Is a Brief Body Scan Helpful for Adolescent Athletes’ Sleep Problems and Anxiety Symptoms?

This study was a quantitative longitudinal randomized controlled trial with a within- and between-group design in a school setting. Six high schools specialized in cross-country skiing and basketball and enrolled in the Swedish national sports programs were contacted. Within each school and class, adolescents were randomly assigned to four arms: 4 or 8 weeks body scan intervention or 4 or 8 weeks relaxation as active control groups. Participants were individually informed about their allocation.

Data was collected at four time points: baseline (T0), after 4 weeks (T1), after 8 weeks (T2), and after 16 weeks (T3). The schools started the interventions at four different timepoints during the schoolyear between September 2016 and February 2017. All schools included in this study have integrated sports training in the adolescent athletes’ schedules. Schools were located all over Sweden and in both rural and urban regions. The schools had been selected through contact with one of the researchers.

Participants had to be at least 15 years old (according to the Swedish law, no consent from parents or guardians is required from 15 years and older for participating in research that is not including treatment) and being (a) semi-elite and/or competitive-elite adolescents who were presently active in their sport, and (b) able to speak, read, and write in Swedish. Most participants did not live together with their family during the school year. Semi-elite athletes were defined as athletes competing below the top standard in their sport (e.g., in talent programs and development programs), while competitive-elite athletes were defined as those who regularly compete at the highest level in their sport (e.g., top divisions) (Swann et al., 2015).

A power analysis was conducted to estimate the sample size by using G*power version 3 for Windows (Faul et al., 2007). The power analysis was carried out using an ANOVA test for repeated measures and within-between interaction with the significance level set to 0.05 and power = 0.8. The number of interventions was 4 and the number of measurements was 4. The correlation between repeated measurements of each outcome variable was assumed to be at least r = 0.5 (default). Assuming small to moderate effect size, i.e., Cohen’s d = 0.1–0.25 for both the anxiety and the sleep problem scales, a total sample size of 40–200 participants is required, or 10–50 participants per intervention category (Cohen, 1992). Anticipating a dropout rate of at least 20%, the required sample size was set to 260 participants.

Among eligible adolescent athletes 261 participants from 7 schools were selected. Fifty-five participants were excluded for the following reasons: one school (basketball) declined to participate (n = 40), incomplete demographic information (n = 7), incomplete data on BDI-I (n = 5), and high values on BDI-I (≥ 25) indicating severe depression (n = 3). The final sample consisted of 206 participants, of whom 98 were girls and 108 boys (mean age 17.0 years (SD = 0.9), age range 15–19 years, 163 basketball players and 43 skiers).

After ethical approval from the Regional Ethics Board, the recruitment started in September 2016 and continued during the school year 2016/17. All coaches at the schools were contacted via e-mail and received a detailed letter explaining the purpose of the study. The first author and/or a research assistant visited the schools and explained the purpose of the study. All adolescent athletes had the opportunity to ask questions and interested participants provided written informed consent to participate in the study. After filling in the information on their age, sex, practicing the sport, years of athletic experience, and years competing (Table 1), the three questionnaires with BDI, ISI, and BAI were administrated in a random order and were filled in a classroom or auditorium. The first author or assistants were present to answer questions. The participants who were excluded from further participation due to high levels on BDI received personal information about the exclusion and, if needed, advice on how to get help from health care facilities.

Body scan was adapted and influenced from Mindfulness Based Stress Reduction (MBSR; Kabat-Zinn, 2005) and integrated into the school context. Neither coaches nor teachers were involved in instructions of body scan practice. The active intervention groups took part of a 4-week or 8-week 8-min audio-guided body scan exercise recorded by the first author. The participants received the audio-guided exercises via e-mail and were asked to do these on their own (see Supplementary Materials for a complete translated transcript). Participants were instructed to sit or lay down, while listening to the recording. Immediately before and after listening to the audio recording, the participants were asked to complete an evaluation of the level of being in the present moment (level 0–100). The body scan was a verbally recorded guidance through the body from the feet to the hands. Instructions included, for example, “Just notice the contact with your left foot against the floor. What do you notice? Just pay attention to what you become aware of.” The participants were not told when or where to do the exercises but were encouraged to practice them regularly. No previous experience of body scan was needed to be able to follow the instructions. No music was added in the background of the recorded body scan practice.

