Mindfulness

, Volume 5, Issue 5, pp 477–486

School-Based Prevention and Reduction of Depression in Adolescents: a Cluster-Randomized Controlled Trial of a Mindfulness Group Program

Authors

    • Faculty of Psychology and Educational SciencesUniversity of Leuven
  • James W. Griffith
    • Department of Medical Social SciencesNorthwestern University
  • Katleen Van der Gucht
    • Faculty of Psychology and Educational SciencesUniversity of Leuven
  • J. Mark G. Williams
    • Department of Psychiatry, Warneford HospitalUniversity of Oxford
ORIGINAL PAPER

DOI: 10.1007/s12671-013-0202-1

Cite this article as:
Raes, F., Griffith, J.W., Van der Gucht, K. et al. Mindfulness (2014) 5: 477. doi:10.1007/s12671-013-0202-1

Abstract

Our objective was to conduct the first randomized controlled trial of the efficacy of a group mindfulness program aimed at reducing and preventing depression in an adolescent school-based population. For each of 12 pairs of parallel classes with students (age range 13–20) from five schools (N = 408), one class was randomly assigned to the mindfulness condition and one class to the control condition. Students in the mindfulness group completed depression assessments (the Depression Anxiety Stress Scales) prior to and immediately following the intervention and 6 months after the intervention. Control students completed the questionnaire at the same times as those in the mindfulness group. Hierarchical linear modeling showed that the mindfulness intervention showed significantly greater reductions (and greater clinically significant change) in depression compared with the control group at the 6-month follow-up. Cohen's d was medium sized (>.30) for both the pre-to-post and pre-to-follow-up effect for depressive symptoms in the mindfulness condition. The findings suggest that school-based mindfulness programs can help to reduce and prevent depression in adolescents.

Keywords

Mindfulness-based cognitive therapyMindfulness-based stress reductionRandomized controlled trialDepressionAdolescents

Introduction

Mindfulness refers to a compassionate and nonjudgmental moment-to-moment awareness of one's experiences. An often-cited definition is that of Kabat-Zinn (1994, p. 4), who describes mindfulness as the awareness that emerges through “paying attention in a particular way: on purpose, in the present moment, and nonjudgmentally.” In the last three decades, interventions have been developed to teach mindfulness skills to reduce physical and emotional complaints. Two predominant mindfulness-based approaches, both using mindfulness meditation, are Mindfulness-Based Stress Reduction (MBSR; Kabat-Zinn 1990) and Mindfulness-Based Cognitive Therapy (MBCT; Segal et al. 2002). Other influential approaches within the family of “mindfulness-oriented interventions” (Keng et al. 2011, p. 1043), such as Dialectical Behavior Therapy (Linaehan 1993) and Acceptance and Commitment Therapy (Hayes et al. 1999), also teach mindfulness but with less focus on meditation. The success of these mindfulness-based approaches suggests that mindfulness skills may help to prevent and reduce emotional distress.

In both clinical and non-clinical samples, mindfulness-based interventions have generally been effective at reducing physical and psychological problems (e.g., anxiety and depression) and increasing health and well-being (for reviews, see: Hofmann et al. 2010; Keng et al. 2011). Most studies that provide support for mindfulness-based interventions have been conducted in adults. Recently, however, there have been more efforts to develop and to examine the effectiveness of mindfulness-based interventions for children and adolescents (Burke 2010; Meiklejohn et al. 2012).

Burke (2010) concludes that there are clear indications now that mindfulness-based programs with children and youth are feasible and acceptable (e.g., Mendelson et al. 2010). There is also emerging evidence on the efficacy and effectiveness of such interventions for children and adolescents, both in clinical and in normative or community samples (for reviews, see Black et al. 2009; Burke 2010; Meiklejohn et al. 2012).

The results of a randomized controlled trial (RCT) by Biegel et al. (2009), for example, reported positive effects of MBSR in an adolescent outpatient psychiatric sample, in terms of decreases in anxiety and depression and increases in self-esteem. Other promising findings on the benefits of mindfulness-based approaches for adolescents and children with clinical problems are reported, for example, for ADHD (e.g., Singh et al. 2010; Van de Weijer-Bergsma et al. 2012) and anxiety (Semple et al. 2005).

