Inflammation mediates the association between visceral adiposity and obstructive sleep apnea in adolescents

Abstract

Only a handful of studies, primarily in clinical samples, have reported an association between obesity, inflammation, and obstructive sleep apnea (OSA) in children and adolescents. No studies, however, have examined the pathogenetic link between visceral adiposity, systemic inflammation, and incident OSA in a large general population sample using objective measures of sleep and body fat. Adolescents (n = 392; mean age 17.0 ± 2.2 yr, 54.0% male) from the Penn State Child Cohort (PSCC) underwent 9-h overnight polysomnography; a DXA scan to assess body fat distribution; and a single fasting blood draw for the assessment of plasma interleukin-6 (IL-6), IL-6 soluble receptor (IL-6 sR), tumor necrosis factor alpha (TNFα), tumor necrosis factor receptor 1A (TNFR1), C-reactive protein (CRP), leptin, and adiponectin levels via ELISA. Visceral fat area was significantly elevated in moderate OSA (AHI ≥ 5), especially in boys. IL-6, CRP, and leptin were highest in adolescents with moderate OSA, even after adjusting for BMI percentile. Mediation analysis revealed that 42% of the association between visceral fat and OSA in adolescents was mediated by IL-6 (p = 0.03), while 82% of the association was mediated by CRP (p = 0.01). These data are consistent with the model of a feed-forward, vicious cycle, in which the release of proinflammatory cytokines by visceral adipocytes largely explains the association between central obesity and OSA; in turn, inflammation is also elevated in OSA independent of BMI. These findings, in a large, representative, non-clinical sample of young people, add to our understanding of the developmental pathogenesis of sleep apnea.

obstructive sleep apnea (OSA) is increasingly recognized in children and adolescents. It is estimated that 1–4% of the general pediatric population has obstructive sleep apnea (3), depending on the definition used (41). A significantly greater number of young people (27–40%) have sleep-disordered breathing (SDB) (4, 15), which includes a spectrum of breathing abnormalities such as snoring, upper airway resistance syndrome, and OSA. Although SDB in young people is commonly thought to result from decreased muscular tone or upper airway structural abnormalities, central obesity has been increasingly recognized as a risk factor for OSA in children and adolescents (3, 7, 41, 59, 62) as in adults (56, 65). Currently, 17% of U.S. children under the age of 20 yr are considered obese (≥95th percentile of the Center for Disease Control’s sex-specific body mass index (BMI)-for-age growth chart) (45), up from 10% in 2007 (46).

In adults, levels of proinflammatory cytokines, such as IL-6 and TNFα, acute-phase proteins, including CRP, and adipocyte hormones, like leptin, are positively correlated with body mass index (35, 47, 65). Visceral fat, in particular, is a metabolically active organ (16), with resident macrophages secreting high levels of cytokines (18). This immune dysregulation has been directly linked to impaired glucose uptake, endothelial dysfunction, and development of other cardiometabolic morbidities (29, 51, 55). Interestingly, and as seen in adults, elevated levels of CRP and IL-6 have also been reported in children (20, 60, 61) and adolescents (40) with SDB, even after adjusting for BMI. Many studies examining the association between obesity, inflammation, and OSA, however, are small, conducted in clinical populations, or do not explore the role of inflammation in the pathogenesis of the disorder.

The aim of this study was to examine the relationship between body fat composition, incident SDB, and systemic inflammation in a general population sample of adolescents, as well as the potential underlying mechanism between these associations. Given well-known sex differences in OSA prevalence (5, 6) and body composition (19, 36) in adults, we were also particularly interested in exploring these potential differences during adolescence. The present study is unique in that we report comprehensive measures of obesity and systemic inflammation in the largest general population sample of adolescents to date. We hypothesized that incident moderate OSA would be associated with 1) visceral adiposity, particularly in boys, and 2) elevated plasma proinflammatory cytokine levels, independent of BMI; and that 3) inflammation is the major pathogenic link in the relationship between central obesity and development of OSA in adolescents.

MATERIALS AND METHODS

Ethical approval.

This research has been reviewed and approved for compliance with the policy of the human subjects Institutional Review Board at Penn State University College of Medicine (IRB protocol no. HY98–228-A). Written informed consents were obtained from participants 18 yr and older. Assent was sought for those younger than 18 yr, and written informed consent was obtained from their parents or legal guardians.

