A Prevention Program for Insomnia in At-risk Adolescents: A Randomized Controlled Study

Participants were recruited from local secondary schools and community in Hong Kong from July 2015 to December 2017. The follow-up assessments were completed in January 2019. Family history of insomnia was collected by self-reported questionnaire used to assess parents’ lifetime and current insomnia disorder on the basis of Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) diagnostic criteria. The inclusion criteria included (1) ethnic Chinese youth aged 12 to 18 years old, (2) having at least 1 biological parent with current or a lifetime history of insomnia disorder, and (3) presence of subthreshold insomnia symptoms (ie, having insomnia symptoms at least once per month but <3 times a week in the past month). Adolescents who reported any of the following conditions were excluded from the study: (1) current or past history of neuropsychiatric disorder(s), (2) medical condition, sleep disorders (eg, delayed sleep phase syndrome), or (3) use of medication(s) known to interfere with sleep continuity or quality. The diagnosis of psychiatric and sleep disorders was assessed by the adolescent version of the Mini International Neuropsychiatric Interview²⁶  and Diagnostic Interview for Sleep Patterns and Sleep Disorders,²⁷  respectively.

Ethical approval was granted by the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee, and the trial was registered with the Chinese Clinical Trial Registry (identifier ChiCTR-IPC-15005966). All subjects provided parental consent and adolescent assent.

Participants were recruited from community and local schools. Eligible participants were randomly assigned to 1 of the 2 groups by using random sequence allocation. For participants who were recruited from schools, randomization was conducted by using the school as a unit rather than the individual to avoid contamination (ie, communication among students of the same school). The randomization method was effective because demographic characteristics and other baseline measures were similar between the 2 groups. Coauthors (J.Z. and M.W.M.Y.) who have extensive sleep medicine training were assessors, and they were masked to the group allocation.

Telephone interviews were conducted to assess insomnia symptoms of adolescents with a parental history of insomnia (Fig 1). Adolescents who fulfilled the initial inclusion criteria were invited to attend a face-to-face clinical interview to ascertain their eligibility for the study. After completion of the baseline assessment, adolescents were randomly assigned to either the control or intervention group. Participants in the intervention group received 4 weekly, 60-minute group-based sessions conducted either in their school or hospital. For the control group, no active intervention was provided, but they were required to attend all the follow-up assessments. All participants were invited to attend either a face-to-face or telephone clinical interview and complete a written or online questionnaire at postintervention, 6-month follow-up, and 12-month follow-up.

The prevention program was adapted and modified according to the key components of CBT-I (see Supplemental Table 3),²⁸  with the particular emphasis on the precipitating and perpetuating factors that led to the onset of insomnia. In particular, behavioral and cognitive strategies including stimulus control, sleep hygiene, constructive worry technique, and relaxation were introduced with the aim to foster good sleep habits and avoid negative behaviors or cognition that might trigger insomnia. The intervention was conducted in a group format with 6 to 8 participants per group. All intervention sessions were led by experienced therapists in sleep medicine, and they were closely supervised by sleep specialists.

The primary outcomes were (1) the incidence of insomnia disorder as ascertained by semistructured clinical interview on the basis of DSM-5 insomnia criteria and (2) the severity of insomnia symptoms as measured by the 8-item Insomnia Severity Index (ISI).

