June 6, 2010. The Effects of Obstructive Sleep Apnea and Visceral Fat on Insulin Resistance: The Icelandic Sleep Apnea Cohort, Associated Professional Sleep Societies, LLC (APSS).

1. The Effects of Obstructive Sleep
Apnea and Visceral Fat
on Insulin Resistance:
The Icelandic Sleep Apnea Cohort
Greg Maislin
Director, Biostatistics Core
Adjunct Associate Professor
of Biostatistics in Medicine
Division of Sleep Medicine
University of Pennsylvania School of Medicine

2. Co-Authors
Dept. of Respiratory Medicine Center for Sleep and
and Sleep Respiratory Neurobiology
Landspitali University Division of Sleep Medicine
Hospital University of Pennsylvania
Reykjavik Iceland School of Medicine
Philadelphia PA
 Bryndis Benediktsdottir
 Erna Sif Arnardóttir  Richard J. Schwab
 Isleifur Olafsson  Allan I. Pack
 Thorarinn Gislason

3. Quick Background I
 Mary Ip et al AJRCCM 2002, 165:670
“OSA is Independently Associated with Insulin
Resistance”
 Hong Kong sleep center population, N=270, excluded
known diabetes, homeostasis model assessment method
(HOMA).
 “Stepwise multiple linear regression analyses showed that
obesity was the major determinant of insulin resistance but
sleep disordered breathing parameters were also
independent predictors of insulin resistance.”

4. Quick Background II
 Naresh Punjabi et al AJRCCM 2002 165:677
“Sleep-disorded Breathing and Insulin Resistance in
Middle-aged and Overweight Men”
 Community population, N=150 without diabetes or
cardiopulmonary disease, PSG, MSLT, OGGT and
fasting insulin and lipids
 “Multivariable linear regression analyses revealed that
increasing AHI was associated with worsening insulin
resistance independent of obesity”

5. Quick Background III
 Naresh Punjabi et al Am J Epid 2004, 160:521.
”Sleep-Disordered Breathing, Glucose Intolerence,
and Insulin Resistance: The Sleep Heart Health Study”
 Community dwelling, N=2656, RDI and O2 sat., fasting
and glucose 2-hour glocuse (OGTT) within 1 year of PSG.
 “The results from this study suggest that SDB in
independently associated with glucose intolerance and
insulin resistance and may lead to type 2 diabetes
mellitus.”

6. Quick Background IV
 Igor Harsch et al AJRCCM 2004, 169:156.
“CPAP Treatment Rapidly Improves Insulin
Sensitivity in Patients with OSA”
 N=40, euglycemic clamp (invasive)
 “The effect of CPAP on insulin sensitivity is smaller in non
obese individuals suggesting that in obese patients
insulin sensivity is mainly determined by obesity, and to a
smaller extent, by sleep apnea.”

7. Potential Mechanisms
 Elevated sympathetic nervous system activity
 Alterations in glucocorticoid regulation induced
by sleep loss
 Recurrent intermittant hypoxemia associated with
sleep-disorded breather

8. Aim of Study
 To compare the relative importance of OSA
and obesity on insulin resistance (IR)
 To assess if the effect of OSA on IR varies
among non-obese, mildly obese, and severely
obese
 To use MR imaging volumetric determinations
of abdominal fat

9. Icelandic Sleep Cohort
 Patients diagnosed with OSA at one of five health
clinics within Iceland and referred for CPAP to the
Landspitali University Hospital in Reykjavik from Sept
2005 - Dec 2009. >90% agreed to participate in the
study.
 Subjects were initially diagnosed as having OSA by
sleep studies. Anthropometric measurements,
medical history, health related quality of life and other
questionnaires, MR imaging, and fasting insulin and
glucose (to determine HOMA) were obtained at
baseline.