Participants.

The Penn State Child Cohort (PSCC) (3, 4) is a general population sample of 700 children between ages 5 and 12 yr, of whom 421 were followed up 8.4 yr later as adolescents (mean age 17.0 ± 2.2 yr, 53.9% male, 21.9% ethnic minority).

The cohort was originally designed as a two-phase study. During Phase I, all children (n = 7,312) from 18 public elementary schools within three Dauphin County, Pennsylvania, school districts were sent home with a questionnaire completed by a parent on their child’s sleep and behavior; 5,740 surveys were returned (79% response rate). During Phase II, based on stratification for grade, sex, and SDB risk based on the questionnaire, 1,000 children were randomly selected to spend a night at the Penn State College of Medicine Clinical Research Center; 700 children participated (70% response rate). The loss before follow-up was mainly due to participants moving out of the central Pennsylvania area; no major differences in the baseline demographic characteristics, however, were observed between those who did and did not participate in the follow-up study (2).

Physical assessment.

During their follow-up visit in the laboratory, all participants underwent a physical examination, during which height (stadiometer model 242, SECA; Hanover, MD), weight (model 758C, Cardinal Manufacturing; Webb City, MO), and waist/hip circumference via tape measure were recorded according to Centers for Disease Control Criteria (10). Body mass index (BMI) was calculated (in kg/m²) and converted to a percentile according to a formula based on the Centers for Disease Control’s sex-specific BMI-for-age growth charts (9). Pubertal development (Tanner staging) was determined via a self-administered rating scale (8). Participants also underwent a dual-energy X-ray absorptiometry (DXA) scan using a Hologic Discovery W scanner (Hologic, Waltham, MA; 195 × 65 cm field of view) to obtain a precise measure of body fat. Regions of interest included android/gynoid, subcutaneous, visceral, and total adipose tissue area. These regions were identified by Hologic APEX 4.0 software (Hologic, Bedford, MA) and visually verified by an experienced investigator; detailed descriptions of these measures can be found elsewhere (25, 33). Participants identified their race/ethnicity from one of six options; “ethnic minority status” was redefined as “non-white/Caucasian” for statistical purposes.

Sleep laboratory protocol.

At both baseline and follow-up time points, all participants underwent a 9-h, single-night polysomnography (PSG) recording in a sound-attenuated, light- and temperature-controlled room with a comfortable, bedroom-like atmosphere. Each subject was continuously monitored from 2200 until 0700 using 14-channel recordings of electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG). Respiration was monitored via nasal pressure (Pro-Tech PTAF Lite, Mukilteo, WA), thermocouple (Salter Laboratories, Lake Forest, IL), and thoracic/abdominal strain gauges (model 1312, Sleepmate Technologies, Midlothian, VA). Hemoglobin oxygen saturation (SpO₂) was assessed using a pulse oximeter placed on the index finger (model 3011 Xpod, Nonin Medical, Plymouth, MN). Snoring sounds were monitored via a sensor attached to the throat. All data were recorded using Twin Recording & Analysis software (Grass-Telefactor; West Warwick, RI). Visual sleep stage scoring was conducted by a registered polysomnography technologist according to standardized criteria (52). Apnea/hypopnea index (AHI; number of apneas and hypopneas summed per hour) was ascertained. An apnea was defined as a cessation of airflow with a minimum duration of 5 s (for those aged < 16 yr) or 10 s (for those ≥ 16 yr) and an associated out-of-phase strain gauge movement; a hypopnea was characterized by a reduction of airflow by ~50% with an associated decrease in SpO₂ of at least 3% or an associated EEG arousal (26).

Blood draw and assay procedures.

