Condition: Sleep Apnea

Across 108 studies and 47,989 participants, for moderate apnea (AHI 15 or more) pooled sensitivity and specificity were 90% and 36% for STOP-BANG, 77% and 44% for the Berlin questionnaire, and 47% and 62% for the Epworth Sleepiness Scale. For mild apnea (AHI 5 or more), STOP-BANG was 88% sensitive and 42% specific, the Berlin 76% and 59%, and Epworth 54% and 65%. Measured in: 47,989 participants across 108 diagnostic accuracy studies, mostly clinic populations referred for suspected apnea. The review found sex to be a significant moderator of both sensitivity and specificity without publishing stratified estimates, which is a stronger version of this page’s point than saying the question went unexamined. On the instrument itself, only the neck circumference and male sex items are unambiguously weighted toward men: by our own cited clinic series, women presenting with apnea are older and heavier, so the age and BMI items score in their favor.

The AASM clinical practice guideline recommends polysomnography, or home sleep apnea testing with a technically adequate device, for uncomplicated adults with signs and symptoms of moderate to severe apnea, and recommends (strong) against using clinical tools, questionnaires or prediction algorithms to diagnose apnea in the absence of a sleep study. If a single home test is negative, inconclusive or technically inadequate, polysomnography should be performed. Measured in: Adults being assessed for obstructive sleep apnea; GRADE assessment of the underlying diagnostic literature. A guideline is a synthesis with expert judgement layered on top, and the underlying evidence for several recommendations was rated low certainty. The home-testing recommendation is limited to uncomplicated patients: significant cardiorespiratory disease, neuromuscular weakness, opioid use, hypoventilation or suspected non-respiratory sleep disorders take you back to the laboratory study.

In a sample of 4,925 employed adults, with in-laboratory polysomnography on a subset of 1,090, an estimated 93% of women and 82% of men with moderate to severe sleep apnea syndrome had never been clinically diagnosed. Measured in: 4,925 employed adults in Wisconsin with good access to healthcare, 1,090 of whom had laboratory sleep studies. What could explain it instead: Diagnosed status came from self-reported doctor diagnosis, so anyone told they had 'a snoring problem' without a formal label counts as undiagnosed. Health-care-seeking behavior also differs by sex independently of symptom severity, which contributes to the gap without being a failure of the clinician.. These are 1997 figures from a single employed US population, and diagnosis has become easier since home testing became routine, so the absolute rates are dated. The 11-point sex gap is the part that later presentation studies keep reproducing. Sex-specific substudies do exist on weight and on glycemia, including one in which women on usual care fared worse. What no trial reports by sex is its CARDIOVASCULAR outcome, which is the narrower and defensible version of this gap.

Among 2,827 patients with diagnosed obstructive sleep apnea (2,052 men, 775 women), men reported witnessed apneas more often, while nocturnal choking, morning headache, fatigue, insomnia symptoms, impaired memory, mood disturbance, reflux and nocturia were all more frequent in women (all p<0.0001). Women were 5.7 years older at diagnosis (56.1 vs 50.4) and heavier (BMI 36.3 vs 31.8) despite lower AHI, with higher REM AHI and lower supine AHI. Measured in: 2,827 consecutive sleep-clinic patients with polysomnography-confirmed OSA. What could explain it instead: Age and body mass index differed substantially between the sexes in this sample, and both independently produce fatigue, nocturia and mood symptoms, so the symptom profile is not cleanly attributable to sex alone.. A sleep-clinic sample describes women who were referred; women who never got a referral are absent by construction, which would if anything widen the real gap. The women were also older and heavier than the men, and some of the symptom differences could follow from that rather than from sex.

