A Randomized Pilot Cognitive Behavioral Sleep Health Trial for Young Adults with Type 1 Diabetes

Abstract

Objectives:

The purpose of this randomized controlled trial was to determine whether a cognitive behavioral sleep health self-management intervention (CB-Sleep Health) would be more effective than a time-balanced attention control (AC) condition in improving multiple dimensions of sleep health (self-reported and objectively derived).

Methods:

Young adults with T1D (ages 18–26 years) were randomly assigned to a 12-week CB-Sleep Health (n = 21) or AC condition (n = 18). They wore concurrent continuous glucose monitors and actigraphy devices and completed daily sleep surveys for 14 days at baseline, post-intervention, and 3-month follow-up.

Results:

Of the randomized participants, 31 (79.5%) completed the post-intervention, while 33 (84.6%) completed the 3-month follow-up. The CB-Sleep Health intervention had a significant effect on alertness and duration compared to the control group. The changes from baseline were −3.21 seconds vs. +0.71, p = .005 and +18 minutes vs. −25.8 minutes, p = .01, respectively. These effects were sustained at the 3-month follow-up.

Conclusions:

Longer sleep duration, higher daytime alertness, and sustained sleep efficiency is possible with this CB-Sleep Health intervention in young adults managing a complex condition.

Introduction

Young adults with type 1 diabetes (T1D) have poor sleep health, including lower satisfaction and daytime alertness, inconsistent regularity and timing, and shorter sleep duration when compared to same-age peers without a chronic condition (Jauch-Chara et al., 2008; Ji et al., 2021). Poor sleep health is associated with a range of medical and psychiatric comorbidities, including emotional distress and impaired glucose regulation (Borel et al., 2013; Freeman et al., 2020). Among young adults with T1D, rates of achieving glycemic targets are low (14% in 18–25 years compared to 29–30% >26 years) (Foster et al., 2019; Miller et al., 2015). Of relevance, melatonin is secreted later from adolescence through late 20’s (Fischer et al., 2017) resulting in a delay in phase and there are recognized impairments in melatonin synthesis with hyperglycemia (Amaral et al., 2014).

There is preliminary evidence supporting that cognitive-behavioral sleep health self-management (CB-Sleep Health) interventions (e.g., sleep extension + timing consistency) improve sleep health in adolescents and young adults in the general population (Griggs et al., 2020), with a few phase I trials on children, adolescents, or adults across a larger age range (18–65 years) with T1D (Jaser et al., 2020, 2021; Martyn-Nemeth et al., 2023a; Perfect et al., 2023). However, the efficacy of CB-Sleep Health (phase I trial) in young adults aged 18–26 years, transitioning to adulthood while managing T1D, has not been examined. The primary aim of this study was to evaluate the preliminary short-term (3-month) efficacy of CB-Sleep Health on measures of sleep health for young adults with T1D. A secondary aim was to determine if these improvements could be sustained 3 months post-intervention.

Materials and Methods

Study design

This study is a randomized, outcome-assessor blind controlled pilot trial conducted at Case Western Reserve University. Eligible participants were randomly assigned 1:1 to a 12-week CB-Sleep Health (n = 21) or time-balanced attention control (AC) condition (n = 18). The study followed the Consolidating Standards of Reporting Trials (CONSORT) recommendations and was registered in ClinicalTrials.gov (NCT04975230). Ethical approval was obtained from the [University Hospitals, IRB # STUDY20211165]. The trial was conducted from March 2022 to October 2023.

Participants

Individuals receiving care at a diabetes specialty clinic in Northeast Ohio were eligible to participate if they (1) were between 18 and 26 years of age; (2) had T1D for at least 6 months; (3) had no other complex comorbid medical (e.g., active cancer) or major psychiatric condition (e.g., schizophrenia or bipolar disorder); (4) were not enrolled in another intervention study; (5) had a most recent glycated hemoglobin (A1C) value ≥ 7% or spent ≤80% time in glucose range. Exclusion criteria were: (1) sleep disorder diagnosis including obstructive sleep apnea (apnea/hypopnea index ≥15/h), periodic limb movement disorder, restless legs syndrome; (2) current pregnancy; night shift or recent trans meridian travel; (3) those who habitually slept > 7 hours.

