Augmenting Usual Care SSRIs with Cognitive Behavioral Therapy for Insomnia to Improve Depression Outcomes in Youth: Design of a Randomized Controlled Efficacy-Effectiveness Trial

Abstract

Importance:

Extant treatments for youth depression are only modestly effective. Alternative approaches are needed to improve health outcomes. A novel approach to improve depression outcomes is suggested by epidemiological studies finding that insomnia often predates and may contribute to depression risk. We test whether treating insomnia among youth starting a new course of SSRI antidepressants improves depression outcomes. This paper describes our study design.

Design:

2-arm randomized controlled efficacy-effectiveness trial.

Setting:

a large non-profit health maintenance organization.

Participants:

165 adolescents aged 12 to 19 with research-confirmed depression and insomnia diagnoses, starting a new episode of selective serotonin reuptake inhibitor (SSRI) antidepressant treatment prescribed by their usual care provider.

Interventions:

Two sleep interventions, each 6-7 sessions, both overlaying “treatment as usual” (TAU) SSRIs: a sleep hygiene (SH) attention control condition, and cognitive-behavioral therapy for insomnia (CBT-I).

Conclusions and Relevance:

If CBT-I improved sleep is shown to improve depression-related outcomes, this may provide an additional, easily tolerated intervention for an important public health target.

Trial Registration:

clinicaltrials.gov, NCT02290496, https://clinicaltrials.gov/ct2/show/NCT02290496.

Unipolar depression is a major public health issue among youth; more than 13% of adolescents age 12-17 have experienced at least one episode of major depression.^1,2 An estimated 75% of youth with previous depression will experience a second episode within 5 years.^3-5 Depression is associated with substantial impairment in school and interpersonal relationships, substance abuse, and an increased risk of attempted or completed suicide.^6-9 Many cases of recurrent depression in adulthood have their first onset in adolescence,^10,11 making this an important developmental period in which to intervene. Unfortunately, commonly available depression treatments such as antidepressants and psychotherapies including cognitive behavioral therapy (CBT) are only modestly effective, especially in real-world complex cases.^12-15 There is significant room for treatment improvement.

Insomnia is an independent risk factor for incident depression, ^16-21 and is present in 70% or more of depressed adolescents⁴, making it a potential target for treatment to improve depression outcomes. Patients with insomnia have a 2- to 3.5-fold increased risk for future depression relative to those without insomnia,²² a pattern evident across the lifespan.^20,23-30 Insomnia independently predicts suicidal behavior in depressed adults³¹ and similar results are reported in teens.^32,33 Youth who report frequent trouble sleeping are more likely than normal-sleeping controls to report anxiety or depression.³⁴ Lastly, insomnia in teens predicts future depression³⁵ but depression does not predict subsequent insomnia³⁶, suggesting a potentially causal and modifiable risk mechanism by which insomnia affects depression.

Beyond the associations described above, insomnia may complicate depression treatment response. Depressed adults with sleep disorder respond less well to depression CBT.³⁷ Similarly, depressed youth with insomnia were less responsive to antidepressants than those without insomnia. ³⁸ In youth, low sleep efficiency and delayed sleep onset both predict the presence of depression following treatment.^39-41

To test whether insomnia treatment improves depression outcomes, we conducted an RCT in adolescents with research-confirmed diagnoses of both major depression and insomnia. Youth just starting treatment as usual (TAU) SSRI antidepressants, prescribed and managed by their usual care providers, were randomized to one of two interventions to augment pharmacotherapy: (a) cognitive behavioral therapy for insomnia (CBT-I) or (b) an active control condition of credible but minimally effective sleep hygiene (SH). Both CBT-I and SH were developmentally adapted from approaches used in several adult insomnia studies.^42-44 We hypothesize that CBT-I + SSRIs will be superior to SH + SSRIs for both depression and sleep outcomes.

This RCT addresses a significant evidence gap by testing the efficacy and effectiveness of a CBT intervention to address sleep dysfunction in youth and evaluating the impact on depression outcomes. CBT-I has few or no side effects and could positively alter a broad range of emotional and physical health domains in vulnerable youth. Moreover, the focus of this CBT-I intervention on behavior change holds potential to assist youth in establishing healthy patterns that will serve them well into the future. This paper presents a detailed description of our methods, including the setting, participants, outcomes of interest, instruments, and analysis plan for this trial.

Methods

Setting

This trial was conducted in Kaiser Permanente Northwest, a large, non-profit health maintenance organization serving more than 620,000 members in the Pacific Northwest region of the United States. Every member of the health plan has a unique, permanent health record number. Every encounter with the medical care system, referrals to outside services, lab results, treatments, and diagnoses are recorded in a comprehensive electronic health record (EHR) under the patient’s health record number. Dispenses from integrated pharmacies are also recorded, including medication, dose, supply, and target diagnosis/condition. This EHR system stores information such as patient demographic characteristics, medical history, and visit summaries. All study participants were covered by the health plan at the time of study enrollment. The study was reviewed and approved by the Kaiser Permanente Northwest Internal Review Board (IRB).

Participants

Participants were adolescents (12-19 years) just starting a new usual care course of SSRI antidepressants for depression, and who had research-confirmed active insomnia and major depressive disorder (MDD). At the baseline interview, participants were required to have a subjective complaint of insomnia for at least one month (ascertained via the Duke Structured Interview for Sleep Disorders [DSISD]^45,46 and a score ≥ 9 on the Insomnia Severity Index [ISI]⁴⁷). The MDD inclusion criteria was confirmed at baseline via the mood disorders module of the Children’s Schedule for Affective Disorders and Schizophrenia (K-SADS-PL)⁴⁸, a semi-structured psychiatric interview. The new antidepressant treatment episode was defined as an SSRI dispense recorded in the health plan electronic health record (EHR) with no other SSRI or other antidepressant dispenses in the prior 6 months.

Individuals were excluded from the study if they (1) had an active, progressive physical illness (e.g., cancer) or neurological disease (e.g., multiple sclerosis) that might impact the onset/treatment of insomnia; (2) had a clinical diagnosis of a sleep condition that was not typically treated by conventional CBT-I, such as sleep apnea, restless legs, periodic limb movement disorder, or delayed sleep phase syndrome; (3) had a major pervasive developmental disorder that might impair participant ability to take part in CBT-I or SH (e.g., mental retardation, severe autism spectrum disorder); (4) were regularly taking medications known to alter sleep (e.g., steroids, high-dose antihistamines) or over-the-counter sleep medications (e.g., melatonin, hypnotics, herbs); or (5) had received six or more sessions of cognitive behavioral therapy (CBT) for insomnia or depression in the past year.