Of 421 participants at follow-up, 392 (93.1%) provided a blood sample upon awakening at 0700. Samples were collected in an EDTA-containing tube, then spun for 10 min at 3,000 RPM. Plasma was aliquoted into cryotubes and stored at −80°C until assayed. Plasma interleukin-6 (IL-6), interleukin-6 soluble receptor (IL-6 sR), tumor necrosis factor alpha (TNFα), tumor necrosis factor receptor superfamily member 1A (TNFR1), high-sensitivity C-reactive protein (CRP), leptin, and adiponectin were measured via enzyme-linked immunosorbent assay (ELISA; R&D Systems, Minneapolis, MN). Due to availability of blood specimens, the number of participants for whom we assayed each cytokine varied. The intra- and interassay coefficients of variation were 4.7% and 5.1%, respectively (IL-6), 5.4% and 6.2% (IL-6 sR), 4.6% and 4.9% (TNFα), 5.1% and 6.4% (TNFR1), 5.8% and 5.3% (CRP), 6.5% and 7.0% (leptin), and 5.6% and 5.6% (adiponectin). The lower detection limits were 0.039 pg/ml (IL-6), 6.5 pg/ml (IL-6 sR), 0.106 pg/ml (TNFα), 0.77 pg/ml (TNFR1), 0.010 ng/ml (CRP), 7.2 pg/ml (leptin), and 0.25 ng/ml (adiponectin). All samples and standards were run in duplicate.

Statistical analysis.

Participants were categorized into four groups based on previous criteria (2, 53): 1) no sleep-disordered breathing (SDB; AHI < 2 and no snoring), 2) primary snoring (AHI < 2 and the presence of snoring at any time during the sleep period), 3) mild OSA (2 ≤ AHI < 5), and 4) moderate OSA (AHI ≥ 5). Those who had AHI ≥ 5 at baseline (n = 6) were excluded from all analyses to examine incident cases of OSA. Differences in sociodemographic and physical characteristics in the four groups were assessed using analysis of variance (ANOVA). A generalized linear model was then used to assess differences in z-transformed visceral fat area and subcutaneous fat area separately between the four SDB groups, controlling for the confounders age, sex, ethnic minority status, and total fat area, with Bonferroni correction to correct for multiple comparisons. To assess sex differences, visceral fat area was also examined separately in males and females within each of the four SDB groups, adjusting for age, ethnic minority status, and total fat area, with Bonferroni correction.

Because all cytokines measured in the blood (IL-6, IL-6 sR, TNFα, TNFR1, CRP, leptin, and adiponectin) were nonnormally distributed, logarithmic transformation was applied. To examine the complex relationship between obesity, systemic inflammation, and OSA, a number of statistical analyses were performed. First, differences in plasma cytokine concentrations between the four groups were examined via generalized linear model, adjusting for age, sex, ethnic minority status, and BMI percentile, with Bonferroni correction.

To examine the relative contributions of SDB, visceral fat, and subcutaneous fat to inflammatory and metabolic markers, we then performed linear regression analyses. Given the lack of significant differences observed in adiposity and biomarker concentrations in the no SDB, primary snoring, and mild OSA groups, we redefined OSA as a binary variable with two levels: AHI < 5 and AHI ≥ 5. Linear regression analyses were ultimately performed with moderate OSA, visceral fat, and subcutaneous fat as predictors for each of the biomarker outcomes separately, adjusting for age, sex, and ethnic minority status.

Logistic regression was then conducted to assess the role of inflammation (IL-6 and CRP) as potential mediators of the relationship between visceral fat area and moderate OSA, adjusting for age, sex, and ethnic minority status. Specifically, visceral fat area was assessed as a predictor of AHI ≥ 5, then IL-6 and CRP were added in separate models to observe the attenuation of the β for visceral fat. Based on the results of these logistic regressions, the R package “Mediation” (version 4.4.4) was then used to quantitatively estimate the mediating effects of inflammation in this association.

The statistical confidence level selected for all analyses was P < 0.05. Except for mediation, all analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 22.0 (IBM, Armonk, NY).

RESULTS

Demographic and physical characteristics of the sample.

Of the 392 participants examined (mean age 17.0 ± 2.2 yr, 54.0% male, 22.3% ethnic minority), 25.77% of the sample were primary snorers, 27.81% had mild OSA (2 ≤ AHI < 5), 11.22% had moderate OSA (AHI ≥ 5), and 35.20% had no SDB (AHI < 2 and no snoring) (Table 1). Compared with those without SDB or with primary snoring, adolescents with at least mild OSA tended to be older and have a higher BMI percentile, waist circumference, android/gynoid fat area ratio, more visceral fat area, and were more likely to be male and ethnic minority (nonwhite; all P < 0.023).

Table 1.

