Well-established treatments have the highest level of empirical support, requiring at least two high-quality (e.g., random assignment, adequate sample size) between-groups trials by different investigative teams showing that treatment is superior to placebo or another treatment, or equivalent to an already established treatment. Probably efficacious treatments require only one high-quality trial comparing treatment to placebo (or alternative treatment) or two trials comparing treatment to no treatment. Finally, possibly efficacious treatments have at least one study showing the treatment to be efficacious but not meet criteria as well-established or probably efficacious.

The second set of criteria, summarized in Table 2, is from Nathan and Gorman (2002, 2007) and was used to evaluate the methodological robustness of a study. Type 1 study designation requires random assignment to treatment conditions, clear inclusion and exclusion criteria, blinded assessments (i.e., assessor or informant was unaware of treatment assignment), “state-of-the-art” diagnostic methods (operationalized here as the use of valid and/or reliable measures), adequate sample size (operationalized as 12 participants per condition; Kazdin & Bass, 1989),² and clearly described statistical methods. Type 2 studies included clinical trials that were missing one or more elements of a Type 1 study. Nathan and Gorman (1998) also described Type 3, 4, 5, and 6 studies; however, these criteria were not applied to this review because such studies have serious methodological flaws (e.g., no comparison group).

To evaluate treatments for ethnic minority youth, several additional factors were considered. These features were established solely for this review and do not represent any organization's (e.g., APA) official guidelines for classifying treatments as evidence-based for ethnic minorities. After EBT criteria were met, an intervention was considered well-established, probably efficacious, or possibly efficacious for ethnic minority youth if supporting studies met one or more of three conditions listed in Table 2 as “additional considerations.” The first was based on the proportion of ethnic minority participants included in the study. Eligibility was met if at least 75% of participants in the EBT study were ethnic minorities (Condition A). Although lower thresholds have been used by some reviewers (e.g., 50% cutoff by Tobler, 1997; 60% cutoff by S. J. Wilson, Lipsey, & Soyden, 2003), the 75% threshold used here (representing a 3:1 ratio of ethnic minority to nonminority participants) provided stronger evidence that treatment effects were applicable to minorities. If most participants were not ethnic minorities, however, a treatment could still meet EBT criteria if either separate analyses with the subset of ethnic minority participants demonstrated superiority of treatment over control/comparison conditions (Condition B), or analyses showed ethnicity did not statistically moderate treatment outcomes (or treatment was efficacious for ethnic minorities despite “ethnicity-as-moderator” effects; Condition C). Thus, statistical evidence that ethnic minority participants benefited from treatment (or did not differ from nonminorities in terms of treatment benefit) was considered when making determinations about EBT status.

Column 5 describes the main findings and corresponding effect size coefficients, but only for those outcomes directly relevant to referral problems (e.g., if youth were referred for anxiety disorders, outcomes representing posttreatment fear or internalizing symptoms would be presented, but externalizing symptoms would not). However, when youth were referred for unspecified and/or a broad array of problems, outcomes for all youth symptoms were presented (e.g., Rowland et al., 2005; Weiss, Harris, Catron, & Han, 2003). Finally, column 6 specifies the EBT classification status, type of study (1 or 2 based on Nathan & Gorman, 2002), and which ethnic minority eligibility criteria were met. Note that no treatments summarized in this review met criteria as well-established for ethnic minority youth.

Four clinical trials support the efficacy of MST with African American juvenile offenders (Borduin et al., 1995; Henggeler, Clingempeel, Brondino, & Pickrel, 2002; Henggeler, Melton, & Smith, 1992; Henggeler, Melton, Brondino, Scherer, & Hanley, 1997). Compared to usual services and individual therapy, MST led to greater reductions in re-arrests and time incarcerated. These effects lasted as long as 13.7 years posttreatment (Schaeffer & Borduin, 2005), and youth ethnicity (African American vs. European American) did not moderate outcomes (Borduin et al., 1995; Henggeler et al., 2002; Henggeler et al., 1992; Schaeffer & Borduin, 2005). Although MST efficacy was also established by independent research teams in the United States and Norway (Ogden & Halliday-Boykins, 2004; Timmons-Mitchell, Bender, Kishna, Mitchell, 2006), neither trial assessed whether ethnic minorities benefited.

