Blues Program

Study 1

Summary

Rohde, Stice, Shaw, & Brière (2014), Rohde et al. (2015), Brière et al. (2014, 2016), and Müller et al. (2015) examined 378 students from five high schools who were randomly assigned to three conditions: cognitive-behavioral group depression prevention, cognitive-behavioral bibliotherapy, or brochure control. The students completed self-report surveys and diagnostic interviews at pretest, posttest, and 6-, 12-, 18- and 24-month follow-ups.

Rohde, Stice, Shaw, & Brière (2014), Rohde et al. (2015), Brière et al. (2014, 2016), and Müller et al. (2015) reported that, compared to participants in one or more comparison groups, participants completing the cognitive-behavioral depression prevention program had significantly:

• Greater reductions in interviewer-rated depressive symptoms at posttest
• Lower rates of major depression onset at six-month follow-up and two-year follow-up.

Evaluation Methodology

Design:

Recruitment /Sample size: Students (approximately 8,020) from five high schools from three school districts in one county (state not reported) were recruited through direct mailings sent to all students and recruitment posters hung at the schools between 2009 and 2011. The mailing included a simplified version of the Center for Epidemiological Studies-Depression Scale (CES-D). Students who completed the CES-D and endorsed two or more depression symptoms were encouraged to participate in the study. Eligible students providing individual and parental consent (n=451) were then asked to schedule an eligibility interview with research staff to be completed at school. Of 421 students that attended the eligibility interview, 43 were excluded for meeting the diagnostic criteria for a major depressive disorder resulting in an original sample of 378 students.

Study type/Randomization/Intervention: Eligible participants (n=378) were randomly assigned, within blocks created by gender and school, to one of three conditions: (1) cognitive-behavioral group depression prevention (n=126), (2) cognitive-behavioral bibliotherapy (n=128), or (3) brochure control (n=124). Participants in the cognitive-behavioral prevention group completed six weekly one-hour program sessions in single-gender groups of five to nine participants at school (usually after normal classes). Students in the cognitive-behavioral bibliotherapy group were given a personal copy of the book Feeling Good, which provides relevant and practical cognitive-behavioral techniques for preventing and reducing negative moods. Participants in the brochure control group were given a National Institute of Mental Health brochure that describes major depressive disorder and recommends treatment for depressed youth, as well as information about local treatment options.

Assessment/Attrition: Participants completed a survey and diagnostic interview at pretest (during the eligibility interview), posttest, and 6-, 12-, 18- and 24-month follow-ups. At posttest, all participants in the cognitive-behavioral prevention group completed assessments, but the cognitive behavioral bibliotherapy and brochure control groups each had an attrition rate of 8%. At six-month follow-up, attrition rates were reportedly 5% for the cognitive-behavioral prevention group, and 12% for each of the other two groups. From pretest to 6-month follow-up, total attrition was 9.5% in Figure 1 and 13% in the text (this total is inconsistently reported). Attrition was 7% at 1-year, 10% at 18-month, and 12% at 2-year follow-up.

Sample characteristics:

The sample included high school students who ranged in age from 13 to 19 years, with a mean age of 15.5 years at pretest. Sixty-eight percent of participants were female, 72% were Caucasian, 6% Hispanic, 2% Asian, 1% African American, 1% Native American, and 18% identified as other or mixed racial/ethnic category. Educational attainment of parents was 39% high school graduate or less, 26% some college, 22% college graduate, and 13% graduate degree. Thirty-seven percent of participants had received treatment services for mental health problems during the 12-month period preceding the study.

Measures:

Validity of measurements: All measures and scales have been used in prior published work and reliability and validity have been established.

Primary outcomes: An adapted version of the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS), a 16-question semi-structured diagnostic interview assessing major depressive disorder (MDD) symptoms, was used to determine whether participants met diagnostic criteria for MDD. In addition to identifying the onset of MDD, K-SADS data were used to calculate a continuous depressive symptom composite by averaging the severity ratings for each symptom.

Secondary outcomes: Substance use was measured with 10 items from a previous study in which participants were asked to report the frequency of alcohol and drug use.

Social adjustment in school, work, peer, spare time, and family domains was measured using an adapted version of the Social Adjustment Scale-Self-Report for Youth.

At pretest, after receiving a description of the three conditions, participants completed four items assessing credibility (this concept is undefined), expected improvement from each condition, satisfaction with regard to the condition to which the participant was assigned, and preferred group assignment.

Analysis:

Statistical methods/baseline control/correct unit of analysis: The primary outcome of major depressive disorder onset was analyzed using Cox proportional hazard regression models with missing data taken into account with right-censoring.

All other outcomes (depressive symptom severity, social adjustment, and substance use) were evaluated using repeated-measures analysis of covariance (ANCOVA) with pretest value of the outcome as a covariate. These analyses used multiple imputation to replace missing values. Long-term 2-year effects from posttest were evaluated with random effects growth models in a hierarchical modeling framework that adjusted for clustering within treatment groups.

F tests were used to investigate the main effect of condition and group-by-time interactions for the depressive symptom severity and the secondary outcomes.

Intention-to-treat: Multiple imputation was used to replace missing values in all ANCOVAs and multilevel models. Additionally, deliberate efforts were made to maintain contact with participants to reduce attrition.

Outcomes:

Implementation fidelity: Program facilitators were trained, observed, recorded (and provided feedback on their performance), and rated for competency in and adherence to the program. Half of the audio-recorded sessions were randomly selected for review by the researchers. Using 1 to 10 point scales, mean adherence was 7.0, and mean competence was 7.1, with less than 2% of items rated as "minimal adherence" or "minimal competence." Students in the cognitive-behavioral prevention group attended an average of 5.3 sessions, with 48% attending all six. Additionally, individual make-up sessions were offered to students who missed scheduled sessions-with 94% of participants missing a session receiving the make-up-and 80% of program homework assignments were recorded by the group leader as completed.

Baseline Equivalence: At baseline, participants in the three conditions did not significantly differ on outcome measures, demographic variables, or receipt of mental health treatment one year prior to enrollment in the study.

Differential attrition: The study reports that attrition did not differ between the three groups at posttest or at any follow-up, and missingness was not significantly related to any of the study variables at pretest with one exception. Participants who completed assessments at all six time points had significantly lower baseline substance use compared to those that did not.

