Professional Documents
Culture Documents
[ Original Research ] 56
2 57
3 58
4 59
5 60
6
7
Q2
Nasal vs Oronasal CPAP for OSA Treatment 61
62
8 Q1 A Meta-analysis 63
9 64
10 65
Q22 Rafaela G. S. Andrade, RPT, PhD; Fernanda M. Viana, MD; Juliana A. Nascimento, RPT; Luciano F. Drager, MD, PhD;
11 66
Q3 Q4 Adriano Moffa; André R. Brunoni, MD, PhD; Pedro R. Genta, MD, PhD; and Geraldo Lorenzi-Filho, MD, PhD
12 67
13 68
14 69
15 BACKGROUND: Nasal CPAP is the “gold standard” treatment for OSA. However, oronasal 70
16 71
masks are frequently used in clinical practice. The aim of this study was to perform a meta-
17 72
analysis of all randomized and nonrandomized trials that compared nasal vs oronasal masks
18 73
on CPAP level, residual apnea-hypopnea index (AHI), and CPAP adherence to treat OSA.
19 74
20 METHODS: The Cochrane Central Register of Controlled Trials, Medline, and Web of Science 75
21 were searched for relevant studies in any language with the following terms: “sleep apnea” 76
22 and “CPAP” or “sleep apnea” and “oronasal mask” or “OSA” and “oronasal CPAP” or 77
23 “oronasal mask” and “adherence.” Studies on CPAP treatment for OSA were included, based 78
24 on the following criteria: (1) original article; (2) randomized or nonrandomized trials; and (3) 79
25 Q8 comparison between nasal and oronasal CPAP included pressure level, and/or residual AHI, 80
26 81
and/or CPAP adherence.
27 82
28 RESULTS: We identified five randomized and eight nonrandomized trials (4,563 patients) that 83
29 reported CPAP level and/or residual AHI and/or CPAP adherence. Overall, the random- 84
30 effects meta-analysis revealed that as compared with nasal, oronasal masks were associated 85
31 with a significantly higher CPAP level (Hedges’ g, –0.59; 95% CI, –0.82 to –0.37; P < .001) 86
32 (on average, þ1.5 cm H2O), higher residual AHI (Hedges’ g, –0.34; 95% CI, –0.52 to –0.17; 87
33 P < .001) (þ2.8 events/h), and a poorer adherence (Hedges’ g, 0.50; 95% CI, 0.21-0.79; P ¼ 88
34 .001) (–48 min/night). 89
35 90
CONCLUSIONS: Oronasal masks are associated with a higher CPAP level, higher residual AHI,
36 91
37
and poorer adherence than nasal masks.
92
38 TRIAL REGISTRY: PROSPERO database; No.: CRD42017064584; URL: https://www.crd.york. 93
39 ac.uk/prospero/. CHEST 2017; -(-):--- 94
40 95
41
Q9 KEY WORDS: CPAP; oronasal mask; OSA 96
42 97
43 98
44 99
45 100
46 101
47 ABBREVIATIONS: AHI = apnea-hypopnea index; NOS = Newcastle- Psychiatry (Mr Moffa), University of New South Wales, Black Dog 102
48 Ottawa Scale Institute, Sydney, Australia. 103
AFFILIATIONS: From the Sleep Laboratory of the Heart Institute (Drs Drs Genta and Lorenzi-Filho are co-last authors.
49 104
Andrade, Viana, Drager, Genta, and Lorenzi-Filho, and Ms Nasci- FUNDING/SUPPORT: The authors have reported to CHEST that no
50 mento), Hospital das Clínicas da Faculdade de Medicina da Uni- 105
funding was received for this study. Q6
51 versidade de São Paulo (HCFMUSP), São Paulo, Brazil; the Service of CORRESPONDENCE TO: Geraldo Lorenzi-Filho, MD, Av. Enéas Car-
106
52 Interdisciplinary Neuromodulation (Dr Brunoni), Laboratory of Neu- valho de Aguiar 44, 05403-000, São Paulo, SP, Brazil; e-mail: geraldo. Q7 107
Q5 rosciences (LIM-27) and National Institute of Biomarkers in Psychi- lorenzi@incor.usp.br
53 atry, Department and Institute of Psychiatry, HCFMUSP, São Paulo, 108
54 Brazil; the Department of Psychiatry and Psychotherapy (Dr Brunoni), Copyright Ó 2017 American College of Chest Physicians. Published by 109
Ludwig Maximilians University, Munich, Germany; and the School of Elsevier Inc. All rights reserved.