Lochman's Coping Power program (in various formats) is similarly efficacious with aggressive, African American youth (Lochman, Curry, Dane, & Ellis, 2001). Coping Power (the child-only version) involves social problem solving, positive play, group-entry skills training, and training for coping with negative emotions. In their first ethnic minority-focused trial, Lochman, Coie, Underwood, and Terry (1993) found that Social Relations Training (an early version of Coping Power) led to greater improvement than no treatment control for aggressive-rejected African American youth. In subsequent trials (Lochman & Wells, 2003, 2004), youth in the Coping Power intervention (adapted to include behavioral parent training) again showed greater improvement than either treatment as usual or no treatment. Moreover, results showed that ethnicity did not moderate treatment effects for most outcomes (Lochman & Wells, 2003, 2004).

Brief Strategic Family Therapy (BSFT; Szapocznik, Hervis, & Schwartz, 2003) may be the only efficacious treatment designed for Latino youth (primarily Cuban) with conduct problems. Based on the family systems work of Salvador Minuchin (Minuchin & Fishman, 1981), BSFT adopts strategies such as joining, reframing, and boundary shifting to restructure problematic family interactions of externalizing youth and their parents. Over the past two decades, Szapocznik and colleagues have carried out an extensive program of research testing the efficacy of various forms of BSFT including one-person BSFT (Szapocznik, Kurtines, Foote, Perez-Vidal, & Hervis, 1983, 1986), Bicultural Competence Training (Szapocznik, Rio, et al., 1986), Family Effectiveness Therapy (Szapocznik, Santisteban, et al., 1989), and standard BSFT (Santisteban et al., 2003; Szapocznik, Rio, et al., 1989). However, only three trials evaluated BSFT's efficacy relative to either a placebo or waitlist control. Two of these studies showed that BSFT was superior to control (Santisteban et al., 2003; Szapocznik, Santisteban, et al., 1989). In a third, process-oriented evaluation, BFST was not superior to a recreational comparison control (Szapocznik, Rio, et al., 1989).

MST, Coping Power (with parent training component), and BSFT all have been validated in two or more clinical trials with ethnic minority youth, although no replications by independent investigators have been carried out with minorities. Thus, MST and Coping Power (with parent training) are probably efficacious for African American youth whereas BSFT is probably efficacious for Hispanic youth.

Similarly, Trauma-Focused Cognitive-Behavioral Therapy (TF-CBT; Deblinger & Heflin, 1996) is efficacious for trauma-exposed ethnic minority youth. TF-CBT is a 12-session parent- and child-focused treatment involving psychoeducation, coping skills training, gradual exposure, cognitive processing of the abuse experience, and parent management training. In a multisite evaluation for sexually abused youth with posttraumatic stress disorder (PTSD), Cohen, Deblinger, Mannarino, and Steer (2004) found TF-CBT led to greater PTSD symptom reduction than child-centered therapy, although ethnicity (White vs. non-White [70% African American]) was not a significant moderator of treatment (Cohen et al., 2004; J. A. Cohen, personal communication, June 2004). Because all well-established criteria were met except replication by an independent investigator, TF-CBT is probably efficacious for ethnic minority youth.

To shed light on this debate, 13 studies were examined that evaluated ethnicity as a treatment moderator in the context of a randomized controlled trial (Table 6). A treatment moderator is defined as a pretreatment variable that has an interactive effect with treatment condition on clinical outcomes (Kraemer et al., 2002). With regard to ethnicity, significant Treatment Condition × Ethnicity interaction effects would generally indicate that treatment was more efficacious for one ethnic group than for another.