Posttest: Cox proportional hazard regression analysis indicated that the onset of major depressive disorder at six-month follow-up was significantly greater in both the brochure control (p=.023, one-tailed) and cognitive-behavioral bibliotherapy (p=.024, one-tailed) comparison groups than in the treatment group. Controls for school, mental health treatment, and intervention expectancies did not change the significance of the hazard regression coefficients.

ANCOVAS revealed that the cognitive-behavioral prevention group had significantly lower depressive symptom severity than the brochure control group at posttest (d=.29), but not six-month follow-up (both without and with controls). Depressive symptom differences between the cognitive-behavioral prevention group and the cognitive-behavioral bibliotherapy group were marginally significant at posttest (p=.06), with lower severity among the prevention group, but non-significant at six-month follow-up. No other significant differences were found for other outcome measures of social adjustment or substance use.

Additionally, less powerful exploratory F tests were used as a check on the ANCOVA group comparisons. The tests found no overall main effect of condition or group-by-time interaction effects for depressive symptom severity, social adjustment, or substance use.

Brière et al. (2014) examined eight moderators of the group-based and bibliotherapy intervention effects on depressive symptoms over the posttest and six-month follow-up. Only one reached statistical significance. It involved the bibliotherapy condition relative to the control condition but not the group-based condition.

Long-term effects: Cox proportional hazard regression analysis indicated that the onset of major depressive disorder through the 2-year follow-up was significantly greater in the cognitive-behavioral bibliotherapy (p=.006, one-tailed) comparison group than in the treatment group but not in the brochure only comparison group. The hazard models showed the same results for the onset of major depressive disorder or minor depression.

No additional effects were found, suggesting that posttest effects were largely equalized during the long-term follow-up period.

Using combined data from both Study 1 and Study 2, Brière et al. (2014) estimated four trajectories in depressive symptoms over the two-year follow-up period. The results showed that the group-based intervention participants were less likely than control group participants to belong to trajectories of high and persistent symptoms and more likely to belong to trajectories of low and declining symptoms.

Using combined data from both Study 1 and Study 2, Müller et al. (2015) tested for eight moderators of the group-based and bibliotherapy intervention effects on depressive symptoms. Over the period from posttest to the 24-month follow-up, the tests found two significant moderation effects: High pretest depressive symptoms and high pretest motivation to change both increased the impact of the intervention.

Study 2

Summary

Stice et al. (2008, 2011), Stice, Rohde, Gau, & Wade (2010), Stice, Rohde, Seeley, & Gau (2010), Gau et al. (2012), Rohde, Stice, Gau, & Marti (2012), Rohde, Stice, & Gau (2012), Marchand et al. (2010), Brière et al. (2016), and Müller et al. (2015) examined 341 students from six high schools who were randomly assigned to four conditions: cognitive-behavioral depression prevention, supportive-expressive intervention, bibliotherapy intervention, or educational brochure control group. The students completed self-report surveys and diagnostic interviews at pretest, posttest, and 6-, 12-, and 24-month follow-ups.

Stice et al. (2008, 2011), Stice, Rohde, Gau, & Wade (2010), Stice, Rohde, Seeley, & Gau (2010), Gau et al. (2012), Rohde, Stice, Gau, & Marti (2012), Rohde, Stice, & Gau (2012), Marchand et al. (2010), Brière et al. (2016), and Müller et al. (2015)found that, compared to participants in one or more comparison groups, participants completing the cognitive-behavioral depression prevention program had significantly:

• Greater reductions in interviewer-rated depressive symptoms at posttest, six-month follow-up, and one- and two-year follow-ups
• Greater reductions in self-rated depressive symptoms at posttest, six-month follow-up, and two-year follow-up
• Lower rates of major depression onset at six-month follow-up and two-year follow-up
• Greater reductions in self-reported substance use through two years.
• Greater improvements in social adjustment at six-month follow-up
• Reduced negative cognitions and increased reports of pleasant activities
• Greater social support from friends.

Evaluation Methodology

Design:

Recruitment /Sample size: Participants were recruited through mass mailings, handbills, and posters targeting students experiencing sadness at six high schools in the greater Austin area in Texas between 2004 and 2007. Interested students were given a depressive symptom screen (the Center for Epidemiological Studies-Depression Scale [CES-D]). Students who scored above a pre-determined cutoff point for inclusion (20 or above) and had individual and parental consent were invited to complete a pretest assessment. Those who met criteria for major depression during pretest were excluded and given treatment referrals. In total, 698 students were assessed for eligibility, and 341 students (40 to 75 per school) were randomly assigned to condition.

Study type/Randomization/Intervention: Using blocks created by gender and school, eligible participants were randomly assigned to one of four conditions: (1) cognitive-behavioral depression prevention (n=89), (2) supportive-expressive intervention (n=88), (3) bibliotherapy intervention (n=80), or (4) educational brochure control group (n=84). Participants in the cognitive-behavioral depression prevention group participated in the six-week cognitive-behavioral program. Students in the supportive-expressive intervention group participated in a six-week program, including weekly one-hour sessions, that provided a supportive group for participants to identify, express, and discuss their feelings. Subjects in the bibliotherapy intervention group were given a copy of the book Feeling Good, and encouraged to use the book as a self-help resource. Students in the brochure control condition were given a brochure that describes major depression, recommends treatment, and provides information about local treatment options.

Stice et al. (2011) used the same sample but examined only three of the conditions (cognitive-behavioral depression prevention, bibliotherapy, educational brochure) and 253 of the randomized participants.

Gau et al. (2012) also used the same sample but examined only two of the conditions (cognitive-behavioral depression prevention, educational brochure) and 173 of the randomized participants.

Facilitators attended a two-day training. During the groups, one of the program developers conducted separate weekly supervision meetings with the CB and supportive-expressive facilitators to provide corrective feedback.

Assessment/Attrition: Measures included diagnostic interviews and participant-completed surveys at pretest, posttest (upon program completion), six-month follow-up (Stice et al., 2008), and one- and two-year follow-ups (Stice, Rohde, Gau & Wade, 2010). Attrition rates were 3% at posttest, 7% or 9% at six-month follow-up (this total is inconsistently reported), 15% at one-year follow-up, and 22% at two-year follow-up.

