HFNC vs. CPAP for Sleep Apnea: What a Small Trial Found

High-flow nasal cannula (HFNC) can reduce obstructive sleep apnea, but the only controlled adult trial found that it controlled breathing events far less completely than CPAP and worsened sleep architecture. That makes HFNC an experimental fallback, not a substitute for a treatment that a patient can tolerate and use.

The appeal is obvious. CPAP is highly effective when patients use it, yet long-term adherence remains a serious problem in sleep medicine. A 2023 prospective cohort study described poor adherence as a critical issue and found that sleep improvements during titration predicted later positive-airway-pressure use. A 2023 systematic review in people with neurodegenerative diseases also found that adherence barriers have several causes. But the often-repeated claim that roughly half of patients abandon CPAP within a year is not quantified by these studies. It should not be treated as a precise estimate.

Still, the practical question is reasonable: could a less effective device help people who truly cannot sustain CPAP? The current evidence does not answer that question. It does show the size of HFNC's performance gap.

What the randomized trial actually tested

In a 2025 open-label randomized crossover trial, Wali and colleagues enrolled 68 treatment-naive adults with confirmed obstructive sleep apnea. Each participant underwent one in-laboratory titration night with HFNC and one with CPAP. HFNC flow ranged from 10 to 60 liters per minute and was titrated in the sleep laboratory.

HFNC reduced the apnea-hypopnea index (AHI) by a median 52% from baseline, with an interquartile range of 18% to 77%; the within-treatment reduction was statistically significant (p<0.001). Its residual AHI was 16.6 events per hour. CPAP produced a residual AHI of 5.8 events per hour, about 2.9 times lower than HFNC, with a statistically significant between-treatment difference (p<0.001).

AHI is a count of breathing disruptions per hour of sleep. It is an important treatment measure, but it is not the same as daytime alertness, blood-pressure control, cardiovascular outcomes, or quality of life. The trial did not measure those patient-centered outcomes. Even so, the residual AHI difference matters: a value below 5 events per hour is commonly used as a near-normal treatment target, while 16.6 events per hour remains in the moderate range.

HFNC also met the study's threshold for clinically acceptable titration in only 48% of participants. Among those who responded, 53% achieved at least a 50% AHI reduction. These figures came from a supervised, single-night lab setting, where fitting and titration receive more attention than they do at home.

Lower AHI did not mean better sleep

The more concerning finding was sleep architecture. CPAP improved sleep-architecture measures, while HFNC worsened them. The publication does not establish whether that change would impair daytime function or long-term health; it collected no symptom, cognition, quality-of-life, or cardiovascular outcomes. But treating sleep apnea aims to restore more than a cleaner respiratory tracing. A device that reduces events while degrading sleep stages needs stronger evidence before clinicians can assume a net benefit.

HFNC also became less effective as OSA severity increased. The 68-person study was not powered to define a reliable severity cutoff, so no one can honestly name an AHI value above which HFNC fails. The direction of the finding, however, argues against using it as a broad alternative for severe disease.

Why HFNC has a built-in ceiling

CPAP delivers pressure through a closed pneumatic system that splints the upper airway open. HFNC sends a high flow of warmed, humidified gas through nasal prongs and generates pressure more indirectly. That distinction helps explain why it can help without matching CPAP.

A 2024 study by Li and colleagues measured intrapharyngeal pressure during fiberoptic bronchoscopy under deep sedation in people without OSA. Pressure reaching the vocal folds was lower than device output, consistent with leakage around the prongs, upper-airway resistance, and anatomy. This is useful mechanistic context, not proof of what happens during natural sleep in people with OSA. The participants were sedated bronchoscopy patients, not sleeping patients receiving long-term OSA treatment.

That limitation matters because HFNC's proposed advantage is often comfort and usability. The randomized trial did not measure preference, comfort, home use, or adherence. It therefore cannot establish the premise that HFNC would be worn more consistently than CPAP, much less that better use would offset weaker AHI suppression.

A device that cuts AHI by 52% could outperform an unused device in real life. That is a plausible hypothesis, not a result of this trial.

Alternatives should match the phenotype

People who struggle with CPAP have options that are better characterized than HFNC, although none fits every form of OSA. For mild-to-moderate positional OSA, a 2025 randomized crossover trial of vibrotactile positional therapy compared positional therapy with CPAP in treatment-naive symptomatic adults. It measured symptoms, use, preference, sleep architecture, and quality of life over sequential eight-week treatment periods. That is a different population and intervention from HFNC, but it illustrates the value of choosing therapy based on the pattern of airway collapse rather than treating all OSA as one condition.

Oral appliances may suit some patients. A 2025 prospective cohort study of mandibular advancement therapy in 93 eligible patients found a dose-response relationship between jaw advancement and respiratory-effort burden during titration. For people with overweight or obesity, a 2022 analysis of a six-month clinical trial in 180 adults with moderate-to-severe OSA found that respiratory and oxygenation measures improved only in participants who lost weight. These approaches are not interchangeable, and weight loss was used alongside prescribed CPAP in that study, not as proof that CPAP can always be replaced.

Drug treatment remains early-stage. In a 2025 randomized, double-blind crossover trial, 20 patients received acetazolamide plus eszopiclone or placebo for three-day phases; 18 later entered an open-label triple-drug phase that added venlafaxine. The study reported preliminary efficacy, but its size and short duration rule out broad clinical conclusions.

Why CPAP remains the reference treatment

CPAP's advantage is not limited to a better AHI in the HFNC trial. In a 2023 study of 34 adults with moderate or severe OSA and obesity, three months of CPAP improved cardiopulmonary exercise tolerance and cardiac and pulmonary measures; the study was not designed to prove fewer heart attacks or longer life. A separate 12-week randomized trial in 72 people with type 2 diabetes and moderate-to-severe OSA found no significant change in average glucose, time in glucose range, hypoglycemia, hyperglycemia, or glucose variability with CPAP. That trial is a useful reminder that a therapy can improve breathing without delivering every hoped-for metabolic outcome.

HFNC has only one controlled OSA study, with 68 treatment-naive participants, one night per treatment, no blinding, and no follow-up at home. It excluded the very group most likely to consider it: people who have already failed or abandoned CPAP. Single-night AHI also varies, and an in-lab titration cannot tell us how device maintenance, changing sleep position, or chronic discomfort will affect results.

The bottom line

For people who can use CPAP, the evidence gives no reason to switch to HFNC. CPAP left a residual AHI of 5.8 events per hour versus 16.6 with HFNC, improved sleep architecture where HFNC worsened it, and achieved better control in the only direct comparison.

For someone who cannot tolerate CPAP, HFNC remains a question for research, not a standard salvage treatment. Trials must enroll CPAP-intolerant patients, track home adherence and comfort, measure symptoms and quality of life, and test outcomes over months rather than one carefully supervised night. Until then, a sleep clinician can help match established alternatives—such as positional therapy, an oral appliance, or weight-management treatment—to the patient's OSA phenotype and medical context.