Recruitment

From March 2015 to May 2017 we regularly ran a case-finding algorithm to search the EHR for youth with depression diagnoses and new SSRI dispenses. Before attempting to recruit the possibly eligible youth identified by this search, we first sent their primary care providers (PCPs) a summary of the study and notified them of our intent to recruit their patient. Providers could ask that we not recruit individuals for any reason (e.g., SSRIs prescribed for pain rather than depression). If providers did not decline within three days, study staff mailed EHR-identified youth a recruitment packet containing a copy of the consent form and a brochure summarizing study information. Individuals could respond to the recruitment materials by contacting the study team directly; otherwise, study staff began recruitment calling five days after materials were mailed.

Interested individuals were screened by phone for depression and insomnia symptoms. Youth meeting preliminary criteria (high levels of self-reported depression and insomnia symptoms; current Kaiser member; still taking prescribed SSRI; available to attend weekly in-person intervention sessions) were scheduled for a baseline assessment to fully determine eligibility. All minors provided assent and were required to have a parent consent, while young adults aged 18 or older consented for themselves.

Randomization

The trigger event to start recruitment was the dispense of the new SSRI medication with no antidepressants dispensed in the previous 6 months; all recruitment and randomization was completed within 21 days of this index date. This delay was necessary for the study to recruit, consent, and conduct baseline assessments with youth to determine study eligibility. Youth had to be taking the prescribed SSRI at the time of study enrollment; if they discontinued the index SSRI they were not enrolled. Because antidepressants require several weeks of use for therapeutic benefit,⁴⁹ the delay between SSRI initiation and collection of baseline measures at randomization was expected to be insufficient for SSRIs to yield much depression improvement.

Participants were randomized 1:1 to one of two behavioral sleep interventions: Sleep Hygiene or CBT-Insomnia (described in more detail below), with non-research treatment as usual (TAU) underlying both conditions. We stratified randomization by depression severity (CDRS-R < 56 or >56). In order to balance the number of study participants randomized by strata and to ensure study staff would not be able to anticipate random allocation, we used permuted blocks within strata (randomized block sizes of 2 or 4, masked from clinical staff).

Study Design

We employed a blended efficacy-effectiveness study design to compare the control (SH) and active (CBT-I) insomnia treatments. Efficacy elements included detailed assessment of outcomes and possible mediators of depression improvement, and tight controls on intervention quality. Effectiveness elements included a clinically complex, real-world sample of patients actively seeking non-research depression treatment, and examining impacts on non-research healthcare. The SH intervention mainly emphasized good sleep habits, while the CBT-I intervention focused on improving sleep by addressing both cognitive and behavioral factors (e.g., discussing/addressing negative beliefs about sleep/insomnia, altering sleep routine). Both participants and assessment staff were blinded to study condition.

Major follow-up assessments occurred at 12-, 26- and 52-weeks and included actigraphy data collection, sleep diary completion, and an interview with several measures (described below). Two brief assessments were also conducted at 4- and 8-weeks to measure potential mediators. The primary mood outcomes are depression response (partial but not complete improvement) based on the Clinical Global Impression-Improvement (CGI-I) score, and time to remission (end of the depression episode and return to near-normal functioning) based on the KSADS-LIFE. The primary insomnia outcomes are total sleep time (from sleep diary and actigraphy measures) and Insomnia Severity Index (ISI) score.

Intervention Arms

Neither CBT-I nor SH focused on depression directly. Both arms were overlaying SSRI antidepressants prescribed by a health plan provider, as well as any other TAU services received. Study therapists had both SH and CBT-I cases in their treatment panels and were supervised weekly with a focus on SH and CBT-I implementation. There were distinct, separate therapist manuals and youth workbooks for each intervention arm (see online supplementary materials to view manuals and workbooks). CBT-I intervention content was not included in the SH arm, but an abbreviated version of the SH content was included in the CBT-I condition, as SH is routinely a part of this approach. Supervision led by Dr. Clarke allowed therapists to discuss individual cases and explore with the intervention team how best to deliver the components of each arm of the intervention while avoiding bleed-through of skills across cases in the two arms. This was achieved through systematic review of the content of each session in each arm of the study and regular supervision discussion of what could or could not be included per the therapy manuals, most often as it related to SH cases. Study therapists did not formally promote medication adherence, but they did communicate that the sleep-focused study interventions were not a substitute for depression medication. Participants were encouraged to communicate with their TAU provider if they wished to change or discontinue their medication.

Both interventions were delivered in-person across 6-7 sessions, as needed to complete content and skill practice. They were designed to be equal in duration across the CBT-I and SH arms to control for therapeutic relationship and contact time. Intervention sessions were ideally weekly but were required to be completed within 12 weeks of randomization, before the first major follow-up time point. Therapists reviewed participant sleep diary entries at each session, using this data to discuss with participants how the elements of SH/CBT-I were successfully implemented and to explore opportunities for improvement. Motivational Interviewing strategies⁵⁰ were typically implemented when participants failed to regularly complete the sleep diary or engage in SH/CBT-I techniques. Table 1 presents an outline of intervention content by study session and condition.

Table 1.

Intervention content by session number and study condition

Session	Both Interventions	SH Specific Content	CBT-I Specific Content
1	Information gathering and orientation to SleepWell program Confidentiality Sleep history and assessment Review sleep diary entries Workbook: Barriers to Sleep Psychoeducation about sleep Personal goals for treatment How to get most out of program (motivational interviewing) Practice/homework for next session	N/A	Basics of Sleep Hygiene (winding down, rapid wake-up)
2	Review sleep diary and homework Practice/homework for next session	Introduce winding down (establishing a bedtime routine) Improving bedroom environment	Introduce sleep plan Introduce stimulus control (electronic curfew, nesting place, regular wakeup time, promoting daytime activity)
3	Review sleep diary and homework Practice/homework for next session Optional: Problem-solving Optional:Parent check-in	Winding down, part two (limiting exposure to technology)	Review sleep plan Introduce Cognitive therapy for sleep Savoring
4	Review sleep diary and homework Practice/homework for next session	Introduce rapid wake-up	Introduce sleep restriction and review implementation guidelines (if appropriate) Cognitive therapy (introduce 5 column form)
5	Review sleep diary and homework Practice/homework for next session	Relaxation and nutrition for better sleep	Sleep restriction/revise sleep plan as necessary Cognitive therapy for sleep (continued practice) Adopting beliefs/attitude that contribute to good sleep
6	Review sleep diary and homework	Refine wind down and wake up plans	Refine sleep plan Cognitive therapy for sleep (continued practice)
7	Review sleep diary and homework (or in-session practice) Review goals and progress Relapse prevention	N/A	Refine sleep plan
Optional Content	Optional: Fear of Missing Something (FOMS; sessions 1-3) Problem-solving(sessions 2-7) Parent check-in (sessions 1-7) Relapse prevention (session 6)	Lifestyle factors (session 6)