In 589 women with laboratory polysomnography, adjusted odds of an AHI of 15 or more were 3.5 times higher after menopause than before it (95% CI 1.4 to 8.8), and 2.6 times higher for an AHI of 5 or more (95% CI 1.4 to 4.8). In a separate two-phase population sample, clinically defined sleep apnea affected 0.6% of premenopausal women and 2.7% of postmenopausal women not taking hormone therapy, against 3.9% of men. Measured in: 589 women in the Wisconsin Sleep Cohort; and 1,000 women plus 741 men in a Pennsylvania population sample. What could explain it instead: Menopausal status travels with age and with weight gain, both of which independently raise apnea risk. The Wisconsin estimates are adjusted for age, body habitus and smoking, and statistical adjustment does not fully separate variables that move together this tightly. Menopausal status was assigned partly from self-reported vasomotor symptoms.. Both are cross-sectional, so this describes an association with menopausal status rather than a transition observed in the same women over time. The hormone therapy finding in the second study is observational, and women who take hormone therapy differ from women who do not in ways that also affect apnea risk.

Positive airway pressure compared with no treatment produced clinically significant reductions in disease severity, sleepiness, blood pressure and motor vehicle accidents, and clinically significant improvement in sleep-related quality of life. Auto-adjusting and bilevel modes gave no clinically significant advantage over standard CPAP; nasal masks gave better adherence and slightly greater severity reduction than oronasal masks. Measured in: 336 studies met inclusion, 184 provided data suitable for meta-analysis; adults with obstructive sleep apnea. The comparator is no treatment, so this measures what the therapy does when used rather than what it delivers across everyone prescribed it. Sleepiness gains are largest in people who were sleepy to begin with, and a large share of people who meet the AHI threshold are not. Trial adherence is generally better than adherence at home.

Across nine observational studies of drivers with obstructive sleep apnea compared before and after CPAP, crash risk fell to about a quarter of the pre-treatment rate (risk ratio 0.278, 95% CI 0.22 to 0.35). Daytime sleepiness improves significantly after a single night of treatment and simulated driving performance improves within two to seven days. Measured in: Drivers with obstructive sleep apnea across nine before-and-after observational studies. What could explain it instead: People starting CPAP have just been diagnosed with a condition they were told affects their driving, and diagnosis alone changes driving behavior independently of the therapy.. Every included study is a before-and-after comparison with no control group, so regression to the mean, changed driving behavior after a diagnosis, and the tendency to crash less in the period following any medical scare are all in the estimate. A risk ratio of 0.278 from uncontrolled data should be read as a clear direction rather than a precise size. No crash data were available to establish how quickly the reduction appears.

Obese adults with type 2 diabetes randomized to an intensive lifestyle intervention lost 23.8 lb (10.8 kg) at one year against 1.3 lb (0.6 kg) with diabetes support and education, and their AHI fell by an adjusted 9.7 events per hour more than the control group's. More than three times as many reached total remission of apnea, and severe apnea was half as prevalent in the intervention group. Weight loss of 22 lb (10 kg) or more gave the largest AHI reductions. Measured in: Obese adults with type 2 diabetes and objectively measured sleep apnea, the Sleep AHEAD sub-study of Look AHEAD. Everyone enrolled had both obesity and type 2 diabetes, so the result does not transfer to people with apnea from craniofacial anatomy at a normal weight. Remission was the exception rather than the rule: most participants still had apnea after losing 23.8 lb (10.8 kg). Lifestyle weight loss at this scale is difficult to maintain outside a trial that funds the coaching and the contact hours.

Apnea-hypopnea index fell by 20.0 and 23.8, about 20 to 24, events per hour against placebo across the two trials. Measured in: 469 adults with moderate to severe OSA and obesity, 142 women and 327 men, over 52 weeks. Placebo-adjusted, which is the comparison that matters; the within-arm changes are larger and are the numbers usually quoted. Sponsored by the manufacturer, entry BMI 30 or above, and 142 of 469 participants were women. A surrogate endpoint: nothing here measures symptoms or events.

126 participants, 83% men, mean age 54.5, mean BMI 28.4. Median AHI fell 68% from 29.3 to 9.0 events per hour at 12 months, and the oxygen desaturation index fell 70% from 25.4 to 7.4. In the randomized therapy-withdrawal phase, the group switched off went from 7.6 at 12 months back to a mean AHI of 25.8, while the maintenance group stayed at 8.9. Procedure-related serious adverse events under 2%. Measured in: 126 adults who could not accept or adhere to CPAP, with BMI under 32 and a favorable airway collapse pattern on drug-induced sleep endoscopy. What could explain it instead: Participants were selected for a favorable airway phenotype and then followed without a comparison group, so the response rate reflects the selection as much as the device.. The 12-month result comes from a single-group design with no control arm, so only the withdrawal sub-study is controlled. The eligibility criteria are the fine print: CPAP failure, BMI under 32, and a specific pattern of collapse on endoscopy exclude a large share of the people in whom apnea is diagnosed. 83% of participants were men. This is a surgically implanted device with a battery that eventually needs replacing.

