Diabetes Care 1

Babak Mokhlesi,1 Karla A. Temple,1 Association of Self-Reported Ashley H. Tjaden,2 Sharon L. Edelstein,2 Kristina M. Utzschneider,3 and Circadian Measures Kristen J. Nadeau,4 Tamara S. Hannon,5 Susan Sam,1 Elena Barengolts,6 With Glycemia in Adults With Shalini Manchanda,5 David A. Ehrmann,1 Prediabetes or Recently Diagnosed and Eve Van Cauter,1 the RISE Consortium* Untreated Type 2 Diabetes https://doi.org/10.2337/dc19-0298

OBJECTIVE Sleep disturbances and circadian misalignment (social , late chronotype, or ) have been associated with worse glycemic control in type 2 diabetes (T2D). Whether these findings apply to adults with prediabetes is yet unexplored. Wehypothesized that self-reported short sleep, poor sleep quality, and/or circadian misalignment are associated with higher glycemia, BMI, and blood pressure (BP) in adults with prediabetes or recently diagnosed, untreated T2D.

RESEARCH DESIGN AND METHODS Our cohort included 962 overweight/obese adults ages 20–65 years with pre- diabetes or recently diagnosed, untreated T2D who completed a 2-h oral glucose tolerance test and validated sleep questionnaires. Independent associations of sleep and circadian variables with glycemia, BMI, and BP were evaluated with regression models. RISK METABOLIC AND CARDIOVASCULAR 1University of Chicago, Chicago, IL 2 RESULTS George Washington University Biostatistics Center (RISE Coordinating Center), Rockville, MD The multiethnic cohort was 55% men, with mean 6 SD age 52.2 6 9.5 years and BMI 3VA Puget Sound Health Care System and Uni- 34.7 6 5.5 kg/m2. Mean sleep duration was 6.6 6 1.3 h. Poor sleep quality was versity of Washington, Seattle, WA 4 reported by 54% and high risk for obstructive by 64%. HbA was University of Colorado Anschutz Medical Cam- 1c pus/Children’s Hospital Colorado, Denver, CO significantly higher in those reporting <5or>8 h sleep per night. Sleep duration >8h 5Indiana University School of Medicine, India- was also associated with higher fasting glucose and <6 h with higher BMI. Shift work napolis, IN was also associated with higher BMI. Social jet lag and delayed chronotype were 6Jesse Brown VA Medical Center, Chicago, IL associated with higher BP. Corresponding author: Babak Mokhlesi, bmokhles @medicine.bsd.uchicago.edu CONCLUSIONS Received 13 February 2019 and accepted 15 April In our cohort, self-reported short and long sleep were both associated with adverse 2019 measures of glycemia, and short sleep and shift work were associated with higher This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/suppl/ BMI. Further research using objective measures of sleep is needed to better doi:10.2337/dc19-0298/-/DC1. delineate the relationship between sleep and glycemia in adults with prediabetes *A complete list of the RISE Consortium Inves- or T2D. tigators can be found in Supplementary Data. © 2019 by the American Diabetes Association. The obesity epidemic has led to an increase in type 2 diabetes (T2D) (1). There are ;30 Readers may use this article as long as the work is properly cited, the use is educational and not million individuals with T2D and nearly 90 million with prediabetes in the U.S. (2). In for profit, and the work is not altered. More infor- parallel, there has been an increase in the prevalence of sleep disturbances (3). mation is available at http://www.diabetesjournals Chronic partial sleep loss due to restriction is increasingly prevalent in our .org/content/license. Diabetes Care Publish Ahead of Print, published online May 2, 2019 2 Sleep and Circadian Measures in Prediabetes Diabetes Care

modern society. Moreover, 24-h access investigators’ clinics, internal and exter- 5 min apart. Only the second measure- to light allows people to engage in be- nal advertising, social and public media, ment was taken as the reported value. At haviors that are inappropriately timed and mailings to individuals identified via screening, blood samples for HbA1c, fast- relative to the endogenous circadian electronic medical records. Prescreening ing plasma glucose, and OGTT 2-h plasma rhythm, leading to circadian misalign- of electronic and clinical medical records glucosewere collected.All bloodsamples ment. Nowadays, nearly 20% of working was used, when possible, to identify were immediately placed on ice, sepa- adults are shift workers, an extreme form patients at risk based on BMI and rated by centrifugation, and frozen at of circadian misalignment (4). HbA1c. Additional details on participant 280°C prior to shipment to the central Over the last decade, there has been recruitment and eligibility criteria have biochemistry laboratory at the University mounting evidence from a large number previously been described (9), and de- of Washington (Northwest Lipid Metab- of prospective epidemiologic studies that tailed information is available on the RISE olism and Diabetes Research Laborato- short and long sleep duration, poor sleep website (https://rise.bsc.gwu.edu/web/ ries, University of