Contrast Set Mining for Actionable Insights into Associations

Between Sleep and Glucose in a Normoglycemic Population

Hoang Huyen Nhung

and Zilu Liang

1,2 a

Ubiquitous and Personal Computing Lab, Kyoto University of Advanced Science (KUAS), 621-8555 Kyoto, Japan

Institute of Industrial Science, The University of Tokyo, 113-8654 Tokyo, Japan

Keywords: Data Mining, Personal Informatics, Ubiquitous Computing, Contrast Set Mining.

Abstract: Prior studies have suggested potential associations between poor sleep and glucose dysregulation among

diabetic patients. However, little is known about the relationship between sleep and glucose regulation in

healthy populations. In this study, we proposed a data mining pipeline based on contrast set mining to identify

significant associations between sleep and glucose in a dataset collected from a normoglycemic population in

free-living environments. Unlike traditional correlation analysis, our approach does not assume a linear

relationship between sleep and glucose and can potentially discover associations when a pair of metrics fall

within certain value ranges. The data mining result highlights the total sleep time as an important sleep metric

associated with glucose regulation the next day, which is characterised by rules with high lift and confidence.

Furthermore, the result suggests that having a higher time ratio in normal glucose range was associated with

better sleep continuity at night. These results may provide insights that people can immediately act on for

better sleep and better glucose control. Future research may leverage the proposed data mining protocol to

develop healthy behaviour recommender systems.

1 INTRODUCTION

Sleep is an essential part of human daily life. There is

consensus that adults need 7 or more hours of sleep

(Watson et al., 2015). Sleep deprivation was reported

to account for a wide spectrum of health problems,

including glycaemic disturbances and the

development of diabetes (Jiawei et al., 2017; Lou et

al., 2015; Xiao et al., 2014; Zuraikat et al., 2020).

Modern lifestyle has led to significant changes in

human sleep patterns, such as delayed sleep onset and

decreased total sleep time. Furthermore, modern

abundance has also caused significant shifts in

people’s eating habits and has consequently led to a

surge in chronic metabolic diseases such as diabetes,

which is characterised by glucose dysregulation. A

few prior studies have demonstrated a strong

connection between sleep quality and glucose

homeostasis, especially in people with health

conditions (Cauter et al., 1991 Sep; Gottlieb et al.,

2005; Kothari et al., 2021; Lou et al., 2015; Wang et

al., 2017). Nonetheless, those studies were limited by

the methods available for data collection and analysis,

https://orcid.org/0000-0002-2328-5016

and the healthy population is often underrepresented

or completely excluded.

With a variety of consumer wearable technologies

being developed, researchers are now able to

perform longitudinal measurements of sleep and

glucose in a more naturalistic environment.

Multidimensional sleep structure can be measured

using a Fitbit wristband, and 24-hour interstitial

glucose can be monitored using a CGM system such

as the FreeStyle Libre. Data collected with modern

wearable technologies make it possible to examine

the reciprocal relationships between sleep and

glucose at a finer resolution. Different from prior

studies that primarily consider the daily aggregates of

sleep and glucose data, our study introduces a

circadian perspective. To be more specific, we are

interested in how sleep metrics in the previous night

associates with glucose during the day and how the

glucose level in the daytime associates with the

subsequent night-time sleep. To the best of our

knowledge, this is the first study investigating the

associations between sleep and glucose using

ecologically valid and high-resolution data. The

contribution of this study is as follow:

522

Nhung, H. and Liang, Z.

Contrast Set Mining for Actionable Insights into Associations Between Sleep and Glucose in a Normoglycemic Population.

DOI: 10.5220/0011783600003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 5: HEALTHINF, pages 522-529

ISBN: 978-989-758-631-6; ISSN: 2184-4305

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

 We designed a data mining pipeline based on

the contrast set mining algorithm STUCCO

(Bay & Pazzani, 2001) instead of traditional

statistical methods. Different from correlation

analysis or linear regression, our method allows

the identification of associations that only

manifest when the metrics are within certain

value ranges.

 The data mining pipeline generates interesting

rules and hypotheses that may help inform the

design of future studies to deepen our

understanding of the relationships between

sleep and glucose regulation.

2 RELATED WORKS

Existing studies have found a link between sleep

quality and glucose regulation in both diabetic and

healthy populations. However, the relationship

between sleep and interstitial glucose is complex,

which makes it difficult to quantify the connection.

Multiple approaches have been proposed to examine

this correspondence within and between persons.