55 110
DOI: https://doi.org/10.1016/j.chest.2017.10.044
chestjournal.org 1
Identification
228 database searching 283
229 revision, five randomized trials28-32 and eight n = 3,870
284
230 nonrandomized trials17,18,33-38 that met our inclusion 285
231 criteria were identified. Quality assessment of 286
Records excluded
232 randomized trials revealed that the risk of selection n = 3,834 287
233 bias was moderate. Among the nonrandomized trials, 288
234 three articles scored 8 stars and five articles scored 6 289
235 stars according to the NOS (e-Tables 1 and 2). The 290
Records screened
Screening
236 randomized trials included 125 patients from four n = 36 291
237 292
different countries (two studies from the United States,
238 293
one from Australia, one from the United Kingdom,
239 Records excluded 294
and one from New Zealand). Overall, the
240 n = 14 295
241
nonrandomized trials included 4,438 patients from five 296
242
Q11 different countries (two studies from Canada, two from 297
243 France, and one each from Italy, Germany, Finland, Eligibility Full text articles assessed
298
for elegibility
251 306
summarized in Tables 1 and 2.
252 307
Randomized Observational
253 The random-effects meta-analysis of randomized and 308
Trials Studies
254 nonrandomized trials revealed that when compared with n=5 n=8 309
255 nasal masks, oronasal masks required a higher CPAP level 310
256 Figure 1 – Study selection flow diagram. 311
(Hedges’ g, –0.59; 95% CI, –0.82 to –0.37; P < .001) (Fig 2;
Q23 Q16
Q17
257 312
Q12 funnel plot of data is shown in Fig 3) (on average, þ1.5 cm
258 313
H2O; 95% CI, –2.24 to –0.93), meaning a significant related to these studies was substantial (Q ¼ 31.43; df ¼ 5;
259 314
260
medium effect size. The test for heterogeneity was also P < .001; I2 ¼ 84.1%). None of the three outcomes 315
261 significant (Q ¼ 27.46, df ¼ 9, P ¼ .001, I2 ¼ 67.2%), presented substantial between-study variance (all 316
262 indicating that the variability in outcome measures s2 < 0.1). Although the total number of studies regarding 317
263 between the studies exceeded what was expected by each outcome was limited, visual inspection of CPAP 318
264 chance. Likewise, oronasal masks maintained a level, residual AHI, and CPAP adherence funnel plots 319
265 significantly higher residual AHI (Hedges’ g, –0.34; revealed that the distribution of the studies was 320
266 95% CI, –0.52 to –0.17; P < .001) (Fig 4; funnel plot of reasonably symmetrical, suggesting no evidence of 321
267 data is shown in Fig 5) (on average, þ2.8 events/h; publication bias (Figs 3, 5, 7). Begg test (P ¼ .929, 322
268 95% CI, –3.76 to –0.61) and a small to moderate effect P ¼ .902, P ¼ .452, respectively) supported this finding, 323
269 324
size. Heterogeneity across these studies did not exceed and Egger test (P ¼ .08, P ¼ .894, P ¼ .098, respectively)
270 325
that expected by chance (Q ¼ 11.21; df ¼ 7; P ¼ .13; I2 ¼ did not show evidence of small study effects. Meta-
271 326
37.5%). Finally, users of oronasal masks had poorer regression analyses using sex, age (# 50 vs > 50 years
272 327
273
adherence (Hedges’ g, 0.50; 95% CI, 0.21-0.79; P ¼ .001) old), baseline BMI, baseline OSA severity (< 30 vs $ 30 328
274 (Fig 6) (on average, –48 min/night; 95% CI, 0.56-1.03) events/h), and CPAP titration mode (manual or using an 329
275 compared with users of nasal masks. The heterogeneity auto device) revealed a nonsignificant trend of CPAP 330
chestjournal.org 3
