Content uploaded by Brydon J. B. Grant
Author content
All content in this area was uploaded by Brydon J. B. Grant
Content may be subject to copyright.
1
Title: 1
Accuracy of Auto-Titrating CPAP to Estimate the Residual Apnea-Hypopnea Index in Patients 2
with Obstructive Sleep Apnea on Treatment with Auto-Titrating CPAP. 3
Authors: 4
Himanshu Desai M.D.1, Anil Patel M.D.2, Pinal Patel M.B.B.S.1, Brydon J.B. Grant M.D.1 and 5
M. Jeffery Mador M.D.1, 3. 6
Institutions: 7
1. State University of New York at Buffalo, Buffalo, NY, USA 8
2. Faxton-St. Luke's Hospital, Utica, NY, USA 9
3. Veteran Affairs Medical Center, Buffalo, New York 10
Corresponding Author: 11
M. Jeffery Mador M.D., 12
Division of Pulmonary, Critical Care and Sleep Medicine, 13
Section 111S 14
3495 Bailey Avenue, 15
Buffalo, New York 14215 16
Phone: 716-862-8635 17
Fax: 716-862-8632 18
Email: mador@buffalo.edu 19
20
21
22
23
2
Abstract: 1
Objective: 2
Auto-titrating continuous positive airway pressure (Auto-CPAP) devices now have a smart card 3
(a pocket-sized card with embedded integrated circuits which records data from the CPAP 4
machine such as CPAP usage, CPAP pressure, large leak, etc.) which can estimate the Apnea-5
Hypopnea Index (AHI) on therapy. The aim of this study was to determine the accuracy of auto-6
CPAP in estimating the residual AHI in patients with obstructive sleep apnea (OSA) who were 7
treated with auto-CPAP without a CPAP titration study. 8
Methods: 9
We studied 99 patients with OSA from 4/2005 to 5/2007 who underwent a repeat sleep study 10
using auto-CPAP. The estimated AHI from auto-CPAP was compared with the AHI from an 11
overnight polysomnogram (PSG) on auto-CPAP using Bland-Altman plot and likelihood ratio 12
analyses. A PSG AHI cutoff of 5 events per hour was used to differentiate patients optimally 13
treated with auto-CPAP from those with residual OSA on therapy. 14
Results: 15
Bland and Altman analysis showed good agreement between auto CPAP AHI and PSG AHI. 16
There was no significant bias when smart card estimates of AHI at home were compared to smart 17
card estimates obtained in the sleep laboratory. An auto-CPAP cutoff for the AHI of 6 events per 18
hour was shown to be optimal for differentiating patients with and without residual OSA with a 19
sensitivity of 0.92 (95% CI: 0.76 to 0.98) and specificity of 0.90 (95% CI: 0.82 to 0.95) with a 20
positive likelihood ratio (LR) of 9.6 (95% CI: 5.1 to 21.5) and a negative likelihood ratio of 21
0.085 (95% CI: 0.02 to 0.25). Auto-CPAP AHI of 8 events per hour yielded the optimal 22
sensitivity (0.94, 95% CI: 0.73 to 0.99) and specificity (0.90, 95% CI: 0.82 to 0.95) with a 23
3
positive LR of 9.6 (95% CI: 5.23 to 20.31) and a negative LR of 0.065 (95%CI: 0.004 to 0.279) 1
to identify patients with a PSG AHI ≥ 10 events per hour. 2
Conclusion: 3
Auto-CPAP estimate of AHI may be used to estimate residual AHI in patients with OSA of 4
varying severity treated with auto-CPAP. 5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
4
1
2
Keywords: 3
Auto-CPAP, Apnea-Hypopnea Index, Obstructive Sleep Apnea, Residual Apnea-Hypopnea 4
Index, Residual Obstructive Sleep Apnea, Smart Card 5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
5
1
2
Abbreviations: 3
AHI: Apnea-Hypopnea Index 4
AASM: American Academy of Sleep Medicine 5
BMI: Body Mass Index 6
CI: Confidence Interval 7
CSA: Central Sleep Apnea 8
CPAP: Continuous Positive Airway Pressure 9
CHF: Congestive Heart Failure 10
COPD: Chronic Obstructive Pulmonary Disease 11
ESS: Epworth Sleepiness Scale 12
OSA: Obstructive Sleep Apnea 13
PSG: Polysomnogram 14
RDI: Respiratory Disturbance Index 15
SD: Standard Deviation 16
17
18
19
20
21
22
23
6
1
2
INTRODUCTION 3
Obstructive sleep apnea syndrome (OSA) is highly prevalent in middle age populations 4
and can interfere with quality of life and increase morbidity and mortality [1, 2]. Continuous 5
positive airway pressure (CPAP) is a standard, safe, and efficacious treatment for patients with 6
OSA [3]. Conventionally, the pressure applied during long term treatment is determined 7
manually by a technician during attended polysomnographic recording. This allows adjusting 8
pressures to find a setting that essentially eliminates apneas and hypopneas in all sleep stages and 9