Although most studies summarized in Table 6 did not report significant moderator effects, five studies did show that ethnicity influenced treatment outcomes. Surprisingly, three studies suggested that identical treatments may show stronger effects for ethnic minority youth compared with European American youth (Arnold et al., 2003; Huey et al., 2004; Weiss, Catron, Harris, & Phung, 1999), whereas two treatments favored European American youth over ethnic minorities (Lochman & Wells, 2004; Rohde, Seeley, Kaufman, Clarke, & Stice, 2006). Yet this summary does not fully convey the complexity of these moderator findings. For example, although Rohde et al. found superior CBT effects only for depressed White youth, ethnic differences were likely a function of the unusually positive response by non-White youth to placebo control (i.e., life-skills training). Thus, neither the ethnic invariance nor ethnic disparity perspective is clearly supported by these findings.

Nonetheless, culture-responsive methods have been identified and utilized by a small but growing number of clinical investigators. The diversity of culture-responsive approaches is reflected in Table 7, which summarizes the different ways that treatments in this review were adapted to address the needs of ethnic minority clients. Unfortunately, with the exception of those studies described next, the clinical impact of such modifications has rarely been tested.

Correlational data provide some evidence linking culture-responsive methods to beneficial responses in treatment outcome studies. Specifically, two studies indicate that ethnic match between client and therapist was associated with positive outcomes following youth- and family-based treatment (Halliday-Boykins, Schoenwald, & Letourneau, 2005; Yeh, Eastman, & Cheung, 1994). For both studies, however, nonrandom assignment to matched therapists leaves open the possibility that factors other than match accounted for the significant findings.

In contrast to correlational studies, experimental evaluations do not support the culture-responsive perspective. Szapocznik and colleagues compared BSFT with Bicultural Effectiveness Training (BET) for 31 Cuban American families with behaviorally disordered youth (Szapocznik, Rio, et al., 1986). BET was identical to BSFT, except that BET also focused on teaching “bicultural skills” to family members (e.g., methods for addressing intercultural conflict between the youth and parents). The treatments differed minimally on posttreatment ratings of behavioral problems, suggesting that bicultural skills training was not associated with additional benefits.

A second study yielded similar results. Specifically, Genshaft and Hirt (1979) evaluated how ethnic matching influenced outcomes in the context of a peer-modeling intervention. Sixty African American and European American youth were randomly assigned to a same-race model, an opposite-race model, or no-treatment control. Regardless of ethnicity, training by “White” models was more successful at ameliorating cognitive impulsivity than training by either “Black” models or no treatment. Thus, neither Szapocznik, Rio, et al. (1986) nor Genshaft and Hirt provide empirical support for the utility of culture-responsive treatment.

Aggregate effect size data were also used to evaluate whether ethnic minority youth fared better with culturally modified approaches. There is no consensus definition in the field about whether or not a treatment is considered culture-responsive or how to decide whether an adaptation is warranted (see Lau, 2006, for an emerging model). Therefore, for this study, two broad methods were used for classifying EBTs as culture-responsive. First, EBTs were defined as culture-responsive only when the clinical trial from which posttreatment effect size estimates were derived identified intervention or clinician characteristics that made treatment more appropriate for ethnic minority participants. Using this conservative approach (κ = .80), 10 treatments were considered culture-responsive and 10 were classified as standard (i.e., treatment has no apparent culture-responsive element; Table 8). However, because investigators sometimes omit such information from published clinical trials, a second more liberal approach (κ = .78) defined treatment as culture-responsive when information from supplementary sources (e.g., treatment manuals, prior clinical trials, book chapters) suggested that treatments were modified for ethnic minority participants. Using this approach, 14 treatments were classified as culture-responsive and 6 as standard. Table 4 shows the resulting effect size estimates. No significant effects were found based on either the first, Q(1) = 0.01, p = .93, or second, Q(1) = 1.79, p = .18, definition. Notably, these findings contrast with results from a recent meta-analysis of culturally adapted interventions (Griner & Smith, 2006).

