Sleep and https://doi.org/10.1007/s11325-018-1631-9

SLEEP BREATHING PHYSIOLOGY AND DISORDERS • ORIGINAL ARTICLE

Intermittent nocturnal hypoxia and metabolic risk in obese adolescents with obstructive sleep

Indra Narang1,2 & Brian W. McCrindle1,2 & Cedric Manlhiot3 & Zihang Lu1,2 & Suhail Al-Saleh1,2 & Catherine S. Birken1,2 & Jill Hamilton1,2

Received: 16 August 2017 /Revised: 6 December 2017 /Accepted: 11 January 2018 # Springer International Publishing AG, part of Springer Nature 2018

Abstract Purpose There is conflicting data regarding the independent associations of obstructive (OSA) with metabolic risk in obese youth. Previous studies have not consistently addressed central adiposity, specifically elevated waist to height ratio (WHtR), which is associated with metabolic risk independent of body mass index. Objective The objective of this study was to determine the independent effects of the obstructive apnea-hypopnea index (OAHI) and associated indices of nocturnal hypoxia on metabolic function in obese youth after adjusting for WHtR. Methods Subjects had standardized anthropometric measurements. Fasting blood included insulin, glucose, glycated hemoglo- bin, alanine transferase, and aspartate transaminase. Insulin resistance was quantified with the homeostatic model assessment. Overnight polysomnography determined the OAHI and nocturnal oxygenation indices. Results Of the 75 recruited subjects, 23% were diagnosed with OSA. Adjusting for age, gender, and WHtR in multivariable linear regression models, a higher oxygen desaturation index was associated with a higher fasting insulin (coefficient [standard error] = 48.076 [11.255], p < 0.001), higher glycated hemoglobin (coefficient [standard error] = 0.097 [0.041], p = 0.02), higher insulin resistance (coefficient [standard error] = 1.516 [0.364], p < 0.001), elevated alanine transferase (coefficient [standard error] = 11.631 [2.770], p < 0.001), and aspartate transaminase (coefficient [standard error] = 4.880 [1.444], p =0.001). However, there were no significant associations between OAHI, glucose , and liver enzymes. Conclusion Intermittent nocturnal hypoxia rather than the OAHI was associated with metabolic risk in obese youth after adjusting for WHtR. Measures of abdominal adiposity such as WHtR should be considered in future studies that evaluate the impact of OSA on metabolic health.

Keywords Nocturnal hypoxia . . Obesity . Metabolic risk

Introduction nocturnal oxyhemoglobin desaturations and sleep disrup- tion [1]. The prevalence of OSA is 1–4% in otherwise Obstructive sleep apnea (OSA) is characterized by snor- healthy school children but alarmingly reported to occur ing and recurrent partial and/or complete obstruction of in 25% of obese youth [2]. Obesity predisposes to met- the upper airway and is associated with intermittent abolic disease, including diabetes and non-alcoholic fat- ty liver disease (NAFLD) in children [3]. In recent years, the presence of OSA in adults has been shown to be a risk factor for insulin resistance and NAFLD * Indra Narang independent of obesity [4, 5]. Effective treatment for OSA [email protected] with continuous positive airway pressure (CPAP) in adults is associated with improvements in glycemic control [6]. Of 1 Department of Paediatrics, Hospital for Sick Children, emerging interest is the relative causal role of OSA and asso- Toronto, Ontario, Canada ciated in potentiating metabolic risk in obese 2 University of Toronto, Toronto, Ontario, Canada youth. However, the current available data are conflicting 3 The Cardiovascular Data Management Centre, The Hospital for Sick and some of the disparities in the reported literature are due Children, Toronto, Canada to different definitions of OSA and obesity (which is typically Sleep Breath defined by body mass index [BMI]), limiting direct compari- score was calculated according to age and sex-specific growth sons and definitive conclusions [7]. curves of the World Health Organization, and obesity was There is recent evidence to suggest that controlling for BMI defined as a BMI z score > 1.96 [11]. Waist circumference may not adequately exclude the influence of increased visceral was measured three times to the nearest 0.1 cm using a non- adiposity [8], which is also known to be a significant risk elastic fiberglass measuring tape (Tech-Med model 4414; factor associated with cardiometabolic risk in adults [9]. Moore Medical Corp., New Britain, Connecticut) placed at Specifically, an elevated waist to height ratio (WHtR), regard- the top of the iliac crest with the subject in a standing position less of BMI percentile category, is associated with an increase [12]. The WHtR was calculated by waist circumference in in cardiometabolic risk in the pediatric population [10]. As centimeters divided by height in centimeters. A WHtR > 0.6 such, WHtR may be an important discriminatory measure- was considered to indicate greater likelihood of metabolic ment associated with metabolic risk. Thus, any models eval- risk [10]. uating the independent effects of OSA on metabolic risk Fasting blood work was drawn via an intravenous catheter should be adjusted for WHtR. Since OSA has modifiable risk inserted into the antecubital fossa. Fasting metabolic variables factors, early diagnosis and targeted interventions for OSA in collected were plasma glucose, insulin, glycated hemoglobin obese youth may be associated with improved metabolic out- A1c (HbA1c), alanine transferase (ALT), and aspartate trans- comes long-term. aminase (AST). Both ALT and AST were used as surrogate The objective of this study was to determine the relation- markers of liver inflammation and possible NAFLD. Insulin ship between the indices associated with OSA, specifically the was measured by chemiluminescence (Siemens Immulite obstructive apnea-hypopnea index (OAHI), the oxygen 2500; range of assay 15–2165 pmol L21, intra- and inter- desaturation index (ODI), and nadir oxygen saturation assay coefficient of variation [CV] < 7.6%). Insulin resistance