Preliminary studies examining mindfulness-based approaches in community children and adolescents suggest that such interventions might also be beneficial for non-clinical individuals for reducing stress, depression, and anxiety, as well as helping to improve attention and emotion regulation (see Meiklejohn et al. 2012, for a review). Positive effects have been reported for school-based mindfulness approaches for elementary- and middle school-aged children (e.g., Napoli et al. 2005; Schonert-Reichl and Lawlor 2010; Van de Weijer-Bergsma et al. 2013) as well as for high-school-aged adolescents (e.g., Broderick and Metz 2009; Huppert and Johnson 2010).

Although the findings to date on mindfulness interventions for children and adolescents are promising, further empirical work is clearly needed. This is particularly so for community samples because existing studies are typically limited either by small sample sizes, lack of adequate control groups, or nonrandomized designs (Burke 2010; also see: Keng et al. 2011; Meiklejohn et al. 2012). Thus, there is a need to advance research on mindfulness for youth using larger scale group RCTs.

This article reports the effects of a mindfulness group program on reducing and preventing depression in adolescents, using a group RCT design with a large school-based sample. The mindfulness program was developed specifically for adolescents (Dewulf 2009, 2013) and integrates components of MBCT (Segal et al. 2002) and MBSR (Kabat-Zinn 1990).

Method

Participants and Design

Fifty schools were invited to participate. The schools were located in Flanders—the northern, Dutch-speaking region of Belgium. Fifteen schools expressed interest in participating, but ten were unable to participate due to practical difficulties that mainly had to do with course timetable incompatibilities. Four of the five schools each offered two or three pairs of parallel classes, spread over two school years (2009–2010; 2010–2011), resulting in nine pairs of classes (n = 315). Within pairs, one class was randomized to the mindfulness condition, the other to the (no intervention) control condition. In the fifth school, classes were too small to be used as mindfulness or control groups. For that particular school, half of the students of each class were randomized to the mindfulness condition, the other half to the control condition, which led to an additional three pairs of mindfulness and control groups (n = 93). This randomization resulted in a total of 12 pairs of mindfulness and 12 control groups (N = 408; mindfulness n = 201; control n = 207). Randomization, using an online random number generator, was done by the first author who did not participate in the assessments and who had no contact with schools, classes, or students. Participating classes were from years 3 to 6 in secondary school, a time that is characterized by the highest percentage of age at onset of first major depressive episode (Zisook et al. 2007). Years 3 to 6 in Flemish secondary schools, roughly referring to ages 14 to 17, correspond to grades 9 to 12 (or high school) in the American educational system. To be more precise, there were three pairs of groups from year 3 (mean age 14), four pairs of groups from year 4 (mean age 15), four pairs of groups of year 5 (mean age 16), and one pair of groups of year 6 (mean age 17). Group size ranged from 10 to 24 (M = 16.8; SD = 4.9) for the mindfulness condition; for the control groups, group size ranged from 12 to 24 (M = 17.3; SD = 4.8), t < 1.

Measures

Depression Anxiety Stress Scales (DASS-21)

The DASS-21 (Lovibond and Lovibond 1995) consists of three 7-item scales designed to assess depression (DASS-21-D), anxiety (DASS-21-A), and stress symptoms (DASS-21-S). In this study, the depression scale was the main outcome. Items were scored on a four-point scale, ranging from 1 (did not apply to me at all) to 4 (applied to me very much or most of the time), for the past week. Good psychometric properties are reported for the original as well as the Dutch version that we used (Beurs et al. 2001; Willemsen et al. 2011). To aid in questionnaire interpretation, all subscales of the DASS were converted to percent-of-maximum-possible (POMP) scores, in which a scale is converted to a 0–100 % metric with 0 being the minimum value of the scale and 100 being the maximum possible value of the scale (Cohen et al. 1999). For the depression scale, a raw score of 14 or above (POMP score of 33.3) was considered clinically significant. Reliability for this scale was good in this study; Cronbach's alphas were >.80 across all three time points.

