A Randomized Placebo-Controlled Trial of a School-Based Depression Prevention Program

SALLY MERRY, F.R.A.N.Z.C.P., HEATHER MCDOWELL, PH.D., CHRIS J. WILD, PH.D., JULLIET BIR, M.SC., RACHEL CUNLIFFE, B.SC.
AND

ABSTRACT Objective: To conduct a placebo-controlled study of the effectiveness of a universal school-based depression prevention program. Method: Three hundred ninety-two students age 13 to 15 from two schools were randomized to intervention (RAP-Kiwi) and placebo programs run by teachers. RAP-Kiwi was an 11-session manual-based program derived from cognitive-behavioral therapy. The placebo was similar but with cognitive components removed. Outcomes were self-rated depression scales, the Reynolds Adolescent Depression Scale (RADS), and the Beck Depression Inventory II (BDI-II). Follow-up was to 18 months. Analysis was done on an intent-to-treat basis. Results: Immediately after the intervention, depression scores were reduced significantly more by RAP-Kiwi than by placebo, with a mean difference in change from baseline between groups of 1.5 on BDI-II (CI > 0.38, p = .01) and 2.24 on RADS (CI > 0.22, p = .04). Categorical analysis confirmed significant clinical benefit with an absolute risk reduction of 3% (95% CI, 111%, McNemar 2, p = .03), with the number needed to treat for short-term benefit of 33. Group differences in depression scores averaged across time to 18 months were significant on RADS but not on BDI-II. Retention rates were 91% at 6 months and 72% at 18 months. Conclusions: The RAP-Kiwi program is a potentially effective public health measure. Confirmation of effectiveness measuring episodes of depressive illness and broader measures of adjustment is warranted. J. Am. Acad. Child Adolesc. Psychiatry, 2004;43(5):538547. Key Words: depression, prevention, adolescent, randomized controlled trial.

Depression is a major cause of morbidity and mortality among adolescents. It is associated with poor academic performance, social dysfunction, substance abuse, suicide attempts, and completed suicide (NHMRC, 1997). It is considered the fourth most important disease in the estimation of disease burden (Murray and
Accepted November 19, 2003. Drs. Merry and McDowell and Ms. Bir are in the Department of Psychiatry, Prof. Wild and Ms. Cunliffe in the Department of Statistics, University of Auckland. The research was supported by the Health Research Council, Grant No. 99/039. Our thanks to the students and teachers who participated in the study; Dr. Shochet for generous support; Ms. Muller, Mr. Wharemate, Mr. Dutt, Dr. Crengle, Mr. Stewart, Mr. Leo, and Dr. Foliaki for advice over cross-cultural issues; Ms. Wills, project coordinator; Dr. Plunket, who designed the graphics; Prof. Merry and Dr. Frampton for advice about design and analysis. Correspondence to Dr. Merry, Department of Psychiatry, University of Auckland, Private Bag 92019, Auckland, New Zealand; email: s.merry@auckland. ac.nz. Manuals from Dr. Merry on request. 0890-8567/04/430505382004 by the American Academy of Child and Adolescent Psychiatry. DOI: 10.1097/01.chi.0000117063.63530.68

Lopez, 1996). Up to 24% of young people will have suffered from at least one clinically significant depressive episode by age 18 years (NHMRC, 1997). Because of this, there has been interest in the development of programs aimed at preventing the onset of depressive disorder.
DEPRESSION PREVENTION PROGRAMS

Depressive disorder usually starts in adolescence, with a marked increase in period prevalence estimates from middle to late adolescence (Fergusson and Horwood, 2001; Hankin et al., 1998), making this a good time to intervene to prevent onset. Depression prevention programs may include programs that are targeted (i.e., delivered to those at risk of disorder) or universal (i.e., delivered to the whole population). Targeted prevention programs can be divided into selective approaches aimed at people at increased risk of the disorder and indicated approaches directed at individuals who have signs or symptoms that fall short of those
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

