A study of glycemic variability in patients with type 2 diabetes mellitus with obstructive sleep apnea syndrome using a continuous glucose monitoring system

Aims

To assess the Glycemic Variability in patients with type 2 diabetes and OSAS. The primary objective was to correlate GV to respiratory disturbance indices in patients of type2 Diabetes with OSAS. The secondary objective was to assess GV in patients with OSAS without DM.

Design and setting

This cross sectional study was conducted by the Endocrinology and Pulmonology services of a tertiary care center in Mumbai, Maharashtra, India, after Institutional Ethics Committee approval (No. ECARP/2018/95). We screened 55 patients with type 2 DM treated with lifestyle measures (LSM) alone or LSM with one or more oral antidiabetic medications (OAD). Patients aged between 18 to 60 years who had a BMI between 23 to 30 kg/m2 were eligible to participate in the study. The major exclusion criteria were patients with type 2 diabetes who were on insulin, type 1 diabetes, upper airway surgery in the past, sinusitis, chronic respiratory diseases, heart, lung, liver and kidney disease, patients who were taking sedative-hypnotic medications, pregnant patients, patients with history of alcohol consumption, smoking, steroid intake and untreated hypothyroidism. After applying inclusion and exclusion criteria, 40 patients with type 2 DM underwent further study. We included 10 patients with OSAS without diabetes from Pulmonology clinic and 10 were normal volunteers without diabetes. After a written informed consent participant’s baseline characteristics, anthropometric and clinical data were collected by a single investigator.

Patients were screened for OSAS with the help of Modified Sleep Apnea Clinical Score (SACS) which is a pre-test probability score- based on snoring (3 points), witnessed episodes of apnea (3 points), neck circumference (in cm) and systemic hypertension (4 points). Based on SACS patients were categorized into low (below 43), medium (43–48) and high risk (above 48) Groups [20, 21]. SACS score was used as a screening tool to identify patients with high probability of OSAS and to avoid unnecessary screening polysomnography (PSG) in a resource limited setting. Patients in moderate and severe groups were subjected to a screening polysomnography (PSG) for confirmation of the diagnosis. Patients in the low risk group willing to participate in the study were also screened with PSG with an intent to classify them into control groups. PSG was done with RESMED’s Apnea link device in the sleep laboratory of the Pulmonology department. This was a limited 5 channel level 3 PSG involving measurement of cardiovascular variables. It measured snoring, respiratory effort, pulse, oxygen saturation and nasal flow. The subjects were monitored for sleep apnea starting at 22:00 h. The sleep monitoring equipment was worn for at least 7 h and was removed by the specialist the next morning after the patient awoke. During the test, the patients did not have access to sedatives, coffee and tea. The Apnea Hypopnea Index (AHI) was calculated based on the total number of sleep apneas and hypopneas per hour. OSAS was defined as per American association of sleep medicine (AASM) criteria [22]. Patients with AHI ≥5 /hour, were classified as having obstructive sleep apnea. Four patients with low SACS score had significant AHI scores and were included in the obstructive sleep apnea group. The patients were categorized into 4 groups based on AHI:

Group A: DM with OSAS (n = 20 patients),

Group B: DM without OSAS (n = 20 patients),

Group C: Non DM with OSAS (n = 10 patients),

Group D: Non DM without OSAS (n = 10 patients).

On day 1, a fasting blood sample was drawn after a 12 h overnight fast between 8 to 9 am for routine biochemistry and HbA1c. HbA1c was determined by High Performance Liquid Chromatography using BioRad D10 Analyzer (Intra and Inter assay coefficient of variation- < 2%).

Continuous glucose monitoring with the iPro2 CGM (Model REF-MMT 7102 W, Medtronic MiniMed, USA) was initiated for 5 days. The sensor was inserted subcutaneously over the anterior abdominal wall. The instrument was calibrated by four capillary glucose values obtained on a glucometer (Freestyle Optium Xceed) prior to three major meals and at bedtime. Blood glucose meters and the test strips were provided to the patient. Patients were instructed to be consistent with their meal timings, pattern and maintain a food diary which was analyzed by a registered dietitian. The patients were subjected to a second PSG on the second night of CGMS insertion. Fig. ​Fig.11.

