Study design
The randomized controlled trial (RCT) uses a multicenter, prospective, randomized, open-label, blinded end point study design (PROBE) (19)to compare eCBT-I with TAU. A total of 48 patients will be randomly assigned to the eCBT-I or TAU group. Assessment will be performed at baseline (T1), one week after the 6-week intervention period (T2) and after 6-weeks follow up (T3). The total duration of participation is 14 weeks. See figure 1 for a flowchart.
Recruitment, randomization, blinding and treatment allocation
Participants will be recruited from five outpatient rehabilitation centres spread over the Netherlands (Heliomare Rehabilitation at Wijk aan Zee, Reade at Amsterdam, Rehabilitation Friesland at Beetsterzwaag, Basalt Rehabilitation at Delft and The Hague and Adelante at Hoensbroek). Patients that are eligible based on the inclusion and exclusion criteria will be asked to participate. After informed consent is given, participants will complete the baseline measurement. Participants will then be randomly assigned to the eCBT-I or the TAU group. All participants will continue with standard rehabilitation care for various complaints, which do not specifically address insomnia (see box 2). As complaints differ between persons, interventions differ as well. This care can include psychotherapy aimed at mood or other psychopathology, therapy aimed at cognitive functioning, physiotherapy, fitness, occupational therapy and social work. The therapy dosage depends on different needs and capacity of participants. The variation in rehabilitation care between participants is expected to be similar for both groups and will be registered at follow up by their clinicians . The TAU group will only receive standard rehabilitation care, and can receive the eCBT-I after the study period. Participants can leave the study at any time for any reason without consequences. In case of withdrawal, participants will be asked to still do the measurements to control for attrition bias and a new patient will be included in accordance with the procedure. The measurements and intervention are not likely to have adverse consequences. Randomization will be done by a research randomizer program (www.randomizer.org) using blocks of four participants to balance participants equal to both groups within centers. Treatment assignment is blinded, as researcher assistants, clinicians and participants are unaware of number of included participants per center and not informed of this randomization procedure.. Treatment allocation is not blinded for therapist and participant, as the nature of cognitive behavioral therapy does not permit blinding, but will be concealed until participants are recruited and baseline measurements are finished. Outcome assessment is self-reported, and therefore not blinded for participant, but potential assessment bias of clinician or data collector on outcomes is excluded. Data collector is blinded for treatment allocation as well. Research assistant responsible for data input is blinded. See table 1 for an overview of all measurements.
Eligibility criteria
Participants are eligible for inclusion if diagnosed with stroke or TBI and insomnia disorder according to DSM-5 criteria. Furthermore, they should be aged 18 or older and capable of using the internet. Exclusion criteria are untreated sleep apnea, current or expected treatment with a main focus on fatigue or sleep during the study, unstable medication regiments, use of medication with insomnia as side effect, alcohol or drug abuse, and a major untreated or unstable medical or psychiatric condition. Users of sleep medication will be encouraged to finish medication before enrolment or to keep intake stable during the study period. Depending on type and dosage of medication, the physician will determine the period needed to exclude the influence of finishing sleep related medication on the outcome measures.
Intervention
The eCBT-I comprises six guided weekly sessions provided online, combined with two face-to-face sessions of 60-minutes and a smartphone diary-app for daily reporting of sleep times and subjective sleep quality. Each session is structured around one topic and contains specific information, assignments and testimonials of two patients with insomnia after brain injury to illustrate sleep problems and homework assignments. All participants will receive online personal feedback after each session and will be encouraged to practice daily with the provided exercises, downloadable within the eHealth intervention on a daily basis. Participants can start with the next session after they have read their personal feedback. They can contact their therapist at any time by means of the integrated email function. Participants will be encouraged to complete the diary every day by their therapist.
The eCBT-I is based on well-established CBT-I components and includes behavioral and cognitive techniques. These techniques contain sleep hygiene education, stimulus control, sleep restriction, cognitive restructuring, activation, relaxation, fatigue- and stress management. The eCBT-I has been adjusted to people with acquired brain injury in general, both with respect to content and the way of conveying information. Content adjustments include specific education about the nature and treatment of insomnia after acquired brain injury, testimonials of patients with insomnia following acquired brain injury, and adaptations to cognitive impairments due to brain damage. Information is given in clear and short texts and is visually supported. An option is included to allow for listening to an audio version of the texts. All sessions follow the same structure, with repetition of key points. Specific feedback suggestions for each session will be provided for the therapists.
The length of our eCBT-I protocol of 6 weeks is comparable with eCBT-I protocols in other studies. The CBT-I protocol in other studies examining efficacy in ABI consisted of 4 to 8 weekly session (12-17), and 6-weeks for eCBT-I (18). In a review on the efficacy of eCBTI in the general population an average length of 5.5 weeks was found, with a range of 2-9 weeks (11).
The eCBT-I will be given by an experienced registered healthcare psychologist, trained in using the eHealth intervention. Adherence will be monitored by checking frequency of registration in the sleep diary app, time spent online, and online assignments done. For a detailed description of the intervention per week see Box 1.
