Cardiorespiratory fitness and circadian rhythms in adolescents: a pilot study

Abstract

Although cardiorespiratory fitness (CRF), an important marker of youth health, is associated with earlier sleep/wake schedule, its relationship with circadian rhythms is unclear. This study examined the associations between CRF and rhythm variables in adolescents. Eighteen healthy adolescents (10 females and 8 males; M_age = 14.6 ± 2.3 yr) completed two study visits on weekdays bracketing an ambulatory assessment during summer vacation. Visit 1 included in-laboratory CRF assessment (peak V̇o₂) using a ramp-type progressive cycle ergometry protocol and gas exchange measurement, which was followed by 7–14 days of actigraphy to assess sleep/wake patterns and 24-h activity rhythms. During Visit 2, chronotype, social jetlag (i.e., the difference in midsleep time between weekdays and weekends), and phase preference were assessed using a questionnaire, and hourly saliva samples were collected to determine the dim light melatonin onset (DLMO) phase. All analyses were adjusted for sex, pubertal status, and physical activity. Greater peak V̇o₂ was associated with earlier sleep/wake times and circadian phase measures, including acrophase, UP time, DOWN time, last activity peak (LAP) time, and chronotype (all P < 0.05). Peak V̇o₂ was negatively associated with social jetlag (P = 0.02). In addition, the mixed-model analysis revealed a significant interaction effect between peak V̇o₂ and actigraphy-estimated hour-by-hour activity patterns (P < 0.001), with the strongest effects observed at around the time of waking (0600–1000). In healthy adolescents, better CRF was associated with an earlier circadian phase and increased activity levels notably during the morning. Future studies are needed to investigate the longitudinal effects of the interactions between CRF and advanced rhythms on health outcomes.

NEW & NOTEWORTHY In healthy adolescents, better cardiorespiratory fitness, as assessed by the gold standard measure [laboratory-based assessment of peak oxygen consumption (V̇o₂)], was associated with earlier circadian timing of sleep/wake patterns, rest-activity rhythms and chronotype, and less social jetlag. These findings highlight the close interrelationships between fitness and rhythms and raise the possibility that maintaining higher cardiorespiratory fitness levels alongside earlier sleep/wake schedule and activity rhythms may be important behavioral intervention targets to promote health in adolescents.

INTRODUCTION

Circadian rhythms are biological, psychological, and behavioral processes that oscillate over the course of a day and are entrained to the 24-h day by environmental cues, or zeitgebers, such as consistent exposure to sunlight and regularly timed physical activity, meal timing, and social interactions (1). Substantial changes in circadian rhythms occur during adolescence. Behavioral manifestations of these age- and puberty-related changes include alterations in sleep/wake patterns. Both survey and actigraphy data suggest a tendency for teenagers to exhibit progressively later bedtimes and waking times as they age (2, 3). Adolescents also show delayed timing in actigraphy-derived 24-h rest-activity patterns, an objective indicator of circadian rhythmicity (4). These observed later timings in sleep/wake patterns and rest-activity rhythms are partly attributable to the natural shift toward a delayed endogenous circadian timing system during puberty, which is evident by a later dim light melatonin onset (DLMO) phase (3). The delayed endogenous circadian phase (i.e., the time where the peak and trough of circadian rhythms occur) often parallels with adolescents’ chronotypic preference for later schedules (3, 5). Chronotype represents individual differences in phase of entrainment and can be computed as the midsleep time between sleep onset and offset on school- or work-free days, accounting for sleep duration (6). Relative to other age groups in the general population, teenagers on average exhibit the most delayed chronotypes (7). Later chronotype and eveningness phase preference have been additionally associated with poorer mental and physical health in adolescents (8, 9). Together, these findings underline the importance of identifying factors that may influence the magnitude of phase delays in developing adolescents to optimize health outcomes.

Sleep/wake timing and sleep duration also become more irregular across teenage years (10). Greater sleep variability during adolescence is linked to poor mental health and altered brain development (11, 12). Circadian rhythmicity and sleep/wake regularity are further impacted by adolescents’ tendency to extend their sleep on weekends, usually through waking up later to compensate for an accumulated sleep debt over school nights. Such discrepant circadian timing of sleep between weekdays and weekends results primarily from a mismatch between the intrinsic biological rhythms and extrinsic behavioral rhythms imposed by social obligations (e.g., waking up early due to school schedules), which is referred to as social jetlag (13). Social jetlag has often been characterized by an average 2 h discrepancy in midsleep time between weekdays and weekends (14, 15). Recent findings suggest that social jetlag is exacerbated by modern technology access, excessive evening light exposure, increased sedentary behavior with screen time, and increased homework demands (16, 17). These psychosocial factors, in turn, may further aggravate delayed rhythms and sleep irregularity that are commonly observed during adolescence.

Disruptions in entraining stimuli and poor behavioral input into the circadian timing system may have detrimental effects on mental and physical health (18, 19). Regular participation in physical activity can facilitate the entrainment of circadian rhythms (19), yet the associated mechanisms are not fully understood. It remains unclear whether the benefits of entrainment are due to consistent engagement in physical activity, per se, or through its possible downstream effects on health, including improved cardiorespiratory fitness (CRF). CRF has been considered an important marker of health in adolescence. Healthy CRF is positively correlated with mental well-being, cardiovascular health, and academic achievement in youth (20). Recent findings from our group showed that CRF is associated with an earlier sleep/wake schedule and a more mature pattern of topographically specific features of sleep electroencephalography known to support neuroplasticity and cognitive processes (21). However, few studies have examined the connections of CRF with rhythm variables in adolescence, with almost all using field tests of fitness or focusing on subjective measures. For instance, a cross-sectional study reported a negative association between CRF, as estimated from a 20-m shuttle run performed on a tennis court or in a gymnasium, and social jetlag in male adolescents (15). Another study of 1,044 healthy adolescents showed that reduced CRF and higher metabolic risk were each independently associated with more fragmented rest-activity rhythms (22). The associations between laboratory-based assessments of CRF and objective measures of rhythm variables in any age group have yet to be studied. Given that fitness and behavioral manifestations of circadian rhythms including sleep/wake patterns and rest-activity rhythms are modifiable, a better understanding of their interrelationships may inform behavioral interventions to promote health during adolescence, a critical stage of life for brain, cognitive, physical, and mental well-being development.

