How to Read Oral Appliance Research

A case series follows a group of patients treated the same way and reports what happened. It can generate hypotheses and reveal side effects, but it has no comparison group — so improvements might reflect the treatment, natural fluctuation, or expectation. Observational studies compare groups who chose different treatments, but because those groups usually differ in other ways too, the comparison is never fully clean.

In a randomized controlled trial (RCT), patients are assigned to treatments by chance — a coin flip, effectively. Randomization is powerful because it balances out the differences that would otherwise bias results: patients in each arm should be comparable in severity, age, and health. The comparison group (the "control") may receive another active treatment, a sham device, or no treatment. RCTs are the workhorse of treatment evidence, and the 2015 AASM/AADSM guideline recommendations rest on a systematic review of them.

A systematic review uses predefined criteria to find and evaluate all studies on a question; a meta-analysis statistically pools their results. These summaries sit at the top of the evidence hierarchy because they reduce the distorting effect of any single study. But they inherit the limitations of the studies they pool — a meta-analysis of small, short trials is still built from small, short trials.

Blinding means keeping participants — and ideally the assessors — unaware of which treatment is being received, so that expectation cannot color the results. Drug trials can use identical placebo pills; oral appliance trials cannot fully hide an appliance in the mouth. Some studies use sham (inactive) dental devices as controls, which helps, but participants can often tell the difference.

This is one reason the type of endpoint matters. Subjective outcomes, such as self-reported sleepiness, can be influenced by expectation when blinding is imperfect. Objective outcomes recorded by equipment — such as breathing events and oxygen levels on a sleep study, explained in AHI, REI and RDI — are far harder to influence by belief. When you read a trial, check both: the strongest studies pair subjective improvement with objective measurement, the approach recommended in our sleep testing center.

A noninferiority trial is designed around a different question than a conventional trial. Instead of asking "Is treatment A better than treatment B?", it asks "Is treatment A not meaningfully worse than B — within a margin the researchers defined in advance?" If the result stays inside that margin, the treatment is declared noninferior: acceptable as an alternative, not proven superior.

The CRESCENT trial, published in the Journal of the American College of Cardiology in 2024, illustrates the concept. Investigators recruited adults over 40 with hypertension and increased cardiovascular risk; of these, 220 with moderate-to-severe obstructive sleep apnea were randomized to a mandibular advancement device or CPAP. The question was narrow and prespecified: does the oral appliance reduce 24-hour mean arterial blood pressure over six months without being meaningfully worse than CPAP, within a margin of 1.5 mm Hg? It did — blood pressure fell in the oral appliance group, and the result stayed within the margin, meeting the noninferiority definition.

Reading it correctly means honoring its boundaries: the trial lasted six months, its endpoint was blood pressure rather than cardiovascular events, and its population was adults over 40 with hypertension and increased cardiovascular risk. Noninferior on that endpoint, in that population, for that duration — nothing more. Our cardiovascular evidence guide places this trial in its full context.

A confidence interval (CI) is the range of values within which the true effect plausibly lies. When a meta-analysis in JAMA (2015) reported that mandibular advancement devices were associated with an average systolic blood-pressure reduction of about 2.1 mm Hg versus no treatment, it reported a 95% CI of roughly 0.8 to 3.4 mm Hg: the true average reduction could plausibly be as small as 0.8 or as large as 3.4, with the best estimate at 2.1.

Two habits follow. A narrow interval means the estimate is precise; a wide interval means real uncertainty remains. And when the confidence intervals of two treatments overlap substantially, a "our number was slightly bigger" comparison is not meaningful — as in that same analysis, where the difference between oral appliances and CPAP was not statistically significant. For how such blood-pressure findings should (and should not) be interpreted, see oral appliances and cardiovascular evidence.

Efficacy is how a treatment performs under ideal trial conditions — selected patients, close monitoring, protocol-driven adjustment. Effectiveness is how it performs in ordinary life. Adherence is how much the treatment is actually used — an appliance only works on the nights it is worn, which is why our adherence guide treats nightly use as the hinge between efficacy and effectiveness. Symptoms are what a patient feels; objective sleep-study outcomes are what equipment measures. A treatment can feel successful while breathing events persist, which is why symptom improvement alone cannot confirm efficacy.

When a study reports improvement, identify which of the five is being reported. An improvement in self-reported sleepiness is a meaningful finding about symptoms — but it is a different claim from a reduction in the apnea-hypopnea index measured on a sleep study. The strongest treatment evidence reports both, and the strongest treatment plans verify both. See how to know if your oral appliance is working and follow-up sleep testing.

Some studies define success as the apnea-hypopnea index falling below a normal-range threshold. Others count a percentage reduction from baseline as success. Others include symptom criteria, or combine objective and subjective measures. The same group of patients can produce a high success rate under one definition and a modest one under another — which is why published rates cannot be compared across studies without matching their definitions, and why our success rates guide explains definitions before numbers. When you read any headline rate, find the definition first.

A result applies to the people studied. CRESCENT enrolled adults over 40 with hypertension and increased cardiovascular risk; its findings travel best to similar patients. Trials conducted mostly in mild sleep apnea should be applied cautiously to severe disease — our guide to outcomes by sleep apnea severity shows how strongly severity shapes results. Before borrowing any result, ask whether you share the studied population’s defining features; when in doubt, treat the finding as a conversation-starter with your provider, not a conclusion about you.

This guide teaches research literacy. It does not diagnose conditions, evaluate any individual’s treatment, or recommend a specific therapy. Reading research well can make you a better-informed partner in care; it does not substitute for evaluation, diagnosis, and treatment decisions by qualified medical and dental providers. Never start, stop, or modify any treatment — including appliance adjustment — based on research you have read, without professional guidance.

American Academy of Sleep Medicine and American Academy of Dental Sleep Medicine — joint clinical practice guideline, built on a systematic review of the oral appliance literature (Journal of Clinical Sleep Medicine, 2015).

The CRESCENT randomized noninferiority trial of mandibular advancement device versus CPAP (Journal of the American College of Cardiology, 2024), used here as the worked example of a noninferiority design.

Bratton DJ, et al. — CPAP vs mandibular advancement devices and blood pressure in obstructive sleep apnea: a network meta-analysis of 51 randomized trials (JAMA, 2015), used here as the worked example of confidence intervals.

For how this site selects and applies its sources, see Evidence & Sources and the Medical Content Policy.