(SaO2), with insulin resistance and hepatic enzyme levels in was quantified with the homeostatic model assessment obese youth. We hypothesized that with adjustment for central (HOMA-IR) with fasting plasma glucose and insulin measure- adiposity, specifically WHtR, the OAHI, ODI, and SaO2 are ments (HOMA-IR = [glucose (nmol/L) × insulin (μU/mL)]/ associated with metabolic risk on obese youth. 22.5). A HOMA-IR value greater than 3.16 was considered abnormal [13]. HbA1c values greater than 6.5% were consid- ered abnormal [14]. Methods Polysomnogram Study population Subjects underwent a standard overnight PSG using Natus The study sample was derived from the cross-sectional High Sleepworks system (Natus Medical Incorporated, CA, Impact Strategies Towards Overweight Reduction in Youth USA), which is a data acquisition and analysis system. PSG (HISTORY) study. The HISTORY study was designed to measurements included electroencephalogram, electro- (1) evaluate baseline metabolic status and (2) screen for oculogram, and submental and bilateral anterior tibialis elec- OSA using a formal overnight polysomnogram (PSG). tromyogram. Chest wall and abdominal movements were Subjects aged 8–18 years who were obese were consecutively measured using chest wall and abdominal belts. Other respi- recruited into this study between 2009 and 2012 using adver- ratory measurements included nasal air pressure transducer by tisements posted in the hospital, at pediatricians’ offices, and Braebon, oronasal thermal sensor, oxygen saturation (SaO2) at community pediatric obesity clinics as well as at our tertiary using a Massimo pulse oximeter (Irvine, CA), transcutaneous care institution. Exclusion criteria included known genetic or carbon dioxide (tcCO2) using a Sentec carbon dioxide sensor syndromic conditions that may contribute to obesity (e.g., (Therwil, Switzerland), and end-tidal carbon dioxide (etCO2) Prader-Willi Syndrome), use of glucose-altering medications, using a BCI Capnocheck unit (Waukesha, USA). Video and known diabetes, known liver abnormalities, use of medica- audio recordings were obtained as well as body position. tions that may significantly impact sleep architecture (e.g., Sleep architecture was assessed by standard techniques [15]. anti-depressants), youth with known OSA at the time of re- All respiratory events were scored according to the American cruitment or on CPAP treatment for OSA, developmental de- Academy of Sleep Medicine scoring guidelines [16] by a reg- lay, pregnancy, and an inability to speak English. istered certified polysomnographic technician who was blinded to the clinical status of the patients. All of the sleep Study procedures and measurements studies were reviewed and interpreted by experienced pediat- ric sleep physicians. An obstructive apnea event was scored Subjects arrived at the hospital following a 12-h overnight when airflow dropped at least 90% from baseline with chest fast. Weight and height were measured by a standard, calibrat- and/or abdominal motion throughout the entire event; the du- ed scale and wall-mounted stadiometer, respectively. BMI z ration of which was at least a minimum of two baseline Sleep Breath breaths. A hypopnea event was scored when airflow dropped Separate multivariate linear regression models were used to at least 50% from baseline, the duration of which was at least a assess the associations of OAHI, ODI, SaO2, and arousal in- minimum of two baseline breaths. The hypopnea event must dex, which is a surrogate marker for sleep fragmentation, with have been accompanied by either (i) a minimum 3% drop in the aforementioned outcome variables. Log-transformation oxygen desaturation, (ii) an arousal, or (iii) an awakening [16]. was applied to OAHI, ODI, and nadir SaO2 as a means to A central apnea was defined as a cessation of airflow with an account for non-linear associations. All models were adjusted absence of respiratory and abdominal effort for a minimum of for age, sex, and adiposity, which was measured by either 20 s or of the duration of two prior baseline breaths in which WHtR or BMI. All tests were two sided and statistical signif- case the event must be accompanied by (i) a minimum 3% icance was set at 5%. All statistical analyses were performed drop in oxygen desaturation, (ii) an arousal, or (iii) an awak- using SAS v9.3 (SAS statistical software, Cary, NC). ening. OSA was diagnosed according to the obstructive apnea-hypopnea index (OAHI), which is defined as the num- ber of obstructive , mixed apneas and obstructive Results hypopneas per hour during sleep. OSA was diagnosed if the OAHI was ≥ 5 events per hour, a value associated with at least A total of 75 subjects who completed an overnight PSG and moderate OSA in pediatric studies and mild OSA in adult fasting blood work were included in the study sample. Table 1 studies [16, 17]. The central apnea index (CAI) was defined summarizes the clinical characteristics of all subjects in the as the number of central apneas during sleep. The Epworth study. Seventeen (23%) subjects were in the OSA group, sleepiness scale for children was used as a self-reported mea- and 58 (77%) were in the no-OSA group. The mean (± SD) sure of sleepiness [18]. Our exposure variables were (1) OAHI age of the subjects in the OSA group and the no-OSA group of ≥ 5 events per hour and (2) measures of nocturnal (sleep) was 15.0 (± 2.2) and 14.2 (± 2.5) years, respectively (p = hypoxia including the ODI (the number of desaturations ≥ 3% 0.23). There were no statistically significant differences in per hour) and nadir SaO2. Our primary outcome of interest height, gender, and ethnicity between the two groups. The was a measure of insulin resistance, HOMA-IR. between-group differences in prior adenoidectomy and/or ton- sillectomy were also non-significant. The OSA group had a Ethics higher BMI z score and WHtR than the no-OSA group (both p =0.06). The study protocol was approved by the Research Ethics Table 1 also shows the median (IQR) OAHI in the OSA Board at the Hospital for Sick Children, Toronto, Canada and no-OSA groups which were 13.0 (7.0–28.2) and 1.3 (0.5– (Research Ethics Board number 1000013173). All subjects 2.6) events per hour, respectively (p = 0.009). As would be and parents provided written informed consent, and assent expected, the OSA group, compared to the no-OSA group, were obtained from guardians and/or participants as had a significantly higher ODI (p < 0.001), a lower mean appropriate. SaO2 (p < 0.001) and a lower nadir SaO2 (p =0.02). Additionally, the OSA group had significantly shorter rapid Statistical analyses eye movement (REM) latency (p < 0.001), increased REM sleep time (p =0.05),decreasedN3(p < 0.001), and a signif- Data were described as means with standard deviations (SD), icantly higher arousal index (p = 0.009). The OSA group also median with interquartile ranges (IQR), and frequencies as had increased sleepiness (i.e., higher Epworth sleepiness appropriate. Variables and categories with low frequency are scale) than the no-OSA group (p =0.06). reported in the descriptive statistics but were collapsed (when Table 2 summarizes the glucose metabolism and liver en- possible) or excluded from further analyses. Basic compari- zymes in the OSA and no-OSA groups. Although the fasting sons between diagnostic groups were performed using insulin and HOMA-IR were higher in the OSA group com- Fisher’s exact tests for all categorical variables and two- pared to the no-OSA group, these differences were not statis- sample t tests adjusting for unequal variance between samples tically significant. for continuous variables. The outcomes considered in the The association of OAHI with glucose metabolism and study were glucose metabolism and liver enzymes. Glucose liver enzymes using multivariate linear regression is summa- metabolism were measured by fasting plasma glucose, fasting rized in Table 3. The results were adjusted for age, sex, and insulin, HbA1c, and HOMA-IR. The liver enzymes were mea- WHtR. There were no significant associations between sured by ALT and AST. OAHI is a global index used to diag- OAHI, glucose metabolism and liver enzymes in these analy- nose OSA. For the descriptive analysis, all subjects were ses. Additional analyses were conducted to assess the associ- grouped into two mutually exclusive categories (No-OSA: ation of OAHI with the outcome variables adjusted for age, OAHI< 5 vs. OSA: OAHI ≥ 5). The clinical characteristics sex, and BMI (the BMI adjusted results not shown), which were then assessed and compared between the two groups. were also not significant. Sleep Breath