The present study aimed to examine the associations between the gold standard measure of CRF [i.e., laboratory-based assessments of peak oxygen consumption (V̇o₂)] with rhythm variables, which include actigraphy-estimated sleep/wake patterns and rest-activity rhythms, chronotype, phase preference, DLMO phase, social jetlag, and sleep regularity in healthy adolescents. We also aimed to explore the associations between behavioral activity rhythms and biological rhythms. We hypothesized that adolescents with greater CRF would show more advanced circadian phase (i.e., earlier actigraphy-estimated sleep/wake times and rest-activity rhythm phase, chronotype, and DLMO phase), reduced social jetlag, and more regular sleep/wake patterns (i.e., consistent sleep/wake times and greater sleep duration regularity). We also hypothesized a positive association between actigraphy-estimated circadian phase measures and endogenous biological rhythms indicated by the DLMO phase.

MATERIALS AND METHODS

Study Participants

Eighteen adolescents (10 females and 8 males) aged 11–17 yr (mean ± SD = 14.6 ± 2.3 yr), who took part in a larger pilot project investigating the effects of CRF and sleep on cognitive performance participated in this study (Table 1). All participants were in good health and reported no evidence of chronic disease or disability that would impact their participation in study procedures. Exclusion criteria included pregnancy or breastfeeding, use of illegal drugs or alcohol in the last month, and regular use of any medication. Participants did not travel across time zones for at least 2 wk before and during their participation in the study. All evaluations were conducted during summer vacation (July–August) in the United States when participants’ sleep was not restricted by school schedules. Ethical approval for this study was obtained from the Institutional Review Board at the University of California, Irvine (UCI), and written informed consent and assent were obtained from parents/legal guardians and participants.

Table 1.

Demographics and anthropometrics of the sample

	All (n = 18)	Females (n = 10)	Males (n = 8)
	Mean (SD)	Mean (SD)	Mean (SD)
Demographics
Age, yr	14.6 (2.3)	14.6 (2.3)	14.6 (2.5)
Sex
Female, n; %	10 (55.6)
Male, n; %	8 (44.4)
Pubertal status
I Prepubertal, n; %	2 (11.1)	1 (10.0)	1 (12.5)
II Early pubertal, n; %	3 (16.7)	1 (10.0)	2 (25.0)
III Midpubertal, n; %	1 (5.6)	0 (0.0)	1 (12.5)
IV Late pubertal, n; %	9 (50.0)	6 (60.0)	3 (37.5)
V Postpubertal, n; %	3 (16.7)	2 (20.0)	1 (12.5)
Anthropometrics
Height, cm	162.45 (13.84)	156.63 (10.57)	169.73 (14.57)
Weight, kg	53.49 (13.23)	51.21 (10.85)	56.35 (16.03)
BMI-for-age percentile	49.43 (23.46)	58.4 (21.91)	38.21 (21.47)

BMI, body mass index.

Study Design

This study involved two laboratory visits on weekdays bracketing an ambulatory assessment. CRF was assessed during Visit 1 at the UCI Pediatric Exercise and Genomic Research Center (PERC) Human Performance Laboratory. Chronotype, social jetlag, phase preference, and endogenous biological rhythms were assessed during Visit 2 at the UCI sleep clinic. A minimum of 7 days and no more than 14 days of ambulatory accelerometry and sleep diary were collected between two visits for the assessment of sleep/wake patterns, rest-activity rhythms, and sleep regularity (see Fig. 1).

Figure 1.

Schematic of the study procedure.

Visit 1.

A 2-h study visit was scheduled in the late morning or early afternoon on a weekday according to the time that was convenient for the parent/legal guardian and the participant to commit to the study protocol.

Anthropometry and pubertal status.

Anthropometric measurements including height and weight were assessed. Body mass index (BMI)-for-age percentile (i.e., BMI relative to other adolescents of the same age and sex) was calculated using the Centers for Disease Control and Prevention (CDC) growth charts for children and teens (23). The Pubertal Development Scale (PDS) (24) was completed by participants to estimate pubertal status, which was categorized into stage I (prepubertal) through stage V (postpubertal).

Cardiorespiratory fitness.

Evaluation of peak V̇o₂ was conducted following standardized cardiopulmonary exercise testing (CPET) protocol (25) by using a ramp-type progressive cycle ergometry with the SensorMedics metabolic system (Vmax 229, Yorba Linda, CA) and breath-by-breath gas exchange measurements. A respiratory exchange ratio (RER) ≥ 1.1 was used as an objective criterion to determine peak V̇o₂, which was calculated as the highest 20-s rolling average in the last 2 min of exercise. All participants achieved a RER ≥ 1.1 at the end of the CPET. The time of assessment ranged from 0930 to 1500 (mean ± SD = 12:14 ± 1:38); half of the sample (n = 9) was scheduled before noontime.

Physical activity.

A modified version of the self-reported Modifiable Activity Questionnaire (MAQ) (26) was completed to assess the frequency and duration of different levels of physical activity engagement. Participants listed up to seven physical activities they had performed over the past week and reported the number of occasions and the average duration of each activity. We used CDC guidelines (27) to categorize each reported physical activity as either moderate (e.g., walking, bicycling, ping-pong, and yoga) or vigorous (e.g., swimming, martial arts, team sports, and spin classes). The average duration of moderate-to-vigorous physical activity (MVPA) engagement (hours/day) was used as a summary variable to characterize physical activity in the present study.

At-Home ambulatory assessment.

Sleep/wake patterns.

