Circadian misalignment impacts the association of visceral adiposity with metabolic syndrome in adolescents

Abstract

Study Objectives

Although insufficient sleep is a risk factor for metabolic syndrome (MetS), the circadian timing of sleep (CTS) is also involved in cardiac and metabolic regulation. We examined whether delays and deviations in the sleep midpoint (SM), a measure of CTS, modify the association between visceral adipose tissue (VAT) and MetS in adolescents.

Methods

We evaluated 277 adolescents (median 16 years) who had at least 5 nights of at-home actigraphy (ACT), in-lab polysomnography (PSG), dual-energy X-ray absorptiometry (DXA) scan, and MetS score data. Sleep midpoint (SM), sleep irregularity (SI), and social jetlag (SJL) were examined as effect modifiers of the association between VAT and MetS, including waist circumference, blood pressure, insulin resistance, triglycerides, and cholesterol. Linear regression models adjusted for demographics, ACT-sleep duration, ACT-sleep variability, and PSG-apnea–hypopnea index.

Results

The association between VAT and MetS was significantly stronger (p-values for interactions < 0.001) among adolescents with a schooldays SM later than 4:00 (2.66 [0.30] points increase in MetS score), a SI higher than 1 hour (2.49 [0.30]) or a SJL greater than 1.5 hours (2.15 [0.36]), than in those with an earlier SM (<3:00; 1.76 [0.28]), lower SI (<30 minutes; 0.98 [0.70]), or optimal SJL (<30 minutes; 1.08 [0.45]).

Conclusions

A delayed sleep phase, an irregular sleep–wake cycle, and greater social jetlag on schooldays identified adolescents in whom VAT had a stronger association with MetS. Circadian misalignment is a risk factor that enhances the impact of visceral obesity on cardiometabolic morbidity and should be a target of preventative strategies in adolescents.

Graphical Abstract

Graphical Abstract.

Statement of Significance.

Insufficient sleep is a recognized risk factor for cardiometabolic morbidity, while circadian misalignment is not despite being highly prevalent in adolescents. This study shows that the association of visceral adiposity with metabolic syndrome is worse among youth with circadian misalignment. These data have the potential to inform targeted prevention by prompting early identification of different forms of circadian misalignment and focusing on aligning school start times to adolescents’ shifting circadian phase. It is plausible that maintaining a regular and/or in-phase circadian timing of sleep may protect adolescents from the adverse cardiometabolic outcomes associated with visceral obesity. Furthermore, adolescents with obesity should be screened for circadian misalignment as they may have a worse cardiometabolic prognosis, and clinical evaluations of circadian sleep–wake patterns should account for different entrainment scenarios.

Introduction

Pediatric obesity, which increases the risk of developing chronic diseases later on in life, has reached epidemic proportions in the United States, with the prevalence of obesity nearly tripling in adolescents since the 1970s [1]. Adolescence is a particularly vulnerable time period for the onset of adverse cardiovascular and metabolic disorders as there is a developmental shift in adipose tissue distribution towards the visceral depot [2]. Visceral adipose tissue (VAT) is hormonally active and has the biochemical characteristics capable of activating pro-inflammatory cytokines, causing metabolic dysfunction [3]. Therefore, VAT is the phenotype of choice to identify metabolically unhealthy individuals, and it is a risk factor for the adverse cardiovascular and metabolic health outcomes associated with weight gain [4]. Excessive accumulation of VAT is strongly associated with cardiovascular diseases [5], type 2 diabetes [6], and other forms of cardiometabolic morbidity [7] as early as adolescence [8, 9], including subclinical insulin resistance [10] or impaired glucose tolerance [11], which are markers of the metabolic syndrome (MetS).

MetS encompasses an array of adverse cardiometabolic factors, including central obesity, elevated blood pressure (BP), insulin resistance or hyperglycemia, elevated triglycerides, and low high-density lipoprotein (HDL) cholesterol levels [12]. In adults, the clustering of three or more of these components into MetS increases the risk of developing CVD above and beyond the individual impact of each component [13]. At a public health level, the early diagnosis of youth with elevated MetS is imperative as data from the US National Health and Nutrition Examination Survey shows that the prevalence of adolescents with MetS has increased from 4% in 1988 to 11% in 2019 [14, 15], and additionally, it has been estimated that 30% to 50% of adolescents with obesity, as per BMI percentile criteria, have MetS [16]. However, unlike adults, the definitions used for MetS and each one of the five cardiometabolic components, except for exceptions like BP, are not universal and agreed upon for adolescents even across the abovementioned studies [17, 18]. To address this gap, epidemiological studies in adolescents have defined MetS through the use of a continuous MetS (cMetS) score that allows examining preclinical MetS impacts before development of clinical levels in each single component, and their full clustering [19].

