Study population
In the process of choosing the population, we opted for two different schools in terms of academic performance and urbanicity, to obtain an adequate sample that could be considered relatively representative of Swedish upper secondary school students. This selection is more thoroughly described in the baseline cross-sectional study [
11], which represents a larger randomised, controlled, single-blinded study (CHAMPS) aimed to investigate whether iMBI could be used to prevent psychiatric and stress-related symptoms.
All Swedish-speaking students enrolled at the schools (adolescents aged 15–19 years) received information, initially through personal letters sent to their home addresses, and thereafter lectures in the schools (10-min oral information from the researchers in groups of 10–100 students with the possibility to ask questions).
We had three inclusion criteria: to be able to read and understand Swedish, to have an e-mail address, and to be willing to participate in an 8-week Internet-based self-help programme. A total of 1404 students were attending the two schools during the study period. The Swedish-speaking inclusion criterion excluded one subject, leaving 1403 possible participants. The second inclusion criterion did not lead to exclusion of any potential participants and the third criterion, willingness to participate in a stress reduction intervention, left 283 individuals all of whom gave written informed consent. The 283 participants were numbered consecutively after the arrival of their written consent form and were divided into seven groups, based on the three grades (first to third year of upper secondary school) and sex, to acquire all grades and both sexes in all intervention groups. This led to the construction of the following seven groups: first-grade male, first-grade female, second-grade male, second-grade female, third-grade male, third-grade female, and others. Others were classified as participants where grade and/or sex was missing. After being categorised in this way, the individuals in each of the seven groups were computer randomised into three intervention groups (A–C) by a statistician, i.e. each individual was assigned the letter A, B or C. The three treatment groups were iMBI, Internet-based music therapy (iMT), and waiting list. The initial seven groups were only created for randomisation purposes.
Of the 283 participants who gave written consent, 202 students—142 female, 50 male, and 10 where the information on sex and/or school was missing—also answered the Web-based questionnaires. The more rural school provided 45 students and the remaining 147 participants came from the more urban school. The mean age was 16.9 years and the median age 17 years (range 15–19 years).
Both intervention groups represented Internet-based self-help programmes and the coordinator at the institution handled all contact with the study subjects. This was to preserve the single-blindness of the intervention to the researchers. The participants were allowed by the schools to do a 10-min daily intervention during lesson time and were asked to do at least one intervention on each school day. There were no reminders given to the students, i.e. they had to participate in their appointed intervention on their own incentive. The 8-week interventions commenced on March 15, 2012. The participants could, via e-mail, contact two of the co-authors of this study (CA or FT), who are also physicians, if they felt that their psychiatric health deteriorated during the interventions.
Power was calculated for the main outcome of the study, improvement in global severity index (GSI) (see below) with the conventional values for significance level of 0.05 (
α), and for power of 0.80 (
β). With three groups to be compared, we used a one-way ANOVA pairwise two-sided equality calculation. We estimated a priori that with a population sample of 73 in each group, the GSI in the iMBI group would differ significantly from the waiting list group. This was based on a face-to-face mindfulness study in adults where GSI, our primary outcome, was measured [
44]. In that study, the authors found an improvement in GSI from 0.91 (SD 0.71) to 0.53 (SD 0.51) and a median change of –43.2% after an 8-week MBSR course. According to the power calculation, we needed at least 219 participants, i.e. 73 in each group. To compensate for a presumed high pre-baseline dropout rate, we estimated the dropout rate before the first questionnaire to be 70%, based on two Swedish studies using GSI as an outcome in adolescents [
45,
46]. We also made an estimation of the post-baseline dropout rate, of 44%, based on the responders that actually complied with an intervention in an adult iMBI study [
43]. These assumptions led to an estimated total sample size of 1140 in the present study. We judged that the population of 1403 in the two chosen schools was adequate to address the study aims.
A recent systematic review of 36 studies [
47] listed the different definitions/measures of dropout and engagement. As it does not seem as if a consensus has been reached on what terminology is most appropriate, we discriminated those who logged in with those who did not as follows: ‘logged in’ and ‘not logged in’. We believe that this represents the most accurate description of those students who enrolled in our study [
47,
48].
Interventions
The two active interventions were Internet-based self-help programmes that were completely computerised without any human interference. They were based on modules of approximately 10 min and contained both video and audio material, but the focus was on the latter. Both interventions logged compliance in terms of log in identification, time at log in, and fulfilment of the session. They were accessible by any device that connects to the Internet, such as computers, smartphones, and tablets. All groups, including the waiting list group, got access to both interventions after all post-intervention questionnaires were completed. No further follow-up was performed.