body positions. In-lab CPAP titration also allows direct observation by trained technologists to 10
guide pressure selection, to adjust mask fit, to eliminate leak, and to help the patient adapt to the 11
initial CPAP experience [4]. However, there are some potential limitations associated with 12
polysomnogram (PSG)-directed CPAP determinations like the cost and inconvenience of repeat 13
PSG, the potential bias of in-laboratory versus in-home environment, and the potential to 14
prescribe pressures that are not optimal due to results based on one night of study [5]. At home 15
auto-CPAP titration has been introduced as a practical strategy that can reduce the time to 16
effective treatment and reduce costs [6, 7]. The American Academy of Sleep Medicine 17
(AASM) has issued practice parameters for the use of auto-CPAP devices for treating patients 18
with OSA and have stated that auto-CPAP devices may be initiated and used in the self-adjusting 19
mode for unattended treatment of patients with moderate to severe OSA without significant co-20
morbidities (congestive heart failure, chronic obstructive pulmonary disease, central sleep apnea 21
syndromes or hypoventilation syndromes). This is an optional recommendation which implies 22
inconclusive or conflicting evidence or conflicting expert opinion [5]. The document also states 23
7
that certain auto-CPAP devices may be used in an unattended way to determine a fixed CPAP 1
treatment pressure for patients with moderate to severe OSA without significant co-morbidities 2
(optional recommendation). 3
AASM also recommended that patients being treated with auto-CPAP must have close 4
clinical follow up to determine treatment effectiveness and safety. A reevaluation or a standard 5
attended CPAP titration should be performed if symptoms do not resolve or if the auto-CPAP 6
treatment otherwise appears to lack efficacy [5]. Approximately two thirds of newly diagnosed 7
patients with OSA at our institution (Veterans Affairs Medical Center at Buffalo) are being 8
started on auto-CPAP treatment to avoid in-lab CPAP titration studies and to improve overall 9
access to sleep lab services. One potential way to determine efficacy of treatment objectively at 10
home is to follow the estimated residual AHI from the smart card which is a part of newer 11
generation auto-CPAP devices. A smart card is a pocket-sized card with embedded integrated 12
circuits which records data from the CPAP machine such as CPAP usage, CPAP pressure, large 13
leak, etc. and can estimate the Apnea-Hypopnea Index (AHI) on therapy using machine specific 14
event detection algortithms.. Sparse data exist to support the finding that the residual AHI 15
obtained from the auto-CPAP devices is a reliable marker of residual OSA [8]. 16
The aim of our study was to determine the accuracy of auto-CPAP in estimating the 17
residual AHI in patients with OSA who are being treated with auto-CPAP without a CPAP 18
titration study. 19
PATIENTS AND METHODS 20
We analyzed data from 99 patients with OSA seen at Veterans Affairs (VA) Medical 21
Center at Buffalo who were being treated with auto-CPAP and who returned for an attended in-22
lab overnight sleep study using auto-CPAP from April, 2005 to May, 2007. All patients had 23
8
obstructive sleep apnea based on an in-lab attended diagnostic PSG. We excluded patients with 1
central sleep apnea (CSA) or combined sleep apnea with predominantly OSA but a central apnea 2
index of ≥ 5 events per hour. The decision to treat the patient with auto-CPAP was made by the 3
treating physician. Patients without a history of snoring were not excluded. Epworth sleepiness 4
scale (ESS) and body mass index (BMI) were recorded at the time of diagnostic PSG for each 5
patient. We also collected information about CPAP compliance, average AHI for the week prior 6
to the study night, mean auto-CPAP pressure, and average leak per minute from the auto-CPAP 7
smart card. 8
9
Polysomnography 10
Out of the 99 patients, 92 diagnostic PSGs were performed at the VA Medical Center at 11
Buffalo and 7 studies were done at different sleep labs with a scanned report in the patient’s 12
electronic medical record. All PSGs done at the VA Medical Center were scored manually by 13
