Prevalence and impact of obstructive sleep apnea in type 2 diabetes mellitus: A descriptive cross-sectional study

Abstract

Background

Obstructive sleep apnea (OSA) has been reported to have a high prevalence in patients with type 2 diabetes mellitus. There is scarcity of literature on relationship between OSA and diabetes in Indian population.

Methods

A cross-sectional observational study was conducted at a tertiary care hospital and 80 consecutive and consenting patients with diabetes were enrolled over 24 months from 01 Sep 2014 to 31 Aug 2016. After a detailed history and clinical examination, all patients were subjected to a level I polysomnography (PSG), and their blood sample was drawn for the assessment of diabetes control, insulin resistance, and microvascular complications.

Results

Out of 80 patients with diabetes, 30 (37.5%) patients had a high-risk score on Berlin questionnaire and 59 (73.8%) patients had evidence of OSA by PSG. The prevalence of OSA in diabetics with normal body mass index, waist circumference, and neck circumference was 65.5%, 64.2%, and 67.2%, respectively. Patients with OSA had a significantly higher mean glycosylated hemoglobin (correlation coefficient 0.53) and higher insulin resistance (correlation coefficient 0.78). Patients with microvascular complications had a higher prevalence of OSA than those without. This included neuropathy (100% versus 62.5%), retinopathy (100% versus 69.6%) and diabetic kidney disease (DKD) (90.9% versus 52.8%). Patients with any microvascular complication were significantly more likely to have OSA (odds ratio 13.66, 95% confidence interval 3.94–47.37, p < 0.001).

Conclusion

Indian patients with diabetes have a high prevalence of OSA, regardless of obesity. Patients with diabetes and OSA have poorer diabetes control, more insulin resistance, and higher prevalence of microvascular complications like nephropathy, neuropathy, and retinopathy.

Introduction

In most parts of the world, non-communicable diseases cause more premature deaths than communicable, nutrition-related, peri-natal, and maternal deaths combined. Diabetes mellitus, along with cancers, chronic respiratory diseases, and cardiovascular diseases comprise the four biggest causes of mortality and morbidity attributable to non-communicable diseases.¹

Obstructive sleep apnea (OSA) is characterized by the repeated occurrences of upper airway obstruction during sleep leading to sleep fragmentation and/or hypoxemia. OSA is an independent risk factor for several clinical consequences, such as systemic hypertension, cardiovascular disease, stroke, and abnormal glucose metabolism.²

The relationship between Type 2 diabetes mellitus (T2DM) and OSA is probably bidirectional, with either increasing the risk of the other.^3,4 Among patients with T2DM, the prevalence of OSA has been estimated as 23% to as high as 86%.^{5, 6, 7, 8}

Indians are particularly prone to T2DM and coronary artery disease due to some unique characteristics in this population like abdominal adiposity, insulin resistance, higher C-reactive protein, and lower adiponectin.⁹ There is a scarcity of data on the epidemiology of OSA in Indian diabetics.^{10, 11, 12} This study aimed to study the prevalence of OSA in patients with T2DM, using ‘‘Berlin Questionnaire” and Polysomnography (PSG) and see whether the association was independent of obesity. We also endeavored to study the impact of OSA on glycemic control and microvascular complications in these patients.

Material and methods

Study design: This was a cross-sectional observational study conducted over 24 months at a tertiary care teaching hospital from 01 Sep 2014 to 31 Aug 2016.

Study participants: Patients with T2DM aged 18–75 years were consecutively recruited from Medical and Endocrinology out-patient departments of the hospital. All diagnosed cases of T2DM who were on stable medications for diabetes for the preceding 3 months were included. Patients with the following characteristics were excluded from the study: (a) unstable cardiopulmonary, neurological, or psychiatric disease, (b) on sedative or psychotropic drugs, (c) have had upper airway surgery, (d) the use of nocturnal oxygen, positive airway pressure therapy, or oral appliances, (e) recent myocardial infarction, (f) congestive heart failure, (g) pregnancy, (h) hypothyroidism on treatment, (i) acromegaly, (j) chronic renal failure, (k) systemic steroid treatment, and (l) hormone replacement therapy.

