Consequences of obstructive sleep apnea on metabolic profile: A Population-Based Survey†
Disclosure: The authors declared no conflict of interest.
See the online ICMJE Conflict of Interest Forms for this article.
Epidemiologic studies that control for potential confounders are needed to assess the independent associations of obstructive sleep apnea (OSA) with metabolic abnormalities. The aim of our study was to evaluate the associations of OSA with metabolic abnormalities among the adult population of Sao Paulo, Brazil.
Design and Methods:
Questionnaires were applied face-to-face, full night polysomnography (PSG) was performed, and blood samples were collected in a population-based survey in Sao Paulo, Brazil, adopting a probabilistic three-stage cluster sample method. The metabolic profile included fasting glucose, insulin, and lipid levels. The hepatic insulin resistance index was assessed by the homeostasis model assessment-estimated insulin resistance (HOMAIR).
A total of 1,042 volunteers underwent PSG. Mild OSA and moderate to severe OSA comprised 21.2% and 16.7% of the population, respectively. Subjects with severe to moderate OSA were older, more obese, had higher fasting glucose, HOMAIR, and triglycerides (TG) levels than did the mild and non-OSA group (P < 0.001). Multivariate regression analyses showed that an apnea-hypopnea index (AHI) ≥15 and a time of oxy-hemoglobin saturation <90% were independently associated with impaired fasting glucose, elevated TG, and HOMAIR.
The results of this large cross-sectional epidemiological study showed that the associations of OSA and metabolic abnormalities were independent of other risk factors.
It has been reported that obstructive sleep apnea (OSA) is related to metabolic and cardiovascular consequences (1-4). The majority of apneic subjects also have abdominal obesity and in this context studies have suggested that OSA should be included as criteria for metabolic syndrome (5).
Clinical studies (6, 7) and community surveys (8-11) have demonstrated that OSA itself, independent of obesity, may lead to abnormalities in glucose homeostasis. The Sleep Heart Health Study, a large epidemiological cross-sectional study, consistently demonstrated an association between OSA and glucose intolerance (8). Recently, a prospective study showed that nocturnal intermittent hypoxia increased the risk for developing type II diabetes (12).
Regarding the relationship between OSA and lipid metabolism, some studies have reported associations between levels of cholesterol and triglycerides (TG) and the number of abnormal respiratory events during sleep (13-15).
Despite growing evidence that OSA is related to metabolic impairment, most large studies to date are limited by lack of polysomnography (PSG) data in close temporal association with measurement of metabolic parameters and did not study the whole population as well as including all ages and both genders. The primary aim of this study is to investigate the association between OSA and metabolic abnormalities in population survey. A secondary aim of this study was to determine the effect of hypoxemia on metabolic abnormalities.
Methods and Procedures
Study population and sampling procedures
The Sao Paulo Sleep Epidemiologic Study is a population-based cross-sectional study of sleep disturbances and their risk factors among adults living in Sao Paulo, the largest city in the Southern Hemisphere and the fifth largest metropolis in the world. There were 10,886,518 inhabitants living within a 1,524 km2 area in January 2008, corresponding to a population density of 7,233 inhabitants per km2.
Participants between 20 and 80 years of age underwent baseline examinations from July to December 2007. The design, survey methods, and laboratory techniques have been described elsewhere (16). Briefly, our single-centre study had a sample of 1,101 individuals living in Sao Paulo. This sample size allowed us to calculate prevalence estimates with a three percent precision error rate (17). To obtain a representative sample of the inhabitants of Sao Paulo, we used a three-stage cluster sampling technique with unequal selection probability (18), generating sample weights for each participant for all of the representative estimates. This is the same conceptual framework that was used in the North American National Health Surveys (19). A total of 1,042 volunteers agreed to undergo a PSG at the Sleep Institute with a very low refusal rate of 5.4%. Age (P = 0.11), gender (P = 0.55), and socioeconomic status (P = 0.38) distributions were not significantly different between the volunteers who agreed to perform the PSG recording and those who refused. The study protocol was approved by the Ethics Committee for Research at the Universidade Federal de Sao Paulo (CEP 0593/06) and was registered with ClinicalTrials.gov (Identifier NCT00596713).