Washington, Seattle, quality, (OSA), rise/collaborators). The study was ap- WA). Plasma glucose concentrations and circadian misalignment are associ- proved by the institutional review boards were measured by the glucose hexoki- ated with T2D (5,6). Whether sleep or of all participating centers, and all par- nase method using Roche reagent on a circadian disturbances impact glycemic ticipants provided written informed Roche c501 autoanalyzer. The method control in individuals with prediabe- consent prior to initiation of any study- interassay coefficients of variation on tes and T2D, independently of BMI and related activities. quality control samples with low, me- other confounders, has been examined in dium, and high glucose were 2.0%, 1.7%, fewer studies (7,8). This question is im- Data Collection and 1.3%, respectively. HbA1c was mea- portant, since these lifestyle factors may Individuals who met preliminary inclu- sured by ion-exchange high-performance compromise the efficacy of treatment sion/exclusion criteria underwent a 75-g chromatography on a Tosoh G8 analyzer and accelerate the progression of the oral glucose tolerance test (OGTT) and (Tosoh Bioscience, South San Francisco, disease and the development of compli- HbA1c measurement (n = 1,069). Fasting CA). The interassay coefficients of var- cations. Two cross-sectional studies from blood samples for HbA1c and plasma iation on low and high quality con- Thailand have reported an association of glucose were obtained; blood was also trol samples were 1.9% and 1.0%, circadian misalignment with higher glu- collected at 2 h of the OGTT for mea- respectively. cose concentrations and BMI in patients surement of plasma glucose. For the with prediabetes (7,8). To date, however, current analyses, we included partici- Sleep and Circadian Assessments these associations have not been ex- pants who met the American Diabetes To assess sleep quality independent of plored in individuals from other ethnic/ Association definition of prediabetes sleep duration, we used a modified Pitts- racial backgrounds. To that end, we (fasting plasma glucose 100–125 mg/dL, burgh Sleep Quality Index (PSQI) ques- aimed to quantify the associations be- 2-h plasma glucose 140–199 mg/dL, or tionnaire to determine self-reported tween self-reported sleep duration, sleep HbA1c 5.7–6.4%) as well as recently di- usual bedtime, wake time, and sleep quality, and circadian misalignment (e.g., agnosed (,1 year), untreated partici- duration on workdays and days off social jet lag, late chronotype, or shift pants with existing T2D (fasting plasma work during the prior month (11,12). work) and dysglycemia, higher BMI, and glucose $126 mg/dL, 2-h plasma glu- Our modified PSQI starts by asking increased blood pressure (BP) in a cohort cose $200 mg/dL, or HbA1c $6.5%) whether the responder is employed or of adults with prediabetes or recently (10). not employed. Those who are employed diagnosed, untreated T2D. Sex and race/ethnicity were self- are asked for how many days per week reported. Anthropometric measurements and whether their schedule involves a RESEARCH DESIGN AND METHODS were performed with participants wear- form of shift work. Sleep duration was Participants ing light clothing without shoes. Height derived from the following question: This is a cross-sectional analysis of data was measured in a fully vertical position “During the past month, how many hours obtained during the screening phase of with heels together using a calibrated of actual sleep did you get at night? (This the Restoring Insulin Secretion (RISE) stadiometer. Weight was measured us- may be different than the number of consortium randomized controlled trials. ing a calibrated electronic scale, zeroed hours you spent in )”; when appro- Between 2013 and 2017 we screened before each measurement. Height and priate, this question was asked sepa- 1,355 overweight/obese men and women weight measurements were performed rately for workdays and days off work, ages 20–65 years for RISE. Participants twice, with the average value reported. and average self-reported sleep duration were recruited from the active patient BP was measured with a calibrated was calculated as the weighted average populations and communities at four automated BP machine with appropri- of reported sleep duration on workdays RISE adult centers 1): University of Chi- ately sized arm cuffs; readings were and days off work. Self-reported sleep cago and the Jesse Brown VA Medical obtained with the participant in a seated duration was analyzed categorically (,5h, Center, 2) Indiana University, 3)Univer- position with feet touching the floor or 5to,6h,6to,7h,7to,8h,and$8h). sity of Washington and the VA Puget otherwise supported after at least 5 min Ascoreof$5 on the PSQI (minimum Sound Health Care System, and 4) Uni- of rest in a quiet room, with outer score of 0 and maximum score of 21) versity of Southern California. Recruit- clothing removed and sleeves rolled to indicated poor sleep quality (13). Sleep ment techniques included referral from the shoulder. The cuff was placed at heart debt was calculated as the difference colleagues and screening from the level, and two measurements were taken between preferred hours of sleep per care.diabetesjournals.org Mokhlesi and Associates 3

