Pharyngeal collapsibility during sleep is elevated in insulin-resistant females with morbid obesity

Abstract

Insulin resistance is associated with sleep apnoea, leading us to hypothesise that it is also associated with elevations in pharyngeal collapsibility, even in the absence of sleep apnoea.

90 bariatric patients were characterised for sleep apnoea, pharyngeal collapsibility and insulin resistance. Patients with a respiratory disturbance index (RDI) >10 events·h⁻¹, diabetes mellitus, tonsillar hypertrophy and pulmonary disease were excluded. The remaining 14 females underwent collapsibility measurements (passive critical pressure, Pcrit_p) during non-rapid eye movement sleep. The homeostasis model assessment (HOMA) index, a measure of insulin resistance, was derived from measurements of fasting glucose and insulin levels, and compared to Pcrit_p.

Groups with high Pcrit_p compared to low Pcrit_p did not differ in age, body mass index or RDI. HOMA and insulin were elevated in the high Pcrit_p group compared to the low Pcrit_p group (p<0.02). Pcrit_p correlated with HOMA (Spearman’s ρ=0.565, 95% CI 0.104–0.862; p=0.035) and insulin (Spearman’s ρ=0.609 95% CI 0.196–0.835; p=0.021).

Obese insulin-resistant subjects without frank diabetes or sleep apnoea demonstrate preclinical elevations in pharyngeal collapsibility, which may increase their susceptibility to sleep apnoea. Our findings suggest that insulin resistance could play a significant role in sleep apnoea pathogenesis by generating requisite elevations in pharyngeal collapsibility.

Introduction

Obstructive sleep apnoea (OSA) is characterised by recurrent episodes of upper airway obstruction that result from increases in upper airway collapsibility during sleep [1]. Recent evidence suggests that increases in airway collapsibility are related to disturbances in pharyngeal structural and neuromuscular control [2]. Structural loads have been linked to obesity and central adiposity [3, 4] which are potent risk factors for OSA [5–7]. Nevertheless, mechanisms linking obesity with increased airway collapsibility remain poorly understood.

In addition, sleep apnoea and obesity have been linked to insulin resistance [8–12]. Insulin resistance has been associated with central adiposity and ectopic fat deposition in a broad array of skeletal muscles [13–16]. Similarly, sleep apnoea has been associated with central adiposity [17, 18] and excess fat deposits in upper airway structures including the tongue [19, 20], which can increase pharyngeal collapsibility and sleep apnoea susceptibility [19, 21, 22]. It is therefore possible that insulin resistance is associated with elevations in upper airway collapsibility in obesity, which could in turn predispose to OSA. However, in OSA patients, intermittent oxyhaemoglobin desaturations and sleep fragmentation can aggravate insulin resistance [23, 24] and thereby confound the relationship between insulin resistance and sleep apnoea susceptibility. Thus, it is not clear whether insulin resistance predisposes to sleep apnoea (by increasing upper airway collapsibility), or vice versa.

The major goal of the present study was to examine the link between insulin resistance and upper airway collapsibility. We hypothesised that insulin resistance is associated with elevations in pharyngeal collapsibility in obesity that is unrelated to concomitant OSA. To minimise potential confounding by sleep apnoea [11, 25], we examined this relationship in a select group of obese patients who were free of significant sleep apnoea by quantifying upper airway dysfunction with measurements of passive critical pressure (Pcrit_p) during sleep.