Sociodemographic, body fat composition, and cytokine values, stratified by SDB group

	No SDB	Primary Snoring	Mild OSA	Moderate OSA	P
n, %	138 (35.20)	101 (25.77)	109 (27.81)	44 (11.22)
AHI	0.79 (0.40)	1.02 (0.47)	3.14 (0.45)d	12.08 (0.70)	<0.001
Male, %	42.75	49.50	66.97	70.45	<0.001
Ethnic minority, %	16.67	17.82	31.19	25.00	0.022
Age, yr	16.37 (0.18)	16.77 (0.22)	17.44 (0.21)	18.41 (0.33)	<0.001
Tanner stage	4.11 (0.07)	4.19 (0.08)	4.21 (0.08)	4.44 (0.12)	0.159
BMI percentile	61.33 (2.32)	63.40 (2.71)	70.91 (2.61)	78.99 (4.11)	<0.001
Waist circumference, cm	76.36 (1.08)	79.18 (1.26)	83.39 (1.22)	90.78 (1.91)	<0.001
Hip circumference, cm	91.32 (1.16)	88.89 (1.36)	89.94 (1.31)	90.91 (2.06)	0.337
Subcutaneous fat area, cm2	197.31 (12.89)	210.85 (15.60)	228.30 (15.37)	261.77 (24.24)	0.121
Android/gynoid fat area ratio	0.33 (0.01)	0.34 (0.01)	0.39 (0.01)	0.42 (0.02)	<0.001
Visceral fat area, cm2	50.94 (3.32)	55.76 (3.92)	67.16 (3.81)	87.25 (5.94)	<0.001
IL-6, pg/ml	1.04 (0.08)	1.16 (0.09)	1.09 (0.09)	1.92 (0.15)	<0.001
IL-6 sR, pg/ml	44.51 (1.18)	43.98 (1.40)	44.12 (1.31)	41.79 (2.03)	0.689
TNFα, pg/ml	1.81 (0.11)	1.85 (0.13)	1.84 (0.12)	2.48 (0.19)	0.026
TNFR1, ng/ml	12.29 (0.28)	12.68 (0.33)	12.62 (0.32)	13.48 (0.49)	0.201
CRP, mg/l	0.70 (0.08)	0.63 (0.10)	0.96 (0.10)	2.10 (0.15)	<0.001
Leptin, ng/ml	12.25 (1.09)	12.22 (1.27)	12.07 (1.22)	16.18 (1.90)	0.249
Adiponectin, µg/ml	8.65 (0.42)	7.89 (0.50)	7.40 (0.48)	6.02 (0.76)	0.013

Data are presented as means (SE). Those with AHI ≥ 5 at baseline (n = 6) were excluded from analysis. No SDB = AHI < 2 and no snoring; primary snoring = AHI < 2 and the presence of snoring at any time during the sleep period; mild OSA = 2 ≤ AHI < 5; moderate OSA = AHI ≥ 5. BMI percentile, body mass index percentile; IL-6, interleukin-6; IL-6 sR, interleukin-6 soluble receptor; TNFα, tumor necrosis factor alpha; TNFR1, tumor necrosis factor superfamily member 1A; CRP, high-sensitivity C-reactive protein; SDB. sleep-disordered breathing; OSA, obstructive sleep apnea.

Body fat distribution and sleep-disordered breathing.

Overall, there were significant differences in visceral fat area (F_3,379 = 4.11, P = 0.007) and subcutaneous fat area (F_3,379 = 4.11, P = 0.007) between the four SDB groups. Visceral fat was significantly highest in the moderate OSA group (0.15 ± 0.05 cm², P < 0.05 vs. all groups), while subcutaneous fat was significantly lowest in the moderate OSA group (−0.04, ± 0.01 cm², P < 0.05 vs. all groups), adjusting for age, sex, ethnic minority status, and total fat area (Fig. 1A). Given known sex differences in body fat composition in adults, particularly in regards to body fat distribution in those with OSA, we then examined sex differences in visceral fat and SDB in adolescents. Overall, boys tended to have higher visceral fat than girls (62.60 ± 2.61 vs. 56.26 ± 2.83 cm², P = 0.10). In boys, visceral fat area was highest in the moderate OSA group (69.06 ± 2.24 cm²) compared with those with no SDB (P = 0.069), primary snoring (P = 0.02), and mild OSA (P = 0.006; F_3,202 = 4.03, P = 0.008). In girls, although visceral fat was also highest in the moderate OSA group (62.05 ± 3.46 cm²), this group did not differ significantly from the other three SDB groups (all P > 0.394; Fig. 1B).