Participants completed a daily sleep diary (28) with items recording information about their sleep/wake schedule and caffeine consumption and wore a watch-like wrist accelerometer, Actiwatch (Actiwatch Spectrum Plus; Philips Respironics, Bend, OR), on their nondominant wrist for 7 to 14 consecutive days and nights (24-h periods), starting from the end of Visit 1. They were instructed to remove the Actiwatch only if they were immersing in saltwater. The accelerometer assesses both the magnitude and the direction of acceleration in 1 min epochs, and the resulting digitized wrist activity counts are used as a proxy measure to estimate sleep/wake patterns (29). The Actiwatch is equipped with colored light sensors and an off-wrist detector; detected off-wrist times are coded as missing data. Sleep diaries and light information were used to cross validate and set the analysis intervals of rest periods according to expert consensus guidelines (29). Participants were included in the analysis if they had at least five consecutive 24-h days of recordings and at least 1 weekend night. Those with > 3 consecutive hours of missing data on more than 2 days were reviewed by the authors (A.B.N. and I.Y.C.) to ensure that these missing data occurred during daytime and did not impact nocturnal sleep outcomes. Actigraphy data from all participants met the abovementioned prerequisites and were all included in the reported analyses. Sleep variables, including sleep onset time, sleep offset/wake time, and total sleep time (TST; minutes of the rest interval scored as sleep), were derived from actigraphy data. Actigraphic midsleep time, an objective measure of chronotype, was computed as the midpoint between sleep onset and offset/wake times. These variables were computed as the average for the 7–14 consecutive nights (weekdays and weekends) before Visit 2 for each participant. The day-to-day regularity of sleep onset and offset/wake times and TST were calculated using the average daily standard deviation for each measure; smaller values indicate higher regularity.

Rest-activity rhythms.

The accelerometry data were also processed to derive estimates of 24-h rest-activity rhythms, using both cosinor modeling (30) and the graphical approach (31). Briefly, minute-by-minute wrist activity counts were transformed using a natural logarithm function, and a cosine curve with a period at or near 24 h was then fitted to the log-transformed activity data using regression analysis (30). Outcome variables derived from the cosinor method included estimates of 1) midline estimated statistic of rhythm (MESOR, which represents the mean activity count of the fitted 24-h period), 2) amplitude (the difference between the model-fitted peak activity and the MESOR), and 3) acrophase (also called phi, which represents time of peak activity).

Although the cosinor model is one of the most commonly used parametric approaches for the evaluation of actigraphy-estimated activity rhythms (31), an indicator of circadian rhythmicity, it is limited by the assumption that the recorded wrist activity data are normally distributed and symmetrical within the 24-h day. We therefore also applied the graphical approach (31) to further investigate activity rhythms. Specifically, we used a generalized additive model (GAM) (32) to fit smoothed nonlinear curves to log-transformed wrist activity counts aggregated minute-by-minute across the entire recording period (7–14 days), resulting in a single average 24-h period for each individual (Fig. 2). This graphical approach generates variables characterizing the slope of activity increase in the morning (UP slope), the time of the maximal rate of morning activity increase (UP time), the slope of evening decrease in activity (DOWN slope), the time of maximal rate of activity decline (DOWN time), and the time of last activity peak (LAP time), providing additional information on the rhythm characteristics that are supplementary to the cosinor model. Computation and psychometric information of the cosinor model and the graphical approach are previously reviewed in detail (31).

Figure 2.

Examples of actigraphy-derived rest-activity rhythms and salivary melatonin profiles for 4 individuals. For each plot, black dots represent aggregated log-transformed actigraphy-derived activity counts. Nonlinear smoothed curve (red line) produced by the generalized additive model (GAM) depicts the modeled activity rhythms over 24 h from aggregated 7–14 days of actigraphy. Last activity peak (LAP) time, DOWN time, and UP time are marked by black vertical dashed lines. Light blue shaded area represents average sleep period derived from actigraphy. Dim light melatonin onset (DLMO) phase (turquoise vertical line) was computed using the 4 pg/mL threshold indicated by the dashed turquoise horizontal line. Each plot depicts data collected from an 11-yr-old boy with early pubertal status (A), an 11-yr-old girl with prepubertal status (B), a 17-yr-old boy with late pubertal status (C), and a 15-yr-old girl with postpubertal status (D). Visual inspections suggest that relative to younger adolescents (A and B), those with more mature pubertal status (C and D) exhibited later sleep/wake patterns and more delayed rest-activity phase (UP time, DOWN time, and LAP time). The mean actigraphy-derived activity counts as measured by the midline estimated statistic of rhythm (MESOR) were similar across these 4 adolescents (MESOR = 1.53, 1.44, 1.57, and 1.56, respectively).

Visit 2.

Participants were scheduled to arrive at UCI sleep clinic on a weekday afternoon, ∼6.5 h before their habitual bedtime, which was determined by the actigraphy data collected during the at-home ambulatory assessment.

Chronotype, social jetlag, and phase preference.

The Children’s Chronotype Questionnaire (CCTQ) (33), a 27-item, mixed-format parent-report questionnaire, was collected. The CCTQ contains questions about a child’s sleep/wake patterns for scheduled (i.e., school/work) and free days of recent weeks, and provides several measures of chronotype, including the midsleep point on free days (MSF; i.e., sleep without social obligations), which was calculated as the midpoint between sleep onset and offset on free days. In the present study, we computed the MSF adjusted for individual average sleep need (MSF_sc = sleep corrected MSF), an indicator of chronobiological phase, and social jetlag (i.e., the difference between the midsleep point on free days and workdays) according to Roenneberg et al. (6). In addition, a multi-item morningness/eveningness (M/E) score (range: 10–49) was derived from the CCTQ, which includes 10 questions about the child’s preferred timing or “feeling best” rhythm for various activities. Morning types were classified by a M/E score ≤ 23, intermediate types by a score of 24–32, and evening types by a score ≥ 33 (33). Finally, participants’ phase preference was determined using a single-item measure, on which parents read a short description of various chronotypic preference categories and selected the one that best represents their child: definitely a morning type, rather a morning type than an evening type, neither/nor type, rather an evening type than a morning type, or definitely an evening type.