the same experienced, certified sleep technician and were reviewed and verified by one of two 14
certified sleep physicians (MJM or BJG). All patients underwent a second attended standard 15
nocturnal polysomnography using the auto-CPAP device in the sleep laboratory at the VA 16
medical center at Buffalo. The Auto CPAP device used by all patients was the Remstar Auto 17
with software version Encore Pro 1.6 (Respironics, Murrysville, PA). All PSG’s were done 18
using the same polysomnography system (Sandman; Nellcor Puritan 19
Bennett:Ottawa,Ontario,Canada). The auto-CPAP device was used from lights off to lights on 20
with the smart card in place and the estimated AHI from the smart card was obtained without the 21
knowledge of the polysomnography scorer until after the manual scoring of the PSG had been 22
completed. Apnea was defined as the absence of airflow for at least 10 seconds. If respiratory 23
9
effort was present during this apnea episode, it was defined as an obstructive apnea and when 1
respiratory effort was absent, it was termed a central apnea. Hypopnea was defined as a 2
reduction in airflow lasting at least 10 seconds and associated with either a 4 percent drop in 3
arterial oxyhemoglobin saturation or an electroencephalogram arousal. An arousal was defined 4
according to the criteria proposed by the Atlas Task Force [9]. The severity of sleep apnea on 5
diagnostic PSG was classified as follows: mild, AHI > 5 to < 15 events per hour; moderate, AHI 6
>15 to < 30 events per hour; and severe, AHI > 30 events per hour. Residual OSA was defined as 7
an AHI > 5 events per hour on PSG while using auto-CPAP. The estimated AHI from the smart 8
card was compared with the AHI from the overnight PSG. The average estimated AHI from the 9
week prior to the study from the smart card was also compared with the AHI from overnight 10
PSG and with the auto-CPAP AHI from the study night. 11
Statistical Analysis 12
Numeric variables are presented as arithmetic means ± SD or medians (25th, 75th 13
percentiles) when the data were not normally distributed. Statistical significance was considered 14
to be present when p < 0.05. The PSG from the study night was considered the gold standard for 15
identifying and quantifying apneas and hypopneas during sleep. The accuracy of the auto-CPAP 16
smart card in detecting residual AHI was based on comparisons of the auto-CPAP AHI and the 17
PSG AHI which was evaluated by Spearman’s coefficient of rank correlation (due to non normal 18
distribution of the data), agreement using the method of Bland and Altman [10] (MedCalc® 19
Statistical Software version 9.3.8), and by constructing receiver-operator characteristic (ROC) 20
curves [11] to determine optimal cutoff values for determining positive and negative likelihood 21
ratios . We log transformed the data to improve the scatter of the differences as the AHI 22
increased [10]. To avoid zero value problems with log transformation to base 10, we added 0.1 to 23
10
the AHI before log transformation (log 10[AHI+0.1]). The average estimated AHI from the week 1
prior to the study night from the smart card was also compared with PSG AHI and auto-CPAP 2
AHI from the study night using Spearman’s coefficient of rank correlation and Bland-Altman 3
analysis. The area under the receiver operating curve was estimated by the c index.[11].We also 4
analyzed the ability of auto-CPAP AHI to identify patients with residual OSA as likelihood 5
ratios (NCSS 2007 statistical software). Patients with and without residual OSA on auto-CPAP 6
were compared with t-test: two sample assuming unequal variances using Microsoft Office Excel 7
2003[12]. 8
The study protocol was approved by the institutional review board of the Western New 9
York Veteran Affairs Healthcare System, Buffalo, New York. 10
11
RESULTS 12
We analyzed data from 99 patients. Patient characteristics are shown in Table 1. The 13
mean age of patients was 56.7 years (range, 25-83) and the mean body mass index (BMI) was 14
35.0 (range, 22.4-52). Data for AHI and Epworth sleepiness scale (ESS) were available in 92/99 15
(93%) patients at the time of initial diagnostic PSG. 20 (22%) (19 male, 1 female) patients had 16