Outcomes
Sleep outcome measures
The primary outcome measure is the change in insomnia severity measured with the Dutch version of the Insomnia Severity Index (ISI) at posttreatment. The ISI consists of 7-items and uses a 5-point scale to measure to which extent participants experience insomnia. The total score ranges from 0 (no insomnia) to 28 (severe insomnia). A cut-off of 10 is used to indicate clinical levels of insomnia in this study, similar to other studies (20-22). The Minimally Clinical Improvement Difference (MCID), which indicates the minimal improvement to be clinically significant, is a reduction of six points (23). The internal consistency is adequate (Cronbach’s alpha = 0.74-0.78). The ISI is selected as it is sensitive to treatment response (22, 24), and used in comparable research worldwide, including the Netherlands (20, 21). Secondary sleep outcome measures include overall sleep disturbances assessed with the Pittsburgh Sleep Quality Index (25) and the following sleep features derived from the sleep diary app: total sleep time, sleep onset latency, number of nocturnal awakenings, sleep efficiency and subjective sleep quality.
Other outcome measures
Secondary outcome measures cover fatigue, anxiety and depression, subjective cognitive functioning and societal participation. Fatigue after acquired brain injury will be assessed with the Dutch Multifactor Fatigue Scale (DMFS). The DMFS measures 5 aspects of fatigue: Impact of fatigue, Mental fatigue, Signs and Direct consequences of fatigue, Physical fatigue and Coping with fatigue. All subscales of the DMFS showed sufficient to good reliability (Cronbach’s alpha = 0.70 to 0.91), good convergent validity with an existing fatigue scale, and good divergent validity with measures of mood and self-esteem (26). Anxiety and depression symptoms will be assessed with the Dutch version of the 14-item Hospital Anxiety and Depression Scale (HADS). The reliability of the HADS is good (Cronbach’s Alpha = 0.71 to 0.90) as is the test-retest reliability (0.86-0.90) (27). Cognitive Failure Questionnaire (CFQ) is a measure of subjective cognitive functioning. Internal consistency is good (Cronbach’s alpha= 0.88) as is the test-retest reliability (0.83) (28). The Utrecht Scale for Evaluation of Rehabilitation – Participation (USER-Participation) is a questionnaire to rate objective and subjective participation after rehabilitation. Internal consistency is satisfactory (Cronbach’s Alpha = 0.70-0.91), and test- retest reliability is 0.65 for the Frequency scale, 0.85 for the Restrictions scale, and 0.84 for the Satisfaction scale (29, 30).
Other study parameters
Demographical, injury related and clinical variables which may influence the treatment effect will be registered: age, gender, diagnosis, time since injury, insomnia duration, use of prescribed sleep medication, use of other medication, comorbid psychiatric and somatic disorders, educational level, and currently being employed. Possible presence of sleep apnea will be screened with the Stop-Bang questionnaire (31) in participants that have not been evaluated for its presence with the gold standard overnight polysomnography (32).
Sample size calculation
Calculation of power and group size in a repeated measures design require an estimate of the expected intraclass correlation coefficient (ICC, within subject correlation). We therefore resorted to a recent randomized trial of Dekker and colleagues (2020) on the effect of eCBT-I on ISI in people with insomnia. We used this intraclass correlation coefficient as indication of follow up of ISI assessments across 6 weeks in our study population with acquired brain injury. That study reported an ICC of 0.54 for ISI assessments repeated across 6 weeks in 175 people suffering from insomnia without brain injury (33). Calculation of power and group size moreover requires an estimate of the expected treatment effect size. Whereas treatment effects of eCBT-I are often reported to be of moderate to large size, we preferred to be somewhat conservative and expect a somewhat smaller than moderate effect in the population of TBI and stroke and therefore used f= 0.20. Calculation of the required sample size using G*Power (34) for ANOVA, with two repeated measures and ICC of 0.54, indicated that 48 completers would provide (2x 24), at a significance of alpha=0.05, sufficient power (1-beta=0.80) for a minimal detectable time-by-group interaction effect of f=0.20 (small to moderate).
Statistical analyses
Data analysis will be performed with SPSS 23 (IBM; Armonck, USA). Means and standard deviations of the demographic, injury related variables and the clinical characteristics collected at baseline will be calculated. For data not normally distributed, median and interquartile range will be reported. Independent T-tests will be used to check for an imbalance between groups. Nominal variables will be checked with χ2- Tests and Ordinal variable with the Mann-Whitney Test. Statistical significance will be set at a p-value of 0.05. To accommodate likely occasional missing days, the repeated measures obtained by diaries will be analyzed with mixed effect models. The treatment effects will be examined using a repeated measures ANOVA. In the case of significant baseline imbalance, post hoc analysis of covariance (ANCOVA) will be used on the change scores of the outcome measures with baseline scores as covariates. Intention-to-treat analyses will be conducted and presented as primary results. Additionally, a per protocol analyses will be performed. Also, we will compare the eCBT-I and TAU group with regard to the percentage of participants who have improved (reduction > 6 points on the ISI) and recovered (ISI<10). We will use mixed effect regression models to explore whether treatment efficacy, i.e. the time-by-treatment effect, is modified by time since brain injury and by diagnosis (stroke versus TBI). As normality assumptions may not be met with this small sample size, data distributions and all model assumptions will be checked for all analyses. If model assumptions are not met, data will either be transformed or analyzed using a non-parametric test, as appropriate.