Hidden Sensor Reveals Sleep Apnea Patients Overestimate Their Own Treatment Use

A tiny thermal sensor hidden inside a dental device has exposed a striking gap between what patients with obstructive sleep apnea believe they are doing and what they actually do at night. In a one-year longitudinal study conducted at Peking University School and Hospital of Stomatology, researchers tracked forty-nine adults with mild to moderate obstructive sleep apnea who were fitted with custom-made mandibular advancement devices, the mouthguard-like oral appliances that hold the lower jaw forward to keep the airway open during sleep. By embedding a microsensor in the appliance that recorded temperature at fifteen-minute intervals, the team could tell precisely when the device was in the warm environment of the mouth and when it sat unused in a drawer. The results, published in the Journal of Clinical Sleep Medicine, suggest that self-reported adherence to oral appliance therapy is substantially inflated, and that the patients most likely to abandon treatment can be identified within the first weeks of therapy.

The clinical stakes are considerable. Obstructive sleep apnea affects hundreds of millions of people worldwide and is linked through sleep fragmentation and repeated drops in blood oxygen to cardiovascular disease, metabolic dysfunction, and daytime impairment. Continuous positive airway pressure, or CPAP, remains the gold standard treatment, but between 46 and 83 percent of patients fail to meet the conventional adherence benchmark of more than four hours of use per night. Oral appliances, particularly mandibular advancement devices, are recommended as a first-line option for mild to moderate disease and as an alternative for patients who cannot tolerate CPAP. Earlier studies suggested that patients wear these appliances roughly an hour longer per night than they use CPAP, but much of that evidence rested on self-reports, which are notoriously prone to overestimation.

The study’s technical approach was elegantly simple. The TheraMon sensor, embedded in the posterior buccal side of the mandibular arch tray, was programmed to log temperature readings between 33.5 and 39.5 degrees Celsius. Because the oral cavity sits within that range and room temperature does not, continuous readings within the window were interpreted as periods of actual device use. Participants were not explicitly told that nightly appliance use was being objectively monitored, nor were they informed of the details of the one-year follow-up plan, a deliberate design choice intended to preserve natural patterns of use. At baseline, each participant underwent overnight polysomnography, sleep questionnaires including the Epworth Sleepiness Scale and the Pittsburgh Sleep Quality Index, a five-factor personality assessment, and cone-beam computed tomography to characterize craniofacial morphology.

The therapy itself worked well when the device was actually worn. Mandibular advancement was initially set at 70 percent of each patient’s maximum protrusion and titrated in half-millimeter increments when needed, and the mean apnea-hypopnea index fell from 16.6 events per hour at baseline to 4.16 events per hour with the appliance in place. But the adherence data told a more complicated story. During the first three months, the median nightly use across all nights was just 3.85 hours, below the four-hour threshold used to define adequate adherence. When patients did wear the device, they wore it for a median of 7.05 hours, a full night of sleep. The problem was frequency: patients used the appliance on a median of only 3.7 days per week, and only 55.7 percent of nights met or exceeded the four-hour mark.

The comparison between objective and subjective measures revealed a specific and revealing pattern of bias. Participants were statistically accurate when reporting how long they wore the device on nights they used it, reporting a median of 7.0 hours against an objective 7.05 hours. But they significantly overestimated how often they used it, claiming a median of five days per week when sensors recorded 3.7 days, a difference that was highly significant. Bland-Altman analysis showed that self-reported nightly use across all nights exceeded objective values by a mean of 0.92 hours, and self-reported usage days exceeded objective records by an average of 1.41 days per week. The researchers attribute this to a combination of recall bias and what they describe as white coat adherence: a temporary uptick in device use shortly before scheduled clinic visits, which then dominated patients’ memories of their own behavior.