Length of Follow-up
Trial/Year Location Design Inclusion Criteria Total No. Mask for Each Mask Group Titration End Point(s) Q19
Residual AHI
FLA 5.5.0 DTD CHEST1475_proof 10 January 2018 1:19 pm EO: CHEST-17-1495
Ebben29 (2012) USA Randomized AHI > 5 e/h 55 Nasal vs oronasal 1 night Manual CPAP level
vs pillow Residual AHI
Bakker30 New Zealand Randomized AHI > 30 e/h 12 Nasal vs oronasal 2 or 3 nights Automatic/ CPAP level
(2012) vs under chin þ CPAP Residual AHI
oronasal standard with
the pressure
of manual
titration
Ebben31 USA Randomized AHI $ 16 e/h 14 Nasal vs oronasal 1 night Manual Residual AHI CPAP
(2014) adherence
Beecroft33 Canada Nonrandomized AHI > 5 e/h 98 Nasal vs oronasal 1 night Manual CPAP level
(2003) vs oral CPAP adherence
Bachour38 Finland Nonrandomized NA 730 Nasal vs oronasal > 3 wk Automatic CPAP adherence
(2013) vs nasal pillows
Borel17 (2013) France Nonrandomized $ 18 y old 2,311 Nasal vs oronasal Database Automatic CPAP level
OSA vs nasal pillows information of CPAP adherence
CPAP users of 1 to CPAP use
24 mo
Casanova34 France Nonrandomized Home care service 761 Nasal vs oronasal Database Automatic Residual AHI CPAP
(2013) users information of adherence
CPAP use
Kaminska35 Canada Nonrandomized OSA not 6 Nasal vs oronasal 1 night Manual CPAP level
(2014) adequately treat
by oronasal CPAP
[
Bettinzoli36 Italy Nonrandomized AHI > 15 e/h 109 Nasal vs oronasal 3 or 4 nights Automatic CPAP level
-#- CHEST - 2017
AHI ¼ apnea-hypopnea index; APAP ¼ automatic positive pressure; ASV ¼ adaptive servo ventilation; e/h ¼ events per hour; NA ¼ not available; RDI ¼ respiratory disturbance index.
]
400
440
404
408
409
406
430
420
405
403
402
434
424
407
438
428
390
388
439
436
429
426
394
398
389
386
399
396
410
401
435
433
432
425
423
422
414
437
427
418
387
419
416
395
393
392
397
431
421
415
413
412
417
391
411
495
494
493
492
491
490
489
488
487
486
485
484
483
482
481
480
479
478
477
476
475
474
473
472
471
470
469
468
467
466
465
464
463
462
461
460
459
458
457
456
455
454
453
452
451
450
449
448
447
446
445
444
443
442
441
chestjournal.org
FLA 5.5.0 DTD CHEST1475_proof 10 January 2018 1:19 pm EO: CHEST-17-1495
Trial/Year
Mortimore32 Teo28 Ebben29 Bakker30 Ebben31 Beecroft33 Bachour38 Borel17 Casanova34 Kaminska35 Bettinzoli36 Westhoff37 Deshpande18
(1998) (2011) (2012) (2012) (2014) (2003) (2013) (2013) (2013) (2014) (2014) (2015) (2016)
Age, y 52 3 51 13 57 14 48 62 15 50 12 58 12 57 12 NA 53 12 58 13 64 13 57 14
Male, No. % NA 18 (75) 33 (60) 11 (92) 11 (78) 69 (70) 569 (78) 1,641 (71) NA 5 (83) 85 (78) 61 (94) 226 (63)
BMI, kg/m2 32 1 34 9 33 7 37 5 30 6 32 9 NA 31 6 NA 28 5 34 6 32 8 36 9
AHI, e/h 34 5 47 15 36 22 60 29 36 14 41 26 NA 41 21 NA 57 40 41 20 NA 47 30
CPAP,cm H2O
Nasal NA 11.4 1.9 7.6 1.5a 11.5 2.8 7.4 2.4 7.7 2.1 NA 8.8 2.6a NA 10.4 3.0 10.0 2.0a 7.3 2.0a 10.4 3.0
Oronasal NA 11.8 2.4 10.4 3.6 11.5 3.2 9.2 4.1 9.7 3.2 NA 9.6 2.4 NA 16.3 5.4 11.2 2.1 9.5 2.2 12.5 4.0
Nasal NA 5.3 3.4a 2.7 2.2 2.0 2.7a 1.6 1.8a 7.0 13.0 NA NA 2.2 1.2a NA 2.6 2.5a NA 9.4 11.6
Oronasal NA 11.0 10.4 3.6 3.1 3.3 3.9 0.7 1.2 10.0 13.0 NA NA 6.5 1.0 NA 4.5 4.0 NA 15.4 15.4
Adherence,
h/night
Nasal 5.3 0.4a NA NA NA 5.1 1.7 5.5 1.8 5.8 2.8a 5.5 3.4a 5.3 2.0a NA NA NA NA
Oronasal 4.3 0.5 NA NA NA 4.6 2.2 4.0 2.3 4.7 2.8 5.0 2.7 4.5 2.1 NA NA NA NA
Data are summarized as means SD. Note: Not all baseline characteristics were available for all studies. See Table 1 legend for expansion of abbreviations.