Sleep Hygiene

The Sleep Hygiene (SH) intervention focused on establishing a relaxing wind-down routine; improving the bedroom environment; avoiding caffeine, nicotine or eating close to bedtime; and an active wake up plan to improve daytime energy. Optional components—content not required to be delivered during the intervention—included breathing exercises to reduce stress and eating healthy for better sleep. Therapists were encouraged to limit use of formal problem-solving methods to addressing implementation of SH strategies. The final session focused on standard relapse prevention techniques, reviewed progress toward goals for treatment, and established a plan for how participants would continue to use the newly acquired SH skills in the future.

CBT-Insomnia

The initial CBT-I session addressed psychoeducation about sleep, consistent with the content from the SH intervention. This included information about the ideal number of hours of sleep for adolescents, description of biological influences, and circadian rhythms. The active elements of CBT-I consisted of three core components: Stimulus Control, Sleep Restriction, and Cognitive Therapy.

Stimulus Control (SC).

SC was introduced in an early session to establish a regular sleep-wake cycle and strengthen the association between bed and sleep by limiting incompatible behaviors in bed (homework, TV, phone/computer use). Regular bed and wake times were set using sleep diary data and borrowed SH strategies of establishing relaxing bedtime and positive wake up routines. Therapists coached participants on reducing use of electronics within 30-60 minutes of bedtime, going to bed only when feeling tired, going to a calming and comfortable “nesting place” if unable to sleep after 15-20 minutes, and resisting afternoon napping while also promoting daytime activity.

Sleep Restriction (SR).

To determine whether SR should be introduced, the therapist and participant reviewed the previous week’s sleep diary. If sleep efficiency (total sleep time/total time in bed) was below 85% for a majority of nights, therapists explored implementing SR. Therapists worked with participants and sleep diary data to identify a maximum amount of time allowed in bed from bedtime to wake time, and set bedtimes and wake times, with a minimum of 6.5 hours of time in bed. Bedtimes were often determined by pre-identified wake times necessary for teens in school, or employment for older youth. Progress was evaluated each week with the intention of advancing the bedtime earlier by 15-30 minutes each week if participants reported improved sleep efficiency, ultimately working toward 8-9 hours in bed per night with high sleep efficiency (SE>= 85% in order to increase goal time in bed each week). Review of the sleep diary was an important tool in monitoring successful implantation of SR.

Cognitive Therapy (CT).

CT was introduced early on (approximately session 3) to all CBT-I participants and carried forward through the duration of the intervention period. CT was used to address dysfunctional beliefs about sleep and bedtime rumination that could be interfering with a participant’s ability to fall or stay asleep. CT involved 1) identifying negative or otherwise unhelpful beliefs about sleep, 2) learning to challenge those unhelpful beliefs and 3) creating more positive and realistic thoughts, with the goal of eliciting a more helpful emotional response to activating events or trigger situations. While the intention was to apply CT strategies to thoughts about sleep, therapists were permitted to employ CT to address depression or anxiety-related thoughts that were identified by participants, but only if these beliefs were interfering with sleep. Therapists reviewed progress made toward participants’ goals for treatment in each session. As with the SH intervention, the final session of the CBT-I intervention focused on relapse prevention, with therapists working with participants to develop a plan for using the CBT-I skills in the future.

Parent Involvement

Adult participants aged 18 and 19 were not required to have parent consent or involvement, although this was encouraged. Parent therapy involvement was therefore not a required element of either the SH or the CBT-I intervention. However, younger participants—especially those who relied on a parent to bring them to appointments—were strongly encouraged to include their parents in the last 5-10 minutes of each counseling session if they were available. The rapists had flexibility regarding parental involvement based on youth preference and parent availability. Many minor participants took public transportation or drove themselves and thus arrived for counseling sessions without a parent, limiting the amount of parental involvement possible. When parents were included, the therapist and youth used the time to review general concepts covered in the session, changes to the sleep plan, and any homework assignments for the week. When parents were available for youth in the CBT-I arm, it was important to get their support for both stimulus control and sleep restriction so that parents would not unintentionally undermine therapy plans. For example, therapists often explained the rationale behind sleep restriction to parents so that they had a clear understanding that setting a later bedtime to consolidate sleep periods was temporary, that there was no sleep deprivation but only sleep consolidation, and that the goal was to adjust the bedtime earlier each week to eventually arrive at a more conventional bedtime. In addition, parents often provided valuable insights into potential barriers to implementing the sleep plan or other elements of the intervention, allowing the therapist to make any necessary adjustments.

Assessments and Measures

Participants were asked to complete major assessments at baseline (week 0), and at 12, 26, and 52 weeks after baseline (all follow-up assessments took place after the intervention had concluded). Each of the major assessments included a period of sleep diary completion, actigraphy data collection, and an interview. Baseline interviews took place either in-person or by phone, based on participant preference and availability. Follow-up interviews were completed by phone, with participants having the option to complete some self-report measures online before the phone interview (seeTable 2). Parents of minor-aged youth were also asked to provide assessment data for some of the measures at the same time points (additional detail provided below).

Table 2.