night and the weighted average of sleep Unadjusted group comparisons were per- measurements of complete fasting and duration.Chronotype quantifies individ- formed using ANOVA for normally distrib- 2-h OGTT plasma glucose concentrations 2 ual preference for bedtime. The metric uted continuous variables. Pearson x test as well as HbA1c values. Validated sleep of chronotype or midsleep time on free was used to compare categorical vari- questionnairedata were available in 1,042 days with adjustment for ables. Multiple linear regression models participants. After exclusion of those with (MSFsc) was derived from midsleep were used to explore the independent normal glucose and HbA1c values, there time on days off work (or weekend nights association of glycemic variables and were 962 participants with complete in unemployed individuals) with further BP with sleep variables (sleep duration, OGTT and sleep data (704 with prediabe- adjustment for the sleep debt taking into sleep quality, sleepiness, and OSA risk) tes and 258 with recently diagnosed, account the sleep duration average of and circadian measures (chronotype, untreated T2D). Table 1 summarizes days off work or weekends and weekdays social jet lag, and shift work) after baseline characteristics of the partici- as follows: MSFsc = midsleep time on adjustment for age, sex, race/ethnicity, pants. The cohort had nearly equal days off work or weekend nights – 0.5 * and BMI, as well as the association of numbers of men and women, from di- [SDF 2 (no. of days off work * SDW + no. BMI with sleep and circadian measures verse racial/ethnic backgrounds. Shift of days off work * SDF)/7], where SDF is after adjustment for age, sex, and race/ work was reported by 24.2% of the the calculated sleep duration on days off ethnicity. All statistical calculations cohort. Table 2 summarizes sleep ques- work or weekend nights and SDW is the were performed without correction for tionnaire measures. There were no sig- calculated sleep duration on workdays or multiple testing. Analyses were per- nificant differences in self-reported sleep weekday nights (14). In individuals who formed using SAS 9.4 (SAS Institute, or circadian measures between partici- were unemployed, we used 2 days for Cary, NC). pants with prediabetes and T2D. The weekends.Inindividualswhoreportedto mean 6 SD habitual sleep duration of be working, we specifically asked for the RESULTS the entire cohort was 6.6 6 1.3 h number of days each individual was off of Of the 1,355 adults who participated in with a sleep debt of 1.5 6 1.3 h. Average work per week. Therefore, the weighing screening OGTTs for RISE, 1,069 had sleep duration of #6 h per night was of “days off work” in the calculation of chronotype was individualized. For mul- — tivariate analyses, chronotype was trans- Table 1 Descriptive characteristics of participants at RISE screening formed into values relative to midnight All Prediabetes Diabetes (212 to 12) with 0 representing mid- N 962 704 258 P† night. Social jet lag, the misalignment of Demographic biological and social time, was calculated Age, years 52.2 6 9.5 51.8 6 9.5 53.2 6 9.3 0.046 based on the absolute difference be- Age category (years) 0.574 tween midsleep time on workdays and 20–39 111 (11.5) 83 (11.8) 28 (10.9) – days off work (15). Absolute social jet lag 40 49 224 (23.3) 169 (24.0) 55 (21.3) 50–66 627 (65.2) 452 (64.2) 175 (67.8) was analyzed as a categorical variable Sex 0.231 (,1h,1–2h,.2 h). The Berlin Ques- Men 525 (54.6) 376 (53.4) 149 (57.8) tionnaire was used to classify the risk of Women 437 (45.4) 328 (46.6) 109 (42.2) OSA into high or low risk (16). The Ep- Race/ethnicity 0.611 worth Sleepiness Scale (ESS) (scores White 400 (41.6) 288 (40.9) 112 (43.4) 0–24) was used to assess subjective Black 348 (36.2) 262 (37.2) 86 (33.3) Hispanic 155 (16.1) 109 (15.5) 46 (17.8) daytime sleepiness. Scores of .10 are Asian 32 (3.3) 23 (3.3) 9 (3.5) considered indicative of daytime sleep- American Indian 27 (2.8) 22 (3.1) 5 (1.9) iness (17). Shift work was assessed based Employment on three questions: starting work before Employed 579 (71.3) 431 (70.3) 148 (74.4) 0.271 6:00 A.M., working overnight shifts, and Start work before 6:00 A.M. 105 (13.7) 80 (13.7) 25 (13.7) 0.977 working rotating night and day shifts. Overnight shifts 82 (10.7) 59 (10.2) 23 (12.5) 0.374 Shift work was analyzed as a dichoto- Rotating night and day shifts 81 (10.6) 53 (9.1) 28 (15.4) 0.016 mous variable (yes/no). Shift worker* 188 (24.2) 136 (23.1) 52 (28.0) 0.173 Screening anthropometrics BMI (kg/m2) 34.7 6 5.5 34.5 6 5.6 35.3 6 5.3 0.053 Statistical Analysis Weight (kg) 100.8 6 20.1 99.9 6 20.0 103.3 6 20.2 0.017 Data were stored and managed