Methods

Participants

A cross-sectional study was conducted in 90 patients presenting to the Johns Hopkins Bayview Medical Center bariatric surgery clinic (Baltimore, MD, USA) between 2005 and 2012. Inclusion criteria included age >21 years and a body mass index (BMI) >35 kg·m⁻². All volunteers were screened for sleep apnoea and were included if they had an overall respiratory disturbance index (RDI) of <10 events·h⁻¹ (n=29). The remaining participants were excluded if they had asthma (diagnosed using pulmonary function tests and/or current use of bronchodilators; n=3), diabetes mellitus (defined by prior clinical diagnosis, use of hypoglycaemic agents or documented fasting glucose >126 mg·dL⁻¹; n=2) or palatine tonsillar hypertrophy (grade ⩾II tonsils; n=5) [26]. One patient met two exclusion criteria (palatine tonsillar hypertrophy and diabetes mellitus). None of the remaining subjects were pregnant or had unstable cardiovascular disease, documented prior use of nasal continuous positive airway pressure (CPAP) within the past 3 months, moderate or severe chronic obstructive pulmonary disease (defined by Global Initiative for Chronic Obstructive Lung Disease criteria [27–29]), current use of systemic steroids, liver disease, history of HIV disease or use of home supplemental oxygen. Of the remaining 20 participants, Pcrit_p was successfully determined in 14 subjects [30], who comprised the final study group (figure 1). Those with and without Pcrit_p measurements did not differ in age (mean±SD 39.2±8.7 years versus 39.8±8.8 years, p=0.743) or BMI (mean±SD 47.5±4.3 kg·m⁻² versus 43.6±4 kg·m⁻², p=0.094). Informed consent was obtained for a protocol that was approved by the Johns Hopkins Medical Institutions institutional review board.

FIGURE 1.

Flow chart illustrating how the final study sample was derived from bariatric patients, after excluding those with sleep apnoea, other medical conditions and loss of nocturnal data.

Study design

The relationship between measures of insulin resistance (fasting glucose and insulin, and the homeostasis model assessment (HOMA) index) was examined in the final patient sample. We also accounted for potential confounds of these relationships using measures of body anthropometry and sleep apnoea severity (RDI) in these patients.

Study procedures and protocols

Anthropometrics

Measurements of obesity and regional adiposity were performed, including weight, height and neck, waist and hip circumference, BMI and waist-to-hip ratio.

Baseline assessment of sleep and breathing patterns

A full-night sleep study was performed from 22:00 h to 06:00 h to characterise sleep and breathing patterns, as previously described [31]. In brief, digitised signals included left and right electro-oculogram, submental electromyogram, electroencephalogram, arterial oxygen saturation, nasal pressure, chest and abdominal plethysmography and video monitoring to determine body position throughout the night. Patients were instructed to sleep in the supine position whenever possible. Sleep staging, sleep disordered breathing episodes and arousals were scored according to standard criteria [31]. Hypopnoea was defined as a ⩾30% decrease in airflow for ⩾10 s that was accompanied by either an oxyhaemoglobin desaturation of 4% or arousal. The RDI was quantified for the entire night, and separately for rapid eye movement (REM) and non-REM (NREM) sleep. A fasting blood sample was drawn at 07:00 h after the baseline sleep study.

Assessing upper airway mechanical loads (Pcrit_p)

Within a month of the initial sleep study, subjects underwent another sleep study to assess mechanical loads from the upper airway pressure–flow relationship in NREM sleep using a validated protocol [30]. In brief, nasal pressure was elevated initially to abolish airflow obstruction completely during NREM sleep. The nasal pressure was then lowered abruptly for five breaths to characterise the level of maximal inspiratory airflow across a range of nasal pressures encompassing the pressure at which airflow fell to zero. Strict criteria were implemented to determine whether sufficient data were obtained to generate an accurate upper airway pressure–flow relationship with minimal extrapolation [30]. In 14 out of 20 eligible patients, these relationships were deemed adequate to derive the Pcrit_p as the nasal pressure at zero flow, a measure of pharyngeal mechanical loads [2, 30, 32].

Blood work

Fasting morning serum was analysed using ELISA to measure insulin (Linco Research Inc., St Charles, MO, USA) and glucose (in the Johns Hopkins Bayview Medical Center clinical laboratory). The HOMA index was calculated as fasting serum insulin (μU·mL⁻¹) × fasting blood glucose (mmol·L⁻¹)/22.5 [33]. The HOMA-IR provides a validated index of insulin resistance against euglycaemic hyperinsulinaemic clamp studies [34] in a broad range of research subjects.

Data analysis

All anthropometric, demographic and sleep study variables are presented as median (interquartile range) (table 1). Our analyses were designed to characterise associations between measures of insulin resistance and its major predictor, Pcrit_p, which was treated as both a dichotomous and continuous variable. Dichotomous comparisons were based on whether Pcrit_p fell above or below the median value of −1.98 cmH₂O for the entire group. Nonparametric Mann–Whitney rank sum tests were used to compare parameters between groups with high and low Pcrit_p. Continuous comparisons of Pcrit_p with metabolic parameters were also examined using Spearman’s ρ (rank correlation coefficient). Significance was inferred for p<0.05. Analyses were conducted using SPSS (version 22; IBM, Armonk, NY, USA).

