Efficacy of Cognitive-Behavioral Therapy for Insomnia Associated With Traumatic Brain Injury: A Single-Case Experimental Design

Participants 

Eleven subjects with mild to severe TBI participated in this study (see table 1 for sociodemographic information and clinical characteristics from the medical file). Participants were recruited with the help of clinicians working in major rehabilitation centers of the Québec City metropolitan area as well as through advertisements sent via mailing lists of TBI associations or support groups. The mean age of participants was 27.3 years (range, 20−46y) and the mean years of education was 14.4 (range, 11−16y). There were 6 men and 5 women; 9 were single (although 5 of these had companions) and 1 was married. Eight were not working at the time of the study, 1 was progressively returning to work, and 2 were working or studying full-time.

Table 1. Sociodemographic and Clinical Characteristics of the Participants

	Participant	Age (Sex)	TBI Severity	Months Since Injury	Glasgow Coma Scale Score	Neurologic Examination	Medication	Intellectual Function⁎	Cognitive Deficits⁎	Insomnia Type	Insomnia Duration
	01	26(M)	Mild	25.5	ND	Negative	None	Average	Significant	Initial	1y
	02	32(M)	Severe	28	6/15	Positive	Trazodone	Average	Significant	Maintenance	2y
	03	22(F)	Moderate-severe	13	6/15	Positive	None	ND	Mild	Maintenance	1y
	04	21(M)	Moderate	9	14/15	Negative	None	High average	Significant	Maintenance	3mo
	05	38(F)	Mild-moderate	25	14/15	Positive	None	High average	Mild	Maintenance	2y
	06	21(M)	Severe	25	3/15	Positive	None	Borderline	Significant	Mixed	2y
	07	21(F)	Moderate	41.5	11/15	Positive	Venlafaxine diphenhydramine	ND	Mild	Mixed	2.5y
	08	20(F)	Moderate-severe	31.5	5/15	Positive	None	Average	Mild	Mixed	2y
	09	46(M)	Moderate-severe	35.5	13/15	Positive	None	Average	Mild	Maintenance	2.5y
	10	23(F)	Severe	36	3/15	Positive	Trazodone	Borderline	Significant	Maintenance	3y
	11	30(M)	Mild-moderate	33	ND	Negative	Venlafaxine gabapentin oxycodone	Low average	Significant	Initial	3y

Abbreviations: F, female; M, male; ND, no data available.

⁎ As estimated in the earliest neuropsychologic report postaccident.

The average time since injury was 25.64 months (range, 9−41mo). The severity of head injury ranged from mild to severe (1 mild, 2 mild to moderate, 2 moderate, 3 moderate to severe, 3 severe). Severity of injury was evaluated by the multidisciplinary team at the rehabilitation center according to standardized criteria29 used routinely in the field of traumatology. Eight participants had positive neurologic scans (ie, visually perceptible neurologic damage) and 3 had negative scans. Four were taking medication regularly (7d/wk) at the time of the study (either trazodone, venlafaxine, or gabapentin). All participants had some form of cognitive deficit (eg, slowing of information processing, attention or memory deficits, word finding difficulties) and at least 1 psychosocial problem (eg, irritability, marital tension, financial stress) as reported in a filed neuropsychologic or psychosocial evaluation. Several had depression or anxiety symptoms.

To be included in the study, participants had to (1) be aged between 18 and 50 years (to control for the confounding effects of age-related sleep difficulties), (2) have sustained a mild, moderate, or severe TBI in the last 5 years, (3) no longer be on an inpatient basis at the hospital or rehabilitation center, and (4) present with an insomnia syndrome as defined below. Excluded of the study were persons (1) with a major untreated or unstable medical or psychiatric comorbid condition (eg, epilepsy, major depression), (2) who were taking a medication known to produce insomnia, (3) who reported having suffered from significant sleep difficulties before the TBI, (4) who presented evidence of another sleep disorder (eg, sleep apnea or periodic limb movements during sleep), as evaluated by key questions from the Diagnostic Interview for Insomnia27 and by polysomnography when possible, (5) who had significant pain that could be established as the cause of the sleep disturbance (eg, persons who reported being unable to fall asleep because of pain or being awakened by pain), or (6) who were unable to complete the questionnaires due to visual, cognitive, or language comprehension deficits. People using hypnotic or sedative medications were not excluded from the study if they suffered from insomnia despite the use of the medication, if they had been on a stable dosage for 6 months prior to the study, and if the dose remained stable throughout the study.