Given prior rigorous evidence demonstrating an association between VAT with MetS [7, 8, 16], research has focused on novel factors, such as inadequate sleep, that may contribute to increase/enhance such a well-established association. As such, the importance of an adequate amount of sleep has now been widely recognized [20], as insufficient sleep is associated with obesity, cardiovascular, and metabolic dysfunction [21, 22]. However, recent data indicates that, beyond insufficient sleep, deviations in circadian or sleep timing may be more strongly associated with VAT and MetS than average metrics alone [8, 23–25]. Circadian misalignment of the sleep–wake cycle is defined as a mismatch between the endogenous circadian system, governed by the suprachiasmatic nucleus of the hypothalamus, and 24-hour environmental and behavioral synchronizers [26, 27]. During adolescence, there are changes in the circadian timing of sleep (CTS) that are naturally brought by the onset of puberty, where youth’s sleep phase becomes more delayed compared to childhood [28]. Due to the potential mismatch between this maturational change in the CTS and the social/academic responsibilities during weekdays, when rising times are early but steady due to school start times (entrained), and during weekends (free days), when bedtime and rising times become more delayed and adolescents sleep-in [28–31], they may experience different forms of circadian misalignment: delayed sleep phase, sleep irregularity, or social jetlag [28]. A misalignment of the sleep–wake cycle has been reported to contribute to increased abdominal obesity, impaired glucose tolerance, reduced insulin sensitivity, and elevated BP [21, 31]. Thus, circadian misalignment has become a new window into understanding the interplay between VAT, sleep and MetS. However, data are greatly lacking in adolescence, despite circadian misalignment being most prevalent during this developmental period.

From a conceptual standpoint, most prior studies have focused on the independent impact that circadian misalignment variables (e.g. delayed sleep phase, irregular sleep, and social jetlag) may have on the known obesity-related cardiometabolic health outcomes [21, 31]. As shown in Figure 1, we have previously proposed a moderating role for alterations in the CTS during adolescence in the increased cardiometabolic risk associated with obesity that identifies “under what conditions” or “in whom” the impact of increased VAT on MetS is greatest [32]. Prior research has shown that circadian misalignment disrupts underlying pathophysiological mechanisms, such as immune pathways, which may lead VAT to over-stimulate the release of pro-inflammatory cytokines, contributing to worsening cardiovascular and metabolic function, and increasing the development of MetS [3, 33]. As compared to prior studies [10, 11, 32], the current study offers a view of MetS as a multidimensional cardiometabolic health risk score in youth. Thus, our current study examines the sum of all five factors of MetS into a single score, and we propose that greater delays (sleep midpoint), deviations (sleep irregularity), and weekday-to-weekend differences (social jetlag) in the sleep–wake cycle may increase the association between VAT and MetS impacts in adolescents.

Figure 1.

Conceptual model tested in this study. It is well-established that (A) increased visceral adiposity is associated with adverse cardiometabolic outcomes (e.g. blood pressure, insulin resistance, triglycerides, and cholesterol); however, most prior research that (B) examined the role of the circadian timing of sleep (CTS) has studied the independent association of metrics of circadian misalignment (e.g. sleep midpoint and its regularity) with either visceral adiposity or cardiometabolic health outcomes, without accounting for the known association between the latter two. Our conceptual model tested in this study posits that (C) deviations in the CTS act as effect modifiers (moderators) of the known association between increased visceral adiposity and adverse cardiometabolic outcomes, helping identify “under what circumstances” or “in whom” visceral adiposity has the strongest impact on cardiometabolic health outcomes (e.g. highly irregular CTS) as well as “under what circumstances” or “in whom” there is resilience for such adverse health outcomes (e.g. a CTS aligned to the biological night). Reproduced with permission from Morales-Ghinaglia and colleagues [32].

Methods

Sample

The data that support the findings of this study are available from the corresponding author upon reasonable request. The Penn State Child Cohort has been previously described in detail elsewhere [34, 35]. In brief, the Penn State Child Cohort was designed as a three-phase study in which 700 randomly selected participants ages 5–12 years from central Pennsylvania underwent an in-laboratory study between 2002 and 2006. The adolescent visit conducted from 2010 to 2013 consisted of 421 participants who returned 6 to 13 years later (median, 7.4 years) for a reexamination (60.1% response rate). No differences in baseline demographic characteristics were observed in the 279 participants who did not return for a follow-up [33, 35]. Participants were examined at the Clinical Research Center at Penn State College of Medicine. The study was approved by The Penn State College of Medicine Institutional Review Board. Written informed consent was acquired from the parents or legal guardians and from participants 18 years or older, and assent from the participants younger than 18 years as well. Participants who completed the study received $200 in compensation.

Visceral adiposity

During their in-lab visit, a certified technician conducted a whole-body dual-energy X-ray absorptiometry (DXA) scan (Hologic QDR 4500W) to measure fat composition and distribution in the fasted state [24, 25, 32, 35, 36]. The DXA scan uses beams of low-energy X-ray that pass by the body tissue and are collected by detectors after attenuation by the body tissue through which they have passed; hence, participants are required to remove all metal, plastic, and rubber materials to avoid impact on x-ray beams [8]. Daily quality control and calibrations are performed on the DXA machine to assure that our standardized operation protocol is followed, assuring data validity. VAT in cm² was selected as the body adiposity region of interest to measure central obesity. VAT was measured by calculating the amount of SAT over the visceral cavity and subtracting SAT from the total abdominal fat using appropriate modeling which was identified by the Hologic APEX 4.0 software (Hologic Inc., Bedford, MA) and visually verified by an experienced investigator [8, 32]. DXA measures of abdominal fat are suitable for use in children and adolescents and provide a good indication of VAT in youth [37]. VAT was treated as the independent variable.

