Social Jetlag, Eating Behaviours, and Body Mass Index among Adolescents in the United States

Abstract

There is a lack of research on associations of social jetlag with eating behaviours and obesity among adolescents. We examined the associations of social jetlag with eating behaviours and body mass index (BMI) in adolescents before and after adjustment for potential confounders. Self-report data were collected from 3,060 adolescents [(48.1% female, mean (SD) age 15.59 (.77) years] from the Fragile Families and Child Wellbeing Study (FFCWS). In regression models, social jetlag predicted odds of consumption of breakfast, fruits/vegetables, fast food, and sweetened drinks and BMI percentile. Primary models adjusted for school night sleep duration, sex, age, household income, and youth living arrangements; secondary models further adjusted for race/ethnicity. In fully adjusted models, greater social jetlag was associated with lower odds of consumption of breakfast (OR = .92, p = .003) and fruits/vegetables (OR = .92, p = .009), and higher odds of consumption of fast food (OR = 1.18, p < .001) and sweetened drinks (OR = 1.18, p < .001). Social jetlag was positively associated with BMI percentile after additional adjustment for eating behaviours (b = .84, p = .037) but this relationship was attenuated after adjustment for race/ethnicity (b = .72, p = .072). Ethnoracial differences in social jetlag may attenuate the association of social jetlag with BMI and should be considered in future studies of circadian misalignment, eating behaviours, and obesity markers.

INTRODUCTION

Obesity is a critical public health issue that affects 40% of American adults, translating to 93 million individuals.⁽¹⁾ Excess body weight increases the risk for Type 2 diabetes, metabolic syndrome, inflammation, and many cancers⁽²⁾ and costs the U.S. healthcare system an estimated $48–66 billion per year.⁽³⁾ In 2015–2016, over 20% of adolescents aged 12 to 19 years met the criteria for obesity,⁽⁴⁾ or body mass index (BMI) at or above the 95^th percentile (based on 2000 data) matched for age and sex.⁽⁵⁾ Research indicates a link between adolescent and adulthood obesity; adolescents with obesity at age 15 are over five times as likely to have adulthood obesity (BMI ≥ 30) by 35 years of age compared to their normal-weight counterparts.⁽⁶⁾ Thus, identifying risk factors for excess weight before adulthood may be critical for early treatment.

Recent studies have indicated “social jetlag” as a risk factor for obesity in both adults^(7–11) and adolescents.⁽¹²⁾ Social jetlag is the misalignment of sleep timing across the week.⁽¹³⁾ During the work or school week, individuals with later preferred timing of daily activities (e.g., sleep), or later chronotype, have later bedtimes but relatively early wake times due to work, school, or other obligations.⁽¹⁴⁾ Late bedtimes and early wake times during the work or school week result in short sleep duration, and individuals may attempt to compensate by delaying wake time and obtaining more sleep on the weekend.⁽¹⁴⁾ Later chronotypes therefore tend to have a sleep interval with a later midpoint on the weekend compared to the work or school week.⁽¹³⁾ This difference in sleep-wake timing between the work or school week and the weekend is known as “social jetlag.”⁽¹³⁾

Due to the combination of both a shift toward later chronotype during adolescence⁽¹⁵⁾ and early school start times,⁽¹⁶⁾ adolescents experience short sleep during the school week and more social jetlag compared to adults.⁽¹³⁾ In addition to short sleep duration,^(17–19) some studies indicate that social jetlag is associated with higher BMI^(7–11) and larger waist circumference^(7,10) among adults, though other studies indicate a null association.^(20–23) One study that examined the relationship between social jetlag and BMI among adolescents found a positive association,⁽¹²⁾ with another demonstrating an association between social jetlag and increased odds of obesity only among females.⁽²⁴⁾ However, one study found a negative association between social jetlag and BMI among adolescents,⁽²⁵⁾ indicating a need for more research in this population. The high prevalence of social jetlag in adolescence⁽¹³⁾ and the evidence for a link with obesity among both adolescents^(12,24) and adults^(7–10) indicate that millions of adolescents may be at risk for excess body weight accompanying social jetlag.

Unhealthy eating behaviours are additional risk factors for obesity that are associated with social jetlag. For example, previous research in adults indicates an association between greater social jetlag and skipping breakfast^(7,26) and higher consumption of sweets and saturated fat,⁽²⁷⁾ which are risk factors for obesity.^(28–35) However, there is a lack of research on the association between social jetlag and different aspects of unhealthy eating in the adolescent population specifically.