Information pertaining to demographic factors, medical history, medication use, and personal health habits were collected by trained interviewers. At the Sleep Institute, questionnaires were administered and the PSG was conducted. Participants went to bed at their usual bedtime. General physical measurements (i.e., body weight (kg), height (m), neck circumference, waist and hips (cm)) were taken immediately before the PSG hook-up; following recommended procedures and using calibrated instruments. The standard overweight and obesity cut-off points were used: a body mass index (BMI) ≥25 kg/m2 and ≥30 kg/m2, respectively. Abdominal obesity was defined as a waist circumference ≥88 cm for women and ≥102 cm for men (20). Blood samples were taken at 8 AM, following a 12-h overnight fast, to obtain plasma glucose, plasma insulin, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol (HDL), and TG measurements.
Dyslipidemia, fasting glucose impairment, insulin resistance, and type 2 diabetes
Fasting plasma glucose was measured using the glucose oxidase method, whereas fasting plasma insulin was assessed using radioimmunoassay. Total cholesterol, low-density lipoprotein cholesterol, HDL, and TG levels were assessed using an auto analyser and the appropriate reagents.
Impaired fasting glucose was defined as positive when the glucose serum value was ≥100 mg/dl in men and women. The criteria for dyslipidemia included HDL levels <40 mg/dl in men and <50 mg/dl in women or TG levels ≥150 mg/dl in men and women (20), as well as the use of antilipaemic medications. Type 2 diabetes was defined as a fasting glucose level ≥126 mg/dl, use of antidiabetic medications or a previous diagnosis of diabetes. The hepatic insulin resistance index was assessed using the homeostasis model assessment-estimated insulin resistance (HOMAIR) and calculated as fasting serum insulin (µU/ml) × fasting plasma glucose (mmol/l)/22.5 (21). We considered the cut-off point for the presence of clinically significant insulin resistance (HOMAIR) to be values >2.71 mmol·µU·mL2 as previously reported for the general Brazilian population (22).
A full-night PSG was performed using a digital system (EMBLA S7000; Embla Systems, Broomfield, CO) at the sleep laboratory during the subject's habitual sleep time. The following physiological variables were monitored simultaneously and continuously: electroencephalogram, electrooculogram, surface electromyogram (submentonian region, anterior tibialis muscle, masseter region, and seventh intercostal space), electrocardiogram, airflow detection via thermocouple and nasal pressure, thoracic and abdominal respiratory effort using inductance plethysmography, snoring, body position, oxy-haemoglobin saturation (SpO2), and pulse rate. Four trained technicians visually scored all of the PSG data according to standardized criteria for investigating sleep (23). Electroencephalogram arousals, sleep-related respiratory events, and leg movements were scored in accordance with the criteria outlined in the American Academy of Sleep Medicine (AASM) Manual for Scoring Sleep and Associated Events (24). Apneas were scored and classified following the recommended respiratory rules for adults: a drop in the airflow amplitude by >90% of baseline lasting ≥10 s. Hypopneas were scored according to alternative rules: a 50% reduction in the airflow amplitude lasting ≥10 s, followed by a decrease of ≥3% in SpO2 or an electroencephalogram arousal. The average number of episodes of apnea and hypopnea per hour of sleep (apnoea-hypopnea index (AHI)) was calculated. Hypoxia was defined as the presence of SpO2 <90% expressed in minutes during the sleep time recorded by the PSG. AHI was categorized into non-OSA (<5/h), mild OSA (5-14.9/h), and moderate to severe OSA (≥15/h).
Total sleep time was considered the time during the night scored as sleep stages of NREM and REM stages.
Categorical variables were expressed in proportions and analyzed by a Pearson's χ2-test. Continuous variables were expressed as the mean ± SD and analyzed by a t test or an analysis of variance, when appropriate. We applied logistic regression models adjusted for age, abdominal obesity, total sleep time (total sleep time), and gender to determine the independent association between metabolic alterations (i.e., dyslipidemia, impairment fasting glucose, and insulin resistance) and the severity of sleep apnoea or hypoxemia. The results were considered statistically significant if the P value was <0.05. All data were analyzed using the statistical software STATA 10.
The characteristics of the subjects according to OSA severity are listed in Table 1. A total of 1,042 subjects were enrolled in the Sao Paulo Epidemiologic Sleep Study. 396 participants (38% of the sampled population) had OSA. 221 participants (21.2%) had mild OSA, and 175 (16.7%) had moderate-to-severe OSA. Subjects with OSA were significantly older, more obese, and had greater mean values for waist and neck circumferences as well as for lipid levels than subjects without OSA (P < 0.05).