TABLE 1.

Demographic, anthropometries and steep disordered breathing parameters

	Entire group	Low Pcritp	High Pcritp	p-value
Subjects	14	7	7
Age years	37.2 (34.0-43.2)	42.3 (34.0-48.4)	34.9 (34.0-43.2)	0.710
Anthropometries
Weight kg	127.8 (122.3-134.2)	125.9 (122.3-128.3)	132.7 (1 19.6-136.9)	0.456
BMI kg·m−2	48.6 (45.9-50.1)	49.1 (44.0-51.3)	48.1 (47.6-50.1)	1
Neck circumference cm	39.7 (36.8-41.3)	39.8 (37.0-42.2)	39.5 (36.0-41.3)	0.710
Waist circumference# cm	126.8 (120.8-134.8)	129.7 (120.8-136.5)	126.8 (1 12.0-132.7)	0.534
Hip circumference# cm	143.8 (136.5-145.8)	140.2 (135.0-145.5)	144.3 (141.0-148.7)	0.628
Waist/hip ratio#	0.91 (0.86-0.94)	0.94 (0.91-0.96)	0.87 (0.85-0.92)	0.138
Respiratory disturbance index events-h−1
NREM	3.1 (1.6-4.5)	2.3 (0.7-4.5)	3.3 (3.0-5.2)	0.165
REM	10.6 (7.5-22.9)	22.9 (2.3-43.8)	10.5 (7.5-10.6)	0.259
Total¶	4.4 (2.5-6.6)	4.2 (2.3-9.5)	4.5 (3.9-6.5)	1
Oxygen desaturation index (4% SpO2 drop)+ events·h−1
NREM	1.7 (0.63-2.23)	1.1 (0.44-2.51)	1.9 (1.84-1.95)	0.291
REM	1.8 (0.63-3.46)	2.6 (0.54-5.94)	1.2 (0.72-1.98)	0.291
Total¶	3.8 (2.03-5.50)	3.8 (0.73-7.56)	3.8 (2.58-3.95)	0.808
Peripheral capillary oxyhaemoglobin saturation %
Baseline SpO2
NREM	96.0 (94.8-96.8)	95.6 (94.5-96.2)	96.6 (94.8-97.7)	0.209
REM	95.8 (95.3-96.6)	95.6 (95.1-95.8)	96.5 (95.8-96.7)	0.128
Total¶	95.8 (95.2-96.9)	95.7 (95.2-95.8)	96.6 (95.1-97.1)	0.259
Average low SpO2
NREM	91.6 (90.2-93.0)	91.0 (90.2-91.6)	92.9 (90.1-93.6)	0.366
REM	91.3 (90.5-91.8)	90.7 (88.8-91.3)	91.5 (90.5-92.7)	0.128
Total¶	91.4 (90.1-92.7)	90.8 (89.2-91.4)	92.5 (90.3-93.0)	0.209
Pcritp cmH2O	−1.98 (−5.22-−1.00)	−5.22 (−8.00-−3.01)	−1.00 (−1.55-1.30)	0.001

Data are presented as n or median (interquartile range], unless otherwise stated. Pcrit_p: passive critical pressure; BMI: body mass index; NREM: non-rapid eye movement (REM); SpO₂: peripheral capillary oxyhaemoglobin saturation.

^#: one patient had no waist and hip measurements;

^¶: NREM and REM values combined;

⁺: oxygen desaturation indices could not be calculated in two patients.

Results

Patient characteristics

Demographic and anthropometric characteristics are shown for the 14 patients in table 1 for the entire group and for those with high and low Pcrit_p. All subjects were severely obese females with a median BMI of 48.1 kg·m⁻². Four were African-American, nine were Caucasian, and one was Latino. Anthropometric and demographic parameters did not differ significantly between groups with high compared to low passive Pcrit_p. Specifically, these groups did not differ significantly in the degree of adiposity (BMI) or in measures of regional fat distribution (waist to hip ratio, and neck and waist circumferences), although contrary to expectations [21], those with low passive Pcrit_p were marginally, but not significantly, older (p=0.710), and had somewhat greater degrees of central adiposity (higher waist-to-hip ratio, p=0.138).