Racial/ethnic differences in social jetlag, eating behaviours, and BMI have been observed. For example, Black⁽³⁶⁾ and Hispanic/Latino⁽³⁷⁾ adolescents exhibit less school night sleep duration compared to White adolescents. Shorter sleep during the school week increases the likelihood of social jetlag,⁽¹³⁾ suggesting racial and ethnic minorities may have more social jetlag compared to their White peers. Blacks and Hispanics/Latinos also exhibit eating behaviours associated with higher BMI. Compared to Whites, Black adolescents skip breakfast more frequently and consume more sweetened drinks⁽³⁸⁾ and fewer fruits and vegetables.⁽³⁹⁾ Hispanics skip breakfast more frequently⁽³⁸⁾ and consume more fast food⁽⁴⁰⁾ and fewer vegetables⁽⁴¹⁾ than Whites. Similarly, more Black and Hispanic/Latino adolescents meet the criteria for excess weight (≥ 85^th percentile) compared to Whites.⁽⁴²⁾ The differences in social jetlag, eating behaviours, and BMI among Whites, Blacks, and Hispanics/Latinos indicate that ethnicity may confound the relationships of social jetlag with eating behaviours and BMI. To our knowledge, two studies have examined the relationship between social jetlag and BMI in adolescents while adjusting for ethnicity, finding a positive relationship remained in all adolescents⁽¹²⁾ or in female adolescents only;⁽²⁴⁾ however, these studies did not examine ethnoracial differences in eating behaviours.

The current study investigated the relationships of social jetlag with eating behaviours and BMI among adolescents. We hypothesized that greater social jetlag would be associated with unhealthier eating behaviours and higher BMI percentile, even after adjustment for potential confounders.

EXPERIMENTAL METHODS

Participants and Procedures

Data for the current analyses come from the Fragile Families and Child Wellbeing Study (FFCWS; www.fragilefamilies.princeton.edu). The original birth cohort consists of 4,898 children born 1998–2000 in 20 U.S. cities.⁽⁴³⁾ To date, data have been collected in six waves; the current study examines cross-sectional survey responses of 3,444 youth at age 15 and their primary caregivers (PCGs). Youth missing demographic, sleep, eating behaviour, or weight/height information were excluded from the current study, yielding an analytical sample of 3,060 youth (88.9% of in-wave total and 62.5% of initial birth cohort). Families were compensated $100 USD for completion of the PCG questionnaire and $50 USD for completion of the youth questionnaire, administered either through phone or in person. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human participants were approved by the Stony Brook University Institutional Review Board (CORIHS B) (FWA #00000125). Written (for in-home interviews) or verbal (for phone interviews) informed consent was obtained from all participants. Verbal consent was witnessed and formally recorded.

Materials

Data were drawn from two sources at the age 15 wave: a questionnaire administered to youth, and a separate questionnaire administered to their PCGs (each administered once).

Youth Age 15 Questionnaire

Teens were asked to report sleep variables, race/ethnicity, height, and weight.

Sleep and Social Jetlag Measures

Youth were asked to report bedtimes and wake times during the school week (Sunday-Thursday nights and Monday-Friday mornings, respectively) and bedtimes and wake times during the weekend (Friday-Saturday nights and Saturday-Sunday mornings, respectively).

The following variables were calculated using bedtime and wake time measures collected from the Youth Age 15 Questionnaire. Sleep duration on school nights was calculated as the interval between reported bedtime and wake time on school nights in hours. Sleep duration on weekend nights was calculated as the interval between reported bedtime and wake time on weekend nights in hours. Social jetlag (in hours) was calculated through the following formula: | sleep midpoint on weekend nights − sleep midpoint on school nights |.⁽¹³⁾

Eating behaviours

Breakfast consumption was assessed with the question, “How many days in a typical school week do you eat breakfast? Do not count the weekend” and ranged from 0 to 5. Literature indicates that eating breakfast 71% of the time during the week is associated with reduced risk of overweight in adolescents.⁽⁴⁴⁾ The current study surveyed breakfast consumption during the 5-day school week rather than all 7 days; we therefore dichotomized breakfast consumption to model the odds of consuming breakfast ≥ 4 days (71% of 5 is 3.55, which rounds to 4 days or at least 80% of the school week) during the school week versus < 4 days.

Vegetable and fruit consumption was assessed with the question, “In a typical week, how many days do you eat at least some green vegetables or fruit?” and ranged from 0 to 7. Based on the recommendations for fruit and vegetable consumption from the United States Department of Agriculture,⁽⁴⁵⁾ fruit and vegetable consumption was dichotomized to model the odds of consumption all 7 days of the week (versus < 7 days).

