For years, the clinical story on CPAP and cardiovascular outcomes has been frustratingly flat. Randomized controlled trials — including the largest one ever run — repeatedly failed to show that CPAP therapy reduces heart attacks and strokes at the population level. Yet cardiologists and sleep physicians continued recommending it. Patients with OSA were told it helped their hearts. The population-level data kept saying: not significantly.
A study published in April 2026 in Nature Communications Medicine offers the most compelling explanation yet for this contradiction — and it changes how data-driven people should think about evaluating CPAP as an intervention.
The answer isn't that CPAP doesn't work. It's that for some people it works extraordinarily well, and for others, the data suggests it may make cardiovascular outcomes meaningfully worse. When you average these extremes together in a trial, the result looks like nothing.
What the Study Did
The Mount Sinai research team applied a causal survival forest model — a machine learning method that estimates individualized treatment effects — to data from the SAVE trial (Sleep Apnea Cardiovascular Endpoints). SAVE is the largest randomized clinical trial ever conducted on CPAP for cardiovascular disease prevention: 2,687 participants from 89 sites across 7 countries, followed for a median of nearly 4 years.
Previous analyses of SAVE had returned the now-familiar null result: CPAP did not significantly reduce the composite endpoint of cardiovascular death, MI, or stroke at the group level. The Mount Sinai team did not re-analyze the primary endpoint. Instead, they asked a different question: is there evidence of individual-level heterogeneity in who benefits from CPAP?
To answer this, they considered more than 100 baseline variables — sleep study metrics, prior medical conditions, medications, smoking history, demographics — and used the model to identify the 23 features that most strongly predicted individual-level response to treatment.
The Finding: 100-Fold Heterogeneity
The model identified two subgroups with dramatically different responses:
Favorable subgroup: Participants predicted to benefit from CPAP experienced a greater than 100-fold improvement in cardiovascular outcomes compared with matched participants in the usual-care group. These were people for whom CPAP appeared to be providing substantial cardioprotection — which the population-level trial result had been completely obscuring.
Harmed subgroup: Participants predicted to be harmed by CPAP experienced a greater than 100-fold increase in cardiovascular disease outcomes — including recurrent strokes and heart attacks — compared with matched usual-care participants.
A 100-fold difference in each direction. Averaged together across all participants: a null result.
This is a hypothesis-generating finding, not a clinical guideline. The model was applied to observational subgroups within a trial that was not powered to detect these subgroup effects, and the findings require prospective validation before they can guide treatment decisions. The researchers are explicit about this limitation.
But the pattern is striking enough that it deserves serious attention from anyone who is currently on CPAP, considering CPAP, or tracking their health data closely enough to evaluate whether a prescribed intervention is actually working for them.
Why This Happens: What the Research Suggests
OSA is not a single disease. The field has increasingly recognized it as a heterogeneous condition with different underlying mechanisms, phenotypes, and physiological consequences in different patients. The research team identified 23 baseline features that predicted individual CPAP response — the specific features are not all published in detail, but the categories include prior cardiovascular conditions, presence of comorbidities, sleep study characteristics, smoking status, and other baseline health parameters.
The leading hypothesis for why some patients may be harmed relates to sleep fragmentation. CPAP use in patients who do not achieve adequate adherence — or in patients whose OSA has a different physiological character — may disturb sleep architecture without providing sufficient airway stabilization. The resulting fragmented sleep could increase sympathetic nervous system activation and nocturnal blood pressure variability. In patients with pre-existing established cardiovascular disease, this may be worse than the disease state CPAP is intended to treat.
This remains a hypothesis, and the mechanism is not settled science. What is established is that the "one-size-fits-all" assumption for CPAP — that it uniformly benefits cardiovascular outcomes in people with OSA — does not appear to hold in the data.
What This Means If You Are Currently Using CPAP
Do not stop using CPAP because of this study.
This cannot be overstated. The study does not identify who is in the harmed subgroup in a way that can currently be applied clinically. The 23 predictive features were identified in a retrospective analysis and have not been validated in a prospective model that could classify individual patients before treatment. Your sleep physician prescribed CPAP based on your specific presentation and the established evidence that untreated moderate-to-severe OSA carries significant cardiovascular, metabolic, and mortality risk. That evidence is not overturned by this finding.
