The Impact of Fear of Hypoglycaemia on Sleep in Adolescents with Type I Diabetes

Abstract

Aims

Fear of hypoglycaemia (FOH) can contribute to impaired sleep for adults with Type 1 Diabetes (T1D) and parents of children with T1D, although it is unknown how fear of hypoglycaemia may affect sleep for adolescents with T1D. This study examines the relationship between adolescent FOH and sleep and assessed the influences of continuous glucose monitor (CGM) and insulin pump use.

Methods

Adolescents ages 14–18 years with T1D completed questionnaires evaluating FOH (Child Hypoglycemia Fear Survey) and sleep (Pittsburgh Sleep Quality Index, PSQI). Analyses included linear and logistic regression, t-tests, and Fisher’s exact tests.

Results

Participants included 95 adolescents (52 female) with a median (IQR) age of 16.5 (15.3–17.7) years and a T1D duration of 5.7 (2.5–9.6) years. Analyses showed increased FOH-Worry subscale scores were associated with reduced sleep duration (β = −0.03, p = 0.042, adjusting for BMI z-score, race and ethnicity) and increased sleep disturbances (OR = 1.1, p = 0.038, adjusting for race and ethnicity). Frequent CGM users had longer sleep duration (average 7.5 hours) compared to infrequent or non-CGM users (average = 6.8 hours; p = 0.029), and pump users had overall improved sleep health as determined by PSQI score (p=0.019). Technology use did not have significant interactions in the relationships between FOH and sleep duration or sleep disturbances.

Conclusions

Worry about hypoglycaemia was associated with impaired sleep for adolescents with T1D. Diabetes technology users have some sleep improvements, but CGM and pump use do little to alter the relationship between FOH and sleep outcomes.

Introduction

Adolescents with type 1 diabetes (T1D) are predisposed to poor sleep health,¹ which includes sleep duration of less than the American Academy of Sleep Medicine recommended 8–10 hours of sleep per night,^2–3 suboptimal sleep efficiency (the percentage of time spent asleep while in bed)³ and increased sleep disturbances compared to healthy age-matched controls.⁴ Moreover, roughly two-thirds of adolescents with T1D report poor sleep quality.⁵ Impaired sleep may contribute to the poorer glycaemic control observed in adolescents with T1D compared to all other age groups,⁶ as worse sleep quality, shorter sleep duration and increased sleep disturbances have been associated with poor diabetes self-management and inferior glycaemic control.^7–9 For adults with T1D, sleep restriction contributes to reduced insulin sensitivity, which may be particularly concerning for adolescents with T1D during puberty, a period of increased insulin resistance.^10,11 Depression and anxiety in healthy adolescents have been linked to poor sleep quality;¹² adolescents with T1D also have elevated rates of depression and anxiety,¹³ likely worsening their overall susceptibility to poor sleep quality.

Fear of hypoglycaemia (FOH) may further contribute to impaired sleep for adolescents with T1D.^14–15 Hypoglycaemia can occur with controlled insulin therapy, and individuals with T1D can develop significant FOH due to frequent, unpredictable, and symptomatic hypoglycaemic events. Parents of children with T1D report their greatest FOH during their child’s sleep, reducing both parental and child sleep quality.^{5, 16–18} In adults with T1D, FOH has been associated with reduced sleep quality.¹⁹ Despite the secular trend toward adolescents assuming increasing self-management responsibilities for their T1D, the effect of FOH on sleep has not yet been fully assessed in adolescents themselves.

Considering the greater precision in diabetes management that technology affords, continuous glucose monitors (CGMs) and insulin pumps have the potential to reduce FOH and improve sleep, although the data in this area are inconsistent. In qualitative interviews with adults with T1D and parents of children with T1D, 81% reported that the use of CGM led them to sleep better and have less awakenings through an increased feeling of safety.²⁰ However, 10% reported detrimental sleep effects due to alarms and technical issues.²⁰ In a cross-sectional study of 154 adolescents and young adults with T1D, greater sleep disturbances were reported with CGM use.²¹ In addition, a study from the T1D exchange involving 515 parents of children with T1D demonstrated no association between sleep quality or duration with CGM or pump use.⁵ As diabetes technology improves, the relationship between sleep and diabetes technology requires further evaluation.