Fast food consumption was assessed with the question, “How many days in a typical week do you eat food from a fast food restaurant, such as McDonald’s, Burger King, Wendy’s, Arby’s, Pizza Hut, Taco Bell, or Kentucky Fried Chicken or a local fast food restaurant?” and ranged from 0 to 7. As most participants had consumed fast food at least one day of the week (85%), fast food consumption was dichotomized to model the odds of consumption ≥ 2 days during the week (versus < 2 days).

Sweetened drink consumption was assessed with the question, “In a typical day, how many regular, non-diet sweetened drinks do you have? Include regular soda, juice drinks, sweetened tea or coffee, energy drinks, flavored water, or other sweetened drinks.” Based on literature indicating that ≥ 2 sweetened drinks per day was associated with weight gain,⁽⁴⁶⁾ sweetened drink consumption was dichotomized to model the odds of consuming ≥ 2 sweetened drinks daily (versus < 2).

BMI Percentile

BMI at age 15 was assessed using self-reported height and weight and calculated through the following formula: weight in kg / (height in m)².⁽⁴⁷⁾ BMI percentile (range 0–100) was calculated based on the 2000 CDC growth charts,⁽⁴⁸⁾ which matches BMI for the adolescent’s sex and age.

Covariate Measures from Youth Questionnaire

Race and ethnicity consisted of four exclusive categories: “White/Caucasian” (non-Hispanic or Latino; reference group), “Black/African American” (non-Hispanic or Latino), “Hispanic and/or Latino” (any race), or “Other, Mixed, or none” (non-Hispanic or Latino).

Sex and age (calculated in exact years) were assessed from administrative records collected at the youth’s time of birth.

Primary Caregiver (PCG) Questionnaire

Covariate Measures from PCG Questionnaire

Covariates for statistical analyses (youth living arrangements and household income) were additionally drawn from a questionnaire administered once to PCGs.

Statistical Analyses

Analyses were conducted in SAS 9.4 (SAS Institute Inc., Cary, NC). Cases were excluded from all analyses if missing any of the following: bedtime or wake time on either school or weekend nights (n = 63), any eating behaviour (n = 57), BMI percentile (n = 224), and/or covariate questions (n = 38) out of 3,444. Two adolescents were excluded due to implausible midpoints of sleep (≥ 12:00 PM) on the weekend (no school nights met this same criterion), yielding 384 excluded adolescents from the age 15 wave and a total analytic sample of N = 3,060. The outcome of BMI percentile met pre-defined criteria for normality (skewness < |3| and kurtosis < |10|).

Pairwise associations between dichotomized eating behaviours were assessed with chi-square analyses (PROC FREQ). We conducted a bivariate Pearson’s correlation analysis (PROC CORR) to test the association between school night school duration and social jetlag. Univariate one-way ANOVAs (PROC ANOVA) or binomial logistic regression analyses (PROC LOGISTIC) were conducted to determine differences in school night and weekend sleep duration, social jetlag, BMI percentile (ANOVAs), and eating behaviours (logistic regression) among White, Black, and Hispanic/Latino adolescents (Other, Mixed, or none were excluded from these analyses due to heterogeneous nature of sample). Significant ANOVA results were followed by post-hoc pairwise comparisons corrected using Tukey’s Honestly Significant Difference (HSD). Separately, social jetlag predicted each outcome of interest (one of four eating behaviours or BMI percentile) in regression analyses. Binomial logistic regression analyses were conducted to model the odds of each eating behaviour outcome, and linear regression analysis was conducted with the BMI percentile outcome (PROC REG). For each outcome, Model 1 included only social jetlag as the predictor. Model 2 added demographic and household covariates selected a priori (age in years, sex, youth living arrangements, and household income) except for ethnicity. Model 3 added school night sleep duration. For BMI percentile only, Model 3e added the four dichotomized eating behaviours. Model 4 added ethnicity (White as reference group; other categories dummy-coded) to the analysis. We examined whether social jetlag interacted with ethnicity on eating behaviours and BMI in Model 4; if the interaction between social jetlag and ethnicity was not significant, the interaction term was dropped from Model 4. The variance inflation factor (VIF) did not exceed 2.5 for any predictor in the regression model, indicating no multicollinearity (i.e., no predictor was strongly correlated with another predictor).⁽⁴⁹⁾ An alpha level < .05 (two-tailed) was accepted as significant.