mild, 30 (33%) (29 male, 1 female) had moderate and 42 (45%) (41 male and 1 female) had 17
severe OSA as per diagnostic PSG. The mean ESS was 13.4±5.2 and median AHI was 26.3 18
(16.0, 63.2) events per hour at the time of diagnosis. Sixteen (16%) patients had COPD and five 19
(5%) patients had CHF. None of the patients had obesity hypoventilation syndrome. 20
Estimated AHI from the auto-CPAP smart card and PSG were obtained and analyzed for 21
all 99 patients. Spearman's coefficient of rank correlation was 0.74 (p<0.0001, 95% CI: 0.64 to 22
0.82) for the auto-CPAP AHI and PSG AHI (Figure 1). All residual apneas and hypopneas were 23
11
obstructive in nature on PSG study. A Bland-Altman plot of log values of auto-CPAP AHI and 1
PSG AHI is shown in Figure 2a. The scatter of AHI differences was non random. The smart 2
card estimate of AHI overestimated PSG AHI at low values of AHI and tended to underestimate 3
PSG AHI at higher AHI levels. A linear regression of the Bland Altman plot revealed an r-value 4
of 0.69 (p<0.0001) as shown in figure 2b. We constructed ROC curves to assess the sensitivity 5
and specificity of different values of auto-CPAP AHI to identify patients with residual OSA, 6
defined as a PSG AHI ≥ 5 events per hour on auto-CPAP. The area under the ROC curve as 7
estimated by the c index was 0.958, 95% CI 0.920 to 0.997. An auto-CPAP AHI of 6 events per 8
hour yielded the optimal sensitivity (0.92, 95% CI: 0.76 to 0.98) and specificity (0.90, 95% CI: 9
0.82 to 0.95) with a positive likelihood ratio (LR) of 9.6 (95% CI: 5.1 to 21.5) and a negative 10
likelihood ratio of 0.085 (95% CI: 0.02 to 0.25) (Table 2). An auto-CPAP AHI of 8 events per 11
hour yielded the optimal sensitivity (0.94, 95% CI: 0.73 to 0.99) and specificity (0.90, 95% CI: 12
0.82 to 0.95) with a positive LR of 9.6 (95% CI: 5.23 to 20.31) and a negative LR of 0.065 13
(95%CI: 0.004 to 0.279) to identify patients with PSG AHI ≥ 10 events per hour on auto-CPAP 14
(Table 2). Auto-CPAP failed to identify one patient with residual OSA (auto-CPAP AHI 4, PSG 15
AHI 5.1). 16
We also had data of average estimated AHI for one week prior to the study night from the 17
auto-CPAP smart card on 88 (89%) patients. We compared average smart card AHI to PSG AHI 18
and to auto-CPAP AHI from the study night. Spearman's coefficients of rank correlation were 19
0.67 (p<0.0001, 95% CI: 0.54 to 0.77) for average AHI and PSG AHI, and 0.86 (p<0.0001, 95% 20
CI: 0.79 to 0.90) for average AHI and auto-CPAP AHI during the study night. Bland-Altman 21
plots of log values of average AHI comparing with PSG AHI, and average AHI comparing with 22
auto-CPAP AHI during the study night are shown in Figure 3a and 3b respectively. The 23
12
comparison of the average AHI and PSG AHI also revealed a non random scatter of AHI 1
differences. The average AHI overestimated PSG AHI at low values of AHI and tended to 2
underestimate PSG AHI at higher AHI values. A linear regression of the Bland-Altman plot 3
revealed an r-value of 0.64 (p<0.0001). Comparing the average AHI and study night AHI from 4
the smart card, there was no significant bias and the scatter of the differences was random. 5
Table 3 shows mean AHI at the time of diagnosis and on auto-CPAP treatment for 6
patients with mild, moderate and severe OSA. 26 (26%) of 99 patients had residual OSA (PSG 7
AHI ≥ 5 events per hour) on auto-CPAP treatment. We compared age, BMI, diagnostic AHI data 8
and auto- CPAP smart card compliance data from patients with residual OSA with those from 9
without residual OSA using t-test: two sample assuming unequal variances (Table 4). Patients 10
without residual OSA were younger (p=0.02) and had lower BMI (p=0.04) compared to patients 11
with residual OSA. Patients without residual OSA tended to have better compliance (% of days 12
with device use, % of days with >4 hours use) than those with residual OSA, which did not quite 13
reach statistical significance. Average leak per minute was higher in patients with residual OSA 14
compared to those without residual OSA (p=0.04). Seventeen patients (17%) had a residual AHI 15
of ≥10 events per hour on auto-CPAP treatment. 16
DISCUSSION 17
In this study, we assessed the accuracy of auto-CPAP to estimate residual AHI in patients 18