102 adults with moderate or severe apnea in whom conventional treatment had failed, randomized to modified uvulopalatopharyngoplasty plus tongue volume reduction or to ongoing medical management. AHI went 47.9 to 20.8 with surgery and 45.3 to 34.5 with medical management, a baseline-adjusted between-group difference of -17.6 events per hour (95% CI -8.4 to -26.8). Epworth scores went 12.4 to 5.3 and 11.1 to 10.5, a difference of -6.7 points. Measured in: 102 adults, 18 of them women, mean age 44.6, all of whom had already failed device treatment. 102 people across a small number of centers, open-label with no sham surgery, so the 6.7-point Epworth difference carries whatever a major operation contributes by expectation. The mean AHI after surgery was still 20.8, which remains moderate apnea. Two of 51 in the surgical arm had a serious adverse event, one a myocardial infarction on day 5. Only 18% of participants were women.

Across 21 comparative epidemiological studies, higher alcohol consumption was associated with a 25% higher risk of sleep apnea (RR 1.25, 95% CI 1.13 to 1.38). The estimate held across different definitions of both exposure and outcome, and across study designs. In a separate eight studies, mean alcohol intake was two units a week higher in people with apnea, which was not statistically significant. Measured in: Adults across 21 observational studies, 1985 to 2015. What could explain it instead: Body weight is the dominant shared cause: heavier drinkers are heavier, and weight is the strongest modifiable risk factor for obstructive apnea. Smoking and shift work also cluster with drinking.. Heterogeneity was high (I2 82%) and Egger's test showed evidence of publication bias (p = 0.001), so treat the exact figure loosely and the direction as reasonably settled. All studies are observational: alcohol travels with weight, smoking and late bedtimes. The pooled estimate cannot separate a drink tonight worsening tonight's breathing from long-term drinking raising underlying risk, and the mechanism (reduced pharyngeal dilator muscle tone) supports both.

Across 82 papers reporting adherence between 1994 and 2015, the overall CPAP non-adherence rate was 34.1%, with no significant improvement over the twenty-year span. Behavioral intervention improved adherence by roughly one hour per night on average. Measured in: 82 studies of CPAP adherence in adults with obstructive sleep apnea, 1994 to 2015. Adherence was defined against a 7-hour sleep time, and definitions vary across the included studies, so the 34.1% figure is a synthesis of measurements that were not made the same way. Study populations are mostly sleep-clinic patients on structured follow-up, which is likely to be better supported than routine care. The null here is about the trend over time, not about whether any individual intervention helps.

145 adults with an AHI of 15 or more and comorbid insomnia were randomized to four sessions of cognitive behavioral therapy for insomnia before starting CPAP, or to treatment as usual. The CBT-I group averaged 61 more minutes of nightly CPAP use at six months (95% CI 9 to 113, d = 0.38) and higher initial acceptance of CPAP (99% vs 89%). Insomnia severity and dysfunctional sleep beliefs improved more in the CBT-I group. Measured in: 145 adults with moderate to severe obstructive sleep apnea and co-occurring insomnia. One trial of 145 people, unblindable by design, with the CBT-I group receiving four extra sessions of clinician contact that the control group did not. There were no between-group differences in sleep outcomes or daytime impairment at six months, so what was shown is better use of the device rather than a better outcome from it.

In a network meta-analysis of 80 randomized trials, CPAP lowered the apnea-hypopnea index most, by 25.27 events per hour against control (95% CI 22.03 to 28.52). A mandibular advancement device lowered it as well but ranked below CPAP, which reduced the index about 10 events an hour more than the device (between-treatment difference 10.06, 95% CI 5.91 to 14.21). Daytime sleepiness on the Epworth Sleepiness Scale did not differ significantly between the device and CPAP, which fits the pattern that an appliance worn all night can match a stronger therapy worn part of the night. The comparison runs partly through indirect links in the network rather than head to head throughout, and the device suits mild to moderate apnea and CPAP-intolerant patients more than severe disease.