The primary aim of this study was to assess the relationships between FOH and sleep health in adolescents with T1D, using total sleep health, sleep duration and sleep disturbances as primary outcomes. Our secondary aim was to determine whether CGM and pump use affects sleep and whether technology use alters the relationship between FOH and sleep health. Our hypothesis was that adolescent FOH would be associated with sleep impairments and that this association would be attenuated among those using diabetes technology.

Materials and Methods

The study was reviewed and approved by the Institutional Review Board of the Children’s Hospital of Philadelphia (protocol 17–013945). Informed consent was obtained from all participants.

Design and Participants

Adolescents with T1D attending the Diabetes Center of the Children’s Hospital of Philadelphia in Philadelphia, Pennsylvania, between May 2018 and September 2019, were consented and recruited to participate in this cross-sectional study. The study’s inclusion criteria were as follows: age 14–18 years old, diagnosis of T1D with one or more positive diabetes autoantibodies (anti-GAD, anti-IA2, anti-insulin, or anti-ZNT8), and diabetes duration of greater than 1 year (to avoid the possible contribution of heightened FOH from a recent diagnosis and honeymoon period). Participants were excluded if they were primarily non-English speaking, had systemic glucocorticoid treatment within the previous month (due to the potential confounding effects on glycaemic control), were known to be pregnant, or had developmental disorders that might affect questionnaire completion. Participants were pre-screened for eligibility and approached while awaiting their outpatient appointment. This study’s aims were secondary analyses as part of a larger study and goal recruitment of 100 participants was determined based on the original study’s primary aim.

Questionnaires

Adolescents completed questionnaires on an iPad (Apple Inc., Cupertino, CA) while in clinic. Survey data were collated using REDCap (Research Electronic Data Capture; Vanderbilt University, Nashville, Tenn., USA).

FOH was measured using the Children’s Hypoglycemia Fear Scale (CHFS), which has been validated in children aged 6–18 years.²² This scale has two subscales: 1) the FOH-Behavior subscale, which measures behaviours used to avoid hypoglycaemia with 10 items and 2) the FOH-Worry subscale, which measures anxiety-provoking aspects of hypoglycaemia with 15 items.²² Questions were answered on a Likert scale format (1 = never to 5 = always), and scores were tallied for the subscales and total score as per similar studies assessing parental FOH^23,24 (range 25–125). Higher scores indicate a greater FOH.^22–24

Sleep health was measured using the Pittsburgh Sleep Quality Index (PSQI) questionnaire^25,26 which assesses seven categories of sleep health: subjective sleep quality, sleep latency (time it takes to fall asleep each night when attempting to sleep), sleep duration, sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction. The global PSQI score is the measure of total sleep health and is the sum of the 7 component scores, each scored 0 (no difficulty) to 3 (severe difficulty), with a lower score indicating better sleep quality.²⁵ A global PSQI score of >5 is correlated with other measures of poor sleep quality, including clinical diagnosis of insomnia and obstructive sleep apnoea.^27,28 Sleep duration was obtained directly from a PSQI question.²⁵ The sleep disturbances score was derived from the PSQI question “During the past month, how often have you had trouble sleeping because you…” with subsequent sub-questions with types of sleep disturbances. Participants rate frequency of these sleep disturbances based on a four-option scale. These sleep disturbance scores are added up for the first nine sleep disturbances listed (sum 0–27) and based on the sum, the sleep disturbance score is put into category score of 0, 1, 2, or 3 (0 is 0, 1–9 is 1, 10–18 is 2, and 19–27 is 3).²⁵

Medical Record Review

Medical records were reviewed to extract demographic and clinical information corresponding to the office visit closest to the date that the questionnaires were completed. Data included: date of birth, date of diagnosis, gender, race, ethnicity, insurance type, anthropometrics (including weight Z-score, height Z-score, and BMI Z-score), haemoglobin A1c (HbA1c), start date of current CGM or insulin pump, and percent time of CGM use, if applicable. Regarding CGM use, we examined CGM data from each participant’s last medical visit, and participants were grouped into those who used the device the majority of the time (CGM use ≥50%) and those who did not (no CGM or used it <50% of the time) since these were the groupings that were available in the medical visit documentation. Participants were also identified as pump users or non-pump users based on documentation from last medical visit.