Biological circadian rhythms.

DLMO was assessed to determine the endogenous circadian phase using seven whole, unstimulated saliva samples collected via passive drool under dim red-light conditions (< 10 lux). Participants did not eat, drink (except water), use dental floss, or brush their teeth for at least 30 min before each sample collection. Caffeinated beverages (coffee, tea, and soda) were prohibited after noontime on the study day. Samples were collected every hour starting from 5 h before each participant’s average habitual bedtime (computed from the actigraphy data) and continued through 1 h past bedtime. Bedtimes ranged from 2130 to 0030; therefore, the earliest sampling window was from 1630 to 1930 and the latest sampling window was from 2230 to 0130. Samples were assayed for melatonin in duplicate, using commercially available kits with enzyme-linked immunosorbent assays (ELISA; Salimetrics, Carlsbad, CA). The lower and upper limit of sensitivity (i.e., LLOS and ULOS) of the assay is 1.37 pg/mL and 50.00 pg/mL, respectively. The intra- and interassay coefficients of variation were 3.9% and 4.4%, respectively. Using a recommended absolute threshold of 4 pg/mL (34), DLMO was operationalized as the time point when melatonin values exceeded the threshold value and remained above the threshold. Linear interpolation of the times between melatonin immediately below and above the threshold value was used to calculate the DLMO clock time. Of all collected samples (n = 124; 7 samples per participant with 2 participants each having 1 missing sample), 20% of assay values were below the LLOS, and Tobit regression was therefore used to provide an imputed value where this occurred. When the linear interpolation to derive DLMO clock time required using data points that had been imputed, substitution was only done if the correlation between predicted and observed values (i.e., the overall correlation yielded by the Tobit) for that participant was r ≥ 0.61. Data from two participants did not meet such criteria with r < 0.61 and were therefore excluded from the analyses involving the DLMO phase (n = 16 participants).

Analyses

Distribution statistics were calculated for demographics, CRF, physical activity, and rhythm measures for the overall sample. Physical activity data (i.e., average daily duration of MVPA) were normalized using log transformation. Sex, pubertal status, and physical activity were included as covariates in all analyses outlined in the results section because of their potential confounding influence on CRF and rhythm variables. A one-way analysis of covariance (ANCOVA) controlling for sex, pubertal status, and physical activity was performed to compare time differences between habitual midsleep time and time of day when CRF was assessed against phase preference (morning, neither/nor, evening types). Bivariate partial parametric (Pearson) and nonparametric (Spearman’s) correlational analyses adjusting for sex, pubertal status, and physical activity were performed to examine 1) the associations between CRF with rhythm variables and 2) the associations between activity rhythms and biological rhythms. Specifically, nonparametric approach was applied to analyses involving MESOR, UP time, DOWN time, and LAP time. Cohen’s conventions were used to interpret effect sizes of the bivariate relationships (35); the effect size was considered small if the absolute value of correlation coefficient (|r|) varied around 0.10, medium if |r| varied around 0.30, and large if |r| varied more than 0.50. In addition, a mixed-model analysis using longitudinal actigraphic activity data that were aggregated hour-by-hour over a single 24 h period was conducted to examine the time of day effects on the relationship between CRF and activity patterns while adjusting for sex, pubertal status, and physical activity. The mixed model included the main effects of CRF (peak V̇o₂) and time (hour of the day), along with a peak V̇o₂ by time interaction. The utilization of mixed models provided an alternative to handle serial dependence and missing actigraphy data. Analyses were conducted using SPSS version 28.0 (IBM SPSS Statistics, Inc., Chicago, IL) and SAS version 9.4 (SAS Institute, Inc., Cary, NC), and an α level of 0.05 was considered statistically significant.

RESULTS

Sample Characteristics

The majority of the sample had average-to-high levels of CRF as indicated by V̇o₂ norms (36), with a median peak V̇o₂ = 45.28 and a mean peak V̇o₂ = 44.75 ± 8.88 mL/kg/min. The relatively high CRF level was evident in both sexes (peak V̇o_2–Males = 50.96 ± 5.13 and peak V̇o_2–Females = 39.79 ± 8.17 mL/kg/min). Between-group comparisons suggested that males exhibited the expected significantly higher peak V̇o₂ than did females, t(16) =3.36, P = 0.004. The average self-reported duration of MVPA was 1.16 ± 1.27 h/day. The data suggested that 33.3% of the sample (n = 6) met the recommended national guidelines (27) of engaging in ≥60 min moderate physical activity daily and 55.6% (n = 10) achieved ≥3 sessions of vigorous activity per week. After controlling for sex and pubertal status, peak V̇o₂ was not correlated with average daily duration of MVPA (r = 0.467, P = 0.088; see Table 2).

Table 2.

Means, standard deviations, and correlation coefficients for cardiorespiratory fitness with physical activity and circadian rhythm variables of the entire sample (n = 18)

	Mean (SD)	Correlation with Peak V̇o2a
Cardiorespiratory Fitness		r	P
Peak V̇o2, mL/kg/min	44.75 (8.88)
Physical Activity
Duration of MVPA, h/day	1.16 (1.27)	0.467	0.088