Fig. 1.

A: mean z-transformed visceral (light gray bars) and subcutaneous (dark gray bars) fat area by SDB groups, adjusting for age, sex, ethnic minority status, and total body fat area. Error bars represent SE. Those with AHI ≥ 5 at baseline (n = 6) were excluded from analysis. *P < 0.05; **P < 0.01. B: mean visceral fat area in boys (light gray bars) and girls (dark gray bars) by SDB groups, adjusting for age, ethnic minority status, and total fat area. Error bars represent SE. Those with AHI ≥ 5 at baseline (n = 6) were excluded from analysis. ^TP < 0.1; *P < 0.05; **P < 0.01.

Sleep-disordered breathing and predictors of inflammation in adolescents.

After adjusting for age, sex, ethnic minority status, and BMI percentile, there were significant differences between the four SDB groups in terms of IL-6 (F_3,360 = 4.01, P = 0.008), CRP (F_3,361 = 13.37, P < 0.001), and leptin (F_3,365 = 5.84, P = 0.001). Specifically, levels of IL-6, CRP, and leptin were elevated in adolescents with moderate OSA compared with those with no SDB (all P < 0.05), primary snoring (all P < 0.001), and mild OSA (all P < 0.05) (Table 2). There were no differences between the four groups in terms of IL-6 sR (all P > 0.32), TNFα (all P > 0.14), TNFR1 (all P > 0.23), and adiponectin levels (all P > 0.30). Further adjusting for health comorbidities that could potentially influence inflammation levels (arthritis, asthma, chronic sinusitis/rhinitis, total number of reported health problems, and use of anti-inflammatory medication) did not significantly alter these associations.

Table 2.

Plasma cytokine concentrations in a general population sample of adolescents, stratified by SDB group

	No SDB	Primary Snoring	Mild OSA	Moderate OSA
log IL-6, pg/ml (n = 371)	−0.08 (0.03)	−0.12 (0.04)	−0.11 (0.04)	0.12 (0.06)a,d,f
log IL-6 sR, pg/ml (n = 379)	3.74 (0.03)	3.74 (0.03)	3.75 (0.03)	3.69 (0.05)
log TNFα, pg/ml (n = 362)	0.19 (0.02)	0.20 (0.03)	0.18 (0.03)	0.29 (0.04)
log TNFR1, ng/ml (n = 383)	2.49 (0.02)	2.51 (0.03)	2.50 (0.03)	2.55 (0.04)
log CRP, mg/l (n = 372)	−0.24 (0.03)	−0.34 (0.04)	−0.23 (0.04)	0.11 (0.06)c,e,g
log Leptin, ng/ml (n = 379)	1.90 (0.07)	1.93 (0.08)	1.86 (0.08)	2.43 (0.13)b,d,g
log Adiponectin, µg/ml (n = 381)	1.86 (0.06)	1.86 (0.07)	1.89 (0.07)	1.81 (0.11)

Data presented as log-transformed means (SE). Adjusted for age, sex, ethnic minority status, and BMI percentile. Those with AHI ≥ 5 at baseline (n = 6) were excluded from analysis. No SDB = AHI < 2 and no snoring; primary snoring = AHI < 2 and the presence of snoring at any time during the sleep period; mild OSA = 2 ≤ AHI < 5; moderate OSA = AHI ≥ 5. IL-6, interleukin-6; IL-6 sR, interleukin-6 soluble receptor; TNFα, tumor necrosis factor alpha; TNFR1, tumor necrosis factor superfamily member 1A; CRP, high-sensitivity C-reactive protein; SDB. sleep-disordered breathing; OSA, obstructive sleep apnea.

^aP < 0.05 vs. no SDB.

^bP < 0.01 vs. no SDB.

^cP < 0.001 vs. no SDB.

^dP < 0.01 vs. primary snoring.

^eP < 0.001 vs. primary snoring.

^fP < 0.01 vs. 2 ≤ AHI < 5.

^gP < 0.001 vs. 2 ≤ AHI < 5.