RESULTS

See Table 1 for demographic information and descriptive statistics for the analytical sample (N = 3,060). The sample was 46.2% Black/African American, 24.1% Hispanic/Latino, 17.5% White/Caucasian, and 12.3% Other, Mixed, or no ethnicity. Mean social jetlag was 2.75 (1.29) hours. Mean BMI was 24.02 (5.73) kg/m² and BMI percentile was 68.0% (27.9%). Consumption of breakfast ≥ 4 days per school week was 52.2%; consumption of green fruits/vegetables 7 days of the week was 30.4%; consumption of fast food ≥ 2 days per week was 48.6%, and consumption of ≥ 2 sweetened drinks daily was 61.9%. Pairwise associations between eating behaviours (according to chi-square analyses) are presented in Table S1 showing each eating behaviour was significantly associated with the other, but with small or small-to-medium effect sizes⁽⁵⁰⁾ (φ ranging from .05 to .22). Social jetlag and school night sleep duration were negatively correlated with a small effect size, r = −.10, p < .001.

Table 1.

Descriptive Statistics from Analytical Sample (N = 3,060) of Age 15 Wave of the Fragile Families and Child Wellbeing Study (FFCWS)

	n or Mean	% or SD
Youth
Sexa
Female	1473	48.1%
Male	1587	51.9%
Race/ethnicity
Black/African American, non-Hispanic/Latino	1413	46.2%
Hispanic and/or Latino, any race	736	24.1%
White/Caucasian, non-Hispanic/Latino	535	17.5%
Other,b mixed, or none	376	12.3%
Age (yrs)	15.59	.77
Household arrangements (youth lives with)
Biological mother	2691	87.9%
Biological father	224	7.3%
Non-parent	145	4.7%
Household income	$62,550	$65,315
Eating behaviours and body mass index (BMI)
Breakfast (≥ 4 days of school week)	1598	52.2%
Green vegetables or fruit (all 7 days of week)	930	30.4%
Fast food (≥ 2 days weekly)	1487	48.6%
Sweetened drinks (≥ 2 daily)	1895	61.9%
BMI	24.02	5.73
BMI percentilec	68.02	27.89
Sleep and social jetlag measures
Bedtime
School night (Sun-Thurs)	22:22	1.13
Weekend night (Fri-Sat)	0:23	1.67
Wake time
School morning (Mon-Fri)	6:21	.86
Weekend morning (Sat-Sun)	9:46	1.79
Sleep duration (hrs)
School night	7.98	1.24
Weekend night	9.38	1.94
Social jetlag (hrs)d	2.75	1.29

^aData collected at birth.

^bOther category includes Asian, Central American/Caribbean, Native American/Alaska Native, and/or Native Hawaiian/Pacific Islander.

^cCalculated based on 2000 Centers for Disease Control and Prevention (CDC) growth charts, matched for age and sex.⁽⁴⁸⁾

^d| Midpoint of sleep on weekend nights − midpoint of sleep on school nights |.⁽¹³⁾

Between-Ethnicity Comparisons of Social Jetlag, Sleep Duration, Eating Behaviours, and BMI Percentile

See Table 2 for analyses of differences in social jetlag, sleep duration, eating behaviours, and BMI percentile among race/ethnicity groups. Omnibus ANOVA analyses indicated significant differences among White, Black, and Hispanic/Latino adolescents in social jetlag (p < .001), school night sleep duration (p < .001), and BMI percentile (p < .001). Pairwise comparisons indicated Whites and Hispanic/Latinos reported significantly less social jetlag (p < .001) and longer school night sleep duration (p < .001) compared to Blacks. Both Hispanic/Latino (p < .001) and Black (p < .001) adolescents had higher BMI percentile compared to Whites. There were no differences in weekend night sleep duration by race/ethnicity (p = .057).

Table 2.

Differences Among Race/Ethnicity Groups in Sleep Duration, Social Jetlag, Eating Behaviours, and BMI Percentile (N = 2,686)