Other Measures

We included several measures in this study that are not reported in this manuscript. For example, we had initially wanted to use the 15-item Five-Factor Mindfulness Questionnaire (Baer et al. 2012) as a mindfulness measure, but the internal consistency for this measure was too low to be trusted (Cronbach's α = .47). We suspect that this was because this measure was not developed for secondary school students. Thus, we do not report this measure. We also included a Mood Disorders Questionnaire (Van der Does et al. 2003), which is an experimental questionnaire that is intended to yield a yes/no diagnosis of major depressive disorder. However, the sensitivity and specificity of this measure are not well known at this time in English or Dutch. Thus, we did not use this questionnaire as an endpoint.

Procedure

All students gave written assent or informed consent if they were already legal adults. For those below age 16, passive parental informed consent was also obtained using a letter in which parents were asked to complete and return an attached form if they did not want their child to participate. Students in the mindfulness condition attended the mindfulness program during school hours for 8 weeks, replacing religious studies, physical education, or another academic course, depending on the class's timetable. One school was a boarding school, so the mindfulness sessions were organized on a weekday evening. Each mindfulness session lasted 100 min. Students were not graded on any aspect of the mindfulness course. Participants in the control groups followed their regular school program; no intervention or attention was provided to them. Mindfulness and control groups of each pair were conducted at the same time. Students in the mindfulness groups completed the questionnaire before the mindfulness program (baseline; T1), after the mindfulness program (post-intervention; T2), and at a 6-month follow-up point (T3). Control students completed the questionnaire at the same times as those in the mindfulness group. All assessments were administered during regular school hours by the third author who was not involved as a mindfulness trainer in any of the groups. Students were provided with help-seeking contacts (e.g., helpline numbers, mental health centers) at each assessment point. The ethical committee of the University of Leuven approved the study.

Mindfulness Intervention

The intervention was a mindfulness group training specifically developed for adolescents (Dewulf 2009, 2013). It integrated elements of MBCT (Segal et al. 2002) and MBSR (e.g., Kabat-Zinn 1990). The program was delivered by an instructor in eight weekly 100-min sessions, which include guided experiential mindfulness exercises (e.g., mindfulness of breathing, breathing space, body scan), sharing of experience of these exercises; reflections in small groups, inspiring stories; psycho-education (e.g., stress, depression, self-care), and review of homework. Homework assignments were 15 min of mindfulness practice each day, suggested reading, and weekly tips on how to bring mindfulness into daily life. Each session focused on a specific theme, and some exercises are repeated throughout the program. Sessions thematically focus on “attention to the breath and the moment” (session 1), “attention to the body and pleasant moments” (session 2), “attention to your inner boundaries and to unpleasant moments” (session 3), “attention to stress and space” (session 4), “attention to thoughts and emotion” (session 5), “attention to interpretations and communication” (session 6), “attention to your attitudes and your moods” (session 7), and “attention to yourself and your heartfulness” (session 8). Participants in this program received the book Mindfulness voor jongeren [Mindfulness for adolescents] (Dewulf 2009) for reviewing the material at home. For formal practice, the book comes with a double CD with mindfulness exercises and several sitting meditations. Participants were also encouraged to apply mindfulness throughout their daily lives. They received a workbook for making notes on their home practice; these notes are then used as input for discussion during the next group session.

The three instructors, two men and one woman, were experienced mindfulness trainers; two of them were psychologists; one was a medical doctor. One of them, D. Dewulf, developed the mindfulness intervention and is the founder and chairman of the Institute for Attention and Mindfulness (I AM). The other two completed their mindfulness training at I AM. They had extensive experience with delivering the program to adults and adolescents and also have an ongoing personal mindfulness meditation practice. The three instructors regularly met eight times during the course of the first mindfulness groups to discuss their experience with the sessions and to evaluate and to maximize or ensure close adherence to the protocol, but no formal adherence measure was used.