A three-step screening process was employed (1) electronic health record screen for eligibility; (2) phone or in-person interview including review of health history including sleep apnea screening via the Berlin Questionnaire (Netzer et al., 1999); (3) 2-week home monitoring period with concurrent actigraphy and CGM. There was an additional screening of a 1–2-night home sleep apnea test (Nox T3) for those scoring high risk on the Berlin Questionnaire (Netzer et al., 1999) (See Figure 1 CONSORT Flow Diagram).

Figure 1.

Study Protocol

Study procedure

After completing screening procedures and the baseline monitoring period, participants were randomized 1:1 to study arms: (1) CB-Sleep Health or (2) time-balanced AC condition, stratified by sex at birth. A study biostatistician with no participant contact generated random allocation using a random number generator. After the 12-week treatment period (post-treatment) and 3 months after posttreatment, participants completed follow-up assessments, including 14 days of actigraphy with daily sleep diaries, 14 days of CGM monitoring (Battelino et al., 2019; Danne et al., 2017), questionnaires, and an A1C point of care at each time point. Both conditions were administered in person or via telehealth and included a 60-minute primary session, followed by 5-minute weekly check-ins (text, call, email) and 30-minute booster sessions every 4 weeks. The follow-up, whether conducted via text or call, was kept brief and focused on the main goal identified for the participant. This method of follow up was used to keep participants engaged and a majority of participants opted for text (87.6%) with fewer opting for a phone call (8.8%) or email (3.5%). Participants were compensated for each assessment ($30 for baseline, $40 for main intervention session, $50 for post, and $50 for follow-up, with additional compensation $30 based on wear compliance as determined with data extraction from devices). We also provided parking passes for study visits.

CB-sleep health condition

The CB-Sleep Health manual, based on established protocols (Griggs et al., 2020; Harvey et al., 2016; Perfect et al., 2023) was delivered by a sleep coach either in-person or through telehealth. The manualized sessions included an initial 60-minute consultation to set goals and develop an action plan, followed by 5-minute weekly follow-ups and 30-minute booster sessions every 4 weeks. The coach focused on areas such as increasing time in bed, maintaining consistency, and improving bed/rise time. Session 1 covered treatment overview, sleep and diabetes information, sleep health assessment, and tips for creating a healthy sleep environment. Progressive muscle relaxation and guided imagery were practiced. Sessions 2 and 3 reinforced healthy sleep behavior. Weekly follow-ups were conducted through text or phone calls, and treatment adherence and sleep diaries were reviewed in all sessions.

Attention control condition

The AC condition received three sessions face to face or telehealth with weekly follow-ups by text or call with time spent analogous to the experimental condition. The time-balanced manualized sessions were centered on participant-driven goals with an initial 60-minute consultation to set goals and develop an action plan, followed by 5-minute weekly follow-ups and 30-minute booster sessions every 4 weeks. The coach focused on diabetes standard of care areas such as current plan of care, lowering risk, healthy eating, problem solving, and healthy coping.

Measures

Sleep diary

Participants completed 14 consecutive days of sleep diaries at baseline (Monk et al., 1994), postintervention, and 3-month follow-up and daily for the 12-week intervention period.

Actigraphy

The Actiwatch Spectrum Plus (Philips Respironics) was used to measure sleep and circadian behavior objectively. Actigraphy has been compared to polysomnography as an accurate measure of sleep duration, sleep efficiency, and wake-after-sleep onset in young adults with T1D (Farabi et al., 2017). Actigraphy-derived circadian parameters have also been reported to be reliable in adults with diabetes (Cavalcanti-Ferreira et al., 2018). Data were collected in 30-second epochs using medium sensitivity settings and scored through the agreement (2 out of 3) of a combination of (1) the event marker, (2) the light meter, and (3) diary (with priority given to the event marker and light meter). The Spectrum Plus collects activity data with a standard spectrum of light and off-wrist detection. Actigraph recordings are valid for identifying circadian disorders, sleep disturbance, and other temporal rhythms that are discernible in the time series data available through actigraphy (Sack et al., 2007). Participants were instructed to continuously wear the actigraph on their non-dominant wrist for 14 days and depress the event marker at “lights out” and “lights on” times to demarcate time in bed. They were asked to record bedtimes and wake times in a daily electronic diary in Research Electronic Data Capture (REDCap). Actigraphy outcomes were set to missing if a participant provided fewer than 3 of the target 14 days; median number of nights provided per participant was 13 at baseline, 13.5 at post-intervention, and 12 at 3 months.