Mindfulness-based intervention (iMBI)
The iMBI (In Swedish:
Mindfulness Grundkurs 2.0 from Mindfulnesscenter AB, Sweden) was designed by Dr. Ola Schenström, a family physician and renowned national expert in Sweden on clinical mindfulness meditation. The programme is an 8-week course consisting of sessions of 10 min of mindfulness meditation twice daily, 6 days a week. The modules consist of standard mindfulness meditation techniques, such as body scan and mindfulness of breath, and other perceptions [
49], and could be defined as an intervention based on mindfulness training [
31]. The intervention incorporates elements from both MBSR [
50] and more cognitively oriented parts from MBCT [
51]. An iMBI, in a study on anxiety in adults, has used the same platform and guided meditations, but with cognitive material focused on anxiety [
40]. For our study, a complete intervention was defined as at least 40 sessions, which is lower than the original programme but was deemed appropriate by the constructor of the programme.
Music therapy intervention (iMT)
We chose iMT as the active control to iMBI, as one meta-analysis on music therapy showed a good effect in stress reduction from listening to music both in itself and in combination with music-assisted relaxation techniques, and concluded that the best effect is on adolescents [
52]. Listening to music is also something that adolescents are generally interested in and tend to do, of their own free will, when feeling stressed or emotionally challenged [
53]. Furthermore, a small study on adolescents targeting depression with music therapy, which included active participation with instruments and interpretation in terms of painting, showed promising results [
53]. In addition, the partial similarity between the two Internet-based self-help programmes makes these two groups comparable. For the present study, a complete intervention was defined as at least 40 sessions.
The iMT, Musikintervention, was designed with the aid of Professor Björn Ejdemo, MD, and visiting professor of music at the Australian National University, who made a preliminary selection of pieces of music. Per Vegfors, MD, specialist in child and adolescent psychiatry, assessed the appropriateness of the music from an adolescent psychiatric perspective. A total of ten non-vocal classical music pieces accessible on YouTube were chosen that met the criteria of (1) accessibility, i.e. being relatively easy to listen to for an untrained ear, (2) being of approximately 10 min in duration, and (3) recognisable as being calming or soothing. The programme Musikintervention (Paxx Media AB, Sweden) used streamed music videos from YouTube. See appendix for list of music and interpretations.
Questionnaires
We used three well-defined, well-validated, and reliable psychometric tests, before and after the intervention, that had previously been used on adolescents: the Symptoms Checklist 90 (SCL-90) [
54], the Perceived Stress Scale (PSS-14) [
55], and the Pittsburgh Sleep Quality Index (PSQI) [
56]. The combination of scales was chosen to give an insight into the students’ perceived stress (PSS-14) and likely outcomes of that stress, expressed as low-quality sleep (PSQI) and increased general psychiatric symptoms (SCL-90), as previous studies have shown bi-directionally stress-associated impairments in psychiatric symptoms and sleep [
11]. The time required to fill in the questionnaires was adjusted to avoid questionnaire fatigue and keep good test–retest reliability.
Statistical analysis
To examine aim 1, i.e. to investigate whether there was a potential effect of the Internet-based self-help programmes on psychiatric and stress-related symptoms assessed before and after the intervention, we aimed to compare the scores of the different scales before and after the intervention among the active participants. We also used Student’s
t test to investigate whether the three groups differed at baseline. These comparisons were performed to check whether the randomisation procedure was successful. To examine aim 2, i.e. to investigate the feasibility in terms of compliance to the interventions with the two Internet-based self-help programmes, we used a CONSORT flowchart to display the dropout rates at the different steps of the study [
65]. The comparison in dropout rate between the two intervention groups (iMBI and iMT) was done using Fisher’s two-tailed exact test calculated with the number of participants that were randomised to either intervention and the participants that finished at least one session of either intervention. The Fisher’s two-tailed exact test was used as we intended to analyse whether the relatively small samples deviated from the null hypothesis that there was a difference between the dropout rates between iMBI and iMT. To examine aim 3, i.e. to investigate whether there is any association between compliance to the intervention and psychiatric and stress-related symptoms at baseline, we compared the participants that logged in and completed at least one session and those who did not log into the self-help programme. We compared whether there was a difference in psychiatric and stress-related symptoms between those two groups, separately in each intervention group (iMBI and iMT). We used a scatter plot to check whether the distribution was skewed and used the non-parametric Wilcoxon rank-sum test if the distribution was skewed and the parametric Student’s
t test when the data were normally distributed.
Statistical analysis was done using Stata IC 12 (Stata Corp, Texas, USA).