Data Collection and Statistical Analyses

Summary statistics were reported using medians and interquartile ranges. Linear regression analyses were used to identify associations between FOH score and continuous sleep outcomes. The Shapiro-Wilk test was used to determine normality of sleep outcome variables. Non-normally distributed sleep outcomes were transformed to become normalised using Tukey’s ladder of powers to decide an appropriate transformation. Sleep duration was the only sleep variable fitting a normal distribution. Sleep health was square root-transformed to be made normally distributed to be used in linear regression. To assess sleep outcomes based on standard sleep recommendations, two continuous sleep outcomes were also transformed into binary variables (“sufficient sleep duration” defined as ≥ 8 hours² and “appropriate sleep health” defined as global PSQI score ≤ 5^25,26). Logistic regression analyses were used to determine the associations between FOH scores and binary sleep variables. As sleep disturbances was a categorical variable with multiple score categories, it was thus evaluated with ordinal logistic regression. Demographic and health variables were included as covariates in regression analyses if associated with their respective sleep outcomes on linear or logistic regression analyses.

Sleep variables were compared between CGM or pump users versus non-users: normal continuous variables were compared using unequal t-tests and Fisher’s exact test was used for categorical sleep variables. Regressions investigating the relationship between FOH and sleep outcomes adjusting for covariates were performed with an interaction analysis with technology use to determine if an interaction was present between FOH and CGM or pump use. Additionally, regressions adjusting for covariates were stratified by CGM and pump use to assess for associations between FOH and sleep outcomes by diabetes technology use. The interaction and stratification analyses were performed only if found to have significant associations between FOH and sleep parameters. Statistical analysis was performed using Stata/IC Version 15 (StataCorp, College Station, TX, USA).

Results

Survey Respondent Characteristics

The adolescents who completed the study (n=95) were predominantly female (55%), of non-Hispanic White race and ethnicity (65%) and had private health insurance (80%) (Table 1). Median age was 16.5 years, T1D duration was 5.7 years, and HbA1c was 66 mmol/mol (8.2%) (Table 1). Further, 73% had used CGMs with 45% using CGM ≥ 50% of the time, and 63% were using insulin pumps (with 5% of participants using low glucose suspend or hybrid closed loop technology) (Table 1). Participants’ median (IQR) FOH scores were: Total score = 61 (54, 70), FOH-Worry = 30 (26, 37), FOH-Behavior = 30 (26, 35) (Table 2). FOH-Worry scores were significantly higher in those with government insurance (median = 36) compared to private insurance (median = 29; p = 0.008) and higher in individuals identifying as Hispanic (median = 45) compared to White participants (median = 29; p = 0.032). FOH total score and subscale scores were not significantly associated with participants’ age, diabetes duration, BMI z-score, or HbA1c. Participants reported median sleep duration of 7 hours with 61% sleeping < 8 hours per night (Table 2). Most participants (87%) reported some degree of sleep disturbances, and based on the distribution of the sleep disturbance scores with only one participant scoring a 3, the sleep disturbance scores of 2–3 were combined into one category for analysis (Table 2). Roughly half of participants (47%) had overall poor sleep health (i.e., PSQI score > 5; Table 2). Sleep duration was negatively associated with BMI z-score (β = −0.44 [95% CI: −0.80, −0.08], p = 0.017) and Black participants had significantly shorter sleep (median = 6 hours) compared to White participants (median = 7 hours; β = −0.76 [95% CI: −1.5, 0.00], p = 0.050), so BMI z-score, race, and ethnicity were included in regression analyses for sleep duration. Additionally, Black participants were more likely to have ratings in a higher sleep disturbance category when compared to White participants (OR = 3.4 [95% CI: 1.2, 9.9], p = 0.025) and Black participants had worse sleep health (median global PSQI score = 8) compared to White participants (median global PSQI score = 5; β = −0.48 [95% CI: 0.08, 0.87], p = 0.020), so race and ethnicity were included in regression analyses for sleep disturbances and sleep health.