Sample characteristics:

Participants ranged in age from 14 to 19 years (with a mean age of 15.6 years) at pretest. Fifty-six percent of participants were women (44% men), 46% were Caucasian, 33% Hispanic, 9% African American, 2% Asian, and 10% identified as other or mixed racial/ethnic category. Educational attainment of parents was 26% high school graduate or less, 17% some college, 35% college graduate, and 18% graduate degree. Twenty-eight percent of the sample had received treatment services for emotional/behavioral problems during the 12-month period preceding the study.

Measures:

Validity of measurements: All measures and scales have been used in prior published work and reliability and validity had been established.

Primary outcomes: An adapted version of the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS), a 16-question semi-structured diagnostic interview assessing major depressive disorder (MDD) symptoms, was used to determine whether participants met diagnostic criteria for MDD. During interviews with trained professionals who were blind to condition, participants reported the peak severity of each symptom over their lifetime (pretest) or since the last interview (posttest and follow-ups) using a four-point scale.

In addition to identifying the onset of MDD, K-SADS data were used to calculate a continuous depressive symptom composite by averaging the severity ratings for each symptom.

Depressive symptoms were also measured using self-reports from the 21-item Beck Depression Inventory (BDI).

Secondary outcomes: Social adjustment (for all assessments except posttest) was measured using an adapted version of the Social Adjustment Scale-Self-Report for Youth assessing social impairment in school, peer, spare time, and family domains.

The following measures are only reported for posttest and six-month follow-up analyses (Stice et al., 2008):

Substance use was measured with 10 items from a previous study in which participants were asked to report the frequency of alcohol and drug use. Items were averaged to form an overall substance use measure.

Bulemic pathology was measured using the Eating Disorder Diagnostic Interview. Items assessing symptoms of bulimia nervosa were summed to create an overall index of the disorder's behavioral symptoms.

Stice, Rohde, Seeley, and Gau (2010) examined several measures of mediating risk and protective factors:

• Negative cognitions with 12 items from the Automatic Thoughts Questionnaire (alpha = .93)
• Pleasant activities with 12 items from the Pleasant Events Schedule (alpha = .73)
• Emotional expression with 9 items developed for the study (alpha = .82)
• Loneliness with 8 items from the Loneliness Scale (alpha = .83)

Stice et al. (2011) focused on self-reported measures of perceived social support from 12 items of the Network of Relationships Inventory. The study examined two subscales, parental and friend social support (alpha =.84 and .88, respectively).

Analysis:

Statistical methods/baseline control/correct unit of analysis: Omnibus repeated measures analysis of covariance (ANCOVA) models (with baseline interviewer-rated depressive symptoms used as a covariate) with group-by-time interactions were used to test for differential changes in the continuous outcomes. For outcomes with significant group-by-time interactions, separate follow-up repeated measures ANCOVA models tested whether groups differed from each other from pretest to posttest and pretest to six-month follow-up (Stice et al., 2008), and from pretest to one-year follow-up and pretest to two-year follow-up (Stice, Rohde, Gau, & Wade, 2010). Stice et al. (2011) used multilevel models for repeated measures, and Gau et al. (2012) used hierarchical linear models. Rohde, Stice, Gau, & Marti (2012) used linear mixed-effects models with random intercepts and slopes, an unstructured covariance structure, individual time scores measured in months from baseline assessment, and baseline depression scores as a covariate.

Cox proportional hazard regression models controlling for interviewer-rated depressive symptoms at baseline were used to analyze the onset of major (or minor (Stice, Rohde, Gau, Wade, 2010)) depression.

The reliable change index was used to analyze change scores-testing for clinically significant change-in interviewer-rated depressive symptoms and social adjustment from pretest to six-month follow-up (Stice et al., 2008), and interviewer-rated and self-rated depressive symptoms from pretest to two-year follow-up (Stice, Rohde, Gau, & Wade, 2010).

Intention-to-treat: The study employs intent-to-treat analyses by using maximum likelihood estimates to impute missing data and include all randomized subjects in the analysis.

Outcomes:

Implementation fidelity: Among students in the cognitive-behavioral prevention group, 44% attended all six sessions and 86% attended at least three (with the average participant attending 3.3 sessions), and 7% completed all homework assignments, 20% completed three or four, 25% completed one or two, and 48% completed none. Relying on checklists and scales used in prior research (with demonstrated reliability), the researchers evaluated each group facilitator during two sessions for their first group, and one session for each subsequent group, and report the following: 96% of the programs' components were rated at full-adherence and 94% of the items were rated at 'good' competence (with only 1% 'poor').

Baseline Equivalence: At baseline, participants in the bibliotherapy intervention group had significantly lower interview-assessed (K-SADS) depressive symptoms than participants in the other intervention groups (but not the brochure control group). Participants did not differ on demographic characteristics, treatment services received for emotional/behavior problems during the year prior to the study, or any other outcome measures at baseline.

Differential Attrition: Attrition did not differ significantly by condition, nor was attrition related to any baseline outcomes or demographic factors.

Posttest: Stice et al. (2008) report significant effects in Table 2 for 17 of 21 tests for the intervention condition at posttest and six-month follow-up:

Participants in the cognitive-behavioral depression prevention group had significantly greater reductions in interviewer-rated depressive symptoms than participants in all comparison groups at posttest, and greater reductions than brochure control group (only) participants at six-month follow-up. Similarly, participants in the cognitive-behavioral depression prevention group had significantly greater reductions in self-rated depressive symptoms (Beck Depression Inventory) than participants in all comparison groups at posttest, and greater reductions than bibliotherapy intervention and brochure control (but not supportive-expressive intervention) group participants at six-month follow-up. Positive program effects on the Beck Depression Inventory and K-SADS were statistically equivalent across race and ethnic groups (Marchand et al., 2010).

Participants in the cognitive-behavioral depression prevention group showed significantly greater improvements in social adjustment than participants in all comparison groups at six-month follow-up (the social adjustment scale was not completed at posttest).

Participants in the cognitive-behavioral depression prevention group had significantly greater reductions in self-reported substance use than did participants in the bibliotherapy intervention and brochure control groups at posttest, and greater reductions than participants in all comparison groups at six-month follow-up. The difference between cognitive-behavioral depression prevention group participants and supportive-expressive intervention group participants was marginally significant at posttest (p=.069).