Measures collected by study time point

	Respondent	Baseline	Active Treatment (target of 6-7 sessions over 12 weeks)											Post-Tx	Follow- Up
Weeks		0	1	2	3	4	5	6	7	8	9	10	11	12	26	52
Primary Outcomes
Clinical Global Impressions Improvement (CGI-I) and Clinical Global Impressions Severity (CGI-S)	I	x												x	x	x
Children’s Schedule for Affective Disorder & Schizophrenia (KSADS) and Longitudinal Interval Follow-up Evaluation (LIFE)	IYP	x												x	x	x
Total Sleep Time (sleep diarya and actigraphy)	YA		x	x										x	x	x
Insomnia Severity Index (ISI)b	YP	x												x	x	x
Secondary Outcomes
Children’s Depression Rating Scale-Revised (CDRS-R)	IYP	x												x	x	x
Patient Health Questionnaire – Depression (PHQ-9)c	IYP	x				x				x				x	x	x
Generalized Anxiety Disorder Scale (GAD-7)b	YP	x												x	x	x
Duke Structured Interview for Sleep Disorders (DSISD)	IYP	x												x	x	x
Other Measures
Anxiety and Preoccupation about Sleep Questionnaire (APSQ)b	Y	x				x				x				x	x	x
Brief Medication Questionnaire (BMQ), Regimen subscaleb	Y	x				x				x				x	x	x
Children’s Emotion Management Scale – Sadness (CEMS-S), Anger (CEMS-A) b	Y	x				x				x				x	x	x
Children’s Global Adjustment Scale (CGAS)	I	x												x	x	x
Credibility-Expectancy Questionnaire (CEQ)b	Y	x
Demographics	IP	x
Dysfunctional Beliefs and Attitudes about Sleep (DBAS-C10)b	Y	x				x				x				x	x	x
EQ-5D Derived Quality Adjusted Life Years (QALYS) b	YP	x												x	x	x
Instrumental Variables Questionnaire(IVQ)b	YP	x												x
Post-Treatment Evaluation Questionnaire (PTEQ)b	YP													x
RAND-36 Fatigue scaleb	Y	x				x				x				x	x	x
Sleep Behavior Self-Rating Scale (SBSRS) Abbreviated b	Y	x				x				x				x	x	x
Youth Social Adjustment Scale – Self-Report (YSAS-SR) b	YP	x												x	x	x
EHR Data Extraction (utilization and pharmacy dispense data)	-															x
Youth therapy comprehension and mastery, attendance, homework completion, contacttime, parent involvement (rated by therapists)	T		x	x	x	x	x	x	x	x	x	x	x	x

I=Interviewer; Y=Youth; P=Parent; T=Therapist; A=ActiGraph device; YP=Youth and parent both rate youth; IP=interviewer with input from parent; IYP=Interviewer with input from both youth and parent

^aSleep diary data collected during the first 12 weeks.

^bParticipants given the option to complete this measure online for follow-up assessments.

^cThe PHQ-9 was interviewer-administered for baseline and the 12-, 26-, and 52-week assessments. The PHQ-8 was administered for the 4- and 8-week assessments, and participants were given the option to self-report at these two time points.

Incentives

Youth received Amazon gift certificates for (a) each assessment ($20 for baseline; $10 for 4-week; $10 for 8-week; $25 for 12-week; $30 for 26-week; $35 for 52-week); (b) online sleep diary completion ($1 per day of data entered); and (c) actigraphy data ($1 per day of data collected). Participating parents also received gift certificates for completing planned assessments ($10 for baseline; $10 for 12-week; $15 for 26-week; $20 for 52-week).

Sleep Diary

Youth were asked to complete a daily online sleep diary during active sleep treatment (i.e., the first 12 weeks after enrollment), and for a two-week period at each follow-up assessment. The sleep diary questions asked youth what time they got into bed, what time they tried to go to sleep, how long it took to fall asleep, how many times they woke up, how long awakenings lasted, the time of their final awakening, whether they tried to fall back to sleep, what time they got out of bed for the day, and whether they woke up earlier than planned (if so, how much earlier). There were also questions related to quality of sleep and how rested/refreshed participants felt upon waking.

If the online sleep diary entry was not completed by 4:00pm each afternoon, participants received a text message reminder that included a link to the sleep diary website. Youth could answer sleep diary questions for ”last night” and the “the night before last” if the latter entry had not been submitted the previous day. Youth were not allowed to submit sleep diary entries going farther back than two nights due to concerns about the accuracy of recall. We will use the sleep diary data to calculate Total Sleep Time (TST), Sleep Onset Latency (SOL), and Sleep Efficiency (SE).

Actigraphy

The ActiGraph GT9X Link (wrist worn device) was used to collect sleep actigraphy data. Devices were programmed to collect data at a rate of 30Hz. To preserve battery life, the idle sleep mode feature was enabled while the wireless options and inertial measurement unit (IMU) settings were disabled. The 12-hr time display option was enabled, but participants did not receive any direct feedback (e.g., step count) from the device. The only initialization parameter used was the participant name field, in which the participant’s study identification number was entered.

Actigraphy data was scored using ActiLife version 6 software. Data was initially downloaded into 10 second epochs, and these data files were converted to 60 second epochs for sleep scoring. Data files were scored by selecting the Sadeh algorithm⁵¹ and the ActiGraph sleep period detection option.

While actigraphy data are typically scored by applying the proper software algorithm and then adjusting based on self-report sleep diary data, we were unable to uniformly apply this approach due to a large amount of missing sleep diary data. Instead, algorithm errors were identified and manually corrected by trained staff based on a scoring manual that was created for this study (see Manual Scoring Procedures for ActiGraph Data in the online supplement). For example, sleep periods detected by the algorithm that contained non-wear were adjusted to remove the non-wear from the sleep period. If cases arose in which the manual did not clearly address whether observed data contained an error, nor provided enough guidance on how to correct an identified error, the case was brought to a consensus meeting and discussed by all three data scorers. Sleep diary entries were not used to inform these decisions because this data was not consistently available. Sleep metrics calculated included sleep efficiency, total minutes in bed, total sleep time, wake after sleep onset, number of awakenings, and average awakening length.

Primary Outcome Measures

Clinical Global Impressions Severity (CGI-S) and CGI Improvement (CGI-I).^52,53

We administered the CGI-S and CGI-I, which are brief, single-item measures of a youth’s depressive disorder. The CGI-S measures the severity of depressive illness on a seven-point scale ranging from 1 (normal, not at all ill) to 7 (among the most extremely ill). The CGI-I is a complementary measure that assesses improvement in depression symptoms after the initiation of treatment and is similarly scored on a scale from 1 (very much improved) to 7 (very much worse). For the CGI-I, scores ≤2 (representing much or very much improved) are considered clinically meaningful. Both measures are scored by the interviewer using all available data from each full assessment.

Children’s Schedule for Affective Disorder & Schizophrenia (KSADS) and Longitudinal Interval Follow-up Evaluation (LIFE).