However, some scholars (e.g., Rogler, Malgady, Costantino, & Blumenthal, 1987) contend that such standard ways of defining culture-responsive practice may be unduly narrow, arguing that conceptualizations of “cultural-sensitivity” should be broadened to encompass mainstream modalities with particular relevance for ethnic minorities. For example, some contend that in contrast to individual psychotherapy, family- or group-based treatments may be ideal for ethnic minority youth because such modalities permit clinicians to better consider the cultural context when planning and conducting treatment (Rogler et al., 1987; Tharp, 1991). Yet empirical support for this perspective is lacking as well. Szapocznik and colleagues tested the relative efficacy of one-person versus conjoint family therapy for conduct-disordered Latino youth and found no outcome differences (Szapocznik & Hervis, 1983; Szapocznik, Kurtines, et al., 1986). Moreover, a recent trial by Rossello et al. (in press) indicated that individual treatments (CBT and IPT) were just as effective for depressed Puerto Rican youth as group-based versions of the same therapies. These findings suggest that, for Latinos, individual treatment is equal to family- and group-based modalities. Unfortunately, because only two suitable studies focused on individual psychotherapy (Garza & Bratton, 2005; Rossello & Bernal, 1999), this hypothesis could not be further tested in the current meta-analysis.

Unfortunately, evaluation of youth treatment mediation is exceedingly rare (Hinshaw, 2002; Kazdin & Nock, 2000; Weersing & Weisz, 2002b). However, the limited research does show that efficacious, minority-focused treatments are often successful at modifying hypothesized mediators of ultimate outcomes, including family functioning (Henggeler et al., 1992; Liddle et al., 2004; Lochman & Wells, 2004; Santisteban et al., 2003; Stuart et al., 1976), parenting competencies (Cohen et al., 2004), peer functioning (Liddle et al., 2004; Lochman et al., 1993), and individual cognitions (Cohen et al., 2004; Hudley & Graham, 1993). Moreover, using more formal analytic tests (Holmbeck, 1997), several investigators have assessed specific mediation effects within ethnic minority samples. Lochman and Wells (2002a) provided a compelling example of mediation testing within the context of a clinical trial with aggressive, predominantly African American youth. They found that intervention effects (i.e., Coping Power vs. control) on drug use, delinquency, and school behavior were partially mediated by changes in parenting behavior and youth cognitions.

Two nonexperimental studies of MST similarly revealed significant outcome mediators. Huey and colleagues found that for rural, mostly African American offenders, changes in family functioning and deviant peer affiliation mediated the relationship between therapist adherence to MST and reductions in delinquent behavior (S. J. Huey, Henggeler, Brondino, & Pickrel, 2000). These results were replicated in a sample of urban, predominantly European American offenders, suggesting that these mechanisms were not ethnic- or region-specific (S. J. Huey et al., 2000). In a larger multisite evaluation of MST, Halliday-Boykins et al. (2005) found that the relations between therapist–client ethnic match on discharge success was partially mediated by higher therapist adherence to MST. Findings from these three studies are encouraging and suggest that clinical change for ethnic minority youth may occur via theory-consistent mechanisms.

However, the mediator framework articulated by Kraemer et al. (2002) suggests that only the Lochman and Wells (2002a) study would serve as an example of treatment mediation. According to Kraemer et al., a treatment mediator must satisfy several conditions including (a) association with treatment condition (e.g., ratings on the mediator variable are higher for treatment vs. control youth), (b) association with the outcome variables, and (c) change during the period of active intervention. Because S. J. Huey et al. (2000) and Halliday-Boykins et al. (2005) included only youth assigned to the MST condition—and thus did not satisfy the first condition—the factors tested in these studies cannot be considered true mediators of MST effects (Hinshaw, 2000; Kraemer et al., 2002).

