Impact of obstructive sleep apnoea on insulin resistance in nonobese and obese children

Subjects

This was an observational study of healthy snoring children, approved by the University of Louisville Human Research Committee (protocol 474.99) and the University of Chicago Institutional Review Board (protocol 09-115-B). Snoring children aged 5–12 years undergoing evaluation for suspected OSA were prospectively recruited. Children with genetic syndromes or chronic medical conditions (excepting well-controlled asthma on no controller medications) and children receiving medications potentially affecting sleep, insulin or glucose homeostasis, or lipids (e.g. systemic glucocorticoids within 1 month of the study) were excluded from participation. Informed consent and age-appropriate assent were obtained.

Anthropometric measurements

All participants underwent anthropometric measurements. An electronic scale assessed weight. Height was assessed using a wall-mounted stadiometer. Body mass index (BMI; kg·m⁻²) was calculated. Age- and sex-adjusted standard scores (z-scores) for BMI were calculated [12]. Study participants with BMI z-score ≥1.64 (BMI ≥95th percentile) were deemed obese [13]; other children were classified as nonobese.

Overnight polysomnography

All participants underwent overnight polysomnography (PSG) at Kosair Children's Hospital (Louisville, KY, USA) or at the Pediatric Sleep Laboratory at the University of Chicago Medical Center (Chicago, IL, USA). Parameters recorded and digitised using commercially available software included: eight-channel electroencephalogram (EEG), chin and bilateral anterior tibial and forearm electromyograms, bilateral electro-oculogram, heart rate by ECG, chest and abdominal wall movement by semicalibrated respiratory inductance plethysmography with derivation of sum from Lisajou analysis, airflow monitoring via side-stream end-tidal capnography, breath-by-breath assessment of end-tidal carbon dioxide levels via capnography (BCi International, Waukesha, WI, USA), nasal pressure transducer (Braebon, Kanata, ON, Canada), oronasal thermistor, pulse oximetry (SET; Massimo, Irvine, CA, USA) with simultaneous recording of the pulse waveform for arterial oxygen saturation (S_pO2), and analogue output from a body position sensor were simultaneously monitored. All measures were digitised using a commercially available PSG system (Nihon Kohden, Tokyo, Japan). Sleep architecture, respiratory events and arousals were scored by the same investigators (L.K.G., R.B. and D.G.) serially at the two sites using standard paediatric criteria [14].

Apnoeas were defined as ≥90% decrement in airflow lasting at least two breaths and hypopnoeas were defined as >50% decrement in nasal airflow accompanied by either a 3% desaturation or an EEG arousal from sleep lasting ≥3 s. Arousals were defined as per the revised American Academy of Sleep medicine guidelines [14]. Mean and nadir S_pO2 were assessed. Children with an obstructive apnoea/hypopnoea index (AHI), defined as number of apnoeas and hypopnoea per hour of sleep, of <1.0 events·h^–1 total sleep time (TST) were deemed to have normal breathing during sleep, children with AHI 1–5 events·h^–1 TST were considered to have mild OSA, and children with AHI ≥5 events·h^–1 TST were considered to have moderate–severe OSA [11].

Metabolic parameters

The morning following the sleep study, participants had fasting blood drawn. Fasting plasma insulin (FPI) levels were measured using a radioimmunoassay kit (Coat-A-Count Insulin; Diagnostics Products, now Siemens Medical Solutions Diagnostics, assay no longer manufactured), sensitivity 1.2 μIU·mL⁻¹, and subsequently using a solid-phase, two-site chemiluminescent immunometric assay (Immulite 2000; Siemens Medical Solutions Diagnostics, Los Angeles, CA, USA), measuring range 2–300 μIU·mL⁻¹. Of note, the initial insulin assay was more sensitive, allowing insulin values between 1.2 and 2 μIU·mL⁻¹ to be determined; children with equivalent insulin sensitivity with insulin measured using the second assay would have insulin values reported as <2 μIU·mL⁻¹, meaning homeostasis model assessment of insulin resistance (HOMA-IR) and the McAuley index could not be calculated or mean HOMA-IR values for the group measured with the second insulin assay. To avoid introducing a potential bias between children whose insulin levels were measured with the two different methodologies due to measurement sensitivity differences, any child with FPI <2.0 μIU·mL⁻¹ was excluded from analyses. Fasting plasma glucose (FPG) was measured using a hexokinase glucose 6-phosphate dehydrogenase method (Flex Reagent Cartridges; Dade Behring, Newark, DE, USA) and using a UV enzymatic method with hexokinase (Roche Cobas 8000 702 platform; Roche, Basel, Switzerland); FPG levels <63 mg·dL⁻¹ were excluded from analyses as artefactual (likely due to excessive specimen processing time). Children with FPG values consistent with T2DM (>125 mg·dL⁻¹) were excluded from analyses, as they constitute a separate metabolic disease population.