centrally, Systolic BP (mmHg) 127.9 6 14.3 127.4 6 14.2 129.3 6 14.3 0.062 and analyses were performed according Diastolic BP (mmHg) 77.2 6 9.8 76.9 6 10.0 77.9 6 9.1 0.151 fi to a prespeci ed analytic plan. All anal- Screening glucose measurements yses were cross-sectional. Outcomes of HbA1c (%) 5.8 6 0.4 5.7 6 0.3 6.1 6 0.5 ,0.001 interest were glycemic variables (HbA1c Fasting plasma glucose (mg/dL) 111.5 6 13.1 106.6 6 8.4 124.8 6 14.4 ,0.001 and fasting and OGTT 2-h plasma glu- 2-h plasma glucose (mg/dL) 160.5 6 48.4 139.8 6 30.6 217.6 6 42.4 ,0.001 cose), BMI, and BP. Continuous variables Data are n (%) or mean 6 SD unless otherwise indicated. †P value comparing prediabetes with were expressed as mean 6 SD for nor- diabetes. *The percentage of shift workers is not the sum of all three categories of shift work, as mally distributed data; categorical vari- some participants had multiple types of shift work (i.e., starting work before 6:00 A.M., working overnight shifts, and working rotating night and day shifts). ables were summarized as percentages. 4 Sleep and Circadian Measures in Prediabetes Diabetes Care

Table 2—Self-reported sleep and circadian measures Sleep duration, sleep quality, and high All Prediabetes Diabetes P† risk of OSA were not associated with BP. Daytime sleepiness, however, was inde- Sleep measures pendently associated with systolic BP (P = Habitual sleep duration (h) 6.6 6 1.3 6.6 6 1.3 6.7 6 1.4 0.673 Sleep duration #6 h 303 (32.1) 221 (31.9) 82 (32.5) 0.850 0.037). Compared with participants with- Sleep efficiency, % 85.8 6 14.2 85.6 6 14.2 86.4 6 14.3 0.485 out daytime sleepiness, those who re- PSQI global score 6.7 6 3.9 6.8 6 3.9 6.4 6 3.9 0.140 ported sleepiness had 2.0 mmHg lower Poor sleep quality 511 (54.1) 386 (55.8) 125 (49.6) 0.092 systolic BP (95% CI 23.867, 20.128). High apnea risk 618 (64.6) 447 (63.9) 171 (66.3) 0.504 Sleep debt (h) 1.5 6 1.3 1.5 6 1.3 1.4 6 1.3 0.304 Association Between Circadian Daytime sleepiness 327 (34.0) 246 (34.9) 81 (31.4) 0.303 Measures and Outcomes 6 6 6 ESS 8.1 4.9 8.3 5.0 7.6 4.7 0.045 Social jet lag and shift work were not Circadian measures associated with HbA , fasting plasma Bedtime workdays (h:min) 22:31 6 2:01 22:31 6 01:59 22:31 6 02:06 0.998 1c Bedtime days off work (h:min) 23:15 6 1:29 23:16 6 01:30 23:15 6 01:24 0.868 glucose, or OGTT 2-h plasma glucose Wake time workdays (h:min) 06:31 6 2:15 06:30 6 02:07 06:36 6 02:38 0.604 after adjustment for age, sex, race/ Wake time days off work (h:min) 07:43 6 1:42 07:43 6 01:44 07:43 6 01:38 0.988 ethnicity, BMI, and chronotype. Chronotype (h:min)* 03:31 6 02:33 03:35 6 02:43 03:20 6 01:57 0.280 Chronotype and social jet lag were not Social jet lag (h) 1.2 6 1.5 1.2 6 1.5 1.3 6 1.7 0.406 associated with BMI. In contrast, shift Shift work 188 (24.2) 136 (23.1) 52 (28.0) 0.173 work was associated with a higher BMI. Data are n (%) or mean 6 SD. †P value comparing prediabetes with diabetes. *The MSFsc was In a fully adjusted model, shift work was derived from midsleep time on days off work (or weekend nights in unemployed individuals) associated with 1.32 kg/m2 higher BMI with further adjustment for the sleep debt taking into account the sleep duration average of days off work or weekends and weekdays as follows: MSFsc = midsleep time on days off work or (95% CI 0.42, 2.23; P = 0.0043). weekend nights – 0.5 * [SDF 2 (no. of days off work * SDW + no. of days off work * SDF)/7], Both later chronotype and social jet where SDF is the calculated sleep duration on days off work or weekend nights and lag were independently associated with SDW is the calculated sleep duration on workdays or weekday nights. higher BP. Chronotype was significantly associated with both systolic (P = 0.0004) reported by 32% of the participants. duration as a continuous variable. For and diastolic BP (P = 0.0120). For every Close to one-third (34.0%) of the cohort each hour of additional sleep, the adjusted hour of later chronotype, systolic BP was reported excessive daytime sleepiness fasting glucose was 0.79 mg/dL higher 1.28 mmHg higher (95% CI 0.58, 1.98) and based on an ESS score of .10. Poor (95% CI 0.15, 1.42; P = 0.015). We did not diastolic BP was 0.66 mmHg higher (95% sleep quality (54.1%) and high risk for detect significant associations between CI 0.15, 1.17). Compared with a social jet OSA (64.9%) were highly prevalent. Bed- self-reported sleep duration and OGTT lag of ,2 h, social jet lag of .2hwas times and wake times occurred later on 2-h plasma glucose levels. Further, our associated with a significantly higher sys- days off work (or on weekends in the measures of sleep quality, daytime sleep- tolic BP (adjusted mean 127.4 mmHg [95% unemployed), leading to an absolute iness, and OSA risk were not significantly CI 125.3, 129.6] and 131.0 mmHg [95% CI social jet lag of 1.2 6 1.5 h. Nearly associated with HbA1c, fasting glucose, or 127.9, 134.2], respectively, P = 0.014). Shift half of the participants (48.5%) had ab- 2-h glucose. work was not associated with BP. solute social jet lag of 1–2 h, while 32.2% Consistent with existing epidemiologic In a sensitivity analysis that excluded had ,1 h and 19.3% had .2 h. Chro- evidence, BMI was inversely associated shift workers, chronotype and social jet notype varied substantially, with 13.3% with sleep duration after adjustment for lag were not associated with measures having a chronotype before 2:00 A.M. and age, sex, and race/ethnicity (P = 0.028). of glycemia and BMI (data not shown). 