Your CPAP machine congratulates you every morning. AHI: 1.2. Therapy working. But you drag yourself out of bed feeling like you slept in a cement mixer. Sound familiar? You are not imagining it. And it is not "just stress." The metric your entire treatment is built on has a fundamental design flaw, and thousands of CPAP users are paying the price.
AHI Was Never Designed to Measure Sleep Quality
Let's be clear about what AHI actually is. The Apnea-Hypopnea Index counts two things: complete breathing stops (apneas) and partial breathing reductions (hypopneas). To count, each event must last at least 10 seconds and meet strict criteria -- typically a 3-4% drop in blood oxygen or a visible arousal on EEG.
That is it. That is the entire scoring system sleep medicine has relied on since the 1990s. It was designed for one job: diagnosing obstructive sleep apnea in a lab setting. It was never intended as a comprehensive measure of how well your therapy is working, or whether your airway is truly open while you sleep.
And yet that is exactly how it is used. Your device summarises an entire night into a single number. Your sleep physician checks that number once a quarter. Your insurance company uses it to decide whether you deserve treatment. A number that, by design, cannot see the majority of breathing problems.
The core problem
A night with hundreds of flow-limited breaths -- your airway narrowed, your body straining to pull air through a partially collapsed passage -- can produce an AHI of zero. The metric says you are fine. Your body knows otherwise.
The Evidence Is Damning
This is not a fringe opinion. The research community has been raising alarms about AHI for over a decade. Here is what the evidence shows:
AHI fails to predict who gets better
Multiple studies have found poor correlation between AHI reduction and symptom improvement. Patients with an AHI of 30 sometimes feel identical to patients with an AHI of 3. The 2016 SAVE trial -- one of the largest randomised controlled trials of PAP therapy -- showed no reduction in cardiovascular events despite successful AHI suppression in over 2,600 patients.
The 10-second minimum is arbitrary
A breathing disturbance lasting 9.5 seconds is invisible to AHI. One lasting 10.1 seconds counts. A 60-second apnea that tanks your oxygen to 70% counts the same as a 10-second event with a 3% dip. Equal events. Wildly unequal consequences.
UARS exists precisely because AHI fails
Upper Airway Resistance Syndrome was defined by Dr. Christian Guilleminault in the 1990s to describe patients with significant sleep disruption from airway resistance that AHI could not capture. The fact that an entire syndrome had to be invented to describe what AHI misses tells you everything about the metric's limitations.
Flow limitation drives symptoms directly
Research from Dr. Avram Gold suggests that inspiratory flow limitation itself -- not just the arousals it causes -- can drive daytime symptoms through activation of the limbic system and HPA axis. Your body mounts a stress response to the effort of breathing through a narrowed airway, even when you never fully "wake up." AHI does not see this. It does not even try.
Even the American Academy of Sleep Medicine (AASM) has acknowledged that AHI has significant limitations as a standalone metric. The field is slowly moving toward multi-dimensional assessment. Slowly being the operative word.
Five Things AHI Cannot See
Your PAP device records detailed breath-by-breath flow waveforms to its SD card every night. But the summary screen boils all of that data down to one number. Here is what it throws away:
1. Flow limitation severity
The Glasgow Index scores inspiratory flow shapes across 9 components, capturing the full spectrum of airway narrowing from subtle flattening to severe obstruction. A high Glasgow score means your breaths are distorted -- your airway is fighting you -- even when AHI sees nothing wrong.
2. Breathing regularity
Sample Entropy measures how chaotic your breathing pattern is. Healthy sleep produces rhythmic, predictable breathing. High irregularity signals an unstable airway or ventilatory control problem that AHI will never flag because no single breath crosses the event threshold.
3. RERAs (Respiratory Effort-Related Arousals)
Sequences of flow-limited breaths that end in a micro-arousal. Your brain wakes just enough to restore airflow, then drops back to sleep. This happens dozens of times per night in some patients -- each one fragmenting sleep architecture. AHI does not count a single one of them.