Regularity (actigraphy-derived)

The raw time series activity data from the wrist actigraphy were exported for this analysis. One circadian rhythm parametric parameter acrophase (timing estimation) and one nonparametric parameter: interdaily stability (IS) (regularity estimation) was derived from the wrist actigraph recordings to address the primary outcomes (Cavalcanti-Ferreira et al., 2018; Cornelissen, 2014). The circadian rhythm for each participant was estimated using the regression model for a single component cosinor model and a least squares approach as

$$Y_i\left(t\right)=M_i+A_i·cos\left(\frac{2\pi t}{\tau }+{\phi }_i\right)+e_i\left(t\right)$$

where $M_i$, $A_i$, and ${\phi }_i$ are the mesor (midline estimating statistic of rhythm: rhythm-adjusted average activity count), the amplitude (a measure of the half range of the predicted variation of a cycle), and the acrophase (time of the high peak of the cycle) for participant $i$ respectively.

Timing (actigraphy-derived)

Timing was derived from acrophase the peak alertness time (actual clock time of the peak amplitude) calculated with the equation, $\varnothing =(-\frac{24}{2\pi })·\phi$ (Grutsch et al., 2011). For the 24-hour cycle of the rhythm, the period ($\tau$) is 1,440 mins (60 minutes/hours × 24 hours). The other parametric circadian parameters (e.g., mesor and amplitude) were used to describe the sample. We also explored the regularity of sleep duration by computing the coefficient of variation.

Satisfaction

Satisfaction was measured at baseline, posttreatment, and 3-month follow-up using the 4-item Patient Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance Scale with higher scores indicating higher sleep disturbance. The satisfaction score performed with good internal consistency at each timepoint (Cronbach’s α = 0.828, 0.782, 0.773).

Alertness (objective)

The data collectors administered the trail making test (TMT), a 3 – 5 minute test on paper consisting of two parts (A & B) (Gaudino et al., 1995). The TMT provides information on visual search, scanning, speed of processing, mental flexibility, and executive functions. For the TMT-A, an individual is instructed to connect lines sequentially for 25 encircled numbers, and for TMT-B an individual is instructed to alternate between numbers and letters (e.g., 1, A, 2, B, 3, C, etc.). Seventy-five seconds is the adult average for the TMT-B, with deficiencies noted > 273 seconds (Gaudino et al., 1995). Higher scores reflect lower sustained alertness.

Efficiency and duration (actigraphy-derived)

Sleep characteristics were computed using Actiware 6.0.9 software from Spectrum Plus data including time in bed (TIB), sleep duration (e.g., total sleep time), sleep efficiency (%), wake after sleep onset, and sleep onset latency. Sleep efficiency percentage was calculated as sleep duration / TIB X 100.

Sleep-health composite

The Sleep Health Composite was derived from objectively collected data for regularity, alertness, timing, efficiency, and duration (actigraphy and Trail-Making Test) and self-reported for satisfaction coded as 1 = good and 0 = poor, with scores ranging from 0–6 and higher scores indicating better sleep health (Dong et al., 2019; Griggs et al., 2023). The cut points for the five dimensions were: regularity (interdaily stability >0.6), satisfaction (PROMIS sleep disturbance < 40, i.e., 1 SD below the population mean), alertness (trail-making test score ≤ 29.6, i.e., 1 SD above the mean), timing (acrophase 14:00 to 18:00), efficiency ≥ 85%, and duration 7–9 hours. The composite was derived based on the extant literature and National Sleep Foundation recommendations (Dong et al., 2019; Ohayon et al., 2017).