Using the reliable change index, analyses from pretest to six-month follow-up found the following: reliable change rates for interviewer-rated depressive symptoms were significantly higher among cognitive-behavioral depression prevention group participants than participants in the brochure control group (OR=2.9); and reliable change rates for social adjustment were significantly higher among cognitive-behavioral depression prevention group participants than participants in the bibliotherapy intervention (OR=3.1) and brochure control groups (OR=2.5).

Cox proportional hazard regression models revealed that participants in the cognitive-behavioral depression prevention group showed a significantly lower risk for the onset of major depression than participants in the brochure control group (OR=2.5) at six-month follow-up.

A dose-response analysis revealed a significant, positive relationship between the number of sessions attended and interviewer-rated depressive symptoms for participants in the cognitive-behavioral depression prevention group at posttest (r=.27).

In formal mediation tests (Stice, Rohde, Seeley, & Gau, 2010), measures of negative cognitions and pleasant activities were improved by the intervention, affected an outcome measure of depression, and when serving as controls reduced the effect of the intervention on depression. However, tests also showed that, in contradiction to the mediation model, changes in depression largely occurred before changes in the negative cognitions and pleasant activities mediators.

Long-term effects: Stice, Rohde, Gau, and Wade (2010) report significant effects in Table 3 for 4 of 12 tests of the intervention group at one- and two-year follow-up:

Participants in the cognitive-behavioral depression prevention group had significantly greater reductions in interviewer-rated depressive symptoms than participants in bibliotherapy intervention and brochure control groups (but not in the supportive-expressive intervention group) at one-year follow-up, and greater reductions than bibliotherapy intervention group (only) participants at two-year follow-up. The difference between cognitive-behavioral depression prevention group participants and brochure control group participants was marginally significant at two-year follow-up (p=.056).

At the two-year (but not one-year) follow-up, participants in the cognitive-behavioral depression prevention group had significantly greater reductions in self-rated depressive symptoms than participants in the bibliotherapy intervention group. The difference between cognitive-behavioral depression prevention group participants and brochure control group participants was marginally significant at two-year follow-up (p=.087).

Using the reliable change index, analyses from pretest to two-year follow-up found the following: reliable change rates for interviewer-rated depressive symptoms were significantly higher among cognitive-behavioral depression prevention group participants than participants in the brochure control group (OR=2.21); and reliable change rates for self-rated depressive symptom were significantly higher among cognitive-behavioral depression prevention group participants than participants in the brochure control group (OR=2.11).

Cox proportional hazard regression models revealed that participants in the cognitive-behavioral depression prevention group showed a significantly lower risk for the onset of major or minor depression than participants in the brochure control group (OR=2.23) at two-year follow-up. A marginally significant difference (p=.093) was observed between participants in the cognitive-behavioral depression prevention group and the bibliotherapy intervention group, with participants in the latter group having a lower risk of major/minor depression onset.

Stice et al. (2011) found significant intervention effects on social support in five of 16 tests. The intervention group had significantly more friend social support than both the bibliographic and brochure control groups at posttest and six months and more than the brochure control group at one year. However, there were no significant effects at two years or for parent social support at any assessment. The significant effect sizes ranged from medium to small (r = .18-.29). The authors also tested for mediation, finding that reductions in depressive symptoms due to the intervention improved friend social support.

Gau et al. (2012) tested for moderation of intervention effects on depression symptoms by motivation to reduce depression, initial depressive symptom severity, negative life events, social support deficits, and substance use. Differential intervention effects emerged for two of the five moderators: the intervention had strong effects for adolescents at low and medium levels of substance use and negative life events, but not at high levels. The authors concluded that high-risk adolescents may require specialized depression prevention efforts.

Rohde, Stice, Gau, & Marti (2012) found that the intervention group reported significantly smaller increases in substance use relative to the control group at one year and relative to the bibliotherapy and control groups at two years. Mediational analyses estimated with Structural Equation Models indicated that reductions in depressive symptoms from baseline to posttest significantly mediated the effects of the intervention on changes in substance use over two years relative to brochure control participants but not the bibliotherapy participants.

Rohde, Stice, & Gau (2012) used Classification Tree Analysis to examine the onset of a depression disorder. They found that for participants with a negative or pessimistic attributional style, the bibliotherapy condition did more to reduce the onset than any of the other conditions.

Using combined data from both Study 1 and Study 2, Brière et al. (2014) estimated four trajectories in depressive symptoms over the two-year follow-up period. The results showed that the group-based intervention participants were less likely than control group participants to belong to trajectories of high and persistent symptoms and more likely to belong to trajectories of low and declining symptoms.

Using combined data from both Study 1 and Study 2, Müller et al. (2015) tested for eight moderators of the group-based and bibliotherapy intervention effects on depressive symptoms. Over the period from posttest to the 24-month follow-up, the tests found two significant moderation effects: High pretest depressive symptoms and high pretest motivation to change both increased the impact of the intervention.

Study 3

Summary

Stice et al. (2006) and Marchand et al. (2010) examined 225 students from two high schools and one college who were randomly assigned to one of six conditions and completed self-report surveys at pretest, posttest, and one-month and six-month follow-ups.

Stice et al. (2006) and Marchand et al. (2010) found that participants in the cognitive-behavioral depression prevention group, relative to the control group, had significantly greater reductions in

• Self-rated depressive symptoms at posttest and one-month follow-up

Evaluation Methodology

Design:

Recruitment /Sample size: Participants were recruited through mass mailings and e-mails, handbills, and posted fliers at two high schools and one college inviting students between the ages of 15 and 22 experiencing sadness to participate in the study. Interested students were given a depressive symptom screen (the Center for Epidemiological Studies-Depression Scale [CES-D]). Students who scored at or above a pre-determined cutoff point for inclusion (20 or above) were invited to complete the Beck Depression Inventory. Interested students with evidence of clinically significant depression were excluded and given treatment referrals. In total, 225 students were randomly assigned to condition.