Sleep difficulties were operationalized as follows. Participants with TBI had to (1) present with a significant sleep complaint (eg, dissatisfaction with sleep or significant distress), (2) report insomnia symptoms defined either as (a) a sleep latency of 30 minutes or more or (b) nocturnal awakenings adding up to 30 minutes or more, (3) present insomnia symptoms at least 3 nights a week, (4) have a complaint of at least 2 negative daytime effects (eg, fatigue, impaired functioning, or mood disturbances) attributed to poor sleep, and (5) have an insomnia duration of at least 1 month. These criteria represent a combination of the diagnostic criteria of the International Classification of Sleep Disorders30 and The Diagnostic and Statistical Manual of Mental Disorders.31

Forty-five TBI patients were referred for the study or contacted us to participate. Of these, 4 participants could not be reached for further information, 21 were excluded after the telephone screening either because they did not meet the inclusion criteria or because they were no longer interested in participating, and 7 decided not to participate after the initial evaluation. Thirteen participants were enrolled but 2 abandoned before starting treatment. Eleven completed the treatment.

Procedure 

We gave potential participants explanations about the goals and implications of the study and completed a brief telephone screening interview to verify the major inclusion criteria pertaining to age, time since injury, presence of major medical or psychiatric conditions, and significance of sleep difficulties.

Once the initial telephone screening interview was completed, we invited participants to a face-to-face clinical interview during which they signed informed consent and were given several self-report measures and sleep diaries. Participants still eligible for the study were asked to fill out a sleep diary during either 3, 5, or 7 weeks (randomly determined). During this baseline period, they were also invited to spend 2 consecutive nights in the sleep laboratory for a polysomnographic evaluation with a standard montage to exclude any co-occurring sleep disorder (ie, restless legs syndrome, sleep apnea). Due to time or transport constraints, only 9 of 11 participants could complete the polysomnographic evaluation. None of these participants showed any evidence of another sleep disorder after this evaluation. All participants were then scheduled to receive 8 weekly sessions of CBT over an 8- to 10-week period, each session lasting approximately 1 hour. Participants were not compensated financially for their involvement in the study but the treatment was offered free of charge and their transportation fees were reimbursed when needed. Once the treatment was completed, a post-treatment evaluation included self-report measures, 2 weeks of sleep diary completion, as well as a second clinical interview conducted over the phone by an independent evaluator. Follow-up evaluations after 1 and 3 months after the end of treatment also included a short telephone interview conducted by an independent evaluator, self-report measures, and 2 weeks of sleep diary completion.

Evaluation Interviews 

Self-Report Measures 

Intervention 

The treatment was administered by a certified psychologist. The intervention was adapted from the protocol presented by Morin27 and has 5 components: (1) the stimulus control instructions42 aim at reassociating the bed, bedroom, and bedtime stimuli with sleep rather than with frustration, anxiety, or tension, (2) the sleep restriction procedure43 consists of limiting the time spent in bed to the actual total time spent sleeping, (3) cognitive therapy of insomnia27 is designed to identify, challenge, and alter a set of dysfunctional beliefs and attitudes about sleep, (4) sleep-hygiene education consists of teaching people about the impact of certain lifestyle habits (eg, diet, drug use, exercise) and the influence of some environmental factors (eg, light, noise, temperature) on sleep,44 and finally (5) the fatigue management skills training component aims at recognizing and managing fatigue more effectively and at revising dysfunctional attitudes about fatigue and rest. Globally, this intervention package is aimed at changing unhealthy habits, regularizing the sleep-wake cycle, and changing dysfunctional thoughts, beliefs, and attitudes about sleep or fatigue.

Main Outcome Measures 

The following outcome variables widely used in the insomnia literature were derived from the sleep diary data: sleep-onset latency, which corresponds to the estimated time taken to fall asleep from lights out; wakefulness after sleep onset (WASO), which includes the total number of minutes estimated to be spent awake during the night; total sleep time, which is the total number of minutes slept from lights out to arising from the bed; sleep efficiency, which is the ratio of total sleep time over the total time spent in bed (from lights out to rising from bed for the last time); and finally total wake time, which includes the time to sleep onset, the time spent awake after sleep onset, and the time between the last awakening and rising from bed.

Design, Data Management, and Statistical Analyses 

A single-case multiple baseline design across subjects was used in this study. The length of the baseline period was determined randomly (3, 5, or 7wk) across subgroups of 3 participants who were enrolled in the study at about the same time.

To examine the efficacy of the intervention in every participant individually, we first used visual inspection of daily sleep diary values of total wake time. Mean changes in sleep efficiency and in the major symptom (either sleep onset latency or WASO) for each participant were also analyzed visually. Visual inspection was completed according to the criteria suggested by Kazdin45: reduction of variability in the daily data, changes in means across study phases, changes in trend, and latency of change.