Table 1. Characteristics of the probabilistic sample (n = 1,042) of inhabitants from Sao Paulo according to the severity of obstructive sleep apnea
The mean levels of HOMAIR and glucose were higher in subjects with mild OSA compared to subjects without OSA, and further increased in subjects with moderate-severe OSA. The percentage of subjects who met the criteria for dyslipidemia and type 2 diabetes were higher in the OSA group (P < 0.05).
After adjusting for gender, abdominal obesity, age, and total sleep time in a logistic regression model, the severity of OSA had a significant effect on impaired fasting glucose levels and HOMA levels and a marginally significant effect on elevated TG but not on HDL levels (Table 2). Subjects with moderate to severe OSA were 1.67 times more likely to have impaired fasting glucose or type 2 diabetes than those without OSA. Subjects with moderate to severe OSA had a two-fold risk of having insulin resistance than subjects without OSA.
Table 2. Multivariable-adjusted associations of sleep apnea severity with the presence of dyslipidemia, impaired fasting glucose or type 2 diabetes and insulin resistance (HOMA≥2.7) in a population-based survey in Sao Paulo (n = 1,042)
The presence of hypoxemia (expressed as the number of minutes with a SpO2 <90% different than zero at any time) was entered in second multiple logistic regression model adjusted for gender, abdominal obesity, age, and total sleep time to assess for the presence or absence of metabolic alterations. This analysis revealed that hypoxemia was independently associated with impaired glucose metabolism, insulin resistance, and TG levels but not with altered HDL. Subjects with hypoxemia were 1.52 times more likely to have altered glucose metabolism, 1.86 times more likely to have insulin resistance, and 1.51 times more likely to have elevated TG levels than those without hypoxemia (Table 3).
Table 3. Multivariable-adjusted associations of hypoxia with the presence of impaired fasting glucose or type 2 diabetes, dyslipidemia, and insulin resistance (HOMA >2.7) in a population-based survey in Sao Paulo city (n = 1,042)
This study demonstrates that OSA associated with altered glucose and lipid metabolism independent of the effects of obesity, gender, age, and total sleep time. The most striking result is the association of OSA severity and hypoxia with increased metabolic abnormalities, independently of gender, obesity, and age. Advantages of our study were data from a large sample of adults (>1,000 subjects) of both genders with wide age range (20-80 years), very low refusal rate (5.4%) and full PSG conducted with close temporal determination of metabolic profile in all participants.
These results support those of clinical studies (6, 7) as well as those of the Sleep Heart Health Study (SHHS) (8, 9), which also detected that OSA is associated with an increased risk of glucose impairment after adjusting for age, gender, smoking status, waist circumference, and self-reported sleep duration. They also used PSG to characterize disordered breathing during sleep. In contrast to SHHS, which conducted the PSG 1 year after the metabolic assessment, we conducted the PSG at the same time as the metabolic assessment, allowing for a more reliable assessment of the relationship between abnormal respiratory events and metabolic dysfunction.
Abdominal obesity is the most important causal factor for metabolic diseases (e.g., type 2 diabetes, dyslipidemia) as well as for OSA (20). In our sample a higher proportion of individuals were overweight or obese (59.9%) and the association with OSAS was 2.6-fold higher among overweight subjects and 10.5-fold higher for obese subjects. (25). However several studies have demonstrated that OSA also may have a role in the pathogenesis of metabolic diseases independent of obesity and (8, 9, 10, 11, 12, 13, 14, 15) these association are most likely complex and bidirectional.
With regard to the lipid profile, several lines of evidence have suggested an association between OSA and dyslipidemia. The Sleep Heart Health Study (26) studied 6,440 men and women over the age of 40 years and found that there was an inverse relationship between AHI and HDL cholesterol levels and a positive association between AHI and TG in younger men and women after adjusting for confounding factors; this was not the case for subjects 65 years of age and older. In contrast, another study that evaluated 255 Chinese adults between 30 and 60 years old did not find an association between OSA and HDL or TG levels, after controlling for confounding variables (10). We found that only hypertriglyceridemia was marginally independently associated with moderate to severe OSA, after controlling for well known confounders; however, this association was not found for low levels of HDL. Furthermore, the presence of intermittent hypoxemia (expressed as the number of minutes with a SpO2 <90% different than zero at any time) affected only elevated TG levels independent of gender, abdominal obesity, age, and total sleep time. This finding suggests that hypoxemia may be a better marker than AHI for determining these lipid parameter abnormalities.