Study type/Randomization/Intervention: Using blocks created by gender and school, eligible participants were randomly assigned to one of six conditions: (1) cognitive behavioral depression prevention (n=50), (2) supportive-expressive intervention (n=19), (3) bibliotherapy intervention (n=28), (4) expressive writing (n=27), (5) journaling (n=34), or (6) waitlist control (n=67). It is stated that conditions 2-5 were "added after the study was underway," but details regarding this process are not provided. Participants in the cognitive behavioral depression prevention group participated in the four-week cognitive behavioral program in groups composed of 6-10 participants. Students in the supportive-expressive intervention group participated in a four-week program, including weekly one-hour sessions, that provided a supportive group for participants to identify, express, and discuss their feelings. Subjects in the bibliotherapy intervention group were given a copy of the book Feeling Good, and encouraged to use the book as a self-help resource. Students in the expressive writing group were provided a quiet, private space three times over a three-week period to write about their intense emotional experiences during 45 minute sessions. The journaling intervention group was provided a journal and pen, and encouraged to write in their free time and/or to write at least once per week. Participants in the waitlist control condition were offered the cognitive behavioral depression prevention group intervention at the end of the study.

Assessment/Attrition: Measures assessing depressive symptoms were completed at pretest, posttest (one-month later), one-month follow-up, and six-month follow-up. The study reports that 33 participants (14%) did not provide complete follow-up data (i.e., complete measures at posttest, one-, and six-month follow-up); however, detailed information on attrition is not provided.

Sample characteristics:

Participants ranged in age from 15 to 22 years (with a mean age of 18.4 years) at pretest. The sample was primarily female (70%, 30% male), and primarily college students (59%, 41% high school students). Fifty-five percent of the participants were Caucasian, 17% Asian, 15% Hispanic, 6% Black, and 7% identified as other or mixed racial/ethnic category. Educational attainment of parents was 20% high school graduate or less, 20% some college, 34% college graduate, and 26% graduate degree.

Measures:

Self-reports of the severity of depressive symptoms experienced in the past week were collected using the Beck Depression Inventory (BDI). Reliability and validity of this measure has been established in prior published work.

Analysis:

Statistical methods/baseline control/correct unit of analysis: Omnibus repeated measures multivariate analysis of variance (MANOVA) models with group-by-time interactions were used to test for differential change in depressive symptoms. Due to the presence of significant group-by-time interactions, separate follow-up repeated measures ANOVA models contrasted pairs of conditions at each of the follow-up periods (from pretest to each follow-up). The pairwise comparisons identify which groups significantly differed, and when.

Operationalized as a Beck Depression Inventory score of 30 or greater, rates of severe depression onset (note that prospective participants scoring in this range at pretest were excluded from the study) were analyzed using a two-by-five chi-square model.

Intention-to-treat: The study employs intent-to-treat analyses by using full information maximum likelihood estimation to impute missing data and analyze all randomized subjects.

Outcomes:

Implementation fidelity: The researchers did not collect data on implementation fidelity. The study reports that program facilitators received training and were provided detailed manuals that included session content, verbatim scripts, as well as common problems and suggested solutions. For participants that missed a program session, a brief (10-15 minute) individual session was conducted with the participant prior to their next session in order to review missed material.

Baseline Equivalence: The groups did not differ (p > .10) on demographic characteristics, treatment-seeking rates during the study, or outcome measures at pretest.

Differential attrition: Attrition was not related to any pretest variables or outcome measures at pretest (p > .10). However, there was a marginal difference in attrition across condition (p=.08) with a higher rate of attrition in the cognitive behavioral depression prevention group (24%) relative to all other conditions.

Posttest: Participants in the cognitive behavioral depression prevention group had significantly greater reductions in self-rated depressive symptoms than participants in the waitlist control group at posttest (r=.48) and one-month follow-up (r=.28) (but not six-month follow-up). Participants in the cognitive behavioral depression prevention group also had significantly greater reductions in depressive symptoms than participants in the journaling intervention group at posttest (r=.23), and a marginally significant difference was observed between the cognitive behavioral and bibliotherapy intervention groups at six-month follow-up (p=.08). Positive program effects on the Beck Depression Inventory were statistically equivalent across race and ethnic groups (Marchand et al., 2010).

The chi-square analysis revealed no significant differences in the rates of severe depression onset across conditions.

Long-term effects: The study reports findings from measures collected through six-month follow-up, and thus does not demonstrate long-term effects.

Study 4

Summary

Rohde, Stice, Shaw, and Gau (2014) examined 82 students from one large state university who were randomly assigned to one of three conditions: cognitive-behavioral group depression prevention, cognitive-behavioral bibliotherapy, or brochure control. The students completed self-report surveys and diagnostic interviews at pretest, posttest, and 6-month and 12-month follow-ups.

Rohde, Stice, Shaw, and Gau (2014) found significantly higher scores among cognitive-behavioral depression prevention group participants when compared to control group participants on

• Cognitive behavioral knowledge at posttest

Evaluation Methodology

Design:

Recruitment /Sample size: Students from a large state university were recruited in 2010-2011 through direct mailings sent to a subset of first and second year students (no information is provided on the identification or characteristics of this subset), and recruitment posters hung on campus. The mailing included a depression screener based on the Center for Epidemiological Studies-Depression Scale (CES-D). Students who completed the CES-D and endorsed two or more depression symptoms were encouraged to enroll in the study. Eighty-five students met both the criteria for enrollment and completed a pretest interview assessing depressive symptoms. Following the pretest interview, three students were excluded from the study due to having a current diagnosis of major depressive disorder or acute suicidal ideation. The number of students (if any) who enrolled in the study but did not complete a pretest interview is not reported.

Study type/Randomization/Intervention: Eligible participants (n=82) were randomly assigned to one of three conditions: (1) cognitive behavioral group depression prevention (n=27), (2) cognitive behavioral bibliotherapy (n=22), or (3) brochure control (n=33). Participants in the cognitive behavioral depression prevention group completed six weekly one-hour program sessions in mixed-gender groups of four to eight participants. Students in the cognitive behavioral bibliotherapy group were given a personal copy of the book Feeling Good, which provides relevant and practical cognitive behavioral techniques for preventing and reducing negative moods. Participants in the brochure control group were given a National Institute of Mental Health educational brochure that describes major depressive disorder symptoms and treatment, and provides referral information.

Three female masters-level graduate students in clinical psychology completed a 4-hour workshop after reading the intervention script and supporting documents.

Assessment/Attrition: Participants completed a survey and diagnostic interview at pretest, posttest, and 6- and 12-month follow-ups. Attrition rates were 8.5% at posttest, 15.9% at 6-month follow-up, and 13.4% at 12-month follow-up.