Although the design of the study did not include a control group, mean comparisons were conducted in order to document the potential efficacy of CBT. Means were computed for each participant for each study phase and were averaged across all participants for different sleep diary variables (total wake time, total sleep time, sleep efficiency) and for scores on the different questionnaires. Paired-samples t tests corrected for multiple comparisons were used to examine averaged data.

To statistically determine whether changes in sleep parameters were due to time and/or treatment, we performed intervention time series analysis (ITSA) using autoregressive residuals. This analysis controls for the serial dependency of the data collected (ie, measures repeated daily) which allows to estimate the magnitude of the effects of treatment with increased power, reliability, and precision.46 The Autoreg procedure of SASa was used to complete the ITSA. Missing data were not included in the analyses. Level and slope effects were estimated as recommended by Huitema and McKean.47 Extreme scores were identified using the auto-regressive integrated moving average outlier procedure and their contribution to error variance was removed by estimating the magnitude of these events. The residuals of the final models were reviewed visually and statistically to ensure that they were normally distributed.

We also evaluated the clinical impact of the intervention at the post-treatment phase, as well as at the 1- and 3-month follow-up assessment phases. A clinically significant treatment response was achieved if (1) the participant presented a sleep efficiency superior to 85%, as computed from at least 2 weeks of sleep diary data, (2) the participant had both a sleep-onset latency and WASO of less than 30 minutes (as computed from at least 2 weeks of sleep diary data), (3) the participant no longer fulfilled the criteria for an insomnia syndrome as evaluated by the Diagnostic Interview for Insomnia by an evaluator independent from the study, or (4) the participant had an ISI score in the subclinical (<15) or no-insomnia range (<8).

Visual Analysis 

Daily sleep diary data for each participant are presented in figure 1. Due to space constraints, we present visually only 1 dependent measure, total wake time, and grouped together participants randomly assigned to the same baseline lengths (3, 5, or 7 weeks).

View Large Image Download to PowerPoint View Large Image Download to PowerPoint View Large Image Download to PowerPoint

Fig 1. Results of daily diary reports of total wake time (in minutes) throughout the study. (Persons with the same baselines [3, 5, or 7 weeks] are presented together.)

Comparisons Across Study Phases 

The average total wake time, sleep efficiency, and total sleep time scores were calculated for all participants in the different study phases (table 2). Paired-samples t tests were used to examine the differences between pre- and post-treatment, pre and 3-month follow-up, and post and 3-month follow-up. A Bonferroni adjustment was applied to control for multiple comparisons (α=.05/3=.017). The t tests revealed a significant reduction in total wake time from pre- (mean, 128.46) to post-treatment (mean, 59.29) (t=5.87, P<.001) and from pre to 3-month follow-up (mean, 49.66) (t=4.47, P<.01). The improvements were well maintained as the difference between post and 3-month follow-up was nonsignificant (t=1.05, P=.06). Sleep efficiency was significantly increased from pre- (mean, 77.20%) to post-treatment (mean, 87.99%) (t=−4.91, P=.01), as from pre to 3-month follow-up (mean, 90.88%) (t=−4.79, P=.01). Again, this improvement was well maintained from post to 3-month follow-up (t=−1.29, P=.24). Total sleep time, on the other hand, was not significantly increased from pre- (mean, 425.68) to post-treatment (mean, 444.43) (t=−.079, P=.44), probably due to sleep restriction procedures, yet it did increase from pre to 3-month follow-up (mean, 496.72) (t=−3.32, P<.015), indicating that further progress was made once treatment was terminated. There was no significant difference from post to 3-month follow-up (t=−2.21, P=.06).