538

EFFECTIVENESS OF DEPRESSION PREVENTION

needed for a diagnosis. A seminal article by Jaycox et al. (1994) reported encouraging results following a targeted depression prevention intervention, with reduction in depressive symptoms that persisted to 24 months (Gillham and Reivich, 1999). Although there were methodological weaknesses in this study, it led to a number of studies published in the past decade. There are some difficulties with a targeted approach. Selective interventions have been delivered to children whose parents have a history of depression (Beardslee et al., 1997b; Clarke et al., 2001) or whose parents have separated (Gwynn and Brantley, 1987). Accessing all those at risk is problematic. A well-designed study by Clarke et al. (2001) highlighted some of the difficulties that may arise. Offspring of parents who had been treated for affective disorder were identified through their health care provider. Of a potential pool of 3,374 youths, 551 youthparent dyads were interviewed, and 94 youths were deemed at risk. Although the results of the study are encouraging, the process of identification would be difficult on a large scale and would miss youths whose parents had untreated mood disorder or whose parents were not depressed. Indicated programs have also been reported (Clarke et al., 1995; Lamb et al., 1998; Petersen et al., 1997; Seligman et al., 1999; Thompson et al., 2000). One of the more promising approaches is to screen for elevated depressive symptoms within schools, but this may result in stigmatization of the high-scoring pupils, and this approach ignores a large group of students who may not be symptomatic at the time of the intervention but may generate more cases of depressive illness than the high-risk group (Rose, 1992). Universal interventions may therefore be warranted, and delivery at school within the school day would ensure access to the program for the majority of youths.
PLACEBO STUDIES

Stevenson, 2001). Studies that did report efficacy compared the prevention program with being on a waitlist (Hains, 1994; Lowry-Webster et al., 2001; Quayle and Dziuraweic, 2001), with no intervention (Cardemil et al., 2002; Jaycox et al., 1994; Lamb et al., 1998; Petersen et al., 1997; Seligman et al., 1999; Spence et al., 2003; Thompson et al., 2000; Yu and Seligman, 2002), and with being in a usual class or having usual care (Clarke et al., 1993, 1995; Shochet et al., 2001), and one compared two interventions (Beardslee et al., 1997a). Because of the absence of a definitive answer on the effectiveness of depression prevention programs, we concluded there was a need for a prospective placebocontrolled trial of an intervention in a format that would allow practical delivery within a population with a sample size large enough to detect change. We report the results of such a trial. In New Zealand the Treaty of Waitangi, signed in 1840, is the founding document of the country. Under the Treaty there is a requirement that ventures, such as this study, be done in partnership between New Zealanders of predominantly British descent, known in this country as Pakeha, and the indigenous Maori people. Our hypotheses were that (1) the intervention program called RAP-Kiwi would reduce depressive symptoms immediately after delivery compared with an active placebo; (2) the effect would persist for 12 months; and (3) the program would be equally effective for Pakeha and Maori people.
METHOD Study Participants Participants were recruited from two different years in two schools in Auckland, New Zealand. One of these schools, school A, was from a lower socioeconomic urban area; the other, school B, was from a middle-class rural district. The schools were selected on the basis of their ethnic mix, almost purely Maori and Pakeha, and the numbers on their rolls. Most other schools in Auckland have a substantial number of students of Pacific Island or Asian ethnicity. All students in year 10 at school A and year 9 at school B were invited to participate. These years were chosen in consultation with the schools so that the study could fit the school curriculum and so that 13- to 14-year-old students could be recruited. Procedure Students and parents/caregivers were informed about the study and students were included if they could speak English and they and their parents/caregivers gave written consent to participation.

A major weakness of studies of depression prevention to date has been the lack of active comparison groups. The strength of the placebo effect in the reduction of depressive symptoms is well established (Shapiro and Shapiro, 1997). Only two reported randomized controlled trials have used an active comparison. Neither showed efficacy, but the condition described as placebo in one included active therapeutic elements (Shatte, 1997) and the other did not have power to detect a difference (Pattison and Lynd-

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

539

MERRY ET AL.

Participating students were given a study number. A research assistant who did not know the pupils used these numbers and randomization tables to assign students to RAP-Kiwi or placebo groups. RAP-Kiwi and the placebo comprised 11 sessions designed to run as regular classes. To fit the school timetable, in one school the program was run twice a week in year 9 (13-year-olds) and in the other it was run weekly in year 10 (14-year-olds). Intervention and Placebo Programs Both the RAP-Kiwi and placebo interventions had group leader and participant manuals. RAP-Kiwi was adapted from the Resourceful Adolescent Program (RAP), designed in Australia. The program incorporates cognitive-behavioral and interpersonal therapy principles and has been shown to be effective in a cohort study comparing the program with no intervention (Shochet et al., 2001). To ensure applicability to New Zealand teenagers, we kept the overall structure but adapted the language, cartoons, and anecdotes in consultation with youths. Titles and key messages were provided in English and Maori. Input was provided from Asian and Pacific advisers to ensure acceptability to the small number of youths from these ethnic groups attending the schools we selected. Teachers were consulted about the content and layout and their suggestions incorporated. The New Zealand version was titled RAP-Kiwi. To control for the nonspecific effects of a novel intervention in the school, a comparison placebo program was devel-