Sample size: Taking the level of confidence in an interval estimation of 95%, anticipated prevalence of OSA in T2DM (based on existing data on the prevalence of OSA in T2DM from western studies in Caucasian population) as 71%, an absolute error of margin of 10%, and finite correction for N = 1000, sample size worked out to be 80 patients.⁶

Methodology: After a written informed consent, detailed history and clinical details were recorded for each included participant according to a pre-designed proforma. This included data pertaining to medications and the presence of diabetic microvascular complications. The microangiopathies were diagnosed by screening tests for peripheral neuropathy, retinopathy, and nephropathy. For diagnosing neuropathy, 10 g mono-filament test, tuning-fork test for vibration sense (128 Hz), and ankle jerk were tested. Retinopathy was tested using fundoscopy by an experienced ophthalmologist. For nephropathy, the estimation of serum creatinine and urinary proteins were conducted. Diabetic kidney disease (DKD) was labeled if a patient had any microalbuminuria, macroalbuminuria, or estimated glomerular filtration rate (eGFR) less than 60 mL/min as estimated by Cockcroft-Gault method. ‘Berlin questionnaire’ for OSA was compiled for each of the participants and anthropometric values including body mass index (BMI), neck circumference (NC), waist circumference (WC), and waist-hip ratio (WHR) were measured and recorded. For defining abdominal adiposity, WC of 102 cm and 88 cm were used as cut-offs for males and females, respectively.¹³ It has been suggested that for Indian ethnic population, a lower WC cut-off should be used (90 cm for males and 80 cm for females), in view of typical fat distribution and related cardiovascular risk.¹⁴ These lower cut-offs were separately analyzed. Insulin resistance was estimated using homeostasis model of assessment of insulin resistance (HOMA-IR), and a value more than or equal to 2.5 was considered significant for insulin resistance.¹⁵ An overnight level I PSG was performed using ALICE 5 PSG system (Respironics, Phillips). Electroencephalogram from frontal, central, and occipital regions, submental electromyogram, and electrooculogram were used for sleep staging. Respiratory efforts were assessed with respiratory inductive plethysmography using chest and abdomen belts. Apneas and hypopneas were assessed by oro-nasal thermistor and nasal pressure transducer, respectively. Pulse oximetry, body position sensors, microphone for snoring, and electrocardiogram were also included. Apnea-hypopnea index (AHI) was measured using the recordings. PSG was reported as normal if the AHI was less than 5 per hour. The diagnosis of mild OSA, moderate OSA, or severe OSA was made if the AHI was between 5 and 14 per hour, between 15 and 30 per hour or, more than 30 per hour, respectively. Blood sample was drawn in the morning after PSG for glycosylated hemoglobin (HbA1c), fasting plasma glucose, and insulin levels. Appropriate treatment was offered to patients with newly discovered OSA.

Statistical analysis: The prevalence of OSA was calculated and stratified according to the BMI. Mean HbA1c levels, insulin resistance, body habitus (weight, BMI, WHR, and AC), and the presence of microvascular complications were studied in diabetics with OSA and compared with those without OSA using standard statistical tests. The effect of common confounding factors (weight, BMI, AC, and WHR) were checked by applying regression models. Data analysis was done by using Statistical package for social sciences (SPSS) Version 20.0. Qualitative data variables were expressed by using frequency and percentage. Chi-square test and Fisher's exact test were used to find the association between the occurrence of OSA with demographic variables, lab parameters, and various risks factors. P-value < 0.05 was considered as significant.

The study was approved by the Institutional Ethics Committee. Informed patient consent was obtained in all cases, and there was no invasive procedure involved in the study. All guidelines as per the Declaration of Helsinki and good clinical practice guidelines were followed.

Results

Out of 80 patients enrolled in the study, 48 were males (60.0%). Demographic profile and parameters pertaining to the severity of diabetes, microvascular complications, and anthropometric measurements are detailed in Table 1. Majority of patients (65%) were in the age group between 41 and 60 years. The commonest co-morbidity was hypertension seen in 13 (16.3%) patients. Twenty-two (27.5%) patients qualified as overweight or obese using BMI as the index (Table 1). Twenty-seven (33.8%) patients had abdominal adiposity taking WC of 102 cm and 88 cm as cut-offs for males and females, respectively. When lower cut-offs (90 cm for males and 80 cm for females) were selected, the prevalence of abdominal adiposity in the study population was 55%.