We then examined the relative contributions of SDB and body fat composition on each inflammatory marker (Table 3). Given the lack of differences observed in adiposity and biomarker concentrations in the no SDB, primary snoring, and mild OSA groups (Fig. 1, Table 2), SDB was redefined as a binary variable with two levels (AHI < 5 and AHI ≥ 5). Neither the interaction terms between binary OSA with visceral fat nor subcutaneous fat significantly predicted any of the markers examined. In general, moderate OSA significantly predicted levels of IL-6 (β = 0.17, P = 0.002), TNFα (β = 0.14, P = 0.015), CRP (β = 0.29, P < 0.001), and leptin (β = 0.16, P < 0.001) in adolescents. Visceral fat predicted IL-6 (β = 0.33, P = 0.017) and CRP (β = 0.41, P = 0.004) levels, and was associated with higher TNFR1 (β = 0.27, P = 0.052) and lower adiponectin (β = 0.24, P = 0.071). Although visceral fat area was associated with lower leptin levels (β = −0.17, P = 0.052), this was compensated for by subcutaneous fat as a strong predictor of elevated leptin (β = 0.78, P < 0.001) (Table 3).

Table 3.

Predictors of plasma cytokine concentrations in a general population sample of adolescents

	log IL-6	log IL-6 sR	log TNFα	log TNFR1	log CRP	log Leptin	log Adiponectin
Moderate OSA	0.17b	−0.05	0.14a	0.06	0.29c	0.16c	−0.03
Visceral fat	0.33a	0.08	0.14	0.27a	0.41b	−0.1T	−0.24T
Subcutaneous fat	−0.06	0.04	−0.07	−0.11	−0.01	0.78c	−0.06
R2	0.14	0.04	0.03	0.08	0.22	0.67	0.16

Linear regression adjusted for age, sex, and ethnic minority status. OSA defined as AHI ≥ 5. Standardized β reported. Those with AHI ≥ 5 at baseline (n = 6) were excluded from analysis. Moderate OSA = AHI ≥ 5; IL-6, interleukin-6; IL-6 sR, interleukin-6 soluble receptor; TNFα, tumor necrosis factor alpha; TNFR1, tumor necrosis factor superfamily member 1A; CRP, high-sensitivity C-reactive protein; OSA, obstructive sleep apnea.

^TP < 0.1;

^aP < 0.05;

^bP < 0.01;

^cP < 0.001.

Mediating role of inflammation in the association of visceral fat and sleep-disordered breathing.

Finally, given the associations observed in Table 3, we then decided to test the potential mediating role of inflammation in the association between visceral adiposity and moderate OSA (Fig. 2). Adjusting for age, sex, and ethnic minority status, one standard deviation increase in visceral fat area was associated with a 65% increased odds of moderate OSA in this general population of adolescents (Table 4, model 1). When IL-6 was introduced in the model (model 2), the β for visceral fat predicting OSA was attenuated by 46.0% (β = 0.50 decreased to β = 0.27). To a greater degree, replacing IL-6 with CRP (model 3) attenuated the regression coefficient by 64.0% (β = 0.50 decreased to β = 0.18). Together, these findings suggest that a large portion of the association between visceral adiposity and OSA can be explained by inflammation, especially IL-6 and CRP. Further adjusting for AHI at the baseline (childhood) sleep laboratory visit had no effect on these associations. Mediation analysis was then conducted to assess the proportion of the total effect of visceral fat on moderate OSA that is attributed to systemic inflammation. According to our model, 42% of the association between central obesity and OSA in adolescents is mediated by IL-6 (P = 0.03) (Fig. 2A), while 82% of the association is mediated by CRP (P = 0.01) (Fig. 2B).

Fig. 2.

Model for the association between visceral adiposity and moderate OSA, as mediated by IL-6 (A) and CRP (B). *P < 0.05.

Table 4.

Logistic regression examining the relationship between visceral fat area and moderate OSA (AHI ≥ 5), with CRP and IL-6 as potential mediators

	Model 1			Model 2			Model 3
	β (SE)	OR (95% CI)	P	β (SE)	OR (95% CI)	P	β (SE)	OR (95% CI)	P
Visceral fat	0.50 (0.15)	1.65 (1.23–2.22)	0.001	0.27 (0.16)	1.31 (0.95–1.80)	0.10	0.18 (0.18)	1.19 (0.95–1.80)	0.33

Model 1: adjusted for age, sex, and ethnic minority status. Model 2: model 1 + log IL-6. Model 3: model 1 + log CRP. OR, odds ratio; CI, confidence interval.