Metabolic syndrome score

Metabolic syndrome was assessed using a previously validated cMetS score derived from the sum of the age-and-sex-adjusted z-scores of the 5 established MetS components: waist circumference (WC), mean arterial pressure, homeostasis model assessment of insulin resistance (HOMA-IR), triglycerides (TG), and HDL [8, 9, 35, 38–40]. WC was measured at the top of the iliac crest. Systolic and diastolic BP levels were taken three consecutive times in the seated position, and the average of the last two measurements were used as the outcome. Mean arterial pressure was calculated as diastolic pressure + 1/3 systolic pressure. Glucose, insulin, and lipid profiles were assayed from a fasting venous blood sample obtained the morning after the in-lab sleep study. HOMA-IR was calculated as the fasting insulin level (in µU/mL) and fasting glucose level (in mmol/L) divided by 22.5 [9]. As HDL level is inversely associated with metabolic risk, it was multiplied by −1 [8]. A higher cMetS score is indicative of a more adverse cardiometabolic profile and higher MetS impacts [8, 9, 35]. Because the cMetS score is based on sex-and-age adjusted z-scores, these standard deviations can be interpreted in a clinically meaningful manner [9]. cMetS score was treated as the dependent variable.

Actigraphy

During the week following the Clinical Research Center visit, participants wore an actigraphy (ACT; accelerometer) device on their non-dominant wrist (ActiGraph GT3X) for 7 consecutive nights before mailing it back to our lab [24, 25, 32]. Actigraphy-estimated sleep variables were used as surrogate measures of circadian misalignment in the current study. Participants additionally completed sleep logs to record bedtime and rising times daily to increase the accuracy of the scoring of ACT-derived sleep estimates using Sadeh’s algorithm (Actlife 6 software, ActiGraph LLC, Pensacola, FL, USA). Table 1 describes each of the key sleep and circadian constructs extracted from actigraphy and described below.

Table 1.

Sleep and Circadian Constructs Used in this Cohort Study

Construct	Source variable	Formula(s)	Calculation	Meaning
Sleep duration	Minutes of total sleep time (TST)	TIB – (SOL + WASO)	Within-participant mean TST across consecutive nights	Amount of sleep obtained during the main sleep episode
Sleep variability	Day-to-day deviations in TST	StDev of TST	Within-participant StDev of TST across consecutive nights	Alternation of insufficient, normative, and long-duration sleep episodes across the week
Sleep midpoint	Clock time of the middle point of the sleep period (SM)	SOT + (SOT—WUT)/ 2	Within-participants mean SM across consecutive nights	Timing of the main sleep episode within the 24 hours
Sleep irregularity	Day-to-day deviations in SM	StDev of SM	Within-participant StDev of SM across consecutive nights	Alternation of advanced, normative, or delayed sleep episodes across the week
Social jetlag	Weekdays-to-weekend deviation in SM	| SMweekends—SMweekdays |	Within-participant absolute difference of SM on weekends minus SM on weekdays	Alternation of advanced, normative, or delayed sleep episodes between weekdays and weekends

SM, sleep midpoint; SOL, sleep onset latency; SOT, sleep onset time; StDev, standard deviation; TIB, time in bed; TST, total sleep time; WASO, wake after sleep onset; WUT, wakeup time.

Sleep duration measured by the ACT was calculated as the intraindividual mean of the 7-night total sleep time (TST; total number of minutes slept measured by actigraphy since bedtime until rising time identified by sleep log). Sleep variability was calculated as the intraindividual standard deviation in TST across the 7 nights [24, 25, 31]. These two sleep variables were treated as covariates in order to adjust for habitual sleep duration and night-to-night sleep variability (i.e. nights of insufficient vs. normative vs. extended sleep).

SM was calculated as the ACT-measured mean midpoint of the sleep period such as in (sleep onset time—wakeup time)/2 across the 7 nights [31]. SI was calculated as the intraindividual standard deviation in SM [31, 32, 41]. These two circadian variables were treated as effect modifiers/moderators in order to examine their role in enhancing the impact of VAT on cMetS score.

All sleep metrics were calculated for weekdays (5 nights), while only sleep duration and SM were calculated for weekends (2 nights) in order to capture any particular changes in the timing of sleep that varied depending on nights of the week [32]. The absolute difference between mean SM on weekdays versus weekends was also calculated as a continuous measure of SJL with the formula: ∆SM = |weekends SM—weekdays SM| [42, 43]. A higher (positive) ∆ SM is indicative of greater SJL, understood as participants delaying their sleep–wake cycle during weekends compared to weekdays. The same formula was used for bedtime, rising times, and sleep duration.

Based on the dates of the study and reports by the participants, we identified participants who were studied while “in-school” versus while “on-break” to examine the influence that entrainment may have on circadian misalignment as a function of free days versus days determined by environmental factors (i.e. schooling) [32].

Polysomnography

All participants partaking in our sleep study were evaluated for one night in the sleep laboratory in a sound-attenuated, light, and temperature-controlled room using 16-channel PSG. Each participant was continuously monitored from 22:00 to 07:00 hours. The sleep recordings collected were scored based on standard criteria, and followed by an evaluation of the parameters of sleep continuity and sleep architecture. The focus of this article was on sleep continuity factors such as sleep onset latency (the number of minutes to fall asleep since lights off), wake after sleep onset (the amount of time awake after the onset of sleep), and TST (the total number of minutes slept since lights off until lights on). The apnea–hypopnea index (AHI) was determined based on the number of apneas and hypopneas per hour of sleep, as previously reported [32, 34–36, 40]. The AHI was treated as a covariate.