By design, patients with significant sleep apnoea (a total RDI ⩾10 events·h⁻¹) were excluded from the study. When comparing RDI between groups with low and high Pcrit_p (table 1), we found no significant differences in NREM RDI, REM RDI or total RDI. Furthermore, four patients in the low Pcrit_p group and five patients in the high Pcrit_p group had a total RDI of <5 events·h⁻¹. The baseline peripheral capillary oxyhaemoglobin saturation (SpO₂) and the average low SpO₂ during NREM, REM and total sleep were not significantly different between groups.

Upper airway function during sleep and metabolic disturbances

Comparing low and high Pcrit_p groups (figure 2), we found significantly increased insulin levels in those with higher compared to lower Pcrit_p (p=0.011), whereas no significant difference in fasting glucose was observed (p=0.902). Similarly, HOMA was elevated in the group with a high Pcrit_p compared to low Pcrit_p (p=0.017). Of note, no significant differences in anthropometric, demographic or sleep apnoea characteristics were found between groups with high and low HOMA values (table 2).

FIGURE 2.

Fasting a) insulin and b) glucose and c) homeostasis model assessment (HOMA) in groups with high and low passive critical pressure (Pcrit_p) during non-rapid eye movement sleep. Significant elevations in insulin and HOMA were observed in the high Pcrit_p group versus the low Pcrit_p group.

TABLE 2.

Passive critical pressure (Pcrit_p) correlations with indices of glucose homeostasis

	Low HOMA	High HOMA	p-value
Subjects	7	7
Age years	37.9 (34.6-42.3)	34.9 (29.6-50.2)	0.85
Anthropometrics
Weight kg	277 (263.0-281.7)	292 (271.5-301.2)	0.23
BMI kg·m−2	48.1 (44.0-49.1)	49.7 (47.6-51.6)	0.34
Neck circumference cm	40 (36.5-42.2)	39.2 (36.8-41.3)	0.57
Waist circumference# cm	129.5 (120.8-136.5)	126.8 (118.2-132.7)	0.67
Hip circumference# cm	141.2 (135.0-143.8)	145.8 (136.5-150)	0.15
Waist-to-hip ratio#	0.94 (0.88-0.96)	0.87 (0.85-0.92)	0.12
Respiratory disturbance index events·h−1
NREM	2.3 (0.7-3.3)	4.5 (3.0-5.2)	0.05
REM	7.7 (2.3-22.9)	10.6 (10.5-23.4)	0.28
Total¶	3.9 (2.3-4.5)	6.5 (4.0-9.5)	0.06
Peripheral capillary oxyhaemoglobin saturation %
Baseline SpO2
NREM	95.6 (94.5-97.7)	96.2 (94.8-96.6)	0.85
REM	95.8 (95.1-97.3)	95.8 (95.6-96.6)	0.95
Total¶	95.8 (95.2-97.1)	95.8 (95.1-96.6)	0.61
Average low SpO2
NREM	92.3 (90.2-94.6)	91.6 (90.1-92.9)	0.48
REM	91.3 (90.0-92.7)	91.3 (90.5-91.8)	0.85
Total¶	91.3 (90.1-94.2)	91.4 (89.8-92.5)	0.66

Data are presented as n or median (interquartile range), unless otherwise stated. HOMA: homeostasis model assessment; BMI: body mass index; NREM: non-rapid eye movement (REM); SpO₂: peripheral capillary oxyhaemoglobin saturation.

^#: one patient had no waist and hip measurements;

^¶: NREM and REM values combined.

In addition, we observed that Pcrit_p was positively correlated with HOMA score (Spearman’s ρ=0.565, 95% CI 0.104–0.862; p=0.035) and with insulin serum level (Spearman’s ρ=0.609, 95% CI 0.196–0.835; p=0.021) (table 3 and figure 3). In contrast, Pcrit_p did not correlate with fasting glucose level. Pcrit_p did not correlate significantly with age, neck circumference, weight or BMI.

TABLE 3.