	White (W), non-Hispanic/ Latino (n = 536)		Black (B), non-Hispanic/ Latino (n = 1,414)		Hispanic (H) and/or Latino (n = 736)		Analyses
ANOVA	M	SD	M	SD	M	SD	F (2, 2683)		r2	Pairwise comparisonsa
School night sleep durationb	8.04	1.14	7.86	1.31	8.16	1.16	15.49***		.011	H, W > B
Weekend night sleep durationc	9.54	1.68	9.32	2.08	9.45	1.82	2.90		.002	---
Social jetlag (hrs)d	2.56	1.27	2.93	1.32	2.60	1.25	24.35***		.018	H, W < B
BMI percentilee	61.37	29.66	7.22	26.99	7.23	27.00	22.17***		.016	H, B > W
Logistic regressionf	n	%	n	%	n	%	ORBW 95% CI	ORHW 95% CI	ORHB 95% CI	Pairwise comparisonsg
Breakfast ≥ 4 days of school week	303	56.5	694	49.1	400	54.4	.75** .61, .91	.92 .73, 1.15	1.23* 1.03, 1.48	H, W > B
Fruits/vegetables 7 days of week	223	41.6	414	29.3	176	23.9	.58*** .48, .72	.44*** .35, .56	.76** .62, .93	W > B > H
Fast food ≥ 2 days of week	187	34.9	769	54.4	365	49.6	2.23*** 1.82, 2.75	1.85*** 1.47, 2.32	.83* .69, .99	B > H > W
≥ 2 sweetened drinks daily	251	46.8	982	69.5	441	59.9	2.57*** 2.10, 3.15	1.69*** 1.35, 2.12	.66*** .55, .79	B > H > W

OR—odds ratio.

BW—Black vs. White; HW—Hispanic vs. White, HB—Hispanic vs. Black.

^aCorrected for multiple comparisons with Tukey’s Honestly Significant Difference (HSD).

^bSunday through Thursday night.

^cFriday and Saturday night.

^dCalculated as | midpoint of sleep on weekend nights − midpoint of sleep on school nights |.⁽¹³⁾

^eCalculated based on 2000 Centers for Disease Control and Prevention (CDC) growth charts, matched for age and sex.⁽⁴⁸⁾

^fWhite=reference group.

^gHispanic/Latino vs. Black comparison obtained by changing reference group to Black.

^*p < .05.

^**p < .01.

^***p < .001.

Analyses of eating behaviours indicated 25% lower odds of consuming breakfast ≥ 4 days per school week in Blacks compared to Whites and 23% lower odds compared to Hispanics/Latinos with no differences between Whites and Hispanics/Latinos. Blacks reported 42% lower odds of consuming fruits or vegetables 7 days per week than did Whites and 24% higher odds than did Hispanics/Latinos, and Hispanics/Latinos had 56% lower odds compared to Whites. Blacks reported 123% higher odds of consuming fast food ≥ 2 days per week compared to Whites and 17% higher odds compared to Hispanics/Latinos, and Hispanics/Latinos reported 85% higher odds compared to Whites. Blacks reported 157% higher odds of consuming ≥ 2 sweetened drinks daily compared to Whites and 34% higher odds compared to Hispanics/Latinos, and Hispanics/Latinos reported 69% higher odds compared to Whites.

Associations of Social Jetlag with Eating Behaviours

Models (1–4) were conducted separately for each eating behaviour (see Figure 1 and Table 3). Social jetlag did not interact with ethnicity on any eating behaviour; therefore, no interaction term was included in these analyses, but ethnicity was included as a covariate.

Figure 1.

Adjusted for school night sleep duration and covariates including ethnicity (White non-Hispanic/Latino, Black non-Hispanic/Latino, Hispanic/Latino, or Other, Mixed, or None) in Model 4, binomial logistic regression analyses indicated that greater social jetlag was significantly associated with lower odds of consumption of breakfast (panel a; OR = .92, p = .003) and vegetables/fruits (panel b; OR = .92, p = .009), and higher odds of consumption of fast food (panel c; OR = 1.18, p < .001) and sweetened drinks (panel d; OR = 1.18, p < .001). Grey bars are “healthy” and hatched bars are “unhealthy.” Error bars indicate standard error of the mean. **p < .01; ***p < .001.

Table 3.

Logistic Regression Models Predicting Eating Behaviours from Social Jetlag (N = 3,060)