The discrepancies reported for the association between the lipid profile and OSA in previous studies could be due to the use of different selection criteria in each local community, which resulted in cohorts, composed of subjects of differing gender, ethnicity, and age distribution. Moreover, it is known that subjects with insulin resistance and abdominal obesity have a characteristic dyslipidemia, with elevated levels of TG and low HDL levels (20).
In previous studies the relationship between OSA and dyslipidemia was controlled for BMI or percentage of body fat and not abdominal obesity. In contrast, our study was controlled for abdominal obesity instead of BMI and we found an effect of OSA on TG. Using abdominal obesity is better than BMI because it takes into account the known association between visceral fat and TG.
OSA mechanisms related to glucose metabolism impairment includes hypoxemic stress, recurrent arousals, sleep fragmentation, and sleep loss (4, 27). Experimental studies have shown that short and long-term exposure to intermittent hypoxia results in increased insulin resistance and upregulation of lipid biosynthesis (28, 29). Chronic intermittent hypoxia has been shown to worsen glucose tolerance and increase sympathetic nervous system activity in humans and mice exposed to intermittent hypoxia (30, 31). As indicated in metabolic studies, intermittent hypoxemia can independently contribute to the development of insulin resistance and type 2 diabetes in subjects with OSA through activation of the sympathetic nervous system, hypothalamus-pituitary-adrenal axis and a proinflammatory state. These alterations have already been demonstrated in OSA subjects (32, 33, 34). With regard to lipid metabolism, some studies have indicated that intermittent hypoxemia may lead to lipid metabolism alterations at the transcriptional level. The excessive production of reactive oxygen species increases the activity of the hypoxia inducible factor 1 (HIF-1) and nuclear factor κB. The activation of these transcriptional factors affects glucose metabolism, lipid biosynthesis in the liver, and inflammation (35, 36). Recent report provided new insights in possible mechanisms by which Intermittent Hypoxia also affects lipid metabolism accelerating atherosclerosis. It may induce dyslipidemia by upregulating lipid biosynthesis in the liver, increasing (ES) adipose tissue lipolysis with subsequent free fatty acid flux to the liver, and inhibiting lipoprotein clearance (37).
Therapy with nCPAP (nasal Continuous positive airway pressure) is the most effective treatment for sleep apnoea, preventing recurrent occlusion of upper airway and hypoxemia during sleep. However, randomized controlled trials reporting the role of CPAP in reducing insulin resistance are controversial (38, 39). This could be explained by different methods used in the evaluation of glucose metabolism, small sample size, CPAP compliance, and the short duration period of OSAS treatment and the presence of sleepy or not sleepy OSA subjects. Further randomized CPAP trials of longer duration are needed before a causal relationship between OSA and insulin resistance/glucose intolerance can be supported.
There are some limitations that should be considered when interpreting our results. One must keep in mind that because of the cross-sectional design of the study, the direction of causality cannot be inferred from the analyses. We also did not evaluate ethnicity because the majority of the Brazilian population is of a mixed race (40). As such, we cannot extrapolate our results to other ethnicities. In addition, medications, nutritional habits, and exercise status may have influenced the subjects' glucose and lipid profile, and these confounding factors were not considered in the analyses. Total sleep time was evaluated by the one night of register during PSG that could not reflect the amount of sleep deprivation, that is reported related to affect glucose impairment.
In conclusion, our findings demonstrated that the severity of OSA and hypoxemia are independently associated with insulin resistance, glucose and lipid metabolism abnormalities. Therefore, health professionals involved in the prevention and treatment of OSA should be attentive to the presence of metabolic dysfunction to properly plan and implement health interventions.
This study was supported by grants from the Associaçao Fundo de Incentivo a Psicofarmacologia (AFIP) and FAPESP (#07/50525-1 to R.S.-S., and CEPID no. 98/14303-3 to ST). J.A.T., L.R.A.B., and S.T. received CNPq fellowships.