10. Iceland Sleep Apnea Cohort (ISAC)
Overall and by Obesity Category
Icelandic
Sleep BMI < 30 BMI 30 -< 35 BMI ≥ 35 p-value
2 2 2 for comparison
Demographics Apnea
Cohort
kg/m kg/m kg/m
between groups
(Ranks or χ2)
n = 826 n = 180 n = 192 n = 126
Age (years) 54.4 ± 10.7 55.8 ± 9.7 55.2 ± 10.6 52.5 ± 11.2 0.001
% of males 81.0 82.9 82.39 78.1 0.28
2
Body mass index (kg/m ) 33.5 ± 5.7 27.4 ± 2.0 32.4 ± 1.5 39.7 ± 4.0 N/A
Abdominal fat vol.
3 4.34 (1.83) 3.21 (1.37) 4.61 (1.54) 5.29 (1.93) <0.0001
(MRI times 10 ) (N=665)
Current smokers (%) 21.1 24.2 19.5 20.0 0.37
Epworth Sleepiness Score 11.6 ±5.1 11.5 ± 4.9 11.3 ± 5.2 12.1 ± 5.1 0.20

11. Why restrict analysis cohort by
AHI 14-80 and ODI 10-65?
To simultaneously estimate the effects of BMI category and OSA
severity, the range of the OSA severity must be restricted to be the
same across the BMI tertiles for both AHI and ODI (called ‘double
overlap cohort’) in order to avoid extrapolation.
All subjects Overlap cohort
60
50
BMI
40
BMI tertiles
30 BMI above 35
BMI 30-35
20 BMI below 30
0 30 60 90 120 150
AHI

12. Description of Overlap Cohort
Overall and by Obesity Category
Demographics
BMI < 30 BMI 30 -< 35 BMI ≥ 35 p-value
All 2 2 2
kg/m kg/m kg/m for comparison
between groups
N = 498 n = 180 n = 192 n = 126
(Ranks or χ2)
Age (years) 54.5 ± 10.4 55.4 ± 9.5 54.8 ± 10.9 52.8 ± 10.8 0.09
% of males 82.1 86.7 83.9 73.0 0.007
2
Body mass index (kg/m ) 32.1 ± 4.7 27.5 ± 2.0 32.3 ± 1.4 38.32 ± 2.9 N/A
Abdominal fat vol.
3 4.25 (1.75) 3.26 (1.37) 4.61 (1.54) 5.11 (1.87) <0.0001
(MRI times 10 )
Current smokers %) 20.5 23.3 19.8 17.4 0.45
Epworth Sleepiness Score 11.6 ±4.9 11.9 ± 5.0 11.1 ± 4.7 12.1 ± 5.1 0.17

13. Description of Overlap Cohort
Overall and by Obesity Category
Sleep disordered breathing
p-value
BMI < 30 BMI 30 -< 35 BMI ≥ 35 for comparison
All 2 2 2
kg/m kg/m kg/m between groups
(Wilcoxon or χ2)
Apnea-hypopnea index 40.7 ± 15.4 39.2 ± 14.2 42.6 ± 16.0 40.0 ± 15.7
0.14
(events/hour) (14.8-79.6) (15.3-79.4) (14.8-79.6) (14.8-75.7)
Oxygen desaturation index 31.2 ± 13.7 28.5 ± 12.4 33.2 ± 14.8 31.9 ± 13.3
0.008
(events/hour) (10.3-65) (10.3-63.7) (10.3-65.0) (10.8-61.2)
Minimum SaO2 (%) 77.4 ± 6.8 78.7 ± 5.9 77.4 ± 7.2 75.5 ± 7.0 <0.0001
55.0 ±
Hypoxia time (minutes) 42.5 ± 60.0 30.4 ± 42.8 45.7 ± 63.7 <0.0001
71.5

