Sleep, Self-Management, Neurocognitive Function, and Glycemia in Emerging Adults with Type 1 Diabetes Mellitus: A Research Protocol

Abstract

Type 1 Diabetes (T1D) affects 1.6 million Americans, and only 14% of emerging adults ages 18–25 years achieve targets for glycemic control (A1C < 7.0%). Sleep deficiency, including habitual short sleep duration (<6.5h total sleep time and high within-person variability in total sleep time), is associated with poorer glycemic control. Emerging adults with T1D have a more pronounced sleep extension on weekends compared with matched controls, consistent with sleep deficiency; however, associations among sleep variability and glycemic control have not been explored in this population. Sleep deficiency may affect the complex higher-order neurocognitive functioning needed for successful diabetes self-management (DSM). We report the protocol for an ongoing study designed to characterize sleep and the associations among sleep deficiency, neurocognitive function, DSM, diabetes quality of life, and glycemia among a sample of 40 emerging adults with T1D. We monitor sleep via wrist-worn actigraphy and glucose via continuous glucose monitoring (CGM) concurrently over 14 days. We are collecting data on self-report and objective sleep, a 10-minute psychomotor vigilance test (PVT) on a PVT-192 device, a 3-minute Trail Making Test on paper, and questionnaires, including twice-daily Pittsburgh sleep diaries using Research Electronic Data Capture (REDCap)™. Results from this study will be used to support the development and testing of the efficacy of a tailored sleep self-management intervention that may improve total sleep time, sleep variability, neurocognitive function, DSM, glycemic control, and glucose variability among emerging adults with T1D. Keywords: sleep deficiency, sleep variability, emerging adults, type 1 diabetes, glycemic control, glucose variability

Sleep, Self-Management, and Glycemia in Emerging Adults with Type 1 Diabetes Mellitus

Type 1 Diabetes (T1D), characterized by beta cell destruction and absolute insulin deficiency, affects 1.6 million Americans (Centers for Disease Control [CDC], 2020). The prevalence of T1D is on the rise, with a projected increase to 5 million by 2050 (CDC, 2017). Only one in seven emerging adults ages 18–25 years with T1D achieve glycemic control (glycosylated hemoglobin, A1C < 7.0%) based on 2010–2014 national data (Foster et al., 2019). Lower A1C levels are associated with reduced risk for both micro- and macrovascular complications (Fowler, 2008) and better diabetes quality of life (DQOL) (American Diabetes Association [ADA], 2020).

Emerging adulthood is a distinct period of development characterized by rapid change and gradual exploration of life possibilities and choices in love, work, and worldviews (Arnett, 2000). Emerging adults with T1D must navigate multiple transitions (e.g., education, occupation, living situation, pediatric to adult diabetes care) while also assuming responsibility for their diabetes self-management (DSM) and overall health (Monaghan, 2015; Peters & Laffel, 2010). These transitions and the developmental issues experienced by this high-risk age group can lead to suboptimal use of health care, deteriorating glycemic control, an increase in the occurrence of acute complications (e.g., diabetic ketoacidosis), the emergence of chronic complications, and psychosocial challenges (Foster et al., 2019; ADA, 2020). Despite these critical health and psychosocial concerns in emerging adults with T1D, there has been limited emphasis on research geared towards this distinct developmental period (Monaghan, 2015; Peters & Laffel, 2011; ADA, 2020).

Sleep in Emerging Adults with T1D

Sleep deficiency is a major public health concern that affects a majority of emerging adults in the United States. Emerging adults are at a higher risk for sleep deficiency due to the demands of social interactions (e.g., social jetlag), the stress of academic or work roles, connectedness via the web and mobile phones, and shifts in circadian physiology, later chronotype – delayed sleep period, coupled with a need to maintain early wake-up times for work or school (Owens, 2014). Social jetlag is a discrepancy between one’s biological rhythm and social constraints (Wittmann, Dinich, Merrow, & Roenneberg, 2006). Later chronotypes fall asleep later in the evening and wake up later in the morning (Owens, 2014).

A majority of emerging adults with T1D do not meet National Sleep Foundation guidelines of 7–9 hours per night (Barone et al., 2015; Borel et al., 2013; Farabi, Carley, & Quinn, 2017), with a mean sleep duration of 6.6 hours (Barone et al., 2015; Borel et al., 2009; Farabi, Quinn, & Carley, 2017). More sleep variability, alternating between short sleep duration and compensation and shifts in circadian timing, is associated with poorer glycemic control in adolescents (Patel et al., 2018) and middle-aged adults with T1D (Larcher et al., 2016). Compared with control participants without chronic conditions, emerging adults with T1D have more pronounced sleep extension on weekends (Barone et al., 2015); however, to our knowledge, the relationships among sleep variability and glycemic control has not been explored in this age group.