Table 1 Baseline characteristics and PSG data for OSA and no- Variable OSA group (n = 17) No-OSA group (n =58) p value OSA groups Male, number (%) 11 (65%) 28 (48%) 0.28 Ethnicity, no. (%) Caucasian 12 (71%) 38 (66%) 0.78 Age (years) 15.0 + 2.2 14.2 ± 2.5 0.23 Height (cm) 167 ± 9 163 ± 11 0.07 Weight (kg) 127.0 ± 38.0 101.5 ± 34.6 0.02 BMI z score 2.66 ± 0.59 2.34 ± 0.54 0.06 WHtR 0.75 ± 0.17 0.66 ± 0.13 0.06 Tonsils removed, no. (%) 4 (24%) 15 (26%) 1.00 Adenoids removed, no. (%) 5 (29%) 14 (24%) 0.75 OAHI 13.0 (7.0–28.2) 1.3 (0.5–2.6) 0.009 CAI 0.1 (0.0–0.4) 0.0 (0.0–0.3) 0.65

Mean SaO2 (%) 96.2 ± 1.5 97.8 ± 0.8 < 0.001

Min SaO2 (%) 81.5 ± 11.8 89.4 ± 5.0 0.02 ODI 3.6 (2.4–22.8) 0.8 (0.2–2.6) < 0.001

Mean tcCO2 45.5 ± 5.4 42.1 ± 4.9 0.11 Sleep efficiency (%) 82.4 ± 6.7 77.8 ± 13.4 0.06 Total sleep duration (min) 368.6 ± 79.8 327.4 ± 69.0 0.07 Sleep onset latency (minutes) 14 (12–24) 18 (8–40) 0.15 REM latency (minutes) 123 (92–157) 167 (126–212) < 0.001 Stage N1 (% TST) 8.8 ± 5.8 7.7 ± 4.4 0.45 Stage N2 (% TST) 51.2 ± 14.1 50.5 ± 9.7 0.85 Stage N3 (% TST) 17.9 ± 11.1 27.4 ± 9.4 < 0.001 REM sleep (% TST) 18.2 ± 6.2 14.4 ± 6.4 0.05 Arousal index 19.0 ± 15.8 10.2 ± 6.4 0.009 Epworth sleepiness scale 8.0 (5.0–10.0) 5.0 (3.0–8.0) 0.06

