INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: obstructive sleep apnea; independently associated; insulin resistance

Obstructive sleep apnea (OSA) is associated with increased cardiovascular and cerebrovascular morbidity (1). It is also recognized that many subjects with OSA have central obesity and other features of the metabolic syndrome (4), which is most widely accepted as being comprised of hyperinsulinemia, glucose intolerance, dyslipidemia, central obesity, and hypertension (7). These factors in the metabolic syndrome, also known as the "insulin resistance syndrome," have been established as independent risk factors for vascular disease (7). Hence there is ongoing controversy regarding the causal versus comorbid relationship between OSA and cardiovascular disease.

It is postulated that the cerebral activation and increased sympathetic output related to sleep-disordered breathing may provide a stress stimulus that triggers or aggravates some of these vascular risk factors, and thus confers independent predisposition to vascular pathogenicity. Notably, recent epidemiological studies provide strong evidence that OSA itself confers independent risks for the development of hypertension (10).

Insulin resistance, as indicated by an impaired biological response to insulin and hence a reduced insulin-mediated glucose disposal (13), has been implicated in the pathogenesis of the metabolic syndrome (7). Furthermore, there is evidence that insulin resistance predisposes to cardiovascular risk (7, 14). It is therefore important to determine if sleep apnea has any effect on insulin resistance.

Previous studies on the relationship between insulin resistance and OSA have yielded conflicting results (17). To test the hypothesis that sleep-disordered breathing is an independent risk factor for insulin resistance, we examined a cross-sectional cohort of subjects with a range of sleep-disordered breathing from absent to severe and analyzed the relationship between sleep-disordered breathing and insulin resistance.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Consecutive subjects admitted to the Sleep Laboratory at the University Department of Medicine, Queen Mary Hospital, for overnight sleep studies in March 1999 to February 2000, either as part of a community-based prevalence study (23) or because of a clinical referral for suspected sleep apnea, were recruited. Exclusion criteria were subjects with known diabetes mellitus on medications, acromegaly, chronic renal failure, on systemic steroid treatment, and on hormonal replacement therapy. A questionnaire on demographics, sleep symptoms, medical history, and medications was completed. Body habitus was measured in light clothing and bare feet using standard anthropometric methods (24). Waist and hip circumferences were measured to the nearest 0.5 cm. Waist circumference was measured midway between the lower costal margin and iliac crest, and the hip circumference as the maximal girth at the greater trochanters. On the morning after the sleep study, blood pressure was taken on waking at 7-8 A.M. in the supine position using Dinamap (Critikon Inc, Florida). Venous blood was then obtained in the fasting state for the measurement of glucose and insulin. All subjects gave written informed consent to blood taking. The study was approved by the Institutional Ethics Committee.

Polysomnogram

The polysomnography (PSG) (Alice 3 System; Healthdyne, Atlanta, GA) consisted of continuous polygraphic recording from surface leads for electroencephalography, electrooculography, electromyography, electrocardiography, thermistors for nasal and oral airflow, thoracic and abdominal impedance belts for respiratory effort, pulse oximeter for oxyhemoglobin level, tracheal microphone for snoring, and sensors for leg and sleep position. PSG records were scored manually. Sleep data were scored according to standard criteria (25). Arousals were scored according to established criteria (26). Respiratory events were scored according to AASM criteria (27): apnea was defined as complete cessation of airflow lasting 10 s or more; hypopnea was defined as either a  50% reduction in airflow for 10 s or more or a less than 50% but discernible reduction in airflow accompanied either by a decrease in oxyhemoglobin saturation of > 3% or an arousal. The average number of episodes of apnea and hypopnea per hour of sleep (the apnea-hypopnea index, AHI) was calculated as the summary measurement of sleep-disordered breathing.

Plasma glucose was measured by the glucose oxidase method on a Beckman autoanalyzer (Beckman Instruments, Bream, CA). Serum insulin was determined with a microparticle enzyme immunoassay on an Abbott IMx system (Abbott, Abbott Park, IL), using a monoclonal mouse anti-human insulin antibody. Intra- and interassay coefficients of variation were < 4% (28).

The estimation of insulin resistance by the homeostasis model assessment method (HOMA-IR) as previously described (29) was calculated by the following formula: fasting serum insulin (µU/ml) × fasting plasma glucose (mmol/L)/22.5.