At each assessment wave interviewers administered the KSADS-PL mood disorder module to generate DSM diagnoses of major depressive disorder (MDD) at baseline and each follow-up time point.⁴⁸ Information from the KSADS was employed to obtain weekly Psychiatric Status Ratings (PSRs) for MDD over the follow-up period, using the Longitudinal Interval Follow-Up Evaluation (LIFE) method.⁵⁴ This method provides a continuous assessment of symptoms and onset and offset of MDD since the last assessment. A score from 1 through 6 on the PSR scale is given for each week of the follow-up period. Scores of 1 to 2 indicate 0, 1, or 2 symptoms with no or mild impairment; a score of 3 reflects at least 3 symptoms with mild to moderate impairment; a score of 4 indicates at least 4 symptoms and mild to moderate impairment; and scores of 5 and 6 indicate the person meets definite MDD criteria. LIFE and KSADS data will be employed to determine the study primary outcome of depression remission as defined by a recent consensus task force:⁵⁵ MDD remission is defined as 3 or more weeks of absence of both the two core depression symptoms of sadness and reduced interest/pleasure, and 3 or fewer of the remaining 7 MDD symptoms (suicidal ideation or behavior; insomnia or hypersomnia; fatigue/lack of energy; difficulty concentrating; change in appetite; psychomotor retardation/agitation; feelings of guilt/worthlessness).

Total Sleep Time (TST).

TST is defined as actual time slept, not including time attempting to fall asleep, middle of the night awakenings, and time awake while in bed at the end of the sleep period. TST will be calculated from both actigraphy and sleep diary data.

Insomnia Severity Index (ISI).

The 7-item Insomnia Severity Index (ISI)⁵⁶ measures subjective insomnia over the last month and has good psychometric characteristics in youth.⁵⁷The seven items encompass severity of insomnia (difficulty falling/staying asleep and waking), satisfaction with current sleep pattern, interference of sleep problem with daily functioning, noticeability to others, and worry caused by the sleep problems. Items are scored on a 5-point Likert scale from 0 (no problems) to 4 (very severe); higher scores indicate greater insomnia severity. A score of ≥9 optimally identifies clinically significant insomnia in teens.⁵⁸

Secondary Outcome Measures

Children’s Depression Rating Scale-Revised (CDRS-R).

The interviewer-rated 17-item CDRS-R generates a continuous measure of unipolar depression symptomatology, and has good psychometric properties in youth age 7-18.^59-61 Due to similarity with questions on the KSADS, CDRS-R questions were embedded within the KSADS to facilitate gathering sufficient information for interviewer ratings. All items are rated on scales of 1 to 5 or 1 to 7, with higher numbers indicating more severe symptomatology. Because most interviews were conducted by telephone, interviewers were often unable to rate the last 3 items (depressed facial affect, hypoactivity, listless speech) as they required direct observation. To maximize retrospective and prospective comparability of our results with other studies, we will provide alternate CDRS-R scoring results when analyzing the data: presenting CDRS-R totals with, and without, the last 3 items. For both approaches, higher total scores indicate greater severity of depression.

Patient Health Questionnaire (PHQ-9).

This 9-item self-report depression scale has excellent psychometric properties in youth and young adults.^62-64 Each of the items is scored on a scale of 0 (not at all) to 3 (nearly every day), with higher scores indicating greater depression symptomology. The main scoring is the continuous PHQ-9 total, calculated as the sum across all items. Additional PHQ-9 outcomes include classifications of the total score as moderate or higher (score≥10) and moderately severe or higher (≥15) depression symptomatology.⁶⁵ This measure was interviewer-administered so that appropriate risk evaluation could be conducted immediately if the ninth item (suicide ideation) was endorsed.

Generalized Anxiety Disorder Scale (GAD-7).

This 7-item anxiety measure has good psychometric properties.^66-68 Each item is rated on a scale of 0 (not at all) to 3 (nearly every day), with higher scores indicating greater anxiety symptomology. A continuous total score is calculated by summing the seven items. Total scores ≥10 indicate moderate or higher anxiety, and scores ≥15 indicate severe anxiety.⁶⁵

Duke Structured Interview for Sleep Disorders (DSISD).

The insomnia module of the DSISD^45,46 was administered by interviewers to identity past and present insomnia. Participants were asked 1) if they currently had difficulty getting to sleep, staying asleep, waking too early, or feeling poorly rested, 2) if the sleep difficulty caused daytime problems, such as fatigue, 3) how long the sleep problems have occurred, and 4) if the sleep difficulty occurs even if they give themselves enough time to sleep. Each item was interviewer rated as 1 (symptom absent), 2 (subthreshold symptomology), or 3 (threshold symptomology). Participants were categorized as having insomnia (yes/no) if they met threshold for all 4 items.⁶⁹

Other Measures

Additional measures were administered to collect data on demographics, potential mediator and moderator variables, and participant perceptions and experiences regarding the interventions. These measures are described in more detail below.

Anxiety and Preoccupation about Sleep Questionnaire (APSQ).⁷⁰

The APSQ is a 10-item psychometrically sound instrument for assessing sleep-related worry in the past month.⁷⁰ Participants rate each item on a scale of 1 (not true)to 10 (very true), with higher scores indicating greater worry. We calculated a total score as the sum of all items and calculated two established factor scores: a six-item factor that summarizes participants’ worries about the consequences of poor sleep and a four-item factor that assesses participants’ worries about the uncontrollability of sleep.

Brief Medication Questionnaire (BMQ)–Regimen Subscale.⁷¹

We asked four questions from the BMQ Regimen subscale to assess antidepressant use: 1) “Are you still taking the antidepressant medication prescribed to you?”, 2) “In the PAST WEEK, how many days did you take this medication?”, 3) “In the PAST WEEK, how many pills did you usually take each time?”, and 4) “In the PAST WEEK, how well did this medication work for you? (not at all, moderately well, very well, don’t know).”

Children’s Emotion Management Scales (CEMS).⁷²

Three items from the CEMS Anger subscale and three items from the CEMS Sadness subscale were measured as potential mediators, to examine whether improved sleep led to improved emotional regulation, and thus to improved depression.^73-75 Items are rated on a 3-point Likert scale (hardly ever, sometimes, and often). These six items comprised the dysregulated expression dimension of the CEMS scale.

Children’s Global Adjustment Scale (CGAS).⁷⁶

The CGAS is an interviewer-rated measure of youth global functioning, considering psychological, social, and occupational functioning, but excluding impairment due to physical or environmental limitations. It is scored on a scale of 1-100, with lower scores indicating greater impairment. Scores above 70 are considered to indicate normal functioning, while scores < 60 typically indicate the need for treatment.

