Sleep in children with type 1 diabetes and their parents in the T1D Exchange.

Introduction

Sleep has physiologic and behavioral impacts on many health outcomes. Sleep disturbances, which include bedtime resistance or difficulty initiating sleep, night wakings, and insufficient sleep [1], are highly prevalent in children, occurring in 20%–30% of the general population [2], and a recent meta-analysis found that children and adolescents with type 1 diabetes had significantly shorter sleep duration than controls [3]. For children with type 1 diabetes, sleep disruptions can be the result of night wakings due to hypo/hyperglycemia and parental nocturnal diabetes caregiving behaviors [4]. Nevertheless, sleep characteristics, such as total sleep time, sleep/wake times, and sleep quality, are not routinely addressed in standards of care for youth with type 1 diabetes [5,6].

Accumulating evidence indicates that short sleep duration and poor quality sleep contribute to problems with glycemic control and adherence in adolescents and adults with type 1 diabetes [7,8]. The effect of sleep on glycemic control likely occurs through a direct physiological pathway (decreased insulin sensitivity) [9–12] and an indirect behavioral pathway (insufficient sleep compromises the cognitive functions needed to effectively manage diabetes) [13]. The only study to examine sleep in youth with type 1 diabetes used actigraphy and polysomnography and found that adolescents with type 1 diabetes (n = 40, aged 10–16) spent less time in slow wave sleep than matched controls. Adolescents with reduced slow wave sleep had worse glycemic control and poorer self-reported quality of life [8]. In this study, poor sleep habits were also associated with behavior problems and academic difficulties, but sleep was not examined in relation to adherence. In a more recent study of adolescents and young adults with type 1 diabetes (aged 13–20), shorter self-reported sleep duration was significantly associated with a lower frequency of blood glucose checks, and sleep quality was significantly related to glycemic control for males but not for females [14]. In a smaller study of adolescents (n = 45), sleep duration was significantly associated with adherence, measured with frequency of blood glucose monitoring (BGM) and insulin bolus [15]. While these studies generally support the association between sleep disturbances and glycemic control, they have not examined the role of sleep on adherence to diabetes management in younger children or the influence of diabetes-related technology, such as continuous glucose monitoring (CGM) or insulin pumps on sleep quality.

Characterizing sleep patterns and disturbances and their impacts in youth with type 1 diabetes will inform future studies designed to improve diabetes management and diabetes outcomes. Previous studies of sleep in type 1 diabetes have been limited by small sample sizes and narrow age ranges, and children with type 1 diabetes are likely to experience unique sleep disturbances related to nocturnal caregiving. We used validated survey measures to evaluate sleep patterns in youth enrolled in the Type 1 Diabetes (T1D) Exchange clinic registry and their parents. In addition, we examined the associations between sleep and diabetes outcomes (hemoglobin A1c [HbA1c], hypoglycemia, and diabetic ketoacidosis [DKA]), use of diabetes-related technology (insulin pump, CGM), and BGM. Finally, we explored relationships between parental well-being, fear of hypoglycemia, nocturnal caregiving, and children’s sleep patterns.

Materials and methods

We used parental surveys to characterize sleep patterns in youth 2–12 years of age with type 1 diabetes and their parents. We administered both the validated Child Sleep Habits Questionnaire (CSHQ) [16] and Pittsburgh Sleep Quality Index (PSQI) [17]. The CSHQ is a parent-reported measure of child sleep habits and sleep disturbances, validated for use in 2–12 years of age [18,19]. CSHQ scores range from 33 to 99, with a higher score indicating greater sleep disturbance. A score of 41 or higher is considered clinically significant. The PSQI is a self-reported measure of sleep quality and was used to assess the parental sleep quality. PSQI scores range from 0 to 21, with a higher score indicating poor sleep quality. A score greater than five indicates a clinically significant sleep disturbance.

In addition to capturing data regarding sleep patterns, we included validated measures for parental fear of hypoglycemia (Hypoglycemia Fear Survey [HFS-P]; range 0–104 with higher score indicating more fear) and parental emotional wellbeing (World Health Organization-Five-Wellbeing Index [WHO-5]; range 0–100 with high score indicating better well-being). Nighttime caregiving habits also were captured from parents and reported as frequency per week of a caregiver checking the child’s blood sugar after bedtime. Information regarding diabetes history and demographic variables (gender, race/ethnicity, parent education, annual household income, and insurance status) were also obtained from the parent. Parents were asked to report on most recent HbA1c value, episodes of severe hypoglycemia (SH), and episodes of DKA over the previous three months.