Predictor	Breakfasta		Vegetables or Fruitb		Fast Foodc		Sweetened Drinksd
	OR	95%CI	OR	95%CI	OR	95%CI	OR	95%CI
Model 1 (unadjusted)
Social jetlag (hrs)e	.90***	.85, .95	.91**	.86, .97	1.19***	1.12, 1.25	1.23***	1.16, 1.30
Model 2 (+ covariatesf)
Social jetlag (hrs)e	.89***	.84, .94	.91**	.86, .97	1.20***	1.13, 1.27	1.21***	1.14, 1.29
Model 3 (+ school night sleep duration)
Social jetlag (hrs)e	.91**	.86, .96	.92**	.86, .97	1.20***	1.13, 1.27	1.20***	1.13, 1.28
School night sleep duration (hrs)	1.25***	1.18, 1.33	1.06	1.00, 1.13	.98	.92, 1.04	.94*	.88, 1.00
Model 4 (+ ethnicityg)
Social jetlag (hrs)e	.92**	.86, .97	.92**	.86, .98	1.18***	1.11, 1.25	1.18***	1.11, 1.25
School night sleep duration (hrs)	1.25***	1.17, 1.32	1.07*	1.00, 1.14	.99	.93, 1.05	.95	.89, 1.01
Black, non-Hispanic/Latino	.83	.67, 1.02	.64***	.52, .79	2.05***	1.65, 2.54	2.33***	1.88, 2.88
Hispanic and/or Latino	.93	.74, 1.17	.46***	.36, .58	1.78***	1.41, 2.25	1.67***	1.33, 2.11
Other,h mixed, or none	.92	.70, 1.21	.68**	.51, .90	1.45**	1.10, 1.90	1.55**	1.11, 1.25

CI, confidence interval; OR, odds ratio.

^aOdds of consuming breakfast ≥ 4 days (versus < 4 days) during the school week.

^bOdds of consuming fruits or green vegetables 7 days (versus < 7 days) weekly.

^cOdds of consuming fast food ≥ 2 days (versus < 2 days) weekly.

^dOdds of consuming ≥ 2 sweetened drinks (versus < 2) daily.

^eCalculated as | midpoint of sleep on weekend nights − midpoint of sleep on school nights |.⁽¹³⁾

^fCovariates include sex, age, youth living arrangements, and household income level.

^gCategories include White, non-Hispanic/Latino (reference group), Black, non-Hispanic/Latino, Hispanic and/or Latino (any race), or Other, Mixed, or none.

^hOther category includes Asian, Central American/Caribbean, Native American/Alaska Native, and/or Native Hawaiian/Pacific Islander.

^*p < .05.

^**p < .01.

^***p < .001.

Breakfast Consumption per School Week

Model 1 (unadjusted) indicated a significant negative association between social jetlag and odds of breakfast consumption ≥ 4 days per school week. This association remained significant after further adjustment for demographic and household characteristics (age in years, sex, youth living arrangements, and household income) in Model 2, school night sleep duration in Model 3, and ethnicity in Model 4 (OR = .92, p = .003; 8% lower odds with each additional hour of social jetlag).

Vegetable/Fruit Consumption per Week

Model 1 (unadjusted) indicated a significant negative association between social jetlag and odds of vegetable/fruit consumption 7 days per week. This association remained significant after further adjustment for demographic and household characteristics (excluding ethnicity) in Model 2, school night sleep duration in Model 3, and ethnicity in Model 4 (OR = .92, p = .009; 8% lower odds with each additional hour of social jetlag).

Fast Food Consumption per Week

Model 1 (unadjusted) indicated a significant positive association between social jetlag and odds of fast food consumption ≥ 2 days per week. This association remained significant after further adjustment for demographic and household characteristics (excluding ethnicity) in Model 2, school night sleep duration in Model 3, and ethnicity in Model 4 (OR = 1.18, p < .001; 18% higher odds with each additional hour of social jetlag).

Sweetened Drinks Consumed Daily

Model 1 (unadjusted) indicated a significant positive association between social jetlag and odds of ≥ 2 sweetened drinks daily. This association remained significant after further adjustment for demographic and household characteristics (excluding ethnicity) in Model 2, school night sleep duration in Model 3, and ethnicity in Model 4 (OR = 1.18, p < .001; 18% higher odds with each additional hour of social jetlag).

Association of Social Jetlag with BMI Percentile

Model 1 (unadjusted) indicated a significant positive association between social jetlag and BMI percentile, which remained after further adjustment for demographic and household characteristics (excluding ethnicity) in Model 2, school night sleep duration in Model 3, and eating behaviours in Model 3e (b = .84, p = .037; .84 higher BMI percentile with each additional hour of social jetlag). Social jetlag did not interact with ethnicity in the analysis of BMI percentile; therefore, no interaction term was included in these analyses, but ethnicity was included as a covariate. After adjustment for ethnicity in Model 4, there was no significant association between social jetlag and BMI percentile (b = .72, p = .072). Figure 2 depicts associations of social jetlag with BMI percentile, unadjusted for ethnicity in Model 3e and adjusted for ethnicity in Model 4 (see also Table 4).

Figure 2.