Values shown are mean (± SD) or median (IQR) unless stated otherwise. p values were calculated using Fisher’s exact test for categorical variables and two-sample t-tests adjusted for unequal variance for continuous variables BMI body mass index, WHtR waisttoheightratio,OSA obstructive sleep apnea, REM , OAHI obstructive apnea-hypopnea index, CAI central apnea index, SaO2 oxygen saturations as measured by a pulse oximeter, ODI oxygen desaturation index, tcCO2 transcutaneous carbon dioxide

Nocturnal hypoxia, glucose metabolism, and liver nadir) and metabolic variables. As shown in Table 4,anin- enzymes crease in ODI was associated with an increase in fasting insu- lin (coefficient [standard error, SE] = 48.076 [11.255], Further analyses were undertaken to explore the association p < 0.001), HbA1c (Coef [SE] = 0.097 [0.041], p =0.02), between oxygenation indices (specifically ODI and SaO2 HOMA-IR (Coef [SE] = 1.516 [0.364], p < 0.001), and an

Table 2 Metabolic variables in OSA and non-OSA groups Variable OSA group (n = 17) Non-OSA group (n =58) p value

Fasting glucose (mmol/L) 4.8 ± 0.5 4.7 ± 0.5 0.47 Fasting insulin (pmol/L) 153 (103–340) 113 (58–163) 0.51 HbA1c 0.05 (± 0.004) 0.05 (± 0.003) 0.77 HOMA-IR 5.0 (3.1–9.5) 3.5 (1.7–5.6) 0.12 ALT (U/L) 29.0 (24.0–36.0) 29.0 (24.0–39.0) 0.50 AST (U/L) 25.0 (22.0–30.0) 24.0 (21.0–29.0) 0.67 MetS, no. (%) 2 (11%) 4 (7%) 0.61

Values shown are mean (± SD) or median (IQR) unless stated otherwise. p values were calculated two-sample t tests adjusted for unequal variance between the two comparison groups OSA obstructive sleep apnea, HbA1c glycated hemoglobin, HOMA-IR homeostatic model assessment for insulin resistance, ALT alanine transaminase, AST aspartate transaminase Sleep Breath

Table 3 Adjusted association of OAHI with metabolic variables of measures of central adiposity, specifically WHtR. Variable Log-transformed OAHI Importantly, adequate adjustment for visceral adiposity rather than adjustments for BMI yielded different results, stressing Coef SE p value the importance of measurements of WHtR. These data suggest that intermittent nocturnal hypoxia may be implicated in the − Fasting glucose (mmol/L) 0.012 0.013 0.34 pathogenesis of OSA-related insulin resistance and NAFLD. − Fasting insulin (pmol/L) 1.02 2.492 0.97 Previous pediatric studies evaluating the presence of OSA HbA1c 0.003 0.008 0.69 and the association with metabolic risk have been conflicting − HOMA-IR 0.012 0.080 0.88 with some suggesting that OSA, particularly the OAHI, po- ALT (U/L) 0.588 0.583 0.31 tentiates metabolic risk [19, 20], whereas other published data AST (U/L) 0.326 0.289 0.26 have suggested that adiposity is the main driver of metabolic OAHI obstructive apnea-hypopnea index, SE standard error, HbA1c risk [21, 22]. Direct comparisons between studies are difficult glycated hemoglobin, HOMA-IR homeostatic model assessment for insu- due to differences in the groups with regard to the presence lin resistance, ALT alanine transaminase, AST aspartate transaminase and severity of OSA, the differing definitions of obesity and OSA, and the differing metabolic variables. In contrast, few elevation in ALT (Coef [SE] = 11.631 [2.770], p <0.001)and studies have addressed the potential negative contributions of AST (Coef [SE] = 4.880 [1.444], p = 0.001) after adjusting for intermittent hypoxia associated with OSA in children and age, sex, and WHtR. A decrease in SaO2 nadir was associated adolescents. with an increase in fasting insulin, HbA1c, HOMA-IR, ALT, In a study by Redline and colleagues [20], where 25% of and AST (Table 4). the adolescent cohort was overweight, both OSA (OAHI > 5 z An additional analysis was performed where BMI events per hour) and intermittent hypoxia (SaO2 nadir and % scores, instead of WHtR, wereusedasameasureofadi- of time SaO2 < 90%) were associated with an increased risk posity, where there were no significant associations of for metabolic syndrome, which was observed in 59% of their OAHI and ODI with glucose metabolism. However, a de- subjects with OSA. Additionally, after adjusting for BMI and crease in the nadir SaO2 was significantly associated with gender, OAHI strongly correlated with abnormal glucose me- higher ALT (Coef [SE] = − 68.22 [26.95], p = 0.01) and tabolism (i.e., fasting insulin, HOMA-IR). However, com- higher AST (− 35.61 [13.65], p = 0.01). pared to the current study, their study was over-represented Finally, after adjusting for age, sex, and WHtR, an increase with regard to the presence of metabolic syndrome, preterm in arousal index was significantly associated with higher ALT status (57% of their cohort), and African-American ethnicity (Coef [SE] = 1.557 [0.205], p < 0.001) and higher AST (Coef (54% of their cohort). This is relevant as prematurity and [SE] = 0.605 [0.117], p < 0.001) but was not associated with African-American ethnicity are risk factors associated with markers of glucose metabolism. insulin resistance and OSA [23, 24]. In a study by Kelly

et al. [25], both OAHI and SaO2 indices were not associated with HOMA-IR in a sample of obese pubertal adolescents Discussion after adjusting for BMI. However, the lack of association in that study may be related to the fact that BMI does not neces- In a cohort of obese subjects with OSA, mild intermittent sarily make adequate adjustments for visceral adiposity, which nocturnal hypoxia that is characteristic of OSA, rather than may have masked the influence of OSA and intermittent hyp- OAHI, may be linked with metabolic risk factors independent oxia on insulin resistance.