Sleep also plays a role in glucose homeostasis (Donga et al., 2010). The adverse impact of sleep deficiency on hormones (e.g., cortisol, thyroid stimulating hormone), glucose metabolism, and body weight regulation in healthy adults without chronic conditions has been documented in multiple studies (Knutson & Van Cauter, 2008; Kohatsu et al., 2006; Spiegel, Tasali, Penev, & Van Cauter, 2004). Sleep deficiency is associated with higher body mass index (BMI) (Kohatsu et al., 2006), impaired body weight hormonal regulation (e.g., lower leptin levels, higher ghrelin levels) (Spiegel, Leproult, & Van Cauter, 1999; Spiegel et al., 2004), poorer insulin sensitivity, and poorer acute insulin response to glucose (Knutson & Van Cauter, 2008; Spiegel et al., 1999). Further, sleep deficiency is associated with higher cortisol secretion that can result in higher glucose levels (Leproult & Van Cauter, 2010).

Sleep deficiency is associated with hypoglycemia (Banarer & Cryer, 2003; Borel et al., 2009), hyperglycemia (Barone et al., 2015; Feupe, Frias, Mednick, McDevitt, & Heintzman, 2013), and poorer glycemic control in middle aged adults with T1D (Barone et al., 2015; Denic-Roberts, Costacou, & Orchard, 2016; Feupe et al., 2013). Better glycemic control is associated with better sleep quality (Barone et al., 2015), longer time in deep sleep measured by polysomnography (PSG) (Feupe et al., 2013), and fewer depressive symptoms (Denic-Roberts et al., 2016). On the other hand, poorer glycemic control is associated with more sleep variability (Chontong, Saetung, & Reutrakul, 2016), shorter sleep duration (Borel et al., 2013; Matejko et al., 2015), Obstructive Sleep Apnea (OSA) (Banghoej et al., 2017), social jetlag, and later chronotype (Larcher et al., 2016).

Sleep and Neurocognitive Function

Complex higher-order neurocognitive skills (e.g., psychomotor vigilance and executive function) are needed for successful DSM (Jacobson et al., 2011; Lyoo et al., 2009). Adequate DSM is critical to achieving glycemic control. Although various diabetes-related variables are associated with neurocognitive dysfunction - e.g., chronic hyperglycemia (Jacobson et al., 2011), microangiopathy (Ferguson et al., 2005), age of onset (Nathan et al., 2013), or disease duration (Brismar et al., 2007) - the pathophysiology of neurocognitive dysfunction is not well understood. Emerging adults with T1D perform poorer than peers without T1D on neuropsychological tests of psychomotor speed, sustained attention, and speed of information processing (Jacobson et al., 2011; Kodl & Seaquist, 2008). Sleep deficiency, including habitual short sleep duration (<6.5h total sleep time) and high variability in total sleep time, is a potential explanation for neurocognitive dysfunction since emerging adults with T1D have poorer self-reported and poorer objectively measured sleep when compared to a control group without chronic conditions (Adler, Gavan, Tauman, Phillip, & Shalitin, 2017; Barone et al., 2015; Jauch-Chara, Schmid, Hallschmid, Born, & Schultes, 2008).

Sleep and DSM in T1D

Sleep deficiency may affect DSM negatively in people with T1D resulting in difficulties with performing DSM activities necessary to optimize glycemic control such as checking blood glucose and responding to results, preparing optimal foods, and engaging in regular physical activity (Barnard et al., 2016). DSM involves complex tasks that require sustained attention and speed of information processing (Caruso et al., 2014; Thomas et al., 2000; Vloemans et al., 2019). The ability to perform these complex neurocognitive tasks decline with sleep deficiency (Durmer & Dinges, 2005; Jacobson et al., 2011; McNally, Rohan, Pendley, Delamater, & Drotar, 2010). Poorer executive function is associated with poorer glycemic control and is moderated by poorer DSM over time (e.g., four years) (Vloemans et al., 2019).

Much of the existing data about sleep in emerging adults with T1D have come from studies with cross-sectional between-subjects descriptive designs (N < 30) and short measurement time frames (e.g., over 3 days), self-reported sleep measures, samples with a broad range of ages with inclusion of middle-aged or older adults, or intermittent blood glucose monitoring rather than current monitoring technologies (e.g., continuous glucose monitors [CGM]) (Barone et al., 2015; Borel et al., 2013; Jauch-Chara et al., 2008). In this study, we address these important scientific gaps. We focus on characterizing sleep and glycemia concurrently over a longer period of time than in previous studies (e.g., over 14 days) with rigorous measures for our primary constructs of sleep, glycemia, and neurocognitive function among emerging adults age 18–25 years in their naturalistic environment. Monitoring over 14-days addresses the relatively short time period used in previous studies that may not fully reflect emerging adults’ habitual sleep and glucose patterns and is also in line with current ADA 2017 consensus reports for monitoring glucose patterns with CGM. Examining sleep and glucose concurrently and both the between-person and within-person associations are critical to inform our understanding of the true nature of the associations between sleep characteristics and glucose levels in this high-risk population.