Table 2. Means for Grouped Data

	Dependent Variable	Pre	Post	1-Month Follow-Up	3-Month Follow-Up
	Major sleep indices
	Total wake time	128.46±47.86	59.29±39.54⁎	71.49±42.97	49.66±27.96⁎
	Sleep efficiency	77.20±8.76	87.99±7.99⁎	86.26±7.92	90.88±5.29⁎
	Total sleep time	425.68±51.20	444.43±58.74	453.43±59.48	496.72±50.24⁎
	Questionnaires
	ISI score	17.55±4.03	8.78±3.38⁎	11.33±3.74	10.30±5.38⁎
	MFI
	Total score	63.45±9.25	55.89±12.65⁎	54.40±14.41	52.67±14.05⁎
	General	15.36±2.33	13.11±3.82	13.60±3.81	12.11±3.33⁎
	Physical	13.18±4.35	10.88±4.19	10.80±4.61	9.44±4.09⁎
	Reduced activity	12.45±2.94	12.45±2.94	9.70±3.80	9.89±3.21
	Reduced motivation	8.50±2.63	8.80±2.78	7.66±2.34	8.67±3.31
	Mental	13.45±4.41	11.8±5.00	12.70±4.11	11.89±5.58
	DBAS score	47.40±9.91	32.86±8.48⁎	37.42±7.92	32.10±9.05⁎
	BDI score	16.09±5.92	13.67±7.02†	13.10±5.88	11.90±6.94
	BAI score	13.64±8.64	11.89±6.43	8.30±5.87	9.60±7.73†

NOTE. Values are mean ± standard deviation.

⁎ Significantly different (P<.017) from pretreatment (note that pre vs 1-month follow-up differences were not included in statistical analyses). † Close to significance (P≤.05) compared with pretreatment.

Intervention Time Series Analysis 

The time series analysis was completed only on baseline, and treatment data (including the post-treatment phase) as follow-up phases did not include enough data points. Table 3 presents the results of the time series analysis for total wake time and sleep efficiency for each participant. Nonstandardized regression coefficients (B) in ITSA are interpreted much like coefficients of a linear regression and can be interpreted directly as changes in minutes for total wake time and percentage for sleep efficiency. On average, the statistical modelization of the daily data explained 66.65% of the variance for total wake time and 61.75% for sleep efficiency. Large significant level changes were observed in terms of total wake time for participants 2, 3, 4, 6, 8, 9, 10, and 11 (8 participants) indicating that a rather swift change occurred in the daily observations after the treatment was introduced. Participants 2, 3, 4, 6, and 11 also showed a level change in sleep efficiency. Only participant 7 had no significant level change, although the level effect for total wake time almost reached significance (P=.09) and the regression coefficient of −100.46 was indicative of a large change in total wake time. Participant 5 did not make progress. According to these analyses, participant 1 did not make any significant progress, a result contrasting with that of the visual analysis. It is important to remember that ITSAs allow the identification of extreme scores and that the contribution of extreme scores to error variance is removed during the analysis. This approach differs significantly from the visual analysis approach where extreme scores are very important clinical markers of internight variability and represent nights of very poor quality.

Table 3. Results of Intervention Time Series Analyses

	Participant and Variable	df	Outliers	R2 (%)	Time (B)	Level (B)	Slope (B)
	Total wake time
	1	97	5	86.8	−1.76†	2.09	1.89†
	2	124	4	56.4	−0.73†	−21.96†	0.82†
	3	108	5	65.1	2.39†	−139.89†	−2.74†
	4	122	5	64.7	−0.66⁎	−32.76†	0.54
	5	107	2	56.8	−1.03†	2.15	1.04†
	6	80	2	54.3	3.73†	−150.19†	−3.43†
	7	74	1	61.3	−1.17	−100.46	−0.08
	8	119	5	68.2	0.34†	−15.54†	−0.68†
	9	98	5	76.6	−2.94†	−47.72†	2.80†
	10	133	5	66.5	−0.38	−31.13†	0.04
	11	57	5	76.5	−3.79⁎	−101.85†	4.9†
	Mean				−0.54	−57.93	0.46
	Sleep efficiency
	1	98	5	81.2	0.29†	−1.36	−0.30†
	2	124	5	59.7	0.15⁎	2.66	−0.16†
	3	108	5	66.7	−0.31†	17.03†	0.44†
	4	04	5	58.5	0.10	4.65⁎	−0.07
	5	109	1	40.6	0.11⁎	2.31	−0.15⁎
	6	80	2	43.5	−0.12	12.29†	0.10
	7	73	1	54.9	0.06	7.54	0.25
	8	121	5	65.8	−0.05	2.14	0.12†
	9	100	5	80.8	0.002	14.6†	0.02
	10	134	4	60.3	0.10†	2.01	−0.01
	11	58	5	67.2	0.93⁎	15.89⁎	−1.24†
	Mean				0.11	7.25	−0.09

⁎ P<.05; † P<.01.