a
P < .05.
5
498
499
496
500
540
504
544
508
548
509
506
549
546
497
550
530
520
505
503
502
545
543
542
534
524
507
547
538
528
539
536
529
526
510
501
541
535
533
532
525
523
522
514
537
527
518
519
516
531
521
515
513
512
517
511
551 Study % 606
552 ID SMD (95% CI) Weight 607
553 608
Randomized
554 609
555 Teo, 2011 –0.18 (–0.75 to 0.39) 8.87 610
556 Ebben, 2012 –1.01 (–1.41 to –0.61) 12.38 611
557 Bakker, 2012 0.00 (–0.80 to 0.80) 5.72 612
558 Ebben, 2014 –0.52 (–1.28 to 0.24) 6.20 613
559 614
Subtotal (I2 = 64.0%, P = .040) –0.48 (–0.99 to 0.02) 33.16
560 615
561 616
562 Nonrandomized 617
563 Beecroft, 2003 –0.90 (–1.83 to 0.03) 4.59 618
564 Borel, 2013 –0.31 (–0.41 to –0.22) 19.34 619
565 620
Kaminska, 2014 –1.25 (–2.53 to 0.04) 2.69
566 621
Bentizzoli, 2014 –0.58 (–0.98 to –0.19) 12.47
567 622
568 Westhoff, 2015 –1.04 (–1.48 to –0.60) 11.37 623
569 Deshpande, 2016 –0.58 (–0.82 to –0.34) 16.39 624
570 Subtotal (I2 = 70.5%, P = .005) –0.62 (–0.89 to –0.36) 66.84 625
571 626
572 627
Overall (I2 = 67.2%, P = .001) –0.59 (–0.82 to –0.37) 100.00
573 628
print & web 4C=FPO
593 648
594 .4 649
595 650
Discussion
596 651
597
.6 The main finding of the present meta-analysis is that, 652
598 compared with nasal CPAP, the treatment of OSA with 653
print & web 4C=FPO
682 737
683 NOTE: Weights are from random effects analysis 738
684 739
–1.33 0 1.33
685 740
686 Figure 4 – Forest plot for residual AHI. Data are presented separately for randomized and nonrandomized studies (random-effects model). 741
The effect size statistic is Hedges’ g. Values below 0 favor lower nasal mask pressure and values above 0 favor lower oronasal mask pressure.
687 742
I2 ¼ 37.5% (P ¼ .130). AHI ¼ apnea-hypopnea index. See Figure 2 for expansion of other abbreviation.
688 743
689 744
690 745
There are several potential mechanisms that help to oronasal masks may allow mouth opening.40 Mouth
691 746
explain why oronasal CPAP is less effective than nasal opening in turn is associated with a significant reduction
692 747
CPAP to treat OSA. One possibility is that oronasal in retropalatal and retroglossal cross-sectional areas in
693 748
694
masks push the chin and the tongue backward, inducing awake subjects41 and with a more positive pharyngeal 749
695 upper airway obstruction.39 Another possibility is that critical closing pressure during sleep.42 We recently 750
696 evaluated 18 patients with OSA who slept with an 751
697 oronasal mask with two sealed compartments (nasal and 752
698 Funnel plot with pseudo 95% confidence limits oral) in an experimental condition that allowed acute 753
699 0 changes of the CPAP route with no patient contact and 754
700 no mandibular position change. Acute changes of CPAP 755
701 .1 flow route during sleep from nasal to oronasal and oral 756
702 757
routes induced obstructive events and narrowing of
SE (SMD)
703 .2 758
oropharyngeal dimensions due to a posterior
704 759
displacement of the tongue demonstrated by
705 .3 760
nasoendoscopy.19 The most attractive explanation is that
706 761
707 .4 oronasal CPAP applies positive pressure not only to the 762
708 posterior pharynx but also in the oral compartment.43,44 763
print & web 4C=FPO
chestjournal.org 7
791 846
792 NOTE: Weights are from random effects analysis 847
793 848
794 –3.08 0 3.08 849
795 Figure 6 – Forest plot for CPAP adherence. Data are presented separately for randomized and nonrandomized studies (random-effects model). The 850
796 effect size statistic is Hedges’ g. Values below 0 favor lower nasal mask adherence and values above 0 favor lower oronasal mask adherence. I2 ¼ 851
84.1% (P ¼ .000), suggesting substantial heterogeneity. See Figure 2 legend for expansion of abbreviation.
