for this systematic review was registered in The Campbell Collaboration on 28.07.2011. The review protocol was registered on 01.09.2012. Both the title registration and the protocol are available at: http://www.campbellcollaboration.org/library.php.

3.2 CRITERIA FOR CONSIDERING STUDIES FOR THIS REVIEW

3.4.6 Dealing with missing data and incomplete data

We assessed missing data and recorded attrition rates in the included studies. We were able to discern reasons for attrition from two of the studies (Liddle, 2009; Rigter 2011). None of the studies reported an intention-to-treat analysis, although Dennis 2004 provided data for all but two of the participants.

3.4.7 Assessment of heterogeneity

Heterogeneity among primary outcome studies was assessed with Chi-squared (Q) test, and the I-squared, and τ-squared statistics (Higgins, Thompson, Deeks, & Altman, 2003). Any interpretation of the Chi-squared test was made cautiously on account of its low statistical power.

3.4.8 Assessment of publication bias

Reporting bias refers to both publication bias and selective reporting of outcome data and results. We were unable to comment on the possibility of publication bias because there were insufficient studies for the construction of funnel plots. Selective reporting has been considered in the risk of bias assessment and any concerns reported in section 4.3.4.

3.5.1 Subgroup analysis, moderator analysis and investigation of heterogeneity

We did not identify sufficient studies to allow any moderator analysis to be conducted.

3.5.2 Sensitivity analysis

Sensitivity analysis was used to evaluate whether the pooled effect sizes were robust across components of risk of bias. We conducted a sensitivity analysis for the incomplete outcome data and other bias components of the risk of bias checklists by removing studies scoring 4 (see section 3.4.3 for a definition).

4.1 RESULTS OF THE SEARCH

We ran the searches in June 2011 and October 2014. The results are summarised in the flow chart on page 107.

We identified 6,519 potential relevant records after excluding duplicates (database search, 1,425; grey search, 898; hand search, snowballing and other resources, 4,196).

All 6,519 records were screened based on title and abstract. Of these, 170 were retrieved and screened in full text. One hundred and fifty four records did not fulfil the screening criteria and were excluded. One paper from the snowball search was included. A total of five unique studies, reported in 16 papers, was included in the review. Further details of the included and excluded studies are provided in section 9.

4.2 DESCRIPTION OF THE STUDIES

4.3.1 Allocation

All five included studies were randomized controlled trials. The methods of randomization were described in four studies (Rigter 2011, Liddle 2008, Dennis 2004 and Liddle 2009) and the method used for sequence generation was judged adequate in each case. Concerning allocation concealment, only one study mentioned the method used to achieve this (Dennis 2004), and one study simply stated that the allocation was concealed (Rigter 2011). It was not possible to judge whether concealment of allocation was adequate in the three remaining studies.

4.3.2 Blinding

As is common in social intervention, it is generally impossible to blind participants or those delivering the interventions. Only two studies clearly stated that outcome assessors were blinded to allocation status (Liddle 2001 and Liddle 2009).

4.3.3 Incomplete outcome data

Two studies had very low levels of missing data (Dennis 2004 and Liddle 2009), and two studies had relatively high levels (Liddle 2001 and Liddle 2008). All except one study (Liddle 2001) dealt with missing data, typically using latent growth curve modelling (however, the estimates of the latent growth curve models are not included in the meta-analyses in this review). Only one study (Dennis 2004) imputed missing data (see section 9.3).

4.3.4 Selective reporting

All studies reported data on the primary outcome reduction of substance abuse. Although Liddle 2009 did not provide data that enabled the calculation of effect sizes because only means and standard deviations of the natural log transformed variables were provided, the non-transformed data were kindly provided by Professor Howard Liddle. Two studies did not report all the secondary outcomes as planned in the protocol (Dennis 2004 and Rigter 2011). We could not locate a protocol for the remaining three studies.

4.3.5 A priori protocol and analysis plan

We were able to locate a protocol and an a priori analysis plan for two studies (Dennis 2004 and Rigter 2011).

