1992; Judd, Akiskal, & Paulus, 1997). In a longitudinal study, subsyndromal depression among adolescents predicted poorer functioning as these individuals became adults (Devine, Kempton, & Forehand, 1994). Subsyndromal depressive symptoms among adolescents predicted MDD later on in adolescence and young adulthood (Pine, Cohen, & Brook, 1999; Rao et al., 1995; Weissman, Warner, Wickramaratne, Moreau, & Olfson, 1997). Lewinsohn, Solomon, Seeley, and Zeiss (2000) found that increasing levels of depressive symptoms among a large sample of nondepressed adolescents (average age of 161 2 ) predicted increased levels of social dysfunction and incidence of MDD, as well as increased substance abuse at age 24. These data indicate that subsyndromal depression renders adolescents at risk for a first episode of MDD, and they are prime candidates for depression prevention programs.

Poverty has been linked with an early onset of depression. It is not clear whether this represents an independent risk factor or can be grouped among the more general examples of diversity that are associated with depression. Results from epidemiological studies have linked lower socioeconomic status with depression and a multitude of other mental health problems (Robins, Locke, & Reiger, 1991). This vulnerability is particularly strong for families living at poverty levels (Bruce, Takeuchi, & Leaf, 1991). This relationship may be explained in part by a phenomenon of selection, whereby those with mental health problems are more inclined to drift toward economic disadvantage and remain there (Dohrenwend et al., 1992). Longitudinal data have also demonstrated that socioeconomic disadvantage is largely a cause of higher vulnerability to psychiatric disorder, particularly for depression (Dohrenwend et al., 1992; Gilman, Kawachi, Fitzmaurice, & Buka, 2002; Johnson, Cohen, Dohrenwend, Link, & Brook, 1999). In a study of over 4,000 Australian families, poverty caused a small but significant increase in risk when other sociological variables were controlled (Spence, Najman, Bar, O'Callaghan, & Williams, 2002); this effect was more pronounced in girls than in boys.

If we consider poverty as a generator for a variety of stressors, the possible mechanisms driving poverty-induced vulnerability appear boundless. A number of mediators between socioeconomic disadvantage and depression have been studied empirically. These include external mediators such as access to health care, quality of social networks and resources, quality of parenting and parent availability, and, of course, level of exposure to violence. Children of families who are of lower socioeconomic status are most likely to witness violence and to be the victims of abuse (Buka, Stichik, Birdthistle, & Earls, 2001; Sedlak & Broadhurst, 1996).

Internal individual mediators include self-esteem, health-risk behaviors, cognitive deficits, interpersonal skills, and academic achievement. Several comprehensive reviews on the consequences of poverty and mediating factors demonstrate the vast amount of knowledge we have accumulated on the relation between poverty and depression (Aber, Bennett, Conley, & Li, 1997; Leventhal & Brooks-Gunn, 2000; Turner & Lloyd, 1999). This literature highlights the importance of two larger factors: (a) the need for universal health care with parity for mental illness and physical illness and parity for services for adults and children; and (b) the need to address large-scale public health risk factors that have a strong effect on the occurrence of adolescent depression (i.e., exposure to violence).

Exposure to violence during childhood is a potent risk factor for future psychological and psychiatric disorders (Kilpatrick et al., 2003; MacMillan et al., 2001) as well as physical health-risk behaviors (Felitti et al., 1998), in both the short and long term. The violence to which children are exposed has many forms. This includes being a victim of sexual or physical abuse as well as witnessing violence in the home (Kilpatrick et al., 2003). A large number of children also frequently witness violence in the community (Buka et al., 2001). Children who are exposed to violence are most often exposed to more than

one type, and evidence suggests that the amount of violence-related adversities a child encounters has a substantial impact on the severity of the outcome (Felitti et al., 1998). The most disturbing illustration of this accumulation phenomenon is the gradation effect of violence-related adversities on risk for suicide attempt. Results from the Adverse Childhood Experiences Study demonstrated that for every additional adversity experienced as a child, the risk of suicide attempts increased from 2-to 5-fold, such that children or adolescents who encounter seven or more adversities are 50 times as likely to attempt suicide as those without violence exposure (Dube et al., 2001).

Although the mental health consequences of violence exposure are diverse, the most prevalent and commonly studied are posttraumatic stress disorder (PTSD) and major depression. This makes sense, particularly if violence exposure is viewed as a form of trauma. The consequences of our country's recent dealings with terrorism have provided an especially graphic picture of how violence-related trauma is linked to PTSD and depression (North et al., 1999; Schlenger et al., 2002). Survivors of the Okla-homa City bombing, for instance, were studied by North and collegues (1999) about 6 months after the disaster. Of these individuals, 45% had some form of post-disaster psychiatric disorder (34.3% had PTSD, and 22.5% had MDD). Psychiatric comorbidity predicted functional impairment and treatment. Fifty-six percent of depressed subjects had never been depressed before the incident.