797 852
798 853
799 46 854
above the critical closing pressure obtained with a nasal adherence rates ranging from 30 to 74%. Common
800 855
mask.45 This mechanism may help to explain the interface-related side effects that may contribute to low
801 856
observation that in some patients the progressive compliance include discomfort, pressure sores,
802 857
increase in oronasal CPAP levels is not able to eliminate unintentional mask leak, skin reactions, and
803 858
804
OSA.19,35 claustrophobia.47,48 The choice of interface is influenced 859
805 by several factors, including the shape of the patient’s 860
Despite the effectiveness and known benefits of CPAP,
806 face and preference, sex, the presence of a mustache, 861
compliance with therapy remains challenging, with
807 breathing pattern (nasal vs oral), and the experience of 862
808 the staff.49 Oronasal masks have been associated with 863
809 Funnel plot with pseudo 95% confidence limits poorer sleep quality, less slow-wave sleep, less 864
810 0 satisfaction, and less comfort when compared with nasal 865
811 masks.30,31,38,50,51 Air leak was reported as a secondary 866
812 .1 867
end point in six studies. Two studies29,35 were neutral
813 868
and four28,30-32 found a significantly higher air leak with
SE (SMD)
814 .2 869
oronasal CPAP as compared with nasal CPAP. Oronasal
815 870
.3 masks are bigger and have higher contact than nasal
816 871
817
masks, which may help to explain the presence of more 872
818 .4 air leaks. All these difficulties, associated with the higher 873
819 CPAP level and higher residual AHI, help to explain the 874
print & web 4C=FPO
884 939
efficacy and adherence. analysis that compared nasal vs oronasal masks in
885 940
patients who are mouth breathers. Nasal CPAP should
886 Our meta-analysis has several strengths. The pooling of 941
be the first option. However, oronasal masks may
887 relevant data from all eligible studies showed 942
888
outperform nasal masks in patients who are 943
consistency and yielded more precise and reliable
889 predominantly mouth breathers or who experience 944
conclusions than those drawn from individual studies.
890 significant air leak through an open mouth. Future 945
Nevertheless, there are several limitations to be
891 physiologic experiments to understand the effect of the 946
addressed: first, we used pooled data of randomized and
892 CPAP interface on upper airway anatomy are 947
nonrandomized studies due to the small number of
893 warranted. The field needs randomized controlled trials 948
patients, and only five randomized controlled trials were
894 comparing nasal and oronasal CPAP, targeting the 949
895
included. The meta-analysis of the randomized trials 950
subgroup of patients who are mouth breathers or who
896 showed the same trends but did not reach statistical 951
experience a large mouth leak while using nasal CPAP.