Factors such as total sleeping time, sleep

structure, time going to bed, age, and eating habits are

widely known to have a strong influence on blood

glucose levels (Frank et al., 1995; Reutrakul et al.,

2013; Tasali et al., 2008). In a study (Cauter et al.,

1991), eight normal men were supervised for a total

of 53 hours (8h of night sleep, 28h of non-sleep

period, and 8h of daytime sleep). Although glucose

was infused into the body at a constant rate, plasma

glucose levels went through large fluctuations

throughout the study. Interestingly, during sleep

deprivation, glucose levels rise gradually to reach the

maximum at roughly the same time as during regular

sleep, then decrease to daytime levels.

Many studies repeatedly report the association of

sleep dept, sleep curtailment, and reduced of sleep

hygiene contribute to the increase of insulin

sensitivity (SI). Even one night of partial sleep

deprivation may decrease glucose tolerance and

insulin sensitivity significantly (Donga et al., 2010).

(Tasali et al., 2008) paid attention to slow-wave sleep

(SWS) when his team tried to reduce SWS proportion

but sustain total sleep time. His findings suggest that

low levels of SWS, as occurs in the elderly and obese

patients, can decrease insulin sensitivity by 25%,

reaching the reported value for high-risk of diabetes.

In the case of sleep debt, the product of insulin

sensitivity and acute insulin response to intravenous

glucose (AIRg) was decreased by an average of about

40% as compared to the fully rested state, indicating

a high risk of diabetes (Xiang et al., 2006).

Despite evidence had been found to prove the

connection, they were reported as weak and did not

reach statistical significance. In the study focusing on

women only, no impact of progressive sleep

curtailment over 4 nights was reported on measures

of glucose tolerance and SI (Bosy-Westphal et al.,

2008). Similarly, (Zielinski et al., 2008) assessed the

impact of 8 weeks sleep curtailment on glucose

tolerance in self-reported older long sleepers (≥8.5

h/night), compared to a control group; throught

OGTT, the authors observed no effect of sleep

restriction on glucose tolerance. Sleep hygiene is

more challenging to measure as it is more difficult to

define than sleep duration.

Just as sleep affects glucose levels, glucose levels

may as well impact sleep quality. For example, one

study found that people with prediabetes have a

higher rate of suffering from poor sleep than people

with normal glucose (Iyegha et al., 2019). Diabetic

patients display shorter sleep duration and worse

sleep quality, demonstrated by both self-report and

objective measures (Yoda et al., 2015).

Figure 1: sleep hours of some of our subject indicated by vertical lines (blue lines for weekdays and red lines for weekends).

Contrast Set Mining for Actionable Insights into Associations Between Sleep and Glucose in a Normoglycemic Population

523

3 PROPOSED DATA MINING

PIPELINE

3.1 Dataset

We used a public dataset which contains sleep and

glucose readings recorded between June and August

2020 (Bertrand et al., 2021). In total there are 228

days of data collected from 12 healthy subjects who

had no glucose dysregulation. Half of the subjects

were female, and the average age was 32.7 years. The

sleep patterns varied a lot across participants, as

illustrated in Figure 1.

Currently, several technologies are available to

quantitatively assess all-day interstitial glucose

concentration. In this study, the glucose readings

were recorded using the FreeStyle Libre 2 system

which is a coin-size continuous glucose monitoring

(CGM) device attached at the back of the upper arms.

Sleep data were measured with the accelerometer and

photoplethysmography (PPG) embedded in the Fitbit

Charge 3 worn on the non-dominant wrists. Both the

FreeStyle Libre and the Fitbit devices were proven to

be able to generate reasonable valid measurements in

free-living conditions (Li & Bao, 2018; Liang &

Chapa-Martell, 2019).

Table 1: Features constructed from sleep and glucose data.

Feature Denotation

Sleep Total Sleep Time

(hours)

TST

Wake After Sleep

Onset

WASO

Number of wakes ≥ 5

minutes

awake5minCnt

Slee

Efficienc

)

Dee

slee

/TST

(

)

dee

Ratio

Rem sleep/TST (%) remRatio

Glucose Mean (mg/dL) mean

Maximum

(

/dL

)

max

Minimum

(

/dL

)

min

Mean glucose level

outside range (mg/dL)

mge

Mean glucose level

inside range (mg/dL)

mgn

Standard Deviation

(

/dL

)

Coefficient of

variation (%)

Time spent in range

(%)

tir

Low glucose index LBGI

lucose index HBGI

J-index

(

2/dL2

)

index

3.2 Data Preprocessing

The data preprocessing protocol includes segmenting

the CGM data, deriving sleep and CGM features,

categorizing numeric features, synchronizing sleep

and CGM features, and removing missing data.