Adjusted for school night sleep duration, demographic and household covariates, eating behaviours, but not ethnicity in Model 3e, linear regression analyses indicated that greater social jetlag (black line) was significantly associated with higher body mass index (BMI) percentile, b = .84, p = .037. After adjustment for ethnicity in Model 4, the association between social jetlag and BMI percentile was attenuated (grey line), b = .72, p = .072. *p < .05; n.s.—not significant.

Table 4.

Linear Regression Model Predicting BMI Percentile^a from Social Jetlag (N = 3,060)

Predictor	b	95%CI	Model R2	ΔR2
Model 1 (unadjusted)
Social jetlag (hrs)b	1.01*	.24, 1.77	< .01	.002
Model 2 (+ covariatesc)
Social jetlag (hrs)b	.98*	.21, 1.75	.01	.002
Model 3 (+ school night sleep duration)
Social jetlag (hrs)b	.93*	.15, 1.70	.01	.002
School night sleep duration (hrs)	−.48	−1.28, .32		< .001
Model 3e (+ eating behaviours)
Social jetlag (hrs)a	.84*	.05, 1.62		.001
School night sleep duration (hrs)	−.25	−1.06, .55		< .001
Breakfast ≥ 4 days of school week	−3.88***	−5.90, −1.85	.02	.005
Fruits/vegetables 7 days of week	−.87	−3.02, 1.29		< .001
Fast food ≥ 2 days of week	−1.20	−3.24, .84		< .001
≥ 2 sweetened drinks daily	.94	−1.15, 3.04		< .001
Model 4 (+ ethnicityd)
Social jetlag (hrs)a	.72	−.06, 1.50		.001
School night sleep duration (hrs)	−.24	−1.05, .56		< .001
Breakfast ≥ 4 days of school week	−3.85***	−5.87, −1.84		.005
Fruits/vegetables 7 days of week	−.27	−2.42, 1.89	.03	< .001
Fast food ≥ 2 days of week	−1.76	−3.79, .28		.001
≥ 2 sweetened drinks daily	.29	−1.81, 2.39		< .001
Black, non-Hispanic/Latino	7.73***	4.85, 1.61		.019
Hispanic and/or Latino	8.44***	5.30, 11.58		.017
Other,e mixed, or none	2.93	−.74, 6.61		.001

b, unstandardized beta; CI, confidence interval.

^aCalculated based on 2000 Centers for Disease Control and Prevention (CDC) growth charts, matched for age and sex.⁽⁴⁸⁾

^bCalculated as | midpoint of sleep on weekend nights − midpoint of sleep on school nights |.⁽¹³⁾

^cCovariates include sex, age, youth living arrangements, and household income level.

^dCategories include White, non-Hispanic/Latino (reference group), Black, non-Hispanic/Latino, Hispanic and/or Latino (any race), or Other, Mixed, or None.

^eOther category includes Asian, Central American/Caribbean, Native American/Alaska Native, and/or Native Hawaiian/Pacific Islander.

^*p < .05.

^***p < .001.

DISCUSSION

We investigated the associations of social jetlag with eating behaviours and BMI percentile in a large, ethnically diverse sample of adolescents. Our findings indicate that adolescents with high social jetlag exhibit patterns of eating that are associated with negative health consequences such as obesity. With each additional hour of social jetlag, adolescents were 8% less likely to engage in healthy eating behaviours, and 18% more likely to engage in unhealthy eating behaviours. Without accounting for ethnicity, each additional hour of social jetlag was significantly associated with a nearly 1-unit increase in BMI percentile, which became attenuated when accounting for ethnicity. Future studies that examine the cross-sectional relationship between types of circadian misalignment (such as social jetlag) and obesity markers should therefore consider examining ethnoracial differences.

Even after adjustment for ethnicity, social jetlag was associated with unhealthier eating behaviours: 8% lower odds of consumption of breakfast and vegetables/fruits, and 18% higher odds of consumption of fast food and sweetened drinks, with each additional hour of social jetlag. Experimental studies have indicated that induced circadian misalignment results in metabolic alterations that are conducive to greater consumption of energy-dense foods, such as sweetened beverages and fast food. For example, forced desynchrony, wherein individuals follow a day/night cycle shorter or longer than the intrinsic circadian period of ~24 hours, leads to lower levels of the satiety hormone leptin.⁽⁵¹⁾ Circadian misalignment induced by a 12-hour phase shift increases post-prandial levels of the hunger hormone ghrelin and increases appetite for energy-dense foods.⁽⁵²⁾ These studies are supported by the current observational study in which adolescents with social jetlag were more likely to consume greater fast food and sweetened drinks, and by previous research indicating an association between social jetlag and higher consumption of sweets and saturated fat.⁽²⁷⁾