Passive critical pressure (Pcrit_p) and homeostasis model assessment (HOMA) correlations with body anthropometrics and demographics

Parameters	Spearman’s ρ (95% CI)	p-value
Pcritp
Glucose homeostasis
Glucose	−0.145 (−0.720-0.584)	0.620
Insulin	0.609 (0.196-0.835)	0.021
HOMA	0.565 (0.104-0.862)	0.035
Anthropometrics/demographics
Age	0.068 (−0.665-0.673)	0.817
Weight	0.235 (−0.510-0.773)	0.418
BMI	−0.037 (−0.676-0.609)	0.899
Neck circumference	−0.134 (−0.668-0.451)	0.648
HOMA
Anthropometrics
Weight	0.385 (−0.192-0.862)	0.175
BMI	0.297 (−0.340-0.806)	0.303
Neck circumference	−0.103 (−0.805-0.646)	0.725
Waist circumference	−0.022 (−0.594-0.695)	0.943
Hip circumference	0.489 (0.021-0.731)	0.090
Waist-to-hip ratio	−0.440 (−0.835-0.362)	0.133

BMI: body mass index.

FIGURE 3.

Scatter plots of insulin resistance metrics versus passive critical pressure (Pcrit_p). a) Homeostasis model assessment (HOMA); b) insulin.

HOMA did not correlate significantly with BMI (Spearman’s ρ=0.297, 95% CI −0.340–0.806; p=0.303), weight (Spearman’s ρ=0.385, 95% CI −0.192–0.862; p=0.175), neck circumference (Spearman’s ρ=−0.103, 95% CI −0.805–0.646; p=0.725), waist (Spearman’s ρ=−0.022, 95% CI −0.594–0.695; p=0.943), hip (Spearman’s ρ=0.489, 95% CI 0.021–0.731; p=0.090) or waist-to-hip ratio (Spearman’s ρ=−0.440, 95% CI −0.835–0.362; p=0.133).

Discussion

The aim of our study was to analyse the association between insulin resistance and upper airway collapsibility in a well-defined sample of obese individuals without overt diabetes or sleep apnoea. Our major finding was that insulin resistance was greater in the group with high compared to low Pcrit_p, and correlated directly with Pcrit_p, independently of demographic and anthropometric characteristics as well as RDI. These findings suggest that insulin resistance could confer underlying defects in pharyngeal collapsibility during sleep and increase sleep apnoea susceptibility in obese individuals.

Obesity imposes mechanical loads on the airway that increase its collapsibility (Pcrit_p) during sleep [2, 35]. As airflow obstruction ensues, active neuromuscular responses are recruited that can mitigate the obstruction and prevent the development of OSA [2, 36]. Current evidence suggests that elevations in mechanical loads and reductions in neuromuscular responses are both required for sleep apnoea pathogenesis [2, 37]. Other factors, including ventilatory demand, controller gain and arousal threshold can also modulate sleep apnoea severity [38]. These findings suggest that despite severe obesity, our subjects were protected from significant sleep apnoea for one of two reasons. In those with a relatively low Pcrit_p, structural stability of the pharynx prevented airway collapse during sleep. In those with a high Pcrit_p, compensatory neuromuscular responses would be required to offset mechanical loads and maintain airway patency during sleep. Thus, elevations in Pcrit_p place subjects at risk of OSA, which will ensue when neuromuscular responses are also lacking.

Two mechanisms can account for the observed associations between insulin resistance and upper airway collapsibility (figure 4). First, OSA and accompanying oxyhaemoglobin desaturations and arousals can exacerbate insulin resistance [23, 24, 39] (figure 4, arrow 1), which decreases with CPAP therapy.

FIGURE 4.

Putative pathogenic mechanisms leading to increased upper airway collapsibility and obstructive sleep apnoea (OSA) in obese insulin-resistant patients. OSA can worsen insulin resistance (arrow 1) and underlying defects in pharyngeal collapsibility (arrow 2), leading to further increases in severity of OSA (arrow 3).