14. Summary of Insulin Resistance Measures
All BMI<30 BMI 30-35 BMI>35
Measure (N=498) (N=180) (N=192) (N=126) P-value
Glucose 5.60 (0.63) 5.43 (0.50) 5.61 (0.61) 5.83 (0.75) <0.0001
Insulin 17.60 (14.24) 11.29 (8.77) 17.17 (11.75) 27.26 (18.25) <0.0001
HOMA 4.52 (4.06) 2.75 (2.09) 4.38 (3.23) 7.27 (5.58) <0.0001
Log(HOMA) 1.26 (0.68) 0.86 (0.54) 1.29 (0.59) 1.78 (0.62) <0.0001
HOMA>3.99 40.4% 12.2% 42.2% 77.8% <0.0001

15. MRI Imaging of the Abdominal Region
BMI: 27.7 kg/m2 BMI: 32.3 kg/m2 BMI: 39.0 kg/m2
SAT: 3568 cm3 SAT: 5926 cm3 SAT: 9153 cm3
VAT: 2556 cm3 VAT: 4318 cm3 VAT: 5902 cm3
= Subcutaneous Adipose Tissue (SAT) = Visceral Adipose Tissue (VAT) = Liver = Kidneys = Spine

16. Correlations between log(HOMA) and
Obesity Measures
Log(HOMA)
Obesity Measures r p
3
Total abdominal fat volume (cm ) 0.56 <0.0001
2
Body mass index (kg/m ) 0.55 <0.0001
Waist circumference (cm) 0.54 <0.0001
3
Subcutaneous fat volume (cm ) 0.47 <0.0001
3
Visceral fat volume (cm ) 0.44 <0.0001
Waist-to-hip ratio (cm/cm) 0.29 <0.0001
Note: Similar correlations were obtained when assessed
using Spearman correlation

17. Bootstrap Tests for Differences between
Correlations with log(HOMA)
Measure 1 Measure 2 Mean 95% non- P-value
difference parametric CI
Total abdominal fat Visceral fat 0.123 (.073 – 0.173) <0.001
BMI Visceral fat 0.110 (.043 - .186) <0.001
Waist circumference Visceral fat 0.102 (.046 - .162) <0.001
Subcutaneous fat Visceral fat 0.036 (-.043 – .128) 0.16
Waist to hip ratio Visceral fat -0.144 (-.212 - -.081) <0.001

18. Correlations between log(HOMA)
and Apnea Severity Measures
Overall and by Obesity Category
BMI < 30 BMI ≥30 and BMI ≥35
All 2 2 2
kg/m <35 kg/m kg/m
(N=498)
Measure (N=180) (N=192) (N=126)
r p r p r p r p
Apnea-hypopnea index 0.05 0.26 0.08 0.27 -0.04 0.55 0.13 0.16
Oxygen desaturation index 0.14 0.002 0.21 0.005 -0.01 0.84 0.16 0.07
Minimum SaO2 (%) -0.18 <0.001 -0.07 0.37 -0.10 0.16 -0.16 0.07
Hypoxia time (minutes) 0.18 <0.001 0.24 0.0009 0.11 0.13 0.03 0.78

19. Correlations between log(IL-6) and
Apnea Severity Measures
BMI < 30 BMI ≥30 and <35 BMI ≥35
All
kg/m2 kg/m2 kg/m2
(N=347)
Measure (N=142) (N=131) (N=74)
r p r p r p r p
Apnea-hypopnea index -0.02 0.70 -0.23 0.005 0.12 0.16 0.02 0.87
Oxygen desaturation index 0.08 0.12 -0.09 0.30 0.16 0.07 0.07 0.57
Minimum SaO2 (%) -0.13 0.01 0.12 0.17 -0.24 0.005 -0.16 0.18
Hypoxia time (minutes) 0.18 0.001 -0.06 0.45 0.36 <0.0001 0.09 0.43

20. Response Surface Methodology
 Y=f(X1, X2, X3,) + ε, ε~N(0, σε 2)
 E[log(HOMA)] was modeled as a third order
linear model with terms defined by apnea (A)
and obesity severity (O)
 Normal and constant variance assumptions
verified for log transformed outcomes