Other key measures

Participants underwent a clinical history prior to the PSG and included recording demographic characteristics, such as age (years), race (black/white), ethnicity (Hispanic/non-Hispanic), sex (assigned male/female at birth), Tanner staging [44], and Morningness–Eveningness Questionnaire to assess chronotype [44]. Age was treated as a continuous variable, while sex (male = 0 and female = 1) and race/ethnicity (non-Hispanic white = 0 and racial/ethnic minority = 1) were treated as binary variables. In our sample, there were 61 participants (21%) who identified as a racial/ethnic minority and, given the inability to stratify with such sample size, we statistically controlled for the potential effect of underlying racial/ethnic disparities. These variables were treated as covariates.

Statistical analyses

Among the 421 participants, 391 had DXA data, 381 had cMetS data and 327 had at least 5 nights of ACT. The effective sample size for this study consisted of 277 participants with complete DXA, cMetS, and ACT data. There were no significant demographic differences (e.g. sex, race/ethnicity, BMI percentile, and all p > 0.340) between the 277 participants included in the analyses and the 144 excluded, except in terms of age with the former 0.8 years younger (15.9 (2.1]) than the latter (16.7 [2.3], p < 0.02).

In our analysis, we first evaluated the cross-sectional association (main effects) of the independent variable VAT and the effect modifiers SM, SI, and SJL with MetS score using multivariable linear regression controlling for sex, age, race/ethnicity, AHI, sleep duration, and sleep variability. Regression coefficients and their standard errors for the independent variable and each effect modifier were computed after all these variables were standardized as z-scores. To test circadian misalignment variables as effect modifiers of VAT on MetS score, the interaction terms between VAT and SM, SI, and SJL (with main effects in the model) were also fitted. If an interaction term was statistically significant, stratified analyses were conducted based on the quartile distribution rounded to clinically meaningful numbers from prior literature [32, 45, 46]. These models were further fitted separately for the in-school group versus on-break group to test the entrainment conditions that participants were evaluated under.

Finally, sensitivity analyses tested whether the study findings were similar and in the same direction when the WC component was removed from the cMetS score for potential overlap with VAT. All analyses were performed using SPSS, version 27 (IBM Corp). The two-sided p-values were deemed statistically significant at p < 0.05.

Results

Characteristics of the sample

As shown in Table 2, among the 277 participants who were 12 to 23 years old (median, 16.2 years), 144 (52%) were male, 61 (22%) identified as a racial/ethnic minority, and 100 (36.1%) participants were studied while on-break from school. These on-break participants were not significantly different in terms of sex (p = 0.457), race/ethnicity (p = 0.279), VAT (p = 0.165), or cMetS score (p = 0.326) as compared to those studied while in-school. There were no significant differences in SI overall (p = 0.156) or on weekdays (p = 0.345) between the two sets of participants. Participants studied while on-break were on average 9.6 months older than those studied while in-school (p = 0.011), had an average SM an hour and 8 minutes later (p < 0.001), slept on average 24 minutes longer on weekdays (p < 0.001) and with 15 minutes greater variability (p = 0.002), but slept on average 39.1 minutes shorter on weekends (p < 0.001). Moreover, participants studied while on-break had a lower SJL (p < 0.001) than those studied while in-school. These sleep and circadian ACT data are commensurate with prior studies [28, 29, 47].

Table 2.

Characteristics of the Sample

	Overall (N = 277)	In-School (N = 177)	On-Break (N = 100)	P-value
Demographics
Female sex, %	48.0%	46.3%	51.0%	0.457
Racial/ethnic minority, %	22.0%	20.0%	26.0%	0.279
Age, years	16.2 (2.2)	15.9 (2.1)	16.7 (2.3)	0.011
Tanner stage, score	4.1 (0.8)	4.1 (0.8)	4.2 (0.8)	0.171
Anthropometrics
BMI, percentile	67.4 (27.7)	65.4 (28.6)	70.9 (25.8)	0.116
VAT, cm2	61.3 (40.3)	58.8 (40.4)	65.8 (40.0)	0.165
SAT, cm2	228.5 (160.1)	218.9 (154.6)	245.5 (168.7)	0.185
cMetS, score	0.1 (3.2)	0.0 (3.2)	0.4 (3.1)	0.326
WC (cm)	80.75 (13.90)	80.42 (14.35)	81.33 (13.13)	0.600
MAP (mm Hg)	82.57 (8.65)	81.53 (8.74)	84.41 (8.22)	0.008
TG (mg/dL)	94.38 (47.98)	90.22 (38.53)	101.74 (60.78)	0.055
HDL (mg/dL)	50.09 (12.99)	50.56 (12.76)	49.26 (13.43)	0.425
Log HOMA-IR	−0.15 (0.83)	−0.12 (0.91)	−0.19 (0.68)	0.479
Actigraphy
Overall
Sleep midpoint, hh:mm	3:56 (1:34)	3:32 (1:38)	4:40 (1:07)	<0.001
Sleep irregularity, hh:mm	0:56 (0:32)	0:58 (0:35)	0:52 (0:25)	0.156
Sleep duration, min	420.7 (49.8)	418.2 (47.9)	425.2 (52.9)	0.260
Sleep variability, min	70.6 (36.5)	67.9 (34.7)	75.3 (39.4)	0.105
Weekdays
Bedtime, hh:mm	23:23 (1:31)	22:58 (1:27)	24:08 (1:21)	<0.001
Rising time, hh:mm	7:53 (1:51)	7:16 (1:49)	8:59 (1:22)	<0.001
Sleep midpoint, hh:mm	3:15 (2:15)	2:53 (2:35)	3:54 (1:15)	<0.001
Sleep irregularity, hh:mm	0:49 (0:48)	0:47 (0:56)	0:52 (0:31)	0.345
Sleep duration, min	419.0 (53.2)	410.7 (49.2)	433.8 (56.8)	<0.001
Sleep variability, min	62.6 (38.8)	57.2 (36.0)	72.2 (41.8)	0.002
Weekends
Bedtime, hh:mm	23:52 (1:31)	23:35 (1:29)	24:21 (1:28)	<0.001
Rising time, hh:mm	8:45 (1:28)	8:37 (1:33)	8:59 (1:15)	0.042
Sleep midpoint, hh:mm	3:53 (2:08)	3:48 (2:30)	4:02 (1:15)	0.371
Sleep irregularity, hh:mm	n/a	n/a	n/a	n/a
Sleep duration, min	446.3 (82.3)	460.3 (81.3)	421.2 (78.5)	<0.001
Sleep variability, min	n/a	n/a	n/a	n/a
Weekends—Weekdays
∆Bedtime, hh:mm	0:28 (1:11)	0:37 (1:14)	0:13 (1:03)	0.007
∆Rising time, hh:mm	0:51 (1:45)	1:20 (1:44)	−0:00 (1:26)	<0.001
∆Sleep midpoint, hh:mm	1:09 (1:03)	1:20 (1:09)	0:49 (0:44)	<0.001
∆Sleep irregularity, hh:mm	n/a	n/a	n/a	n/a
∆Sleep duration, min	27.0 (84.7)	49.6 (77.1)	−13.4 (82.9)	<0.001
∆Sleep variability, min	n/a	n/a	n/a	n/a
Morningness–Eveningness
Morning-type, %	33.2%	31.6%	36.0%	0.208
Intermediate-type, %	35.7%	39.5%	29.0%
Evening-type, %	31.0%	28.8%	35.0%
Polysomnography
Sleep duration, min	444.7 (59.5)	453.0 (53.2)	430.0 (67.0)	0.002
AHI, events/ hour	2.7 (6.3)	2.3 (2.9)	3.4 (9.7)	0.166