Statistical Analysis

For each of the three DASS subscales, if fewer than 20 % of the items were missing (i.e., not more than one item missing), the score was prorated. Ten participants did not provide sufficient data on the DASS-21-D to compute baseline scores. Those participants who did not complete the pre-intervention assessment were excluded from analyses. Our analysis plan was as follows: We first compared the mindfulness versus control groups to ensure that they were comparable at baseline; we predicted that there would be no meaningful differences. We next used hierarchical linear modeling (e.g., Raudenbush and Bryk 2002) to examine the differential trajectories of depression. Hierarchical linear modeling allowed us to determine individual trajectories in individual change while accounting for the fact that individuals within groups might show dependencies by virtue of the fact that they are in the same group. Cases with missing data at T2 or T3 were still included in the analyses, which still allows for estimation of change over time with full maximum likelihood estimation. We hypothesized that mindfulness would result in larger reductions in depression at T2 and T3 above and beyond gender and school as covariates, relative to the control condition. We also used Cohen's d statistic (Cohen 1988) to calculate within-group effect sizes. In addition, we examined the clinical significance of the effects using a normative cutoff point for our primary outcome measure (DASS-21-D).

Results

Participant Flow

Figure 1 presents the flowchart of the recruitment and retention of participants in the trial. None of the students from the classes involved in the project declined to participate.
https://static-content.springer.com/image/art%3A10.1007%2Fs12671-013-0202-1/MediaObjects/12671_2013_202_Fig1_HTML.gif
Fig. 1

Flowchart of the recruitment and retention of participants in the trial. Note: For four schools, classes were randomized (a). For one school, because of classes being too small, individuals were randomized (for more details, see “Participants and Design” section)

Four participants had missing data at the T1 in the mindfulness condition versus one in the control group. Table 1 presents the percentage of missing data across the follow-ups. The main reason for missing data was participants being absent from school on the day of testing. No students formally withdrew from the study. There were significantly higher rates of missing data for the control group versus the mindfulness group at T2 (12 versus 6 %; Table 1), but rates of missing data were not significantly different at T3 (16 versus 14 %; Table 1). The average age in the sample was 15.4 years (SD = 1.2, range 13–20). In terms of year in school, 20 % were in their third year of secondary school (typically age 14), 40 % were in their fourth year, 29 % were in their fifth year, and 10 % were in their sixth year.
Table 1

Baseline and outcome characteristics (N = 393)

Baseline characteristics

Mindfulness

n

 

Control

n

 

Gender

131 females/63 males; 68 %/33 %

194

 

119 females/80 males; 60 %/40 %

199

 
 

M (SD) of POMP scores

N

 

M (SD) of POMP scores

n

 

T1 DASS-21-D

19.4 (16.6)

194

 

21.2 (19.2)

199

 

T1 DASS-21-A

22.2 (17.5)

194

 

22.2 (16.3)

199

 

T1 DASS-21-S

34.9 (18.5)

194

 

34.9 (19.1)

199

 

Outcome data

M (SD) of POMP scores

N

Missing data

M (SD) of POMP scores

n

Missing data

T2 DASS-21-D

14.1 (14.6)

182

12 (6 %)*

22.2 (21.4)

175

24 (12 %)*

T3 DASS-21-D

12.9 (17.3)

167

27 (14 %)

21.0 (21.3)

168

31 (16 %)

Note: In some cases, percentages do not sum to 100 % exactly due to rounding error. POMP scores are percent-of-maximum-possible scores; each DASS subscale was rescaled to a 0–100 % metric with the minimum of the scale at 0 % and the maximum at 100 %. The percentages of data lost (i.e., missing data) are presented above next to each follow-up time point

DASS Depression Anxiety Stress Scales (DASS-21) for depression (D), anxiety (A), and stress (S)

*p < .05 for the differences between the mindfulness and control groups: At T2, rates of missing data were different on the DASS-21-D

Pre-intervention Comparisons

Table 1 shows comparisons of the mindfulness versus control groups at baseline. Clinical significance of depression was assessed using the DASS-21-D scores of moderate depression or greater (Lovibond and Lovibond 1995). Rates of clinically significant depression are shown in Table 2. No significant differences were observed between the two groups at baseline.
Table 2

Number and percentage of students scoring above the clinical cutoff for depression symptoms (as assessed with the DASS-21-D) at baseline, post-intervention, and follow-up

Condition

Pre (T1)

Post (T2)

FU (T3)

n

%

n

%

n

%

All

Control

47

24

47

27*

52

31*

Mindfulness

41

21

28

15

26

16

Clinical cases at T1 only

Control

47

 