Table 1.

Characteristics of Survey Respondents

	PARTICIPANTS (n, %)
Sample Size	95
Female	52 (54.7%)
Race/ethnicity
Non-Hispanic White	62 (65.3%)
Non-Hispanic Black	24 (25.3%)
Hispanic	6 (6.3%)
Non-Hispanic Asian	1 (1.1%)
Undisclosed	2 (2.1%)
Private Insurance	76 (80.0%)
CGM Use	69 (72.6%)
CGM Use > 50% of the time	43 (45.3%)
Pump Use	60 (63.2%)
Use of low-glucose suspend pump	2 (2.1%)
Use of hybrid closed loop pump	3 (3.2%)
PARTICIPANT CHARACTERISTICS	Median (IQR)
Age (years)	16.5 (15.3–17.7)
Duration of Diabetes (years)	5.7 (2.5–9.6)
HbA1c (mmol/mol [%])	66 [8.2] (57 [7.4] −88 [10.2])
BMI (Z-score)	0.94 (0.29–1.57)
Duration of Current CGM Use (months)	11 (6–24)
Duration of Current Pump Use (months)	15 (8–32.5)

Table 2.

Fear of Hypoglycaemia (FOH) and Sleep Characteristics

CHARACTERISTICS FOR CONTINUOUS VARIABLES	Median (IQR)
FOH Scores (HFS total score, range 25–125)*	61 (54–70)
HFS-Worry Score (range 15–75)*	30 (26–37)
HFS-Behavior Score (range 10–50)*	30 (26–35)
Sleep Health Score (range 0–21)**	6 (3–10)
Sleep Duration (hours)	7 (6–8)
Sleep Efficiency (%)	89.4 (80–100)
CHARACTERISTICS FOR CATEGORICAL VARIABLES	PARTICIPANTS (n, %)
Sufficient Sleep Duration (≥ 8 hours)	37 (38.9%)
Appropriate Sleep Health (global PS QI score ≤5)	50 (52.6%)
Sleep Disturbance Score
None (Score 0)	12 (12.6%)
Moderate (Score 1)	67 (70.5%)
Severe (Score 2–3)	16 (16.8%)

^*Higher scores reflect worse FOH

^**Measured as global Pittsburgh Sleep Quality Index score. Higher scores reflect worse sleep health.

FOH with Primary Sleep Measures: Sleep Health, Sleep Duration and Sleep Disturbances

In analyses prior to adjusting for covariates, FOH was not associated with total sleep health as measured by the global PSQI score in linear or logistic regression. However, increasing FOH was associated with reduced sleep duration (β = −0.03 [95% CI: −0.06, −0.01], p = 0.011) and increased sleep disturbances (OR = 1.0 [95% CI: 1.0, 1.1], p = 0.022). In addition, prior to adjusting for covariates, the FOH-Worry subscale was associated with reduced sleep duration (β = −0.05 [95% CI: −0.08, −0.02], p = 0.003), lower odds of sufficient sleep duration (OR = 0.95 [95% CI: 0.91, 1.0], p = 0.045), increased sleep disturbances (OR = 1.1 [95% CI: 1.0, 1.1], p = 0.007), and with worse square root-transformed sleep health by linear regression (β = 0.02 [95% CI: 0.01, 0.04], p = 0.01). The FOH-Behavior subscale was not significantly associated with total sleep health, sleep duration or sleep disturbances. Adjusting for race, ethnicity, and BMI z-score in linear regression, the FOH-Worry subscale remained associated with reduced sleep duration (β = −0.03 [95% CI: −0.06, 0.00], p = 0.042) (Table 3). Adjusting for race and ethnicity in ordinal logistic regression, the FOH-Worry subscale was associated with increased odds of sleep disturbances (OR = 1.1 [95% CI: 1.0, 1.1], p = 0.038) (Table 3). FOH-Worry did not remain associated with sleep health (global PSQI score) when adjusting for race and ethnicity.