Although treatment mediation effects are rarely studied in youth, formal tests of moderation are more prevalent. Moderator evaluations test the extent to which a specified variable influences treatment efficacy, and address the question for whom does treatment work and under what conditions (Hinshaw, 2000; Kraemer et al., 2002). Perhaps the clearest examples are the studies noted earlier testing ethnicity as a treatment moderator. Additional research suggests that other demographic and clinical factors may also moderate youth treatment effects within ethnic minority samples. The programs of research on Coping Power and narrative treatment best illustrate such effects.

Lochman et al. (1993) found that Social Relations treatment was successful at reducing aggression and peer-rejection for some African American youth but not others. Youth who were both aggressive and peer-rejected at pretreatment benefited from treatment whereas rejected-only youth did not (Lochman et al., 1993). In a subsequent study, Lochman and Wells (2003) evaluated the extent to which Coping Power reduced delinquency/aggression and prevented drug use in aggressive, ethnic minority youth, and whether effects were moderated by gender, age, neighborhood status (problem vs. nonproblem neighborhood), or initial problem severity (moderate vs. high). At the 1-year follow-up, preventive effects on tobacco, alcohol, and marijuana use were strongest for youth who were older and evidenced moderate initial risk. Neighborhood status and gender did not moderate drug use outcomes. Also, none of the moderator effects were significant for delinquency or aggression outcomes. Thus, although Coping Power outcomes were influenced by several significant moderators, no clear pattern of effects emerged.

In contrast, Costantino and colleagues (Costantino, Malgady, & Rogler, 1986, 1994; Malgady, Rogler, & Costantino, et al., 1990) identified age as a consistent moderator of outcomes for narrative treatments with Latino youth. Cuento Therapy is a 20-session, narrative intervention involving Puerto Rican cuentos, or folktales. During treatment, bilingual/bicultural therapists read cuentos to youth, promote group discussion of prominent themes, facilitate role-play and dramatization of themes, and verbally reinforce youth for adaptive responses. In an initial evaluation (Costantino et al., 1986), 208 kindergarten to fourth-grade Puerto Rican youth with below-median ratings of problem behavior were randomly assigned to original cuento therapy (i.e., stories were consistent with the original Puerto Rican cuentos), adapted cuento therapy (i.e., stories were modernized to match the mainland U.S. context), art/play therapy, or no-treatment control. Costantino and colleagues found that grade level moderated the effect of treatment condition on trait anxiety outcomes. For first-grade children only, adapted cuento therapy led to greater reductions in trait anxiety than all other treatment conditions (Costantino et al., 1986). This moderator effect was not found at the 1-year follow-up.

Based on these moderator findings, Costantino and colleagues modified this narrative approach to match the developmental needs of older youth. Yet curiously, age continued to moderate treatment effects (Costantino et al., 1994; Malgady et al., 1990). Malgady et al. randomly assigned eighth- and ninth-grade Puerto Rican students with below-median ratings on a behavior checklist to Hero/Heroine Modeling (a variation of cuento therapy designed for adolescents) or attention-placebo control. Moderator analyses showed that for eighth- but not ninth-grade youth, treatment led to significantly lower trait anxiety than control. Similarly, Costantino et al. (1994) found that the efficacy of their Tell-Me-A-Story Intervention (a variation of cuento therapy using pictorial stimuli and designed for multiracial Hispanic youth) varied as a function of both grade level and gender among Hispanic youth with conduct, anxious, or phobic symptoms. Compared with placebo control, Tell-Me-A-Story Intervention led to fewer school conduct problems for sixth graders only, and fewer phobic symptoms for fifth-grade boys and fourth- and fifth-grade girls only.

Thus, across three “prevention” trials, Costantino and colleagues found evidence that narrative treatment shows its greatest success in ameliorating anxiety-related symptoms among younger children. However, narrative therapy did not meet the APA Task Force criteria (Chambless et al., 1998; Chambless & Hollon, 1998; Chambless et al., 1996) because (a) outcome effects did not clearly match the target behavior (e.g., treatment ameliorated anxiety problems but youth often showed above median levels of externalizing behavior; Costantino et al., 1986; Malgady et al., 1990), (b) treatment had the purported goal of increasing ethnic identity and self-concept rather than decreasing symptomatology (Malgady et al., 1990), and (c) none of the trials reported treatment main effects.