4.3.6 Other potential sources of bias

The predictability of treatment assignment is an issue for all constrained randomization algorithms. This is particularly the case when blocked randomization or an urn procedure is used, and when the assignments are revealed subsequent to the person recruiting into the trial, it is sometimes possible to predict a future assignment, which implies a risk of bias ¹⁰ (Higgins & Green, 2008; Lachin, Matts & Wei, 1988; Berger, 2005). Consequently, none of the studies were rated 1 (low risk of bias) on this domain; Dennis 2004, Liddle 2008 and Rigter 2011 because block randomization was used, Liddle 2009 because an urn procedure was used, and Liddle 2001 because the randomisation procedure was not described. Confounding was not relevant in the review since we did not find any NRCTs meeting the inclusion criteria.

The RoB tool is provided in Section 12.3 and the detailed RoB assessment is given in Section 9.3. In the 5 point scale, 1 corresponds to Low risk of bias and 5 corresponds to High risk of bias.

4.4.1 Primary outcome results

It was possible to measure reduction of drug abuse by drug abuse frequency reduction as well as by reduction in drug abuse problem severity. One study, Liddle 2001, provided a drug abuse classification scheme as the only measure of drug abuse reduction. We judged that the drug abuse classification scheme compared best with the measures of drug abuse problem severity provided in the other studies.

All five studies thus provided data that enabled the calculation of comparable effect sizes on drug abuse problem severity reduction, while four studies provided data that enabled the calculation of comparable effect sizes on drug abuse frequency reduction. Drug abuse frequency and problem severity reduction are measured as decreases; hence a negative effect size favours MDFT.

Drug use problem severity 6 months post-intake

Three studies found statistically significant differences between treatments, all favouring MDFT (Liddle 2001, Liddle 2008 and Liddle 2009). In two studies there was no evidence of a statistically significant difference between treatments (Rigter 2011, and Dennis 2004). Pooled results showed a statistically significant effect of MDFT for drug abuse problem severity reduction. The random effects weighted standardised mean difference (SMD) for analysis 1A was -0.35 (95% CI -0.59 to -0.11, p=0.004), for analysis 1B SMD=-0.33 (95% CI -0.59 to -0.08, p=0.01), for analysis 2A SMD=-0.31 (95% CI -0.53 to -0.10, p=0.004) and for analysis 2B SMD=-0.30 (95% CI -0.53 to -0.07, p=0.01). There was a statistically significant heterogeneity of effects between studies in analysis 1A (τ²=0.04, Q= 10.12, df=4, p=0.04), analysis 1B (τ²=0.05, Q= 11.67, df=4, p=0.02) and analysis 2B (τ²=0.04, Q= 9.64, df=4, p=0.05). In analysis 2A there was no statistical significant heterogeneity (τ²=0.03, Q= 8.18, df=4, p=0.09). The forest plots are displayed in Figure 4.1–Figure 4.4.

OPEN IN VIEWER

4.1

Drug abuse problem severity. 6 months post-intake 1A

OPEN IN VIEWER

4.2

Drug abuse problem severity. 6 months post-intake 1B

OPEN IN VIEWER

4.3

Drug abuse problem severity. 6 months post-intake 2A

OPEN IN VIEWER

4.4

Drug abuse problem severity. 6 months post-intake 2B

Drug abuse problem severity 12 months post-intake

Two studies, Rigter 2011 and Liddle 2009, found a statistically significant difference between treatments, favouring MDFT, whereas in Dennis 2004, Liddle 2001, and Liddle 2008, the effect was not statistically significant. Rigter 2011 found an effect on drug abuse problem severity at 12 months, but not at 6 months; we have no explanation for this anomalous result.