DISCUSSION

This is the first large study in a general population sample of adolescents to examine associations between body fat composition, incident OSA, and systemic inflammation. Our findings suggest that, compared with other types of adipose tissue, visceral fat is most strongly associated with moderate OSA, especially in boys, and appears to precede rather than follow the development of OSA. Furthermore, levels of the inflammatory cytokines IL-6, and CRP, as well as the hormone leptin are elevated in adolescents with moderate OSA, even when adjusting for BMI percentile. Finally, our results suggest that inflammation is the major mediating link in the relationship between visceral adiposity and OSA. These findings, in a large, representative, nonclinical sample of young people, add to our developing understanding of the pathogenetic link between obesity, systemic inflammation, and development of OSA.

Research over the last two decades has begun to hone in on the metabolic impact of fat tissue localization on endocrine and immune function. While all adipocytes produce hormones, cytokines, and other proteins involved in energy homeostasis and cardiovascular function, visceral adipose tissue, in particular, is associated with more severe effects on health compared with subcutaneous (beneath the skin layer) adipose tissue. Packed between organs of the abdominal cavity, increased visceral adiposity is associated with elevated risk of developing insulin resistance and Type II diabetes, dyslipidemia, hypertension, cerebro- and cardiovascular disease, and early death (11, 12, 28, 34, 38, 44). In children and adolescents, there is ample evidence to suggest that poor diet, lack of physical activity (58), genetic factors (32, 54), and male sex (22) are significant predictors of elevated visceral fat area. In the present study, we report a strong association of visceral fat with OSA in boys, but not girls (Fig. 1B). Interestingly, this finding is in accordance with our previous work in middle-aged men and postmenopausal women demonstrating that OSA is strongly associated with visceral fat in males, but global adiposity (total fat area) in females (37). A recent follow-up of the Cleveland Children’s Sleep and Health Study reported that obesity in childhood predicts OSA in adolescence (57), and a follow-up report in the present cohort found that childhood waist circumference and greater ΔBMI percentile from baseline to follow-up predicted OSA in adolescence (2). Similarly, a recent small pilot study in adolescents (n = 20) reported an association of AHI with visceral fat area, but not with BMI or subcutaneous fat area (24). This study, however, only included obese adolescents, and potential sex differences in body fat composition or SDB prevalence were not explored. Previous studies, primarily in adults, have proposed opposing views of the association of visceral adiposity, a component of the metabolic syndrome, with sleep apnea (i.e., visceral adiposity leads to OSA vs. OSA causes visceral adiposity) (13, 63). Our findings suggest that visceral adiposity specifically, present even as young as adolescence with newly-diagnosed OSA, appears to be a causative factor rather than a result of the disorder.

Significant elevations in systemic inflammation are observed in adolescents at the level of moderate OSA (AHI ≥ 5), which give validity to suggested clinical cutoffs for treatment of the disorder (1) as well as the presence of cardiometabolic sequelae previously demonstrated (66) in adolescents. Similar to the present study, reports in children and adolescents have described elevated CRP and IL-6 with increasing AHI, arousal index, and/or SpO₂ nadir (20, 40, 60, 61), even after controlling for BMI, suggesting an inflammatory response to intermittent hypoxia. We also observed an association of TNFα and leptin in adolescents with moderate OSA, which is in line with studies in both children and adults (7, 23, 27, 62, 65).

When we explored the role of inflammation in the mechanistic relationship between obesity and OSA, we found that 42% of the association between visceral fat and OSA in adolescents is mediated by IL-6, and 82% of the association is mediated by CRP. Thus our findings point to a model of a feed-forward, vicious cycle (63), in which the release of proinflammatory cytokines by visceral adipocytes plays a causative role in the development of OSA. It is important to note that, of course, IL-6 and CRP are not independent of one another. IL-6, secreted by adipocyte macrophages (among other body systems), stimulates the synthesis of the acute-phase reactant CRP in the liver (67). Cytokines resulting from other chronic inflammatory conditions, tissue injury, or infection also act to trigger CRP synthesis and secretion (49). Interestingly, since we are examining incident cases of SDB (only those with AHI < 5 at baseline) and adjusting for baseline AHI, we know that the inflammation observed is not a result of a prolonged childhood exposure to SDB. In our study, the mediating effect of CRP was stronger than that of IL-6. IL-6, TNFα (17, 64), leptin (39), and adiponectin (21) exhibit a circadian pattern of secretion over a 24-h period. CRP, on the other hand, which was most strongly associated with both visceral fat area and OSA, is not subject to diurnal variation, which may explain why it is the strongest mediating factor. This may also explain why CRP has been proposed as a good marker for cardiovascular risk (43).