Sample characteristics:

Participating students, all enrolled in college at a state university, were between 17 and 22 years of age (with a mean age of 19.0), and primarily female (69.5%). The sample was composed of 81% Caucasians, 11% Asian Americans, 3% African Americans, and 5% other or mixed race.

Measures:

Validity of measurements: All measures and scales have been used in prior published work and reliability and validity has been established. Assessors were blinded to subject condition.

Primary outcomes: Diagnostic interviews, using an adapted version of the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS), were used to assess the severity of major depressive disorder (MDD) symptoms and to determine whether participants met diagnostic criteria for MDD.

Secondary outcomes: Substance use was measured with 10 items from a previous study in which participants were asked to report the frequency of alcohol, drug, and cigarette use. Responses were averaged to form a substance use measure.

Social adjustment was measured using 13 items adapted from the Social Adjustment Scale-Self-Report for Youth.

Manipulation check: A manipulation check of the degree to which participants in the cognitive behavioral depression prevention and the cognitive behavioral bibliotherapy conditions acquired knowledge of the intervention concepts was completed using 15 True/False items designed for this study to assess knowledge regarding adolescent depression and cognitive behavioral interventions.

Analysis:

Statistical methods/baseline control/correct unit of analysis: Multilevel random effects growth models-with level one including individual change in outcomes and level two the predictor of condition-with pretest scores as a covariate were used to analyze program effectiveness. The examined parameters included a condition variable reflecting change in the outcome measure at posttest, and a group-by-time interaction variable reflecting change in the outcome measure from posttest to 12-month follow-up. The study appeared to take account of the clustering of intervention subjects within program groups of 4-8 subjects with random effects for condition.

Intention-to-treat: Multiple imputation was used to replace missing values so that all subjects were included in the analysis.

Outcomes:

Implementation fidelity: Among cognitive behavioral depression prevention participants, four students (14.8%) attended no sessions, 70% attended four to six sessions, and the mean attendance was 3.9 of 6 sessions. All group sessions were audiotaped and half were randomly selected for review for program adherence and facilitator competence. Using 1 to 10 point scales, mean adherence was 7.9, and mean competence was 7.3, with 1% of items rated as "minimal adherence" and 3% as "minimal competence."

Baseline Equivalence: At baseline, the groups did not differ on any demographic variables; however, the groups did differ on the measure of social adjustment at baseline.

Differential attrition: Attrition was not related to condition, nor did attriters significantly differ on any of the pretest study variables from participants who provided data.

Posttest and 12-month follow-up: No significant program effects were observed at posttest or 12-month follow-up (as measured at posttest by group and at follow-up by group-by-time interactions) for the primary or secondary outcomes.

Analyses of the manipulation check revealed that at posttest participants in the cognitive behavioral depression prevention program had significantly higher scores on the measure of cognitive behavioral knowledge when compared to participants in the brochure control condition.

Study 5

The program examined in the study had four weekly one-hour sessions facilitated by a female clinical graduate student and co-facilitated by a female undergraduate.

Summary

Burton et al. (2007) used a randomized controlled trial to examine 145 female high school and college students with feelings of depression who were assigned to the intervention group or a waitlist control group. Assessments of depressive symptoms, bulimic symptoms, and substance use occurred through a six-month follow-up.

Burton et al. (2007) found that the intervention group relative to the control group reported significantly greater reductions in

• Depressive symptoms (through the six-month follow-up)
• Bulimic symptoms (through the one-month follow-up).

Evaluation Methodology

Design:

Recruitment: The sample included 145 female students recruited from two high schools and one college. The recruitment used mass mailings (high school students) and emails (college students), handbills distributed after classes, and posted fliers that invited students between the ages of 14 and 23 suffering from feelings of depression to participate. Those eligible had to report elevated depressive symptoms on a screening measure.

Assignment: The study randomly assigned the 145 participants within blocks created by school to either the intervention condition or a waitlist control condition (but did not present the study-specific sample sizes). 

Assessments/Attrition: Assessments occurred at baseline, posttest, one-month follow-up, and six-month follow-up. About 14% of the participants "dropped from the study before providing follow-up data."

Sample:

The sample was composed of 18% Asians, 6% blacks, 52% whites, 19% Hispanics, 1% Native Americans, and 4% who specified ''other'' or mixed racial heritage. Educational attainment of parents, a proxy for socioeconomic status, was distributed as 22% high school graduate or less, 22% some college, 35% college graduate, and 21% graduate degree.

Measures:

The three outcome measures came from self-reports:

• Depressive symptoms from the Beck Depression Inventory-I (alpha = .88).
• Bulimic symptoms from the Eating Disorder Examination-Questionnaire (alpha = .83).
• Substance use from items on the frequency of consumption during the past month of beer/wine/wine coolers, hard liquor, cigarettes, marijuana, stimulants, downers, inhalants, and hallucinogens (alpha = .81).

In addition, the study reported test-retest reliabilities for the control group over the initial one-month period of .85 for depressive symptoms, .95 for bulimic symptoms, and .98 for substance use.

Analysis:

The analysis used repeated-measures analysis of variance models. Separate models tested for effects from pretest to posttest, pretest to one-month follow-up, and pretest to six-month follow-up. Full information maximum likelihood estimation, based on the expectation-maximization algorithm, was used to impute missing data.

Intent-to-Treat: The authors noted that "Participants who provided complete data attended all sessions and completed all homework assignments." This statement suggests that only those attending the sessions provided data, which may violate the intent-to-treat criterion.

Outcomes

Implementation Fidelity:

Not examined.

Baseline Equivalence:

The authors reported that the two conditions did not differ significantly on the three baseline outcomes or on age, ethnicity, or parental education.

Differential Attrition:

Attrition rates were similar across conditions (13% for the control group and 15% for the intervention group). Based on the overall attrition rate and the difference in attrition rates between conditions, the study met both the cautious and optimistic WWC standards. In addition, the authors stated that dropouts did not differ significantly from completers on any of the baseline variables. Still further, the use of FIML estimation should help moderate any attrition bias.

Posttest:

Depressive symptoms. Tests indicated significantly greater reductions in the intervention condition than the control condition at all three time points (from pretest to posttest, one-month follow-up, and six-month follow-up).

Bulimic symptoms. Tests indicated significantly greater reductions in the intervention condition than in the control condition at two time points (from pretest to posttest and one-month follow-up but not six-month follow-up).

Substance use. There were no significant differences across conditions in changes for any of the three time points.