	Mean (SD)	Correlation with Peak V̇o2b
Sleep/Wake Patterns		r	p
Sleep onset time, hh:mm	23:37 (1:10)	−0.656**	0.008
Sleep offset time, hh:mm	7:52 (1:18)	−0.603*	0.017
TST, min	424.68 (56.71)	0.062	0.825
Sleep Regularity
SD of sleep onset time, hh:mm	00:55 (0:29)	−0.446	0.096
SD of sleep offset time, hh:mm	00:57 (0:28)	−0.349	0.202
SD of TST, min	65.79 (32.58)	−0.492	0.062
Rest-Activity Rhythms
Cosinor approach
MESOR	1.45 (0.12)	0.098	0.729
Amplitude	1.07 (0.15)	0.006	0.983
Acrophase	15.60 (1.20)	−0.631*	0.012
Graphical approach
UP Time	8.44 (1.76)	−0.603*	0.017
DOWN Time	23.02 (0.99)	−0.548*	0.035
LAP Time	19.77 (1.53)	−0.531*	0.042
UP slope	0.66 (0.15)	−0.043	0.880
DOWN slope	−0.60 (0.13)	−0.147	0.601
Chronotype, Social Jetlag, and Phase Preference
CCTQ
MSFsc, hh:mm	3:45 (0:57)	−0.538*	0.039
Social jetlag, hh:mm	1:25 (0:28)	−0.590*	0.021
M/E score	30.94 (6.10)	0.009	0.978
Actigraphy
Midsleep time	3.75 (1.15)	−0.698**	0.004
Endogenous Biological Rhythms
DLMO phase, hh:mm	20:30 (1:30)	−0.533	0.060

^aPartial correlation adjusted for sex and pubertal status. ^bPartial correlation adjusted for sex, pubertal status, and physical activity. CCTQ, Children’s Chronotype Questionnaire; DLMO, dim light melatonin onset; LAP, last activity peak; M/E score, morningness/eveningness score; MESOR, midline estimated statistic of rhythm; MSFsc, midsleep point on free days adjusted for sleep debt; MVPA, moderate-to-vigorous physical activity; V̇o₂, oxygen consumption; TST, total sleep time. *P < 0.05, **P < 0.01.

Sleep duration from actigraphy suggested insufficient sleep (TST = 424.7 ± 56.7 min; see Table 2), with only three adolescents (16.7%) sleeping on average for more than 8 h per night. The sleep duration variability (i.e., an individual’s standard deviation in TST across 7–14 days) was 65.8 min on average, ranging from 23.0 to 155.4 min. One-third of the sample (n = 6) had >1 h variability in daily TST over the assessment period and one participant had >2 h variability in TST. No significant differences were observed for sleep variables (i.e., sleep onset time, sleep offset time, and TST) between weekdays and weekends (all P’s > 0.05). Only one participant reported daily caffeine consumption (averaging 2 cups of caffeinated beverage per day), and an additional five reported occasional use of caffeine (averaging <1 serving per day).

As shown in Table 2, the mean chronotype phase derived from the CCTQ (i.e., MSF_sc) was at 0345 ± 0057, and the average weekly social jetlag was 1:25 ± 0:28 h. The mean M/E score of the sample was 30.9 ± 6.1 (range: 20 to 40). Eight parents (44.4%; definitely morning type: n = 1, rather morning type: n = 7) classified their children as morning types, four (22.2%) as neither/nor types, and six (33.3%; rather evening type: n = 3, definitely evening type: n = 3) as evening types. The mean DLMO phase of the sample was at 2030 ± 0130 (range: 1813 to 2243). Figure 2 provides examples depicting the characteristics of rest-activity rhythms of four participants including both females and males with early or late pubertal status alongside their individual salivary melatonin levels, DLMO phase, and actigraphy-estimated sleep periods. Figure 3 illustrates the aggregated actigraphy-derived wrist activity data and average characterization of the DLMO phase across the entire sample.

Figure 3.

Aggregated actigraphy-derived rest-activity rhythms and average characterization of salivary melatonin profiles across individuals with valid dim light melatonin onset (DLMO) data (n = 16). Mean of activity were calculated using the aggregated 7–14 days of actigraphy-derived wrist activity counts across all participants, which then were logarithmically transformed (black dots). Nonlinear smoothed curve (red line) produced by the generalized additive model (GAM) depicts the modeled activity rhythms over 24 h. Last activity peak (LAP) time, DOWN time, and UP time are generated from the modeled activity rhythms of the sample and marked by black vertical dashed lines. Light blue shaded area represents average sleep period derived from actigraphy. Turquoise line depicts average salivary melatonin profiles of the sample, and average DLMO phase was indicated by the turquoise vertical line.

CRF Assessment Time, Midsleep Time, and Phase Preference

The average time difference between habitual midsleep time and CRF assessment time was 8.49 ± 2.25 h (range: 4.55 to 12.79 h). We did not observe any statistically significant effect of phase preference (morning, neither/nor, and evening types) on these time differences while controlling for sex, pubertal status, and physical activity, [F(2,12)] = 1.15, P = 0.349. This result suggests that the elapsed time between habitual midsleep time and time of day, when CRF testing occurred, was comparable among participants with early, intermediate, and later circadian systems.

Association between CRF with Sleep/Wake Patterns and Regularity

As reported previously (21), when adjusting for sex, pubertal status, and physical activity, peak V̇o₂ was significantly associated with actigraphy-estimated sleep onset and offset times (see Table 2), suggesting that adolescents with higher CRF had an earlier sleep schedule (falling asleep and waking up earlier). Peak V̇o₂ was not correlated with the regularity measures (i.e., SD) of sleep onset time (r = −0.446, P = 0.096) or sleep offset time (r = −0.349, P = 0.202). However, a large effect size was observed for the negative association between peak V̇o₂ and TST regularity, which trended toward significance in our small sample (r = −0.492, P = 0.062), indicating that those with higher fitness levels may have had a tendency toward more consistent sleep duration. Peak V̇o₂ was not related to any of the summary sleep variables such as sleep onset latency and sleep efficiency (Supplemental Table S1).