Unadjusted for ethnicity, we found that social jetlag was associated with higher BMI percentile, similar to previous research^(7–12) (about a one-unit increase in BMI percentile with each additional hour of social jetlag). Experimental studies of circadian misalignment have demonstrated metabolic dysregulation that may explain the link between social jetlag and obesity. For example, 28-hour forced desynchrony results in insulin resistance,⁽⁵¹⁾ simulated night-shift work reduces energy expenditure,⁽⁵³⁾ and a 12-hour phase shift lowers insulin sensitivity.⁽⁵⁴⁾ Indeed, research has demonstrated that higher social jetlag is associated with obesity and related indices, such as higher triglycerides and fasting insulin.⁽⁵⁵⁾ These studies, along with the present research, indicate that social jetlag increases the risk for excess body weight and related adverse health outcomes. After adjustment for ethnicity, the relationship between social jetlag and BMI percentile was attenuated and no longer statistically significant. To date, most observational studies of the relationship between social jetlag and obesity markers have either been conducted in relatively ethnically homogenous populations (e.g., mostly White/Caucasian or East Asian) and/or have not mentioned adjustment for ethnicity.^{(7,8,10,20–22,25,56)} We found significant differences in social jetlag, eating behaviours, and BMI among ethnic groups that may account for the attenuation of the relationship between social jetlag and BMI percentile observed in the current study. Future studies that examine the associations of social jetlag with eating behaviours and BMI should study individuals of varying ethnoracial identities in consideration of the findings in the current study. In contrast to social jetlag, we did not observe an association between short school night sleep duration and BMI percentile in the current study (Table 4), though research demonstrates short sleep is associated with higher BMI.^(17–19) Short sleep during the school week was associated with higher social jetlag in the current study, but the present analyses support the distinction between short sleep and social jetlag as separate concepts that denote different aspects of sleep deficiency.

Our findings indicate that individuals with social jetlag are at greater risk for unhealthy eating behaviours. Youth with later bedtimes during the school week who must wake early for school or other obligations will likely have higher social jetlag,⁽¹³⁾ which is associated with lower odds of consumption of breakfast and fruits/vegetables and higher odds of consumption of fast food and sweetened drinks while adjusting for race/ethnicity. We also found a significant relationship between social jetlag and BMI that was attenuated after adjustment for race/ethnicity. One strategy to reduce social jetlag is through lengthening sleep on school nights to reduce sleep debt across the week and prevent compensatory delayed wake time on the weekends.

Limitations of the current study include a cross-sectional study design, restricting our ability to determine temporal order of social jetlag, eating behaviours, and BMI. It is plausible, for example, that high BMI induces changes to social jetlag, as a high-fat diet can alter metabolites in the suprachiasmatic nucleus (the “master clock”), leading to changes in sleep-wake patterns and social jetlag.⁽⁵⁷⁾ Future longitudinal research should investigate whether social jetlag precedes, succeeds, or has a reciprocal relationship with eating behaviours and BMI in adolescents. Additionally, as the study was observational, we are unable to determine whether it is ethnicity or an unmeasured related variable that attenuates the association between social jetlag and BMI percentile. The current study used self-reported sleep data rather than objective measures; however, research has shown a strong correlation between subjective and objective measures (i.e., actigraphy) of sleep duration.⁽⁵⁸⁾ Furthermore, though BMI correlates highly with body composition,⁽⁵⁹⁾ research indicates this correlation is stronger for White/Caucasian versus Black individuals.⁽⁶⁰⁾ Additionally, the eating behaviour questions did not capture quantities of food and beverage consumption. Strengths of the current study are its large sample size, racial and ethnic diversity, and the inclusion of demographic and household covariates in regression models (e.g., age, race/ethnicity, sex, household structure) that may have been confounding variables.⁽⁶¹⁾

In conclusion, we found that social jetlag was associated with unhealthy eating behaviours in adolescents while adjusting for socio-demographic characteristics. Furthermore, social jetlag was associated with higher BMI in unadjusted analyses but not after adjustment for race/ethnicity. Our findings indicate ethnoracial differences in social jetlag may attenuate the relationship between social jetlag and BMI observed in some previous studies.^(7–10) Future experimental and longitudinal research should further probe the associations between social jetlag, eating behaviours, and BMI in adolescents of diverse ethnoracial backgrounds.