In addition to self-reported survey data, supplementary demographic and clinical data were obtained from T1D Exchange clinic registry medical chart extraction for survey respondents and non-respondents. Demographic T1D Exchange variables used to compare respondents and non-respondents included the following: age, sex, race/ethnicity, health insurance type, use of insulin pump, CGM use, most recent HbA1c, and age at type 1 diagnosis.

Self-reported HbA1c values were used for the main respondent analysis because a significant number of respondents did not have clinic-reported HbA1c values within six months of survey administration. However, a subanalysis of participants who had recent T1D Exchange clinical HbA1c data (93 (18%) participants) revealed that self-reported HbA1c was only slightly lower than clinic-reported values (median difference 0.1% (inter-quartile range −0.2% to 0.5%)). Further, the difference in self-reported and clinic-reported HbA1c was not associated with CSHQ score or PSQI score (P = 0.76, 0.19, and 0.98, respectively, from separate linear regression models).

The T1D Exchange Clinic Network has enrolled over 30,000 individuals with type 1 diabetes across 74 US-based pediatric and adult endocrinology practices. Details on the eligibility criteria, informed consent process, and data collection for the T1D Exchange Clinic Network have been previously published [20]. Parents of T1D Exchange clinic registry participants who were 2–12 years of age with type 1 diabetes duration of at least one year who consented to receive emailed information regarding T1D Exchange studies were invited to participate through two email waves between June 2015 and September 2015. After signing an institutional review board-approved electronic informed consent form, parents of eligible participants completed the online surveys. Study data were collected and managed using Research Electronic Data Capture (REDCap) tools hosted at the Jaeb Center for Health Research [21]. REDCap is a secure web-based application designed to support data capture for research studies, providing [1] an intuitive interface for validated data entry [2], audit trails for tracking data manipulation and export procedures [3], automated export procedures for seamless data downloads to common statistical packages, and [4] procedures for importing data from external sources. Survey respondents received their choice of a $20 electronic gift card or donation to a diabetes charity.

The study was designed to include a feasibility sample (not based on statistical principles) of 500 participants. Summary statistics were calculated for duration of sleep according to age group (2–4 years old, 5–12 years old, and parents) and for CSHQ score. The proportion of CSHQ scores meeting the definition of poor sleep quality (score >41) was calculated. The association between each of the following parent-reported outcomes and CSHQ score was examined: most recent HbA1c, frequency of BGM, use of insulin pump, CGM use, WHO-5 score, PSQI score, and occurrence of ≥1 SH events and ≥1 DKA events in the three months prior to questionnaire completion. For continuous outcomes (HbA1c, frequency of BGM, WHO-5 score, and PSQI score), separate multivariable linear regression models were used to assess the association with continuous CSHQ score; for categorical outcomes (insulin modality, CGM use, and occurrence of ≥1 SH events and ≥1 DKA events), separate multivariable logistic regression models were used to assess the association with CSHQ score. The associations between CSHQ (outcome) and parental fear of hypoglycemia and nocturnal caregiver behavior were assessed through separate multivariable linear regression models. The association between parental nocturnal caregiver behavior and parental fear of hypoglycemia was examined through an ordinal logistic regression model. Methods assessing the association between outcomes and CSHQ score were repeated to assess the association between outcomes and parental report of child sleep duration.

Summary statistics of the PSQI score were calculated, and the proportion of PSQI scores meeting the definition of poor sleep quality (score >5) was determined. The association between PSQI score and parental fear of hypoglycemia and nocturnal caregiver behavior was assessed through separate multivariable linear regression models.

Results are expressed as mean ± standard deviation for normally distributed variables or median (interquartile range) for non-normally distributed variables. To account for possible confounding, the following covariates were assessed for the association with each outcome through univariate analysis and selection models: race/ethnicity, age, sex, age at T1D diagnosis, insurance status, clinic center, and insulin modality (assessed whether insulin modality was not the outcome of interest). If an association with an outcome was present, the covariate was included in the model for the outcome.

Data analyses were performed using the SAS software version 9.4 (SAS Institute Inc., Cary, NC). In view of the multiple comparisons, only p-values <0.01 were considered significant.

Results

Between June 2015 and September 2015, 2356 parents of eligible participants from 54 pediatric diabetes centers were sent a survey invitation through email; 598 (25%) accessed the survey and 515 (22%) from 50 of the 54 clinics completed the survey (all who completed the survey were analyzed).