22.9% having a chronotype after 4:00 A.M. For each hour of additional sleep, the The remaining 63.8% had chronotypes be- adjusted BMI was 0.3 kg/m2 lower (95% CONCLUSIONS tween 2:00 and 4:00 A.M. CI 20.56, 20.03). Poorer sleep quality, In the present cross-sectional analysis measured by the global PSQI score, and of a large, ethnically diverse cohort of Association Between Sleep Measures excessive daytime sleepiness, as as- overweight/obese adults with prediabe- and Outcomes sessed by the ESS, were also significantly tes or recently diagnosed, untreated T2D, Associations of measures of glycemic con- associated with higher BMI (P = 0.048 we demonstrated that both short and trol, BMI, and BP with sleep and circadian and P = 0.0024, respectively). For each long self-reported sleep durations were measures are shown in Fig. 1. After ad- 1-point increase on the ESS score, the associated with higher measures of gly- justment for age, sex, race/ethnicity, and adjusted BMI increased by 0.1 kg/m2 cemia after we controlled for BMI and BMI, there was a U-shaped relationship (95% CI 0.04, 0.18), and similarly, for other demographic characteristics. Short between categories of sleep duration and each 1-point increase on the global PSQI sleep duration and shift work were also HbA1c, with those reporting ,5 h (mean score, the adjusted BMI increased by associated with higher BMI. Chronotype 5.84%[95% CI5.74, 5.93]) and .8h(mean 0.09 kg/m2 (95% CI 0.001, 0.18). Com- and social jet lag, on the other hand, were 5.85% [95% CI 5.78, 5.93]) of sleep having pared with participants who reported not associated with measures of glyce- significantly higher HbA1c values com- good sleep quality, those who reported mia or BMI but instead were indepen- pared with those with 7–8 h of sleep poor sleep quality had a trend toward a dently associated with BP. (mean 5.74% [95% CI 5.67, 5.80]). Fasting higher BMI (0.6 kg/m2 [95% CI 20.09, Multiple laboratory-based studies in- glucose was directly associated with sleep 1.28]; P = 0.09). volving experimental sleep manipulations, care.diabetesjournals.org Mokhlesi and Associates 5

Figure 1—Association between self-reported sleep measures and outcomes. Adjusted means from multiple linear regression models. Data are adjusted means and 95% CIs. “Quality” is sleep quality, “sleepiness” is daytime sleepiness, and “duration” is sleep duration. Models adjusted for age, sex, and race/ethnicity. HbA1c, fasting glucose, 2-h glucose, and BP models are also adjusted for BMI. The association between apnea risk and BMI was not quantified because we used the Berlin Questionnaire to asses for the risk of sleep apnea. In this questionnaire, a BMI .30 kg/m2 is one of the three categories to assign high risk of apnea. *ANOVA P , 0.05. including sleep restriction and sleep frag- At a population level, several cross- glycemic control in prediabetes. More- mentation, have been performed to eval- sectional studies from various geo- over, most studies have focused on either uate the role of sleep on the control of graphic regions have reported an fasting glucose or HbA1c, without per- energy balance and glucose metabolism. association between self-reported short forming an OGTT. These studies have revealed that short- sleep duration and impaired fasting glu- Several studies have experimentally term sleep restriction can decrease leptin cose (22224). Other studies have found induced extreme circadian misalignment levels, increase ghrelin levels, and increase associations between self-reported short to better elucidate its role in glucose endocannabinoid levels, leading to in- sleep duration and prevalent prediabetes metabolism dysregulation. These studies creases in hunger, appetite, and hedonic or progression to T2D (25229). Although demonstrated that circadian misalign- food intake and, simultaneously, leading most epidemiological and laboratory- ment mimicking shift work led to reduced to a decrease in glucose tolerance (18). based studies have focused on the as- glucose tolerance in healthy humans Although the pathophysiologic and causal sociation between short sleep duration (34236). In these studies, both fasting links between sleep disturbances and and dysglycemia or obesity, a few pro- and postprandial glucose concentrations glucose dysregulation are not fully un- spective studies have suggested that self- increased. Moreover, circadian mis- derstood, multiple mechanistic pathways reported sleep duration has a U-shaped