Study Purpose

We developed our Sleep Self-Management Framework for Emerging Adults with Type 1 Diabetes (Figure 1) from a combination of the Revised Self- and Family Management Framework (Grey, Schulman-Green, Knafl, & Reynolds, 2015) and our integrative review on sleep in young adults with T1D (Griggs, Redeker, and Grey, 2018). The purpose of this study is to characterize sleep and the associations among sleep deficiency (total sleep time and sleep variability), neurocognitive function (executive function and psychomotor vigilance), DSM, DQOL, glycemic control, and glucose variability among emerging adults age 18–25 years with T1D. The study aims are to: (1) characterize sleep using self-report (questionnaires, diaries) and objective (actigraphy) methods, glycemic control, and glucose variability among 40 emerging adults with T1D over 14 days to capture weekend and weekday differences; (2) explore the associations among sleep deficiency, neurocognitive function, DSM, DQOL, and glucose variability between-subjects; and, (3) evaluate the extent to which daily variations in objective sleep characteristics (total sleep time, bed/rise times) predict subsequent glucose variability (time in range, J index, high and low blood glucose indices) and the extent to which daily variations in glucose are associated with subsequent sleep characteristics within-subjects. The purpose of this paper is to report a protocol for an ongoing study designed to characterize sleep and the associations among sleep deficiency, neurocognitive function, DSM, diabetes quality of life, and glycemia among a sample of 40 emerging adults with T1D.

Figure 1.

Sleep Self-Management Framework for Emerging Adults with Type 1 Diabetes

Methods

Design and Setting

We are using a descriptive repeated-measures design with assessments over 14-days. We obtained human subjects’ approval from the university and letters of support from the medical directors at the recruitment sites (pediatric and adult diabetes centers). We are recruiting emerging adults with T1D from the Yale-New Haven Health System (the Yale Diabetes Center, Yale Children’s Diabetes Program, Yale-New Haven Health).

Target Population

Emerging adults are eligible if they: (1) are between the ages of 18–25 years; (2) have been diagnosed with T1D for at least six months; (3) have no other major health problems (e.g., chronic medical condition or major psychiatric illness); (4) are not currently participating in any intervention studies; and (5) read/speak English. We are excluding those with a previous OSA diagnosis, those who are pregnant, and night shift workers. The age range was chosen because it encompasses a key developmental stage during which emerging adults are transitioning into either college or their careers. Emerging adults diagnosed with diabetes for at least 6 months are included to avoid the confounding of the initial adjustment period after diagnosis. We are including both emerging adults attending college and those in the workforce to increase the generalizability of our findings. We exclude night shift workers as night shift is one of the most frequent reasons for disruption of circadian rhythms, causing significant alterations of sleep and biological functions, which in turn affect physical and psychological well-being (Vetter, 2015). The sampling goal for this study is 70–80% White, Non-Hispanic, 9% Hispanic, 9% Black, and 2% other race with equal numbers of men and women, in line with the study population (CDC, 2017).

We use the Berlin Questionnaire (BQ) to screen participants for inclusion in the study (Table 1). The BQ (Cronbach’s α = 0.86–0.92) is a 9-item scale and is used to identify patients at risk for sleep apnea. The items address the presence and frequency of snoring, wake time sleepiness or fatigue, and history of obesity or hypertension ((Netzer, Stoohs, Netzer, Clark, & Strohl, 1999). Patients with persistent and frequent symptoms in any two of these three domains are considered to be at high risk for sleep apnea (Netzer et al., 1999) and are referred for treatment and not included in the study.

Table 1.

Summary of Measures

Measures	Format	Data Source	Schedule
Demographics
Health
Duration of Diabetes *	Nominal	Chart review	Baseline
Treatment Regimen *	Dichotomous	Chart review	Baseline
Individual
Age, * Gender, * Race/ethnicity, * Income *	Nominal, Categorical	Self-report and chart review	Baseline
Environment
Alcohol/caffeine intake	Categorical	Self-report	Twice-daily
Work/academic hours	Nominal	Self-report	Baseline
Illness characteristics/symptoms/response
Problem-focused coping	Likert scale	Self-report	Baseline
PROMIS v1.0 Emotional Distress-depression*	Likert scale	Self-report	Baseline
Diabetes Distress Scale*	Likert scale	Self-report	Baseline
Self-reported sleep characteristics
The Pittsburgh Sleep Quality Index (PSQI)	Open and Likert scale	Self-report	Baseline
Pittsburgh Sleep Diary	Open and categorical	Self-report	Twice-daily
The PROMIS SF v1.0-Sleep Disturbance (8 a)	Likert scale	Self-report	Baseline
Berlin Questionnaire (Sleep apnea risk)	Categorical and Likert scale	Self-report	Baseline
Epworth Sleepiness Scale	Likert scale	Self-report	Baseline
Objective sleep characteristics
Spectrum Plus™	Actigraph	Objective	Continuous
Neurocognitive function
Psychomotor Vigilance Test (PVT)	PVT-192 device	Investigator collected	Baseline
Trail-Making Test (TMT)	Paper	Investigator collected	Baseline
Diabetes Self-management (DSM)
Self-Care Inventory	Likert scale	Self-report	Baseline
Diabetes quality of life (DQOL)
PedsQL 3.2	Likert scale	Self-report	Baseline
Glycemic Control and Glucose Variability
Glycemic control	Lab value	Chart review - A1C	Baseline
Glucose variability	CGM	CGM report	Continuous

Note:

^*indicates potential covariates; Continuous Glucose Monitor (CGM).