Violence-related trauma experienced during childhood can have particularly devastating effects, because the trauma is inflicted during a critical period of development. Neurobiological and neuroendocrine studies of depressed women, which look at the volume of certain brain regions and at hormonal stress-response mechanisms, provide evidence that violence-related trauma experienced during childhood can have profound and lasting effects on brain structure and function (Heim, Newport, Bonsall, Miller, & Nemeroff, 2001; Vythilingam et al., 2002). These alterations, in turn, increase vulnerability to stress-related disorders like depression.

Depression that is comorbid with PTSD or other disorders, as well as depression that has an established neurobiological etiology like that experienced by childhood victims of trauma, are forms of the disorder that are particularly resistant to treatment and are associated with increased levels of impairment (Mervaala et al., 2000; Petersen et al., 2001). Thus, it is essential that prevention strategies attend to violence exposure.

The role of ethnocultural factors has been understudied. Some ethnic groups appear to have higher rates of adolescent depression than others. For example, Mexican Americans and African-American adolescents appear to have higher risk for depression whereas American adolescents of Chinese descent may be at lesser risk (Barrera & Craighead, 2004; Roberts, Roberts, & Chen, 1997). It appears that some groups (such as African-American or Hispanic adolescents) may not show the gender disparity following puberty that is seen among Caucasian adolescents (Hayward, Gotlib, & Schraedley, 1999).

In the Penn Prevention Program, Seligman and colleagues (Gillham & Reivich, 1999; Gillham, Reivich, Jaycox, & Seligman, 1995; Jaycox, Reivich, Gillham, & Seligman, 1994) developed and evaluated a school-based “indicated” prevention program targeting 10-to 13-year-old children in school districts in the Philadelphia suburbs. The youth were defined as at risk for depression on the basis of elevated self-reported depressive symptomatology, self-reported parental conflict, or both. This prevention program was based on a model of explanatory style introduced by Seligman and colleagues (Nolen-Hoeksema, Girgus, & Seligman, 1992) and on research identifying core cognitive deficits associated with youth depression, including negative self-evaluation, dysfunctional attitudes, poor interpersonal problem solving, and low expectations for self-performance (Garber, Weis, & Shanley, 1993; Kaslow, Rehm, & Siegel, 1984; Quiggle, Garber, Panak, & Dodge, 1992). Participants recruited for the treatment group were assigned to one of three treatment programs: a cognitive training program, a social problem-solving program, or a combined program. Eighty-eight students, whose scores were matched to prevention participants, were recruited from nearby schools and comprised the no-participation control group. Assessments included child self-report, teacher-report, and parent-report questionnaires.

Results indicated that relative to control participants, children who participated in any of the treatment groups reported significantly fewer depressive symptoms immediately following the program and at the 6-month and 2-year follow-ups, but not at the 12-month and 3-year follow-ups. Moreover, teacher reports at follow-up revealed better classroom behavior in treatment participants than in control participants. Finally, overall treatment effects were more significant for children who, at the screening phase of the study, reported more significant depressive symptomatology and more significant parental conflict at home. The major limitations of the study are the lack of randomization to intervention conditions, the use of only self-report measures, attrition of approximately 30% of partici

pants during follow-up, and the failure to include diagnoses for clinical depression.

More recently, Seligman and his group at Penn have focused on “positive psychology” programs, which are likely to have an indirect effect of preventing episodes of MDD. These programs are included in Part VII of this book.

Clarke and colleagues (Clarke et al., 1995) in Oregon were among the first to study prevention of MDD among adolescents. In an excellent study, 150 adolescent students from 9th and 10th grades were assigned randomly to either a “prevention” or “usual-care” group. The prevention program, entitled “Adolescent Coping with Stress Course” was delivered in groups and was a prevention-focused version of this group's “Adolescent Coping with Depression Course” (Clarke, Lewinsohn, & Hops, 1990). The 5-week intervention was conducted within the adolescents' school setting and comprised fifteen 45-minute group sessions (3 after-school meetings per week). The usual-care youngsters were free to continue with preexisting treatment or seek new treatment. This program employed both behavioral and cognitive coping techniques designed to reduce vulnerability to future depressive episodes.

Participants were followed for 1 year, and the results were positive. Namely, significantly (p < .05) fewer prevention group (14.5%; 8 of 55) subjects were diagnosed with MDD or dysthymia than control group subjects (25.7%; 18 of 70). The major strengths of this program include random assignment of subjects, adequate sample sizes, diagnoses of clinical mood disorders, and encouraging outcomes. It is important to note, however, that approximately 36% of their participants had suffered a prior episode of MDD. Because 30% to 50% of adolescents who have had a prior episode can be expected to have a relapse or recurrence of the disorder during the time (18 months) of this study (Hart, Craighead, & Craighead, 2001; Lewinsohn, Rohde, Seeley, & Fischer, 1993; Rao et al., 1995), it very well may be that Clarke and colleagues actually had their biggest impact on preventing relapse or recurrence rather than on preventing a first episode of MDD. In addition, it should be noted that there was differential dropout between conditions in this study (Clarke et al., 1995), but as the authors suggested, this probably operated against their favorable outcomes.