Table 1.

Demographic parameters (n = 80).

Parameter	N (%) or mean ± SD
Males, n (%)	48 (60.0)
Age (years), mean ± SD	49.4 ± 10.0
Age groups
≤40 years, n (%)	16 (20.0)
41–50 years, n (%)	27 (33.8)
51–60 years, n (%)	25 (31.3)
>60 years, n (%)	12 (15.0)
Co-morbidities
Hypertension, n (%)	13 (16.3)
Coronary Artery Disease, n (%)	2 (2.5)
Anthropometric data
BMI (kg/m2), mean ± SD	24.6 ± 3.4
Normal weight: BMI 18.5 to 24.9 kg/m2, n (%)	58 (72.5)
Overweight: BMI 25 to 29.9 kg/m2, n (%)	15 (18.8)
Obese: BMI ≥30 kg/m2, n (%)	7 (8.8)
Waist circumference (cm), mean ± SD	89.7 ± 8.6
Increased Waist circumference, n (%) (male ≥102 cm and female ≥88 cm)	27 (33.8)
Waist-hip ratio, mean ± SD	0.9 ± 0.1
Neck circumference (cm), mean ± SD	39.5 ± 2.4
Increased Neck circumference, n (%) (male ≥43 cm and female ≥41 cm)	16 (20.0)
Severity of Diabetes
HbA1c (%), mean ± SD	7.1 ± 0.7
HbA1c ≤ 6.5%, n (%)	18 (22.5)
HbA1c 6.6–7%, n (%)	24 (30.0)
HbA1c 7.1–8%, n (%)	32 (40.0)
HbA1c > 8%, n (%)	6 (7.5)
Fasting Plasma Glucose ≥110 mg/dl	25 (31.3)
Insulin resistance (HOMA-IR ≥2.5)	26 (32.5)
Microvascular complications
Neuropathy, n (%)	24 (30.0)
Retinopathy, n (%)	11 (13.8)
Microalbuminuria, n (%)	26 (32.5))
Macroalbuminuria, n (%)	16 (20.0)
Azotemia, n (%) Serum Creatinine >1.1 mg/dl	19 (23.8)
eGFR <60 mL/min	18 (22.5)
Diabetic Kidney Disease	44 (55.0)

SD: Standard deviation, BMI: Body Mass Index, HbA1c: Glycosylated Hemoglobin, HOMA-IR: Homeostatic Model Assessment of Insulin Resistance eGFR: estimated glomerular filtration rate by Cockcroft-Gault formula.

A total of 38 patients (47.5%) had HbA1c level more than 7%. Insulin resistance was noted in 26 patients (32.5%). Forty-nine (61.3%) patients had at least one microvascular complication.

Diagnosis of OSA: The prevalence of OSA in diabetics in this study was 73.8% by PSG and 37.5% by Berlin questionnaire score (Table 2). Thirty patients (37.5%) had either moderate or severe OSA which was hitherto undiagnosed. Thirty-two (66.7%) males and twenty-seven (84.4%) females had OSA by PSG. The gender difference in the prevalence of OSA among diabetics was not statistically significant (Fisher's exact test, p = 0.365).

Table 2.

Diagnosis of obstructive sleep apnea (OSA).

High-risk score on Berlin questionnaire), n (%)	30 (37.5)
OSA diagnosis on PSG, n (%)	59 (73.8)
Mild OSA, n (%)	29 (36.3)
Moderate OSA, n (%)	25 (31.3)
Severe OSA, n (%)	5 (6.3)

PSG: Polysomnography.

Correlation of anthropometric measures with OSA: The distribution of OSA in various BMI groups and the relationships with WC, and NC are detailed in Table 3. All three parameters had a significant association with AHI with Pearson correlation coefficients of 0.66, 0.62, and 0.70 for BMI, WC, and NC, respectively (p < 0.001). The prevalence of OSA in diabetics with normal BMI was 65.5%. Similarly, the prevalence of OSA in diabetics with normal WC and normal NC was 64.2% and 67.2%, respectively. The prevalence of OSA was higher in patients with BMI≥25 kg/m2, increased WC, or increased NC (95.5%, 92.6%, and 100%, respectively).

