Health effects of obstructive sleep apnoea and the effectiveness of continuous positive airways pressure: a systematic review of the research evidenceBMJ 1997; 314 doi: http://dx.doi.org/10.1136/bmj.314.7084.851 (Published 22 March 1997) Cite this as: BMJ 1997;314:851
- John Wright (), consultant in epidemiology and public health medicine,a,
- Rachel Johns, research development managerb,
- Ian Watt, consultant in public health medicinec,
- Arabella Melville, research fellowc,
- Trevor Sheldon, directorc
- a Bradford Royal Infirmary Bradford West Yorkshire BD9 6RJ
- b North Yorkshire Health Authority York YO3 4XF
- c NHS Centre for Reviews and Dissemination University of York York YOU DD
- Correspondence to: Dr Wright
- Accepted 30 January 1997
Additional data from this article are available on the www.bmj.com/
Objective: To examine the research evidence for the health consequences of obstructive sleep apnoea and the effectiveness of continuous positive airways pressure.
Design: A systematic review of published research, studies being identified by searching Medline (1966-96), Embase (1974-96), and CINAHL (Cumulative Index to Nursing and Allied Health Literature) (1982-95); scanning citations; and consulting experts. Studies in all languages were considered which either investigated the association between obstructive sleep apnoea in adults and key health outcomes or evaluated the effectiveness of treatment of obstructive sleep apnoea with continuous positive airways pressure in adults.
Main outcome measures: Mortality, systematic hypertension, cardiac arrhythmias, ischaemic heart disease, left ventricular hypertrophy, pulmonary hypertension, stroke, vehicle accidents, measures of daytime sleepiness, and quality of life.
Results: 54 epidemiological studies examined the association between sleep apnoea and health related outcomes. Most were poorly designed and only weak or contradictory evidence was found of an association with cardiac arrhythmias, ischaemic heart disease, cardiac failure, systemic or pulmonary hypertension, and stroke. Evidence of a link with sleepiness and road traffic accidents was stronger but inconclusive. Only one small randomised controlled trial evaluated continuous positive airways pressure. Five non-randomised controlled trials and 38 uncontrolled trials were identified. Small changes in objectively measured daytime sleepiness were consistently found, but improvements in morbidity, mortality, and quality of life indicators were not adequately assessed.
Conclusions: The relevance of sleep apnoea to public health has been exaggerated. The effectiveness of continuous positive airways pressure in improving health outcomes has been poorly evaluated. There is enough evidence suggesting benefit in reducing daytime sleepiness in some patients to warrant large randomised placebo controlled trials of continuous positive airways pressure versus an effective weight reduction programme and other interventions.
Obstructive sleep apnoea is claimed to be an important cause of premature death and disability
There is increasing pressure to provide sleep services for the treatment of patients with sleep apnoea
Epidemiological evidence suggests that sleep apnoea causes daytime sleepiness and possibly vehicle accidents
Evidence for a causal association between sleep apnoea and other adverse health outcomes is weak
There is a paucity of robust evidence for the clinical and cost effectiveness of continuous positive airways pressure in the treatment of most patients with sleep apnoea
Obstructive sleep apnoea is the periodic reduction (hypopnoea) or cessation (apnoea) of breathing due to narrowing of the upper airways during sleep. The main symptom is daytime sleepiness, and it is thought to be a cause of premature death, hypertension, ischaemic heart disease, stroke, and road traffic accidents.1 2 Prevalence surveys estimate that 4% of middle aged men and 2% of middle aged women are affected by sleep apnoea.3 4 The high prevalence of the syndrome and the morbidity and mortality thought to be associated with it have led to the view that sleep apnoea may be as big a public health hazard as smoking.5 The recommended initial treatment of choice is nasal continuous positive airways pressure,6 and purchasers are increasingly being urged to fund sleep services.2 7
Most discussion on the topic is based on selective and at times uncritical examination of the available research. We conducted a systematic review to examine (a) the evidence of a causal association between sleep apnoea and morbidity and mortality and (b) evidence for the effectiveness of continuous positive airways pressure.
We conducted the review using national structured guidelines.8 A computerised search of Medline (1966 to January 1996), Embase (1974-96), and CINAHL (Cumulative Index to Nursing and Allied Health Literature) (1982 to December 1995) was undertaken (see Appendix 1). Existing reviews were sought, reference lists of identified papers scanned, and experts in the United Kingdom approached. All studies in any language that included adults were considered for review. Epidemiological studies of any design examining the association between sleep apnoea and mortality, hypertension, pulmonary hypertension, cardiovascular disease, and accidents were identified. They were classified as prospective cohort (A1), retrospective cohort (A2), case-control (B), or cross sectional (C).9 Additional grading was based on the adequacy of case ascertainment, adjustment for confounding variables, and validity of the measurement of disease possibly caused by sleep apnoea. All experimental studies were classified according to an internationally established hierarchy of design, in which randomised controlled trials are regarded as the least susceptible to bias.10 All case definitions used in studies were considered.