Statistical Analysis

All clinical parameters were summarized by descriptive statistics. Comparisons of continuous clinical parameters between subjects with different categorization by AHI were made by Mann-Whitney test, one-way ANOVA, and Kruskal-Wallis H test, and categorical parameters were compared by Chi-square test. Spearman's rank correlation coefficient was used to examine the association of two parameters.

Severity of OSA was measured by AHI, minimum oxygen saturation (Sa_O2), time duration with Sa_O2 < 90%, and arousal index. To determine if they were individually associated with insulin resistance independent of obesity and central obesity, stepwise multiple regression was used with either fasting insulin or HOMA-IR as the dependent variable. The corresponding set of independent variables included obesity (BMI), central obesity (either waist/hip ratio [WHR] or waist circumference), OSA (either AHI, minimum Sa_O2, time duration with Sa_O2 < 90%, or arousal index), sex, age, and smoking (never smoker/ chronic or ex-smoker). Fasting insulin and HOMA-IR were logarithmic transformed before they were used as the dependent variable and deleted studentized residuals of all regression models were examined for the validity of model assumptions.

To determine whether a significant association of sleep parameters with insulin resistance was similar in obese and nonobese subjects, subjects were categorized into obese and nonobese groups based on the WHO recommendation for Asians, i.e., BMI < 25 and BMI  25 (30). Regression analyses were then pursued with the interaction between this categoric variable and AHI, the corresponding main effects, and other identified significant confounders of insulin resistance considered as independent variables.

Apart from considering sex as a covariate in the multiple linear regression for insulin resistance, the influence of sex on the association between insulin resistance and AHI/BMI was further examined by considering their interactions (insulin levels or HOMA-IR and AHI or BMI) in the regression analysis.

The clinical importance of insulin resistance in this cohort was evaluated by performing a stepwise logistic regression of blood pressure on fasting insulin or HOMA-IR, and other confounding variables including BMI, WHR, age, smoking, and sex. Hypertension was defined as either a known history of hypertension on drug treatment or blood pressure (BP) measurements of systolic BP  140 mm Hg or diastolic BP  90 mm Hg.

Two-tailed p values of less than 0.05 were considered to indicate significance.

Statistical analysis was performed by SPSS for Windows software (Version 10.0.7).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Table 1 summarizes the study characteristics of the 270 subjects. None of them had known diabetes mellitus on medications. Seventeen had hypertension on antihypertensive medications, and three had both hypertension and hyperlipidemia on medications.

Fasting insulin levels and HOMA-IR, unadjusted for any covariate, were significantly higher in the group with OSA, defined by AHI  5 (p < 0.001) (Table 1). Most of the sample characteristics, including the indicators of insulin resistance and its major determinants, obesity, central obesity, and male sex, were significantly different over the range of AHI stratum, with an apparent increase with increasing severity of sleep-disordered breathing (Table 2).

The relationship between AHI and fasting insulin or HOMA-IR was nonlinear (Figure 1). The fasting insulin and HOMA-IR values from one patient were found to be outlying from other observations, and data from this one subject have been removed before proceeding to regression analysis.

View larger version (12K): [in this window] [in a new window]   Figure 1.   Scatter plot of apnea-hypopnea index (AHI) and corresponding fasting insulin levels (µU/ml). Number of subjects = 270.  AHI < 5;  AHI  5.

In the regression models using fasting insulin level as the dependent variable, significant predictors were BMI, age, AHI, and minimum oxygen saturation (Table 3). In the models using HOMA-IR as the dependent variable, significant predictors were similarly BMI, AHI, and minimum oxygen saturation when WHR was the parameter for central obesity, whereas AHI was no longer significant when waist circumference was used as the parameter for central obesity (Table 4).

The  coefficients of AHI for insulin level and HOMA-IR were 0.005 (Tables 3 and 4). Because logarithmic transformation of insulin /HOMA-IR values have been used, a one unit increase in AHI would result in a 0.5% increase in fasting insulin levels or HOMA-IR values.

In the analysis for a differential effect of the association between OSA and insulin resistance in obese and nonobese subjects, the interaction term for obesity and AHI was not significant (Table 5), suggesting that the association was similar in both obese and nonobese subjects.

Analysis for influence of sex on the association between insulin resistance and AHI/BMI did not reveal any substantial effect of sex on the identified associations (p > 0.15 for all interaction terms).