Sample Size and Power Analysis

We determined a target sample size of 40 participants based in part on previous studies on sleep and glycemia in emerging adults and adolescents with T1D (Banarer & Cryer, 2003; Barone et al., 2015) and sleep intervention pilot studies in emerging adults (Fucito et al., 2018; Gellis, Arigo, & Elliott, 2013). Detectable associations presented below are based on our data collected thus far. For analyses of participant-specific summaries across 14-days (Aim 2), this sample provides 95% confidence intervals (CI) for Pearson correlations of [−0.01, 0.56], [0.10, 0.63], and [0.22, 0.70] for medium-sized correlations of 0.30, 0.40, and 0.50 respectively between sleep and glucose variables. Additional analyses of correlations between concurrent daily measures have an effective sample size of 40 participants × 14 days divided by the design effect, where the design effect indicates the impact of within-participant correlation (Kish, 1965). Using percent time in glucose range as an example, the intraclass correlation coefficient (ICC) from our preliminary data was 0.51, yielding a design effect of 1 + 13 × 0.51 = 7.63, and an effective N=73; corresponding 95% CIs for Pearson correlations between daily sleep and glucose measures are [0.08, 0.49], [0.19, 0.58], and [0.31, 0.65] for medium Pearson correlations of 0.3, 0.4, 0.5, respectively in assessing baseline associations. In multi-level models for daily measures (Aim 3), the detectable regression slope with Type 1 and 2 errors of 0.05 and 0.2 respectively, is d=[1.96 + .842] (1-ICC)^1/2 σ/[m n Sx²]^1/2, where σ=outcome standard deviation, m=40 participants, n=13 observations per participant (because the predictor is lagged by 1 day), and Sx²=within-person variance of the predictor (Diggle et al., 2002). For example, for determining the association of minutes of sleep with prior-day percent glucose time in range, we estimate ICC=.40, σ=98 minutes, and Sx²=within-person variance of % glucose in range=251, yielding a detectable d=0.59 – that is, the detectable increase in sleep minutes associated with an absolute increase of 1% in % glucose time in range. Similarly, for the association of percent glucose time in range as a function of prior-day sleep minutes, ICC=.51, σ=22.1%, and Sx²=within-person variance of sleep minutes=3905.5, the detectable increase in % glucose time in range associated with an increase of 1 sleep minute =0.03.

Procedures

We are using multiple recruitment methods: flyer, verbal referral from clinic staff, a letter endorsed by the medical directors, electronic messages sent through MyChart™, and face-to-face recruitment in the clinic. We present a summary of the study procedures in table 2. The Principal Investigator (PI) screens potential participants by phone. After explaining the study and obtaining written informed consent, the PI provides detailed verbal and written instructions for the actigraphy and sleep diaries and provided CGM if applicable. Emerging adults are instructed to wear the Respironics Spectrum Plus™, a wrist-worn sleep/wake data recorder, continuously for 14-days, removing it only for bathing, and depressing the event marker at “lights out” and “lights on” times to demarcate time in bed.

Table 2.

Study Procedure

Encounter 1	Encounter2	Encounter 3
Baseline survey and chart review	Download CGM and actigraphy data	Exit Interview
Neurocognitive tests	$35 incentive upon return of device/s	Share sleep report
Start CGM/actigraphy monitoring		$40 incentive
$25 incentive

Participants wear a CGM (their own or a blinded Dexcom G4™ provided by the study team) concurrently with the actigraph. The blinded CGM does not alarm for low or high glucose, and participants are instructed to use their own glucose meter as prescribed and for any signs or symptoms of high or low blood glucose (e.g., usual DSM). Their own CGM device is not blinded (standard of care). Participants who own their CGM have had time to acclimate to the technology and there is less potential for an intervention effect. Also, blinding a device owned by a participant would result in an increased risk and potential for harm in cases of unrecognized severe hypoglycemic episodes as they have come to rely on this technology for glucose monitoring (Ahn, Pettud, & Edelman, 2016).

Participants complete sleep and symptom diaries daily in the mornings and evenings to track daytime sleep-related behaviors (e.g., caffeine use, exercise) and nocturnal sleep characteristics (e.g., bedtime, awakenings). The PI contacts participants the day after enrollment to address any problems and elicit understanding. Throughout the study period, the PI monitors REDCap™, a secure web-based software, for twice-daily completion of the sleep and symptom diaries, sends twice-daily reminders to complete the data collection, and reminds participants to return the Spectrum Plus™ and, if applicable, Dexcom G4™, in a prepaid mailer at the end of the 14 days. All self-report or chart review data are collected or entered directly into REDCap.™

The PI manages the data on REDCap™ and exports the data into the Statistical Package for the Social Sciences (SPSS version 26) for analysis. Interviews are audio-recorded and transcribed verbatim. All transcripts are organized and coded with NVivo 12 for Mac. Actigraphy data are scored with Actiware™ v. 6.0.9 software. We score actigraph rest intervals for bedtime and waketime from a combination of the event marker, diary, and light level. Priority is given for the event marker and light level (2 out of 3 agreement) as these events are in real time. CGM data are calculated with Glyculator 2.0™ software. The baseline questionnaires take about 20–30 minutes to complete, the daily sleep and symptom diaries take about 5–10 minutes per day to complete, the PVT takes 10 minutes, and the Trail Making Test (TMT) for executive function takes 3–5 minutes. A summary of measures is presented in Table 1.