Pooled results showed a statistically significant effect of MDFT for drug abuse problem severity reduction. The random effects weighted standardised mean difference (SMD) for analysis 1A was -0.25 (95% CI -0.39 to -0.10, p=0.0007), for analysis 1B SMD=-0.23 (95% CI -0.39 to -0.06, p=0.007), for analysis 2A SMD=-0.27 (95% CI -0.43 to -0.11, p=0.001) and for analysis 2B SMD=-0.25 (95% CI -0.43 to -0.07, p=0.007). Heterogeneity of effects among studies was not statistically significant in analysis 1A (τ²=0.00, Q= 4.19, df=4, p=0.38), analysis 1B (τ²=0.01, Q= 5.26, df=4, p=0.26), analysis 2A (τ²=0.01, Q= 4.97, df=4, p=0.29) and analysis 2B (τ²=0.01, Q= 6.17, df=4, p=0.19). The forest plots are displayed in Figure 4.5–Figure 4.8.

OPEN IN VIEWER

4.5

Drug abuse problem severity. 12 months post-intake 1A

OPEN IN VIEWER

4.6

Drug abuse problem severity. 12 months post-intake 1B

OPEN IN VIEWER

4.7

Drug abuse problem severity. 12 months post-intake 2A

OPEN IN VIEWER

4.8

Drug abuse problem severity. 12 months post-intake 2B

Drug abuse frequency 6 months post-intake

Four studies reported data on the drug abuse frequency reduction. Three studies found a statistically significant effect that favoured MDFT (Liddle 2009, Rigter 2011 and Liddle 2008). In Dennis 2004, MDFT was not significantly different from the MET/CBT5 treatment or from ACRA treatment. Pooled results showed a statistically significant effect of MDFT for drug abuse frequency reduction. The random effects weighted standardised mean difference (SMD) for analysis A was -0.24 (95% CI -0.43 to -0.06, p=0.01), and for analysis B SMD=-0.25 (95% CI -0.40 to -0.11, p=0.0007). Heterogeneity of effects among studies was not statistically significant in analysis A (τ²=0.01, Q= 4.63, df=3, p=0.2) and analysis B (τ²=0.00, Q= 3.11, df=3, p=0.37). The forest plots are displayed in Figure 9 and Figure 4.10.

OPEN IN VIEWER

4.9

Drug abuse frequency. 6 months post-intake A

OPEN IN VIEWER

4.10

Drug abuse frequency. 6 months post-intake B

Drug abuse frequency 12 months post-intake

Two studies (Rigter 2011 and Liddle 2009) found a statistically significant effect that favoured MDFT, whereas the differences between treatments in Liddle 2008 and in Dennis 2004 were not statistically significant. Pooled results showed no statistically significant effect of MDFT for drug abuse frequency reduction. The random effects weighted standardised mean difference (SMD) for analysis A was -0.28 (95% CI -0.63 to 0.07, p=0.11), and for analysis B SMD=-0.28 (95% CI -0.63 to 0.07, p=0.11). There was a statistically significant heterogeneity of effects among studies in analysis A (τ²=0.10, Q= 15.43, df=3, p=0.001) and analysis B (τ²=0.10, Q= 15.45, df=3, p=0.001). The forest plots are displayed in Figure 4.11 and Figure 4.12.

OPEN IN VIEWER

4.11

Drug abuse frequency. 12 months post-intake A

OPEN IN VIEWER

4.12

Drug abuse frequency. 12 months post-intake B

4.4.2 Secondary outcome results

Family functioning

It was not possible to perform a meta-analysis on family functioning. One study, Liddle 2001, used a rating scale that assessed the degree of family functioning from 1 (optimal functioning) to 10 (severely dysfunctional). Family functioning was measured as a decrease; hence a negative effect size favours MDFT. Results showed no significant differences between MDFT and the MEI treatment either at 6 months post-intake, SMD =0.25 (95% CI -0.23, 0.73), or at 12 months post-intake, SMD =-0.34 (95% CI -0.82, 0.15). Results showed no significant differences between MDFT and the AGT treatment at 6 months post-intake but a significant difference which favored MDFT at 12 months post-intake; SMD =-0.30 (95% CI -0.80, 0.21) at 6 months post-intake and SMD =-1.26 (95% CI -1.81, -0.70) at 12 months post-intake, respectively.