Data are mean and (standard deviation) for continuous variables and n (%) for categorical variables. p-value in the last column is for comparing in-school and on-break groups; bold p-values indicate a statistically significant difference between groups at p < 0.05. Note: the absolute ∆ Sleep midpoint for weekends—weekdays identifies social jetlag (SJL). Abbreviations: BMI, body mass index; cMetS, continuous metabolic syndrome score; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; MAP, mean arterial pressure; TG, triglycerides; WC, waist circumference.

SM, SI, and SJL as effect modifiers

As shown in Table 3, a significant interaction was found between VAT and ∆ SM from weekdays to weekends (social jetlag) on cMetS score (p = 0.017), but not between VAT and SM (p = 0.146) or VAT and SI (p = 0.055). Among the 72 adolescents with a ∆ SM 1 hour and 30 minutes or later on weekends than weekdays, each one standard deviation increase in VAT was associated with a 2.20 (0.29) points higher cMetS score (p < 0.001), an association that was weaker among the 124 adolescents with a ∆ SM 30 to 1 hour and 29 minutes later on weekends than weekdays (β = 1.78 (0.27), p < 0.001] in addition to among the 80 adolescents with a ∆ SM less than 30 minutes [β = 1.85 (0.26), p < 0.001] (Table 4).

Table 3.

Association of Visceral Adiposity and its Interaction With Circadian Timing of Sleep on cMetS Score

	Overall (N = 277)	In-school (N = 177)	On-break (N = 100)
Sleep midpoint
Visceral adiposity	2.07 (0.16) < 0.001	1.90 (0.23) < 0.001	2.34 (0.22) < 0.001
Sleep midpoint	0.20 (0.16) 0.217	0.26 (0.23) 0.257	−0.41 (0.28) 0.139
Visceral adiposity × Sleep midpoint	0.22 (0.15) 0.146	0.53 (0.24) 0.028	0.02 (0.24) 0.933
Sleep irregularity
Visceral adiposity	2.07 (0.16) < 0.001	1.90 (0.23) < 0.001	2.34 (0.22) < 0.001
Sleep irregularity	−0.01 (0.16) 0.936	−0.02 (0.21) 0.940	0.01 (0.25) 0.984
Visceral adiposity × Sleep irregularity	0.24 (0.13) 0.055	0.36 (0.15) 0.016	−0.21 (0.25) 0.401
Social jetlag
Visceral adiposity	2.07 (0.16) < 0.001	1.90 (0.23) < 0.001	2.34 (0.22) < 0.001
Social jetlag	−0.12 (0.14) 0.396	−0.07 (0.18) 0.714	−0.26 (0.32) 0.410
Visceral adiposity × Social jetlag	0.38 (0.16) 0.017	0.59 (0.21) 0.005	0.18 (0.31) 0.567

Data are regression coefficients (standard error) p-value for one standard deviation increase associated with each independent variable (main effects without interaction in the model) and their interaction (with main effects in the model). Adjusted for sex, age, race/ethnicity, polysomnography-measured apnea–hypopnea index, actigraphy-measured total sleep time, and actigraphy-measured variability in total sleep time. Bold p-values indicate a statistically significant interaction effect at p < 0.05.

Table 4.