Mediation. Results showed that change in depressive symptoms mediated the effect of the intervention on change in bulimic symptoms.

Long-Term:

Not examined.

Study 6

This French-language version of the Blues Program added a web-based homework component that included daily assignments over the six weekly sessions.

Summary

Brière et al. (2019) used a randomized controlled trial to examine 74 high school students with depressive symptoms who were assigned to the intervention group or a brochure control group. Assessments of depressive symptoms occurred at posttest and six-month follow-up.

Brière et al. (2019) found that the intervention group relative to the control group reported significantly

• Fewer diagnosed major depressive disorders (at six-month follow-up)
• Fewer diagnosed and self-reported depressive symptoms (posttest only)
• More pleasant activities (posttest only)
• More positive interactions with parents (posttest only).

Evaluation Methodology

Design:

Recruitment: Participants included 74 students from three large public French-speaking secondary schools located in disadvantaged areas of Montreal, Canada. Participants were recruited in Spring 2016 by describing the study to classrooms and setting up an information booth. Those eligible scored high on a screening measure of depressive symptoms but did not meet the criteria for a major depressive disorder.

Assignment: Eligible participants were randomized with stratification by sex and school to an intervention condition (n = 37) or control condition (n = 37). Control participants received a brochure on depression from a non-profit mental health organization.

Assessments/Attrition: Assessments occurred at pretest, posttest, and six-month follow-up. Attrition was 3% at posttest and 4% at the six-month follow-up.

Sample:

The sample of students ages 14-18 was 66% female, 71% Canada-born, and 29% born elsewhere.

Measures:

The eight outcome measures came from self-reports and a diagnostic interview conducted by research assistants who were blind to condition. The authors divided the measures into three groups.

Primary Outcome:

• Major depressive disorder diagnosis from pretest to six-month follow-up using the Structured Clinical Interview for DSM-IV Disorders.
• Depressive symptoms from the Structured Clinical Interview for DSM-IV Disorders (alpha = .78-.80
• Depressive symptoms from the self-reported Center for Epidemiological Studies - Depression (alpha = .88-.91).

Intermediate Outcomes:

• Negative thoughts from the automatic thoughts questionnaire (alpha = .91-.94).
• Pleasant activities from a subset of items of the pleasant events schedule (alpha = .68-.80).

Secondary Outcomes:

• Social adjustment from the Évaluation sociale de soi chez les jeunes adultes (Social self-evaluation in young adults), which assessed feelings of social acceptance, popularity, and interest in social relationships (alpha = .77-.84).
• Anxious symptoms from the social phobia and generalized anxiety subscales of the Spence Children Anxiety Scale (alpha = .78-.84).
• Interactions with parents from the Mesures de l'Adaptation Sociale et Personnelle des Adolescents Quebecois (Measure of the Social and Personal Adjustment of Quebec Adolescents) (alpha = .73-.80).

Analysis:

The analysis used logistic regression of the binary measure of major depressive disorder and a directional one-sided test at alpha = .05. The analysis of the other continuous measures used linear regression with the baseline outcome as a covariate. The models used maximum likelihood estimation with robust standard errors that adjusted for clustering within intervention groups.

Intent-to-Treat: The analysis retained all participants with the maximum likelihood estimation.

Outcomes

Implementation Fidelity:

Two participants (5%) dropped out of the program, and the participants who completed the program attended 85% of sessions on average. Participants completed 8.2 daily homework on the Web site out of a possible 35 during the program (23%). On a scale of 1-3, the average session adherence score was 2.3 (slightly above a rating of "presented with minor omission or errors"). On a scale of 1-5, the average facilitator competence score was 4.1 (near a rating of "very good").

Baseline Equivalence:

The authors stated that "conditions did not differ on sociodemographic variables, as well as primary, intermediate, or secondary outcomes at pretest."

Differential Attrition:

Attrition was minimal.

Posttest:

By the six-month follow-up, one intervention participant (3%) had developed a diagnosed major depressive disorder, compared to five control participants (15%). Logistic regression indicated that control participants were significantly more likely than intervention participants to develop the disorder (OR = 6.0).

The intervention group scored lower than the control group at posttest on both primary measures of diagnosed and self-reported depressive symptoms (d = -.51 and -.40), but the effects dropped to insignificance by the six-month follow-up. Of the five additional measures, two reached significance, with the intervention group scoring higher than the control group at posttest on pleasant activities and interaction with parents, (d = .49 and .34, respectively), but these effects also fell to insignificance by the six-month follow-up.

Long-Term:

Not examined.

Study 7

Summary

Shomaker et al. (2016) used a randomized controlled trial to examine 119 adolescent girls with elevated depression symptoms who were assigned to the cognitive-behavioral intervention group or health education group. Assessments of depression symptoms and physical health occurred at posttest.

Shomaker et al. (2016) found that, relative to the control group, the intervention group showed greater

• Cortisol reactivity at posttest.

Evaluation Methodology

Design:

Recruitment: The study recruited a sample of 119 adolescent girls (ages 12-17) whose parents were concerned about their daughter developing type 2 diabetes. Recruitment occurred within a 60-mile radius of Bethesda, Maryland, via websites, community postings, letters/flyers to physician offices and schools, advertisements, and direct mailings to homes. Inclusion criteria included being overweight or obese, being at risk for type 2 diabetes, having mild-to-moderate depressive symptoms, and otherwise having good general health. Exclusion criteria included current psychiatric symptoms that necessitated treatment; a major medical problem; medication use that could affect insulin resistance, body weight, or mood; and current involvement in structured weight loss or psychotherapy, or pregnancy.

Assignment: The study randomized the 119 girls using permuted blocks and stratification by age and race/ethnicity to the cognitive-behavioral intervention (n = 61) or a health-education control group (n = 58). Both interventions met once a week over six weeks.

Assessments/Attrition: Assessments occurred at baseline and posttest (within 2 weeks of group completion). The posttest had a 91% completion rate.

Sample:

The participants averaged 15 years of age and were 62% Black, 16% white, 11% Hispanic, 3% Asian, and 8% multiracial.

Measures:

The measures came from physiological assessments, adolescent self-reports, and objective counts of calories consumed and exercise completed. Assessors of all key outcomes were blind to group assignment. The key self-report measure of depression used a previously validated scale, but reliability information was not reported for the study sample.

