CPAP machines keep airways open during sleep, and for millions of people with obstructive sleep apnea, they reduce daytime fatigue and improve quality of life. But whether CPAP actually protects the heart has been surprisingly hard to prove. Large clinical trials have produced mixed results, leading to an uncomfortable question: does the therapy help everyone equally?
A new study from Mount Sinai, published in Nature Communications Medicine, suggests the answer is no — and offers a tool that may eventually help clinicians figure out who benefits and who doesn't.
What the Model Does
The research team built a machine learning algorithm that estimates how CPAP therapy will affect an individual patient's cardiovascular risk. It is, according to the researchers, the first tool to provide personalized predictions rather than population-level averages.
The model analyzes 23 baseline characteristics — including prior medical conditions, smoking status, and sleep study metrics — drawn from a larger set of more than 100 potential predictors. It then generates an individualized treatment effect score estimating whether CPAP is likely to improve, worsen, or have no meaningful impact on that person's cardiovascular trajectory.
The Data Behind It
The team trained and validated the model using data from the Sleep Apnea Cardiovascular Endpoints (SAVE) trial, the largest randomized clinical trial evaluating CPAP for cardiovascular disease prevention. SAVE enrolled more than 2,600 participants across 89 sites in seven countries.
When the model was applied to SAVE participants:
- Patients in the subgroup predicted to benefit from CPAP who actually received the therapy experienced a 100-fold improvement in future cardiac risk compared to similar patients on usual care alone
- Patients in the subgroup predicted to be harmed by CPAP who received the therapy experienced a greater than 100-fold increase in cardiovascular events
These are dramatic effect sizes that reflect the model's ability to separate patients whose underlying biology makes them responsive to CPAP from those for whom the therapy may not address — or may even mask — their primary cardiovascular risk drivers.
Why CPAP Doesn't Help Everyone's Heart
The findings help explain why past clinical trials of CPAP for cardiovascular prevention have been inconclusive. If some patients benefit dramatically while others are harmed, the average effect across a trial will appear small or null — even though the treatment is highly effective for a subset.
Several factors may determine whether CPAP protects the heart:
- The dominant mechanism driving a patient's sleep apnea (anatomical vs. neurological) may influence whether keeping the airway open addresses the cardiovascular stress
- Existing cardiovascular damage may be too advanced for CPAP alone to reverse
- CPAP adherence patterns — since many patients use the device inconsistently — could create different exposure levels
- Competing risk factors like uncontrolled hypertension or diabetes may dominate cardiovascular outcomes regardless of sleep apnea treatment
From Research to Clinic
The model is not yet ready for clinical use. It was developed and validated on a single trial dataset, and prospective validation in diverse populations is needed before it could guide treatment decisions.
Mount Sinai has received a $3.32 million grant to further develop and test the tool, with the goal of eventually integrating it into clinical workflows. The vision is that a sleep specialist could input a patient's data and receive a recommendation about whether CPAP is likely to provide cardiovascular benefit — not just symptom relief.
What This Means for Patients
For now, CPAP remains the standard of care for moderate to severe obstructive sleep apnea, and most patients should continue using it as prescribed. The primary benefit of CPAP — reducing apnea events, improving oxygen levels, and alleviating daytime sleepiness — is well established regardless of cardiovascular effects.
But this research signals a shift in how sleep medicine may evolve. Rather than prescribing the same therapy to every patient and hoping it works, clinicians could eventually use prediction tools to match patients with the treatments most likely to benefit them — whether that's CPAP, oral medications, nerve stimulation, or a combination approach.
The era of precision sleep medicine is getting closer. The question is no longer just "Should this patient use CPAP?" but "Will CPAP actually change this patient's cardiovascular trajectory?"