High-density lipoprotein (HDL), low-density lipoprotein (LDL), total cholesterol and triglycerides were measured using homogenous enzymatic colorimetric assays (Roche Cobas 8000 502 platform). Analytical measuring range and assay coefficients of variations (CVs) were: 1) total cholesterol: range 3.86–800 mg·dL⁻¹, CV ≤1.6%; 2) HDL cholesterol: range 3–120 mg·dL⁻¹, CV ≤1.5%; 3) LDL cholesterol: range 3.86–548 mg·dL⁻¹, range ≤2.7%; and 4) triglycerides: range 8.85–885 mg·dL⁻¹, CV ≤2.0%. Total cholesterol/HDL, LDL/HDL and triglyceride/HDL ratios were calculated. Higher values for total cholesterol/HDL [15], LDL/HDL [16] and triglyceride/HDL [17] ratios are associated with increased cardiovascular disease risk.

Calculated insulin sensitivity parameters

HOMA-IR was calculated as: (FPI (μIU·mL⁻¹)×FPG (mg·dL⁻¹))/405 [18, 19]. HOMA-IR values >2.5 were considered “abnormal” (i.e. consistent with insulin resistance) [20].

The McAuley index, which incorporates a weighted combination of FPI and triglycerides [21], was calculated as: exp(2.63–0.28×ln(FPI (μIU·mL⁻¹)–0.31×ln(triglycerides (mmol·L⁻¹)). Lower values denote greater insulin resistance.

Statistical analyses

Statistical analyses were performed using SPSS version 22.0 (IBM, Armonk, NY, USA). Skewed variables were natural log-transformed to normalise distributions. Correlation analyses were used to examine associations between OSA measures and anthropometric and metabolic variables. Stepwise multivariate linear regression models were constructed to assess relationships between sleep measures (e.g. TST or AHI) and metabolic variables (e.g. FPI) while controlling for BMI and, if significantly associated on correlation analyses, age. A p-value<0.05 was used as the cut-off for statistical significance. No differences emerged between the two study sites; as such, the data were merged and the cohort analysed as one group. As obese participants could be more susceptible to OSA's metabolic sequelae, and as severe versus mild OSA could have a differential metabolic impact, subjects were divided into three groups of OSA severity, and subdivided into nonobese and obese within OSA groups. Chi-squared tests were used to compare categorical characteristics, while ANOVA or Kruskal–Wallis tests, as appropriate, were used to compare continuous metabolic and sleep characteristics among the groups.

To further assess the impact of OSA upon insulin resistance, we compared sleep measures among children at extremes of insulin sensitivity. Children with intermediate insulin sensitivity were excluded from this stage of the analysis in order to see whether more subtle differences in sleep architecture and sleep disturbances between children with greatest and lowest insulin sensitivity would emerge. HOMA-IR quartiles were calculated for the entire study population and children in the lowest and highest quartiles (quartile 1, HOMA-IR <0.91; quartile 4, HOMA-IR >2.33) within each AHI and BMI group were included in a further subanalysis. t-tests or Mann–Whitney U-tests were performed to compare differences in OSA elements, age and degree of obesity between children in the lowest versus highest HOMA-IR quartiles within each subgroup.