The first summary review of the effects of youth psychotherapy appeared over 50 years ago. The author, Eugene Levitt, reviewed the existing evidence base of 18 studies treating “neuroses” in youth and came to a startling conclusion. There was little to no empirical evidence to suggest that psychotherapy was beneficial, and, indeed, the recovery rate for child and adolescent psychotherapy might be marginally worse than the simple improvement observed with the passage of time. In combination with the Eysenck (4) review of adult therapy, the Levitt (5) review and followup report (6) produced intense debate in the field. Many of the critiques focused on the methodological weaknesses of the psychotherapy studies that served as a basis for the author’s negative conclusions. Early therapy efficacy studies typically: a) failed to randomly assign youths to treatment and control conditions; b) used nonequivalent comparison groups, such as therapy dropouts, as control conditions; c) failed to specify what therapy procedures were used in the intervention being tested; d) allowed therapists or other nonblind raters to assess outcome; and e) enrolled very heterogeneous samples of youth in terms of diagnoses and developmental level (see Kazdin (7)
for review). These characteristics of early efficacy research substantially weakened the internal validity of the designs and made it difficult to impossible to interpret results, whether positive or negative.

In response to these critiques, the design of the prototypical therapy efficacy study evolved into that of the experimental clinical trial, characterized by explicit inclusion and exclusion criteria, random assignment, blinded and standardized diagnostic assessment, and manualized treatments. Within two decades, the evidence base of 18 studies available to Levitt had expanded to include over 100 clinical trials, and this explosion of therapy research coincided with the advent of modern metaanalytic methods for summarizing research findings. Together, these two developments provided a unique opportunity to revisit the conclusions of Levitt (5,6), and the 1980s and 1990s were marked by the publication of several major metaanalyses on the efficacy of therapy in youth.

In meta-analysis, traditional narrative reviews of a research area are supplemented by a quantitative analysis of empirical findings across studies. As in a narrative review, meta-analysis begins with an exhaustive, well documented literature search, with predetermined criteria for study inclusion and exclusion (e.g., requiring a minimum sample size). Following this collection of studies, researchers develop a coding scheme to capture the critical characteristics of each study, establish the reliability of the system, and code the findings from each investigation. Next, the empirical results of each investigation are transformed into a common metric of effect sizes, and these effect sizes form the unit of analysis for subsequent statistical tests. Statistical analyses range from simple estimates of a population effect size in a set of homogenous studies to multivariate models designed to explain variability in effect sizes across a complex literature.

In psychotherapy meta-analyses, the most common effect size metric is Cohen’s d (8). If the relevant summary statistics are reported in the published study, d is very simply calculated by taking the mean of the treated group on a measure of interest (e.g., depression symptoms at post-treatment), subtracting it from the mean of the control group, and then dividing this difference by the standard deviation of the control group (see Smith, Glass, and Miller (9) for other estimation techniques). This process creates a score indicating how “far apart” in outcomes a therapy condition is from a control group, expressed in standard deviation units. By convention, a d of 0.2 is considered a small effect size, while a d of 0.8 is a large effect (10).

Casey and Berman (11) were the first to apply meta-analytic methods of this kind to the child psychotherapy literature. They reviewed and coded 64 controlled studies of therapy for youth (age 12 and under) and found that the average effect size for psychotherapy was a very respectable 0.71. A series of more comprehensive and complex meta-analyses followed the publication of the Casey and Berman (11) report, and the findings told a similar, positive story (12,13,14). Overall, youth therapy reliably had medium to large effects on symptoms, with results similar in magnitude to effect size evidence from the adult literature (9). Furthermore, the confidence intervals around these population estimates did not cross zero, suggesting that, as a whole, psychotherapeutic interventions for youth were more efficacious than control conditions (typically a wait list or no treatment control).

In addition to supporting the overall benefit of child treatment, meta-analyses examining moderators of efficacy suggested that some youths may profit from therapy more than others, and that some treatments may have larger effects. Notably, type of youth problem did not emerge from these analyses as being significantly related to the magnitude of therapeutic improvement (12,14). Youths with internalizing problems, such as depression and anxiety, showed the same positive symptom gains as youths with externalizing behavior problems. More recent meta-analyses have added some caveats to this finding (15), but the preponderance of evidence across the literature indicated that therapy “worked” for most major childhood psychiatric problems. Controlling for problem type and other confounding factors, Weisz and colleagues (14) found that adolescent girls were particularly likely to do well in psychotherapeutic treatment, although this analysis was, necessarily, at the level of the study rather than at the level of the individual. Gender was coded as the proportion of youths in the sample who were male, and developmental level keyed in a similar fashion. Thus, as with all meta-analytic findings, results are best interpreted as indicating that studies that include large numbers of adolescent girls have stronger findings than studies that do not. Notably, other theoretically interesting youth predictors of treatment success have been difficult to test, due to their spotty reporting in the clinical trial literature. For example, in a recent review of the methodological characteristics of youth treatment research, it was reported that less than half of investigations provided information on the ethnicity of the study sample and a mere 25% provided any information on socioeconomic status (16).