oped with a focus on having fun but without elements thought to be active in preventing depression. Manuals for intervention and placebo were designed to look alike and could be distinguished easily only by color. A summary of the content of sessions in both programs is outlined in Table 1. The placebo group had the experience of being in a study and of meeting weekly in a group with a supportive adult, and had time away from regular classes. They worked from participant manuals similar to RAP-Kiwi, but the activities were around arts and crafts rather than cognitivebehavioral therapy. Program Integrity The Australian team who developed the program trained our research team. We then trained teachers from the two schools over 2.5 days. Teachers running the placebo program were trained separately. When running the programs, all teachers completed a weekly integrity checklist and met a research team member weekly to ensure compliance. Measures Used The Beck Depression Inventory II (BDI-II) (Beck et al., 1996) and the Reynolds Adolescent Depression Scale (RADS) (Reynolds, 1986) were given before and after the program and at 6, 12, and 18 months of follow-up. Both are well-established measures of depres-

TABLE 1 Content of Programs Intervention Session 1 Introduction Rate self-esteem Complete depression rating scales Define self-esteem Discuss ways of building self-esteem Include why doing things improves mood Identify negative cognitions Link self-awareness and mood Introduce RAP model Stress management Relaxation techniques Personalize stress management Introduce cognitive restructuring Identify negative cognitions Use of Socratic questioning Teach cognitive restructuring Group activity to solve a problem Problem-solving techniques made explicit Accessing support Use of humor Conflict resolution 1 Link to the overall RAP model Conflict resolution 2 Role plays Review RAP model and party Certificate presentation Placebo Introduction Complete depression rating scales Identify activities that are fun Spend time doing collage Use of body language in communication

Session 2

Session 3

Session 4

Design a chill-out room Discuss effect of environment on mood Design a family coat of arms

Session 5

Session 6 Session 7 Session 8 Session 9 Session 10 Session 11

Changing your minda debate Group activity to solve a problem Art activity Discuss how to give support and help others Community service Use activities to identify differences in beliefs Group decision making in designing a planet Fun activity Fun activity and party Certificate presentation

Note: RAP = Resourceful Adolescent Program.

540

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

EFFECTIVENESS OF DEPRESSION PREVENTION

sive symptoms with acceptable reliability and validity for adolescents (Beck et al., 1996; Reynolds and Mazza, 1998). So that students with significant levels of depressive symptoms were not left untreated, students were referred to school guidance counselors for appropriate intervention if they scored (1) 23 or more with a score of 2 or 3 on items 2 or 9 on the BDI-II or (2) 30 or more on the BDI-II or (3) 3 on question number 9 on the BDI-II or (4) 77 or more or scored 4 of the critical items on the RADS. These students still participated in the program, but their results were not included in the analysis. Demographic data of age, gender, and ethnicity were collected from the students at the start of the program. Students rated the program for enjoyment and usefulness on Likert scales at the end of the intervention. Teachers provided verbal feedback in a meeting with the researchers. Blinding Participants were blind to which intervention they received. To check the effectiveness of the blinding, students were asked whether they thought they were in the active or the comparison program at the end of the study. However, the teachers who ran the program were not blind. In one of the schools, extra teachers were recruited so that usual classes could be divided in two. One of the conditions of running the program in this school was that the teachers on the permanent staff were trained to run RAP-Kiwi. In the other school we were able to use two of the usual teachers, one for the intervention and one for the placebo, as they normally divided their classes into two to teach the health syllabus. Teachers had access only to the program they were delivering. They were trained separately and asked not to discuss the content of the programs with each other. Students used manuals during the sessions but were not allowed to take them away until the end of the program. Statistical Analysis Analysis of the results was done using SPSS Version 11.0 and R (Ihaka and Gentleman, 1996). The differences from baseline were compared between the two programs by an independent-samples t test. It was assumed that the intervention would be better than placebo; therefore, one-tailed t tests were used. The joint effects of gender, ethnic group, and school were investigated by analysis of variance. The McNemar 2 test was used to test categorical changes between BDI-II categories of minimal or mild and moderate or severe (Beck et al., 1996). The area under the curve of the two depression scales was calculated for the period from the start of the study to the 18-month follow-up. Mean areas under the curve were compared by independent-samples t tests. All analyses were on an intent-to-treat basis. Power Calculations Power calculations based on results of studies to date indicated that 75 subjects and 75 controls gave 80% power to detect one unit of change on the CDI (equivalent to the BDI-II) at 5% level of significance (one-sided). An overall sample size of 200 in each group, of whom 70 were Maori, gave a reasonable chance of detecting a positive treatment effect in Maori and allowed some protection from the inevitable within-subject variability. Ethics The study was approved by the Human Subjects Ethics Committee of the University of Auckland.