The dataset is not in the appropriate format to

which the rule induction algorithms may be applied.

Therefore, the next step in the pre-processing

protocol is to categorise the numeric features into

intervals before feeding them into a mining

algorithm. Discretization helps improve the

performance of contrast set mining algorithms and

improve the interpretability of the results. The

features were discretized using the methods shown in

Figure 2. We mainly focused on features that have

well-defined clinical cut-offs. For sleep features, we

used medical cut-offs for young adults (18-25 years)

reported by (Ohayon et al., 2017). The National Sleep

Foundation concluded that sleep quality was a matter

of sleep continuity and sleep architecture. The

optimal sleep architecture for good sleep quality for

adults was agreed to be <5% stage 1 sleep, <81%

stage 2 sleep, 16-20% slow wave sleep (SWS), and

21-30% rapid eye movement (REM) sleep. However,

stage 1 and stage 2 were combined as light sleep in

Fitbit device, therefore the cut-off limits for this

parameter remain unknown. Consequently, we did

not include light sleep ratio in the sleep feature set.

It is important to properly handle the missing data

and ensure the timestamp matches correctly between

Figure 2: Discretization of sleep and glucose level metrics.

HEALTHINF 2023 - 16th International Conference on Health Informatics

524

the sleep features and the glucose features. Rows that

did not meet the required conditions were removed.

Criteria for removing data are nights that lack glucose

data, non-consecutive nights (isolated nights without

any data recorded the day before or after), features

that are dominated by one component (appear >90%

number of nights). After pre-processing, the cleaned

dataset contains 74 days of data from 8 subjects.

3.3 Contrast Set Mining Algorithm

Contrast set mining is a data mining technique that

helps identify contrast patterns between two groups.

The output of contrast set mining are logical rules

with 1 or n antecedents on the left-hand-side and the

consequent on the right-hand-side.

A typical contrast set mining is the Search and

Testing for Understandable Consistent Contrast

(STUCCO) algorithm developed in (Bay & Pazzani,

2001). The STUCCO algorithm follows a basic

pipeline of identifying a basic idea, controlling the

error, filtering the results, evaluating the results, then

drawing conclusions based on the problem at hand.

For this study, we expect to generate some rules as:

TST=low ∩deepRatio=poor → Glucose_tir=low

Figure 3: Contrast sets were distributed into two groups. Group A reveals the relationship between sleep metrics and

interstitial glucose during nocturnal sleep. Group B reveals the relationship between sleep metrics and interstitial glucose the

following day.

Contrast Set Mining for Actionable Insights into Associations Between Sleep and Glucose in a Normoglycemic Population

525

Given a dataset D, let Y = {Y

, Y

, … ,Y

} be a set

of consequences and X = {X

, X

, … , X

} be a set of

antecedents that consist of 1 or more attributes.

Support, confidence, and lift are some of the measures

widely used to assess the quality of the generated

rules. Support for X1 → Y1 is the percentage of

example in D containing X1∪Y1. Confidence is the

ratio of example X1 in total X that contributes to Y1.

Lift is the threshold defined by the user to find the

contrast set. Long rules with too many features in the

antecedents are often hard to interpret and could be

redundant. Therefore, we limit the length of the rules

to 2 features. We consider a contrast set as validate if

it meets the requirements: support ≥ 0.02, confident ≥

0.75, and lift ≥ 2. After contrast sets generation, rules

which share the same attribute were grouped together

(Figure 3 & 4) for better interpretation.

4 RESULTS

4.1 Case 1: Antecedence = Sleep of Day

N, Consequence = Glucose of Day N

The contrast set mining results show connection

between the sleep quality at night and the glucose

patterns the next day, as well as the sleep quality at

night and the concurrent glucose characteristics

during sleep. The identified rules are presented in

Figure 3. TST is a major factor that associates to

affect the TIR of glucose both in sleep and during

active periods. It is shown that the time the glucose

levels stayed within the normal range was shorter

when the participants slept less than 7 hours at night.

In contrast, people who had long sleep periods (> 9

hours) spent a longer time in normal glucose level

Figure 4: The contrast sets show how interstitial glucose during the day connected with the following night sleep (group C)

and interstitial glucose level during sleep (group D).