Table 4 Adjusted association of Variable Log-transformed ODI Log-transformed SaO nadir ODI and SaO2 nadir with 2 metabolic variables Coef SE p value Coef SE p value

Fasting glucose (mmol/L) 0.029 0.064 0.65 0.209 0.618 0.74 Fasting insulin (pmol/L) 48.076 11.255 < 0.001 − 280.3 117.7 0.02 HbA1c 0.097 0.041 0.02 − 0.863 0.396 0.03 HOMA-IR 1.516 0.364 < 0.001 − 8.838 3.796 0.02 ALT (U/L) 11.631 2.770 < 0.001 − 67.46 28.88 0.02 AST (U/L) 4.880 1.444 0.001 − 28.03 14.68 0.06

ODI oxygen desaturation index, SaO2 oxygen saturations as measured by a pulse oximeter, SE standard error, HbA1c glycated hemoglobin, HOMA-IR homeostatic model assessment for insulin resistance, ALT alanine trans- aminase, AST aspartate transaminase Sleep Breath

Among 7-year-old pre-pubertal lean and obese children, the no-OSA subjects. Although the reasons for this are un- Gozal and colleagues did not report an association between clear, it is possible that N3 was more impacted than REM both OAHI and SaO2 with measures of altered glucose me- sleep in this obese cohort. Indeed, in a study with an adult tabolism in children with OSA [22, 26]. They determined that population, reduced N3 was associated with both obesity both obesity and sleep fragmentation were predominantly as- and severe OSA, where even after adjusting for obesity, the sociated with insulin resistance [22]. However, compared to association between severe OSA and reduced N3 remained their subjects, our cohort was older by an average of 7 years [36]. However, these findings deserve further investigation and consisted of all obese subjects so it is likely that obesity in future studies. and OSA co-existed for a longer duration in our subjects. The strength of this study was that the subjects were not Other considerations include genetics as one study observed selected on the basis of symptoms and represent a group of that genetic predisposition to hyperinsulinemia may con- unselected obese youth population, so our results are likely to tribute to sleep disordered breathing in obese subjects be generalizable to a wider obese community based cohort. [27]. Moreover, since both OSA and obesity are recog- Additionally, this was a prospective study which incorporated nized as inflammatory conditions, OSA in the context of fasting blood samples as well as a formal full overnight PSG. obesity may magnify the contributions of obesity that pre- We adjusted all our analyses for WHtR, for which an elevated dispose to metabolic risk. Finally, our findings concur level is known to be a robust marker of central adiposity and is with other data showing that the presence of hypoxia in- independently related to metabolic dysfunction. Some of the dependent of abdominal adiposity is associated with liver limitations of this study include the lack of a lean control injury and inflammation [28]. group with OSA. However, in the pediatric population, there Our study did not find an association between apnea- are very few lean subjects with OSA who are greater than hypopnea index (AHI) and insulin resistance as observed 12 years of age. As such, the recruitment of a lean adolescent in an adult study with over 10,000 subjects [29]. control group is not feasible. We also did not record respira- Interestingly, the study showed that the mean AHI in those tory effort related arousals (RERAs) which may have more with diabetes was 25.7 events per hour compared to 14.7 accurately identified respiratory events. We did not have an events per hour in adults without diabetes [29]. One possi- accurate evaluation of pubertal status which is known to ble explanation for the lack of association between OAHI influence insulin resistance [37]. Although more than and insulin resistance in our study was the mean AHI was 20% of our cohort had OSA, there are no data on the impact only 13 events per hour in a young cohort and chronic of treatment for OSA on HOMA-IR or liver enzymes. severe OSA over many years may exert greater influence Further, the cross-sectional design of this study does not on insulin resistance downstream. permit us to draw conclusions on any causal relationship Although our study was unable to delineate mecha- between intermittent hypoxemia and possible metabolic nisms by which hypoxia may be a risk factor for insulin risk. This may also be relevant in the context that given resistance, there are relevant data from animal and human the sample size of the study, regression models can only models as highlighted elegantly in a recent review [30]. accommodate a limited number of covariates and as such, OSA is associated with episodes of intermittent hypoxia we were unable to adjust for all the potential confounders. and re-oxygenation which can lead to intermittent tissue More large-scale longitudinal research is needed that is hypoxia and contribute to the formation of reactive oxy- aimed at elucidating the potential mechanisms of OSA- gen species that increase oxidative stress and can activate related nocturnal hypoxia and long-term metabolic risk. redox-sensitive signaling pathways. This can lead to acti- Chronic sleep deprivation has been associated with insulin vation of the sympathetic nervous system (SNS) and in- resistance [38], but we did not have objective data over a flammation [31]. Chronic SNS activation has been shown period of time on sleep duration or sleep disturbance. Our to be associated with inflammation including adipose tis- surrogate marker of sleep fragmentation, the arousal index, sue inflammation, hepatic apoptosis [32], and insulin re- derived from one night was not associated with altered glu- sistance downstream [30, 33]. In one study, mice exposed cose metabolism but was associated with hepatic enzymes. to 14 days of intermittent hypoxia (60 desaturations per As such, multiple nights of sleep monitoring may have hour) had impairments in insulin sensitivity, beta cell provided us with more accurate measure of sleep architec- function, and glucose tolerance tests, which improved af- ture and sleep fragmentation. Finally, measures used to as- ter normoxia was re-established [34]. In limited human sess NAFLD are indirect; hepatic enzymes are elevated experimental data, acute hypoxia for 30 min (oxygen sat- only in a subset of individuals with NAFLD [32, 33]. urations of 75%) in 14 adult healthy males was associated Future studies will need to address the impact of sleep du- with impaired glucose tolerance [35]. ration and disturbance in the absence of OSA on metabolic Moreover, we observed normal REM sleep time but signif- risk in an adolescent cohort especially given our knowledge icantly reduced N3 time in the OSA subjects compared with that > 50% of adolescents are chronically sleep deprived. Sleep Breath