The primary outcomes were depressive symptoms and insulin sensitivity, and the exploratory outcomes related to blood sugar, eating behavior, physical fitness, and cortisol stress. The measures included behavioral outcomes of depression, obesity, and physical health and well-being, activity, while the risk and protective factors related to food intake and physiological traits.

Analysis:

For continuous measures, the study used analyses of covariance with change in outcome as the dependent variable and multiple covariates (including the baseline outcome, sociodemographic characteristics, diabetes family history, and body fat). For binary outcomes, the study used logistic regression and the same covariates but without a change score as the dependent variable.

Intent-to-Treat: The study used multiple imputation for missing data (8.6% of all data points) so that "The intent-to-treat sample consisted of all participants who were randomized regardless of whether they withdrew or were excluded from the study after randomization." Five participants were withdrawn because they no longer met the eligibility criteria but were included in the analysis.

Outcomes

Implementation Fidelity:

For the cognitive-behavioral intervention, the median program attendance was five of six sessions. Median expert ratings on a scale of 1 to 10 were 7.8 for facilitator competence and 7.7 for fidelity.

Baseline Equivalence:

Table 1 showed two significant condition differences in 22 tests. Participants randomized to the cognitive-behavioral intervention had lower mean scores on the body mass index and the standardized body mass index.

Differential Attrition:

Attrition rates did not differ significantly across conditions (11% versus 7%). The authors also reported that those completing the posttest did not significantly differ from those who did not on any variable (p>0.08).

Posttest:

The conditions did not differ significantly at posttest on the two primary outcomes - depression and insulin resistance. One of 11 exploratory outcomes - peak cortisol reactivity - differed significantly across conditions and appeared to favor the cognitive-behavioral intervention group.

Tests for moderation found a program effect on depression for girls with moderate (vs mild) baseline depressive symptoms.

Long-Term:

Not examined.

Study 8

This study sought to evaluate a mindfulness program for depression (Learning to BREATHE) and used the Blues Program as a control or comparison group.

Summary

Shomaker et al. (2017, 2019) used a randomized controlled trial to examine 33 adolescent girls with elevated depression symptoms who were assigned to the cognitive-behavioral intervention group or a mindfulness intervention group. Assessments of depression symptoms and physical health occurred at posttest, six-month follow-up, and one-year follow-up.

Shomaker et al. (2017, 2019) found no beneficial program effects.

Evaluation Methodology

Design:

Recruitment: Participants were 33 girls ages 12-17 with overweight/obesity, elevated depression symptoms, and a family history of Type 2 diabetes. They were recruited for a diabetes trial, conducted in the western region of the United States between January 2014 and August 2016. The recruitment used flyers to schools and physicians, community postings, advertisements on e-mail listservs, and direct mailings to local-area homes. The exclusion criteria included pregnancy, a major depressive disorder or diagnosis warranting treatment, a major medical problem (e.g., Type 2 diabetes), medication affecting mood/insulin, or psychotherapy.

Assignment: A randomization using permuted blocks and stratification by age and race/ethnicity assigned the 33 girls to a six-session mindfulness-based intervention group (Learning to BREATHE; n = 17) or six-session cognitive-behavioral depression prevention group (Blues Program; n = 16). One clinical psychologist led sessions for both conditions, thereby minimizing facilitator effects.

Assessments/Attrition: Assessments came at baseline, posttest, six-month follow-up, and one-year follow-up. Completion rates of 85%, 82%, and 76% occurred at posttest, six months, and one year, respectively.

Sample:

The girls in the sample were ages 12-17 (mean 15) and came from several racial/ethnic backgrounds (70% non-Hispanic White, 21% Hispanic, 9% American Indian).

Measures:

The outcome measures came from blood samples taken by trained research associates who were blind to condition or from self-reports given to assessors who were not consistently blinded. The study offered validity and reliability information from other studies but presented no figures for the current sample.

• Depression symptoms from the Center for Epidemiologic Studies-Depression scale.
• Anxiety symptoms from the State-Trait Inventory for Children-Trait Version.
• Global perceived stress from the 14-item Perceived Stress Scale.
• Insulin resistance from blood samples and lab analysis.
• Fasting insulin from blood samples and lab analysis.
• Fasting glucose from blood samples and lab analysis.
• Dispositional mindfulness from the Mindful Attention Awareness Scale.
• Body fat from dual-energy x-ray absorptiometry.
• BMI score and z-score from weight and height measures. 

Analysis:

The analyses used analysis of covariance models with change scores serving as the outcome and with covariates for race/ethnicity, baseline outcome, age, weight status, puberty, adiposity, and one-year adiposity change. Multiple imputation (Shomaker et al., 2017) or expectation-maximization imputation (Shomaker et al., 2019) were used for missing data. Shomaker et al. (2017) adjusted outliers (3% of all data points) to 1.5 times the interquartile range below or above the 25th or 75th percentile.

Intent-to-Treat: Imputation of missing data allowed for use of all randomized participants.

Outcomes

Implementation Fidelity:

Median attendance was five sessions in the cognitive-behavioral intervention, with 87% attending 80% or more of the sessions.

Baseline Equivalence:

Shomaker et al. (2017) reported that the mindfulness condition had more adolescents with obesity (70.6%) than the cognitive-behavioral intervention (37.5%) - a large difference that was not statistically significant (p = .08). The conditions did not differ significantly on any other baseline characteristic (p > .10). Shomaker et al. (2019) listed the baseline means for the sociodemographic and outcome measures in Supplemental Table 1 but did not include tests for significance or effect sizes.

Differential Attrition:

Shomaker et al. (2017) reported large condition differences in retention at posttest (76% versus 94%) and at six-month follow-up (71% versus 94%). Shomaker et al. (2019) reported a somewhat smaller difference at the one-year follow-up (71% versus 81%) and stated that "Baseline characteristics were not significantly related to retention."

Posttest:

In Shomaker et al. (2017), the cognitive-behavioral intervention participants did significantly worse than the mindfulness alternative at posttest on three of seven outcomes: depressive symptoms, fasting insulin, and insulin resistance. Only depressive symptoms remained significant at the six-month follow-up.

Long-Term:

The cognitive-behavioral intervention did significantly worse than the mindfulness alternative on two of six outcomes. It had smaller decreases in depression and smaller decreases in insulin resistance.