22

55

21

54*

Mindfulness

41

 

18

51

9

27

New clinical cases

Control

(0)

 

25

19*

31

24*

Mindfulness

(0)

 

10

7

17

13

*p < .05 for difference between mindfulness versus control group

Intervention Outcome

Hierarchical linear modeling was used to conduct the analyses of the DASS-21-D. The model is presented in Table 3 and the overall results are presented in Fig. 2. We first examined a “null model,” in HLM parlance, to estimate intraclass correlations (ICCs) for level 2 and level 3 variabilities. These ICCs represent the proportion of variance in the dependent variable (DASS-21-D) that is accounted for by level 2 (subject level, e.g., gender) and level 3 (e.g., school) factors. The ICCs for levels 2 and 3 were .35 and .08, respectively.
Table 3

Hierarchical linear model of DASS-21-D

Level 1 model

Y = P0 + P1(T2) + P2(T3) + E

Level 2 model

P0 = B00 + B01(GENDER)

P1 = B10 + B11(GENDER) + R1

P2 = B20 + B21(GENDER) + R2

Level 3 model

B00 = G000 + G001(CONDITION) + G002(X1) + G003(X2) + G004(X3) + G005(X4) + U00

B01 = G010 + G011(CONDITION) + G012(X1) + G013(X2) + G014(X3) + G015(X4)

B10 = G100 + G101(CONDITION) + G102(X1) + G103(X2) + G104(X3) + G105(X4)

B11 = G110 + G111(CONDITION) + G112(X1) + G113(X2) + G114(X3) + G115(X4)

B20 = G200 + G201(CONDITION) + G202(X1) + G203(X2) + G204(X3) + G205(X4)

B21 = G210 + G211(CONDITION) + G212(X1) + G213(X2) + G214(X3) + G215(X4)

Note. E, R1, R2, and U00 are error terms. Gender was dummy coded (female = 0, male = 1). T2 and T3 are dummy variables for the differences from baseline. X1–X4 are dummy variables for the five different schools. Condition is mindfulness versus control, dummy coded (mindfulness = 1, control = 0). G101 and G201 are the main effects of condition on T2 and T3, respectively. G111 and G211 are the interaction effects of condition and gender on change from baseline to T2 and T3, respectively. Group was included as the identifier variable at level 3, so group-to-group variability is captured as error at that level. The notation used is that of the HLM software

https://static-content.springer.com/image/art%3A10.1007%2Fs12671-013-0202-1/MediaObjects/12671_2013_202_Fig2_HTML.gif
Fig. 2

Depression Anxiety Stress Scales (DASS)-21 Depression Scale over T1 (baseline), T2 (posttreatment), and T3 (follow-up). The DASS is in percent-of-maximum-possible (POMP) scores, which range from 0 to 100 %. Only complete cases, with non-missing data at T1, T2, and T3, are included in the figure. Error bars indicate SE

Level 1 of the model contained the within-subject relationship between time point and the DASS-21-D. Time point was dummy coded such that one dummy variable measured the change from baseline to T2 (post-intervention), and another variable measured the change from baseline to T3 (see Table 3). Level 2 of the model contained the regression of the level 1 parameters on gender, which was the only participant-level variable that we examined. At level 2, error terms were included for the two level 1 slopes as dependent variables to capture participant-to-participant differences that might exist for the relationship between time point and depression. At level 3, group was used as the identifier to capture differences across groups; experimental condition and school were used as covariates. School was represented by four dummy variables. At level 3, the dependent variables were the level 2 regression parameters. Gender and school were not our main variables of interest. Gender is often related to depression, so it was entered to reduce potential nuisance variance. School was entered to represent the nested structure of the data.