Liddle 2009 measured family functioning in terms of family interactions, measuring them separately as positive and as negative family interactions respectively. Results showed no significant differences between MDFT and the peer group comparison; SMD =-0.09 (95% CI -0.52, 0.34) at 6 months post-intake, SMD =0.13 (95% CI -0.30, 0.56) at 12 months post-intake for positive family interactions, and SMD = -0.26 (95% CI -0.69, 0.17) at 6 months post-intake, and SMD =-0.25 (95% CI -0.69, 0.18) at 12 months post-intake for negative family interactions.

Rigter (2011) reported on family functioning measured in terms of conflict and cohesion ¹¹ . Results showed no significant differences between MDFT and the TAU comparison; SMD =0.14 (95% CI -0.07, 0.35) at 6 months post-intake, SMD =0.19 (95% CI -0.01, 0.41) at 12 months post-intake for family conflict, SMD =-0.15 (95% CI -0.36, 0.06) at 6 months post-intake, and SMD =-0.16 (95% CI -0.36, 0.05) at 12 months post-intake for family cohesion.

OPEN IN VIEWER

4.13

Family functioning.

OPEN IN VIEWER

4.14

Positive family interaction.

OPEN IN VIEWER

4.15

Negative family interaction.

OPEN IN VIEWER

4.16

Family conflict.

OPEN IN VIEWER

4.17

Family cohesion.

Grade point average 6 months post-intake

Grade point is measured as a decrease; hence a negative effect size favours MDFT. One study (Liddle 2009) found statistically significant differences that favored MDFT, whereas the differences in Liddle 2001 study were not statistically significant. Pooled results did not show a statistically significant effect of MDFT for school grade improvement when using the AGT comparison in the Liddle 2001 and a marginal statistically significant effect when using the MEI comparison in Liddle 2001. The pooled estimate were SMD =0.38 (95% CI -0.25, 1.01) and SMD =0.47 (95% CI 0.01, 0.92). There was no statistically significant heterogeneity between studies in any of the analyses (τ²=0.15, Q= 3.54, df=1, p=0.06 and τ²=0.05, Q= 1.96, df=1, p=0.05). However, effect sizes differ markedly and with only two studies the power to detect heterogeneity is very low.

OPEN IN VIEWER

4.18

Grade point average. 6 months post-intake 1

OPEN IN VIEWER

4.19

Grade point average. 6 months post-intake 2

Retention

We used the information reported in all five studies that came closest to the proportion of cases with full completed treatment. Results were measured as odds ratio (OR) non-event, implying that an OR of less than one favours MDFT.

Three studies (Liddle 2008, Liddle 2001 and Dennis 2004) found no difference between retention rates. In the remaining two studies, the difference between retention rates between treatments favored MDFT (Rigter 2011 and Liddle 2009) and was statistically significant. Note that the magnitudes of the effect sizes of these two studies and the width of the confidence intervals were quite distinct from the three other studies (Liddle 2001, Liddle 2008 and Dennis 2004).

Pooled results showed a statistically significant effect of MDFT for retention when using the AGT comparison in Liddle 2001, but the effect was not statistically significant when using the MEI comparison in Liddle 2001.

The random effects weighted odds ratio (OR) for analysis 1A was 0.44 (95% CI 0.21 to 0.94, p=0.03), for analysis 1B OR=0.45 (95% CI 0.21 to 0.95, p=0.04), for analysis 2A OR=0.48 (95% CI 0.22 to 1.05, p=0.07) and for analysis 2B OR=0.49 (95% CI 0.22 to 1.07, p=0.07). There was statistically significant heterogeneity of effects among studies in analysis 1A (τ²=0.54, Q= 21.92, df=4, p=0.0002), analysis 1B (τ²=0.55, Q= 22.15, df=4, p=0.0002), analysis 2A (τ²=0.60, Q= 23.80, df=4, p<0.0001) and analysis 2B (τ²=0.60, Q= 23.97, df=4, p<0.0001). The forest plots are displayed in Figure 4.18–Figure 4.21.