Association of Visceral Adiposity With cMetS Score Across Different Levels of Circadian Misalignment

	Overall (N = 277)	In-School (N = 177)	On-Break (N = 100)
Sleep midpoint
< 3:00	n/a	1.76 (0.28) < 0.001 (n = 72)	n/a
3:00–3:59	n/a	1.18 (0.63) 0.064 (n = 60)	n/a
≥ 4:00	n/a	2.66 (0.30) < 0.001 (n = 45)	n/a
Sleep irregularity
< 0:30	n/a	0.98 (0.70) 0.174 (n = 38)	n/a
0:30–0:59	n/a	1.75 (0.34) < 0.001 (n = 69)	n/a
≥ 1:00	n/a	2.49 (0.30) < 0.001 (n = 70)	n/a
Social jetlag
< 0:30	1.85 (0.26) < 0.001 (n = 80)	1.08 (0.45) 0.022 (n = 41)	n/a
0:30–1:29	1.78 (0.27) < 0.001 (n = 124)	1.67 (0.38) < 0.001 (n = 82)	n/a
≥ 1:30	2.20 (0.29) < 0.001 (n = 72)	2.15 (0.36) < 0.001 (n = 54)	n/a

Data are regression coefficients (standard error) p-value for one standard deviation increase in visceral adipose tissue adjusted for sex, age, race/ethnicity, polysomnography-measured apnea–hypopnea index, actigraphy-measured total sleep time, and actigraphy-measured variability in total sleep time.

n/a = not applicable, as there was no significant interaction between visceral adiposity and the circadian-related metric that supported stratification at different levels of circadian misalignment. Bold p-values indicate a statistically significant association at p < 0.05.

Impact of participants being in-school versus on-break

As shown in Table 3, a significant interaction was found between VAT and SM on cMetS score in participants studied while in-school (p = 0.028), but not in those studied while on-break (p = 0.933). Among adolescents with the most delayed SM while in-school (≥ 4:00 am; n = 45), each one standard deviation increase in VAT was associated with 2.66 (0.30) points higher cMetS score (p < 0.001), an association that was weakest among adolescents with the earliest SM (<3:00 am; n = 72) in whom each one standard deviation increase in VAT was associated with a 1.76 (0.28) points higher cMetS score(p < 0.001) (Table 4).

Also, a significant interaction was found between VAT and SI on cMetS score in participants studied while in-school (p = 0.016), but not in those studied while on-break (p = 0.401) (Table 3). Among adolescents with the highest SI while in-school (≥1 hour; n = 70), each one standard deviation increase in VAT was associated with a 2.49 (0.30) points higher cMetS score (p < 0.001), whereas among those adolescents with the lowest SI (< 30 minutes; n = 38) VAT was not significantly associated with a higher cMetS score (β = 0.98 [0.70], p = 0.174) (Table 4).

As it pertained to SJL, a significant interaction was found between VAT and SJL on cMetS score in participants studied while in-school (p = 0.005), but not in those studied while on-break (p = 0.567) (Table 3). Specifically, among adolescents with the greatest SJL while in-school (≥1 hour and 30 minutes; n = 54), each one standard deviation increase in VAT was associated with a 2.15 (0.36) points higher cMetS score (p < 0.001), an association that was weaker among the 82 adolescents with a ∆ SM 30 to 1 hour and 29 minutes later on weekends than weekdays (β = 1.67 [0.38], p < 0.001) in addition to among the 41 adolescents with a ∆ SM less than 30 minutes (β = 1.08 [0.45], p = 0.022) (Table 4).

Sensitivity analyses

As shown in Supplementary Tables S1 and S2, the patterns of association reported above held well even after removing the WC component from the cMetS score because of its potential overlap with VAT. For example, the overall interaction reported above for SJL remained statistically significant (β = 0.25 [0.12], p = 0.040), as did the significant interactions reported above for SM (β = 0.38 [0.17], p = 0.023), SI (β = 0.30 [0.10], p = 0.005), and SJL (β = 0.39 [0.15], p = 0.009) among participants studied while in-school (Supplementary Table S1). Similarly, stratification showed similar patterns of association (Supplementary Table S2). All these data are reported in Supplementary Materials, while we keep the results including WC in the MetS score given that it represents the observable, easily measurable degree of central obesity considered part of MetS, while VAT represents the non-readily observable degree of adiposity with endocrine and pro-inflammatory characteristics underlying the adverse clustering of cardiovascular and metabolic outcomes, as discussed below. Finally, we also examined the significant interaction found while stratifying by VAT using a median split (<45 cm² vs. ≥45 cm²) and examining circadian misalignment metrics as the independent variables of MetS. As shown in Supplementary Table S3, we did not find evidence that VAT was the effect modifier in the significant interaction terms.

Discussion

This study demonstrates that the association of visceral adiposity with MetS is worse among youth with circadian misalignment. Specifically, our novel findings showed that in adolescents with a delayed sleep phase, an irregular sleep–wake cycle or greater social jetlag, VAT had a greater impact on MetS than in adolescents with an in-phase or regular sleep–wake cycle or without social jetlag. These associations were independent of relevant demographic factors, such as age, sex, and race/ethnicity, in addition to clinical factors, such as sleep apnea or sleep duration and its variability. Prior studies examining the direct association of circadian misalignment with cardiometabolic health [21, 26, 31, 48–50] support our novel findings on its role in obesity-related MetS. These studies have examined delayed SM [32, 50–53], SI [32, 50–53], and SJL [46, 52, 54–58] as meaningful phenotypes in youth. The findings of our population-based study also indicate that assessing these three meaningful circadian phenotypes and taking into consideration whether youth are in-school or on-break can better capture the role of circadian misalignment in increasing obesity-related cardiometabolic risk in youth.