Abstracts and letters were included if they contained enough methodological information and results. We excluded case reports, studies with no clinical outcome measures, studies which examined only acute or physiological changes during sleep, and studies on sleep apnoea in children. Each paper was evaluated independently by two assessors using a series of predetermined validity criteria on a data extraction form. Disagreements were resolved by a third assessor. Summary tables of each epidemiological study were drawn up by using the grading described above. Summaries of intervention studies with continuous positive airways pressure were included in a table only if they contained some form of control group.
We found 54 epidemiological studies of the association of obstructive sleep apnoea with mortality (n=6),11 12 13 14 15 16 hypertension (n=18),17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 cardiac arrhythmias (n=8),35 36 37 38 39 40 41 42 coronary heart disease and left ventricular failure (n=6),43 44 45 46 47 48 49 pulmonary hypertension (n=6),49 50 51 52 53 54 stroke (n=3),55 56 57 and road traffic accidents (n=7).14 58 59 60 61 62 63 64 65 Disagreements over study design classifications arose in eight papers and were satisfactorily resolved after discussion. Most epidemiological studies were limited in their ability to establish a causal association because of failure to take sufficient account of the potential effects of confounding by such variables as measures of obesity and smoking (which are correlated with both sleep apnoea and poor health) or because they failed to establish a causal time sequence, sleep apnoea being established after the poor health outcome had been diagnosed.9
Two prospective cohort studies examined the association between apnoea-hypopnoea scores and mortality in the general population (table 1). One found no significant association11 and the other found a significant association in women.12 A four year follow up of non-demented retired older people found that the respiratory disturbance index was not a predictor of mortality.13 One prospective study followed up patients with diagnosed sleep apnoea syndrome and examined the death rate relative to that expected for such age and sex groups.14 Multivariate analysis showed that age, hypertension, and body mass index (weight (kg)/height (m)2) had the largest and most significant effects on excess mortality. Apnoea index (but not apnoea duration) was also a predictor of excess mortality but not of excess deaths due to heart or lung causes.
A previous review of daytime blood pressure and obstructive sleep apnoea based on seven observational studies concluded that the evidence of a causal association was still lacking and the confounding influence of body weight had not been assessed adequately.66 Eighteen additional cross sectional studies were identified and are listed in table A (tables A-C are available from JW and on the http://www.bmj.com). Six of these studies found no association of sleep apnoea with raised blood pressure.20 21 27 28 31 33 Four found statistically significant associations with early morning blood pressure,18 19 22 26 but this may be a marker of nocturnal blood pressure.29 67 Eight studies found a significant positive correlation of sleep apnoea with daytime blood pressure but none adjusted for the effects of smoking, alcohol, or antihypertensive drugs.17 18 19 20 21 22 23 24 25 29 30 32 One of these studies, in truck drivers, found that obstructive sleep apnoea was associated with blood pressure over and above body mass index only in obese drivers.17
Arrhythmias, ischaemic heart disease, and left ventricular hypertrophy
Eight studies investigated the prevalence of nocturnal arrhythmias in patients with sleep apnoea.35 36 37 38 39 40 41 42 Two were prospective studies which followed up consecutive referrals and included a control group.35 41 The study with the most valid measurement and classification of arrhythmias found no difference between the groups.35 The prevalence of arrhythmias in both prospective studies was similar to that observed in healthy adults (table B). Three studies, two using a case-control and one a cross sectional design, found an association between the apnoea index and coronary heart disease.43 44 45 Two did not adjust for the effects of all important confounding factors.43 44 In all studies the diagnosis of sleep apnoea was made after the diagnosis of coronary artery disease. Two of the three cross sectional studies which examined the relation with left ventricular hypertrophy46 47 48 found no association.46 47
Pulmonary hypertension and right heart failure
Six cross sectional studies reported a high prevalence of pulmonary hypertension in patients with obstructive sleep apnoea (table 2).49 50 51 52 53 54 Only one used multiple regression to adjust for confounding,49 and only one took smoking into account. All associations could be explained by pre-existing obstructive airways disease, smoking, and obesity.