Table 3.

Correlation between Anthropometric parameters and OSA.

Anthropometric parameters	OSA n (%)		Total	P-value
	Present (59)	Absent (21)
Body Mass Index (BMI) -kg/m2				0.024
BMI 18.50–24.99 (Normal)	38 (65.5)	20 (34.5)	58
BMI 25.0–29.99 (Overweight)	14 (93.3)	1 (6.7)	15
BMI ≥30 (Obese)	7 (100)	0 (0)	7
Waist circumference (WC)-cm				0.007
Normal WC	34 (64.2)	19 (35.8)	53
Increased WC	25 (92.6)	2 (7.4)	27
Neck Circumference (NC)-cm				0.008
Normal NC	43 (67.2)	21 (32.8)	64
Increased NC	16 (100)	0 (0)	16

OSA: Obstructive sleep apnea.

Increased Waist circumference: male ≥102 cm and female ≥88 cm.

Increased Neck circumference: male ≥43 cm and female ≥41 cm.

Correlation of severity and control of diabetes with OSA: HbA1c, fasting plasma glucose, and HOMA-IR values were significantly higher in diabetics with OSA than those without (Table 4). Pearson correlation coefficients for these three variables with AHI were 0.53, 0.69, and 0.78, respectively (p < 0.001).

Table 4.

Correlation between Diabetes severity and Obstructive sleep apnea.

Diabetes severity	Obstructive sleep apnea		P-value
	Present (59)	Absent (21)
HbA1c, %	7.24 (±0.7)	6.60 (±0.6)	<0.001
Fasting plasma glucose, mg/dl	110.37 (±16.0)	95.67 (±8.4)	<0.001
HOMA-IR	2.37 (±1.2)	1.13 (0.3)	<0.001

Values are represented as mean (±SD).

HbA1c: Glycosylated hemoglobin.

HOMA-IR: Homeostatic Model Assessment of Insulin Resistance.

Correlation between microvascular complications of diabetes and OSA

Relationship of OSA with diabetic neuropathy, retinopathy, and DKD is detailed in Table 5. OSA had a significantly higher prevalence in patients with microvascular complications than those without.

Table 5.

Correlation between Microvascular complications and OSA.

Microvascular complications	OSA prevalence n (%) when		OR	95% Confidence Interval	P-value
	Complication Present	Complication Absent
Neuropathy	24 (100)	35 (62.5)	NC	NC	<0.001
Retinopathy	11 (100)	48 (69.6)	NC	NC	0.058
Proteinuria	38 (90.5)	21 (55.3)	7.69	2.29–25.86	<0.001
eGFR<60	18 (100)	41 (66.1)	NC	NC	0.002
DKD	40 (90.9)	19 (52.8)	8.95	2.65–30.26	<0.001
Any Microvascular complication	45 (91.8)	14 (45.2)	13.66	3.94–47.37	<0.001

OR: odds ratio.

NC: Not calculated, as 100% of patients with the complication had OSA.

eGFR: estimated glomerular filtration rate by Cockcroft-Gault formula.

DKD: Diabetic Kidney Disease.

Discussion

In this study, involving 80 patients with diabetes mellitus, the prevalence of OSA in T2DM was 73.8%. The prevalence of OSA was high even in patients with normal BMI, WC, and NC. Diabetics with OSA had a higher prevalence of microvascular complications and worse indices of diabetes control and severity.

The prevalence of OSA in general population is considered to be around 24% in men and 9% in women.¹⁶ A more recent study utilizing the current definition and PSG technology, found a much higher prevalence of OSA in general population (49·7% in men and 23·4% in women).¹⁷ Various community-based studies conducted on Indian population have an estimated OSA prevalence in men from 4.4% to 19.7% and that in women from 2.5% to 7.4%.^{18, 19, 20, 21} Patients with T2DM have a much higher prevalence of OSA as has been shown in several studies, most of which have been conducted in western population. A meta-analysis of eight longitudinal studies, that included 16,101 patients, has suggested that the pooled relative risk of incident diabetes with OSA is 1.35.²² Thus, compared to traditional risk factors of diabetes, the effect size of OSA is more than physical inactivity and less than family history of diabetes. The high prevalence of OSA among diabetes in our study (73.8%) is similar to reports from western population and adds to the limited data available for Indian population.^{5, 6, 7,10, 11, 12} The differences in the reported prevalence may be attributable to differences in the modality of OSA diagnosis, the AHI thresholds used when diagnosed by PSG and the differences in the race and ethnicity.