Obesity is well-established as a major risk factor for chronic conditions [6, 7], particularly when accumulations occur in the visceral depot [3, 5]. However, obesity is a highly complex multifactorial condition [59], and identifying novel risk factors that may enhance its impact on cardiovascular and metabolic morbidity is essential for the development of preventive strategies and early interventions. While adequate sleep duration has been recognized widely as such risk factor [20, 48, 60, 61], prior evidence has overlooked the role that CTS may play in increasing cardiometabolic health risk beyond average sleep duration. Furthermore, studies that have examined the timing and irregularity of the sleep–wake cycle have focused mostly on adults [26, 49], rather than youth, in whom circadian misalignment is highly prevalent [28]. During schooldays, youth are imposed to early high-school start times leading to early rising times despite later sleep onset, while during leisure days, adolescents sleep-in and shift their circadian sleep phase on weekends, an effect that becomes more evident and persistent during breaks from school [28–30, 62]. Indeed, the adolescents studied herein while in-school slept on average less than 7 hours during schooldays with a SM at around 3:00, while they slept on average 50 minutes longer with a SM an hour later (4:00) on the weekends. These latter data were mimicked by those adolescents who were studied while on-break from school, as they showed a SM at 4:00 regardless of weekdays or weekends. Thus, youth are a population of choice for early interventions that can ameliorate the adverse effects brought forth by the impact that an abnormal CTS can have on the known association of obesity with MetS.

Our first novel finding in adolescents studied while attending school showed that a delayed SM and high SI both increased the impact of VAT on MetS (Table 3). Specifically, in adolescents with a sleep phase at 4:00 or later or with a SI of 1 hour or more during schooldays, the impact of VAT was associated with about 2.5–2.7 standard deviations increase in MetS score. Because the MetS score used is based on sex-and-age adjusted z-scores, these standard deviations can be interpreted in a clinically meaningful manner. In contrast, in adolescents with a normative sleep phase or with optimal sleep regularity during schooldays, the impact of VAT was associated with about 1.0 standard deviation increase in MetS score (Table 4). Therefore, both SM and SI are key markers of circadian misalignment under entrainment conditions imposed by attending school. Although studies have not examined how circadian misalignment impacts obesity-related MetS, studies on adults have suggested that delays in SM and greater SI are associated with markers of metabolic dysfunction, such as reduced insulin sensitivity, lower HDL, and higher triglyceride levels even after adjusting for sleep quality and insufficient sleep [50]. This finding has also been shown in youth [51], particularly in overweight adolescents as per BMI percentile criteria [52], and this effect seems to be more prevalent when circadian misalignment occurs during school [32, 53]. Furthermore, studies of phase alignment using robust circadian biomarkers (dim light melatonin onset) in adolescents have found that an endogenous circadian delay, determined by a wider phase angle between bedtime and melatonin onset, was associated with insulin resistance [51]. Our current study builds upon those prior findings and offers a view of the multidimensional cardiometabolic health risk, via a standardized MetS score, brought forth by circadian misalignment in adolescents attending school. These data have the potential to inform targeted prevention by prompting early identification of different forms of circadian misalignment and focusing on aligning school start times to adolescents’ shifting circadian phase [62]. Prevention efforts would additionally promote a regular circadian timing to maintain the sleep–wake cycle anchored to the biological night, as well as other circadian-related behaviors (e.g. nutrition) [63, 64], all of which may act as protective factors against MetS and future development of CVD or diabetes when optimally aligned.

Our second novel finding showed that SJL increased the impact of VAT on MetS costs overall, particularly when adolescents were studied while attending school (Table 3). Specifically, in adolescents evaluated while in-school with a SJL greater than 1 hour and 30 minutes between weekdays (2:58 in Supplementary Table S4) and weekends (4:48 in Supplementary Table S4), the impact of VAT was associated with more than 2.0 standard deviations increase in MetS score. In contrast, in adolescents with less SJL between weekdays (2:39 in Supplementary Table S4) and weekends (2:45 in Supplementary Table S4), the impact of VAT was associated with less than 1.1 standard deviations increase in MetS score (Table 4). These findings are supported by evidence suggesting that this level of SJL is associated with markers of metabolic dysfunction, independent of insufficient sleep [46, 55, 56], even in overweight adolescents [52, 57, 58]. Stratified analyses in the overall sample also revealed that in adolescents with a “low SJL” between weekdays and weekends, VAT was associated with 1.9 standard deviations increase in MetS score (Table 4). Interestingly, our data showed that they already had a delayed weekday SM (Supplementary Table S4); suggesting that “low SJL” once again identified those adolescents with a delayed sleep phase, in whom VAT was strongly associated with MetS impacts, as reported above. These data further confirm that the three circadian phenotypes play a role in cardiometabolic health [33] and are distinct phenomena: those with persistent delayed sleep phase while in-school, those with high SJL, who delay their sleep phase by an hour and a half from weekdays to weekends overall but particularly while in-school, and those with high SI, whose circadian sleep phase shifts over an hour across nights while in-school. Our data shows that the three circadian phenotypes provide key and distinct markers of misalignment of the sleep–wake cycle when evaluating adolescents under forced entrainment to school versus while on free/leisure days.