RESULTS

Of 540 students, a total of 392 (72.6%) could speak English, agreed to participate, and returned written consent forms; of these, 189/318 (59%) were from school A and 203/222 (91%) were from school B. The main reason for nonparticipation was failure to return a signed consent form from parents. We had 7 active refusals from school A and 12 from school B. The 392 students were randomized to intervention (207) and control (185). Eleven students in each group were referred to the school guidance counselors as their depression scores were over the predetermined cutoffs. Although these students continued to take part in the study, their results were excluded from the analysis. Follow-up of over 90% of participants to 6 months and over 70% to 18 months, with over 80% in school B, was achieved (Fig. 1). Some students were absent for one or more administrations of the depression rating scales.

Fig. 1 Participant flow chart. *Reasons = No consent form or did not wish to participate. **Missed first administration (n = 4), referred to school guidance counselor (n = 11). ***Missed first administration (n = 2), referred to school guidance counselor (n = 11). Missing = withdrew, absent, or lost to follow-up. Throughout diagram % is percentage of the group originally recruited

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

541

MERRY ET AL.

The groups were well balanced on key demographic variables, and although the subjects reported more symptoms of depression than the placebo group at baseline, these differences were not significant (Table 2). Mean scores, standard deviations, and mean change from score at entry to the study on the two depression measures are shown in Table 3.
Immediate Effect of Intervention

The mean reduction from baseline score on the BDI-II was 1.82 for RAP-Kiwi and 0.32 for placebo; on the RADS it was 2.31 for RAP-Kiwi and 0.07 for placebo. This was a significant difference (independent-samples t test: BDI-II, df = 329, p = .01; RADS, df = 324, p = .04) (Table 3). With 95% confidence we
TABLE 2 Demographics RAP-Kiwi (n = 192) Age at start Mean years (SD) Gender Male Female Ethnicity Pakeha Maori Pacific people Asian Other School School A Decile 2b School B Decile 6a BDI-II at start, mean (SD) RADS at start, mean (SD) 14.2 (0.6) 103 (53.6%) 89 (46.4%) 115 (59.9%) 47 (24.5%) 19 (9.9%) 2 (1.0%) 9 (4.7%) 96 (50.0%) 96 (50.0%) 9.3 (7.0) 55.4(13.4) Placebo (n = 172) 14.2 (0.7) 73 (42.4%) 99 (57.6%) 99 (57.6%) 51 (29.7%) 10 (5.8%) 2 (1.2%) 10 (5.8%) 82 (47.7%) 90 (52.3%) 8.5 (6.7) 53.8(11.7)

estimated that the difference between groups for BDIII was at least 0.38; for RADS it was at least 0.22. There was no evidence of a difference in effect by gender, ethnicity, or school. Using the categories defined in the manual (Beck et al., 1996) for the BDI-II, 156 students remained unchanged after RAP-Kiwi, 16 students changed from moderate/severe to minimal/mild categories, and 5 students changed from minimal/mild to moderate/severe categories (net improvement of 11 students). After placebo, 139 students remained the same, 6 improved, and 9 deteriorated (net deterioration of 3 students). This was a significant difference (McNemar 2, p = .03). The absolute risk reduction was 3% (95% CI = 111%), and number needed to treat (NNT) (the reciprocal of the absolute risk reduction (Sackett et al., 1997, p. 136) was 33 (CI 9100) for short-term benefit. This means that to move one student from the moderate/severe to the minimal/mild category on the BDI-II, on average 33 students would need to receive the intervention.
Effect Over 18 Months

Depression scores for both groups remained below baseline over time to 18 months (Table 3). The change from baseline on the RADS was consistently greater following RAP-Kiwi than placebo (Fig. 2a). The mean difference for the RAP-Kiwi group was 1.55 (SD = 1.03); for the placebo group it was 1.31 (SD = 1.02). Using log-transformed data, the area under the curve was significantly greater for the RAP-Kiwi group (p = .033, df = 341, CI = 0.020.46). There were no longterm significant differences on BDI-II (Fig. 2b). There was no evidence of a differential effect by gender, ethnicity, or school on these results.
Participant Feedback on the Program

Note: BDI-II = Beck Depression Inventory II; RADS, Reynolds Adolescent Depression Scale. a Decile 6 schools are those that serve a population in which 20.6% of households falls into the lowest 20% equivalent income, 24.2% of parents have no formal qualifications, 18.7% of parents receive income support, 16.1% of parents are in low-skilled occupational groups, and 19.5% of students are of Maori or Pacific ethnicity. b Decile 2 schools are those that serve a population in which 38.1% of households fall into the lowest 20% equivalent income, 39.5% of parents have no formal qualifications, 31.8% of parents receive income support, 26.1% of parents are in low skilled occupational groups and 64.3% of students are of Maori or Pacific ethnicity.