Statistical analysis

Intention to treat analyses were conducted with no imputation for missing data and statistical tests were two-sided with statistical significance set at p < 0.05 (SAS version 9.4 [SAS Institute, Cary, NC] and SPSS version 29.0). Sleep outcomes – and equivalently, within-participant change from baseline – at postintervention and 3-month follow-up were modeled as a function of fixed effects for timepoint (Level 1, nested within-person), treatment condition (Level 2), and their interaction (the latter term was included to assess whether intervention differences observed immediately postintervention were maintained at 3 months), adjusting for baseline outcome value to account for possible regression to the mean (Chuang-Stein & Tong, 1997). Linear mixed models with a participant-specific intercept were used to account for within-participant correlation over time (Fitzmaurice et al., 2011). Separate models were conducted for each outcome. This was a small sample with no preselected covariates of interest, nor were there between-group differences at baseline, due in part to study design – sex-stratified randomization and narrow inclusion criteria (narrow age range, T1D for at least 6 months without complex medical or psychiatric comorbidities); therefore, analyses were conducted without adjustment for covariates. Effect sizes for within-participant change from baseline were calculated with Cohen’s d (0.20 small, 0.50 medium, and 0.80 large). Model assumptions (normally distributed residuals, no overly influential observations) were satisfied for all outcomes.

Results

Participant characteristics

We present the pretreatment characteristics in Table 1. The final sample comprised 39 young adult participants with a mean age of 21 years (CB-Sleep Health: 21, AC: 18) (Table 1). Approximately 41% of the sample were women, nearly half were full time college students, and a majority were able to cover their household expenses. The racial composition was predominantly Non-Hispanic White, followed by a smaller percentage of Non-Hispanic Black and Hispanic individuals. A majority of participants used an insulin pump (7.7% were closed loop systems) and the AC group had a slightly higher prevalence of insulin pump use (77.8% vs. 66.7%). For continuous glucose monitoring (CGM) devices, the majority used their own device with 74.4% using Dexcom G6, 10.3% using FreeStyle Libre and 7.7% using Medtronic. The remainder of the population opted for the Dexcom G6 study device (8.6%).

Table 1.

Pretreatment Clinical Characteristics of Randomized Participants by Condition (N = 39)

Demographic Characteristics	Total	(N=39)	CB-Sleep Health	(N=21)	Attention Control	(N=18)
	M	(SD)	M	(SD)	M	(SD)
Age	21.08	(2.24)	20.95	(2.16)	21.22	(2.39)
	n	(%)	n	(%)	n	(%)
Gender
Man	23	(59.0)	13	(61.9)	10	(55.6)
Woman	16	(41.0)	8	(38.1)	8	(44.4)
Race
White	29	(74.4)	16	(76.2)	13	(72.2)
Black	7	(17.9)	3	(14.3)	4	(22.2)
Asian	1	(2.6)	1	(4.8)	0	(0)
Bi/Mixed Race	2	(5.1)	1	(4.8)	1	(5.6)
Ethnicity (Hispanic)	2	(5.1)	0	(0)	2	(11.1)
Race x Ethnicity Non-Hispanic White	28	(71.8)	16	(76.2)	12	(66.7)
Non-Hispanic Black	6	(15.4)	3	(14.3)	3	(16.7)
Non-Hispanic Asian	1	(2.6)	1	(4.8)	0	(0)
Hispanic White	1	(2.6)	0	(0)	1	(5.6)
Hispanic Black	1	(2.6)	0	(0)	1	(5.6)
Clinical Characteristics	M	(SD)	M	(SD)	M	(SD)
Body mass index (kg/m2)	25.87	(5.64)	25.02	(5.36)	25.70	(6.10)
T1D Duration	11.68	(4.76)	11.38	(4.91)	12.06	(4.68)
A1C (%)	8.46	(1.79)	8.43	(1.57)	8.51	(2.06)
Time in Glucose Range	44.08	(18.65)	43.11	(19.66)	45.40	(17.82)
Coefficient of Variation	36.31	(4.80)	36.02	(5.10)	36.69	(4.53)
	n	(%)	n	(%)	n	(%)
Insulin pump (%yes)	28.0	(71.8)	14	(66.7)	14	(77.8)

Note: Time in Range 70–180 mg/dL measured by continuous glucose monitor

There were no significant between group differences in demographic or clinical variables. At baseline, the mean sleep regularity (interdaily stability) score was 0.45 with 10.1% of scores < 0.6, satisfaction was 47.05 with 92% reporting low satisfaction, alertness was 29.62 seconds with 36.8% observed to have low alertness, timing (acrophase) was 16:01 with 34.1% early or late timing, efficiency 85.17% with 40.9% low efficiency, duration 6 hours 37.2 minutes with 65.9% short duration.

Sleep duration had Spearman correlations of magnitude 0.31 – 0.39 with glucose parameters such as time in range and coefficient of variation. However, there were no associations of moderate or higher (Spearman correlation 0.30+) between the remaining baseline sleep health dimension parameters and the baseline glucose parameters of interest.