Association between CRF and Rest-Activity Rhythms

Mixed-model analyses, adjusting for sex, pubertal status, and physical activity, showed a significant interaction between peak V̇o₂ and actigraphy-estimated hour-by-hour activity patterns, [F(23,368)] = 3.16, P < 0.001. As can be seen in Fig. 4, decomposition of the interaction effects suggested that the influence of CRF on activity (i.e., participants with higher peak V̇o₂ exhibited greater actigraphy-assessed activity) was observed throughout the day (0500–2300), with the largest effects noted at around time of waking (0600–1000). These findings are further supported by statistically significant associations between peak V̇o₂ and circadian phase parameters generated both from the cosinor and graphical approaches. Peak V̇o₂ was significantly associated with acrophase (r = −0.631, P = 0.012), UP time (r = −0.603, P = 0.017), DOWN time (r = −0.548, P = 0.035), and LAP time (r = −0.531, P = 0.042) while adjusting for sex, pubertal status, and physical activity (see Table 2 and Fig. 5). Participants with comparatively higher CRF levels showed a more advanced phase in rest-activity rhythms characterized by earlier time of day when cosine-fitted peak activity occurred (acrophase), earlier times at which morning activity increased and evening activity declined (UP time and DOWN time), as well as earlier time of day when the last activity peak was observed (LAP time). No significant associations were detected between peak V̇o₂ with MESOR, amplitude, UP slope, or DOWN slope (all P > 0.05).

Figure 4.

Mixed-model estimates of time of day effects on the correlation between peak V̇o₂ and actigraphy-estimated activity levels. Model estimated t values represent an index of the correlation between cardiorespiratory fitness and activity level at the time specified. Color coding of the bars represent the magnitude and statistical significance of the effect where p(t) is the P value of the estimate (t) based on a t probability distribution. V̇o₂, oxygen consumption.

Figure 5.

Scatter plots showing the relationships between CRF and actigraphy-estimated circadian phase measures. Greater peak V̇o₂ was significantly associated with earlier acrophase (A), UP time (B), DOWN time (C), and LAP time (D). Blue circles represent males and pink circles represent females. The diagonal line represents a linear fit of the data. CRF, cardiorespiratory fitness; LAP, last activity peak; V̇o₂, oxygen consumption.

Association between CRF with Chronotype, Social Jetlag, Phase Preference, and DLMO Phase

As reported in Table 2, when controlling for sex, pubertal status, and physical activity, we found that peak V̇o₂ was negatively related to chronotype phase variables derived from the CCTQ (i.e., MSF_sc) and actigraphy (i.e., midsleep time), indicating that adolescents with greater CRF had an earlier chronotype. There was a negative association between peak V̇o₂ and social jetlag (r = −0.590, P = 0.021). Importantly, after removing two participants who had morning sport commitments, which may preclude them from sleeping in during weekends, the observed association remained statistically significant (r = −0.744, P = 0.006; see Supplemental Table S2). Together, these results suggested that adolescents who were more aerobically fit showed less discrepant sleep/wake schedules between weekdays and weekends relative to those who were less aerobically fit. No significant association between peak V̇o₂ and circadian phase preference (i.e., M/E scores) was detected. Regarding the endogenous biological rhythms, we observed a large effect size for the correlation between peak V̇o₂ and DLMO phase, albeit the association trended toward statistical significance (r = −0.533, P = 0.060), suggesting that those with better CRF may have a tendency toward advanced DLMO phase.

Association between Actigraphy-Estimated Circadian Phase Measures and DLMO Phase

After adjusting for sex, pubertal status, and physical activity, exploratory analyses showed that acrophase, UP time, and DOWN time were each robustly correlated with DLMO phase (r = 0.787, P = 0.001; r = 0.815, P = 0.001; and r = 0.757, P = 0.003, respectively; see Fig. 6). The association between LAP time and DLMO phase did not reach statistical significance (r = 0.483, P = 0.094).

Figure 6.

Scatter plots showing the relationships between dim light melatonin onset (DLMO) phase and actigraphy-estimated circadian phase measures. Earlier DLMO phase was significantly associated with earlier acrophase (A), UP time (B), and DOWN time (C). Blue circles represent males and pink circles represent females. The diagonal line represents a linear fit of the data.

DISCUSSION

This study examined the associations between the gold standard measure of CRF (i.e., laboratory-based assessment of peak V̇o₂) with rhythm variables, which include actigraphy-estimated sleep/wake patterns and rest-activity rhythms, chronotype, phase preference, endogenous biological rhythms (i.e., DLMO phase), social jetlag, and sleep regularity in healthy adolescents. Consistent with our primary hypotheses, results suggest that better CRF, an important marker of health (20), was associated with more phase-advanced circadian systems across a variety of indicators. We also found that CRF was inversely correlated with social jetlag and trended to be associated with sleep duration regularity. Exploratory analyses additionally showed that actigraphy-estimated circadian phase measures including acrophase, UP time, and DOWN time were each positively associated with the DLMO phase. Together, these findings suggest that better CRF is related to earlier behavioral manifestations of circadian rhythms, which have been shown to have positive impacts on physical health, mental health, and school functioning in adolescents (9, 10, 37).

CRF assessment was conducted in the late morning or early afternoon on weekdays according to the time that was convenient for the parent/legal guardian and the participant. Despite that the assessment time was not standardized across participants, our analyses showed that the time differences between the time of day when testing occurred and participants’ habitual midsleep time, a reliable indicator of individual chronotype (6, 13), did not differ significantly according to their morningness-eveningness phase preferences. This result suggests that the scheduled CRF assessment time in the present study did not seem to favor participants with an earlier circadian system over those with a later one, or vice versa. In addition, no significant association was found between CRF and TST (r = 0.062, P = 0.825), suggesting it may be less likely that participants with delayed circadian systems performed worse than others due to less sleep before CRF assessment. Although it has been documented that anaerobic exercise performances fluctuate with time of day, with early morning nadirs and peak performances in the late afternoon (38), the effects of time of day on aerobic and cardiorespiratory performances remain inconclusive (38, 39), indicating that additional concomitant factors may need to be considered when assessing CRF. Indeed, two recent reviews address the importance of including the potential influence of circadian rhythms on exercise performance with a specific focus on an individual’s chronotype (40, 41). It would be interesting for future studies to examine the effects of personalized assessment time on CRF in relation to other circadian rhythm variables.