with OSA being treated with auto-CPAP without a CPAP titration study. This is the first 19
reasonably large study (n=99) conducted to address this question to the best of our knowledge. 20
Woodson et al [8] (n=24 of which only 8 patients had simultaneous attended PSG) compared 21
auto-CPAP AHI with PSG AHI in patients undergoing unattended home auto-CPAP titration and 22
found that auto CPAP overestimated AHI by an average of 1.4 events per hour when compared 23
13
to PSG AHI. The results of our study showed that there is reasonable clinical agreement between 1
auto-CPAP AHI and PSG AHI and auto CPAP cutoffs can be determined which predict with 2
accuracy which patients have residual disease on therapy as defined by either a PSG AHI of ≥ 5 3
events per hour or ≥ 10events per hour. Bland and Altman plots demonstrate that the difference 4
between auto-CPAP AHI and PSG AHI was not uniform with auto-CPAP overestimating the 5
AHI at lower values of AHI and underestimating the AHI at higher values of AHI. There was a 6
marked underestimation of AHI by auto-CPAP when compared to PSG in two patients. One 7
patient had an auto-CPAP AHI of 9 events per hour and PSG AHI of 41.1 events per hour, the 8
other had auto-CPAP and PSG AHIs of 29 events per hour and 52.7 events per hour, 9
respectively. The first patient had an unsatisfactory PSG study as he slept poorly during the study 10
night. Auto-CPAP identified a similar number of events during the night but since it cannot 11
differentiate between sleep and wakefulness, it assumed a longer sleep time than was actually 12
present, leading to the discrepancy in AHI. In the other patient, no factors were identified that 13
would explain why the auto-CPAP underestimated the actual AHI. Our study showed that an 14
auto-CPAP AHI of 6 identified patients with residual OSA (AHI ≥ 5events per hour) with strong 15
positive and negative likelihood ratios [13]. This suggests that the auto-CPAP device can be used 16
to assess the adequacy of therapy with a reasonable level of accuracy. 17
Our study also showed good clinical agreement between the average smart card AHI 18
from the week prior to the study night and PSG AHI as well as auto-CPAP AHI from the study 19
night. The average AHI from the prior week gives perhaps a better estimate of effectiveness of 20
treatment as it measures the AHI at home and over a period of seven days instead of a one night 21
value obtained in lab. It also replicates the way clinicians will use this modality in the real world. 22
Authors would like to emphasize that in-lab PSG titration has some advantages over home 23
14
strategies like pressure selection under direct observation by a trained technician, adjusting mask 1
fit, eliminating leak and educating patients; and confirming effectiveness during supine or REM 2
sleep. 3
Though no widely-accepted definition of residual OSA exists, we used the traditional 4
cutoff point of AHI ≥ 5 events per hour on PSG [2]. The positive airway pressure titration task 5
force of the AASM defines an optimal CPAP titration as an RDI < 5 events per hour for at least 6
15 minutes which should include supine REM sleep [14]. In this study, we found that almost one 7
in four patients (26%) on auto-CPAP treatment had residual OSA. Although the prevalence of 8
residual OSA in our study may seem high, our findings are consistent with prior investigations. 9
Torre-Bouscoulet et al [15] reported a prevalence of 29% in 279 patients with residual OSA 10
defined as a residual RDI >10 events per hour using an auto-CPAP device. Stammnitz et al [16] 11
also reported a residual OSA prevalence of 17% in a small number of patients, but he defined 12
residual OSA as RDI > 5 events per hour from the auto-CPAP device. Baltzan et al [17] reported 13
a prevalence of 17% using a cut off of AHI ≥ 10 events per hour on PSG in patients treated with 14
fixed pressure CPAP after a CPAP titration study and who had persistent resolution of 15
symptoms. Another study showed that almost one in five patients with good CPAP compliance 16
had residual moderate to severe OSA on their prescribed setting [18]. The long term effects of 17
residual OSA are not fully understood. One study showed that subtherapeutic CPAP did not 18