Outcomes

Objective sleep health dimensions

Descriptive and inferential statistics for outcomes are depicted in Table 2. Except for regularity, improvements from baseline in sleep health were consistently better for CB-Sleep Health than for AC, with statistically significant main effects – calculated combining data from the two follow-up visits – for alertness, efficiency, duration. There were no statistically significant interactions between randomization group and visit, consistent with stable between-group differences over time. Regarding visit-specific comparisons of the two conditions, which are based on smaller sample sizes than the main effect for condition (which pools visits), between-condition differences were statistically significant for alertness and duration at immediate post-intervention, and no statistically significant differences at 3 months.

Table 2.

Descriptive and inferential statistics for primary sleep health dimension outcomes

	Baseline (a)	Post (b)	3-Month (b)			Group	Time	G x T	Cohen’s d (95% CI) (c)
	M	SE	M	SE	M	SE	F	F	F	BL to Post	BL to 3-Month
Regularity 1								4.73 *	0.40	0.26
CB-Sleep Health	0.41	0.03	0.31 W	0.05	0.31 W	0.05				−0.94 (−1.84, −0.39)	−0.89 (−1.82, −0.32)
Attention Control	0.50	0.08	0.37	0.05	0.43	0.05				−0.18 (−0.79, 0.36)	−0.01 (−0.64, 0.62)
Satisfaction 2							0.02	0.00	0.33
CB-Sleep Health	47.05	1.82	49.76	1.56	48.95	1.60				0.12 (−0.36, 0.63)	0.02 (−0.48, 0.54)
Attention Control	50.98	1.15	48.66	1.84	49.44	1.80				−0.02 (−0.62, 0.58)	0.14 (−0.41, 0.74)
Alertness 3							7.24 *	4.43+	3.20+
CB-Sleep Health	18.39	1.43	15.46 B, W	0.79	15.30 W	0.89				−0.47 (−1.24, 0.10)	−0.49 (−1.51, 0.22)
Attention Control	20.30	1.25	19.38	0.84	17.33	0.91				0.13 (−0.51, 0.82)	−0.24 (−1.15, 0.51)
Timing 4								2.77	0.47	1.70
CB-Sleep Health	16:04	0:26	16:48 B	0:36	16:21	0:38				0.33 (−0.22, 1.09)	0.11 (−0.49, 0.77)
Attention Control	15:57	0:26	15:01	0:35	16:25	0:38				−0.59 (−1.31, −0.06)	0.15 (−0.45, 0.82)
Efficiency (%) 5								4.29 *	0.16	0.07
CB-Sleep Health	85.55	0.89	85.79	0.81	85.93	0.87				0.20 (−0.34, 0.82)	0.23 (−0.36, 0.92)
Attention Control	84.72	1.06	83.98	0.81	84.58	0.84				−0.21 (−0.82, 0.33)	−0.08 (−0.69, 0.51)
Duration (hr) 5								7.69**	0.39	0.06
CB-Sleep Health	6.36	0.22	6.77 B	0.17	6.83	0.18				0.17 (−0.37, 0.78)	0.23 (−0.36, 0.91)
Attention Control	6.94	0.23	6.17 W	0.17	6.30	0.18				−0.43 (−1.10, 0.10)	−0.29 (−0.96, 0.27)
Composite 6								5.63 *	0.18	0.14
CB-Sleep Health	2.95	0.30	2.86 B	0.31	2.82	0.33				0.11 (−0.61, 0.90)	0.33 (−0.46, 1.38)
Attention Control	3.06	0.32	1.88 W	0.31	1.90 W	0.30				−0.59 (−1.58, 0.07)	−0.45 (−1.57, 0.32)

Note:

¹Interdaily Stability

²PROMIS Sleep Disturbance

³Trail Making Test A

⁴actigraphy-derived acrophase

⁵actigraphy-derived efficiency and duration

⁶composite of regularity, satisfaction, alertness, timing, efficiency, and duration.

^(a)Observed mean and standard error (SE).

^(b)Model-based estimated means and SEs from linear mixed modeling.

^(c)Calculated as mean within-participant change divided by group-specific baseline standard deviation. Hedges’ bias correction applied.