One case-control study found a relation between both self reported history of snoring and apnoea-like symptoms and the risk of stroke (table 3).55 In addition to the possibility of recall bias in the diagnosis of apnoea, body mass index was poorly adjusted for, being included as a binary variable (body mass index >27.0) rather than as a continuous or more finely graded variable. One cross sectional study also reported that the prevalence of obstructive sleep apnoea was higher in people with recent stroke than in controls.50 However, the sleep apnoea was diagnosed after the stroke, and a recent study has shown that stroke can cause sleep apnoea.57
Road traffic accidents
Six cross sectional studies examined the association between obstructive sleep apnoea and reported car accidents (table C).14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 None made adequate adjustment for potential confounding variables such as age, sex, drinking, obesity, annual milage, shiftwork, and social activities. Of the two studies looking at driving records, one (using state driving records) found an association in patients with severe sleep apnoea58 59 whereas the other (using the records of a cohort of general truck drivers) did not.60 Two of the three studies which relied on self reports of accidents14 61 62 found a higher rate of accidents in people with sleep apnoea.14 61 Three studies using film or computer driving simulators found that sleep apnoea patients made significantly more errors than controls.63 64 65 An association between simulator performance and accident history has been shown in some studies,63 though others have reported that performance is related to age, education, and cognitive function rather than to markers of sleep apnoea.68 69
Evaluation of continuous positive airways pressure
Forty five evaluations of continuous positive airways pressure were identified, of which one was a truly randomised controlled trial70 and five non-randomised controlled trials (table 4).71 72 73 74 75 Thirty eight were simple before and after studies without any control group.47 65 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 Because of not being able to attribute moderate effects to interventions without a proper comparison group the uncontrolled studies are highly unreliable.9 Clinical outcomes used in the seven controlled studies identified were principally sleepiness, mood, psychometric performance, blood pressure, and general health.
The randomised controlled crossover trial of Engleman et al followed up 32 patients and was the only one to compare continuous positive airways pressure directly with a placebo.70 The researchers found a significant improvement in the multiple sleep latency time, vigilance, and Nottingham health profile part 2 scores but no significant difference in patient preference after one month of follow up. Improved performance on a computer driving simulator after treatment was also reported. That study, however, had important weaknesses. A pill was used as the placebo, so it was impossible completely to attribute the reported difference to positive pressure ventilation. Because there was no washout between the periods there was an increased probability of carryover, so underestimating the effect. Significant differential carryover was reported for one psychological outcome. The test for differential carryover has low power, so lower than conventional significance levels should be used. No information, however, was provided about the critical significance values used in the tests of differential carryover for the other variables. Differential carryover can bias the results, and in the one variable for which it was reported an analysis of the first period as a parallel trial showed no significant difference between the groups. Another randomised crossover trial was found, but only an oxygen intervention was directly compared with a placebo (air).71 In the third period of the study all patients received continuous positive airways pressure. We therefore categorised it as a non-randomised controlled trial.
All trials which reported changes in sleepiness found that the multiple sleep latency time114 increased in the treated group compared with controls70 71 72 74 by around one minute in the randomised controlled trial, to up to seven minutes.74 Other measures of daytime sleepiness also improved in the treated arm.71 75 Some studies also found improvement in psychological outcomes such as the hospital anxiety and depression scale,70 attention and recall,71 and general health as measured by the Nottingham health profile part 270 or SF-36 score.75 Only one study found no difference in psychometric performance between the continuous positive airways pressure and comparison groups.72 The two studies which examined blood pressure found no effect of continuous positive airways pressure compared with either a weight loss74 or oxygen control group.71 Though these studies were often poorly designed and the continuous positive airways pressure and control groups often not comparable at baseline, they strongly suggest that continuous positive airways pressure may be effective in reducing sleepiness. This is supported by “switch back” studies, which show a resumption of symptoms on removing continuous positive airways pressure.104 111
Compliance has been studied extensively.105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 Between 50% and 81% of patients accepted continuous positive airways pressure machines, which were switched on for 3.7-6.0 of the 24 hours115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 and used at a “therapeutic pressure” for between 3.4 and 4.5 hours a night.121 123 124
This systematic review indicates that the evidence for a causal association between sleep apnoea and a range of poor health outcomes is generally weak, with the exception of daytime sleepiness and possibly vehicle accidents, for which the evidence is more convincing. Obstructive sleep apnoea is closely associated with obesity130 131 and aging.132 This raises the question of the extent to which sleep apnoea is a separate disease entity or a marker or a symptom of obesity and aging. A major difficulty in investigating the independent health effects of sleep apnoea is in adjusting out the effect of confounding factors. Many epidemiological studies found no association between sleep apnoea and cardiovascular morbidity after adjustment for age and obesity.