What this study does provide is a compelling scientific rationale for precision medicine approaches to OSA treatment evaluation — and a framework for thinking about what to track if you want to evaluate whether your CPAP is working.
What to Track: The Prova Angle
For men who use wearables and track longitudinal health data, this study creates a concrete experiment framework. If you are starting CPAP, adjusting CPAP settings, or evaluating whether your current setup is optimal, these are the metrics worth tracking before and after:
Heart Rate Variability (HRV)
HRV is the metric most sensitive to autonomic nervous system state and cardiovascular recovery quality. Men with well-treated OSA typically see HRV improve over weeks to months of effective CPAP use, as the chronic hypoxic stress on the autonomic system is reduced. If HRV trends downward or fails to improve after several weeks of consistent CPAP use, that is a signal worth discussing with your sleep physician.
Oura Ring, Garmin devices, and Apple Watch all provide nightly HRV metrics. The key is tracking the trend over at least 4–6 weeks, not individual nights.
Resting Heart Rate
Effective OSA treatment tends to reduce resting heart rate over time. This is a less granular metric than HRV but more stable and easier to interpret from consumer devices.
Sleep Stage Architecture
Consumer wearables provide estimated sleep stage data (light, deep, REM, awake time). In well-treated OSA, REM duration often improves as sleep fragmentation resolves. Persistent fragmented sleep architecture despite CPAP use — especially reduced REM or high wake-time percentages — may indicate suboptimal treatment effectiveness.
Morning Resting Blood Pressure
Nocturnal BP dipping is a normal pattern in healthy cardiovascular function — BP drops during sleep and rises in the morning. OSA disrupts this pattern. Men with effectively treated OSA often see improved morning BP over time. Tracking morning BP with a validated home cuff provides another data point that is difficult to game or placebo-effect away.
Energy and Cognitive Performance
Subjective but important. The intended benefit of CPAP is reduced daytime sleepiness and improved cognitive function. If you track daily energy ratings or cognitive performance metrics (reaction time, working memory tasks, self-rated focus), baseline vs. post-CPAP trends are informative.
Be the first to try Prova
We're building an app to track whether CPAP actually works. Join the waitlist.
The Conversation Worth Having With Your Sleep Physician
The Mount Sinai finding gives you a legitimate evidence-based reason to ask your sleep physician about treatment effect monitoring rather than simply accepting CPAP as set-and-forget. Questions worth raising:
- What metrics should I expect to improve, and over what timeframe, if this is working for me?
- At what point would you consider alternative treatments (positional therapy, oral appliances, hypoglossal nerve stimulation) if tracked metrics suggest CPAP isn't achieving the intended effect?
- Should I have a follow-up sleep study to confirm my AHI is actually controlled on my current settings?
These are not adversarial questions — they are the natural extension of evidence-based shared decision-making. Your physician may not have specific answers for all of them yet, because the precision medicine evidence in OSA treatment is still developing. But the conversation is worth having.
The Bigger Picture: Precision Medicine in Sleep
The CPAP heterogeneity finding is part of a broader pattern emerging across medicine: interventions that show modest or null effects at the population level often contain dramatic responder and non-responder subgroups. GLP-1 non-responders, antidepressant response variation, and now CPAP cardiovascular heterogeneity all point to the same conclusion: average treatment effects can obscure the individual variation that determines whether any given person benefits.
For men who track longitudinal health data with wearables and structured self-experimentation, this is actually the most favorable possible framing. The tools to measure whether an intervention is working for you specifically — not for the average trial participant — are increasingly accessible. HRV trends, resting HR, sleep stage architecture, morning BP: these are not clinical-grade measurements, but they are consistent and sensitive enough to detect signal when a meaningful biological change is occurring.
Effective CPAP should produce improvements across several of these metrics within weeks. If it does not, that is data worth discussing with your physician — not a reason to stop treatment unilaterally, but a reason to have a more specific conversation about whether your current setup is optimized.
The research does not yet tell us who is in which subgroup. Tracking your own data is, right now, one of the better ways to gather evidence about which side of that heterogeneity you are on.
This article is for informational purposes only and does not constitute medical advice. If you use CPAP or are considering treatment for obstructive sleep apnea, discuss any changes to your treatment with a qualified sleep physician or cardiologist. Do not discontinue or modify prescribed CPAP therapy based on this or any other article.