On 5-point Likert scales, where 5 was the most positive score and 1 the most negative, the programs were rated by the students as reasonably enjoyable (RAPKiwi mean score 3.0, SD = 1.1; placebo mean score 3.7, SD = 1.0, p = NS) and useful (RAP-Kiwi mean 2.9, SD = 1.1; placebo mean 3.1, SD = 1.1, p = NS).
Teacher Feedback on the Program

Teachers were less positive about the program than the students. They did not like being tied down to delivery as prescribed in the manual and felt the proJ. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

542

EFFECTIVENESS OF DEPRESSION PREVENTION

TABLE 3 Comparison of Depression Scores From Baseline to Follow-Up RAP-Kiwi n BDI (pre) BDI (post)b BDI (6 mo) BDI (12 mo) BDI (18 mo) RADS (pre) RADS (post)c RADS (6 mo) RADS (12 mo) RADS (18 mo) 177 177 178 153 136 178 178 179 153 136 Mean (SD) 9.40 (7.14) 7.58 (7.21) 8.41 (9.47) 7.99 (10.08) 7.90 (10.26) 55.60 (13.17) 53.28 (16.38) 53.97 (18.38) 52.32 (19.18) 50.89 (16.67) Mean Change (SD) from Baselinea 0 1.82 (6.62) 0.95 (8.72) 1.48 (10.31) 1.15 (10.43) 0 2.31 (14.33) 2.04 (15.47) 2.99 (17.01) 3.26 (16.12) n 154 154 155 142 126 154 154 155 142 127 Mean (SD) 8.21 (6.59) 7.90 (7.20) 6.62 (7.19) 7.85 (9.28) 6.10 (7.01) 53.60 (11.54) 53.53 (14.20) 51.94 (15.56) 53.59 (17.31) 51.69 (16.79) Placebo Mean Change (SD) from Baselinea 0 0.32 1.48 0.23 2.15 (5.79) (6.62) (9.99) (7.23)

0 0.07 (10.77) 1.55 (13.36) 0.21 (16.64) 1.80 (17.11)

Note: Abbreviations are explained in the first footnote of Table 2. a Baseline is the score at entry to the study. b Significant differences between groups in change from baseline (p = .01). c Significant differences between groups in change from baseline (p = .04).

gram would be greatly enhanced if they had been able to adapt it to the needs of the individual classes. They also felt they could have improved on the way some of the concepts were taught.
Effectiveness of Allocation Concealment

At the end of the study we asked students to guess whether they were in the active intervention group or not. Overall 11% guessed which group they were in correctly, 12% guessed incorrectly, and 77% said they did not know. In the RAP-Kiwi group, 14% guessed correctly, 15% guessed incorrectly, and 71% did not know; in the placebo group, 8% guessed correctly, 9% guessed incorrectly, and 82% did not know.
DISCUSSION Summary of Findings

Fig. 2 A: Change in Reynolds Adolescent Depression Scale score from baseline. Mean difference for RAP-Kiwi = 1.55 (SD 1.03), for placebo = 1.31 (SD 1.02). p = .033, df = 341, CI 0.020.46 for log-transformed area under the curve. B: Change in Beck Depression Inventory II score from baseline. There was no significant difference between groups.

In this, the first substantial study to compare a depression prevention program with a placebo, we have shown a significant effect for RAP-Kiwi immediately after intervention on both measures of depression. Following RAP-Kiwi the scores remained lower at all time points at follow-up to 18 months on the RADS but not on the BDI-II when averaged across time (the area under the curve calculations). This is suggestive of a persisting effect that is small but statistically significant. We consider this an important finding. The reduction in depression scores following RAP-Kiwi compared
543

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

MERRY ET AL.