Age, as well as, weight and other anthropometric parameters could confound this association, as central adiposity is a well-known risk factor for both diabetes and OSA.⁴ In this study, though the prevalence of OSA was higher in patients with high BMI, increased WC, or increased NC, even in patients with normal BMI, WC, and NC the prevalence of OSA was higher than that in population studies. Thus, the association between OSA and diabetes appears to be independent of obesity.

In our study, OSA showed association with traditional risk factors like age, BMI, WC, and NC, with the strongest correlation with NC. However, there was no difference in OSA prevalence or AHI severity between the genders. This may be explained by the fact that the mean age of the population was 49.4 years. It has been shown, previously, that the gender differences in OSA prevalence get attenuated with increasing age, particularly in post-menopausal females.²³

Based on answers to the Berlin questionnaire, 37.5% of patients were at ‘‘high’’ risk of OSA. Berlin questionnaire, though a validated and robust tool for OSA screening has demonstrated a lower sensitivity to pick up mild OSA and asymptomatic OSA.²⁴ In a diabetic population, as shown in our study, Berlin questionnaire may be a frail screening tool and may miss several cases that may benefit from treatment. A “modified Berlin questionnaire” has been developed and validated for Indian population.²⁵ The modifications consist of changes in certain questions appropriate for Indian population and reduction in BMI cut-off to 25 kg/m². The modified questionnaire was shown to have a better accuracy than Berlin questionnaire and had the best negative predictive value and negative likelihood ratios among nine questionnaires analyzed.²⁶

There is growing evidence that the relationship between diabetes and OSA is bidirectional.^3,4 It has been shown that sleep fragmentation and intermittent hypoxia can adversely affect glucose metabolism and insulin sensitivity.²⁷ The mechanisms involved may be oxidative stress and systemic inflammation, increasing sympathetic neural activity, activation of hypothalamus–pituitary axis, and alteration of adipokines, altogether leading to pancreatic β-cell dysfunction and insulin resistance.^3,4 In our study, diabetics with OSA had significantly poorer glycemic control and higher insulin resistance than those without OSA.

In this study, mean HbA1c was higher by 0.64% in diabetics with OSA than those without. If we compare patients with moderate to severe OSA (AHI≥15) with those with no or mild OSA, that difference between the mean HbA1c becomes a glaring 0.75%. Similar findings have earlier been shown in a recent large multinational study. In the analysis of the 6117 participants of the European Sleep Apnea Cohort (ESADA), the adjusted mean HbA1c levels were 0.72% higher in patients with severe OSA than patients without OSA.²⁸ Reverse causality has also been hypothesized and it has been shown that autonomic neuropathy in diabetics may affect the central control of breathing during sleep and upper airway patency, leading to sleep-disordered breathing. A high prevalence of OSA in lean patients with diabetes and neuropathy supports this hypothesis.²⁹ We screened our study population only for peripheral neuropathy. Twenty-four patients had features of peripheral neuropathy, and all these patients also had OSA. Screening for autonomic neuropathy can be done by assessing for resting tachycardia, or heart rate variability with deep breathing, Valsalva maneuver and standing from sitting position. Similarly, blood pressure fall (at least 20 mmHg fall in systolic pressure, or 10 mmHg fall in diastolic pressure) after 3 min of standing from sitting posture may indicate the presence of cardiac autonomic dysfunction.