Even more striking was the divergence in trajectories between adherent and non-adherent users. Of the forty-three participants who completed the three-month follow-up, twenty-one, or 48.8 percent, met the adherence criterion. The adherent group wore the device a median of 6.23 hours per night across all nights and used it on roughly six days per week, with stable, sustained use throughout the observation period. Non-adherent users averaged just 2.38 hours per night on 2.32 days per week, and their weekly use fluctuated, declining when unsupervised and spiking before appointments. Notably, on the nights when both groups actually wore the appliance, the duration of use was statistically indistinguishable, at roughly seven hours. The difference between success and failure was not how long patients slept with the device, but whether they chose to put it in at all.

By the twelve-month follow-up, conducted by telephone with all forty-three completers, the split had hardened. Nineteen participants, or 44.2 percent, remained adherent, and their habits had become exemplary: a median of seven hours per night on seven nights per week. Most non-adherent users, by contrast, had effectively discontinued therapy, with median use of zero hours and zero days per week. Multivariate logistic regression identified two independent predictors of twelve-month adherence: older age, with each additional year raising the odds by about 7 percent, and a smaller mandibular plane angle measured relative to the Frankfort horizontal plane. Patients with hyperdivergent, steeply angled facial skeletons, a craniofacial type already known to be a risk factor for sleep apnea and to require greater mandibular advancement for efficacy, were significantly less likely to keep using the device.

The three-month picture had its own predictors. Older age again emerged as protective, alongside lower baseline Epworth Sleepiness Scale scores, with each point of sleepiness reducing the odds of adherence by roughly 26 percent. Spearman correlations showed that sleepiness scores were negatively associated with every objective adherence measure, including nightly use, weekly frequency, and the percentage of nights exceeding four hours. The authors speculate that patients with excessive daytime sleepiness may initially experience limited symptomatic improvement, only encountering positive feedback later as accumulated use eases their symptoms, and they suggest that greater mandibular advancement combined with enhanced early support might help this subgroup. Personality traits measured by the NEO-FFI, including conscientiousness and neuroticism, showed no significant association with adherence, indicating that behavior was shaped more by clinical and anatomical factors than by stable psychological disposition.

Several caveats frame the findings. Objective sensor data were collected only during the first three months, so the twelve-month comparison relied on self-report, the very measure shown to be unreliable. Side effects and discomfort were not systematically recorded, sleep diaries were not used, and the cohort consisted exclusively of Chinese patients with a mean age of about thirty-nine, limiting generalizability to other populations and to the very elderly. The authors also note that generally shorter sleep duration in the Chinese population may partly explain the relatively low nightly wear time, and that adherence to this particular appliance design should not be assumed to generalize to all oral appliances.

Even so, the study carries a clear practical message. Adherence behavior in oral appliance therapy, as in CPAP therapy, appears to be established early: patients who went on to succeed had already formed stable wearing habits within the first week of treatment. Because the low overall adherence in this cohort was driven primarily by infrequent use rather than short wear sessions, the authors recommend embedding reminders, whether through alarms, mobile applications, or email, into routine care, citing prior evidence that app-based monitoring significantly increases wearing days. They argue for long-term, objective, and individualized adherence monitoring as a standard component of oral appliance management, particularly if these devices are to deliver the cardiovascular benefits that have proven elusive in CPAP trials. For clinicians, the takeaway is that the first three months, and perhaps the first few weeks, offer a critical window to identify struggling patients and intervene before quiet abandonment of therapy becomes permanent.

Subject of Research: Objective versus subjective adherence to oral appliance therapy in adults with obstructive sleep apnea

Article Title: Objective and subjective adherence to oral appliance therapy in adults with obstructive sleep apnea: pilot data from a one-year longitudinal study

Article References: Zhang, W., Yu, M., Gong, X., Dang, W., & Gao, X. (2026). Objective and subjective adherence to oral appliance therapy in adults with obstructive sleep apnea: pilot data from a one-year longitudinal study. Journal of Clinical Sleep Medicine, 22(1), Article 115. https://doi.org/10.1007/s44470-026-00123-6

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00123-6

Keywords: obstructive sleep apnea, oral appliance therapy, mandibular advancement device, adherence, thermal microsensor, objective monitoring, self-report bias, craniofacial morphology, Epworth Sleepiness Scale, CPAP, sleep medicine, longitudinal study