SANSA HSAT

Performance Metrics

Clinical Evidence

The evidence comes from two separate studies. The pivotal one is a prospective, multicentre validation (NCT06070389) published in Frontiers in Neurology (June 2025). An earlier algorithm-development study appeared in the Journal of Clinical Sleep Medicine (May 2025). Across seven American Academy of Sleep Medicine accredited laboratories, adults referred for polysomnography (PSG) for suspected obstructive sleep apnoea wore SANSA for the same night. Three registered technologists scored each PSG and their consensus was the reference; SANSA's output was the raw automated score with no technologist review. Of 441 enrolled, 340 were analysed after pre-specified exclusions. Reincluding the 35 excluded records that had both scores did not change the results. The pre-specified primary endpoint was moderate-to-severe OSA (AHI of 15 or more) scored with the 4 per cent desaturation rule (Rule 1B): sensitivity 88 per cent (95 per cent CI 81 to 93) and specificity 87 per cent (82 to 91), with the study powered to show both above 80 per cent. Site PSG scorers, measured against the same consensus, reached 89 and 93 per cent; neither difference was significant. The site scorers did agree more closely on the AHI itself: their Rule 1B correlation was 0.96, significantly higher than SANSA's 0.91, with narrower limits of agreement. Under the 3 per cent or arousal rule (Rule 1A) SANSA's sensitivity at the same threshold was 93 per cent, higher than the site scorers' 79 per cent, but specificity fell to 74 per cent against their 85 per cent, and both differences were significant. At the AHI 5 threshold specificity was 71 per cent (Rule 1B) and 55 per cent (Rule 1A); at AHI 30 it was 95 and 89 per cent with sensitivity of 82 and 87 per cent. AHI correlation with PSG was 0.91 (Rule 1B) and 0.90 (Rule 1A), with limits of agreement of about plus or minus 19 events per hour under Rule 1B and minus 20 to plus 24 under Rule 1A. Total sleep time correlated at 0.82, but SANSA overestimated it by 20 minutes on average with limits of agreement from minus 81 to plus 122 minutes, and its sleep-epoch specificity was 63 per cent against 95 per cent sensitivity. The device recognises sleep readily and wake poorly, which inflates the AHI denominator in patients who lie awake. The earlier JCSM paper is a separate feasibility study used to build the algorithm. It enrolled 164 people at five sleep laboratories and analysed 153. The model was trained on 86 records and tested on 67. Across the severity thresholds, sensitivity was 91 to 97 per cent and specificity 78 to 97 per cent for the whole dataset. Site scorers had significantly higher specificity and accuracy at the AHI 5 threshold. In that version only the sleep/wake step used machine learning; apnoea events were scored by fixed rules. Its second finding is not about apnoea detection. A cardiac technician contracted by Huxley read the stored ECG and found significant arrhythmias, such as frequent ectopy or atrial fibrillation, in 32 per cent of participants. They had a higher AHI (27.5 against 15.8 events per hour) and more time below 88 per cent oxygen saturation (47.5 against 20.5 minutes). The cardiac indication added in October 2025 rests on a different test: a comparison with a simultaneous Holter recording, published without figures. The December 2025 clearance (K252497) added an obstructive-versus-central classifier and REM staging, validated on 325 records from a multisite in-lab study. The summary does not say when those recordings were made; the cohort closely resembles the Frontiers cohort, with the same mean age of 55.4. Twelve patients had a central AHI of 10 or more on PSG; SANSA found all twelve (sensitivity 100 per cent, 95 per cent CI 73.5 to 100) with specificity of 98.7 per cent (96.8 to 99.7). REM-versus-non-REM classification at epoch level reached 74.6 per cent sensitivity and 89.5 per cent specificity. The central-apnoea figure is a twelve-patient result and neither classifier has a peer-reviewed publication. The same submission also updated the core algorithm, but its summary reports no new AHI figures. Both papers were funded by Huxley Medical together with the US National Science Foundation and the Georgia Research Alliance. Each has a Huxley employee as co-author, and several investigators received advisory fees from Huxley. There is no independent validation, no study in the home setting, and no multi-night or outcome data. The FDA comparison table lists the target population as adults aged 22 and over while the studies enrolled from 18. AurisMark reasoning: FDA clearance across four submissions, a peer-reviewed prospective multicentre validation of this device with a pre-specified primary endpoint, and a clean safety record meet the two-mark bar. A third mark is withheld because there is no verifiable approval outside the United States, no randomised or outcome evidence, and no independent or real-world data. Re-evaluate on an independent home-setting study or a non-US register entry.

Inside the Auris+ Listing

Regulatory Approvals

Safety Record

No recall, enforcement action or MAUDE adverse-event report naming SANSA or Huxley Medical was found in the openFDA device databases, queried directly on 13 September 2026. The validation study recorded two skin reactions to the adhesive, without sequelae. The clinical risk sits in the operating points: at the AHI 5 threshold roughly three in ten patients without disease (Rule 1B) and more than four in ten (Rule 1A) are scored positive, and sleep-time overestimation can push a borderline AHI below a treatment threshold. The ECG channel is a recording, not a monitor: no arrhythmia is flagged automatically, and the labelling excludes pacemaker patients and third-party ECG algorithms.

Intended Use & Indications

Per the FDA indications for use (K252497): The Huxley Home Sleep Apnea Test (SANSA) is a wearable device intended for use in the recording, analysis and storage of biophysical parameters to aid in the evaluation of sleep-related breathing disorders of adults suspected of sleep apnea. The device is intended for the clinical and home use setting under the direction of a Healthcare Professional (HCP). The parallel October 2025 clearance (K250882) extends the wording to "sleep-related breathing and cardiac disorders" and adds that the device records and stores an ECG for up to 10 hours for manual annotation and analysis, provides no automated analysis of the ECG, is not intended for use with a third-party automated algorithm and is not intended for pacemaker analysis. Prescription use only.