^*p<0.05

^WSignificant (p < .01) within-group change from baseline.

^BSignificant (p < 0.01) between-group difference at time point.

More specifically, the immediate post-treatment effect sizes on the objective sleep health dimensions ranged from small (0.13) to moderate (0.59), with statistically significant greater improvements in the CB-Sleep Health group for daytime alertness (Trail Making Test A) and duration (actigraphy). From baseline to posttreatment, participants in the CB-Sleep Health group demonstrated greater baseline-adjusted improvements in daytime alertness (−3.21 seconds, p = 0.0014, d = −0.47) and sleep duration (+ 11 minutes, p = .0227, d = 0.217) than those in the AC condition (alertness +0.71 seconds, p = .4132, d = 0.13; duration −25.8 minutes, p = .0227, d = −0.43) at the immediate post-intervention assessment. At 3 months post-intervention, participants in the CB-Sleep Health group continued to have greater baseline-adjusted improvement in baseline compared with the AC group, but between-group differences were smaller and not statistically significant. Specifically, the effect sizes at 3 months ranged in magnitude from −0.89 to 0.33 in the CB-Sleep Health group and from −0.56 to 0.35 in the AC group, suggesting that between-group differences were consistent at immediate post-intervention and 3 months post-intervention, consistent with the lack of statistically significant interactions between group and visit.

CB-Sleep Health participants had consistently better (lower) scores in sleep satisfaction than AC participants (Table 2), although not statistically significant either for the main effect or for visit-specific differences. Effect sizes for group differences in within-participant change (Supplemental Table 1) were small at both visits. The sleep health composite scores for CB-Sleep Health participants remained relatively stable over time, while the AC group experienced declines from baseline at both T2 and T3.

Effect sizes for the between-group differences in change since baseline (Supplemental Table 1) also highlight that sleep health improvements tended to be better for CB-Sleep Health participants, ranging in magnitude at immediate post-intervention from 0.55 (efficiency) to −0.95 (alertness; lower scores are better), all moderate to large effect sizes. At 3 months, there was essentially no between-group difference in within-participant change in timing (effect size=−0.02); for other objective sleep outcomes other than regularity, effect sizes ranged from 0.41 (efficiency) to 0.93 (composite score).

In contrast, both groups exhibited declines from baseline in regularity (interdaily stability), where higher scores are preferred; however, the decline was statistically significant in the CB-Sleep Health group at both visits, with a statistically significant main effect for group (but no statistically significant group x time interaction and no visit-specific statistically significant group differences). Parallel analyses of intradaily variability and coefficient of variation in sleep duration (Supplemental Table 2), found little in the way of between-group differences, however. Corresponding effect sizes for group differences in within-participant change in regularity (Supplemental Table 1) were small to moderate for interdaily stability, and small for intradaily variability and coefficient of variation. In addition, we conducted a sensitivity analysis on a subset of participants who had 13–14 days at baseline and 13–14 days at least in one of the follow ups. The sample size was decreased to 13 participants (8 in CB-Sleep Health and 5 in AC). Thus, the power was diminished but the pattern of the results was consistent with the full sample.

Discussion

This study represents the first RCT to examine the change in sleep health factors following a CB-Sleep Health intervention in young adults aged 18 to 26 years with T1D. We demonstrated the preliminary efficacy of the intervention over three months in a sample of 39 young adults with T1D achieving a retention rate of 85%, with all but six participants remaining engaged through follow-up visits. Sleep health metrics demonstrated varied results. Regularity (0.45 vs. 0.46), satisfaction (5.36 vs. 5.91), timing (16:01 vs. 15:08), and efficiency (85.17% vs. 85.02%) were comparable between groups. However, the alertness score for participants in the CB-Sleep Health group (29.62 seconds vs. 18.68 seconds) was lower compared to findings from another study involving young adults with T1D (Griggs, Hickman, et al., 2021; Griggs, Strohl, et al., 2021). Notably, one dimension (regularity measured via interdaily stability) was worse in the experimental when compared to the control condition, although both groups declined in stability immediately post intervention.