Table 3 presents the hierarchical model that we tested. Table 4 presents the results for each parameter in the model. To test our main hypotheses, we used DASS-21-D as the dependent variable in the regression; the coefficients and standard errors for parameters G101 and G201 (see Table 3) were, respectively, −6.0 (3.1) and −6.8 (3.3). The p value for parameter G101 was .050, but the observed t statistic (t = 1.958, 391 df) was very slightly below the critical value needed for statistical significance. Parameter G201, however, was significant at p < .05, which indicates that, as predicted, the mindfulness intervention had more of an impact on changes in depression than the control group at T3. The mindfulness group had lower levels of depression at T2 and T3 relative to baseline, whereas little change was observed in the control group (see Table 1). Again, the unit of measurement for these coefficients is POMP scores. The effect of the mindfulness intervention did not significantly interact with gender. Another way to think about G101 and G201 are as cross-level interactions of condition and time—these parameters capture the effect of treatment condition on the changes from baseline in depression. Although they were not the main focus of the study, the G201 parameters were also negative and significant at p < .05 when the anxiety scale and stress scale were used as dependent variables. The results of the depression scale are also graphically depicted in Fig. 2; note that only complete cases are included in the figure, so the values will not match the tables exactly.
Table 4

Final estimation of fixed effects: with DASS-21-D as the dependent variable

Fixed EFFECT

 

Coefficient

SE

t(df)

Description of effect

For P0

 

For B00

 

G000

24.0

3.0

7.9 (18)***

Intercept

G001

0.4

2.8

0.1 (18)

Cond. on baseline DASS

G002

−3.3

4.5

0.7 (18)

Schoola on baseline DASS

G003

6.0

4.4

1.3 (18)

Schoola on baseline DASS

G004

−5.8

4.3

1.4 (18)

Schoola on baseline DASS

G005

−5.0

4.1

1.2 (18)

Schoola on baseline DASS

For B01

 

G010

−5.4

5.9

0.9 (1,049)

Gender on baseline DASS

G011

−4.5

3.5

1.3 (1,049)

Cond. × gender on baseline DASS

G012

6.9

7.2

1.0 (1,049)

Schoola × gender on baseline DASS

G013

0.7

6.6

0.1 (1,049)

Schoola × gender on baseline DASS

G014

5.7

6.8

0.8 (1,049)

Schoola × gender on baseline DASS

G015

0.7

6.9

0.1 (1,049)

Schoola × gender on baseline DASS

For P1

 

For B10

 

G100

−1.1

3.3

0.3 (391)

Change from baseline to T2

G101

−6.0

3.1

2.0 (391)

Cond. on T2 change in DASS

G102

3.6

5.0

0.7 (391)

Schoola on T2 change in DASS

G103

−6.6

5.3

1.2 (391)

Schoola on T2 change in DASS

G104

9.0

4.4

2.1 (391)*

Schoola on T2 change in DASS

G105

0.7

4.2

0.2 (391)

Schoola on T2 change in DASS

For B11

 

G110

−1.8

8.9

0.2 (391)

Gender on T2 change in DASS

G111

0.6

5.3

0.1 (391)

Cond. × gender on T2 change in DASS

G112

1.2

11.0

0.1 (391)

Schoola × gender on T2 change in DASS

G113

8.3

10.1

0.8 (391)

Schoola × gender on T2 change in DASS

G114

−4.4

10.0

0.4 (391)

Schoola × gender on T2 change in DASS

G115

7.3

10.3

0.7 (391)

Schoola × gender on T2 change in DASS

For P2

 

For B20

 

G200

−2.9

3.5

0.8 (391)

Change from baseline to T2

G201

−6.8

3.3

2.1 (391)*

Cond. on T3 change in DASS

G202

2.1

5.3

0.4 (391)

Schoola on T3 change in DASS

G203

−5.2

5.8

0.9 (391)

Schoola on T3 change in DASS

G204

7.9

4.7

1.7 (391)

Schoola on T3 change in DASS

G205

0.2

4.5

0.1 (391)

Schoola on T3 change in DASS

For B21

 

G210

26.5

10.8

2.5 (391)*

Gender on T3 change in DASS

G211

3.0

5.7

0.5 (391)

Cond. × gender on T3 change in DASS

G212

−32.3

12.9

2.5 (391)*

Schoola × gender on T3 change in DASS

G213

−18.5

12.1

1.5 (391)

Schoola × gender on T3 change in DASS

G214

−22.1

11.9

1.9 (391)

Schoola × gender on T3 change in DASS

G215

−22.4

12.1

1.8 (391)