Variables and Measures

Demographic and clinical characteristics.

We use a combination of interview and the Electronic Medical Record (EMR) to collect and cross-validate clinical and demographic data, including age, body mass index (BMI, kg/m²), duration of diabetes, most recent A1C, and medical history (diabetes, sleep, last menstrual period for females, and other). The following data are collected by self-report survey: ethnicity, education, primary caregiver, employment status, military status, full-time student status, work hours, marital status, residence, household count, income, cigarette smoking, alcohol or other substance use, insulin therapy regimen (e.g., insulin injections or pump), CGM pump brand (if applicable), and last menstrual period for females. Based on methods used in previous studies, we are collecting multiple indicators of self-reported sleep quality, sleep disturbance, daytime sleepiness, and morning and evening symptoms - fatigue, nocturia, pain, hypo/hyperglycemia.

Illness Symptoms/Characteristics/Response.

We ask participants to report on how they manage sources of general stress with the Problem-Focused Coping Scale (Vitaliano, Russo, Carr, Maiuro, & Becker, 1985), diabetes-related emotional distress with the Diabetes Distress Scale (Polonsky et al., 2005), and depressive symptoms with the PROMIS Emotional-Distress scale (Pilkonis et al., 2011). Scores on the 11-item Problem-Focused Coping Scale (Cronbach’s α = 0.88) range from 0–33 with higher scores indicating better coping (Vitaliano et al., 1985). The 17-item Diabetes Distress Scale (Cronbach’s α = 0.88 to 0.93) was validated across four clinical sites with scores ranging from 17–102 with higher scores indicating higher diabetes-related emotional distress) (Polonsky et al., 2005). Scores on the 8-item PROMIS Emotional-Distress Scale (Cronbach’s α = 0.95) range from 8–40 with higher scores indicating more depressive symptoms (Pilkonis et al., 2011).

Self-reported sleep characteristics.

Participants self-report their sleep characteristics, sleep disturbance, sleepiness at baseline, and total sleep time, bedtime, and symptoms twice daily in a diary. Total sleep time, sleep efficiency, latency, and global sleep quality is reported using the 19-item Pittsburgh Sleep Quality Index (PSQI) (Cronbach’s α = 0.87; diagnostic sensitivity 89.6%; specificity 86.5%) (Backhaus, Junghanns, Broocks, Riemann, & Hohagen, 2002). Each item on the PSQI is grouped into 1 of 7 components, and each component yields a score from 0–3, with 3 indicating the most dysfunction (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989). PSQI component scores are summed and range from 0–21 with higher scores indicating poorer sleep quality (Buysse et al., 1989). Participants rate their sleep disturbance on the 8-item PROMIS Sleep Disturbance Scale (Cronbach’s α = 0.90 or above) with higher scores indicating greater sleep disturbance (Yu et al., 2012). Their general level of daytime sleepiness is measured with the 8-item Epworth Sleepiness Scale (Cronbach’s α = 0.88) (Johns, 1992). Scores range from 0–24 with higher scores being indicative of more sleepiness on the Epworth Sleepiness Scale (Johns, 1992). Participants record sleep and symptoms twice daily for 14-days while wearing the wrist actigraph (Monk et al., 1994). Participants log timing and frequency of awakenings and bed/rise times in the Pittsburgh Sleep Diary and related etiology on a visual analogue scale (fatigue, nocturia, pain, hypo/hyperglycemia).

Objective Sleep Characteristics.

The Spectrum Plus™ device collects activity data with a standard spectrum of light and off wrist detection. Five days or longer of monitoring reduces inherent measurement errors and increases reliability (Sadeh, 2011). Actigraphs provide measures of sleep characteristics including total sleep time, sleep efficiency (%), wake after sleep onset, and sleep onset latency, sleep fragmentation index (movement index + fragmentation index), and bed/rise times. Actigraphs provide the greatest agreement and least bias compared with PSG in emerging adults with T1D for sleep parameters (ICC 0.38 to 0.78) with non-significant associations (P > 0.65) for all sleep parameters (total sleep time, sleep efficiency, wake after sleep onset, and sleep fragmentation index) except for sleep onset latency (P = 0.04) (Farabi et al., 2017). Sleep onset latency is considered a less precise measure compared to PSG (Farabi et al., 2017).

Neurocognitive Function.