Credibility-Expectancy Questionnaire (CEQ) - Adapted.⁷⁷

The CEQ is a measure of treatment credibility (i.e., how believable/convincing participants view the CBT-I and SH treatments to be) and expectancy (i.e., how much improvement participants believe will occur). The purpose of the CEQ was to confirm that participants viewed both the CBT-I and SH interventions as equally credible and effective. Two sets of questions were administered to participants at baseline. In set I, participants were asked six questions about what they think will happen regarding the treatment effectiveness in reducing their insomnia and depression symptoms. In set II, participants were asked four questions about what they feel will happen. The original CEQ focused on trauma symptoms; we modified it to focus on depression and insomnia symptoms.⁷⁸ Higher scores indicate greater perceived credibility or higher expectations of positive outcomes.

Demographics.

At baseline, youth were asked about their age, gender, and race/ethnicity. Participating parents reported their highest educational attainment, marital status, and income bracket. If no parent was participating, young adults were asked to report on parent education, marital status, and income bracket.

Dysfunctional Beliefs and Attitudes about Sleep-Short version for children (DBAS-C10).

We administered the 10-item version of the original DBAS-16, adapted for youth to assess beliefs and attitudes about sleep.⁷⁹ Each item is rated on a scale of 1 (strongly disagree) to 5 (strongly agree) and items are grouped into three factors: (1) beliefs about the immediate negative consequences of insomnia; (2) beliefs about the long-term negative consequences of insomnia; and (3) need to control the insomnia. Higher scores indicate more dysfunctional beliefs about sleep. The DBAS-C10 demonstrates high internal consistency⁷⁹ and is associated with sleep disturbance and depression.⁸⁰

EuroQol (EQ-5D).

This general health-related quality of life measure asked participants to rate their health state in the domains of mobility, self-care, usual activity, pain/discomfort, and anxiety/depression. Responses were transformed into a utility score, which could range from −0.11 (worse than death) to +1.00 (perfect health).^81,82

Post-Treatment Evaluation Questionnaire (PTEQ).

The PTEQ assessed how acceptable, useful, and logical participants found the therapy sessions in improving their insomnia and depression (two separate sets of questions), and whether they still suffer from depression or insomnia. It also assessed how satisfied participants were with the treatment, how likely they were to follow recommendations, and how likely they were to recommend the therapy to friends who experience similar problems. Items were rated on a scale of 1 (not at all useful) to 10 (very useful). Lastly, participants were asked to rate their insomnia and depression improvement separately on a scale of 0% to 100%.

RAND-36 Energy/Fatigue Scale.

We administered the 4-item energy/fatigue subscale of the RAND-36 instrument that assesses health-related quality of life in several domains.⁸³ Each item was rated on a scale of 1 (all of the time) to 6 (none of the time). A summary score is calculated by rescaling each item to a 0 to 100 scale⁸⁴and averaging these four items; higher scores indicate lower energy/fatigue.

Sleep Behavior Self-Rating Scale (SBSRS)–Abbreviated⁸⁵

We measured the frequency of nighttime sleep incompatible behaviors with a modified 13-item version of the SBSRS, with daytime behavior items removed from the original 20-item scale. This measure assesses the frequency of sleep incompatible behaviors (e.g., “Watching television around sleep time”). Items are rated on a scale of 1 (never) to 5 (very often). Higher scores indicate a higher level of sleep incompatible behaviors. The SBRS has high test-retest reliability and internal consistency.

Youth Social Adjustment Scale–Self-Report (YSAS-SR).

Youth completed a 23-item version of the YSAS-SR, a measure of the youth’s adjustment in school behavior, relations with friends, spare time, family behavior, and dating.⁸⁶

Parent Assessments

Participating parents were asked to complete assessments at baseline (week 0), and 12, 26, and 52 weeks after baseline. Parent assessments comprised a subset of the measures that youth were asked to complete (see Table 2). In addition, they were asked to report on the youth (rather than themselves). For some of the measures (e.g., KSADS, CDRS-R), parent report is built into the administration of the measure. For other instruments, the instructions and/or questions were adapted slightly for parent reporting purposes. This usually consisted of replacing “you” with “your child”. For example, the PHQ-9 instructions that state “…how often have you been bothered…” were changed to read “…how often has your child been bothered…”.

Electronic Health Record (EHR) Data Extraction

We will extract all health care utilization for each study participant, including outpatient, inpatient, and remote (e.g., telephone, email) encounters as well as pharmacy dispenses. We will calculate medical expenditures using comprehensive estimates of health services costs from the EHR and other electronic administrative data during the 12 months prior to and following randomization. These data, used in previous studies, accurately represent services paid for by the health plan.⁸⁷ Expenditures will include outpatient, inpatient, and pharmacy categories, and we will adjust all values to constant year US dollars using the consumer price index multiplier from the medical-services category.⁸⁸ We will calculate expenditures accrued during the years prior to and following study entry.

Instrumental Variable (IV) Methods

We will examine the dose-response relationship between intervention uptake and study outcomes as part of secondary analyses using instrumental variable (IV) methods, which require an “instrument:” a measure that is unrelated to confounders for which we are trying to control but is predictive of the number of CBT-I or SH sessions attended (dose).^89,90 We assessed study participants at study entry and at the end of the acute treatment phase about potential instruments,⁹¹ including travel disruptions (e.g., traffic jams) and travel burden (e.g., total travel time). We asked participants to report how frequently they anticipated having difficulties traveling to study sessions at baseline, as well as their actual experiences at the 12-week assessment. IV analytical methods are described in more detail below.

Therapy Process Data

At the end of each intervention session, therapists completed a process checklist form that captured date, number, location and duration of the session; homework completion; and any topics/content covered. After the final intervention session, therapists also rated the participant’s comprehension, implementation, and mastery on a 5-point Likert-type scale ranging from very poor (1) to very good (5). Comprehension refers to a basic understanding of the concepts, underlying rationale, techniques/skills (e.g., getting out of bed when unable to sleep), and “model” (e.g., understanding of why certain behaviors are likely to help insomnia). Implementation refers to whether the participant put the learned skills into practice; this rating captures frequency but does not imply mastery. Mastery refers to how well the participant puts the learned skills into practice, as well as how successful the participant is at adapting the skills to meet individual goals.