The participants in the active control groups received an 8-min audio-guided relaxation exercise recorded by the first author and modified from Öst’s brief applied relaxation (Öst, 1987). The brief relaxation was applied from a CD produced by the first author (Johles, 2002), a psychologist with 35 years’ experience in Cognitive Behavioral Therapy (CBT). The method and recording have been frequently used in clinical and non-clinical contexts since 2002. The participants received the audio-guided exercises via e-mail and were instructed to practice regularly. They were instructed to sit or lie down when they were going to do the exercises. The participants were instructed as follows: “Relax your left hand and arm. Every time you breathe out you will relax your arms more. Let the relaxation continue up to your shoulders.” During the last minute of the program, participants were told to sit in silence and relax. No previous experience of relaxation techniques was needed to be able to follow the instructions. No music was added in the background (see Supplementary Materials for a complete translated transcript).

Measures of sleep problems and of anxiety symptoms were collected at baseline (T0) and 1 month (T1), 2 months (T2), and 4 months (T3) after baseline. At baseline, we also assessed a scale for depressive symptoms in order to identify and exclude pupils with severe depression. The written results were reported by each participant and collected by the first author or the research assistants visiting each school.

The Kruskal-Wallis tests as well as chi-square tests were used to compare baseline characteristics between the four intervention categories. The temporal change in anxiety or sleep problems was the main effect measure that is compared by type and length of intervention. We applied linear repeated measures models to compare the time trends in outcome from baseline to 4 months between the four intervention categories, while accounting for correlation of outcome measures within individual participants. Specifically, we used a random coefficient model with an unstructured covariance matrix for intercept and slope describing the linear time dependence of each outcome score (anxiety symptoms and sleep problems). To this aim, we introduced product (interaction) terms for time and intervention category and performed an overall F-test for differences between any of the intervention categories. In addition, we tested whether the time dependence differed between the mutual intervention and control groups (Brown & Prescott, 2015). The score describing sleep problems was approximately normally distributed at all four timepoints. Because the anxiety symptom scores were skewed to the right, they were log-transformed in order to assure normally distributed residuals in the linear regression model. A small number of participants had scores with values equal to 0, which did not allow for the calculation of the logarithm. For this reason, we added an offset of 0.5 points to the anxiety symptom scores before taking the logarithm. The regression results hardly changed for values of the offset between 0.2 and 1 (not shown). All regression models were adjusted for age, sex, and type of sport. Because of their strong mutual correlation, school and type of sport could not be included in the same model. For this reason, we presented results for models adjusting for type of sport in the main text, and school-adjusted results in the supplement. Time trends in units of months were derived from beta values together with 95% confidence intervals (CI). For outcome variables on logarithmic scale, we calculated the relative change in percent given by (exp(beta) − 1) × 100%. The internal consistency of items presenting each psychometric scale was assessed in terms of Cronbach’s alpha and McDonald’s omega. Mc Donald’s omega was calculated using the package psych for R 4.2.2 (http://​www.​R-project.​org). All other analyses were performed using IBM SPSS Statistics for Windows (Version 22.0), SAS (version 9.4), and MATLAB (R2016b). Statistical significance was set at p-value < 0.05 (two-sided tests).

Table 1 shows baseline characteristics by intervention category for age, sex, type of sport, hours practicing per week, years of performing, and years at elite level, as well as for anxiety symptoms and sleep problems. The numbers of participants were equal after the randomization. However, differences in numbers of participants in each intervention group arose due to various reasons for exclusion (see flow chart in Fig. 1). Proportions of female participants or type of sport (basketball vs. cross-country skiing) did not differ between categories of intervention. Also, mean values for continuous variables did not differ by intervention category, with the exception of the anxiety symptom scale, which showed slightly higher values for athletes assigned to 4 weeks of body scan compared to the other categories (p = 0.03). Overall, 87% of athletes showed minimal (56%) or mild (31%) anxiety levels, and absence of insomnia (55%) or sub-threshold insomnia (33%) sleep problems. The correlation between the two scales was of moderate size, r = 0.52 (0.41, 0.62).

Figure 2 shows the mean values for anxiety symptoms and sleep problems from baseline (T0) to 4 months after baseline (T3). Mean values with standard deviation (SD) for the two scores are also given in Supplementary Table S1. The data suggest a linear time dependence (Fig. 2), including baseline as well as pre- and post-intervention data.