Schoola × gender on T3 change in DASS

Note: t statistics are absolute values. Degrees of freedom (df) are approximate. Parameters that were hypothesized to be significant and negative are in bold; these parameters, G101 and G201, are the effects of experimental condition on change from baseline at T2 and T3, respectively

Cond. experimental condition (mindfulness versus control)

*p < .05, ***p < .001

aThe five schools are coded with four dummy variables

Because gender and school were not hypothesized to influence the efficacy of treatment, we also compared the full model in Tables 3 and 4 to a simplified model that did not contain the parameters for gender and school. The deviance statistic for the simplified model (11 parameters) was 9,330.0; for the full model (Tables 3 and 4; 41 parameters), the deviance statistic was 9,288.5. Despite our large sample, the difference between these two models was not statistically significant (χ2 (df = 30) = 41.5, ns). This suggests that these additional variables did not have a significant effect as a group. Among the other 34 t tests produced by the model (not counting our two hypothesized effects), there were, however, four individual t tests that were significant as shown in Table 4. Because the block of variables was not significant, and because it is not surprising to see some significant tests when testing a large block, these were not interpreted further and no additional modifications were made to the model.

To further examine the magnitude of the effect of the mindfulness intervention, we used repeated measures t tests as well as Cohen's d effect sizes (Cohen 1988). We compared the differences from T2 and T3, respectively, from baseline. In the mindfulness group, for the reduction from baseline to T2 in DASS-21-D, there was a significant decrease in depression (t(181) = 4.4, p < .001, Cohen's d was .32). For the reduction from baseline to T3, there was also a significant reduction (t(166) = 4.0, p < .001, Cohen's d was .31). In terms of Cohen's (1988) conventions, these are small- to medium-sized reductions. No significant changes over time were observed in the control condition (ts < 1.2, ns). And whereas the mindfulness group and control group did not differ in terms of levels of depression at baseline (t(391) = 1.0, p = .32), the mindfulness group had lower levels at T2 and T3 than the control condition (t(363) = 4.1 and p < .001 and t(342) = 3.9 and p < .001).

Clinical Significance: Clinical Cutoff Depression Symptoms (DASS-21-D)

Table 2 shows the number and percentage of students scoring above the clinical cutoff on the DASS-21-D at baseline (T1), post-intervention (T2), and at follow-up (T3). At T2 and at T3, a significantly smaller percentage of students in the mindfulness group scored above the clinical cutoff, compared to the control condition.

Focusing only on participants (n = 88) who scored above the clinical cutoff at baseline, one sees that at T3, in the control condition, a significantly larger percentage of these students were still scoring above the cutoff, as compared to the mindfulness condition. In the control condition, 46 % had recovered versus 73 % in the mindfulness condition. We also examined whether these changes exceeded the reliable change index (RCI; Jacobson and Truax 1991) and whether participants moved from above to below the cutoff. We calculated the RCI using the formula from Jacobson and Truax (1991, p. 14); we used the SD from the DASS-21-D at T1 in the calculations; we used .80, which is Cronbach's α from the T1 DASS-21-D, as the reliability estimate. The RCI was 4.7 in raw units and 22.2 in POMP units. Of the control condition, 36 % were reliably improved at follow-up versus 62 % of the mindfulness condition (χ2(1) = 4.87, p < .05).

For those starting below the DASS-21-D cutoff (n = 305), a larger percentage of control participants were above the clinical cutoff at both T2 (χ2(1) = 8.89, p < .01) and at T3 (χ2(1) = 5.54, p < .05). Looking at both an increase that exceeded the RCI and moving from below to above the cutoff, 13 % of the controls had more depression at T2 versus 1 % of the mindfulness condition (χ2(1) = 16.71, p < .05) from baseline to follow-up, and 18 % of the control condition had more depression at T3 versus 9 % in the mindfulness condition (χ2(1) = 4.39, p < .05).