Neurocognitive function data are collected by the investigator with two tests, the psychomotor vigilance test [PVT] and the trail making test [TMT]. The PVT is a 10-minute test administered by an investigator on a PVT-192^R device used to measure simple reaction time to stimuli that occur at random intervals measuring vigilant attention. Vigilant attention is important when making food and exercise choices, deciding insulin dosing based on carbohydrate intake, and adjusting insulin dosing based on exercise or illness (Zhang, Marmor, & Huh, 2017). The PVT assesses fatigue and poorer alertness associated with sleep loss, extended wakefulness, circadian misalignment, and time on task, and performance does not improve as a function of repeated administration (Dinges et al., 1997). The test has been validated with emerging adults and has high reliability, with intra-class correlations measuring test-retest reliability above 0.80 (Dinges, Rogers, & Dorrian, 2004).

The TMT is a 3–5 minute test administered by an investigator consists of two parts (A & B) (Gaudino, Geisler, & Squires, 1995) and provides information on visual search, scanning, speed of processing, mental flexibility, and executive functions. The TMT-A requires an individual to draw lines sequentially connecting 25 encircled numbers on a paper and the TMT-B requires an individual to alternate between numbers and letters (e.g., 1, A, 2, B, 3, C). The score represents the amount of time required to complete the task. The adult average for the TMT-B is 75 seconds, with deficiencies noted > 273 seconds (Gaudino et al., 1995). The TMT has high reliability and has been validated in emerging adults with test-retest reliability of TMT A and B of 0.76 and 0.89 and 0.86 and 0.94 respectively (S. Wagner, Helmreich, Dahmen, Lieb, & Tadić, 2011).

DSM.

The Self Care Inventory-Revised is a 15-item questionnaire used to assess a person’s ability to perform DSM behaviors (diet, exercise, blood glucose monitoring, insulin administration, and attending medical appointments) (Weinger, Butler, Welch, & La Greca, 2005). The Self Care Inventory-Revised was validated with U.S. and Dutch patients with T1D with very good internal consistency (Cronbach’s α =0.87), and scores range from 15 to 225 with higher scores indicating better diabetes self-care (Weinger et al., 2005).

DQOL.

The Pediatric Quality of Life Inventory™ 3.2 Diabetes Module for Young Adults (PedsQLä Diabetes Module 3.2) is used to assess diabetes-specific Health-Related Quality of Life. It consists of five scales measuring Diabetes Symptoms (15 items), Treatment Barriers (5 items), Treatment Adherence (6 items), Worry (3 items), and Communication (4 items). The PedsQL has been previously validated for ages 8–45 with T1D with very good internal consistency (Cronbach’s α = 0.88 – 0.89) (Varni et al., 2018).

Glycemic Control.

Glycosylated hemoglobin (A1C) is routinely measured at quarterly clinic visits using the Siemens Vantage Glucose Analyzer ® (range = 2.5–14%) (Arsie et al., 2000).

Glucose Variability.

CGM systems provide real-time, dynamic glucose information every five minutes — up to 288 readings in a 24-hour period (J. Wagner, Tennen, & Wolpert, 2012). CGM data are downloaded directly from each participant’s existing or the provided blinded Dexcom G4™ CGM to capture glucose patterns. Participants insert a small sensor wire just under their skin using an automatic inserter (Kovatchev, Patek, Ortiz, & Breton, 2015). CGMs are accurate across a wide range of levels. Test-retest reliability ranges from 0.77–0.95 (Terada et al., 2014), with diagnostic sensitivity 78% and specificity 96% (Jensen et al., 2013). Glucose variability indices are calculated from CGM as Mean±SD, time in range (calculated as % in target range 70–180 mg/dL, % time in hypoglycemia < 70 mg/dL & % time in severe hypoglycemia <54 mg/dL and % time in hyperglycemia >180 mg/dL & % time in severe hyperglycemia >250 mg/dL), J index - overall quality of glucose variability (calculated as 0.001 × [mean + SD]²), low and high blood glucose risk indices, mean amplitude of glucose excursion (average amplitude of upstrokes or downstrokes with magnitude >1 SD), and mean of daily differences (Suh & Kim, 2015).

Planned Analyses

We screen for outliers and assess missing data on questionnaires, actigraphy, and CGM data. Outliers will be identified with visual (e.g., normal curves, control charts, and box plots) as well as parametric and nonparametric analytical techniques. We will code implausible values as missing. A minimum of five days of actigraphy data will be required to characterize sleep reliably (Sadeh, 2011).

To address aim 1, descriptive statistics will be used to summarize each of the self-report (diary, questionnaire) and objective (actigraphy, CGM, neurocognitive tests, chart data) variables. Objective sleep characteristics will be described including total sleep time, sleep onset latency, sleep efficiency, wake after sleep onset, and sleep fragmentation index over the 2-week interval. To evaluate the extent to which there are differences in weekday vs. weekend total sleep time, sleep variability or glucose variability, the mean within-participant differences between weekdays (Sunday through Thursday) and weekends (Friday and Saturday) will be calculated (Cohen’s D effect size) and the difference will be tested using both parametric (paired t-test) and nonparametric (Wilcoxon Rank Sum test) tests.