Increased CRF was associated with earlier behavioral manifestations of circadian rhythms, i.e., earlier sleep/wake patterns and a more advanced phase in rest-activity rhythms. These observations were not replicated by the endogenous circadian timing system as assessed by DLMO phase, which showed a trending (P = 0.060) negative association with peak V̇o₂, with a large effect size (|r|=0.533). Interestingly, based on our small sample, the relation between CRF and 24-h activity rhythms was more localized to morning behavior, as the most significant interaction effects were observed between 0600 and 1000. It is possible that this association is modulated by circadian phase preference, such that morning-type adolescents prefer to go to bed and wake up early and engage in more activity in the early morning, thereby contributing to better CRF. Indeed, it has been reported that morning-type adolescents are more physically active and have better attitudes toward physical activity (42), whereas those with more evening circadian preferences tend to exhibit more sedentary behavior, lower physical activity, and report poorer subjective fitness levels (43, 44). Although these data suggest potential interrelationships among CRF, rest-activity rhythms, and phase preference, larger longitudinal studies are necessary to determine causality.

We found significant negative correlations between CRF and social jetlag. This result is consistent with a recent study reporting that a 1 h increase in social jetlag was associated with a 0.72 mL/kg/min decrease in peak V̇o₂ in adolescents, independent of any sleep characteristics (15). Notably, the magnitude of social jetlag observed in our sample was 1:25 ± 0:28 h, which is slightly less than the typically reported average (2 h) in similar age groups (14, 15). The smaller magnitude of social jetlag in our sample was likely due in part to the study period occurring during summer vacation, a time when adolescents’ self-selected sleep/wake schedule might be more aligned with their internal circadian clock. A previous study found that adolescents’ weekday sleep in the summer was not significantly different from their weekend sleep during the school year (45). Moreover, Crowley and colleagues (46) reported that the time interval from salivary DLMO phase to bedtime was shorter during summer vacation as compared with school year in healthy children and adolescents, suggesting a reduced discrepancy between sleep timing and endogenous circadian phase during summer vacation months. Disrupted rhythms resulting from a mismatch between sleep schedule and the biological clock likely impede an individual’s ability to achieve sufficient sleep, which may negatively impact CRF. Furthermore, individuals with greater social jetlag have increased cardiometabolic and adiposity risk (47, 48), which potentially leads to reduced CRF. Alternatively, it is possible that worse CRF precedes social jetlag, such that less aerobically fit individuals may tend to delay bedtime and/or wake time on the weekends, resulting in later midsleep time and greater social jetlag. Future research conducted both during the school year and summer vacation is warranted to elucidate the precise directionality of this association.

Chronic sleep restriction is common in adolescents, and as shown by the at-home ambulatory actigraphy assessment, less than 16.7% of our sample achieved the recommended 8 h of sleep. In healthy young adults, acute sleep deprivation and restriction have been shown to reduce muscle protein synthesis (49), which may cause loss of muscle mass and impair recovery from exercise (50), although data from youth are lacking. Sleep loss suppresses the anabolic activity of insulin-like growth factor I (IGF1) (51), which is involved in the regulation of body composition, bone density, muscle mass, and neural plasticity (52, 53). Inadequate sleep has also been linked with altered functioning in the hypothalamic-pituitary-adrenal (HPA) axis (54), the major endocrine system that controls and modulates stress reactions. Interestingly, in male adolescents and older adults, higher CRF was associated with lower cortisol reactivity to stress (55, 56), indicating a potential protective role of CRF in buffering the HPA-axis dysregulation. It would be important for future studies to investigate the mechanisms by which sleep and CRF may interact to modulate stress responses in adolescents. In addition to sleep duration, sleep can be assessed and characterized through a multidimensional approach. Sleep health, an emerging construct that consists of various sleep and circadian functions including sleep duration, timing, regularity, efficiency, satisfaction, and alertness, has been proposed as a more holistic conceptualization of sleep as part of overall health (57). Research has shown that sleep health is an important correlate of mental and physical health outcomes. In adolescents, greater sleep health was associated with reduced odds of mood and anxiety disorders and obesity (58). A recent presidential advisory from the American Heart Association additionally identifies sleep health as one of the vital components of cardiovascular health (59). Research including sleep health as a multidimensional approach to assessing sleep and circadian rhythms may provide additional findings to help advance the current understanding of the various compositions of adolescent health.

Adolescents are susceptible to experiencing disruptions in activity rhythms due to both psychosocial and biological influences including a puberty-related shift toward delayed circadian timing (3). Studies in adolescents show that blunted actigraphy-derived rest-activity rhythms are associated with higher BMI and C-reactive protein (60, 61), a proinflammatory marker that has been known to link to adiposity and lower CRF (62, 63). In the present study, we found inverse associations between CRF with circadian systems across various modes of assessments including actigraphy-based sleep/wake patterns and rhythm estimations, chronotype measures assessed by a questionnaire, and to a lesser extent, endogenous biological rhythm findings from DLMO, which showed a borderline significant association with CRF. Furthermore, social jetlag is highly prevalent in adolescents (13), and its magnitude may be exacerbated by the modern lifestyle, including increased electronic media use and social networking in the evening (16, 17). In studies of adolescents, both delayed rhythms and social jetlag are associated with cardiometabolic risk factors such as greater BMI percentile and adiposity (47, 48, 64, 65), and risk for psychiatric disorders such as depression and anxiety (9, 14, 66). Similarly, reduced CRF is linked to worse cardiometabolic and mental health in youth (20, 67, 68). Despite these convergent negative associations with poor health outcomes, the mechanisms through which reduced CRF, delayed rhythms, and social jetlag interact and potentially synergistically contribute to disease risk have not been fully elucidated and warrant future investigation.

