Who Responds Best to a MAD? What Predicts Success With Oral Appliance Therapy

By Dr. Boris Zusin · Published August 30, 2026

Not everyone with obstructive sleep apnea responds equally well to a mandibular advancement device. Research has identified a fairly consistent cluster of traits associated with a better response — but it's worth being upfront about the honest limit here: even the best-known predictors only weakly forecast how any one individual will actually do, and roughly a third of patients see little to no meaningful improvement despite a well-fitted appliance. That's exactly why an objective efficacy sleep study, not a prediction on paper, is what ultimately confirms whether a MAD is working for you.

Demographic and Anthropometric Predictors

Across the largest available meta-analyses, responders to MAD therapy tend to be younger, female, and have a lower BMI and smaller neck circumference than non-responders — severe obesity in particular is a recurring unfavorable feature. Milder disease also predicts success: a lower baseline AHI (mild-to-moderate OSA) is associated with a better response, which lines up with current guidelines recommending MADs first-line for mild-to-moderate OSA and for patients who can't tolerate CPAP. That said, severe OSA can still respond well, particularly in patients with a lower BMI. One more marker worth knowing: patients who need a lower therapeutic CPAP pressure tend to do better with a MAD, while a CPAP pressure above roughly 10.5 cmH₂O — a sign of a more collapsible airway — is associated with a poorer MAD response.

Craniofacial and Anatomic Predictors

Certain jaw and airway shapes show up repeatedly in responders: a retracted (set-back) upper and lower jaw, a narrower airway, a shorter soft palate, and a smaller minimum airway cross-sectional area. In one pilot study, the strongest single predictor on cephalometric X-rays was the degree of jaw retrognathia. Vertical facial pattern also changes how much advancement is actually needed — patients with a low-angle (flatter) facial pattern often reach a meaningful AHI reduction at as little as 20% of their maximum jaw protrusion, while patients with a high-angle (steeper) facial pattern typically need to be advanced much further, often 50% or more, and may still hit a ceiling of benefit beyond 60–70% no matter how much further the device is advanced.

Physiologic and Sleep-Study Predictors

Beyond anatomy, how your breathing control system behaves during sleep matters too. Patients with low “loop gain” — a measure of how unstable your breathing drive is — respond more consistently than patients with high loop gain, even after accounting for AHI and BMI. Airway collapsibility follows the same pattern: mild-to-moderate collapsibility with good muscle compensation favors response, while a highly collapsible airway predicts a poorer one. OSA that's worse lying on your back (positional OSA) responds meaningfully better than non-positional OSA — and for patients with a residual supine-dependent component even on a well-fitted MAD, adding positional therapy can close that gap — while OSA that's concentrated in REM sleep — a pattern with its own distinct risk profile, discussed in our post on REM-predominant sleep apnea — or driven mainly by very low oxygen dips, tends to respond less well. On a sleep endoscopy exam, a collapse pattern centered at the base of the tongue has been associated with a good MAD response specifically. Interestingly, your lowest oxygen saturation before treatment turns out to be a much stronger predictor of whether oxygenation will improve than your overall AHI is.

Imaging and Dental Predictors

On 3D airway imaging (CBCT), a higher starting OSA severity paired with a narrowing located higher up in the airway, and an obstruction that traces more to jaw/facial skeletal proportions than to soft tissue alone, has been associated with a better response — while flat 2D cephalogram measurements alone haven't reliably predicted outcomes. A narrower distance between the lower back teeth has also correlated with treatment success in one small phenotyping study.

The Strongest Signal Yet: What the Airway Actually Does When It Collapses

The most reliable predictors identified so far don't come from a still X-ray at all — they come from watching the airway collapse in real time, whether on sleep endoscopy (DISE) or dynamic imaging during sleep. The clearest physiologic phenotype combines two things: an obstruction centered at the base of the tongue, plus an airway that isn't excessively collapsible to begin with. In one study built specifically around this combination, predicted responders saw their AHI drop by about 83%, compared to roughly 48% in predicted non-responders — a considerably wider gap than demographic factors alone produce. On DISE specifically, tongue-base collapse more than triples the odds of a good response, while certain other collapse patterns point the other way: complete concentric collapse at the level of the palate, collapse involving the epiglottis, and side-to-side (rather than front-to-back) collapse in the throat all predict a poorer response or, in some cases, that a MAD could make things worse. That last point matters clinically — a MAD reliably opens the airway at the palate, throat, and tongue base, but it doesn't reliably help at the epiglottis, so a collapse pattern that's concentrated there is a signal to consider a different or combined approach rather than assuming an appliance will resolve it. Dynamic MRI adds a further layer: patients whose tongue moves forward more, and whose airway opens up more, for each millimeter the jaw is advanced tend to do better, and combining that movement data with baseline AHI has correctly classified responders and non-responders in roughly 7 out of 10 cases — still imperfect, but meaningfully better than guessing from anatomy alone.

Does the Appliance's Own Design Matter as Much as Your Anatomy?

Given how much of the story above comes down to individual anatomy and airway behavior, it's a fair question whether the appliance itself — its design, not just how far it advances the jaw — changes the odds. The short answer is: less than you'd think. The clearest design-related finding is that custom-fitted appliances, made from an impression or scan of your own teeth, outperform prefabricated, boil-and-bite-style devices — modestly better AHI reduction, meaningfully better adherence, and fewer side effects, which is exactly why custom fabrication is standard here. Whether a device is titratable (adjustable in small increments after delivery) versus fixed at a single setting hasn't shown a significant difference in AHI reduction on its own, though titratability still matters practically, since it's what allows the jaw position to be fine-tuned to each patient's individual dose-response curve. Beyond that, head-to-head comparisons of specific mechanical designs — one-piece (monobloc) versus two-piece (bibloc) devices, or different coupling mechanisms between the upper and lower pieces — show some differences on individual measures (one-piece designs, for instance, have ranked somewhat better for minimum oxygen levels and for AHI specifically while lying on the back, likely because they allow less mouth opening and less of the jaw rotating backward overnight, but at the cost of more jaw-muscle soreness and no ability to adjust the fit chairside), but a broader analysis comparing appliance types found no single design comes out ahead across the board. In practice, this means device design is a secondary lever — worth choosing thoughtfully once other factors point toward likely candidacy, but not something that overrides an unfavorable anatomic or physiologic phenotype.

The Honest Caveat: Why We Still Rely on a Trial, Not a Prediction

Even with all of the above, validation studies of pre-treatment prediction tools — cephalometric X-rays, sleep endoscopy, nasopharyngoscopy, CPAP pressure requirements — have generally been disappointing and inconsistent from study to study. The best-performing tool identified so far involves remotely adjusting jaw position during an actual sleep study to directly observe the response, rather than inferring it indirectly from anatomy or demographics. In practical terms, that means the field still relies on a real therapeutic trial with objective confirmation (a follow-up sleep study with the appliance in place) rather than predicting success from your exam alone — which is exactly why we build that confirmation step into every course of MAD therapy rather than treating the initial fitting as the finish line.

If you're wondering whether your own anatomy and sleep study findings make you a good candidate, that's a conversation worth having at a dental sleep evaluation — we'll walk through what your case suggests, while being clear that the real answer comes from how you respond once treatment starts. For more on why sleep apnea itself varies so much from person to person — beyond just the traits that predict MAD response — see our overview of OSA phenotypes.