 

Please note that this image might contain sensitive content. Click to unblur.

Open in a separate window

Fig. 1

Flowchart of Methodology

CGM data was downloaded with the CareLink Ipro1 software (MMT-7340) (Medtronic, Minneapolis, MN, USA) and this data was used to calculate the variability parameters by an automated Software EasyGV version 9.0.R2. Glycemic variability was defined as intraday glycaemic excursions, including episodes of hyperglycemia and hypoglycemia. Following variables were calculated from CGM readings for each patient: Time In Range (TIR), Time Above Range (TAR) and Time Below Range (TBR), Mean glucose, Standard deviation (SD), Night SD, Coefficient of Variation (CV), Night CV, Mean Amplitude of Glycemic excursion (MAGE) [23] and Night mean amplitude of glycemic excursion (NMAGE). NMAGE was calculated as MAGE for the night period and was recorded from 10 pm to 6 am.

Sample size was calculated based on the study of glycaemic variability and OSA by Nakata K, et al [24]; at 80% power and 5% alpha error by comparison of the mean method, the minimum sample size was found to be 18 per group in diabetes cohort.

Statistical analysis was done using the SPSS v 20.Demographic data was analyzed using descriptive statistics. Statistical significance was set to P <  0.05. Quantitative Data was expressed as mean, standard deviation and median. The comparison of GV indices between different groups was performed by one-way analysis of variance (ANOVA) for normally distributed data. For data parameters which failed the normality test, Kruskal Wallis test was applied. Post hoc analysis for multiple intergroup comparisons was done using Tukey HSD after ANOVA or Dunn’s method after Kruskal Wallis test. Correlation analysis of AHI with SD/ MAGE/NMAGE was done by Pearson correlation.

Demographic data

We had 20 patients each in group A and group B and10 patients in group C and group D. The four study groups were adequately matched for age, sex, BMI, waist circumference and blood pressure. Table 1 shows the baseline characteristics of the cohort.

Patients with Diabetes Mellitus who were on lifestyle modifications alone, or on Lifestyle modification with monotherapy or combination therapy with 2 or 3 oral antidiabetic drugs. Fifty percent patients had diabetes duration of more than 5 years. 47.5% (19 out of 40) had retinopathy and 60%(24 out of 40) had peripheral neuropathy (Table ​Table22).

HbA1c was comparable in Group A and B (8.87 ± 2.54% vs. 8.18 ± 2.16%) P value > 0.05. SACS was higher in the group A (45.10 ± 4.24) and was statistically significant (p <  0.00012) when compared with rest of the three groups. Four patients with low SACS had positive polysomnography. The BMI of our patients were in the category of overweight and obese as per the Indian definition of BMI [25]. This is well described as thin fat Indian phenotype [25].

CGMS parameters:

Mean glucose (7.77 ± 2.42 mmol/L), glucose SD (2.44 ± 1.10 mmol/L), Night CV (31% ± 4.5) MAGE (5.20 ± 2.15 mmol/L) and NMAGE (5.78 ± 2.16) were highest in Group A as compared to rest of the three groups which was statistically significant (p values 0.006, 0.008, 0.008, 0.01 respectively) (Table 3). Post hoc analysis was then done to compare group A with group B and group C with group D.

Glycemic variability in group a and group B: (Table 4)

Mean glucose (7.77 ± 2.42 vs. 6.81 ± 2.31 mmol/L), glucose SD (2.44 ± 1.10 vs. 1.91 ± 0.86 mmol/L), MAGE (5.20 ± 2.15 vs. 2.67 ± 1.04 mmol/L) and NMAGE (5.78 ± 2.16 vs. 2.22 ± 1.00 mmol/L) were higher in Group A than Group B. Post hoc analysis with multiple comparisons (Dunn’s method) in subgroup analysis, showed MAGE and NMAGE were significantly different between group A and group B (p <  0.05). The measures of % CV and SD were not significant but the night CV was significantly different (P < 0.05).