Discussion

These results show that a mindfulness program, as compared with a control condition, was able to result in lower levels of depression at a 6-month follow-up in adolescents. Moreover, these results appear to be clinically significant when examining the effect size from the DASS-21-D. The effects for depressive symptoms from baseline to post-intervention and from baseline to follow-up were small to medium (both Cohen's d were >.30). These values correspond to an approximate r value of .15, which is similar to the average pre-to-post (r = .15) and average pre-to-follow-up effect (r = .11) reported for depression prevention programs for youth in a recent meta-analytic review (Stice et al. 2009).

Further, in the mindfulness condition, a significantly smaller percentage of students scored above the clinical cutoff at follow-up than in the control condition, reflecting a combination of a curative and preventive effect of the intervention: A greater proportion of depressed students (i.e., scoring above the clinical cutoff at baseline) were reliably recovered at follow-up (i.e., scoring now below the clinical cutoff) in the mindfulness condition; and a smaller proportion of non-depressed students (i.e., scoring below the clinical cutoff at baseline) reliably deteriorated at follow-up (i.e., scoring now above the clinical cutoff). Our results, thus, extend earlier findings on the positive effects of mindfulness-based interventions for depressive symptoms in clinical samples (see Hofmann et al. 2010) to a community sample (of adolescents). Also, our findings add to the growing body of research that mindfulness-based programs can be successfully integrated in education (see Meiklejohn et al. 2012).

There are several limitations and caveats to note about this study. There was a disproportionately high percentage of females versus males in this study (see Table 1). Although we did not find a significant gender × condition interaction (see parameters G111 and G211 in Table 4), we may have lacked sufficient power to detect such effects. Future studies should aggressively recruit large subsamples from both genders (and potentially other subgroups) to determine definitively whether treatment efficacy interacts with other participant characteristics.

In terms of methodological limitations, we did not assess the fidelity of the intervention by recording (e.g., videotaping) the sessions and having the sessions coded. We did not measure any psychotherapy process variables (e.g., how much homework was completed, what specific topics were discussed), and our control group did not receive any active ingredient of psychotherapy (e.g., attention, supportive listening). Thus, the mechanism by which the mindfulness intervention was effective is unclear.

With regard to our assessments, our measure of depression was a self-report questionnaire. A superior methodology would have been to use a psychometrically strong, clinician-administered interview. Unfortunately, in this study, it was not logistically possible to administer an in-depth clinical interview to our sample of over 400 participants. Future studies should seek to replicate our findings using interview-based assessments. Finally, we stress once more that we only had a no-treatment control, so we cannot rule out the effect of nonspecific aspects of the mindfulness intervention (e.g., attention). Thus, future studies should include active control groups as much as possible. Assuming that future studies show that mindfulness can be effective for preventing depression, other studies could then examine the mechanisms of change so that these future interventions could target these mechanisms. Potential mechanisms or key processes in this respect might be, for example, cognitive reactivity (e.g., Kuyken et al. 2010; Raes et al. 2009), self-compassion (Kuyken et al. 2010), and experiential acceptance (Keng et al. 2011).

In summary, the current study, to our knowledge, is the first group RCT to examine the effects on depression of a mindfulness-based intervention for adolescents in a community setting. Previous research had found that mindfulness interventions could be adapted to younger populations, but evidence for the efficacy was limited by weak study designs (e.g., low sample sizes, no randomization). Our results provide evidence in support of the efficacy of a mindfulness-based approach to reduce depression symptoms in adolescents. This effect was statistically and clinically significant and appears to reflect a combination of a curative and preventive effect.

Acknowledgments

This research was supported by a grant from the Foundation “Go for Happiness.” Dr. Griffith was supported by the Research Foundation-Flanders (FWO; GP.035.11N). We sincerely thank David Dewulf, Inge De Leeuw, and Lieven Vercauteren, and the schools and students of Belgium who participated in this project. Participating schools were: Sint-Maarten Bovenschool (Beveren-Waas), Burgerschool (Roeselare), Abdijschool van Zevenkerken (Brugge), Groenendaalcollege (Campus Sint-Jan; Merksem-Antwerpen), and Sint-Ludgardisschool (Antwerpen Stad).We would also like to acknowledge the assistance provided by Tineke Vandenbroucke and Margot Bastin. And special thank goes to Jan Toye and Ghita Kleijkers for their sincere and motivating interest.

Copyright information

© Springer Science+Business Media New York 2013