Recent findings from several genome-wide association studies provide evidence supporting potential causal associations between sleep and circadian rhythms with disease risk. A large-scale meta-analysis reports that genetic liability to insomnia is associated with an increased risk of a broad range of cardiovascular diseases (69). Using actigraphy-derived sleep and rhythms measures, Li and Zhao (70) identified shared genetic loci between sleep and circadian system with BMI, physical activity levels, depression, and neurological disorders including Alzheimer’s disease. Chronotype has also been shown to be genetically associated with various psychiatric traits (e.g., depressive symptoms, schizophrenia), such that morning chronotype is positively correlated with better mental health (71). These data suggest that our findings of better CRF and a more advanced circadian system may be related to shared genetic influences on sleep, circadian rhythms, and fitness. The addition of assessing molecular components of circadian rhythms including clock genes and clock gene expression will be valuable to supplement the existing literature on the circadian rhythm-fitness relationships.

Strengths of the present study include the use of a gold standard measure of CRF (i.e., peak V̇o₂ assessed by progressive ergometer exercise in the laboratory) along with objective measures of sleep/wake patterns and rest-activity rhythms (i.e., actigraphy), and biological circadian phase (i.e., DLMO phase). In addition, we used both cosinor and graphical approaches to characterize actigraphy-estimated activity rhythms and tested their associations with CRF; such an approach provides a more comprehensive estimation of different domains of rest-activity rhythm characteristics. However, several limitations should be noted. The small sample size might have left us underpowered to detect statistically significant associations or provided less reliable associations between CRF with rhythms and sleep regularity. However, the reported significant statistical findings were observed even with a small sample size and remained significant when controlling for confounding variables. This suggests large effect sizes between CRF and some rhythm variables that have not been examined previously in the literature. The generalizability of findings may also be limited as our sample was small, self-selected, and comprised of healthy adolescents, most of whom were in the later stages of puberty and were aerobically fit. Prior research supports age- and puberty-related, as well as sex-related differences in underlying sleep and circadian physiology changes during adolescence (3, 72). Our preliminary analyses showed results with evidence of sex and puberty effects on the associations between CRF and circadian rhythms (Supplemental Tables S3 and S4). For instance, it was observed that the correlations between CRF and rhythm variables were stronger in males (−0.74 to −0.95) than in females (−0.18 to −0.67), and in late puberty group (−0.39 to −0.85) than in early puberty group (−0.09 to 0.43). However, due to the limited sample size (e.g., there were only 2 females with early pubertal status), we are unable to further delineate a potential interaction of sex and pubertal status in this relationship. Future studies including a larger sample size with adolescents across different stages of pubertal status and a broader range of CRF levels are needed to examine the effects of puberty and sex and their interactions on the observed associations between CRF with sleep and rhythms. The current study was conducted during summer vacation, thus aiming to characterize adolescents’ self-selected sleep/wake patterns during free-living conditions. Of note, two participants reported having morning sport practice/training commitments, which might have impacted light exposure timing and sleep schedule; however, removing these individuals did not meaningfully change the overall results (see Supplemental Table S2). Since the CCTQ questionnaire was based on parent reports, the reliability of the derived chronotype variables may not be as accurate as self-reported data. Nevertheless, we also computed an objective measure of chronotype (midsleep time) using the actigraphy data collected directly from participants and found consistent results. Classification criteria for phase preference using the CCTQ-derived M/E scores were originally validated in children aged 4–11 (33), which may raise concerns about age- and puberty-related adjustments as our sample included adolescents aged 11–17. A recent study (73) validated the questionnaire in adolescents (age: 9–18 yr) and reported significant correlations between the M/E scores of CCTQ and the adolescent version of Morningness/Eveningnss Scale (MES) developed by Carskadon et al. (74). Finally, the present study used a cross-sectional design and was intended to be descriptive. Future studies are needed to further investigate the longitudinal temporal relationships among CRF, circadian rhythms, and sleep regularity, both in healthy adolescents and those with sleep disturbances.

In conclusion, our data from healthy adolescents showed that better CRF was associated with earlier actigraphy-derived sleep and wake times, more advanced circadian activity phase (i.e., acrophase, UP time, DOWN time, and LAP time) and chronotype, and reduced social jetlag. These findings highlight the close interrelationships between CRF and behavioral manifestations of circadian rhythms and raise the possibility that maintaining higher CRF levels alongside earlier sleep/wake patterns and rest-activity rhythms may be important behavioral intervention targets to promote health in adolescents. Future longitudinal studies are necessary to better understand the mechanisms by which CRF and rhythms interact, either indirectly through behavioral changes or directly through modifications of circadian physiological systems, to influence health outcomes.

DATA AVAILABILITY

The data that support the findings of this study may be available to verified researchers upon reasonable request by contacting the corresponding author.

SUPPLEMENTAL DATA

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.22720711.

Supplemental Table S2: https://doi.org/10.6084/m9.figshare.22720708.

Supplemental Table S3: https://doi.org/10.6084/m9.figshare.24270766.

Supplemental Table S4: https://doi.org/10.6084/m9.figshare.24270775.

GRANTS

This work was supported by the National Center for Advancing Translational Sciences (NCATS) under Grant No. UL1TR001414 and the Pediatric Exercise and Genomics Research Center (PERC) Systems Biology Fund.

DISCLOSURES

Dr. Neikrug consults for Edwards Lifesciences. These data are not related to this consulting activity. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

I.Y.C., S.R.-A., B.A.M., R.M.B., and A.B.N. conceived and designed research; I.Y.C., S.R.-A., K.K.L., A.D., M.G.C.-F., D.G., and A.L. performed experiments; I.Y.C., A.S., J.R.P., A.D., and A.B.N. analyzed data; I.Y.C., S.R.-A., A.S., B.A.M., R.M.B., and A.B.N. interpreted results of experiments; I.Y.C., A.S., J.R.P., and A.B.N. prepared figures; I.Y.C. and A.B.N. drafted manuscript; I.Y.C., S.R.-A., J.R.P., M.G.C.-F., K.G.V., B.A.M., R.M.B., and A.B.N. edited and revised manuscript; I.Y.C., S.R.-A., A.S., J.R.P., K.K.L., A.D., M.G.C.-F., K.G.V., D.G. A.L., B.A.M., R.M.B., and A.B.N. approved final version of manuscript.