CPAP's cardiovascular benefit concentrates in patients you can identify from the sleep study

Design

Sleep-clinic cohort linked to the French national health database, Cox proportional hazards with interaction testing

Population

3370 patients on positive airway pressure for moderate to severe obstructive sleep apnoea, median follow-up 9 years

Primary outcome

First major adverse cardiovascular event

Effect

Adherence associated with adjusted HR 0.53 (95 per cent CI 0.46 to 0.62); stronger in the high-risk group (interaction HR 0.61, 0.43 to 0.87, interaction P = 0.006)

Randomised trials of positive airway pressure have repeatedly failed to show cardiovascular benefit, which is usually explained by poor adherence and by enrolling patients whose risk was not modifiable. This cohort tested the second explanation directly. Three thousand three hundred and seventy patients on positive airway pressure for moderate to severe obstructive sleep apnoea in a French sleep clinic cohort, linked to national health data, were stratified by two physiological markers taken from the original sleep study: sleep apnoea-specific hypoxic burden above 40.6 per cent-minutes per hour, or event-related heart rate response above 21.7 beats per minute.

Over a median nine years, 740 patients had a major adverse cardiovascular event. Adherence, defined as four hours a night or more, was associated with a lower risk overall (adjusted hazard ratio 0.53, 95 per cent CI 0.46 to 0.62). But the association was far stronger in the 71.7 per cent classified high risk than in the rest (interaction hazard ratio 0.61, 0.43 to 0.87, interaction P = 0.006), and stronger still in patients who were not sleepy - the group in whom treatment is hardest to justify on symptoms alone. A simplified version of both markers, derived automatically from the oximetry signal, performed similarly.

The caveat has to be stated plainly, because the effect size invites overstatement: adherence was not randomised, and people who use a machine four hours a night differ from those who do not in ways no adjustment removes. A hazard ratio of 0.53 for an unrandomised behaviour is the classic shape of healthy-adherer bias. What survives that objection is the interaction, which is a comparison between two groups of adherent and non-adherent patients rather than a raw contrast - and it says that if positive airway pressure has a cardiovascular effect, hypoxic burden and heart rate response identify who has it. Since the simplified markers come from oximetry that is already recorded, this is worth computing before the next trial rather than after it.

• Look at hypoxic burden and heart rate response on the diagnostic study, not just the apnoea-hypopnoea index.
• Use the high-risk markers to prioritise adherence support, which is where the association was concentrated.
• Do not tell a patient that using CPAP will halve their cardiovascular risk; adherence was not randomised.
• Note the effect appeared stronger in non-sleepy patients, who are the hardest group to keep on treatment.
• The simplified oximetry-derived versions performed similarly, so this does not require full polysomnography scoring.

The statistics, in plain English

An adjusted hazard ratio of 0.53 for adherence versus non-adherence in an observational cohort should not be read as a treatment effect: adherent patients are systematically different, and healthy-adherer bias produces effects of about this size for placebos. The interaction term, 0.61 with an interval of 0.43 to 0.87, is more resistant to that objection because both compared groups contain adherent and non-adherent patients. The thresholds were derived by percentile within this cohort, so they need validating elsewhere before being used as cut-offs.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.