Analysis Plan

Before proceeding with major analyses, we will check for imbalance between the two study arms on demographic characteristics, attrition, and baseline assessments using descriptive statistics and measures of effect size. We will not use the results of this examination as the basis for including covariates in the primary intent-to-treat analysis.^92-94 However, if significant between condition differences are detected for key variables (e.g., days supply of TAU antidepressant medication), subsequent exploratory analyses will examine main outcomes with adjustments for these variables (e.g., as a covariate).

Primary Outcomes

Our primary depression outcomes are (a) depression response based on the CGI-I (score ≤ 2) and (b) time to MDD diagnostic remission as determined by LIFE symptom ratings.⁵⁵ Our primary sleep outcomes are (a) actigraphy TST and (b) total score on the Insomnia Severity Index (ISI).

To determine whether the CBT-I arm had more improvement over time compared to the SH arm on the sleep and depression primary outcomes, we will test differences in the trajectories across time between arms with two-level HLMs in a growth curve framework.^95-97 Depression response will require using the generalized form of the HLM with a logit link and the binomial distribution, given that this outcome is binary. The first level of the model will include time as a predictor (number of weeks since baseline), thus modeling the within-person variation. Assuming that the trajectories are nonlinear as is common in clinical trials (e.g., initial intervention effect upon completing the intervention and maintenance period), we will use fractional polynomials to estimate the most appropriate function for time.^98,99 The second level will include a dummy variable for arm as the predictor variable for both the random effects of the intercept and slope for time. A significant coefficient for arm on the slope of time (i.e., the cross-level interaction of arm by time) would indicate that there are different trajectories across time for each arm. We will probe any significant interactions by graphing the simple-effects equations to determine whether the observed pattern is consistent with what we hypothesized. A pattern in which the CBT-I+SSRIs condition demonstrates a greater increase in sleep (TST and ISI) and greater likelihood of depression response (CGI-I <= 2) over time than the control arm (SH+SSRIs) supports the effectiveness of the intervention.

Because the LIFE symptom ratings record the week in which depression remission occurs, we will use survival analysis to examine this primary outcome. We will examine the data using Kaplan-Meier curves, and we will test efficacy with a Cox proportional regression model using arm as a predictor. We will handle any ties in follow-up times using Efron’s method. A significant hazard ratio such that participants in the experimental arm (CBT-I+SSRIs) are more likely to achieve remission earlier than those in the control arm (SH+SSRIs) indicates intervention effectiveness.

Power Analyses

Using a combination of the initial estimates from a pilot study we conducted⁷⁸ and past research findings that defined clinically meaningful effects, we conducted a series of power analyses for the primary outcomes. Our working sample size (n=165) is based on the number of participants we would feasibly be able to recruit during our funded timeframe after accounting for an anticipated maximum 15% attrition (N=136). We used PASS 2008¹⁰⁰ software for all power analyses.

Continuous Outcomes

For the continuous outcomes of sleep (actigraphy TST and ISI), we used the pattern of standardized means (Table 3) and estimates for the random-error component from the pilot study results in a simulations-based (5000 samples) power analysis for the mixed-models PASS module. At an alpha level of .05, we will have 88% and 81% power to detect differences in the trajectories across time for TST and ISI, respectively.

Table 3.

Estimated Effect Size Differences between Arms Across Time for Sleep Outcomes

Sleep Outcome	Week 0	Week 12	Week 26	Week 52
Actigraphy TST	0	0.80	0.75	0.65
ISI	0	−0.73	−0.66	−0.56

Binary Outcomes

Power for generalized mixed models is less developed, and thus we estimated power using standard logistic regression for the CGI response based on the expected difference in the proportions between arms at 26 weeks. If 61% of the control arm (SH + SSRIs) responds at this point,⁷⁸ we will have 80% power to detect a difference in CGI of 21.5% between the arms (OR=3.00). Although the required difference is higher than that estimated in the cited pilot study (16.7%), it is an achievable and clinically meaningful estimate that is within the margin of error.¹⁰¹

Time to Event Outcome

For diagnostic remission, the pilot estimated the hazard ratio to be 1.80 and the overall event rate to be .64. Using these estimates for a power analysis based on the Cox Proportional regression model, we will have 79% power with N=136 to detect a difference between the arms in survival times at an α level of .05.

Missing Data

For wholly missed measures we will handle missing data using full information maximum likelihood (aka direct maximum likelihood, which uses all available data to compute the parameter estimate based on maximum likelihood (which undergirds the computation of the hierarchical linear models), which provides unbiased parameter estimates and standard errors assuming the data are at least missing at random. Note that basing the power analyses on the expected sample size after attrition results in conservative power estimates. The increase in standard errors as a result of the uncertainty in the maximum likelihood handling of missing values using all available data should be less than the standard errors from a complete case analysis.

For missing items within measures, we will use mean item substitution if 20% or less of the items that make up a measure are missing. For example, for the PHQ-9 we will use mean substitution for no more than one missing item. Otherwise, we will consider the scale unscorable and therefore missing. For sleep diary data, we will require a minimum of four nights of data over a two-week period, at each wave of assessment. Otherwise, the assessment will be considered missing.

We will compare attrition rates between arms across time to determine whether there is evidence for differential attrition. In the unlikely event that differential attrition is detected, we will investigate whether the source of the difference is caused by a non-random mechanism, in which case we will consider carrying out a series of sensitivity analyses using models for data not missing at random (e.g., selection and/or pattern mixture models).^102-105

Secondary Outcomes

Secondary analyses will examine effects of the intervention arms on likely mediators (e.g., sleep change as a mediator for depression improvement), and on all secondary measures of sleep, depression and other psychiatric conditions (e.g., anxiety), and health risk behaviors. We will use the same analysis approach for all continuous secondary outcomes as for the primary sleep outcomes. For binary secondary outcomes, we will use the same framework as described for analyzing depression response.