alignment led to a change in appetite- are likely to be involved. Sleep restriction association, with increased risk of de- regulating hormones, a decrease in can increase sympathetic nervous system veloping T2D withboth short (,5–6h per energy expenditure, and an increase in activity, leading to a decrease in insulin night) and long (.8–9 h per night) sleep BP (34,36). However, it is important to sensitivity (19221). Activation of the hy- (30,31). The mechanisms by which long point out that these studies used short- pothalamic-pituitary-adrenal axis with sleep duration leads to increased risk of term extreme circadian misalignment. A elevation of can also decrease obesity and T2D are not fully understood. few population-based studies have re- insulin sensitivity (5,18). However, it is Long sleep may reflect a more sedentary ported an association between milder important to note that the studies of lifestyle and, similar to short sleepers, forms of circadian misalignment, such as sleep manipulation have been short- long sleepers engage in more snacking later chronotype or social jet lag, with term in nature and performed primarily (32). Although both long and short sleep prediabetes and T2D (37,38), but the ef- in healthy young individuals. It is less clear have been associated with worse glyce- fect of circadian misalignment on actual whether sleep manipulation impacts mic control in patients with established measures of glycemia in people with glycemic measures in patients with pre- T2D, (33) there is a paucity of data on the prediabetes has not been well studied. diabetes or T2D. impact of sleep duration or quality on Only one study explored the association 6 Sleep and Circadian Measures in Prediabetes Diabetes Care

– between chronotype and HbA1c in pre- about treatment of OSA was not col- Nordisk sponsored clinical trial. T.A.B. has re- diabetes. This study was performed in a lected during the screening phase of ceived research support from Allergan and clinical cohort of participants with pre- RISE, and as such, we cannot account Apollo Endosurgery. K.J.M. holds an investigator- initiated research grant from Novo Nordisk. No diabetes in Thailand and found a small for any potential effect of apnea treat- other potential conflicts of interest relevant to effect of chronotype on HbA1c (7). For ment on glycemic control. We also did this article were reported. each 2 h of later chronotype, HbA1c in- not collect data on antihypertensives Author Contributions. Members of the RISE creased by only 0.04% (0.4 mmol/mol). at the time of screening. Lastly, lack of a Consortium recruited participants and collected fi study data. B.M. interpreted data and wrote the There was not a signi cant relationship control group not at increased risk for first draft. K.A.T., K.M.U., K.J.N., T.S.H., S.S., E.B., between social jet lag and HbA1c.This diabetes is another limitation of our S.M., D.A.E., and E.V.C. reviewed and edited the may have been in part due to a very study. manuscript. B.M. and E.V.C. proposed the anal- narrow range of social jet lag in their In summary, sleep duration was in- ysis. A.H.T. and S.L.E. performed the analysis. The RISE Steering Committee reviewed and edited patient population (7). Similarly, we did dependently associated with HbA1c in fi the manuscript and approved its submission. not nd an association between later adults with prediabetes/recently diag- B.M. and A.H.T. are the guarantors of this work chronotype or social jet lag and measures nosed, untreated T2D. This relationship and, as such, had full access to all of the data in of glycemia. However, we did find an was most pronounced in those who the study and take responsibility for the integrity of association between these variables and reported ,5hor.8 h of sleep per the data and the accuracy of the data analysis. higher BP. Although in our cohort shift night (U-shaped relationship). Both short Prior Presentation. Parts of this study were presented in abstract form at the 78th Scientific work was not associated with measures sleep duration and shift work were also Sessions of the American Diabetes Association, of glycemia, it was associated with higher independently associated with higher Orlando, FL, 22–26 June 2018. BMI, as has been shown by others. Shift BMI in this population. Later chronotype work has been associated with increased and social jet lag were associated with References risk of obesity (39), particularly abdom- higher BP. Further research using objec- 1. Tabak´ AG, Herder C, Rathmann W, Brunner EJ, inal obesity (40), and the risk of devel- tive measures of sleep and circadian Kivimaki¨ M. Prediabetes: a high-risk state for – oping T2D (6). markers is needed to better delineate diabetes development. Lancet 2012;379:2279 2290 Our study has several strengths. First the relationship between sleep distur- 2. Centers for Disease Control and Prevention. and foremost, we studied a large number bances, circadian misalignment, and National Diabetes