result in a decrease in mean blood pressure [19]. Peker et al [20] reported that incomplete 19
treatment was not found to be associated with a reduction in the incidence of cardiovascular 20
complications. This high prevalence of residual OSA in patients treated with auto-CPAP or 21
fixed-pressure CPAP after a CPAP titration study stresses the importance of follow-up to 22
determine treatment effectiveness. Clinical follow up alone may not be enough as prevalence of 23
15
residual OSA was found to be high even in patients without symptoms [17, 18]. Thus, institution 1
of auto CPAP with no further evaluation if symptoms improve will not be sufficient if the 2
therapeutic goal is to avoid residual OSA. However, evaluation of residual AHI by smart card 3
estimate may be a satisfactory method to avoid residual OSA in treated patients. 4
In this study, we have not excluded patients with OSA who also had other co morbidities 5
like CHF, COPD or hypoventilation syndromes. The AASM does not recommend use of auto 6
CPAP in patients with such co morbidities [5]. None of our patients in this study had 7
hypoventilation syndromes likely because sleep physicians at our institiution did not start 8
autoCPAP in such patients. ,Sixteen (16%) patients had COPD and 5 (5%) patients had CHF. 9
Since patients with central events and/or Cheynes-Stokes breathing during the diagnostic study 10
were excluded from the study, it is perhaps not surprising that smart card estimates of AHI in the 11
small number of patients with CHF were reasonably accurate. It is interesting that smart card 12
estimates of AHI were no less accurate in patients with COPD than in the rest of the study group. 13
However, the small number of such patients precludes definitive conclusions and further study is 14
required in this patient population. 15
A limitation of our study is that only one manufacturer’s auto-CPAP device was used for 16
the study. Different auto CPAP devices use different algorithms to detect AHI and they may 17
have different detection rates [21,22] . The CPAP device employed in this study adjusts pressure 18
based on analysis of flow (looking for flow limitation) and presence of snoring [22]. Some 19
devices also measure upper airway resistance using the forced oscillation technique [21,22]. 20
Thus, the results of this study may only be applicable to the auto-CPAP device employed in this 21
study. Another limitation is that the smart card only provides an estimate of the AHI on therapy. 22
There is no assessment of oxygen saturation levels. If hypoxemia was prominent in the 23
16
diagnostic study, further assessment may be required to ensure that hypoxemia has been resolved 1
with therapy. 2
In conclusion, auto-CPAP AHI may be used to estimate residual AHI in patients with 3
varying severity of OSA being treated with auto-CPAP without a CPAP titration study provided 4
its limitations in accuracy are understood. Average auto-CPAP AHI from prior week’s use at 5
home was not less accurate than the estimate of AHI obtained from auto-CPAP during the 6
overnight sleep study. Based on the likelihood ratios, auto CPAP yielded a large increase in the 7
probability of residual OSA (PSG AHI ≥ 5 events per hour) when the auto CPAP AHI was ≥ 6 8
events per hour and a large reduction in the probability of residual OSA when the auto CPAP 9
AHI was < 6 events per hour [13]. Thus, auto-CPAP AHI can be used to assess the adequacy of 10
therapy in patients with OSA with reasonable level of accuracy. 11
12
13
14
15
16
17
ACKNOWLEDGEMENTS 18
Funding: None 19
20
Work was performed at: Western New York Veteran Affairs Healthcare System, Buffalo, New 21
York 22
23
17
Himanshu Desai, M.D., has no financial or personal conflict of interest in presenting this 1
manuscript. 2
Anil Patel, M.D., has no financial or personal conflict of interest in presenting this manuscript 3
Pinal Patel, M.B.B.S., has no financial or personal conflict of interest in presenting this 4
manuscript 5
Brydon J. B. Grant, M.D., has no financial or personal conflict of interest in presenting this 6
manuscript 7
M Jeffrey Mador, M.D., has no financial or personal conflict of interest in presenting this 8
manuscript 9
10
11
12
13
14
15
16
17
References 18