Acute metabolic consequences of apnoeic activity can explain the high prevalence of sleep apnoea in type 2 diabetics [40]. However, even in the absence of sleep apnoea we found that insulin resistance was associated with underlying defects in upper airway mechanical function. This finding in a select group of obese nondiabetic, nonapnoeic subjects suggests that insulin resistance can predispose to upper airway obstruction during sleep (figure 4, arrow 2), thereby contributing to OSA pathogenesis. Its effect on upper airway collapsibility may be related to ectopic fat deposition, which has been demonstrated in heart [41], abdominal viscera [42] and skeletal muscle [13]. Ectopic fat accumulation in the tongue and peripharyngeal tissues in sleep apnoea patients [14, 19] can also account for observed elevations in upper airway collapsibility, and increase sleep apnoea susceptibility (figure 4, arrow 3).

In addition to ectopic fat deposits in upper airway structures, insulin resistance has been linked with central adiposity. Obesity and central adiposity have been associated with decreases in end-expiratory lung volume [43, 44], known to increase pharyngeal collapsibility during sleep [22]. Taken together, these findings suggest that in addition to obesity, insulin resistance may play a significant role in the pathogenesis of OSA by generating requisite elevations in pharyngeal collapsibility. Once sleep apnoea develops, insulin resistance can further increase upper airway collapsibility and perpetuate a vicious cycle of sleep apnoea progression.

Several limitations should be considered when reviewing our results. First, our inferences are based on a small sample of obese females, potentially limiting the generalizability of our findings. The exclusion of males from our sample was to be expected, given that patients with sleep apnoea were strictly excluded and that sleep apnoea is so highly prevalent in obese males. As noted, we sought to minimise comorbidities, including sleep apnoea, which could have confounded our assessment of metabolic determinants of upper airway function. Second, we acknowledge that we studied a convenience sample of patients presenting to a bariatric clinic, which may not accurately reflect characteristics of a community-based obese population. Nonetheless, our sample reflects the spectrum of obese patients, in whom differences in metabolic function may confer either protection or susceptibility to sleep apnoea. Third, although no anthropometric differences in obesity and regional adiposity were observed between groups, it is still possible that residual confounding by differences in body anthropometry exists, and/or that anthropometric differences would emerge with a greater sample size. It is also possible that imaging modalities, which offer greater precision in delineating regional fat, could have revealed subtle differences in fat content and distribution to account for variability in Pcrit_p. Fourth, we recognise that our study design was cross-sectional in nature, which precludes us from implicating causal mechanisms linking metabolic and upper airway dysfunction. Specifically, it is possible that visceral adiposity could lead to elevations in both HOMA-insulin resistance and Pcrit_p. If so, we would expect that these parameters would correlate with measures of regional adiposity, which was not the case. Fifth, it is possible that those with elevations in airway collapsibility (high Pcrit_p) were protected from developing sleep apnoea by active neuromuscular responses that preserved airway patency during sleep. Nevertheless, challenges in assessing these active responses experimentally precluded us from addressing this possibility. Sixth, we acknowledge that low levels of REM apnoea were present, which could account for differences in HOMA-insulin resistance between low and high Pcrit_p groups. Paradoxically, the amount of REM apnoea was greater in the low compared to the high Pcrit_p group, which would have only served to attenuate observed differences in HOMA-insulin resistance between these groups [11, 12, 39]. Finally, elevations in Pcrit_p could be related to disturbances in ventilatory control in insulin-resistant patients [45, 46], which could further reduce neuromuscular tone in the hypotonic (passive) pharyngeal airway. Nevertheless, Pcrit_p has been commonly used to integrate anatomical impediments to the maintenance of pharyngeal patency during sleep [38], and further work will be required to determine definitively whether insulin resistance confers further disturbances in pharyngeal neuromuscular control during sleep.

Our findings have important implications for clinical practice and research. First, the findings suggest that impaired glucose homeostasis, even in the absence of frank diabetes mellitus is a risk factor for OSA. They also imply that insulin resistance may play a causal role in sleep apnoea pathogenesis by generating requisite elevations in pharyngeal collapsibility. As sleep apnoea ensues, concomitant increases in insulin resistance could further increase upper airway collapsibility, perpetuating a vicious cycle of sleep apnoea progression (figure 4). Second, it is possible that measures which reduce insulin resistance could also lower pharyngeal collapsibility during sleep. Specifically, agents that target ectopic fat deposition in insulin-resistant patients may ultimately reverse defects in upper airway mechanical function and improve OSA.