Summary and conclusions Funding The Canadian Institute of Health Research provided financial support in the form of research funding for childhood obesity. The spon- sor had no role in the design or conduct of this research. In this cross-sectional study of obese youth who had an overnight PSG and fasting blood work performed, OSA- Compliance with ethical standards related intermittent nocturnal hypoxia, rather than OAHI, itselfmaybelinkedwithmetabolicriskinobeseadoles- Conflict of interest All authors certify that they have no affiliations with cents independent of visceral adiposity. Although causal- or involvement in any organization or entity with any financial interest ’ ity cannot be inferred from this study, early screening and (such as honoraria; educational grants; participation in speakers bureaus; membership, employment, consultancies, stock ownership, or other eq- targeted therapeutic interventions for OSA should be op- uity interest; and expert testimony or patent-licensing arrangements), or timized in obese youth to minimize any potential long- non-financial interest (such as personal or professional relationships, af- term complications of OSA. This is particularly important filiations, knowledge, or beliefs)inthesubjectmatterormaterials as childhood obesity tracks into adulthood; alarmingly, discussed in this manuscript. 75% of obese children will seamlessly transition to adult- Ethical approval All procedures performed in studies involving human hood as obese adults with associated chronic cardiovascu- participants were in accordance with the ethical standards of the institu- lar and metabolic morbidities [39, 40]. tional and/or national research committee and with the 1964 Helsinki Importantly, measures of intermittent hypoxia may be declaration and its later amendments or comparable ethical standards. more useful and relevant in future studies evaluating long-term metabolic risk in OSA subjects. Specifically, Informed consent Informed consent was obtained from all individual participants included in the study. SaO2 indicesasanoutcomevariablemaybemoreuseful as they are of clinical relevance and are easy to measure and interpret due to the availability of normative data sets on SaO2. Additionally, SaO2 has less variance than the References OAHI. Surrogate measures of abdominal adiposity such as WHtR should be considered in future studies evaluating 1. Bradley TD, Floras JS (2009) Obstructive sleep apnoea and its the independent effects of OSA on cardiovascular and met- cardiovascular consequences. Lancet 373(9657):82–93. https:// abolic health in obese individuals. doi.org/10.1016/S0140-6736(08)61622-0 2. Verhulst SL, Van Gaal L, De Backer W, Desager K (2008) The prevalence, anatomical correlates and treatment of sleep- Sources of support This work was supported by a Canadian Institute of disordered breathing in obese children and adolescents. Sleep Health Research team grant in childhood obesity (HISTORY: High Med Rev 12(5):339–346. https://doi.org/10.1016/j.smrv.2007.11. Impact Strategies Toward Overweight Reduction in Youth). JH is sup- 002 ported with unrestricted research funds by the Mead Johnson Chair in Nutritional Science. 3. Friedemann C, Heneghan C, Mahtani K, Thompson M, Perera R, Subject Category List for Authors: 14.5 Pediatric Sleep and Control of Ward AM (2012) risk in healthy children Breathing. and its association with body mass index: systematic review and meta-analysis. BMJ 345(sep25 2):e4759. https://doi.org/10.1136/ bmj.e4759 Authors’ contributions Dr. Indra Narang contributed substantially as the 4. Musso G, Cassader M, Olivetti C, Rosina F, Carbone G, Gambino first author by designing the study, acquiring the data, reviewing the data R (2013) Association of obstructive sleep apnoea with the presence and statistical results in depth, writing the manuscript, and approving the and severity of non-alcoholic fatty liver disease. A systematic re- final version of the manuscript. view and meta-analysis. Obes Rev 14(5):417–431. https://doi.org/ Dr. Brian McCrindle contributed substantially as a co-author by de- 10.1111/obr.12020 signing the study, reviewing the data, revising the manuscript critically, and approving the final version of the manuscript. 5. Punjabi NM, Shahar E, Redline S, Gottlieb DJ, Givelber R, Resnick Cedric Manlhiot contributed substantially as a co-author by designing HE (2004) Sleep-disordered breathing, glucose intolerance, and the study, reviewing the data, undertaking the statistical analyses, revising insulin resistance: the Sleep Heart Health Study. Am J Epidemiol – the manuscript critically, and approving the final version of the 160(6):521 530. https://doi.org/10.1093/aje/kwh261 manuscript. 6. Martinez-Ceron E, Barquiel B, Bezos AM, Casitas R, Galera R, Zihang Lu contributed substantially as a co-author by reviewing the Garcia-Benito C, Hernanz A, Alonso-Fernandez A, Garcia-Rio F data, undertaking the statistical analyses, revising the manuscript critical- (2016) Effect of continuous positive airway pressure on glycemic ly and approving the final version of the manuscript. control in patients with obstructive sleep apnea and type 2 diabetes. Dr. Suhail Al-Saleh contributed substantially as a co-author by A randomized clinical trial. Am J Respir Crit Care Med 194(4): – reviewing the data, revising the manuscript critically, and approving the 476 485. https://doi.org/10.1164/rccm.201510-1942OC final version of the manuscript. 7. Narang I, Mathew JL (2012) Childhood obesity and obstructive Dr. Catherine Birken contributed substantially as a co-author by sleep apnea. J Nutr Metab 2012:134202. https://doi.org/10.1155/ reviewing the data, revising the manuscript critically, and approving the 2012/134202 final version of the manuscript. 8. Carotenuto M, Bruni O, Santoro N, Del Giudice EM, Perrone L, Dr. Jill Hamilton contributed substantially as the senior author by Pascotto A (2006) Waist circumference predicts the occurrence of designing the study, acquiring the data, reviewing the data and statistical sleep-disordered breathing in obese children and adolescents: a results in depth, writing the manuscript, and approving the final version of questionnaire-based study. Sleep Med 7(4):357–361. https://doi. the manuscript. org/10.1016/j.sleep.2006.01.005 Sleep Breath