From a conceptual standpoint, the models resulting from the statistically significant interaction terms found in this study are not equally plausible in the role that circadian misalignment and VAT would play (i.e. independent variable or moderator/effect modifier). Opposed to our conceptual model in Figure 1, treating circadian misalignment as the independent variable and VAT as the effect modifier/moderator would imply that it is among adolescents with high VAT that the association between circadian misalignment and MetS is stronger. Despite being statistically plausible, our sensitivity analyses stratified by VAT did not support it as the effect modifier (Supplementary Table S3). In contrast, the significant interactions found to support the more parsimonious model presented in Figure 1 that posits that an abnormal CTS is an effect modifier and that interventions that focus on those with greater circadian misalignment will help diminish the well-known impact of VAT on MetS. From a mechanistic standpoint, given that a misaligned CTS disrupts the underlying physiological mechanism (e.g. immune pathways) that leads VAT to produce MetS [3, 33], it is plausible that circadian misalignment enhances the impact of VAT on MetS by further dysregulating those mechanistic pathophysiologic factors. For example, youth with circadian misalignment may have an impaired immune system, releasing more pro-inflammatory cytokines, and making them more vulnerable to the effects of VAT on cardiovascular and metabolic function, leading to worsening their cardiovascular and metabolic function and putting them at risk of developing MetS. Thus, future studies using longitudinal designs should also include biomarker data to examine whether the effect modification of circadian misalignment on the association of VAT with MetS occurs via an enhancement of the pro-inflammatory and/or oxidative stress pathways.

It is important to note that there are some potential limitations to the current study. First, we only had 7 nights of ACT data available, preventing us from calculating SI and sleep variability during weekends. Second, these data also precluded using other metrics of SI, as the accuracy of these measures of deviation in habitual sleep patterns varies based on the study design [41]. Third, these ACT-estimated sleep timing variables were used as surrogate measures of circadian misalignment in the current study and, since biomarkers of endogenous melatonin phase (e.g. dim light melatonin onset) were not assessed, true circadian misalignment could not be ascertained. Fourth, since this study was cross-sectional, it limited causal inference regarding the associations found. Fifth, while we were able to study these associations in adolescents who were evaluated either in-school or on-break, we did not have repeated measures allowing us to examine within-participant effects when adolescents shift from being in-school to on-break. Sixth, the five components making up cMetS were equally weighted and treated independently when computing the sum of their individual Z-scores [9]. Although the cMetS derived from this study is sample-specific, it represents the demographic characteristics of adolescents of central Pennsylvania and is therefore comparable to other studies targeting the same population [9]. Finally, future studies should account for social (e.g. income and culture), behavioral (e.g. anxiety/depression, substance use), lifestyle (e.g., social media use, sedentary behavior, and diet quality), or contextual (e.g. parenting, bedtime routines, and household stress) factors as either potential covariates or mediating/mechanistic factors that may also vary by sex, socioeconomic status, and/or race/ethnicity. Future studies should also examine these associations longitudinally as adolescents transition to young adulthood to establish temporal causality and uncover the impact of circadian misalignment across the lifespan. Randomized clinical trials should investigate whether implementing a more regular CTS in adolescents with obesity can improve their cardiometabolic health in order to reinforce causal inference.

Conclusion

In summary, this cohort study found that circadian misalignment of the sleep–wake cycle may increase the impact of visceral adiposity on MetS in adolescents. Our data supports that three phenotypes of circadian misalignment, including delayed sleep phase, sleep irregularity, and social jetlag, may be key determinants of the cardiometabolic sequelae associated with visceral obesity in adolescents beyond insufficient sleep, sleep variability, and sleep apnea. Furthermore, when considering the role of entrainment, our findings suggest that a delayed sleep phase, high sleep irregularity, and greater social jetlag during the school year best identify those adolescents with greater cardiovascular and metabolic risk associated with visceral adiposity, emphasizing the need to examine these circadian misalignment metrics under different real-world entrainment scenarios in adolescents. Consequently, our data suggests that maintaining an aligned and regular CTS may protect these adolescents from adverse cardiometabolic outcomes related to central obesity, informing public health prevention as circadian misalignment may be amenable to contextual (e.g. school start times), behavioral (e.g. sleep–wake scheduling), non-pharmacological (e.g. timed bright light therapy) and pharmacological (e.g., properly formulated timed melatonin) interventions in youth. Future research should investigate whether circadian misalignment in adolescence is associated with other lifestyle factors, as evidence indicates that misaligned sleep is associated with poor diet quality, increased screen time, and lack of physical activity, to name a few [31].

Funding

Research reported in this publication was supported in part by the National Heart, Lung, and Blood Institute, National Institute of Mental Health, and the National Center for Advancing Translational Sciences of the National Institutes of Health under Awards Number R01HL136587 (Fernandez-Mendoza), R01MH118308 (Fernandez-Mendoza), and UL1TR000127 (Penn State University). Ms. Morales-Ghinaglia is funded by the American Heart Association through a predoctoral fellowship (Award Number 23PRE1011962). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Disclosures Statement

Financial disclosure: Dr. Fernandez-Mendoza reported receiving grants from the National Heart, Lung, and Blood Institute (NHLBI) and the National Institute of Mental Health (NIMH) during the conduct of the study; and grants from Pfizer Inc, the American Heart Association, the National Institute on Drug Abuse, and the Patient-Centered Outcomes Research Institute outside the submitted work. Dr. Vgontzas reported receiving grants from the National Foundation for Research and Innovation EU/Greece outside the submitted work. Nonfinancial disclosure: None declared.

Data Availability

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