Multiple objectively derived sleep health dimensions (alertness, efficiency, and duration) were consistently better in the CB-Sleep Health group than the AC group, reflected in the statistically significant main effect for group, particularly at immediate post-intervention assessment. Participants in the CB-Sleep Health condition had consistently longer actigraphy-derived sleep duration than the AC condition who experienced a significant decline from baseline to immediate post-intervention. Importantly, the experimental and control conditions did not significantly diverge regarding the total sample’s data availability, further reinforcing the reliability of our findings. Importantly, the experimental and control conditions did not significantly diverge with respect to the total sample’s data availability.

We determined it to be feasible to collect actigraphy and CGM data at each timepoint—baseline, immediate post, and 3 months post. A total of 100% (39/39), 79.3% (23/29), and 88.9% (24/27) had at least 7 days of actigraphy data at each timepoint respectively. Further, a total of 43.6% (N = 17/39), 48.3% (N = 14/29), and 40.7% (N=11/27) of the total sample had 14 days of actigraphy data at each timepoint, respectively. Continuous glucose monitoring (CGM) followed a similar trend, with 84.6% (33/39), 79.3% (23/29), and 82.3% (n = 26/32) of participants providing at least 7 days of complete CGM data and 76.9% (30/39), 72.4% (21/29), 59.4% (19/32) providing 14 days of complete CGM data at each timepoint respectively. One limitation of using the Spectrum Plus is that the research team did not have access to their data in real time limiting the ability to monitor trends or whether data were being successfully acquired. Furthermore, the device was not always returned leading to potential gaps in data collection and subsequent loss of data.

Our findings align with two prior randomized controlled trials focusing on sleep promotion in T1D populations—one involving 39 adolescents (13–17 years) (Jaser et al., 2020) and another with 14 adults (18–65 years) (Martyn-Nemeth et al., 2023a). Specifically, there were improvements in actigraphy-derived sleep duration in the trial of adolescents with T1D (Jaser et al., 2020). Also, the increase of 18 minutes in sleep duration in the experimental condition in the current study was comparable to the 25-minute increase in the other pilot sleep promotion study of 14 adults aged 18–65 years with T1D (Martyn-Nemeth et al., 2023b). In a subgroup analysis in that study, significant improvements were found only for short or variable sleepers (Martyn-Nemeth et al., 2023b) highlighting the complexity of sleep interventions across different populations. Overall, these findings underscore the potential for the CB-Sleep Health intervention in enhancing sleep duration among individuals with T1D, particularly when compared to previous studies.

Our analysis also revealed improvements in objective alertness for the CB-Sleep Health condition when compared to the AC condition as measured by the Trail Making Test A. Although there are no direct studies for comparison, findings from a sleep extension study of 121 short-sleeping adults with obesity (mean age = 41.8 years); showed no statistically significant improvements in daytime alertness as measured by the Trail Making Test A and B test (Lucassen et al., 2014). Our findings were consistent with experimental sleep deprivation studies of adults without chronic conditions. Specifically, it was previously reported that there is a dose-response effect with sleep deprivation where larger doses of sleep deprivation result in lower daytime alertness as measured by psychomotor vigilance performance testing (Drake et al., 2001; Van Dongen et al., 2003). Thus, these objective measures of alertness may add additional insight into the effects of these behavioral sleep interventions and the dose of sleep duration may mediate this sleep health outcome.

Several limitations warrant discussion when interpreting these results. As a pilot study, our primary aim was to establish preliminary efficacy; therefore, we were not adequately powered to draw definitive conclusions about efficacy. We were unable to examine racial/ethnic differences given the small sample size. Additionally, the delivery modality was predominantly via telehealth and only 11% of participants opted for face-to-face delivery making it challenging to assess whether efficacy varied by delivery mode. Effect sizes reported may not translate well to larger samples, and several confidence intervals crossed zero, preventing conclusive insights about the direction of the effects compared to powered studies. Despite these limitations, our findings suggest that the CB-Sleep Health intervention is potentially more effective than usual care at improving sleep health dimensions (including daytime alertness and sleep duration) in this underserved and high-risk group of young adults with T1D, particularly those struggling to meet glycemic targets. These findings should be confirmed with a powered efficacy trial that includes a racially and ethnically representative sample. Furthermore, studies should explore potential sex differences in the efficacy of the CB-Sleep Health intervention, as previous literature indicates differing neural processes associated with sleep deficiencies and their comorbidities (Li et al., 2023). Studies with a larger sample should investigate whether or how the efficacy of CB-Sleep Health intervention may vary between sexes.

Figure 2.

CONSORT Diagram