The 515 surveys were completed by parents of children who had a mean age of 9 ± 3 years, 240 (47%) children were female, and 442 (86%) were non-Hispanic White. The mean age of the child at diagnosis was 4 ± 2 years. The mean parent-reported HbA1c was 7.8% ± 0.9% (62 ± 9.8 mmol/mol); median number of blood glucose checks per day was 7.0 (6.0, 9.0). Occurrence of at least one SH event in the 3 months prior to questionnaire completion was reported by parents of 15 children (3%), and occurrence of at least one DKA event was reported by parents of 30 children (6%). Additional characteristics of the cohort are shown in Table 1.

Compared with participants who did not complete the survey, participants who completed the survey were more likely to be non-Hispanic White (86% vs. 78%), use an insulin pump (77% vs. 68%), use a CGM (27% vs. 19%), and have lower clinic-reported HbA1c (mean 8.0% vs 8.4% [64 vs 68 mmol/mol]) (Supplemental Table 1).

The mean duration of sleep per night was 10.9 ± 1.2 h in children 2–4 years old, 9.5 ± 1.0 h in children 5–12 years old [22], and 6.5 ± 1.2 h in parents (Supplemental Fig. 1). Among the 511 children with sleep duration reported, 103 (20%) reported sleep duration below the recommended amount (<9 h/night); among 504 parents with sleep duration reported, 259 (51%) reported sleep duration below the recommended amount (<7 h/night) (Table 2).

A total of 346 (67%) children met criteria for poor sleep quality (CSHQ score >41). Among the 501 parents with PSQI score measured, 266 (53%) met criteria for poor sleep quality (score >5). Approximately one-third of parents (163 [32%]) met the WHO-5 criteria for low mood (score ≥50) [23]. Most participants (355 [69%]) often or always had a caregiver check the child’s blood glucose value after the child’s bedtime; only 17 participants (3%) never checked the child’s blood glucose value after bedtime.

Children with poor sleep quality (high CSHQ score) had higher HbA1c than those with non-poor sleep quality (P < 0.001 for continuous score adjusted for race/ethnicity, insurance status, and clinic center) (Fig. 1A). Frequency of BGM per day was not associated with child sleep quality (P = 0.56 adjusted for race/ethnicity, insurance status, diagnosis age, insulin modality, and clinic center). Children with poor sleep quality were more likely to experience at least one SH event in the three months prior to questionnaire completion (P = 0.05 adjusted for race/ethnicity, gender, type of insurance, and HFS score; Fig. 1B) and at least one DKA event in the 3 months prior to questionnaire completion (P < 0.001 adjusted for race/ethnicity, type of insurance, and use of an insulin pump; Fig. 1C). Poor child sleep quality was associated with worse parental wellbeing (P < 0.001 adjusted for race/ethnicity and child age; Fig. 1D) and worse parental sleep quality (P < 0.001 adjusted for race/ethnicity and sex; Fig. 1E). Insulin modality was not associated with child sleep quality (P = 0.13 adjusted for race/ethnicity, sex, insurance status, and diagnosis age); the use of CGM also was not associated with child sleep quality (P = 0.96 adjusted for race/ethnicity, child age, sex, insurance status, age at diagnosis, and insulin modality).

Children of parents with more fear of hypoglycemia (high HFS score) had worse sleep quality than children of parents with less fear of hypoglycemia (P < 0.001 adjusted for race/ethnicity, child age, insurance status, clinic center; Fig. 2A). Parents with more fear of hypoglycemia were more likely to more frequently check their child’s blood glucose level after the child’s bedtime (P = 0.005 adjusted for race/ethnicity, child age, age at diagnosis, and insulin modality). However, child sleep quality was not associated with the frequency of nocturnal checking of blood glucose (P = 0.66; Fig. 2A).

Associations with sleep duration were similar to associations with child sleep quality. Participants with shorter sleep duration had: higher HbA1c (8.0% in participants with <9 h/night vs 7.8% in participants with ≥9 h/night; adjusted P = 0.02), worse parental wellbeing (WHO score 54 for participants with <9 h/night vs 58 with ≥9 h/night; adjusted P = 0.01), and worse parental sleep quality (PSQI score 7 in participants with <9 h/night vs 6 in participants with ≥9 h/night; adjusted P = 0.01). Sleep duration was not associated with frequency of blood glucose checks (adjusted P = 0.66) or CGM use (P = 0.31); however, there was a trend for participants with longer duration of sleep to be more likely to use an insulin pump (68% in participants with <9 h/night vs 82% in participants with ≥9 h/night; adjusted P = 0.06).