1. Young, T., et al., Sleep disordered breathing and mortality: eighteen-year follow-19 up of the Wisconsin sleep cohort. Sleep, 2008. 31(8): p. 1071-8. 20 2. Young, T., et al., The occurrence of sleep-disordered breathing among middle-21 aged adults. N Engl J Med, 1993. 328(17): p. 1230-5. 22 3. Kushida, C.A., et al., Practice parameters for the use of continuous and bilevel 23 positive airway pressure devices to treat adult patients with sleep-related 24 breathing disorders. Sleep, 2006. 29(3): p. 375-80. 25 4. Littner, M.R., et al., Practice parameters for clinical use of the multiple sleep 26 latency test and the maintenance of wakefulness test. Sleep, 2005. 28(1): p. 113-27 21. 28
18
5. Morgenthaler, T.I., et al., Practice parameters for the use of autotitrating 1 continuous positive airway pressure devices for titrating pressures and treating 2 adult patients with obstructive sleep apnea syndrome: an update for 2007. An 3 American Academy of Sleep Medicine report. Sleep, 2008. 31(1): p. 141-7. 4 6. Cross, M.D., et al., Comparison of CPAP titration at home or the sleep laboratory 5 in the sleep apnea hypopnea syndrome. Sleep, 2006. 29(11): p. 1451-5. 6 7. Masa, J.F., et al., Alternative methods of titrating continuous positive airway 7 pressure: a large multicenter study. Am J Respir Crit Care Med, 2004. 170(11): p. 8 1218-24. 9 8. Woodson, B.T., et al., Nonattended home automated continuous positive airway 10 pressure titration: comparison with polysomnography. Otolaryngol Head Neck 11 Surg, 2003. 128(3): p. 353-7. 12 9. EEG arousals: scoring rules and examples: a preliminary report from the Sleep 13 Disorders Atlas Task Force of the American Sleep Disorders Association. Sleep, 14 1992. 15(2): p. 173-84. 15 10. Bland, J.M. and D.G. Altman, Statistical methods for assessing agreement 16 between two methods of clinical measurement. Lancet, 1986. 1(8476): p. 307-10. 17 11. Hanley, J.A. and B.J. McNeil, The meaning and use of the area under a receiver 18 operating characteristic (ROC) curve. Radiology, 1982. 143(1): p. 29-36. 19 12. Ruxton, G.D., The unequal variance t-test is an underused 20 alternative to Student’s t-test and the 21 Mann–Whitney U test. Behavioral Ecology, 2006: p. 688-690. 22 13. Katz, D., Odds and Probabilities: Bayes Theorem and Likelihood Ratios, in Clinical 23 Epidemiology and Evidence-based Medicine: A Primer for the Clinician. 2001, 24 SAGE. p. 57-62. 25 14. Kushida, C.A., et al., Clinical guidelines for the manual titration of positive airway 26 pressure in patients with obstructive sleep apnea. J Clin Sleep Med, 2008. 4(2): p. 27 157-71. 28 15. Torre-Bouscoulet, L., et al., Autoadjusting positive pressure trial in adults with sleep 29 apnea assessed by a simplified diagnostic approach. J Clin Sleep Med, 2008. 30 4(4): p. 341-7. 31 16. Stammnitz, A., et al., Automatic CPAP titration with different self-setting devices in 32 patients with obstructive sleep apnoea. Eur Respir J, 2004. 24(2): p. 273-8. 33 17. Baltzan, M.A., et al., Prevalence of persistent sleep apnea in patients treated 34 with continuous positive airway pressure. Sleep, 2006. 29(4): p. 557-63. 35 18. Pittman, S.D., et al., Follow-up assessment of CPAP efficacy in patients with 36 obstructive sleep apnea using an ambulatory device based on peripheral 37 arterial tonometry. Sleep Breath, 2006. 10(3): p. 123-31. 38 19. Becker, H.F., et al., Effect of nasal continuous positive airway pressure treatment 39 on blood pressure in patients with obstructive sleep apnea. Circulation, 2003. 40 107(1): p. 68-73. 41 20. Peker, Y., et al., Increased incidence of cardiovascular disease in middle-aged 42 men with obstructive sleep apnea: a 7-year follow-up. Am J Respir Crit Care 43 Med, 2002. 166(2): p. 159-65. 44 21. Series,F., et al. Reliability of home CPAP tiration with dfferent automatic CPAP 45 devices. Respiratory Research. 2008. July 24;9:56. 46
19
22. Rigau J. et al. Bench model to simulate upper airwauy obstruction for analyzing 1 automatic continuous positve airway pressure devices. Chest. 2006. 130: p. 350-2 361. 3 4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table 1.Characteristics of the Study Population 35
Mild Moderate Severe Total
Number of Patients* 20 30 42 99
Male 19 29 41 96