9. Shah RV, Murthy VL, Abbasi SA, Blankstein R, Kwong RY, 26. Gozal D, Capdevila OS, Kheirandish-Gozal L (2008) Metabolic Goldfine AB, Jerosch-Herold M, Lima JA, Ding J, Allison MA alterations and systemic inflammation in obstructive sleep apnea (2014) Visceral adiposity and the risk of metabolic syndrome across among nonobese and obese prepubertal children. Am J Respir body mass index: the MESA Study. JACC Cardiovascular Imaging Crit Care Med 177(10):1142–1149. https://doi.org/10.1164/rccm. 7(12):1221–1235. https://doi.org/10.1016/j.jcmg.2014.07.017 200711-1670OC 10. Khoury M, Manlhiot C, McCrindle BW (2013) Role of the waist/ 27. Carotenuto M, Santoro N, Grandone A, Santoro E, Pascotto C, height ratio in the cardiometabolic risk assessment of children clas- Pascotto A, Perrone L, del Giudice EM (2009) The insulin gene sified by body mass index. J Am Coll Cardiol 62(8):742–751. variable number of tandemrepeats (INS VNTR) genotype and sleep https://doi.org/10.1016/j.jacc.2013.01.026 disordered breathing in childhood obesity. J Endocrinol Investig 11. (WHO) WHO Child Growth Standards http://www.who.int/ 32(9):752– 755. https://doi.org/10.1007/BF03346531 childgrowth/en/ 28. Nobili V, Cutrera R, Liccardo D, Pavone M, Devito R, Giorgio V, 12. de Onis MOA, Van den Broeck J, Chumlea WC, Martorell R (2004) Verrillo E, Baviera G, Musso G (2014) Obstructive sleep apnea Measurement and standardization protocols for anthropometry used syndrome affects liver histology and inflammatory cell activation in the construction of a new international growth reference. Food in pediatric nonalcoholic fatty liver disease, regardless of obesity/ Nutr Bull 25(1 Suppl):S27–S36 insulin resistance. Am J Respir Crit Care Med 189(1):66–76. 13. Tresaco B, Bueno G, Pineda I, Moreno LA, Garagorri JM, Bueno https://doi.org/10.1164/rccm.201307-1339OC M (2005) Homeostatic model assessment (HOMA) index cut-off 29. Kendzerska T, Gershon AS, Hawker G, Tomlinson G, Leung RS values to identify the metabolic syndrome in children. J Physiol (2014) Obstructive sleep apnea and incident diabetes. A historical – Biochem 61(2):381 388. https://doi.org/10.1007/BF03167055 cohort study. Am J Respir Crit Care Med 190(2):218–225. https:// — 14. Alberti KG, Zimmet P, Shaw J (2006) Metabolic syndrome anew doi.org/10.1164/rccm.201312-2209OC world-wide definition. A consensus statement from the – 30. Ryan S (2017) Adipose tissue inflammation by intermittent hypox- International Diabetes Federation. Diabet Med 23(5):469 480. ia: mechanistic link between obstructive sleep apnoea and metabol- https://doi.org/10.1111/j.1464-5491.2006.01858.x ic dysfunction. J Physiol 595(8):2423–2430. https://doi.org/10. 15. Rechtschaffen AA Manual of standardized terminiology: tech- 1113/JP273312 niques and scoring system for sleep stages. UCLA brain informa- 31. Lavie L (2008) Intermittent hypoxia: the culprit of oxidative stress, tion service. Brain Research Institute, Los Angeles vascular inflammation and dyslipidemia in obstructive sleep apnea. 16. Iber CA-I, Chesson A, Quan S (2007) The AASM manual for the Expert Rev Respir Med 2(1):75–84. https://doi.org/10.1586/ scoring of sleep and associated events: rules, terminology and tech- 17476348.2.1.75 nical specifications. 