Clinical Significance of the Results 

Table 4 illustrates how participants fulfilled different criteria of clinical significance at post-treatment and at the follow-up phases. At post-treatment, 7 of 10 participants for whom we had data had achieved a sleep efficiency of 85% or higher. This proportion increased to 8 of 9 at the 3-month follow-up. In terms of major symptoms, 6 of 10 had decreased either their sleep-onset latency or WASO below the clinical cutoff of 30 minutes at post-treatment and this proportion remained stable after 3 months. All participants who completed the ISI at post-treatment obtained scores in the subclinical range (<15) and 3 of these obtained a score lower than 8 indicating the absence of insomnia. At post-treatment, when evaluated by an independent evaluator, 7 participants no longer fulfilled the criteria for an insomnia syndrome. For at least 6 of these, this remained the case after 3 months, but information is lacking on the remaining participants. Four participants (36%) fulfilled all criteria of clinical significance at evaluations they completed and could thus be considered to have become good sleepers.

Table 4. Clinical Significance of Results at Post-Treatment and 1- and 3-Month Follow-Ups

	Participant	Sleep Efficiency >85%			Sleep Onset Latency and WASO <30 Minutes			Subclinical ISI Score (<15)			No Longer Fulfills Criteria for Insomnia Syndrome
		Post	FU 1	FU 3	Post	FU 1	FU 3	Post	FU 1	FU 3	Post	FU 1	FU 3
	01	Yes	Yes	Yes	Yes	No	Yes	Yes	No	No	Yes	No	ND
	02	Yes	Yes	Yes	Yes	Yes	No	Yes	No	No	Yes	Yes	Yes
	03	Yes	ND	Yes	Yes	ND	Yes	Yes⁎	ND	Yes⁎	Yes	ND	Yes
	04	Yes	No	No	No	Yes	No	Yes	No	Yes	No	Yes	ND
	05	No	No	ND	No	No	ND	ND	Yes	Yes	ND	No	ND
	06	No	Yes	Yes	No	Yes	Yes	Yes⁎	Yes	Yes	Yes	Yes	Yes
	07	No	No	Yes	No	No	No	Yes	Yes⁎	Yes⁎	No	Yes	ND
	08	Yes	Yes	Yes	Yes	Yes	Yes	Yes⁎	Yes	Yes⁎	Yes	Yes	Yes
	09	Yes	ND	Yes	Yes	ND	Yes	Yes	ND	Yes⁎	Yes	ND	Yes
	10	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes
	11	ND	Yes	ND	ND	ND	Yes	ND	Yes	ND	ND	Yes	ND
	Total yes	7	6	8	6	5	7	9	6	7	7	7	6

Abbreviations: FU 1, 1-month follow-up; FU 3, 3-month follow-up; ND, no data.

⁎ ISI score <8 (no insomnia range).

Questionnaires 

The average scores for all questionnaires were calculated for all participants in the different study phases and are presented in the lower part of table 2. Paired-samples t tests were used to examine the differences between pre- and post-treatment and pre and 3-month follow-up, and post and 3-month follow-up. A Bonferroni adjustment was applied to control for multiple comparisons (α=.05/3=.017). A significant reduction was found for the ISI score from pre- to post-treatment (t=4.63, P <.01) and from pre to 3-month follow-up (t=3.43, P<.01). A similar pattern was found for the DBAS score with a significant reduction from pre to post (t=5.22, P<.01) and from pre to 3-month follow-up (t=4.28, P<.01). The MFI total score was also significantly reduced from pre- to post-treatment (t=3.28, P<.012), and from pre to 3-month follow-up (t=3.18, P<.014). In the MFI subscales, the only differences noted were for general fatigue from pre to 3-month follow-up (t=3.18, P<.013), and for physical fatigue again from pre to 3-month follow-up (t=3.00, P=.017). It may be noted that at post-treatment, as a group, participants were approaching the cutoff point of 12 on the MFI, which has been previously used to differentiate fatigued from nonfatigued patients. No significant differences were noted between treatment phases for the BDI or BAI scores, although these scores were all reduced after treatment, compared with pretreatment levels. BDI differences were almost significant from pre- to post-treatment (P=.04), as was the difference from pre to 3-month follow-up for the BAI score (P=.04).

Study Limitations 

Although rigorous single-case research studies can provide valuable information on treatment efficacy, it remains that the generalizability of these results is somewhat limited. The characteristics of the participants treated in the present study are comparable with those of people with TBI in other studies of sleep disturbances after TBI in terms of age and injury severity (mixed levels), but women were slightly overrepresented. Furthermore, people willing to participate in a long treatment research protocol such as this one may be those who are most motivated to change. Nonspecific factors related to therapy (motivation to change, client-therapist relationship, parallel improvements in depression or anxiety) were not accounted for by this research protocol. Future investigations of insomnia in TBI should be conducted with larger samples in randomized and controlled group studies using both subjective and objective (either polysomnography or actigraphy) measures of sleep quality.