In this cohort, hypertensive subjects had significantly higher insulin and HOMA-IR values compared with normotensive subjects (insulin: 15.1 ± 31.8 versus 7.7 ± 8.1 µU/ml, p < 0.001; HOMA-IR: 3.9 ± 7.7 versus 1.8 ± 2, p < 0.001). Multiple logistic regression showed that insulin resistance was a significant independent determinant of blood pressure status (Table 6).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The findings of this study suggested that sleep-disordered breathing has an independent adverse effect on insulin resistance. Results demonstrated that obesity was the major determinant of insulin resistance in this cohort, but despite controlling for obesity and other important confounding factors of insulin resistance, AHI and/or minimum oxygen saturation were significant determinants of fasting insulin level and HOMA-IR.

The relationship of insulin resistance and other metabolic variates is one of a complex interactive regulation. It is established that central obesity leads to insulin resistance via increased lipolysis and fatty acid availability (31, 32). Furthermore, it has been hypothesized that a stress reaction activating the hypothalamic-pituitary-adrenal axis leading to release of cortisol and other hormones may be a trigger for mechanisms generating insulin resistance and preferential abdominal fat accumulation (33) and may thus contribute to insulin resistance.

In the context of subjects with OSA, the strong association of sleep-disordered breathing and atherosclerotic vascular diseases has long been observed (1). Many subjects with OSA are known to have coexisting risk factors for cardiovascular and cerebrovascular diseases, in particular the factors that comprise the metabolic syndrome of central obesity, hypertension, dyslipidemia, and insulin resistance or glucose intolerance. However, OSA itself is characterized by increased nocturnal sympathetic output (34) and other pathophysiological mechanisms that may provide an independent predisposition to the generation of atherosclerosis, either directly or through aggravation of some of the risk factors such as hypertension (35, 36). Sympathetic activation also raises circulating free fatty acid via stimulation of lipolysis and promotes insulin resistance (37). Furthermore, these risk factors may also interact and develop a vicious cycle culminating in increased vascular risks. There is now substantial epidemiological evidence to support the finding that OSA has an independent adverse effect on hypertension (10).

The association between OSA and insulin resistance remains controversial. Previous studies have used different methodological approaches and target populations, and results have been conflicting. Our study has the strength of having a relatively large sample size with good representation of both subjects with and without OSA. Furthermore, the documentation of sleep-disordered breathing was based on full in-laboratory polysomnograms.

In the evaluation of insulin sensitivity (or its corollary, insulin resistance), the gold standard is the euglycemic clamp method (38). However, this method is invasive and labor intensive, thus hindering its application as a research tool when investigating a large number of subjects. Instead, we have used the absolute fasting insulin level as well as the computer-solved index, HOMA-IR. The fasting insulin levels in an individual is determined by both insulin secretion and insulin resistance, although the former is usually stable in nondiabetic subjects. It has been shown to be a useful guide to insulin resistance in normoglycemic individuals, although less so in subjects with established diabetes mellitus (13). The HOMA of insulin sensitivity was proposed as a simple and inexpensive alternative to more sophisticated techniques (29). The method derives an estimation of insulin sensitivity from mathematical modeling of fasting blood glucose and insulin concentrations and has been compared with several other methods of measuring insulin sensitivity, including different glucose clamp techniques and intravenous glucose tolerance tests (39). Results have indicated the close correlation between the degree of insulin resistance estimated by HOMA and these methods. Recently, HOMA has been validated against the gold standard method of euglycemic clamp in 115 subjects, and there was a strong correlation between clamp-measured total glucose disposal and HOMA-estimated insulin sensitivity, with a correlation coefficient of 0.820 (39). Furthermore, this close correlation did not show any substantial difference between men and women, young and older subjects, nonobese and obese subjects, nondiabetic and diabetic subjects, and normotensive and hypertensive subjects (39). Due to its simplicity, HOMA has been used in large clinical or epidemiological studies (41).

In a previous study that utilized the euglycemic clamp to investigate insulin resistance in a group of 50 healthy subjects of whom one-third had an AHI  10 on a sleep study with a portable device MESAM-IV, no correlation was identified between insulin resistance and sleep-disordered breathing after adjusting for BMI (18). The conclusion was that insulin resistance preceded rather than followed sleep-disordered breathing. However, the small number of subjects with sleep apnea in the cohort and the use of limited polysomnographic assessment could limit the ability to detect any independent effect of sleep-disordered breathing on insulin resistance.

On the other hand, in a study involving 261 men in which the relative contributions of body weight and sleep apnea to blood pressure, fasting insulin, and fasting glucose were examined, there was evidence for an independent association between sleep apnea and fasting insulin levels in those with a BMI > 29 (17).