with placebo is encouraging and occurred in both schools, despite their different socioeconomic status, and across gender. The depression scores in the placebo group also fell, indicating an expected placebo response. Normally a rise in depressive symptoms is expected at this age (Fergusson and Horwood, 2001; Hankin et al., 1998). While significant differences were shown on both depression measures immediately after intervention, it is interesting that a difference between RAP-Kiwi and placebo was demonstrated on the RADS and not on the BDI-II when averaged across the follow-up period. This may be because the two scales have different time frames, or because they measure different aspects of depression. The failure to show a significant effect on the BDI-II may also be due to the variability within the data, as the standard deviation for the BDI-II was nearly as great as the mean score, making it harder to demonstrate statistical significance. Whether a reduction in subsyndromal depressive symptoms constitutes treatment or prevention has been debated. The Institute of Medicine report on prevention of mental illness recommends that the term prevention be used for interventions for symptomatic individuals whose symptoms are still early and are not sufficiently severe to merit a diagnosis of a mental disorder (Mrazek and Haggerty, 1994, p. 25). The assumption that a reduction in scores postpones or moderates the emergence of a full disorder is supported by the study reported by Clarke et al. (2001), where reduction in depressive symptoms was associated with a reduction in the emergence of depressive disorder. Students in our study whose scores fell into the clinical range at the start of the study were referred for care, and although they continued to participate in the program, their results were not included in the analysis. The dangers of not controlling for placebo effects are clear from comparisons of controlled and uncontrolled medication trials. One review showed that 83% of treatments in uncontrolled trials were effective compared with 25% in controlled studies (Foulds, 1958, cited in Shapiro and Shapiro, 1997). The importance of considering the placebo effect in trials of psychotherapy is well argued by Shapiro and Shapiro, who caution that psychological interventions may appear to be effective not because of the theories or therapeutic procedures but because of underlying, unspecified or not clearly determined non-specific effects (Shapiro and Shapiro, 1997, p. 103). This may well apply to
544

prevention programs where interventions designed to be appealing are introduced by an enthusiastic research team and could lead to a lifting of spirits in the participants, at least in the short-term. Improvement in mood may then be attributed to the content of the program. Although the depression prevention programs described to date seem benign, the process of implementing targeted interventions with the potential for stigmatization may not be, and introducing ineffective programs is a waste of resources. Difficulties in controlling appropriately for psychological interventions are acknowledged, but our study shows that it is possible. Our placebo program fits a number of recommendations made by Shapiro and Shapiro (1997, p. 205) in that it is an active placebo that resembles intervention with an expected effect from having time off regular classes, the group interactive process, and interest from an adult. The placebo group did not receive any sessions on the elements thought to be active in preventing depression, including the overall RAP model. The research team checked integrity of both programs weekly, and our participants were blind. It is also possible that the change in the subjects may have been transmitted to the students in the placebo group, reducing the difference between them. The biggest change in the placebo group came after the end of the program. It is unlikely that the students would have discussed the two programs in such detail that the content of the 11 sessions would have been passed on in any depth. However, a change in cognitive set, conflict resolution, or problem solving in the subjects may have led to changes in the placebo group. If this was so, this is an encouraging notion with the potential that interventions may lead to herd immunity. The initial categorical change with an NNT of 33 indicates that this is potentially an effective public health intervention. One of the factors that contribute to bias in randomized controlled trials is whether or not there is allocation concealment (Schulz et al., 1995). Our method of randomization ensured that the identity of individuals was unknown when students were assigned to groups. We recruited and retained an adequate number of participants according to our power calculations. Overall our participation rates were high in school B, with over 90% retention to 6 months and over 80% to 18 months. The rates were lower in school A and students
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

EFFECTIVENESS OF DEPRESSION PREVENTION

were harder to trace, as families seemed to move house a great deal. However, there was no difference in effectiveness between the two schools, indicating that the program was effective for students from different socioeconomic groups and that the different dropout rates did not change the findings.
Comparison With Other Studies

Limitations

Seven published reports of randomized controlled studies of universal interventions (Clarke et al., 1993; Hains, 1994; Lowry-Webster et al., 2001; Pattison and Lynd-Stevenson, 2001; Quayle and Dziuraweic, 2001; Spence et al., 2003) and one report of a cohort study (Shochet et al., 2001) were identified. Two of these, published in one paper, were of short interventions that were ineffective (Clarke et al., 1993). Two studies, reported in one paper, were on ethnic minority groups, one Latino and one African American (Cardemil et al., 2002) and showed initial reduction in depression scores in the Latino but not the African American group, with no long-term effectiveness. The study by Hains (1994) is hard to interpret, as recruitment procedures made it possible that the study group was highly self-selected. The study of 594 students who participated in a universal anxiety prevention program (Lowry-Webster et al., 2001) was randomized by school, and analysis of data was on individual students. It showed reduction of anxiety but not depression immediately after intervention compared with no intervention. Two studies failed to show effectiveness in depression prevention compared with placebo, but one study was small and thus had low power (Pattison and Lynd-Stevenson, 2001), while in the other the placebo included active therapeutic elements (Shatte, 1997). The large trial by Spence et al. (2003) on the effectiveness of problem solving showed an early reduction in depressive symptoms compared with no intervention that was not sustained over time. The cohort study of the universal program entitled RAP (Shochet et al., 2001) showed a reduction on depressive symptoms after intervention that increased at 10-month follow-up. Our study results are similar to results from the cohort study conducted by Shochet et al. (2001) and some of the larger trials (Cardemil et al., 2002; Spence et al., 2003), but this is the first time a significant difference compared with a placebo program has been shown, and unlike Cardemil and Spence we have some evidence of a persistent effect.
J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