In terms of treatment factors, it appears that for children and adolescents, behavioral interventions may be more effective, on average, than nonbehavioral psychotherapies (11,12,14). In the metaanalyses making this comparison, “behavioral treatments” typically include such direct behavioral techniques as teaching parents more effective discipline styles or developing anxious children’s relaxation skills, as well as cognitive-behavioral therapies (CBT), such as helping depressed youth to label and correct unrealistically negative thinking and self-talk. Nonbehavioral psychotherapy has been conceived as a broad category including traditional psychodynamic-based approaches, client-centered therapies, and discussion groups. Notably, this division does not characterize well several treatment programs of recent vintage that have significant empirical support, such as interpersonal therapy (IPT) for adolescents with depression (see Chapter 6.2.3) or multisystemic therapy (MST), a behavioral–family systems approach for juvenile offenders (see Chapter 6.3.2). Also note that some treatment modalities used with youth, such as family (Chapter 6.2.6) or group (Chapter 6.2.5) work, could be coded as either behavioral or nonbehavioral depending on the content of the intervention.

Not surprisingly, the finding of superior outcomes for behavioral treatments has been hotly contested, and a host of alternate explanations proposed (17). However, this result has been robust in analyses controlling for other potentially confounding differences between behavioral and nonbehavioral studies that might spuriously produce this effect, such as differences in problem type, severity of symptoms, or methodology (see Weiss & Weisz (18) for discussion). Two critiques of this effect cannot be tested in the current clinical trial literature base and remain as possible alternate explanations of the superiority of behavioral treatments: a) far more clinical trials of behavioral therapies have been published, and additional research on nonbehavioral therapies may yet yield more positive results; and b) nonbehavioral therapies have been used as control conditions in some studies by investigators, and these treatments may not have been implemented with the same care and vigor as the main behavioral treatment being tested by the study, artificially lowering estimates of nonbehavioral therapy effect sizes (19). These possibilities speak to the central thesis of Chapter 6.2.4 on psychoanalysis and psychodynamically
informed psychotherapies, and the growing efforts within the analytic research community to conduct high-quality clinical trials of insight-oriented therapeutic approaches (20,21). However, psychoanalysis and other nonbehavioral therapies may have a high hill to climb, given the volume of research on behavioral treatments and the consistency of positive effects, particularly for exposure-based treatments for the anxiety disorders (22,23) and behavioral parent training for child oppositional behaviors (24). The hill also may have been made steeper by the trend in the early 1990s for professional organizations in the mental health field to move beyond broad examination of the efficacy of therapy to begin identifying specific evidence-based treatment (EBT) programs for targeted diagnostic clusters (see Chapter 2.1.2).

In the early 1990s, evidence-based medicine was broadly defined as the practice of weighing the available scientific evidence when making decisions regarding clinical care (25). At around the same time, the Society of Clinical Psychology (American Psychological Association [APA] Division 12) formed a task force charged with “educating clinical psychologists, third party payors, and the public about effective psychotherapies” (26). This effort resulted in a series of reports identifying evidence-based psychosocial treatments (EBT) for adults (26,27,28). Parallel efforts by the Society of Clinical Child and Adolescent Psychology (APA Division 53) and the Society of Pediatric Psychology (APA Division 54) led to the publication of major EBT reviews for children and adolescents (29,30). In addition to these efforts driven by professional organizations, several other research teams have endeavored to identify EBTs for psychological problems (22,31,32,33).

As reviewed by Chambless and Ollendick (34), different groups use varying criteria when evaluating the strength of the support for an intervention’s efficacy. In general, however, treatments are considered a well established EBT if they have shown positive effects in a series of carefully controlled, prospective studies by at least two independent teams of investigators. Most often, this is defined as clinical trials in which: a) participants were randomly assigned to conditions, and b) treatment was compared to a placebo or other established treatment. However, some groups consider a large series of well controlled single case studies to be sufficient for this designation. A second class of EBTs, identified as “probably efficacious” (28), is supported by scientific evidence but has been subjected to less rigorous tests (e.g., comparison to a wait-list control group).

Based on these criteria, EBTs have been identified for a wide range of social, emotional, and behavioral difficulties commonly experienced by children and adolescents. To date, well established psychosocial interventions have been developed for conduct disorder, enuresis, and phobias, and probably efficacious EBTs have been identified for other anxiety disorders (e.g., separation anxiety) and depression (34). As discussed in the section on metaanalysis, behavioral treatments are eight to 10 times more likely to be tested systematically than nonbehavioral approaches (16), and it is not surprising that the majority of EBTs identified by the various work groups are behaviorally focused. Although the majority of therapeutic interventions focus on the child, many efficacious programs target the systems in which youths function, such as their families (functional family therapy) or the juvenile justice system [MST (35)].

Currently, the EBT movement has reached beyond the professional exercise of identifying therapies with promising outcomes and moved into the domain of policy, by funding and shaping the content of youth mental health care in practice. Policymakers at the national (36,37,38) and international (39) levels have endorsed the importance of evidence-based mental health care, and U.S. federal funding has been made available to clinical care providers to support the training of community therapists in EBT programs (40,41). Family and patient organizations have begun to advocate for access to mental health interventions with demonstrated effectiveness (42), and states have developed initiatives to support the use of EBT services (43). However, while these efforts may be seen as laudable in their attention to findings on the efficacy of psychotherapy, they have proceeded, in large part, with very little evidence on the actual effectiveness of therapy in practice.