Statistics Report, 2017 [Inter- of adults with prediabetes/recently di- cardio-metabolic factors in prediabetes net], 2017, p. 1–20. Available from https:// agnosed, untreated T2D. By virtue of our and T2D in order to determine whether www.cdc.gov/diabetes/pdfs/data/statistics/ identification of participants in a study intervention studies targeting these novel national-diabetes-statistics-report.pdf. Accessed 1 February 2018 screening program, none of the partic- lifestyle factors to decrease the rate of 3. Heinzer R, Vat S, Marques-Vidal P, et al. ipants had been previously treated with conversion from prediabetes to diabetes Prevalence of sleep-disordered breathing in any confounding glucose-lowering med- are warranted. the general population: the HypnoLaus study. ications or medications known to affect Lancet Respir Med 2015;3:310–318 glucose metabolism. The cohort was eth- 4. McMenamin T. A time to work: recent trends in shift work and flexible schedules. Mon Labor nically diverse, and both sexes were well Acknowledgments. The RISE Consortium Rev 2007;130:3–15 represented, thereby increasing the gen- thanks the RISE Data and Safety Monitoring 5. Reutrakul S, Mokhlesi B. Obstructive sleep eralizability of our findings. Moreover, Board and Barbara Linder, the National Institute apnea and diabetes: a state of the art review. the proportion of participants who re- of Diabetes and Digestive and Kidney Diseases Chest 2017;152:1070–1086 Program Official for RISE (Rockville, MD), for 6. Anothaisintawee T, Reutrakul S, Van Cauter E, ported shift work was similar to that in support and guidance. The Consortium also Thakkinstian A. Sleep disturbances compared to the U.S. workforce (4). Second, all partic- thanks the participants who, by volunteering, traditional risk factors for diabetes development: ipants underwent a 2-h OGTT as well as are furthering the ability to reduce the burden systematic review and meta-analysis. Sleep Med of diabetes. Rev 2016;30:11–24 HbA1c measurements on the same day the sleep questionnaires were com- Funding. RISE is supported by grants from the 7. Anothaisintawee T, Lertrattananon D, National Institute of Diabetes and Digestive and Thamakaison S, Knutson KL, Thakkinstian A, pleted. Plasma glucose and HbA1c from Kidney Diseases, National Institutes of Health Reutrakul S. Later chronotype is associated all participating centers were measured (U01-DK-094406, U01-DK-094430, U01-DK-094431, with higher hemoglobin A1c in prediabetes pa- by a centralized laboratory. Lastly, we U01-DK-094438, U01-DK-094467, P30-DK-017047, tients. Chronobiol Int 2017;34:393–402 used standardized sleep questionnaires P30-DK-020595, P30-DK-045735, P30-DK-097512, 8. Anothaisintawee T, Lertrattananon D, across all centers. Notwithstanding the UL1-TR-000430, UL1-TR-001082, UL1-TR-001108, Thamakaison S, Thakkinstian A, Reutrakul S. UL1-TR-001855, UL1-TR-001857, UL1-TR-001858, The relationship among morningness-eveningness, strengths, our study has several important and UL1-TR-001863), and the Department of sleep duration, social jetlag, and body mass index limitations. Due to its cross-sectional de- Veterans Affairs. In addition, the National in Asian patients with prediabetes. Front Endo- sign, the direction of causality cannot be Heart, Lung, and Blood Institute provided sup- crinol (Lausanne) 2018;9:435 ascertained. Another important weak- port for the RISE Sleep Ancillary study to B.M. 9. RISE Consortium. Restoring Insulin Secretion fi b ness is lack of objective measures of (R01HL119161). Additional nancial and mate- (RISE): design of studies of -cell preservation in rial support from the American Diabetes Associ- prediabetes and early type 2 diabetes across the sleep and circadian markers. We did not ation was received. life span. Diabetes Care 2014;37:780–788 measure several important confounders Duality of Interest. RISE is also supported by 10. American Diabetes Association. Standards such as total calorie consumption, mac- Kaiser Permanente Southern California. Addi- of medical care in diabetes–2014. Diabetes Care ronutrient dietary composition, meal tional financial and material support from Aller- 2014;37(Suppl. 1):S14–S80 gan, Apollo Endosurgery, Abbott Laboratories, 11. Knutson KL, Ryden AM, Mander BA, Van timing, and amount of physical activity and Novo Nordisk was received. S.A.A. and S.E.K. Cauter E. Role of sleep duration and quality in and, as such, cannot control for these serve as paid consultants on advisory boards for the risk and severity of type 2 diabetes mellitus. covariates in our analysis. Information Novo Nordisk. S.A.A. is a participant in a Novo Arch Intern Med 2006;166:1768–1774 care.diabetesjournals.org Mokhlesi and Associates 7