Mediation Analyses

We will test whether sleep (TST, ISI) mediates the relationship between arm and depression (CDRS-R) using a cross-lagged panel model (CLPM) in a structural equation modeling framework.^106,107 In a CLPM,¹⁰⁷ the mediator and outcome at time t (for this study, timepoints are baseline, 4, 8, and 12 weeks) is regressed on arm, the mediator at t-1 and the outcome at t-1. Prior to estimating a final model, we will test for stationarity for the structural paths by comparing whether a model in which the paths are invariant fits significantly worse than one in which the paths are free to vary.¹⁰⁷ We will assess model fit using root mean square error of approximation (RMSEA) and the comparative fit index (CFI). Our criteria for good model fit will include an RMSEA of .08 or below^108,109 and a CFI of .95 or above.¹¹⁰ A significant indirect effect, based on bootstrapped standard errors,¹¹¹ would provide evidence for mediation.^106,112 We must use a secondary measure of depression (PHQ-9), as estimating a CLPM with binary outcomes is not feasible. However, we will also test the mediating effect of sleep on depression response (CGI<=2) using the product coefficients approach,¹¹³ which has been consistently shown to be a more powerful test of mediation^106,112,114 than the causal-steps approach.¹¹⁵ This will be accomplished by regressing (ordinary least squares; OLS) sleep (mediator) at 12 weeks on sleep at baseline and treatment arm, and by regressing (logistic) response (outcome) at 26 weeks on sleep at 12 weeks, response at 12 weeks, and arm. The product of the regression coefficient from the mediator regressed on treatment and the regression coefficient from the outcome regressed on the mediator provides the point estimate of the indirect effect. Because OLS and logistic regressions result in coefficients measured in different scales, we will standardize coefficients by multiplying each by the SD of the predictor variable divided by the SD of the outcome variable prior to estimating the indirect effect. We will statistically test the indirect effect using bootstrapped standard errors.¹¹¹ We will test whether emotional dysregulation mediates the relationship between sleep and depression by expanding the CLPM and causal steps analysis described above to include this variable and examining the respective indirect effects.

Moderation Analyses

We will also test whether there is intervention effect modification by demographic characteristics (age, gender), youth independence (youth living independent, parent involvement), and sleep or psychiatric factors (comorbid anxiety, baseline sleep, depression severity). We will test for effect moderation by adding the moderator and the product of study arm and the moderator variable in the level-two equation described in the primary outcome analysis.^116,117 A significant coefficient for the interaction term on the slope of time (i.e., 3-way cross-level interaction) would indicate a moderator effect and will be followed up with graphs to determine the nature of the effect modification.¹¹⁶ Each moderator will be tested separately.

Economic Analyses

We will conduct an economic evaluation to assess the cost-effectiveness of CBT-I, relative to SH, to inform decision-makers about the relative value of the intervention. We will conduct incremental cost-effectiveness analyses (ICEA) from a healthcare organization perspective, following methodology recommended by the Panel on Cost-Effectiveness in Health & Medicine and methods that we have used in previous cost effectiveness analyses.^118-120 We will use EQ-5D derived quality adjusted life years (QALYs) over 12 months as the primary clinical outcome. While QALYs are recommended by experts and facilitate comparability to other medical interventions, changes in symptoms related to specific mental health disorders targeted by this intervention may be missed (e.g., depression and insomnia).¹¹⁹ Thus, we will also conduct secondary ICEA analyses using depression-free days (DFDs) and insomnia-free nights (IFNs). As part of the economic evaluation, we will also use instrumental variable methods to attempt to better estimate the effect of intervention dose on QALYs, DFDs, IFNs, and cost, using potential instruments^89-91 collected as part of the baseline and post-treatment assessments. If the potential instruments collected through direct assessment fail to meet statistical thresholds,¹²¹ we will search for other sources of exogenous variation.¹²²

Discussion

This paper presents the design and methodology for a randomized controlled efficacy-effectiveness trial of CBT-I for adolescents with depression and insomnia who were also starting a new course of SSRIs. Depression is a very common psychiatric illness in youth and is often severe, chronic, and impairing. Unfortunately, outcomes from traditional treatments (e.g., CBT, antidepressants) have been modest at best; barriers such as cost, difficulty accessing care, and discomfort with treatments or side effects all erode the potential for treatment benefit. We focus on the treatment of comorbid insomnia as an innovative strategy for substantially improving depression outcomes. This approach is relevant to a majority of depressed youth, as insomnia is comorbid with depression in the majority (>70%) of cases.¹²³ Further, improving sleep is a generally desirable outcome itself, and psychosocial sleep treatments are relatively brief and very tolerable. This study has important public health implications and could yield a safe, simple, easily disseminated, inexpensive intervention with the potential to enhance depression outcomes across many critical emotional, cognitive, social, and physical domains for youth.

If this approach proves effective, an incident dispense of a new antidepressant could be a salient trigger event at which practitioners can screen for insomnia and, if present, offer sleep treatment. This “screenable moment” offers a simple, clear, and potentially highly effective path to facilitate dissemination of appropriate insomnia treatment to improve depression outcomes. Furthermore, given that sleep issues have less stigma attached to them than mental-health issues, as well as potential for rapid relief from fatigue, treating sleep issues as a path to improve depression symptoms may be attractive to adolescents and parents.

When antidepressants are delivered in real world settings to patients with complex presentations, they typically have reduced benefits compared to results seen in highly controlled efficacy RCTs. This shrinkage in benefit from controlled trials to real world implementation is due to poor patient adherence¹²⁴ and prescriber factors.¹²⁵ Nonetheless, antidepressants may be the best real-world context in which to test sleep treatments. This is because they are the most frequently delivered treatment for youth depression.^126,127 And yet, there is substantial room for improvement even when adherence and dosing are maximized.¹²⁸

We will also conduct economic analyses of the incremental cost-effectiveness of CBT-I relative to SH, from the healthcare organization perspective. Understanding costs and benefits is especially important in situations where we augment an existing, established treatment (TAU antidepressants) with an additional intervention. The cost-effectiveness analysis (CEA) will determine whether the added cost of CBT-I is justified by the observed additional benefits. Should we find the predicted CBT-I benefits, CEA results will be important for rapid progression to next-step dissemination and implementation research, which relies heavily on being able to demonstrate the value of this approach. In addition, as part of economic analyses, we will examine the dose-response effect of intervention dose on outcomes, applying instrumental variable methods to help dampen the confounding effects inherent in analysis of non-random elements of a controlled trial.

Our design compares active (CBT-I) and control (SH) insomnia treatments, both augmenting TAU SSRI antidepressants. This design combines elements of effectiveness research (cost analyses, embedded in real world samples/settings, TAU care) with features more typical of efficacy trials (indepth assessments, randomization, research delivery of the sleep treatments). We chose this approach rather than a pure efficacy trial with tightly-controlled, study-administered SSRIs because a blended design allows for a controlled test of CBT-I benefit but will also yield effectiveness results that can help speed the cycle of research and more quickly move evidence-based approaches into practice—one of the key objectives of the NIH Roadmap.^129,130 Should the hypothesized results be supported, this trial has the potential for meaningful depression treatment improvement.