To assess whether treatment effects differed by type or duration of treatment, we compared the effect sizes of temporal decline between the four treatment categories using linear repeated measures models adjusted for age, sex, and type of sport. The results for linear time trends by intervention category are shown in Table 2. In sensitivity analyses, time trends were estimated based on follow-up restricted to 2 months after baseline, i.e., omitting the post-treatment follow-up at 4 months after the start of interventions (Supplementary Table S2). Table S3 shows the results for the linear model adjusted for school instead of type of sport.

The raw data on sleep problems (ISI) indicate decreasing levels with both 4 weeks body scan and 4 weeks and 8 weeks relaxation, but less so with 8 weeks of body scan from baseline to the last follow-up 4 months after baseline (Fig. 2). Negative time trends were confirmed in multivariable linear models, but time trends were not different from zero (Table 2). None of the pairwise comparisons indicated that the effect sizes differed between intervention groups (p > 0.5), which is consistent with overlapping CIs for intervention-specific effect sizes.

Figure 2 shows that the mean values of the anxiety scale decreased over time in all four intervention categories. The repeated measures model showed that these negative trends were all significantly different from zero (Table 2). For instance, anxiety symptoms decreased by about 16% per month in the group of participants who received a 4-week body scan. Slightly smaller temporal changes were observed in the other intervention groups. An overall test showed no differences in effect size between the four intervention categories (p = 0.8). Also, pairwise tests indicated no differences between the intervention and active control groups regarding the temporal decline, neither between 4-week interventions (p = 0.5) nor between 8-week interventions (p = 0.8). The largest contrast was observed among pupils receiving body scan, where the effect of the 4-week intervention (− 15.8%/month) was larger than that of the 8-week intervention (− 10.7%/month) but the effect size did not differ by length of intervention (p = 0.3). These results were largely confirmed upon the exclusion of the last examination at 4 months after baseline (Supplementary Table S2).

Overall, female adolescent athletes scored higher on the sleep problem and anxiety scale. Age was not associated with either of the scales, and type of sport showed a weak association with sleep problems indicating more problems among basketball players compared to skiers (not shown). Adjustment for school instead of type of sport hardly affected the results given in Table 2 (Supplementary Table S3).

Supplementary Fig. S1 shows how anxiety symptoms (BAI) and sleep problems (ISI) develop over time for participants starting in September, October, November, and February. For this analysis, the observations from all participants were combined irrespective of type of intervention. We see that levels decreased over time irrespectively of month of start but that the initial values for anxiety and sleep problems were larger for participants who started in October or November compared to participants starting in September or in February. When levels were generally highest, i.e., for participants starting in October and November, the level of sleep problems remained stable over time.

To our knowledge, this is the first longitudinal RCT study to investigate whether a brief body scan intervention can reduce adolescent student athletes’ sleep problems and anxiety symptoms. The study’s RCT design, in combination with close cooperation with coaches and teachers in the school and athletic setting, is a strength.

One limitation of this study is the high dropout at follow-up, which is similar to clinical research with adolescents where dropout commonly was between 40 and 60% (Wierzbicki & Pekarik, 1993). A plausible explanation for the dropout in our study could be that the prerecorded audios were sent to the participants via e-mail and the downloading and application of the audio were self-administrated. Compliance had several practical problems in this study. Ranking and performance among skiers were collected before and after the intervention. However, the majority of students did not complete the questionnaire; therefore, this data is left out. Moreover, there were practical limitations for controlling if, when, and for how long the participants did the practice with body scan or relaxation. Some participants might have been doing other activities while listening to the audio recordings or might not have done the practice at all. Upcoming research might benefit from collecting all data with digital instruments including measuring compliance and skills in performing. By including baseline values in the estimation of time trends, the differences in baseline values are only indirectly adjusted for, and it is possible that the higher starting point in the 4 weeks body scan group explains that this intervention also showed the largest effect on anxiety. In addition, as the participants generally reported a low degree of sleep problems and anxiety symptoms, the small variation in outcome values may have hampered the differentiation of intervention effects on the two scores. Lastly, we acknowledge the lack of an untreated control group, as mental problems might have decreased over time independent of intervention. However, sensitivity analyses showed that levels of anxiety symptoms and sleep problems generally decreased during intervention although the absolute levels varied considerably during the year (Supplimentary Fig. S1). For instance, the relatively high level of anxiety among pupils starting in October suggests that anxiety symptoms among pupils starting in September also had increased, if they not had received an intervention.