Participants were asked to complete a 7-day sleep diary that recorded total sleep time, time in bed, sleep-onset latency, and wake after sleep onset. Social jet lag was defined as a difference between weekday and weekend midpoint sleep. Sleep hygiene practices were measured by the Adolescent Sleep Hygiene Scale revised.²⁹  A higher score indicated a better sleep practice. Adolescents’ sleep-related beliefs and expectations were measured by the 16-item Dysfunctional Beliefs and Attitudes about Sleep (DBAS).³⁰  The average score is computed, with a higher score representing a greater level of dysfunctional belief. The vulnerability to stress-related insomnia was measured by the Ford Insomnia Response to Stress Test (FIRST).³¹  A total score was generated to indicate the vulnerability of sleep disturbance. Depressive and anxiety symptoms were measured by the 14-item Hospital Anxiety and Depression Scale (HADS).³²  Daytime sleepiness and fatigue were captured by the Pediatric Daytime Sleepiness Scale³³  and Multidimensional Fatigue Inventory,³⁴  respectively. Quality of life was measured by KIDSCREEN-27.^35,36  It consists of 5 dimensions including physical well-being, psychological well-being, autonomy and parents, peers and social support, and school environment. The total score of each subscale was computed on the basis of the Rasch model and was transformed into values with a mean of 50 and an SD of 10.

Sample size calculation was based on our previous school-based sleep education study.³⁷  It reported that the incidence rate of insomnia among high-risk adolescents ranged from 14% to 20%. With regard to the meta-analysis on the general efficacy of CBT-I, the number needed to treat (NNT) for CBT-I was 3.5.³⁸  With a power of 80% and 95% confidence level, a planned total sample size of 250 is needed for an estimated attrition rate of 30%.

Analyses were conducted on the basis of the intention-to-treat principles. Baseline characteristics were compared by using χ² and analysis of variance. The incidence of insomnia was defined as the presence of frequent insomnia symptoms at least 3 times a week. Chronic insomnia was defined as the presence of frequent insomnia symptoms for at least 3 months, whereas acute insomnia was defined as having frequent insomnia symptoms for <3 months. Kaplan-Meier survival analysis was used to compare the cumulative incidence of insomnia between the 2 groups, and Cox proportional hazard regression analysis was employed to estimate the hazard ratios (HRs) against the control group.

To test for significant differences in the continuous outcome measures between the 2 groups, multilevel mixed models were employed to examine the interaction effects of group (intervention and control) by time (baseline, postintervention, 6 and 12 months after intervention). Participants with missing data at ≥1 follow-up assessment were also included in the analyses. This approach allowed us to consider subjects as a random effect. Because a large proportion of the participants were recruited from local schools, another separate sensitivity analysis was conducted and additionally controlled for the potential school cluster effect. The significant results observed between 2 sets of analyses were identical. The findings without controlling for school effect were presented. Between-group effect sizes were calculated to indicate the degree of change across time by using Cohen’s d, in which values of 0.20 and 0.50 represent small and medium effects, respectively. Statistical analyses were performed by using SPSS (IBM SPSS Statistics, IBM Corporation) version 20.0 for Windows. A P value of <.05 was considered as statistically significant.

Among 2262 adolescents with a parental history of insomnia, 953 reported subthreshold insomnia symptoms (Fig 1). A total of 335 subjects attended the face-to-face clinical interview, and 242 met the study criteria and consented to participate. A total of 121 adolescents (mean age = 14.7 ± 1.8; female: 51.2%) were randomly assigned to the intervention group and 121 (mean age = 15.0 ± 1.7; female: 62.0%) were allocated to the control group. A total of 218 adolescents (90.1%) attended the postintervention assessment, and 219 (90.5%) and 205 (85%) attended the 6-month and 12-month follow-up, respectively (Fig 1). The rates of missing data were similar between 2 groups at all time points (percentage of missing data in the intervention versus control group: postintervention: 13.2% vs 6.6% [P > .05]; 6 month: 9.9% vs 9.1% [P > .05]; 12 month: 18.2% vs 12.4% [P > .05]). For the intervention group, 82% of them attended all 4 treatment sessions, 15 of them (12%) attended 3, whereas the remaining subjects (n = 7; 6%) attended <3 sessions. There were no significant differences in the age and sex between the 2 groups. At baseline, both groups had a similar level of insomnia symptoms and other measures (all P >.05). In addition, participants who completed the 12-month follow-up did not differ from the dropouts in terms of age, sex, and other characteristics.