In an expansion of this program, Clarke and associates (2001) applied this approach to a health maintenance organization (HMO) population of adolescents of parents with diagnosed depression and youngsters already manifesting symptoms. They screened all those at risk and divided them into three groups: low or no depressive symptomatology, medium symptomatology, and those already in episode. Those already meeting criteria for MDD were referred for treatment, and those with no depressive symptomatology were excluded. In this trial, those adolescents (ages 13–18) with moderate symptomatology were randomized into a usual-care condition (N = 49) or their cognitive behavioral intervention group (N = 45). As in their previous study, prevention group subjects participated in 15 group sessions.

This intervention yielded substantial preventive effects, with significant treatment-by-time effects in the expected direction on the Center for Epidemiological Studies–Depression (CES-D) and the Global Assessment of Functioning scales—i.e., adolescents in the prevention condition did much better than those in the usual-care condition. Survival analysis indicated that over a 15-month follow-up period, there was a cumulative rate of major depression of only 9% in the experimental group in contrast to 28% in the usual-care condition (Clarke et al., 2001). Even though this is the most sophisticated prevention study to date, its specific implications for prevention are limited by the choice to include adolescents who had previously suffered from an episode of depression; these subjects comprised 67% of the adolescents in this study. Thus, as with the prior study, it is impossible to determine if the study prevented first episodes or relapse and recurrence of prior episodes of MDD. Currently, Clarke, Garber, Beardslee, and Brent are conducting a four-site effectiveness study of this preventive intervention.

Two prevention programs in Australia have recently been described and evaluated: one by Shochet and colleagues, and the other by Spence and associates. Shochet and colleagues (2001) evaluated a “universal” prevention program applied in a school setting in Australia. This was a skills-based program of 11 sessions offered by a psychologist and based on a downward extension of principles of cognitive behavior therapy (CBT) and interpersonal therapy (IPT). The student sessions could be supplemented with three parental sessions, but not many parents took advantage of this offer. There were 240, 12-to 15-year-old subjects, who were assigned to assessment-only control, prevention without parental sessions, or prevention with parental sessions.

Students who completed either prevention program (no differences were obtained between the two prevention groups) showed fewer depressive symptoms on one measure of depression (but not another) than controls. The prevention program subjects also reported less hopelessness at the end of the project. All of these effects were maintained at a 10-month follow-up. The limitations of this program include the lack of random assignment of subjects (controls participated in one academic year, and intervention subjects in the next academic year), a small sample size, and assessments conducted at different times of the year in different conditions. The findings, though limited, were encouraging for such a short program and short follow-up period.

A very sophisticated study evaluated the long-term impact of a universal, teacher-implemented, and school-based prevention program that was developed by Spence, Sheffield, and Donovan (2003) in Australia. The program, Problem Solving for LIFE (PSFL), is a combination of cognitive restructuring and problem-solving approaches, and it is designed to prevent a first episode of depression. Subjects were 1,500 eighth-grade (ages 12–14; mean = 12.9) students attending 16 participating high schools in the Brisbane region of Queensland, Australia. The eight 1-hour classroom-session program was implemented by 28 teachers in eight randomly as signed schools (N = 751), while the control subjects (N = 749) attended the other eight schools. There were approximately equal numbers of girls and boys.

Appropriate data analyses indicated that the program significantly decreased depressive symptoms between the beginning and end of the program. This finding, however, was only true for those adolescents who had elevated (“high risk,” defined as 13 or higher on the Beck Depression Inventory) depression scores at the beginning of the study. Unfortunately, this difference was not maintained at a 12-month follow-up.

These two well-conducted studies are not particularly encouraging for the effectiveness of “universal” prevention programs for MDD. There are some hints, however, that similar studies conducted with at risk samples (i.e., selected or indicated prevention programs) might be more effective. For example, in the Spence study (replicating the Penn Study) it appears that students with greater depressive symptomatology (but not MDD) may respond better to prevention programs.

Because of the stable population of Iceland and because most of the citizens live in one city, Reykjavik, Arnarson and Craighead have spent the past several years translating and standardizing the assessment instruments and developing a manualized, developmentally based, behavioral and cognitive program designed to prevent depression. It has been labeled the “Thoughts and Health” program, and it includes a student workbook as well as a program manual. The program consists of 15 group sessions (6–8 students per group) that are delivered in the school setting by the local school psychologists.

In their most recent report of this work (Arnarson & Craighead, 2004), they had studied 72 students at risk for MDD. At risk was defined as scoring above 13 on the Children's Depression Inventory or at the 75th percentile or higher on the negative composite of the Children's Attributional Style (CAS) Questionnaire (see Alloy et al., 2000). Thirty-two of the subjects from five