Participants are asked if they had any issues with the devices during the study or with any study procedures, if the sleep report data look accurate, and facilitators and barriers to obtaining sufficient sleep during an exit interview. Verbal and written qualitative data from the diaries and exit interviews will be explored for themes using qualitative description (Sandelowski, 2010). This method is preferred in a beginning analysis as it allows the researcher to stay close to the data and to events (Sandelowski, 2010). We will use criteria for determining validity in qualitative research described by Whittemore, Chase, and Mandle (2001). The four main criteria of trustworthiness, credibility, transferability, dependability, and confirmability will be used throughout our systematic process (Lincoln & Guba, 1985). The analysis will be inductive and begin during data collection to allow for ongoing modification of the interview guide. Interviews will be coded using an in vivo approach (D. Thomas, 2006). Sampling will continue until data redundancy is reached (approximately 20–30 participants). We will use an audit trail to track decision-making and to triangulate the data. Meaning units will be identified and condensed. The condensed meaning units will then be abstracted based on codes that emerge from the data. Two investigators will use a conceptual coding method to develop the final coding structure by synthesizing and collapsing the preliminary in vivo codes until consensus is reached. Codes will be collapsed into categories, and categories will be collapsed into themes and subthemes.

To address aim 2, we will use bivariate correlations to explore the relationships among participant-level summaries of sleep deficiency, neurocognitive function, DSM, DQOL, glycemic control (A1C) and glucose variability (CGM data) between subjects. Descriptive statistics will be used to summarize each of the variables including the scores for multi-item scales. Using actigraphy data, sleep deficiency will be calculated from total sleep time, and sleep variability will be calculated using the Mean of Squared Successive Differences of total sleep time across the 14 nights, representing the variation within-subjects in sleep night-to-night. This approach for analyzing sleep variability has been used in prior research (Meltzer, Sanchez-Ortuno, Edinger, & Avis, 2015). A1C levels will be used for glycemic control, and time in range and J index from CGM data will be used for glucose variability (Suh & Kim, 2015). A series of t-tests, ANOVA, and correlations will be conducted to determine differences in sleep, DSM, neurocognitive function, glycemic control, and glucose variability by age, gender, race, income, and BMI (potential covariates). If sample sizes permit, additional covariates will include household composition/size, student and employment status, and bed partner status.

To address aim 3, we will use a series of multilevel models to evaluate the extent to which daily variations in objective sleep characteristics (bed/wake times, total sleep time, sleep efficiency, wake after sleep onset, awakenings, and sleep fragmentation index) predict subsequent glucose variability (J index, high and low blood glucose risk indices, and time in range), as well as the extent to which daily variations in glucose are associated with subsequent sleep characteristics within-subjects. This analytic approach is analogous to what was used by Fucito and colleagues (2018). Daily glucose reports will comprise level-1 variables nested within person at level-2. Models will be created using all available data, and missing level-1 data will be handled using full information maximum likelihood estimation. Any missing level-2 data, e.g., participant age or race, will be filled in via sequential regression multivariate imputation (Raghunathan et al. 2001); to date, we have no missing data for participant-level measures. Models will include a random intercept and random effects for predictors (sleep and glucose variables) to allow regression coefficients to vary across individuals. A zero-order autoregressive structure will be used to place the fewest restrictions on the models and allow variances and covariances to be freely estimated from the data. Models may include a contrast indicator of weekend = 1 and weekday = 0 at level 1 and may include gender or other covariates (e.g., duration of diabetes, BMI) if applicable.

Discussion

Sleep deficiency is a major barrier to effective DSM and is associated both directly and indirectly, through DSM, with glycemic control and glucose variability in emerging adults with T1D. The present study will provide important information to support future research to test our central hypothesis that more severe sleep deficiency is associated with poorer neurocognitive function, poorer DSM, poorer DQOL, and poorer glycemic control, and higher glucose variability, and that there is a reciprocal relationship between sleep deficiency and glucose variability among emerging adults with T1D. We expect to use the results first to adapt and then test a sleep self-management intervention tailored for this high-risk population.

This study is innovative in three ways. The primary innovation is in our use of concurrent objective measures of actigraphy for sleep and CGM for glucose over 14 days. This approach will allow us to see habitual sleep and glucose patterns and determine weekday and weekend differences considering that sleep extension on weekends has been noted in our population (Barone et al., 2015). We will also gain insight into the direction of the associations between sleep and glucose levels. We plan to determine the extent to which sleep deficiency predicts glucose variability the next day and how glucose variability is associated with sleep the following night. Second, our use of rigorous objective measures to assess neurocognitive function (PVT and Trail A and B) will allow us to capture aspects of objectively measured fatigue that are often not picked up with self-report measures. Third, our focus on early emerging adults will allow us to capture this high-risk period. In previous studies, a broader range of ages have been included, and this range can underestimate the associations between sleep deficiency and glycemic control or glucose variability.