21. Response Surface Methodology
 The specific model used was:
E[log(HOMA] =
β0 +
β 1*A + β 2*(A)2 +
β 3*O + β 4*(O)2 +
β 5*(A*O) +
β 6*A*(O)2 +β 7*O*(A)2
 A = apnea severity measure; O = obesity severity
measure

22. RSM: log(HOMA)=f(ODI, BMI)
R2=33.3%
(p<0.0001)
4.0
Partial R2’s
3.5 Obesity = 32.6%
3.0 (p<0.0001)
log(HOMA)
2.5
2.0 Apnea = 3.1%
1.5 (p=0.01)
1.0
0.5 O x A = 2.2%
45
0.0
60 40 (p=0.01)
55
50 35 I
45
40
35 30
BM Non-linear = 5%
30
25
OD
I
20
15
10
25
(p=0.0005)

23. Response Surface Methodology
4.0
4 3.5
3.0
log(HOMA)
3
2.5
log(HOMA)
2 2.0
1.5
1 1.0
0.5
0 45
45 0.0
80
75 40 60 40
70 55
65
60 35
MI
50 35 I
55
50
45 30
B 45
40
35 30
BM
40
35 30
AH 30
25 OD 25
20
I 25
20
15 I 15 25
10
4.0 4.0
3.5 3.5
3.0 3.0
log(HOMA)
log(HOMA)
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
45 45
0.0 0.0
40 40
72
74 125 I
76 35 I 35
Min
78
80 30
BM 100
75 BM
Min 30
imu 82
84 ute 50
mS 86 25 sO 25 25
AO 88
2 <90
0
2 %

24. RSM: log(HOMA)=f(ODI, Total Abd. Fat)
R2=34.1%
(p<0.0001)
3.0 Partial R2’s
2.5
Obesity = 32.8%
log(HOMA)
2.0
(p<0.0001)
1.5
1.0 Apnea = 1.4%
0.5 (p=0.20)
14
0.0
60
12
I) O x A = 0.9%
50 10 (MR
40
8 lume (p=0.21)
30
Vo )
OD 20 at
I 6 l F 1000
10
mi
na (x Non-linear = 3.4%
do
Ab (p=0.002)

25. 3.0
2.5
Response Surface Methodology 3.0
2.5
log(HOMA)
log(HOMA)
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 9.0 9.0
0.0
)
7.5
RI 7.5 )
RI
75
70
65 6.0 (M 60 6.0 M
me e(
60 50
55 4.5 4.5
olu
50
45
40
35
3.0
tV
40
3.0 o lum
Fa 000)
30
AH 30 1.5 1.5 tV
I ral OD Fa 000)
25 20
20 0.0
15
e 1 I 0.0 l
isc (x 10
era (x 1
V
Visc
3.0
3.0
2.5
2.5
log(HOMA)
log(HOMA)
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
9.0 9.0
0.0 0.0
7.5 7.5
72 6.0 125 6.0 I)
MR
74
76 4.5 100 4.5 (
me
78 75
M in 80 3.0 I) M in 3.0
imu 82
84 1.5 ( MR ute
sO
50
25
1.5
Vo
lu
mS
AO
86 0.0 me 2 <90
0 0.0
Fa
t )
88 lu l 00
2 tVo %
era 10
Fa ) c (x
l 00 Vis
c era (x
10
Vis

26. Conclusions
 Obesity is a much more important determinant of insulin
resistance than obstructive sleep apnea.
 There is a complex interaction between obesity and OSA
and insulin resistance.
 Different metrics produce different results which may be
related to differences in what these metrics actually
measure.
 Depending upon which metrics are examined, our study
provides additional confirmation that among non-obese, OSA
increases insulin resistance.
 There may also be differentially important amplification of
OSA effects among the most obese.