12. Pilz LK, Keller LK, Lenssen D, Roenneberg T. 23. Engeda J, Mezuk B, Ratliff S, Ning Y. Asso- 32. Tan X, Chapman CD, Cedernaes J, Benedict C. Time to rethink sleep quality: PSQI scores reflect ciation between duration and quality of sleep Association between long sleep duration and sleep quality on workdays. Sleep (Basel) 2018;41 and the risk of pre-diabetes: evidence from increased risk of obesity and type 2 diabetes: 13. Buysse DJ, Reynolds CF III, Monk TH, Berman NHANES. Diabet Med 2013;30:676–680 a review of possible mechanisms. Sleep Med Rev SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: 24. Kim CR, Song YM, ShinJY,Gim W. Association 2018;40:127–134 a new instrument for psychiatric practice and between sleep duration and impaired fasting 33. Lee SWH, Ng KY, Chin WK. The impact of research. Psychiatry Res 1989;28:193–213 glucose in Korean adults: results from the Korean sleep amount and sleep quality on glycemic 14. RoennebergT,Wirz-JusticeA,MerrowM.Life National Health and Nutrition Examination Sur- control in type 2 diabetes: a systematic review between clocks: daily temporal patterns of human vey2011-2012.KoreanJFamMed2016;37:51–56 and meta-analysis. Sleep Med Rev 2017;31:91– chronotypes. J Biol Rhythms 2003;18:80–90 25.YanM,FuZ,QinT,etal.Associationsofsleep 101 15. Roenneberg T, Allebrandt KV, Merrow M, duration and prediabetes prevalence in a middle- 34. Scheer FA, Hilton MF, Mantzoros CS, Shea Vetter C. Social jetlag and obesity. Curr Biol aged and elderly Chinese population with regard SA. Adverse metabolic and cardiovascular con- 2012;22:939–943 to age and hypertension: the China Health and sequences of circadian misalignment. Proc Natl 16. Netzer NC, Stoohs RA, Netzer CM, Clark K, Retirement Longitudinal Study baseline survey. Acad Sci U S A 2009;106:4453–4458 Strohl KP. Using the Berlin Questionnaire to J Diabetes 2018;10:847–856 35. Leproult R, Holmback¨ U, Van Cauter E. identify patients at risk for the sleep apnea 26. Kim CW, Chang Y, Sung E, Ryu S. Sleep Circadian misalignment augments markers of syndrome. Ann Intern Med 1999;131:485–491 duration and progression to diabetes in people insulin resistance and inflammation, indepen- 17. Johns MW. A new method for measuring with prediabetes defined by HbA1c concentra- dently of sleep loss. Diabetes 2014;63:1860– daytime sleepiness: the Epworth sleepiness tion. Diabet Med 2017;34:1591–1598 1869 scale. Sleep 1991;14:540–545 27. Nakajima K, Suwa K, Toyama K. Age- 36. McHill AW, Melanson EL, Higgins J, et al. 18. Reutrakul S, Van Cauter E. Sleep influences dependent changes in the association between Impact of circadian misalignment on energy on obesity, insulin resistance, and risk of type 2 sleep duration and impaired glucose metabo- metabolism during simulated nightshift work. diabetes. Metabolism 2018;84:56–66 lism. World J Diabetes 2017;8:397–406 Proc Natl Acad Sci U S A 2014;111:17302– 19. Buxton OM, Pavlova M, Reid EW, Wang W, 28. Nuyujukian DS, Beals J, Huang H, et al.; 17307 Simonson DC, Adler GK. Sleep restriction for Special Diabetes Program for Indians Diabetes 37. Koopman ADM, Rauh SP, van ’t Riet E, et al. 1 week reduces insulin sensitivity in healthy Prevention Demonstration Project. Sleep dura- The association between social jetlag, the met- men. Diabetes 2010;59:2126–2133 tion and diabetes risk in American Indian and abolic syndrome, and type 2 diabetes mellitus in 20. Robertson MD, Russell-Jones D, Umpleby Alaska native participants of a lifestyle interven- the general population: the New Hoorn Study. AM, Dijk DJ. Effects of three weeks of mild sleep tion project. Sleep 2016;39:1919–1926 J Biol Rhythms 2017;32:359–368 restriction implemented in the home environment 29. Kowall B, Lehnich AT, Strucksberg KH, et al. 38. Yu JH, Yun CH, Ahn JH, et al. Evening on multiple metabolic and endocrine markers in Associations among sleep disturbances, noctur- chronotype is associated with metabolic disor- healthy young men. Metabolism 2013;62:204–211 nal sleep duration, daytime napping, and in- ders and body composition in middle-aged 21. Broussard JL, Ehrmann DA, Van Cauter E, cident prediabetes and type 2 diabetes: the adults. J Clin Endocrinol Metab 2015;100: Tasali E, Brady MJ. Impaired insulin signaling in Heinz Nixdorf Recall Study. Sleep Med 2016; 1494–1502 human adipocytes after experimental sleep re- 21:35–41 39. Liu Q, Shi J, Duan P, et al. Is shift work striction: a randomized, crossover study. Ann 30. Cappuccio FP, D’Elia L, Strazzullo P, Miller associated with a higher risk of overweight or Intern Med 2012;157:549–557 MA. Quantity and quality of sleep and incidence obesity? A systematic review of observational 22. Rafalson L, Donahue RP, Stranges S, et al. of type 2 diabetes: a systematic review and studies with meta-analysis. Int J Epidemiol 2018; Short sleep duration is associated with the de- meta-analysis. Diabetes Care 2010;33:414–420 47:1956–1971 velopment of impaired fasting glucose: the 31. Shan Z, Ma H, Xie M, et al. Sleep duration and 40. Sun M, Feng W, Wang F, et al. Meta-analysis Western New York Health Study. Ann Epidemiol risk of type 2 diabetes: a meta-analysis of pro- on shift work and risks of specific obesity types. 2010;20:883–889 spectivestudies.DiabetesCare2015;38:529–537 Obes Rev 2018;19:28–40