The gold standard for trials of interventions is a prospective double-blind randomized controlled design with a sample size to give adequate statistical power to show change. We randomized and controlled for placebo effect in this study, but the study was singleblind. Although we took care to ensure that teachers conducting the two interventions did not have access to each others materials and were asked not to discuss the two programs, this has the potential to bias the results. Using teachers who have no background of cognitive-behavioral therapy to deliver the program is a potential limitation. Although we trained the teachers, the training was brief, especially compared with standard training in cognitive-behavioral therapy, and focused on the material to be covered in each session as outlined in the manuals. The personal characteristics of the teachers delivering the program may also have affected the results, and we did not address this issue in this study. Although the teachers participated in the development of the program at the outset, they felt they could have delivered the program more effectively if they had been able to adapt material to individual classes. Teaching the teachers basic cognitive-behavioral concepts and providing them with teaching resources that could be adapted may make the program more acceptable, but efficacy using flexible delivery would need to be evaluated. It is important to interpret results in light of the small shifts in score compared with the variability in the data. The standard deviations for the scores were large compared with the means, especially for the BDIII. For example, before intervention the RAP-Kiwi group had a mean BDI-II score of 9.40, with a standard deviation of 7.14. The RADS was more acceptable, with a mean of 55.6 and a standard deviation of 13.17. The ability to show a difference between groups is compromised where there is wide variability in data; this was the case for both measures but was worse for the BDI-II than the RADS. This may have contributed to the statistically significant effects in one measure of depression but not the other. It is clear that a large amount of variation in the scores was not related to the intervention. This is to be expected but highlights a limitation of the study. Depression prevention programs are ideally aimed at reduction of longer-term mood changes seen in depressive disorder, and therefore data on the prevalence of depressive disorder should be collected. Our study would have been
545

MERRY ET AL.

strengthened had we been able to collect prevalence data on depressive disorder following intervention. The study by Clarke et al. (2001) indicates that a relatively small effect seen on depression rating scales can translate into a worthwhile reduction in episodes of depressive disorder, with an effect size of 0.46 associated with NNT of 6that is, six people would need to receive the intervention to prevent one person from developing a depressive episode. The study would also have been strengthened by depression ratings from independent evaluators blind to intervention status, such as parents or teachers or a member of the research team. An assessment of overall function would have allowed a better evaluation of the practical implications of our findings. Unfortunately we did not have sufficient resources to do either of these. A program delivered at school inevitably misses those who have dropped out and is inadequately delivered to those who are absent a great deal. These youths are at risk for many difficulties, and a different process of delivery would be needed for this particularly vulnerable group.
Clinical Implications

We have shown that a universal depression prevention program delivered by teachers is effective in reducing depressive symptoms in the short term, with some evidence that the effect persists to 18 months. Clarification of factors that promote long-term improvement would be worthwhile, and the effect of booster sessions is worth exploring to try to maximize change. It should be confirmed that reduction of depression scores is associated with reduced episodes of depressive disorder and a meaningful impact on broader adjustment, including social adjustment and academic performance. Further study on optimal delivery of depression prevention programs by teachers would be worthwhile.
REFERENCES
Beardslee W, Versage E, Wright E et al. (1997a), Examination of preventive interventions for families with depression: evidence of change. Dev Psychopathol 9:109130 Beardslee W, Wright E, Salt P et al. (1997b), Examination of childrens responses to two preventive intervention strategies over time. J Am Acad Child Adolesc Psychiatry 36:196204