Table 3.

Results of Regressions between Fear of Hypoglycaemia Total and Worry Scores and Sleep Duration and Sleep Disturbances Adjusted for Potential Confounders

Outcome	Primary Exposure	β	95% CI	P-value
Sleep Duration (hours)	FOH-Worry Score	−0.03	−0.06, 0.00	0.042
Sleep Duration (hours)	FOH-Total Score	−0.02	−0.05, 0.00	0.063
		odds ratio (OR)	95% CI	P-value
Sleep Disturbances Score	FOH-Worry Score	1.1	1.0, 1.1	0.038
Sleep Disturbances Score	FOH-Total Score	1.0	1.0, 1.1	0.069

Associations determined by regression modeling as following:

• Linear regression models for sleep duration adjusting for BMI z-score, race and ethnicity
• Ordinal logistic regression models for sleep disturbances adjusting for race and ethnicity

Abbreviations:

• FOH = Fear of hypoglycaemia
• CI = Confidence interval
• OR = Odds ratio
• β = Beta coefficient

Sleep Relationships with Technology Use

When comparing sleep outcomes between technology users and non-users, those on insulin pumps had overall better sleep health (median PSQI score in pump users = 5 vs. median PSQI score in non-pump users = 7; p = 0.019). Sleep duration was higher among those participants using CGM > 50% of the time compared to those without CGM use or using CGM < 50% of the time (mean 7.5 hours in CGM users vs. mean 6.8 hours in non- or minimal CGM users, p = 0.029). Sleep disturbances did not differ between CGM or pump users versus non-users.

There was no significant interaction between FOH and the use of diabetes technologies (CGM or pump use) with respect to any sleep outcomes. However, regression analyses adjusting for covariates when stratified by diabetes technology use revealed that the FOH-Total score was only associated with lower sleep duration in those without CGM use or with CGM use <50% of the time (β = −0.03 [95% CI: −0.06, −0.00], p = 0.049).

Discussion

This study examined the relationship between FOH and various sleep parameters among adolescents with T1D. Consistent with our hypothesis, increased FOH was associated with reduced sleep duration and increased sleep disturbances among adolescents with T1D. These associations were strongest in those scoring high on the FOH ‘worry’ subscale. We also found a relationship between worse overall sleep health (as measured by the total score of the Pittsburgh Sleep Quality Index) only with higher scores on the FOH ‘worry’ subscale. Conversely, there were no associations between hypoglycaemia avoidance behaviours and any aspects of sleep that we measured. These novel findings demonstrate that FOH negatively influences sleep outcomes in adolescents with T1D, congruent with reports in adults with T1D as well as parents of young children with T1D.^5,16–19 Moreover, this study is the first to suggest that, in adolescents with T1D, worry about hypoglycaemia alone appears to relate to abbreviated and more disrupted sleep. It is reassuring that the current findings complement prior studies on parents of children with T1D, which demonstrated that despite parental FOH’s correlation with inferior child and parent sleep quality, behaviours which may be appropriate to prevent hypoglycaemia (e.g. insulin reductions when concerned about trending towards hypoglycaemia, ensuring hypoglycaemia treatment is available, or even increased frequency of nocturnal blood glucose monitoring) were not associated with reduced sleep quality or duration.⁵ The present findings demonstrate the importance of screening for and addressing adolescents’ worries about nocturnal hypoglycaemia when managing sleep difficulties in T1D.