OPEN IN VIEWER

4.20

Retention.1A

OPEN IN VIEWER

4.21

Retention.1B

OPEN IN VIEWER

4.22

Retention.2A

OPEN IN VIEWER

4.23

Retention.2B

Risk behavior

One study, Rigter 2011, provided data on externalizing disorders/symptoms (e.g. aggression and delinquency) measured by the Youth Self-report and the Child Behavior Checklist. Results showed no significant differences between MDFT and TAU on either of the scales; SMD =-0.02 (95% CI -0.23, 0.18) at 6 months post-intake and, SMD =-0.02 (95% CI -0.22, 0.19) at 12 months post-intake for the YSR scale; SMD =-0.01 (95% CI -0.20, 0.22) at 6 months post-intake and SMD =-0.02 (95% CI -0.23, 0.19) at 12 months post-intake for the CBCL scale.

OPEN IN VIEWER

4.24

Externalizing symptoms.

One other study, Liddle 2009, provided delinquency measures. No significant differences between MDFT and peer group were found at either at 6 months post-intake, SMD =-0.28 (95% CI -0.71, 0.16), or at 12 months post-intake, SMD =-0.33 (95% CI -0.76, 0.10).

OPEN IN VIEWER

4.25

Delinquency

4.4.3 Sensitivity analysis

Sensitivity analyses were performed for the primary outcomes; drug abuse problem severity and drug abuse frequency. We examined the robustness of conclusions when the studies scoring 4 on the incomplete outcome data item (Liddle 2001 and Liddle 2008) and the other bias item (Liddle 2008 and Rigter 2011) respectively were removed from the analyses. The results are displayed in table 4.4.1 and 4.4.2.

For drug abuse problem severity, the SMD remains statistically significant and still favours MDFT for most of the comparisons at the 12 post-intake when the studies scoring 4 are removed. However, the effect becomes no longer significant at 6 months post-intake when studies scoring 4 on the incomplete outcome data item are removed. The relative reduction remains significant and becomes larger for two of the contrasts (1A & 2A) at 6 months post-intake when studies scoring 4 on the other bias item are removed.

In both cases the relative reduction becomes marginally smaller at 12 months post-intake. As expected, when studies are removed from the analysis the confidence intervals become wider; there is however considerably overlap between confidence intervals (a graphical display is given in section 11). At 6 months post-intake, the results (with one exception ¹² ) are no longer statistically significant whereas they are still statistically significant at the 12 post-intake (with two exceptions ¹³ ).

For drug use frequency, the SMD ceases to be statistically significant for most of the comparisons when studies scoring 4 are removed.

At 6 months post-intake, the results are no longer statistically significant when the comparison used in Dennis 2004 is MET/CBT5 whereas they are still statistically significant when the comparison used in Dennis 2004 is ACRA.

5.1 SUMMARY OF THE MAIN RESULTS

Our main objective was to evaluate the current evidence on the effect of MDFT on drug abuse reduction for young people in treatment for non-opioid drug abuse. Further objectives of this review were to examine the moderators of drug abuse reduction effects and to examine whether MDFT works better for particular groups. Unfortunately it was not possible to assess this second review objective because of the limited number of studies.

Five randomized controlled trials of MDFT met the inclusion criteria for this review. All five studies compared MDFT to other treatments, mainly CBT or ‘CBT-informed’ interventions, thus it was not possible to analyze the absolute effect of MDFT. Two studies, Liddle 2001, and Dennis 2004, had two comparison groups with different individuals, and the Rigter 2011 study was carried out in five different countries with treatment as usual as the comparison condition, which varied across countries. The comparison treatments in the five studies were: Cognitive Behavioral Therapy (CBT), peer group, treatment as usual (TAU), adolescent group therapy (AGT)/multifamily educational therapy (MEI) and Motivational Enhancement Therapy/Cognitive Behavioral Therapy (MET/CBT5)/Adolescent Community reinforcement approach (ACRA).