Although African Americans and Latinos are underrepresented, Asians, Pacific Islanders, and Native Americans are nearly excluded from the youth treatment outcome literature, and future clinical trials should include these groups in adequate numbers to permit appropriate outcome evaluation. The need is particularly acute for Native American adolescents given the high prevalence of serious mental health problems (e.g., “hard” drug abuse, completed suicide) in this ethnic group (Hawkins, Marlatt, & Cummins, 2004; National Institute of Drug Abuse, 2003). Although prevention work with Native American youth is in ample supply (Hawkins et al., 2004), no evidence-based therapies for Native American youth with preexisting mental health problems have been developed as yet. (For one such effort see Carpenter, Lyons, & Miller, 1985.)

However, some scholars warn against such comparative approaches, recommending instead that research with ethnic minorities focus on within-group evaluations. For example, Yali and Revenson (2004) advised caution when using between-group designs, because ethnic comparisons could inadvertently encourage “minority-deficit” models. Similarly, Bernal and Scharron-Del-Rio (2001) contended that because ethnic comparisons often have weak conceptualizations, “it is best to focus on specific ethnic groups, unless there is a clear theoretical basis for a comparative approach” (p. 338). Thus, an alternative approach would eschew ethnic comparisons and instead explore whether acculturation status, exposure to discrimination, and other culture-related factors serve as treatment moderators for ethnic minority youth (Alvidrez, Azocar, & Miranda, 1996; Hall, 2001). Indeed, some research suggests that immigrant minorities may respond less favorably than nonimmigrants to Western therapies (Martinez & Eddy, 2005; Telles et al., 1995) and that country of origin may affect treatment outcomes for Latino youth (Kataoka et al., 2003). Another important demographic variable rarely reported (Weisz et al., 2005) or considered when examining treatment moderation is socioeconomic status. To our knowledge, treatment outcome studies have not been conducted which examine the differential efficacy of EBTs across youth from ethnic minority families of both low and high socioeconomic status groups.

Yet designs of this sort may be of little theoretical value if cultural adaptations reflect only surface changes in treatment structure or content. Although cultural content differed dramatically across studies in this review, many treatments made “surface” modifications (e.g., ethnic match) that required minimal attention to cultural issues (Kumpfer, Alvarado, Smith, & Bellamy, 2002), and only a few were based on conceptual models of cultural sensitivity. Given the broad definition of culture-responsiveness adopted for this review, one could argue that the true influence of cultural adaptation was not adequately tested here. Thus, future efforts should focus on developing and testing more theoretically compelling adaptations.

Similarly, the two experimental efforts to isolate cultural adaptation effects for youth treatment (Genshaft & Hirt, 1979; Szapocznik, Rio, et al., 1986) likely lacked adequate power. With a two-group comparison (culture-responsive treatment vs. standard treatment), sample size requirements differ dramatically depending on the anticipated strength of the culture-responsive component. If small effects (e.g., d = .20) were expected, sample size requirements would readily exceed 800 (i.e., approximately 400 per condition; see Murphy & Myors, 1998). However, even if moderate effects (e.g., d = .50) were anticipated, as suggested by promising work in the adult treatment literature (S. J. Huey & Pan, 2006; Kohn, Oden, Munoz, Robinson, & Leavitt, 2002; Wade & Berstein, 1991), at least 130 participants (i.e., 65 per group) might be needed (Murphy & Myors, 1998). By contrast, both Genshaft and Hirt and Szapocznik, Rio, et al. (1986) included samples with fewer than 20 participants per condition.

Preparation of this article was supported by AHRQ grant PO1 HS1087 and NIMH grant K08 MH069583. We thank John Weisz for his conceptual and technical assistance with the meta-analysis.