Beck AT, Steer RA, Brown GK (1996), BDI-II Beck Depression Inventory, 2nd ed. San Antonio: The Psychological Corporation, Harcourt Brace & Company Cardemil EV, Reivich KJ, Seligman M (2002), The prevention of depressive symptoms in low-income minority middle-school students. Prevention Treatment 5 Clarke GM, Hawkins W, Murphy N, Sheeber L (1993), School-based primary prevention of depressive symptomatology in adolescents: findings from two studies. J Adolesc Res 8:183204 Clarke GN, Hawkins W, Murphy MS, Sheeber LB, Lewinshohn PM, Seeley JR (1995), Targeted prevention of unipolar depressive disorder in an at-risk sample of high school adolescents: a randomized trial of a group cognitive intervention. J Am Acad Child Adolesc Psychiatry 34:312321 Clarke GN, Hornbrook M, Lynch F et al. (2001), A randomized trial of a group cognitive intervention for preventing depression in adolescent offspring of depressed parents. Arch Gen Psychiatry 58:11271134 Fergusson DM, Horwood LJ (2001), The Christchurch Health and Development Study: review of findings on child and adolescent mental health. Aust N Z J Psychiatry 35:287296 Gillham J, Reivich K (1999), Prevention of depressive symptoms in school children: a research update. Psychol Sci 10:461462 Gwynn C, Brantley H (1987), Effects of a divorce group intervention for elementary school children. Psychology in the Schools 24:161164 Hains AA (1994), The effectiveness of a school-based, cognitive-behavioral stress management program with adolescents reporting high and low levels of emotional arousal. School Counselor 42:114125 Hankin B, Abramson L, Moffitt T, Silva P, McGee R, Angell K (1998), Development of depression from preadolescence to young adulthood: emerging gender differences in a 10-year longitudinal study. J Abnorm Psychol 107:128140 Ihaka R, Gentleman R (1996), A language for data analysis and graphics. J Comput Graphic Stat 5:299314 Jaycox LH, Reivich KJ, Gillham J, Seligman M (1994), Preventing depressive symptoms in school children. Beh Res Ther 32:801816 Lamb JM, Puskar KR, Sereika SM, Corcoran M (1998), School-based intervention to promote coping in rural teens. Am J Maternity Child Nurs 23:187194 Lowry-Webster H, Barrett P, Dadds M (2001), A universal prevention trial of anxiety and depressive symptomatology in childhood: preliminary data from a Australian study. Behavior Change 18:3650 Mrazek PJ, Haggerty RJ (1994), Reducing Risks for Mental Disorders. Frontiers for Preventive Intervention Research, 1st ed. Washington, DC: National Academy Press Murray C, Lopez A (1996), The Global Burden of Disease. A Comprehensive Assessment of Mortality and Disability from Diseases, Injuries, and Risk Factors in 1990 and Projected to 2020. Harvard School of Public Health NHMRC (1997), Depression in Young People: Clinical Practice Guidelines. Australian Government Publishing Service Pattison C, Lynd-Stevenson R (2001), The prevention of depressive symptoms in children: the immediate and long-term outcomes of a schoolbased program. Behavior Change 18:92102 Petersen A, Leffert A, Graham B, Alwin J, Ding S (1997), Promoting mental health during the transition into adolescence. In: Health Risks and Developmental Transitions during Adolescence, Hierrelmann AK, ed. New York: Cambridge University Press, pp 471497 Quayle D, Dziuraweic S (2001), The effect of an optimism and lifeskills program on depressive symptoms in preadolescence. Behavior Change 18:194203 Reynolds WM (1986), Adolescent Depression Scale-High School Version. Odessa, FL: Psychological Assessments Resources Reynolds WM, Mazza JJ (1998), Reliability and validity of the Reynolds Adolescent Depression Scale with young adolescents. J School Psychol 36:295312 Rose G (1992), The Strategy of Preventive Medicine, 1st ed. Oxford: Oxford University Press Sackett DL, Richardson WS, Rosenberg W, Haynes RB (1997), EvidenceBased Medicine. How to Practice and Teach EBM, 1st ed. New York: Churchill Livingstone

546

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

EFFECTIVENESS OF DEPRESSION PREVENTION

Schulz KF, Chalmers I, Hayes RJ, Altman D (1995), Empirical evidence of bias. JAMA 273:408412 Seligman ME, Schulman P, DeRubies RJ, Hollon SD (1999), The prevention of depression and anxiety. Prevention Treatment 2 Shapiro AK, Shapiro E (1997), The Powerful Placebo: From Ancient Priest to Modern Physician, 1st ed. Baltimore: John Hopkins University Press Shatte AJ (1997), Prevention of depressive symptoms in adolescents: issues of dissemination and mechanisms of change. Dissertation Abstracts International, Section B: The Sciences and Engineering 57 Shochet IM, Dadds MR, Holland D, Whitefield K, Harnett PH, Osbarby

SM (2001), The efficacy of a universal school-based program to prevent adolescent depression. J Clin Child Psychol 30:303315 Spence S, Sheffield J, Donovan C (2003), Preventing adolescent depression: an evaluation of the Problem Solving for Life program. J Consult Clin Psychol 71:313 Thompson E, Eggert L, Herting J (2000), Mediating effects of an indicated prevention program for reducing youth depression and suicide risk behaviors. Suicide Life-Threat Beh 30:252271 Yu DL, Seligman M (2002), Preventing depressive symptoms in Chinese children. Prevention and Treatment Volume 5. Article 9, posted May 8, 2002

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 43:5, MAY 2004

547