Progress to Date

The protocol elements are feasible based on our work to date. To determine the feasibility of characterizing sleep, neurocognitive function (executive function, psychomotor vigilance), DSM, DQOL, glycemic control, and glucose variability among emerging adults with T1D, we recruited emerging adults with T1D from the Yale-New Haven Health System over a period of 6 months. Out of 138 identified in the Yale system through MyChart™ who met inclusion criteria, 50 expressed interest, 16 declined, and 60 read the message but did not respond. Of those who expressed interest, 40 were contacted after excluding six for sleep apnea, one for a comorbid chronic medical condition, two for major psychiatric illness, and one for pregnancy. Thirty-five were screened over the phone and four additional people were excluded with two screening high-risk for sleep apnea and the other two self-identifying as in early pregnancy. Those who declined to participate did so due to the time commitment of the study, a lack of interest, or refusal/inability to wear a CGM. Twenty-one who met inclusion criteria completed the study successfully over a period of 6 months thus far. Only 19% (n = 4) did not currently have a CGM; however, all reported having a CGM in the past and were familiar with the technology. The study procedures were well tolerated by participants with 6–9 (M = 7.43 ± 0.63) days/nights of actigraph data at 100% (n = 21) and 6–14 days (M = 8.7 ± 3.4) days/nights of CGM data at 95.2% (n = 20) and exit interviews at 85.7% (n = 18). We anticipate reaching our target enrollment with a study period over 12–18 months based on our previous rate of recruitment. Thus far, participants have been recruited through MyChart™ (90.5%), referral (4.8%), and face to face (4.8%).

Lessons Learned

We have learned a number of lessons that will inform our future work in this project. We believe our approach has facilitated participation and successful data acquisition. The procedures and time commitments for the one in-person visit and phone exit interview have been well tolerated. A strength is that all other procedures are conducted remotely via text, email, or phone, depending on the participant’s preferred method of contact. Most of the emerging adults in our study prefer to communicate via text and reminders about the study (e.g., first day check in, reminder to send equipment back) are best communicated via text. We structured the visits to include all self-report baseline measures and the neurocognitive tests during the in-person encounter. Participants typically complete the baseline survey during this time; however, they are provided the option to complete them at home if there is limited time for the visit with priority placed on the objective neurocognitive testing. These activities are all completed prior to wearing the actigraph and CGM for the next 14 days. For participants who have their own CGM, strategies and instructions for data sharing in REDCap™ are covered in the in-person encounter. Also, there is a queue for twice-daily surveys in REDCap,™ and it works best for emerging adults to work from one link sent via email compared to having twice-daily reminders sent via text or email. Twice daily reminders tend to be burdensome for participants, and once they acclimate to study procedures (typically 1–2 days), there is not a need for further reminders to complete electronic diaries.

We began our study by measuring sleep and glucose for seven days to capture weekday and weekend differences. We revised the protocol to collect data over 14-days to be in line with ADA consensus statements (ADA, 2020) based on our experience that procedures were well tolerated. When we designed the study, we were concerned about the subject burden of using CGMs due to lower usage in this age group, and the potential for sleep disruption was also a concern. CGM use increased in this population nationally from 7% in 2010–2012 to 30% in 2016–2018 (Foster et al., 2019; Miller et al., 2015). We also learned the importance of programming the provided CGMs in a blinded mode so that they will not alarm for low or high glucose to help decrease participants’ day-to-day burden and reduce the potential for intervention effects with the introduction of the CGM (Muchmore, Sharp, & Vaughn, 2011). Participants continue to use their own intermittent glucose meter (4x/day and as needed by fingerstick). Participants with their own CGM continue to use their device unblinded as mentioned previously. Capturing and downloading CGM data sometimes poses a challenge. CGMs need to be calibrated; sensors need to be replaced; and there may be difficulty in uploading/downloading data for retrospective analysis. We ask participants to provide data in a .csv or .xls format so that data can be analyzed using the Glyculator 2.0™ software.

We provide a simple instruction sheet for the actigraph to reinforce care of the device, the button to push to demarcate “lights out” and “lights on”, and pictures of the screen for troubleshooting. At the end of the 14-days, we conduct phone interviews to collect quantitative and qualitative data about the tolerability of study procedures and to go over sleep reports with each person to ensure the data look accurate and to share feedback about the sleep variables. The sleep reports are reports for clinicians generated from Actiware™ software with summary statistics, daily statistics, and a visual actogram. We add recommendations from the National Sleep Foundation for total sleep time, sleep efficiency, wake after sleep onset, sleep fragmentation, and light exposure to the report in the notes section. Participants are particularly interested in the recommendations and actograms depicted in the reports. We ask about their sleep goals along with barriers and facilitators of a good night’s sleep during the interview. We refer participants who request assistance with sleep to their primary care providers. We acknowledge their interest in supporting sleep health and remind them that obtaining knowledge about sleep is a benefit of the study.

Summary

The focus of standard type 1 diabetes care is on diet, glucose monitoring, insulin administration, and physical activity (ADA, 2020) and not on sleep, despite the extensive literature on sleep’s role in glucoregulation and the performance of complex neurocognitive skills (e.g., psychomotor vigilance and executive function) that foster better DSM. Results from this study will build on the limited literature on sleep characteristics, self-management, neurocognitive function, and glycemic control in emerging adults with T1D. Results will be used to inform future development and testing of sleep self-management interventions tailored for emerging adults with T1D with a goal to improve DSM, neurocognitive function, glycemic control and, ultimately, long-term clinical outcomes, in this high-risk age group.