Female 1 1 1 3
Age at Diagnostic PSG 54.4 ± 11.4 56.3 ± 13.9 57.4 ± 9.7 56.7 ± 11.3
20
BMI at Diagnostic PSG 33.0 ± 5.1 33.1 ± 5.4 36.8 ± 6.7 35.0 ± 6.3
ESS at Diagnostic PSG 13.2 ± 4.9 13.0 ± 5.4 13.8 ± 5.2 13.4±5.2†
AHI at Diagnosis 11.0 ± 2.5 20.3 ± 4.3 66.1 ± 25.2 26.3 (16.0, 63.2)
Hypertension 13 (65%) 24 (80%) 32 (76%) 74 (75%)
Diabetes Mellitus 6 (30%) 13 (43%) 17 (40%) 40 (40%)
CAD 5 (25%) 6 (20%) 12 (29%) 24 (24%)
COPD 3 (15%) 4 (13%) 8 (19%) 16 (16%)
Hypothyroidism 2 (10%) 1 (3%) 5 (12%) 10 (10%)
Obesity 9 (45%) 9 (30%) 26 (62%) 45 (45%)
CHF 0 (0%) 1 (3%) 4 (10%) 5 (5%)
1 PSG (Polysomnogram), BMI (Body Mass Index), ESS (Epworth Sleepiness Scale), CAD (Coronary Artery 2
Disease), COPD (Chronic Obstructive Pulmonary Disease), CHF (Congestive Heart Failure) 3
Age, BMI and ESS passed normality test and are presented as means ± 1SD. AHI for mild, moderate and severe 4
OSA passed normality test, but total AHI did not and hence presented as median with 25th and 75th percentiles. 5
† Missing data of ESS in 7 patients, * the AHI of patients with outside diagnostic studies not included in Table.. 6
7
8
9
10
11
12
13
Table 2.Sensitivity, Specificity, and Likelihood ratios 14
PSG
AHI
APAP
AHI
Sens Spec + LR Lower
CI
Upper
CI
- LR Lower
CI
Upper
CI
5 6 0.92 0.90 9.6 5.1 21.5 0.085 0.02 0.25
10 8 0.94 0.90 9.6 5.23 20.31 0.065 0.004 0.279
21
APAP (auto-CPAP), Sens (sensitivity), Spec (specificity), + LR (positive likelihood ratio), - LR (negative likelihood 1
ratio), Lower CI (lower 95% confidence interval), Upper CI (upper 95% confidence interval). 2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Table 3.AHI at Diagnosis and on Auto-CPAP Treatment 21
Mild Moderate Severe Total*
AHI at Diagnosis 11.0±2.5 21.0±4.3 66.0±25.2 26.3(16.0, 63.2)
22
AHI on auto-CPAP
from smart card
5.9±6.1 6.4±6.7 7.3±6.9 4.1(3.0, 8.5)
AHI on auto-CPAP
from PSG
2.5±3.9 4.8±10.3 6.6±10.3 1.6(0.3, 6.1)
Presented as mean ± SD for mild, moderate and severe; median (25th, 75th percentile) for total 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Table 4.Characteristics of Patients With and Without Residual OSA on Auto-CPAP 19
23
Patients with
residual
OSA, AHI ≥
5 on auto-
CPAP (n=26,
26%)
Patients without
residual OSA,
AHI <5 on auto-
CPAP (n=73,
74%)
t Critical
two-tail
P value
Age 58.92±12.13 55.32±12.03 2.00 0.02
BMI 37±6.6 34±6.1 2.02 0.04
AHI at Diagnosis 38.67±24.56 37.6±31.36 2.01 0.29
PLMI at Diagnosis 17.44±27.03 20.56±31.17 2.01 0.52
% of Days with Device
Use
73.63±29.29 83.73±24.2 2.04 0.07
Average use for days
used in Min.
298.83±129.0
9
384.45±309.49 1.99 0.08
Average Leak (
milliliters per Minute)
38.06±55.1 16.47±31.82 2.06 0.04
Mean Auto CPAP
pressure from prior
week (cm H2O)
8.34±2.32 7.94±2.14 2.09 0.56
% of Days with >4 Hrs.
Use
49.91±36.32 64.79±31.15 2.04 0.06
Characteristics Presented as means ± SD. AHI: Apnea-Hypopnea Index, BMI: Body Mass Index, PSG: 1
Polysomnogram, PLMI: Periodic Leg Movement Index 2
3
4
5
6
7
8
9
10
11
24
1
Figure 1.Scatter plot of PSG AHI and auto-CPAP AHI 2
3
Spearman’s coefficient of rank correlation is 0.74 (p<0.0001, 95% CI 0.64 to 0.82) for PSG AHI and auto-CPAP 4
AHI. 5
6
7
8
9
10
11
12
13
14
25
Figure 2a.Bland-Altman plot of auto-CPAP AHI and PSG AHI (Logarithmic 1
transformation) 2
3 Log 10(AHI+0.1) values used. The scatter of the differences was non random. 4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
26
Figure 2b. Linear regression of difference between log values of auto-CPAP and PSG AHI 1
with average of log values of auto-CPAP and PSG AHI 2
3
4 5 Log 10(AHI+0.1) values used. r=0.69 (p<0.0001) 6
7
8
9
10
11
12
13
14
15
16
17
18
19
27
Figure 3a.Bland-Altman plot of average smart card AHI and PSG AHI (Logarithmic 1
transformation) 2
3
Log 10(AHI+0.1) values used. The scatter of the differences was non random The difference between auto-CPAP 4 AHI and PSG AHI was significantly correlated with the average of auto-CPAP AHI and PSG AHI ( r=0.64, 5 p<0.0001) 6 7
8
9
10
11
12
13
14
15
16
17
18
19
28
Figure 3b.Bland-Altman plot of average smart card AHI and auto-CPAP AHI 1
(Logarithmic transformation) 2
3
Log 10(AHI+0.1) values used. Mean difference (bias) of 0.2 (95% CI: 0.17 to 0.25). The scatter of the differences 4 was random (r=0.15, p=0.15) 5
6 7