The outcomes were reported at varying time points. We grouped the outcomes at 6 and 12 months post-intake. As duration of treatment in two studies was 3-4 months and in three studies was 4-6 months, these time points were approximately equivalent to termination and 6 month follow-up respectively.

In summarising these results, we have chosen to interpret a value of the standardised mean difference, SMD=0.20 as a small effect size, as is the general practice (Cohen, 1988).

5.2 OVERALL COMPLETENESS AND APPLICABILITY OF EVIDENCE

The number of included studies in this systematic review is small; only five trials met the inclusion criteria and examined whether MDFT reduced youth drug use. Three of these were conducted by MDFT program developers, one was conducted by an independent investigator with a program developer as a co-author and one study was conducted by independent investigators.

It was not possible to analyze the absolute effect of MDFT. All five studies compared MDFT to other interventions, mainly CBT or ‘CBT-informed’ treatments. Two studies had two comparison groups with different individuals and different treatments, while the study which was carried out in five different European countries had variations in comparison treatments across countries.

The five trials were reported in 16 papers, and we looked for and extracted relevant outcome data in all 16 papers. Studies of secondary analyses of the included trials (Hogue, Dauber, Samuolis, & Liddle, 2006; Hogue et al. 1998; Henderson, Dakof, Greenbaum, & Liddle, 2010; Henderson, Rowe, Dakof, Hawes, & Liddle, 2009; Shelef, Diamond, Diamond, & Liddle, 2005) did not provide any additional data for this review because they analysed subgroups of participants and/or explored for factors which moderate treatment effects.

It was possible to analyse outcomes at 6 and 12 month post-randomization (approximately equivalent to 6 and 12 months post-intake), although not all studies provided data at the longer follow up which would have allowed for documentation of accumulated or longer-term effects.

Whilst it was possible to include all studies in the analysis of the primary outcome reduction of substance abuse, reporting of secondary outcome data was very limited.

Four of the five studies originated from North America. This may limit the applicability of the evidence to a specific social and cultural setting, which in turn may be difficult to translate to other settings. This indicates a need for more well-conducted studies of CBT interventions in countries other than the USA.

5.3 QUALITY OF THE EVIDENCE

All five included studies were randomized controlled trials but none can be characterised as a robust RCT with low risk of bias on all assessed risk of bias items. One of the five included studies provided insufficient information on core issues to allow us to assess the risk of bias and had a significant level of missing data, making us question its validity.

Where effects were statistically significant, there was overall consistency in the direction of treatment effects regarding primary outcome (drug problem severity and drug abuse frequency) with all treatment effects favouring MDFT. Similarly, where effects on retention were statistically significant, MDFT was the favoured treatment. However the overall result must be interpreted with caution as two studies stand out from the others, with large effect sizes and very wide confidence intervals, and the estimated between study variation is quite large.

5.4 POTENTIAL BIASES IN THE REVIEW PROCESS

The narrow search strategy performed in this review may limit the likelihood of finding all relevant studies. However, we attempted to minimize the risk of missing relevant studies by conducting an extensive search for grey literature, an extensive hand searching and by contacting international experts within the field of MDFT; indeed, the large number of grey literature and hand searched literature that has been assessed for relevance attests to this effort.

5.5 AGREEMENTS AND DISAGREEMENTS WITH OTHER STUDIES OR REVIEWS

We identified five narrative reviews and three quantitative reviews on MDFT treatment for drug-using youth.

The identified narrative reviews (Thompson, Pomeroy, & Gober, 2005; Austin, Macgowan, & Wagner, 2005; Szapocznik, Lopez, Prado, Schwartz, & Pantin, 2006; Liddle, 2010; and Rowe, 2010) report a general pattern of a positive effect for MDFT treatment for drug-using youth. Thompson et al., 2005; Austin et al., 2005, and Szapocznik et al., 2006 review several interventions for drug-using youth, and with the exception of Thompson et al., 2005, their conclusions concerning MDFT are based on one study (Liddle 2001). In contrast, Liddle, 2010 and Rowe, 2010, focus on MDFT and base their conclusions on a greater number of studies.