1st ed. edn. American Academy of Sleep Medicine Westchester 32. Alkhouri N, Kheirandish-Gozal L, Matloob A, Alonso-Alvarez 17. Uliel S, Tauman R, Greenfeld M, Sivan Y (2004) Normal polysom- ML, Khalyfa A, Teran-Santos J, Okwu V, Lopez R, Gileles-Hillel nographic respiratory values in children and adolescents. Chest A, Dweik R, Gozal D (2015) Evaluation of circulating markers of – hepatic apoptosis and inflammation in obese children with and 125(3):872 878. https://doi.org/10.1378/chest.125.3.872 – 18. Johns M (2015) The assessment of sleepiness in children and ado- without obstructive sleep apnea. Sleep Med 16(9):1031 1035. lescents. Sleep Biol Rhythms 13(Suppl 1):97 https://doi.org/10.1016/j.sleep.2015.05.002 19. Gozal D, Kheirandish-Gozal L (2008) Cardiovascular morbidity in 33. Carlson JT, Hedner J, Elam M, Ejnell H, Sellgren J, Wallin BG obstructive sleep apnea: oxidative stress, inflammation, and much (1993) Augmented resting sympathetic activity in awake patients – more. Am J Respir Crit Care Med 177(4):369–375. https://doi.org/ with obstructive sleep apnea. Chest 103(6):1763 1768. https://doi. 10.1164/rccm.200608-1190PP org/10.1378/chest.103.6.1763 20. Redline S, Storfer-Isser A, Rosen CL, Johnson NL, Kirchner HL, 34. Polak J, Shimoda LA, Drager LF, Undem C, McHugh H, Polotsky Emancipator J, Kibler AM (2007) Association between metabolic VY, Punjabi NM (2013) Intermittent hypoxia impairs glucose ho- syndrome and sleep-disordered breathing in adolescents. Am J meostasis in C57BL6/J mice: partial improvement with cessation of Respir Crit Care Med 176(4):401–408. https://doi.org/10.1164/ the exposure. Sleep 36(10):1483–1490; 1490A-1490B. https://doi. rccm.200703-375OC org/10.5665/sleep.3040 21. Canapari CA, Hoppin AG, Kinane TB, Thomas BJ, Torriani M, 35. Louis M, Punjabi NM (2009) Effects of acute intermittent hypoxia Katz ES (2011) Relationship between sleep apnea, fat distribution, on glucose metabolism in awake healthy volunteers. J Appl Physiol and insulin resistance in obese children. J Clin Sleep Med 7(3):268– (1985) 106(5):1538–1544. https://doi.org/10.1152/japplphysiol. 273. https://doi.org/10.5664/JCSM.1068 91523.2008 22. Koren D, Gozal D, Philby MF, Bhattacharjee R, Kheirandish-Gozal 36. Mokhlesi B, Pannain S, Ghods F, Knutson KL (2012) Predictors of L (2016) Impact of obstructive sleep apnoea on insulin resistance in slow-wave sleep in a clinic-based sample. J Sleep Res 21(2):170– nonobese and obese children. Eur Respir J 47(4):1152–1161. 175. https://doi.org/10.1111/j.1365-2869.2011.00959.x https://doi.org/10.1183/13993003.01430-2015 37. Kelsey MM, Zeitler PS (2016) Insulin resistance of puberty. Curr 23. Wang G, Divall S, Radovick S, Paige D, Ning Y, Chen Z, Ji Y, Diab Rep 16(7):64. https://doi.org/10.1007/s11892-016-0751-5 Hong X, Walker SO, Caruso D, Pearson C, Wang MC, Zuckerman 38. Donga E, van Dijk M, van Dijk JG, Biermasz NR, Lammers GJ, B, Cheng TL, Wang X (2014) Preterm birth and random plasma van Kralingen KW, Corssmit EP, Romijn JA (2010) A single night insulin levels at birth and in early childhood. JAMA 311(6):587– of partial sleep deprivation induces insulin resistance in multiple 596. https://doi.org/10.1001/jama.2014.1 metabolic pathways in healthy subjects. J Clin Endocrinol Metab 24. Ferdinand KC, Nasser SA (2015) Understanding the importance of 95(6):2963–2968. https://doi.org/10.1210/jc.2009-2430 race/ethnicity in the care of the hypertensive patient. Curr 39. McCrindle BW (2015) Cardiovascular consequences of childhood Hypertens Rep 17(4):526. https://doi.org/10.1007/s11906-014- obesity. Can J Cardiol 31(2):124–130. https://doi.org/10.1016/j. 0526-9 cjca.2014.08.017 25. Kelly A, Dougherty S, Cucchiara A, Marcus CL, Brooks LJ (2010) 40. Guo SS, Wu W, Chumlea WC, Roche AF (2002) Predicting over- Catecholamines, adiponectin, and insulin resistance as measured by – weight and obesity in adulthood from body mass index values in HOMA in children with obstructive sleep apnea. Sleep 33(9):1185 childhood and adolescence. Am J Clin Nutr 76(3):653–658 1191. https://doi.org/10.1093/sleep/33.9.1185