Prospective cohort studies which adequately adjust for the effects of confounding factors are the most reliable design for investigating these links.9 The association between sleep apnoea and morbidity found in some retrospective studies may be an artefact of other, coexisting medical conditions or may be explained by sleep apnoea resulting from rather than causing the disease under study. Similar conclusions were arrived at in a review of the relation between snoring and vascular disease.133 Uncontrolled studies of continuous positive airways pressure are also unreliable.9 For example, some patients with sleep apnoea reported improved subjective assessments and increased multiple sleep latency times without any intervention134 and others showed reduced blood pressure with placebo.135 Weight loss has also been shown to lead to significant improvements in symptoms.136 Only comparison with an appropriate (preferably randomised) control group can eliminate these sources of bias.
The quality of the controlled trials was poor. In particular, we cannot be confident that the control groups were sufficiently comparable to eliminate bias, and none included an adequate placebo. A reliable estimate of the true size of any treatment effect of continuous positive airways pressure is likely to be obtained only if the control group receives a placebo which adequately controls for any effect on sleep or breathing patterns, or both, which can occur when appliances are used during sleep.71 137 The feasibility of using continuous positive airways pressure machines set at a low, non-therapeutic pressure as a comparable placebo has been shown.138 The results from these experimental studies do not therefore provide sufficiently robust evidence for the effectiveness of continuous positive airways pressure. The poor standard of evaluative research in sleep apnoea has also been commented on in other reviews, which have examined orthodontic139 and surgical140 interventions.
Daytime tiredness and reduced attention
Anecdotal evidence from clinicians suggests that some patients obtain dramatic benefit from treatment. There are several examples in the history of medicine, however, in which health care interventions, when rigorously evaluated in randomised controlled trials, have been shown to be less effective than anticipated.141 142
The evidence from epidemiological studies suggests that possibly the only significant adverse effect of obstructive sleep apnoea is daytime tiredness and a reduction in attention. Almost all the intervention studies showed some improvement in measures of sleepiness, though the multiple sleep latency test measures the tendency to fall asleep rather than the ability to stay awake, and other measures may be more appropriate.143 Probably the large benefits claimed by some observers are confined to the minority of patients with very severe sleep apnoea who also display obvious symptoms of profound daytime sleepiness. However, these benefits are unlikely to be generalisable to those with less severe sleep apnoea.
High quality research on sleep apnoea in general and continuous positive airways pressure in particular is needed, not in order to deny the validity of clinically apparent benefits in profoundly apnoeic patients but in order to determine which subgroups of patients may derive benefit, how much benefit, at what cost, and how these patients can be identified simply. The results are sufficiently suggestive, however, to justify conducting well designed, large scale, randomised controlled trials to assess objectively the effectiveness and cost effectiveness of treatment with continuous positive airways pressure relative to a suitable placebo. Because obesity is a cause of sleep apnoea and an important determinant of several purported negative effects of sleep apnoea, greater emphasis should be placed on evaluating the impact of effective programmes of weight loss instead of or as adjuncts to more invasive approaches discussed above. It is also important that other treatments for sleep apnoea (such as surgery140 and dental orthoses144) which are rapidly diffusing are evaluated as part of the same research programme so that unified multidisciplinary guidelines can be established. Patients' needs can then be assessed accurately and managed scientifically rather than according to the vagaries of the referral system and the particular enthusiasms of the clinician the patients consult. Calls for widespread investment in health service provision in this topic may be premature until this research has been carried out.
In the Medline search the following terms were used to retrieve items on (a) sleep apnoea and continuous positive airways pressure and (b) sleep apnoea and its health effects and epidemiology.
Thesarus terms (medical subject headings; MeSH)
Sleep apnoea syndromes (exploded)
Positive-pressure respiration (exploded)
Cerebrovascular disorders (exploded)
Accidents, traffic (exploded)
Myocardial infarction (exploded)
Heart failure, congestive (exploded)
Coronary disease (exploded)
Obstructive sleep apn?ea
SAHS [sleep apnoea hypopnoea syndrome]
OSA [obstructive sleep apnoea]
Continuous positive airways pressure
CPAP [continuous positive airways pressure]
High blood pressure
Coronary artery disease
Isch?emic heart disease
Falling asleep at the wheel
The Medline thesaurus terms were translated across into the equivalent Embase thesaurus terms for searching on Embase.
Tables A-C may be obtained by writing direct to JW and are also available on the BMJ's home page (http://www.bmj.com). We thank the four anonymous referees for helpful comments, Diedre Fullerton for help in data abstraction, Olwen Jones for support in literature searching and document acquisition, and Paula Press and Sally Baker for secretarial work.
Funding The initial phase of this review was supported by the Yorkshire Collaborating Centre for Health Services Research.
Conflict of interest None.