HEALTHINF 2023 - 16th International Conference on Health Informatics

526

range throughout the night. Under the confounding

effect of other sleep metrics, TST was also associated

with the TIR of glucose level during active time the

next day. For example, short TST combined with a

good ratio of REM sleep were associated with higher

TIR the next day.

In addition to sleep duration, sleep continuity and

sleep structure as characterised by the ratio of sleep

stages, were also found to be associated with glucose

regulation during sleep and during active time the

next day. Long awake5minDuration strengthens the

association between short TST and low TIR during

sleep, with the lift increased from 2.31 to 4.30.

However, the same combination was associated with

a low level of HBGI the next day. Long WASO,

together with the confounding effect of sleep

structure, were associated with higher LBGI and

HBGI the next day, indicating impaired glucose

regulation.

The associations between sleep structure (i.e., the

ratio of deep sleep and REM sleep) and glucose

regulation were complex and were often confounded

by sleep duration and sleep continuity.

4.2 Case 2: Antecedent = Glucose

of Day N, Consequent = Sleep

of Day N+1

There is a vicious cycle between sleep and glucose

levels when one affects the other and the impact

repeats days and nights. While finding the rules

between glucose levels at day and sleep metrics at

nights, we found that a long time in range for glucose

level helps to keep sleep WASO within good range

(less than 20 minutes of awakening per night). People

with long tir glucose are highly likely to sleep less

than 7 hours the following night. In addition, long

time in range link with good WASO while short time

in range link with good deep sleep ratio.

It is noticeable that the relationship between

active glucose and in-sleep glucose are more complex

and extremely difficult to interpret. Interestingly, all

the right-hand-side components belong to the “low”

section of the features and the rules are mainly

associated with glucose mge and mgn. From figure 4,

there is a contrast set with 2 completely opposite

components: if active glucose time in range is high,

then in sleep glucose time in range is low. Several

studies have suggested multiple pathways in the

possible connection between sleep quality and the

fluctuation of glycaemic variability. But whether

daytime glucose impacts in-sleep glucose remains a

research gap that needs further examination.

5 DISCUSSIONS

In this study, we analysed the potential connections

between sleep and glucose in the series of consecutive

days. Our finding revealed that there are bi-direction

relationships between TST and glucose TIR. Total

sleep time is the feature that dominated most of the

contrast sets. Short sleep time was associated with the

fluctuation of in-sleep glucose, whereas long sleep

time was associated with more stable glucose

regulation. On the flip side, an association was found

between glucose fluctuation and sleep continuity,

which is consistent with previous findings (Griggs et

al., 2022). Our result indicated that long TIR and low

HBGI correlate to shorter WASO. This echoes

findings in prior studies that long WASO occurred on

days with high J index, high HBGI, and less time in

hypoglycemia (Griggs et al., 2020).

This study has several limitations that should be

addressed in future studies. First, the size of the

dataset is limited. Future studies should collect data

from a larger cohort and potentially cover a wide

spectrum of demographic characteristics.

Furthermore, not only nocturnal sleep, but also

napping may be linked to glucose regulation, as

suggested in previous lab-based studies (Kothari et

al., 2021). Since napping is a common habit of the

young population and people in tropical areas, this

line of research is likely to generate new insights.

Another promising aspect for future examination is

the impact within and between personal variations in

sleep patterns. With a large dataset, the impact of

interpersonal differences should be studied when

each person has different lifestyles and circadian

rhythms. Finally, with the abundance of rules found

by STUCCO, it is necessary for a post-mining method

to select quality rules.

6 CONCLUSIONS

We have designed a data mining pipeline featuring

the contrast set mining algorithm STUCCO to

identify reciprocal associations between sleep and

glucose levels. The finding highlights the total sleep

time as an important sleep metric associated with

glucose regulation the next day, which is

characterised by rules with high lift and confidence.

Reversely, associations were also found between the

glucose fluctuation in wake time and the continuity of

the subsequent sleep at night. Compared to sleep-

>glucose relation, glucose has a weaker association

with nocturnal sleep as the lift of the identified rules

Contrast Set Mining for Actionable Insights into Associations Between Sleep and Glucose in a Normoglycemic Population

527

was lower. These findings added to the existing

knowledge looking at the glucose profile in the

normoglycemic population and helped generate

actionable insights for the holistic management of

sleep and metabolic health. With a better method to

remove abundant rules and interpret information, this

contrast set mining algorithm can be applied to a

recommendation system for adjusting human

behaviour.