Survival analysis of incident insomnia indicated a significant advantage for the intervention group compared with the control group. The cumulative incidence of insomnia (including both acute and chronic insomnia) over the 12-month period for the intervention and the control group were 5.8% (n = 7) vs 20.7% (n = 25), respectively (P = .002; NNT = 6.7) (Fig 2). The cumulative percentage of chronic insomnia between intervention and control groups was 2.5% (n = 3) vs 8.3% (n = 10), respectively (P = .049; NNT = 17.2). The HRs were significantly lower for the overall incidence of insomnia (acute and chronic) (HR= 0.29; 95% confidence interval: 0.12–0.66; P = .003) but not for chronic insomnia (HR = 0.32; 95% confidence interval: 0.09–1.13; P = .075) in the intervention group compared with the control group.

The severity of insomnia symptoms was significantly reduced in the intervention group at postintervention (P = .02) and throughout the 6-month (P = .02) and 12-month (P = .004) follow-up assessments (Table 1, Fig 3). Between-group effect sizes were −0.32 at postintervention, −0.34 at the 6-month follow-up, and −0.36 at the 12-month follow-up.

There was a significant interaction effect on the vulnerability to stress-related insomnia scales (P = .03) (Table 1). Adolescents in the intervention group had reduced FIRST scores at postintervention (P = .04; between-group Cohen d = −0.23) and 12-month follow-up (P = .01; Cohen d = −0.33). No significant interaction effect was documented in DBAS (P = .06), but a significant effect was observed at postintervention (P = .02; Cohen d = −0.26). We recorded a significant effect of the preventive intervention on sleep hygiene (P = .02). The overall sleep hygiene practice was improved at postintervention (P = .003).

The percentage of participants who returned a sleep diary was 72%, 60%, and 68% at postintervention, 6-month follow-up, and 12-month follow-up, respectively. Adolescents who received the preventive intervention had a longer total sleep time compared with the control group at postintervention (P = .01; Cohen d = 0.42), and the improvement was maintained throughout the study period (Table 2). There were no significant interaction effects on sleep diary parameters, including social jetlag (all P >.05).

Daytime sleepiness was reduced in the intervention group at postintervention (P < .05; Cohen d = -0.26) but not at the 6- and 12-month follow-ups (all P >0.05) (Table 1). There was no overall interaction effect on the depressive measure (P = .08). However, the intervention group had significantly lower depressive score at the 12-month follow-up (P = .02; Cohen d = −0.31) but not at other time points (all P >0.05) (Table 1). There were no significant effects on anxiety, fatigue, and quality-of-life measures between the intervention and the control groups.

To the best of our knowledge, this is the first randomized controlled study in which authors examine the efficacy of a brief preventive program for insomnia in at-risk adolescents who have positive family history of insomnia and subthreshold insomnia symptoms. This study demonstrated that a brief modified CBT-I preventive program could reduce the incidence rate of insomnia disorder, severity of subthreshold insomnia symptoms, and vulnerability to stress-related sleep disturbance. There was a 71% risk reduction in the development of insomnia disorder (both acute and chronic) over the 12-month follow-up period.

Individual variations in vulnerability to sleep disturbance have been suggested as predisposing factors for the development of insomnia.³⁹  The reduction in the FIRST score in the intervention group suggested that this preventive program was able to reduce individual sleep reactivity toward stress, which in turn lessened the risk of insomnia. However, the DBAS score, which reflects another commonly noted cognitive perpetuating factor in chronic insomnia,^18,28  was only transiently improved at postintervention, and the effect decayed steadily over time. Although individual vulnerability to stressful life events and dysfunctional sleep beliefs are 2 factors that often predict the onset of insomnia, they are considered as 2 domains with different weighting in the development and maintenance of insomnia.³¹  Thus, we postulated that lowering sleep reactivity toward stress is an important approach to prevent the onset of insomnia, whereas restructuring dysfunctional beliefs might be more important in the treatment of chronic insomnia.