Glycemic variability in group C and group D: (Table 5)

Comparing the Non Diabetes groups (group C and D), we found the mean glucose (5.85 ± 2.10 vs. 4.83 ± 1.95 mmol/L), glucose SD (0.87 ± 0.16 vs. 0.68 ± 0.11 mmol/L),MAGE (2.47 ± 0.42 vs. 0.77 ± 0.19 mmol/L) NMAGE (2.85 ± 0.42 vs. 0.62 ± 0.15 mmol/L) and Night CV (37 ± 2.5 vs. 29 ± 2.3) were higher in group C (with OSAS) as compared to those in group D. MAGE, NMAGE and Night CV were statistically significant (p value < 0.05). The TAR in group C was 8% ± 1.2 whereas it was 0% in group D.

Glycemic profile in group A and group B from 10 pm to 6 am

We compared the mean glucose profile curves from 10 pm to 6 am in group A and group B. There was a higher variability in group A as compared to group B. The mean peak glucose in group A and group B was 12.76 mmol/L and 11.65 mmol/L whereas the mean nadir glucose value was 2.83 mmol/L and 3.1 mmol/L respectively Fig. ​Fig.22.

 

Please note that this image might contain sensitive content. Click to unblur.

Open in a separate window

Fig. 2

Mean glucose profile curves of all groups from 10 pm to 6 am

Patients showing increased glycaemic variability in group A than group B and increased variability in group C than group D.

Glycemic profile in (group C and group D) from 10 pm to 6 am

We also compared the mean glucose profile curves from 10 pm to 6 am in group C and group D. Group C showed a higher variability as compared to group D. The mean peak glucose level of group C and group D was 6.2 mmol/L and 5.5 mmol/L whereas mean nadir glucose value was 4.1 mmol/L and 3.99 mmol/L respectively.

Correlation of AHI with GV parameters in patients with Diabetes (Group A + Group B) Fig. ​Fig.33a

Please note that this image might contain sensitive content. Click to unblur.

 

Open in a separate window

Fig. 3

a: Pearson’s correlation between AHI and GV parameters in diabetes groups (OSAS + Non OSAS) (n-40). b: Pearson’s correlation between AHI and GV parameters in non-diabetes groups (OSAS + Non OSAS) (n-20)

AHI positively correlated with Glucose SD, MAGE and NMAGE with correlation coefficient of 0.219, 0.464, and 0.521 respectively in patients with diabetes (group A and B) with/without OSAS on Pearson’s correlation. This finding denotes moderate correlation of OSAS severity with glycemic variability indices.

Correlation of AHI with GV parameters in patients without diabetes (Group C + Group D) Fig. ​Fig.33b

AHI also correlated positively (moderate correlation) with glycemic variability parameters of Glucose SD, MAGE and NMAGE with correlation coefficient of 0.494, 0.438 and 0.495 respectively on Pearson’s correlation.

No

Authors’ information

Suhas S. Khaire – MD (Pediatrics), DM (Endocrinology); Assistant Professor.

Jugal V. Gada – MD (Internal Medicine), DM (Endocrinology); Assistant Professor.

Ketaki V. Utpat- MD (Pulmonary Medicine), DNB (Respiratory Medicine), EDRM; Assistant Professor.

Nikita Shah- MD (Internal Medicine), DM (Endocrinology).

Premlata K. Varthakavi – MD (Internal Medicine), Diplomate of National Board – DNB (Endocrinology); Professor.

Nikhil M. Bhagwat- MD (Internal Medicine), DM (Endocrinology); Additional Professor and Head of Department.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Associated Data

Data Availability Statement