ACKNOWLEDGEMENTS

This study was supported by JSPS KAKENHI Grant

Number 21K17670.

REFERENCES

Bay, S. D., & Pazzani, M. J. (2001). Detecting Group

Differences: Mining Contrast Sets. Data Mining and

Knowledge Discovery, 5, 213–246.https://doi.org/

10.1023/A:1011429418057

Bertrand, L., Cleyet-Marrel, N., & Liang, Z. (2021).

Recognizing Eating Activities in Free-Living

Environment Using Consumer Wearable Devices.

Eng. Proc. , 6, 58.

Bosy-Westphal, A., Hinrichs, S., Jauch-Chara, K., Hitze,

B., Later, W., Wilms, B., . . . Muller, M. J. (2008).

Influence of partial sleep deprivation on energy balance

and insulin sensitivity in healthy women. Obesity facts,

1(5),266–273. https://doi.org/10.1159/000158874

Cauter, E. V., Blackman, J. D., Roland, D., Spire, J. P.,

Refetoff, S., & Polonsky, K. S. (1991). Modulation of

glucose regulation and insulin secretion by circadian

rhythmicity and sleep. J Clin Invest., 88(3), 934-942.

https://doi.org/10.1172/JCI115396

Donga, E., Dijk, M. v., Dijk, J. G. v., Biermasz, N. R.,

Lammers, G.-J., Kralingen, K. W. v., . . . Romijn, J. A.

(2010). A single night of partial sleep deprivation

induces insulin resistance in multiple metabolic

pathways in healthy subjects. The Journal of clinical

endocrinology and metabolism, 95(6), 2963–2968.

Frank, S. A., Roland, D. C., Sturis, J., Byrne, M. M.,

Refetoff, S., Polonsky, K. S., & Van Cauter, E. (1995).

Effects of aging on glucose regulation during

wakefulness and sleep. The American journal of

physiology, 269(6 Pt1)(E1006–E1016). https://doi.org/

10.1152/ajpendo.1995.269.6.E1006

Gottlieb, D. J., Punjabi, N. M., Newman, A. B., Resnick, H.

E., Redline, S., Baldwin, C. M., & Nieto, F. J. (2005).

Association of sleep time with diabetes mellitus and

impaired glucose tolerance. Archives of internal

medicine,165(8),863–867.

https://doi.org/10.1001/archinte.165.8.863

Griggs, S., Barbato, E., Hernandez, E., Gupta, D.,

Margevicius, S., Grey, M., & Jr., R. L. H. (2022).

Glucose and unstructured physical activity coupling

during sleep and wake in young adults with type 1

diabetes. Scientific reports,12(1)(5790). https://doi.org/

10.1038/s41598-022-09728-2

Griggs, S., Redeker, N. S., Jeon, S., & Grey, M. (2020).

Daily variations in sleep and glucose in adolescents

with type 1 diabetes. Pediatric diabetes, 21(8), 1493–

1501. https://doi.org/10.1111/pedi.13117

Iyegha, I. D., Chieh, A. Y., Bryant, B. M., & Li, L. (2019).

Associations between poor sleep and glucose

intolerance in prediabetes. Psychoneuroendocrinology,

110(104444). https://doi.org/10.1016/j.psyneuen.2019.

104444

Jiawei, Y., Xiaoling, J., Zhilei, S., Shuzhen, L., Hao, H.,

Peiyun, L., . . . Liegang, L. (2017). Relationship of Sleep

Duration With All-Cause Mortality and Cardiovascular

Events: A Systematic Review and Dose-Response Meta-

Analysis of Prospective Cohort Studies. Journal

of the American Heart Association,6(9)(e005947).

https://doi.org/10.1161/JAHA.117.005947

Kothari, V., Cardona, Z., Chirakalwasan, N.,

Anothaisintawee, T., & Reutrakul, S. (2021). Sleep

interventions and glucose metabolism: systematic review

and meta-analysis. Sleep medicine, 78, 24-35.

https://doi.org/10.1016/j.sleep.2020.11.035

Li, M., & Bao, Y. (2018). Methods for Interpreting

Continuous Glucose Monitoring Graphs. In W. Jia (Ed.),

Continuous Glucose Monitoring. https://doi.org/

10.1007/978-981-10-7074-7_5

Liang, Z. (2022). Mining associations between glycemic

variability in awake-time and in-sleep among non-

diabetic adults. Front Med Technol. https://doi.org/

10.3389/fmedt.2022.1026830

Liang, Z., & Chapa-Martell, M. A. (2018). Validity of

consumer activity wristbands and wearable EEG for

measuring overall sleep parameters and sleep structure in

free-living conditions. Journal of Healthcare Informatics

Research, 1-27.