The magnitude of preventive effect in our current study was comparable to that of previous studies using a similar high-risk approach that aimed to prevent depression among adolescents.⁴⁰  A recent meta-analysis study on the impacts of group-based preventive intervention for depression in children and adolescents reported an average of −0.31 to −0.11 effect size on the end-point outcomes,³  which was comparable to our current findings in the prevention of insomnia (effect size: −0.36). Interestingly, in the current study, we also report a lower level of depressive symptoms at 1-year follow-up among adolescents in the intervention group. Given the high comorbidity and reciprocal relationship between insomnia and depression, previous studies have explored the possibility of treating insomnia to prevent the subsequent onset of depression.^41,42  Although existing evidence suggested that targeting established insomnia might be a promising strategy to prevent depression,^23,41  the current finding supports a further possibility that prevention of insomnia might also reduce depression. Nonetheless, a longer follow-up period would be needed to determine the sustainability of this positive effect on depression.

The improvement in daytime sleepiness was in accordance with the improvement of total sleep time in the intervention group. Total sleep time was extended by 39 and 7 minutes for the intervention and control group, respectively, and the gain was maintained throughout the 12-month follow-up period. Some may argue that adolescents might have reprioritized their sleep after the intervention with consequent behavioral modification.³⁷  However, a previous sleep education program suggested that simply changing one’s attitude and increasing knowledge toward sleep is not sufficient enough to change adolescent sleep habits.³⁷  Thus, it is likely that specific behavioral and cognitive components in the preventive program would have contributed to these positive behavioral changes. In contrast, the nonsignificant effects on other sleep parameters were understandable because our adolescents had relatively normal sleep quality (eg, the average baseline sleep efficiency was 94%).

The current study had several strengths. First, there was a high program completion rate, indicating that this program was feasible and acceptable to adolescents. Secondly, the incidence of insomnia was confirmed by clinical diagnostic interview according to DSM-5 diagnostic criteria, which provided a better and complementary estimation than sole reliance on the self-reported questionnaire.⁴³  Finally, the longitudinal follow-up period of this study was up to 12 months, which allowed us to assess the medium-term effects of this prevention program.

However, the study had some limitations. First, we incorporated a nonactive control condition. Therefore, it remained possible that nonspecific factors, such as the placebo (eg, peer support) or Hawthorne effect (research participation effect) could potentially contribute to the positive outcomes. Nonetheless, this was the first trial of its kind in which an inactive or waitlist control was commonly used.²²  Second, the sample size in this study appeared to be moderate in comparison to other preventive trials. A larger sample size will be needed to increase the generalizability of the study findings. Third, functional outcomes and sleep variables were measured by self-reported inventories. Nevertheless, self-rated inventories are generally considered to have sufficient validity in assessing sleep habits in adolescents.^9,13  In addition, although the attendance rate of the follow-up assessment (∼85%) was satisfactory, the loss of the follow-up data, particularly the sleep diary, might lead to certain bias to the study findings. Taken together, we recommended that in future studies, researchers should incorporate an active comparison group with a focus on other health education to control for the potential placebo or Hawthorne effect. A longer follow-up period is also needed to explore the sustainability of the intervention effects, especially on depression.

With this study, we provided initial novel evidence that a brief modified CBT-I program was effective in preventing future onset of insomnia, improving the associated vulnerability factors and functional outcomes in at-risk adolescents. Further studies with more robust study design (such as incorporation of an active control group) and a larger sample size are warranted to further confirm the effectiveness of the insomnia prevention trials in high-risk adolescents and assess the impact of insomnia prevention program in reducing the health care burden in the future.

CBT-I

cognitive behavioral therapy for insomnia

DBAS

Dysfunctional Beliefs and Attitudes About Sleep Scale

DSM-5

Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition

FIRST

Ford Insomnia Response to Stress Test

HADS

Hospital Anxiety and Depression Scale

HR

hazard ratio

ISI

Insomnia Severity Index

NNT

number needed to treat