The second objective of this study was to identify the potentially mitigating influence of diabetes technology use on sleep duration and sleep disturbances and their relationships with FOH. We observed that pump users had better overall sleep health, while those effectively using CGM had on average 42 minutes longer sleep duration than those who were not, which has been shown in prior studies to be a sufficient difference to affect daytime sleepiness, blood pressure and insulin sensitivity.^27,28 Although the more robust interaction analysis did not reveal a statistical interaction between FOH and diabetes technology, stratified analyses did reveal that the association between reduced sleep duration and greater worry about hypoglycaemia was only detected among those not using CGM more than 50% of the time (a reasonable surrogate of effective use). We suspect that the interaction analysis was not adequately powered to detect an interaction. Based upon our stratified analyses, there is an indication that diabetes technology use plays a role in ameliorating the negative effects of FOH on sleep, and future research should be sufficiently powered to examine this question. Diabetes technology use in adolescents with T1D may reduce FOH-related anxiety and its effects on sleep by providing a feeling of increased safety, which is supported by prior qualitative studies.²⁰ Contrary to other studies^20,21, we did not observe any associations between technology use and increased sleep disturbances. This may be due to the newer CGM and pump models used by our participants compared to prior studies, as well as our population’s increased familiarity with technology use (average CGM use: 14 months, average pump use: 25 months).

The results of this study highlight the importance of addressing FOH-related anxiety and its effects on sleep in adolescents with T1D. Cognitive behavioural therapy (CBT) has been shown to reduce FOH in adults with T1D but results of these trials are inconsistent.²⁹ CBT has also been shown to improve sleep health in adolescents overall.¹² Intensive blood glucose and hypoglycaemia awareness training programmes have successfully reduced FOH in adults with T1D²⁹ but remain largely unexamined in adolescents with T1D and high FOH. As the current study suggests, CGM and insulin pumps may benefit adolescents with high levels of FOH and associated sleep difficulties. Further research is needed to define this relationship as well as which individual or technology-related aspects may contribute the most to FOH-related sleep health. For example, hybrid closed-loop therapy (i.e. pumps regulated by CGM) holds the promise of significantly reducing FOH, as did an insulin pump with a low glucose suspend in a recent clinical trial of 144 adolescents with T1D.³⁰

The strengths of this study include use of a well characterised clinical sample of adolescents with T1D from a large diabetes centre. This study is limited by its cross-sectional design, as a longitudinal design would be required to fully understand the evolution of the relationship between FOH and sleep impairment, as well as the effects of diabetes technology introduction. The use of a questionnaire to measure sleep impairment relies on subjective report and leaves out the quantification of naps that could improve overall 24-hour sleep. Future studies should employ actigraphy and/or polysomnography to measure sleep parameters more objectively. This study was also limited by CGM use metrics available in chart review as thresholds other than >50% use may be more appropriate to indicate frequent CGM use. Future studies might also seek to determine the threshold in FOH measurements that is needed to negatively impact sleep, to decide when clinical interventions to reduce FOH should be introduced. Lastly, future work should screen for mental health disorders, as it is possible that generalised anxiety or mood symptoms, not assessed in this study, may have affected sleep health in our population.^12–13

Conclusions

Increased worry about FOH, as opposed to behaviours undertaken to reduce FOH, is specifically associated with reduced sleep duration and increased sleep disturbances in adolescents with T1D. Diabetes technology use is helpful in improving sleep outcomes, as evidenced by better sleep health for pump users and longer sleep duration for participants with CGM use > 50% of the time. Diabetes technology use may ameliorate the negative effects of FOH on sleep, and future research should further examine this question.

Novelty statement.

• What is already known?

  • Adolescents with type 1 diabetes (T1D) have poor sleep health.
  • Fear of hypoglycaemia (FOH) is related to impairments in sleep for adults with T1D and parents of children with T1D.
• What this study has found?

  • Worry-specific FOH is related to shorter sleep duration and increased sleep disturbances in adolescents with T1D.
  • Behaviours to avoid hypoglycaemia were not related to sleep health.
  • Diabetes technology use was related to some sleep benefits, but did not alter the relationship between FOH and worsened sleep.
• What are the implications of the study?

  • This study emphasizes the importance of evaluating and addressing FOH in adolescents with T1D.
  • Diabetes technologies did not alleviate the negative effects of nocturnal FOH on sleep.

Data Availability Statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.