We identified three quantitative reviews (Bender, Tripodi, Sarteschi, & Vaughn, 2011; Waldron & Turner, 2008; Vaughn et al., 2004). Vaughn et al., 2004, examined several interventions for drug-using youth and the conclusions concerning MDFT were based solely on Liddle 2001 with the authors reporting “evidence of clinically meaningful effect (Evidence Summary [ES] > .20) with at least 1 year follow-up or replication and using relatively strong designs”.

Waldron & Turner, 2008, used meta-analysis to evaluate family therapy, CBT (individual and group), and ‘minimal treatment control conditions' for drug-using youth. They concluded that “MDFT, FFT and group CBT emerged as well-established models for substance abuse treatment.” based on the studies by Liddle 2001, Liddle 2008, Liddle 2009 and Dennis 2004.

Bender et al., 2011 used meta-analysis to evaluate family therapy and individual therapy for drug-using youth. The authors reported that “Interventions with large effect sizes (>.80) included several family-based or multisystem approaches (Integrated Family, CBT, MDFT, Teaching family).” based on the studies by Liddle 2001, Liddle 2008 and Liddle 2009.

Consistent with our expectations, the apparent feedback from the three reviews is that more research is needed, not least with regard to moderators and identification of which particular subgroups of youth may be more likely to respond to specific interventions, as well as how treatments can be adapted or tailored to the individual needs of a young person to improve drug abuse outcomes. These are similar issues to those we planned to assess in our review. However, the lack of empirical evidence obscured the possibility of assessing moderators of effect and effects on subgroups.

6.1 IMPLICATIONS FOR PRACTICE

Although most of the few available studies on effectiveness are characterized by methodological problems, the results of this review suggest that MDFT seems to ‘work’ in the sense that the intervention results in a slightly higher reduction of drug abuse for the average individual compared to the average individual receiving other active treatments. Besides knowledge of whether a certain intervention ‘works’, in the sense that it is effective for the average individual, practitioners need knowledge about potential differential effects on treatment of highly relevant participant characteristics such as age, gender, minority background, family composition (e.g., single parents), and co-occurring conditions.

However, the participant characteristics listed above are potential predictors of treatment outcome, and practitioners need to be able to tailor the program to particular types of young drug abusers. It is not possible to identify which particular subgroups of youth may be more likely to respond to specific interventions, and subsequently how treatments could be adapted or tailored to the individual needs of a young person until results of additional MDFT outcome studies are available.

Programmes for drug using youth are costly, and it is also possible that such initiatives have potential cause damage to some individuals. The available evidence points to small effect sizes of MDFT in comparison to other treatments. Taking the individual variation in treatment effects into consideration, we cannot rule out the possibility that MDFT may be counterproductive for some individuals. It is therefore crucial to know more about its effectiveness in order to determine where money is best allocated as well as to understand which form of support would benefit young drug abusers most.

6.2 IMPLICATIONS FOR RESEARCH

It is important to address the need for more research in the field. A small body of evidence exists in relation to the treatment of young drug abusers, with only a very modest number of controlled evaluations of treatments for this group. Most of the few available studies of effectiveness have methodological problems, such as small sample size and varied methods of assessing drug abuse; such problems make definitive conclusions difficult, if not impossible. Well-designed, randomized controlled trials within this population are needed and should be reported clearly in accordance with the principles of the CONSORT 2010 statement. In addition, longer follow up data should be available in future studies.

It is also important to consider the possibility of any adverse effects these interventions might have. The popular belief is that MDFT, as well as other family therapy approaches, is harmless, but very little research has been conducted that focuses on the potential harm of such family therapy approaches.