Similar to child sleep quality, parents with more fear of hypoglycemia had worse sleep quality than parents with less fear of hypoglycemia (P < 0.001 adjusted for race/ethnicity and sex; Fig. 2B). However, frequency of nocturnal checking of blood glucose was not associated with parental sleep quality (P = 0.35; Fig. 2B).

Discussion

The current study is the largest descriptive study of sleep in children with type 1 diabetes and their parents and yielded several important findings. A significant percentage of children and parents met clinical criteria for sleep disturbances. Further, 20% of children and 50% of parents did not meet the recommended duration for sleep. Child sleep disturbances were significantly associated with glycemic control, such that those who met the clinical criteria for sleep disturbances had significantly poorer glycemic control than those who did not have clinically significant sleep disturbances (parent-reported HbA1c = 7.9% vs 7.6% [63 vs 60 mmol/mol]), and children with poor sleep quality were more likely to have experienced at least one event of SH and DKA. In addition, child sleep disturbances were significantly related to parents’ own reported sleep quality and parental well-being.

As it was in our hypothesis, child sleep quality and duration were significantly related to HbA1c. While the relationship between sleep and glycemic control is likely bidirectional, experimental studies offer support for the effect of sleep restriction or disturbances to sleep quality (suppressing slow wave sleep) on insulin resistance and elevated blood glucose [9,11,12]. However, contrary to our hypothesis, we did not find a significant association between child sleep and a measure of adherence (frequency of BGM). Given that parents are primarily responsible for diabetes management in younger children, it may be that associations between sleep disturbances and adherence will be evident in older individuals, who engaged in more self-management (adolescents and adults). Further, as we only assessed one adherence behavior, it remains to be determined whether other adherence behaviors (eg, mealtime insulin bolus, corrections for hyperglycemia) are associated with sleep quality or sleep duration in children.

Child sleep was also associated with parent sleep and parental well-being, in line with findings from smaller studies of sleep in young children with type 1 diabetes [4,24]. Child sleep disturbances are a source of stress for parents, and parents’ own lack of sleep and poor sleep quality are likely to have a negative impact on their emotional wellbeing [25]. Even though parents’ fear of hypoglycemia was associated with greater nocturnal caregiving and poorer sleep quality in children, there was no association between nocturnal caregiving and child sleep. Further, the lack of association between CGM use and parental or child sleep quality was surprising, given that many parents express interest in using CGM as a way to reduce anxiety around nocturnal hypoglycemia and reduce the need for nocturnal caregiving. It may be that objective measures of sleep are needed to demonstrate the impact of CGM on sleep rather than self-report [26]. Alternatively, it is possible that parents with the greatest fear of hypoglycemia were early adopters of CGM, but given the cross-sectional design of the study, we cannot know whether fear of hypoglycemia changed with CGM use. Finally, the potential for alterations in insulin delivery based on CGM data (ie, closed loop systems) may have greater potential to improve parents’ sleep quality than data alone.

The current study is limited by the use of parent-report measures of child sleep and HbA1c and the cross-sectional design. Future studies should use more objective measures of sleep (actigraphy and polysomnography), longitudinal designs, and consider including a matched sample of children without diabetes. It is also important to note that the HFS-P was validated for use in children aged 8–15, and our sample included children aged 2–12. In addition, given the demographic and clinical differences in parents who completed the survey as compared to the total clinic registry sample, findings may not generalize to other samples with lower use of diabetes-related technology or poorer glycemic control. However, this study was the first to examine sleep in a large, nationally distributed sample of young children with type 1 diabetes.

Conclusions

Sleep is an understudied and potentially important influence on glycemic control, and sleep quality may be a viable target for interventions to improve outcomes in youth with type 1 diabetes. In children without diabetes, behavioral interventions have resulted in significant improvements in sleep disturbance [2]; a recent study demonstrated that an increase of even 15–20 min of sleep was associated with an additional BGM check and additional insulin bolus [15]. Further, sleep disturbances observed in young children are likely to persist into adolescence, so it may be important to intervene at an early age [27]. Given that most children with type 1 diabetes are not meeting treatment goals [28], and that parents of children experience high levels of distress related to diabetes management [29], the current study offers support for targeting sleep as a potentially modifiable risk factor for these outcomes in children with type 1 diabetes.