Eight randomized studies, 323 participants. Against inactive control, positional therapy reduced AHI by 7.38 events per hour (95% CI 4.7 to 10.06, low certainty) and Epworth scores by 1.58 points (95% CI 0.29 to 2.89, moderate certainty). Against CPAP, CPAP reduced AHI by a further 6.4 events per hour (95% CI 3.00 to 9.79, low certainty), while subjective adherence favored positional therapy by 2.5 hours a night (moderate certainty). Epworth scores did not differ between CPAP and positional therapy. Measured in: 323 participants, mostly with supine-predominant obstructive sleep apnea; devices ranged from vibration alarms to backpacks and pillows. Cochrane rated the AHI comparisons low certainty and the trials are small and short. The devices pooled are not equivalent to each other: a vibrotactile alarm and a tennis ball sewn into a shirt are different interventions. Adherence data came from one study and was self-reported. This only applies to people whose events are supine-predominant, and none of the trials followed people long enough to say whether the devices are still worn after a year.

Across nine adult studies and 120 patients, AHI fell from a mean of 24.5 to 12.3 events per hour (mean difference -14.26, 95% CI -20.98 to -7.54), about 50%. Lowest oxygen saturation improved from 83.9% to 86.6%, snoring fell from 14.1% to 3.9% of total sleep time, and Epworth scores fell from 14.8 to 8.2. Measured in: 120 adults across nine studies, plus 25 children across two studies. What could explain it instead: Before-and-after designs in a condition whose severity varies night to night will overstate benefit, since people enrol during a bad period and are re-measured later.. Nine studies and 120 adults, most of them before-and-after with no control arm, so regression to the mean and the effect of being observed are not excluded. A 50% fall from a mean AHI of 24.5 still leaves 12.3, which is mild to moderate apnea rather than resolution. The therapy protocols differed between studies and adherence to a daily exercise regimen over months is its own problem.

Nine randomized trials, 584 participants, covering manual acupuncture and electroacupuncture. AHI fell by 6.18 events per hour (95% CI -9.58 to -2.78) and Epworth scores by 2.84 points (95% CI -4.80 to -0.16). Lowest oxygen saturation improved by 5.29 percentage points. Larger AHI reductions were seen in the moderate (-9.44) and severe (-10.09) subgroups. Measured in: 584 adults with obstructive sleep apnea across nine trials, most conducted in China. The authors rate the quality of evidence low to very low. The sleepiness result reversed direction under sensitivity analysis, which means it does not survive removing the weakest studies. Most included trials come from Chinese-language databases where acupuncture trials report positive results at an unusually high rate. A 6-point fall in AHI does not move someone with severe apnea out of the range where airway treatment is indicated.

58 randomized trials, 4,590 participants. Against placebo, Chinese herbal medicine reduced AHI by 7.10 events per hour (95% CI -11.95 to -2.25) across 7 studies and 583 participants. Added to CPAP, it reduced AHI by a further 4.71 events per hour against CPAP alone (95% CI -5.62 to -3.80) across 28 studies and 2,267 participants. Sleepiness, quality of life, body weight, oxidative and inflammatory markers, cognition and blood pressure also improved. Measured in: 4,590 adults with obstructive sleep apnea, the great majority in trials from Chinese-language databases. Only 7 of the 58 trials had a placebo comparison, and the 51 that did not are the ones carrying most of the participants. The formulas differ from study to study, so the pooled figure describes a category of treatment rather than a prescription, and nothing in it identifies which pattern a given formula suited. Trials from this literature carry known risks of selective reporting and inadequate allocation concealment. The effect size is smaller than any device produces.