Liang, Z., & Chapa-Martell, M. A. (2019). Accuracy of Fitbit

wristbands in measuring sleep stage transitions and the

effect of user-specific factors. JMIR Mhealth Uhealth,

7(6), e13384. https://doi.org/10.2196/13384

Lou, P., Zhang, P., Zhang, L., Chen, P., Chang, G., Zhang,

N., . . . Qiao, C. (2015). Effects of sleep duration and

sleep quality on prevalence of type 2 diabetes mellitus: A

5-year follow-up study in China. Diabetes research and

clinical practice, 109 (1), 178–184.

https://doi.org/10.1016/j.diabres.2015.04.012

Ohayon, M., Wickwire, E. M., Hirshkowitz, M., & al., e.

(2017). National Sleep Foundation's sleep quality

recommendations: first report. Sleep Health, 3(1), 6-19.

Reutrakul, S., Hood, M. M., Crowley, S. J., Morgan, M. K.,

Teodori, M., Knutson, K. L., & Cauter, E. V. (2013).

Chronotype is independently associated with glycemic

control in type 2 diabetes. Diabetes care, 36(9),2523–

2529. https://doi.org/10.2337/dc12-2697

Tasali, E., Leproult, R., Ehrmann, D. A., & Cauter, E. V.

(2008). Slow-wave sleep and the risk of type 2 diabetes

in humans. Proceedings of the National Academy of

Sciences of the United States of America,

HEALTHINF 2023 - 16th International Conference on Health Informatics

528

Wang, H., Leng, J., Li, W., Wang, L., Zhang, C., Liu, H., . . .

Yang, X. (2017). Sleep duration and quality, and risk of

gestational diabetes mellitus in pregnant Chinese women.

Diabetic medicine : a journal of the British Diabetic

Association, 34(1), 44–50. https://doi.org/10.1111/

dme.13155

Watson, N. F., Badr, M. S., Belenky, G., Bliwise, D. L.,

Buxton, O. M., Buysse, D., . . . Tasali, E. (2015).

Recommended Amount of Sleep for a Healthy Adult: A

Joint Consensus Statement of the American Academy of

Sleep Medicine and Sleep Research Society. Sleep,

38(6):843-4.

Xiang, A. H., Peters, R. K., Kjos, S. L., Marroquin, A., Goico,

J., Ochoa, C., . . . Buchanan, T. A. (2006). Effect of

pioglitazone on pancreatic beta-cell function and diabetes

risk in Hispanic women with prior gestational diabetes.

Diabetes, 55(2), 517–522. https://doi.org/10.2337/

diabetes.55.02.06.db05-1066

Xiao, Q., Keadle, S. K., Hollenbeck, A. R., & Matthews, C.

E. (2014). Sleep duration and total and cause-specific

mortality in a large US cohort: interrelationships with

physical activity, sedentary behavior, and body mass

index. American journal of epidemiology, 180(10),997–

1006.

Yoda, K., Inaba, M., Hamamoto, K., Yoda, M., Tsuda, A.,

Mori, K., . . . Yamada, S. (2015). Association between

Poor Glycemic Control, Impaired Sleep Quality, and

Increased Arterial Thickening in Type 2 Diabetic

Patients. PLoS One, 10(4)(e0122521). https://doi.org/

10.1371/journal.pone.0122521

Zielinski, M. R., Kline, C. E., Kripke, D. F., Bogan, R. K., &

Youngstedt, S. D. (2008). No effect of 8-week time in

bed restriction on glucose tolerance in older long

sleepers. Journal of sleep research, 17(4), 412–419.

https://doi.org/10.1111/j.1365-2869.2008.00673.x

Zuraikat, F. M., Makarem, N., Redline, S., Aggarwal, B.,

Jelic, S., & St-Onge, M. P. (2020). Sleep Regularity and

Cardiometabolic Heath: Is Variability in Sleep Patterns a

Risk Factor for Excess Adiposity and Glycemic

Dysregulation?. Current diabetes reports ,, 20 (8), 38.

https://doi.org/10.1007/s11892-020-01324-w

Contrast Set Mining for Actionable Insights into Associations Between Sleep and Glucose in a Normoglycemic Population

529