27. BMI, Total Fat, and Visceral Fat
by BMI Category
N Std
BMI3GRP Obs Variable Label N Mean Dev Min Max
---------------------------------------------------------------------------------------------
<30 180 BMI Body Mass Index (k/m-sq) 180 27.5 2.0 20.0 29.9
MR54 ab total fat vol. 180 7.2 2.1 1.2 12.4
MR56 ab visc fat vol. 180 3.3 1.4 0.2 8.0
30-<35 192 BMI Body Mass Index (k/m-sq) 192 32.3 1.4 30.0 35.0
MR54 ab total fat vol. 192 10.6 2.2 4.7 16.9
MR56 ab visc fat vol. 192 4.6 1.5 0.7 10.1
>=35 126 BMI Body Mass Index (k/m-sq) 126 38.3 2.9 35.1 51.2
MR54 ab total fat vol. 126 14.2 2.7 7.4 22.7
MR56 ab visc fat vol. 126 5.1 1.9 1.4 11.2
---------------------------------------------------------------------------------------------

28. Response Surface Methodology
using BMI as Measure of Obesity
No BMI or No No
No BMI No OSA
OSA BMI*OSA Non-
Null hypothesis tested Effect Effect
effects interaction linearity
(df=5) (df=5)
(df=7) (df=3) (df=4)
AHI p-value for rejecting Ho: <0.0001 <0.0001 0.01 0.02 <0.0001
2 2
R or Partial R 33.9% 32.8% 3.0% 1.9% 4.7%
ODI p-value for rejecting Ho: <0.0001 <0.0001 0.01 0.01 0.0003
2 2
R or Partial R 33.9% 32.6% 3.1% 2.2% 5.0%
Min SAO2 p-value for rejecting Ho: <0.0001 <0.0001 0.10 0.11 0.001
2 2
R or Partial R 33.1% 30.7% 1.9% 1.2% 3.7%
Minutes O2 p-value for rejecting Ho: <0.0001 <0.0001 <0.0001 0.001 0.0003
2 2
< 90% R or Partial R 33.8% 31.3% 2.9% 2.1% 3.6%

29. Response Surface Methodology
Using Abdominal Visceral Fat Volume
No VisFAT No No No
No OSA
or OSA VISFAT VISFAT*OSA Non-
Null hypothesis tested Effect
effects Effect interaction linearity
(df=5)
(df=7) (df=5) (df=3) (df=4)
AHI p-value for rejecting Ho: <0.0001 <0.0001 0.25 0.77 0.003
2 2
R or Partial R 21.9% 26.0% 1.2% 0.1% 3.2%
ODI p-value for rejecting Ho: <0.0001 <0.0001 0.36 0.73 0.02
2 2
R or Partial R 21.7% 25.3% 1.1% 0.3% 2.4%
Min SAO2 p-value for rejecting Ho: <0.0001 <0.0001 0.01 0.62 0.02
2 2
R or Partial R 23.2% 20.5% 1.9% 1.2% 3.7%
Minutes p-value for rejecting Ho: <0.0001 <0.0001 0.06 0.12 0.07
2 2
O2< 90% R or Partial R 22.5% 20.6% 2.1% 1.1% 1.8%

30. Description of Overlap Cohort
Overall and by Obesity Category
Medical history
p-value
All BMI < 30 BMI 30 -< 35 BMI ≥ 35 for comparison
2 2 2
(N=498) kg/m kg/m kg/m between groups
(Wilcoxon or χ2)
Hypertension (%) 41.0 31.3 42.6 52.4 <0.001
Cardiovascular disease (%) 16.0 17.4 16.3 13.5 0.65
Obstructive lung disease (%) 18.2 16.2 16.8 23.2 0.25
Diabetes (%) (current users of diabetes
1.4 0.6 1.1 3.2 0.14
medication excluded from cohort)
Statin use (%) 19.5 18.3 19.8 20.6 0.87
Participate in exercise (%) 65.5 76.3 63.0 54.4 <0.001

31. Construction of the ‘Double Overlap Cohort’
AHI 14-80 and ODI 10-65