Across 51 trials and 4,888 patients, CPAP lowered systolic blood pressure by 2.5 mmHg (95% CI 1.5 to 3.5) and diastolic by 2.0 mmHg against inactive control. Mandibular advancement devices lowered systolic by 2.1 mmHg and diastolic by 1.9 mmHg, with no significant difference between the two treatments. Each additional hour of nightly CPAP use was associated with a further 1.5 mmHg systolic reduction. Measured in: 4,888 patients across 51 randomized trials of CPAP or mandibular devices in obstructive sleep apnea. 2.5 mmHg is a small average, roughly a third of what a first-line antihypertensive delivers, and the dose-response with hours of use means the average hides people who got considerably more and people who got nothing. Most included trials were short, so this does not establish that the reduction persists for years. The two treatments matching is likely to be reduced airway benefit from the device offset by more hours of use.

2,717 adults aged 45 to 75 with moderate to severe apnea and established coronary or cerebrovascular disease were randomized to CPAP plus usual care or usual care alone. Mean AHI fell from 29.0 to 3.7 events per hour. Over 3.7 years the primary composite (cardiovascular death, myocardial infarction, stroke, hospitalisation for unstable angina, heart failure or TIA) occurred in 17.0% on CPAP and 15.4% on usual care, hazard ratio 1.10 (95% CI 0.91 to 1.32). Snoring, daytime sleepiness, quality of life and mood all improved significantly. Measured in: 2,717 adults with moderate to severe OSA and established cardiovascular disease, across seven countries. Mean CPAP use was 3.3 hours a night, below the 4 hours conventionally used to define adherence, so this is a test of CPAP as people actually used it. Participants were selected for minimal sleepiness, because randomizing a sleepy patient to no treatment for four years is not permissible, and sleepiness may be the marker of who benefits. This is secondary prevention in people who already had cardiovascular disease, not a statement about untreated apnea in the general population.

Ten randomized trials, 7,266 participants, 80.5% men, mean age 60.9. Positive airway pressure showed no significant association with major adverse cardiovascular events (RR 0.77, 95% CI 0.53 to 1.13), cardiovascular death (RR 1.15, 95% CI 0.88 to 1.50) or all-cause death (RR 1.13, 95% CI 0.99 to 1.29). Meta-regression found no association between outcome and apnea severity, follow-up duration or adherence to the device. Measured in: 7,266 participants across 10 trials (9 CPAP, 1 adaptive servo-ventilation), predominantly men with existing cardiovascular disease. The absence of an adherence signal in meta-regression is the part that complicates the usual explanation for these nulls. The confidence interval on major events (0.53 to 1.13) is wide enough to include a meaningful benefit as well as no effect, so this is an unresolved question rather than a settled negative. Every included trial enrolled people with existing cardiovascular disease and low sleepiness.

In 709 Wisconsin Sleep Cohort participants with baseline polysomnography, adjusted odds of hypertension at four years rose with baseline AHI: 1.42 (95% CI 1.13 to 1.78) at 0.1 to 4.9 events per hour, 2.03 (1.29 to 3.17) at 5.0 to 14.9, and 2.89 (1.46 to 5.64) at 15 or more, relative to no events. Adjusted for baseline hypertension, body mass index, neck and waist circumference, age, sex, alcohol and cigarettes. Measured in: 709 adults in the Wisconsin Sleep Cohort followed four years, 184 of them for eight. What could explain it instead: Obesity causes both sleep apnea and hypertension, and it is the standing objection to this whole literature. Adjusting for body mass index plus neck and waist circumference is the strongest available response and leaves residual confounding by body composition and visceral fat distribution untouched.. Observational with a four-year window, so this establishes a dose-response association rather than causation. The dose-response and the adjustment for three separate measures of body habitus are what make it more persuasive than the cross-sectional literature it replaced, and neither removes the objection.

The CPAP treatment effect was greater in high-risk apnea, defined by heart-rate response and hypoxic burden, than in low-risk apnea: interaction hazard ratio 0.69 (95% CI 0.50 to 0.95). The same analysis reports HARM in low-risk apnea, with the authors concluding the harm may counteract the benefit. The 0.69 is the differential effect between the two groups, not the within-high-risk-group treatment hazard ratio. Measured in: 3,549 participants pooled from three randomized CPAP cardiovascular outcome trials. Both halves come from one post-hoc analysis with tertiles defined after the fact, so this is a hypothesis about who benefits rather than a demonstration. Quoting the favorable half alone would misrepresent the paper.