897 significance (Figs 2, 4, 6). Therefore, larger randomized 952
898 controlled trials would contribute to the field. Second, In conclusion, our meta-analysis indicates that oronasal 953
899 the risk of heterogeneity was moderate to high on CPAP masks are associated with a higher CPAP level, higher 954
900 level and adherence, possibly due to different study residual AHI, and poorer adherence than nasal masks. 955
901 designs (randomized and nonrandomized) and diverse Therefore, patients using oronasal masks should be 956
902 characteristics of included patients. Third, the methods closely monitored. Nasal interfaces should always be the 957
903 958
were heterogeneous between studies. For instance, first choice. There is evidence that the prolonged use of
904 959
CPAP was manually titrated in two randomized nasal CPAP reduces mouth opening and oral
905 960
studies29,31 while automatic CPAP titration was breathing.38,53,54 Therefore, even patients with OSA who
906 961
performed in three studies.28,30,32 The reported CPAP claim to be mouth breathers could be initiated with nasal
907 962
908
levels varied from the 90th percentile36 to the 95th masks. Another clinical implication is that sleep studies 963
909 percentile.28,30,38 Unfortunately, standard definitions for CPAP titration must be performed with the same 964
910 about CPAP nomenclature, adherence, signals, and type of mask that the patient will use at home. 965
911 966
912 967
913 2. Young T, Peppard PE, Gottlieb DJ. 8. Marin J, Carrizo S, Vicente E, Agusti A. 968
914
Acknowledgments Epidemiology of obstructive sleep apnea: Long-term cardiovascular outcomes in 969
Author contributions: Each author had full a population health perspective. Am J men with obstructive sleep apnoea-
915 access to the data and takes responsibility for 970
Respir Crit Care Med. 2002;165(9): hypopnea with or without treatment with
916 Q15 the integrity and accuracy of the analysis. All 1217-1239. continuous positive airway pressure: an 971
917 authors contributed to and approved the final observational study. Lancet. 972
3. Tufik S, Santos-Silva R, Taddei JA,
submitted manuscript. R. G. S. A.: study Bittencourt LRA. Obstructive sleep apnea 2005;365(9464):1046-1053.
918 design, data collection, analysis, and 973
syndrome in the Sao Paulo Epidemiologic 9. Luciano F, Drager VYP, Lorenzi-Filho G.
919 manuscript preparation; F. M. V.: data Sleep Study. Sleep Med. 2010;11(5): Obstructive sleep apnea: an emerging risk 974
920 collection and analysis; J. A. N.: data 441-446. factor for atherosclerosis. Chest. 975
collection and analysis; L. F. D.: manuscript 2011;140(2):534-542.
921 preparation; A. M.: analysis, manuscript 4. Heinzer R, Andries D, Bastardot F, et al. 976
Prevalence of sleep disordered breathing 10. Jenkinson C, Davies RJ, Mullins R,
922 preparation; A. R. B.: manuscript 977
in middle-aged general population: the Stradling JR. Comparison of therapeutic
923 preparation; P. R. G.: study design, analysis, HypnoLaus study. Lancet Respir Med. and subtherapeutic nasal continuous 978
and manuscript preparation; G. L.-F.: study 2015;3(4):310-318. positive airway pressure for obstructive
924 design and manuscript preparation. 979
5. Young T, Palta M, Dempsey J, Skatrud J, sleep apnoea: a randomised prospective
925 parallel trial. Lancet. 1999;353:2100-2105. 980
Financial/nonfinancial disclosure: None Weber S, Badr S. The occurence of sleep-
926 Q14 declared. disordered breathing among middle-aged 11. Giles TL, Lasserson TJ, Smith BJ, White J, 981
927 adults. N Engl J Med. 1993;328(17): Wright J, Cates CJ. Continuous positive 982
Additional information: The e-Tables can 1230-1235. airways pressure for obstructive sleep
928 be found in the Supplemental Materials 983
6. Baranchuk A. Sleep apnea, cardiac apnoea in adults. Cochrane Database Syst
929 section of the online article. Rev. 2006;3:CD001106. 984
arrhythmias, and conduction disorders.
930 J Electrocardiol. 2012;45(5):508-512. 12. Pedrosa RP, Drager LF, de Paula LKG, 985
931 References 7. Engleman HM, Kingshott RN, Wraith PK, Amaro ACS, Bortolotto LA, Lorenzi- 986
1. American Academy of Sleep Medicine Mackay TW, Deary IJ, Douglas NJ. Filho G. Effects of OSA treatment on BP
932 in patients with resistant hypertension: a 987
Task Force. Sleep-related breathing Randomized placebo-controlled crossover
933 disorders in adults: recommendations for trial of continuous positive airway randomized trial. Chest. 2013;144(5): 988
1487-1494.
934 syndrome definition and pressure for mild sleep apnea/hypopnea 989
measurementtechniques in clinic al syndrome. Am J Respir Crit Care Med. 13. Weaver TE, Grunstein RR. Adherence to
935 research. Sleep. 1999;22(5):667-689. 1999;159(2):461-467. continuous positive airway pressure 990
chestjournal.org 9