More recently, a study on cytokines, insulin resistance, and visceral obesity in OSA indicated that mean plasma insulin levels in 14 obese subjects with OSA were significantly higher than BMI-matched controls with no OSA, suggesting that sleep-disordered breathing was an independent risk factor for hyperinsulinemia (22). The same group of workers also studied women with polycystic ovary syndrome compared with premenopausal controls, and insulin resistance was shown to have a stronger association with sleep-disordered breathing than with BMI or testosterone, supporting a close independent link between insulin resistance and OSA in these subjects (45).

The association of diabetes mellitus and OSA has been evaluated in a sample of 116 age-stratified men with hypertension selected from subjects in a population-based study in Sweden. It was shown that although obesity was the main risk factor for diabetes mellitus, coexistent severe OSA may add to the risk independently (46).

Studies looking at the effect of treatment of OSA on insulin resistance also showed conflicting results. In a study that compared fasting insulin levels in 15 subjects with OSA with body habitus-matched controls, before and after 3 mo of continuous positive airway pressure (CPAP) treatment in the OSA group, no difference was detected in either comparison (20). Similarly, in a recent report on leptin and visceral fat in OSA, no significant change was seen in insulin levels in 12 subjects treated with CPAP (47). However, in another study, CPAP treatment of 10 subjects with diabetes and OSA resulted in a reduction of fasting insulin despite maintenance of BMI (19). The independent effect of OSA on insulin resistance, if any, is anticipated to be of modest magnitude only, and may not be easy to detect. The compliance of the use of CPAP in these studies was probably based on self-reporting, which made it difficult to evaluate the "true effectiveness" of CPAP. Furthermore, OSA is a chronic disorder and metabolic changes such as fat redistribution to visceral component may have occurred (47), rendering treatment of OSA, especially in the short term, apparently ineffective in modifying insulin sensitivity.

The major confounding factor in analysis of insulin resistance in OSA is obesity. Measures of obesity including BMI, waist circumference, and WHR have been demonstrated to correlate closely with insulin resistance, and central or abdominal obesity has been shown to be more predictive of insulin resistance than general obesity (48, 49). The regression models indicated that insulin resistance in this cohort was highly dependent on BMI. Nevertheless, AHI and minimum oxygen saturation emerged consistently as independent determinants of insulin resistance, although their effects were much smaller than that of obesity. In the HOMA regression models in which waist circumference was considered, only BMI, waist circumference, and minimum oxygen saturation were significant determinants of insulin resistance and not AHI. It is known that waist circumference is closely correlated with AHI and this may render it very difficult to see any effect of AHI on insulin resistance (50).

In this cohort, age was also noted to have a negative correlation with fasting insulin levels, which would not be expected of a general population in which age per se marginally increases insulin resistance. However, insulin levels are also affected by insulin secretion and the reduction with age may be more obvious in those with increased demand on -cell reserve such as these subjects with OSA. This may explain the inverse relationship seen between age and insulin level but not HOMA-IR, which is a more reliable index of insulin resistance.

In a previous study, it has been reported that the effect of OSA on insulin resistance was seen only in obese subjects but not in nonobese subjects with OSA (17). In our subjects, there was no significant effect of interaction between AHI and BMI, suggesting that this association between sleep-disordered breathing and insulin resistance was present in both obese and nonobese subjects.

Premenopausal women are more insulin sensitive then men because women have less visceral fat despite a higher total body fat mass, whereas postmenopausal women have insulin sensitivity similar to men (51). Sex has been considered as one of the covariates, and no independent effect on insulin resistance was seen. There was also no evidence of a different susceptibility to insulin resistance in males and females with the same degree of obesity and sleep-disordered breathing, although the sample sizes of the two sexes may not be large enough to exclude this definitively.

Although our findings support an independent association between sleep-disordered breathing and insulin resistance, the clinical relevance of this modest magnitude of increased insulin resistance due to OSA is difficult to define. Because insulin resistance is a known risk factor for hypertension, we analyzed the relationship between blood pressure and insulin resistance in this cohort. Results showed that insulin resistance was an independent determinant of hypertension despite controlling for major confounding variables of obesity and age. No direct quantitative interpretation can be given to these findings, but it would be reasonable to surmise that severe OSA with a very high AHI or very low minimum oxygen saturation would not be devoid of adverse clinical relevance.

Hence, our results suggest that sleep-disordered breathing is independently associated with insulin resistance. The role of increased insulin resistance as one of the intermediary mechanisms by which sleep apnea predisposes to vascular pathogenicity is worthy of further exploration.