Our mediational model is also supported by a recent study by Kheirandish-Gozal and colleagues demonstrating that anti-inflammatory therapy decreased sleep apnea severity in over 80% of children with mild OSA (1 < AHI < 5) (36). Furthermore, the Child Adenotonsillectomy Trial (CHAT) reported no significant change in cardiometabolic parameters, including CRP, in the 7-mo follow-up period after adenotonsillectomy in children with 2 ≤ AHI < 30 (50), suggesting that inflammation in OSA is systemic and not a result of a local inflammatory process. Our findings also add to evidence that dietary and exercise interventions resulting in weight loss, particularly in the abdominal region, are likely to improve OSA. Indeed, a longitudinal study in adults, the Wisconsin Sleep Cohort, reported that while a 10% weight gain predicts a 32% increase in AHI and 6-fold odds of developing moderate-to-severe OSA, a 10% weight loss predicts a 26% decrease in AHI (48). Recent work suggests that high-intensity exercise, specifically, is effective in reducing visceral adiposity (28).

There are several limitations to the current study. For one, only a morning (0700) blood sample was collected. As previously mentioned, while CRP lacks circadian rhythmicity and levels remain relatively stable regardless of sample collection time (43), this is not true of IL-6, TNFα (17, 64), leptin (39), and adiponectin (21). All participants were in bed from 2200 until 0700 with a blood sample taken immediately upon awakening, however, so any external influences affecting inflammation (such as wake time or a state of sleep loss/restriction) were carefully controlled. Second, there is evidence suggesting that computed tomography (CT) may be a superior method to DXA for the measurement of, specifically, visceral adiposity (30). We elected to use DXA, however, due to the higher radiation doses associated with CT scan (14) and concern that fewer subjects may be willing to participate in the body scan. Finally, while we have demonstrated that inflammation explains, to a large degree, the association between visceral adiposity and OSA in adolescents, the cross-sectional nature of this study does not allow us to assign directionality to this relationship. However, given the young age and new onset of OSA in this sample, as well as recent longitudinal studies demonstrating that obesity (57) and waist circumference (2) in childhood predict adolescent OSA, our findings suggest that visceral obesity precedes the disorder.

Our study is unique in several ways. For one, our findings from a large nonclinical general population sample of adolescents are more generalizable to public health compared with smaller clinical studies. Furthermore, the fact that we are observing a robust relationship between visceral adiposity and OSA, similar to adults (65), 1) at this young age, and 2) in a nonclinical population further suggests the causative role of abdominal fat in this association. This is also the only large study in this age group to include a more precise measure of obesity, such as a DXA scan, in addition to measures of BMI and waist circumference.

In sum, our findings in a large, representative, nonclinical sample suggest that 1) visceral fat is most strongly associated with moderate OSA, even as young as adolescence; 2) IL-6, CRP, and leptin levels are elevated in moderate OSA; and 3) inflammation is the major mediating link in the relationship between central obesity and sleep apnea, adding to our developing understanding of the pathogenesis and potential treatments for OSA.

GRANTS

This research was supported by National Institutes of Health Grants R01-HL-63772, R01-HL-97165, UL1-RR-033184, and C06-RR-6499.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

J.G., A.N.V., J.F.-M., S.L.C., D.L., and E.O.B. conception and design of research; J.G. and M.D.S. performed experiments; J.G., A.N.V., F.H., M.D.S., and E.O.B. analyzed data; J.G., A.N.V., J.F.-M., S.L.C., F.H., D.L., M.D.S., and E.O.B. interpreted results of experiments; J.G. prepared figures; J.G. drafted manuscript; J.G., A.N.V., J.F.-M., and E.O.B. edited and revised manuscript; J.G., A.N.V., J.F.-M., S.L.C., F.H., D.L., M.D.S., and E.O.B. approved final version of manuscript.