We found a significant association between the presence of OSA and evidence of microangiopathic complications. This association was seen consistently for all complications-neuropathy, retinopathy, and DKD. Several studies in the past have shown association between OSA and diabetes-related microvascular complications.³ A recent study compared a cohort of 3667 patients with diabetes and with OSA and 10,450 age, sex, BMI, and diabetes duration–matched control participants. Patients with T2DM who develop OSA were found to be at an increased risk of peripheral neuropathy (adjusted HR 1.32; 95% CI 1.14, 1.51), diabetic foot disease (1.42; 1.16, 1.74), CKD (1.18; 1.02, 1.36), and all-cause mortality (1.24; 1.10, 1.40) compared with patients without diagnosed OSA.⁵

We found a significant association between OSA and retinopathy. Previously, OSA has been shown to be an independent predictor of proliferative retinopathy.³⁰ This association is driven by increased, insulin resistance, inflammation, and endothelial dysfunction in diabetics.³¹ A recent systematic review and meta-analysis evaluated 15 cross-sectional studies and concluded that OSA was not associated with DR but had an association with an increased severity of DR.³² It has been hypothesized that OSA does not directly cause DR but once DR sets in, episodes of nocturnal hypoxia in patients with OSA causes accelerated the progression of DR. The same was supported in a longitudinal study of 230 patients that showed that OSA was an independent predictor for the progression to pre-proliferative and proliferative DR.³³ Smaller sample size in our study did not allow us to stratify the DR subset with severity.

DKD, proteinuria and reduced eGFR had a significant association with OSA in our diabetic population. A recent systematic review of two longitudinal and ten cross-sectional studies along with meta-analysis of seven studies found an association between OSA and DKD (pooled OR: 1.73, 95% CI: 1.13–2.64).³⁴ Other studies have shown that the severity of OSA appears to be directly correlated to the degree of loss of renal function.³⁵ A longitudinal study of 224 patients with a median follow-up of 2.5 years, showed that CKD prevalence was higher in patients with T2DM who also had OSA than those who did not (49.3 versus 23.8%), and the association remained even after adjusting for confounders (adjusted OR 2.64).³⁶ Patients with OSA had lower eGFR, more microalbuminuria, macroalbuminuria, and greater decline in eGFR over the years with higher AHI being associated with larger decline in eGFR.³⁶ Oxidative and nitrosative stress in patients with diabetes, along with the increased intraglomerular pressures have been hypothesized to drive the pathogenesis of DKD.³⁷ Sleep-disordered breathing and nocturnal hypoxia in OSA causes further sympathetic stimulation and inflammation, and possibly accelerates the evolution of DKD.³⁸ In diabetics with OSA, the severity of hypoxia correlates better with DKD than AHI does.³⁹

Similar to studies from other populations, diabetic neuropathy was found to have a significant association with OSA in Indians, in our study.³⁶ Tahrani et al. have reported an independent association between diabetes and OSA with an odds ratio of 2.82.⁴⁰ Various mechanisms have been postulated for the bidirectional relationship between diabetic polyneuropathy and OSA including inflammation, endothelial dysfunction, oxidative and nitrosative stress, advanced glycation end products, and protein kinase C signaling.⁴¹

This study adds to the scarce literature available on Indian population regarding the relationship between two growing pandemics of diabetes and OSA. The strength of the study is that unlike previous Indian studies, a level I PSG was conducted on every diabetic patient included in the study, irrespective of scores on screening questionnaires or control of diabetes. Also, a comprehensive assessment of the three microvascular complications was done in all the patients.

There were, however, limitations to our study. As with all cross-sectional studies, cause and effect relationships must be interpreted with caution. The role of unmeasured confounders affecting the interpretation cannot be undermined. With two chronic diseases analyzed in a cross-sectional study, errors in interpretation may arise out of prevalence-incidence (Neyman's) bias. We might have missed few cases of autonomic neuropathy without peripheral neuropathy, as we did not screen for the same. Small sample size in our study precluded detailed subgroup analysis.

Conclusion

The prevalence of OSA is high in Indian diabetic patients, regardless of obesity. Standard questionnaire-based screening tools may miss a significant number of diabetics with OSA, and a lower threshold may be kept for a PSG in these patients. Diabetics with OSA have poorer diabetes control, higher insulin resistance, and more microvascular complications. The relation between the two diseases may be bidirectional and long-term prospective studies are required to establish a cause–effect relationship.

Patients/ Guardians/ Participants consent

Patients informed consent was obtained.

Ethical clearance

Institute/hospital ethical clearance certificate was obtained.

Source of support

This paper is based on Armed Forces Medical Research Committee Project No. 4509/2014 granted and funded by the office of the Directorate General Armed Forces Medical Services and Defence Research Development Organization, Government of India.

Disclosure of competing interest

The authors have none to declare.

Acknowledgment

None.