4. Negative Effort Dependence
NED measures whether airflow decreases as respiratory effort increases -- the hallmark of a collapsing airway. A high NED value means your airway is actively working against you, breath by breath. The per-breath Flatness Index and M-shape detection add further nuance that a single AHI number cannot express.
5. How the night changes over time
Your first half and second half of sleep are physiologically different. REM sleep clusters in the second half, and airway tone changes throughout the night. The H1/H2 split in flow limitation metrics often reveals that your therapy works fine early but breaks down during the critical REM-heavy hours. AHI averages everything into one flat number, erasing the story.
The Pattern AHI Cannot Detect
There is one more blind spot worth calling out: periodicity. Some patients show cyclical patterns of breathing disturbance -- their airflow waxes and wanes on 30 to 100-second cycles. This periodic breathing pattern suggests ventilatory instability, a marker of loop gain problems that standard AHI scoring ignores entirely.
The Periodicity Index (derived via FFT analysis of flow data) can detect these cycles automatically. Combined with the other metrics above, it paints a picture of respiratory health that no single number can capture. Your device has the data. It just never shows you.
So What Do You Actually Do?
First: do not stop PAP therapy. AHI is a flawed metric, but it is not useless. It catches the big events -- complete airway closures and significant partial collapses. Those still matter. The problem is not that AHI is measured. The problem is that it is the only thing measured.
Look beyond the summary screen
Your PAP machine's SD card contains detailed flow waveform data from every night -- far more granular than the summary your app shows you. Tools that analyse this raw data can score flow limitation, estimate RERAs, measure breathing regularity, and detect periodic patterns that AHI will never show.
Track trends, not single nights
A single night can be noisy. Position, alcohol, allergies, stress -- all affect results. What matters is the trend across weeks or months. Is your flow limitation improving or worsening? Are your breathing patterns more regular over time? Trend data gives you and your clinician a baseline to measure progress against.
Bring data, not complaints
"I still feel tired" is easy for a clinician to dismiss. A report showing elevated Glasgow scores, high FL percentage, and estimated RERAs despite a low AHI is not. Objective data changes the conversation from "your numbers look fine" to "let's look at pressure adjustments."
This Is a Systems Problem, Not a Patient Problem
If you have been told your numbers look great while you still feel terrible, understand: the system failed you, not the other way around. Sleep medicine has relied on a single metric for three decades because it is simple, standardised, and easy to bill against. Not because it is sufficient.
The research has moved on. The clinical guidelines are slowly catching up. But your device firmware and your insurance company's compliance algorithm have not. Until they do, the gap between what AHI says and how you feel will keep widening for a significant percentage of treated patients.
The good news: the data to close that gap already exists on your SD card. You just need the right tools to read it.
References
McEvoy et al. (2016). "CPAP for Prevention of Cardiovascular Events in Obstructive Sleep Apnea (SAVE Trial)." New England Journal of Medicine, 375(10), 919-931.
Guilleminault et al. (1993). "A cause of excessive daytime sleepiness: The upper airway resistance syndrome." CHEST Journal, 104(3), 781-787.
Gold et al. (2003). "The symptoms and signs of upper airway resistance syndrome." CHEST Journal, 123(1), 87-95.
Azarbarzin et al. (2019). "The Hypoxic Burden of Sleep Apnoea Predicts Cardiovascular Disease-Related Mortality." European Heart Journal, 40(14), 1149-1157.
Punjabi et al. (2008). "Sleep-disordered breathing and mortality: A prospective cohort study." PLoS Medicine, 5(8), e173.
Related articles
Beyond AHI: Why Your Sleep Apnea Score Might Be Misleading You -- the research case for multi-dimensional sleep assessment.
Your AHI Is Normal But You're Still Exhausted -- a practical guide to identifying flow limitation in your own data.
Understanding Flow Limitation -- what flow limitation is, why it matters, and how to detect it.
AirwayLab Glossary -- definitions of all metrics mentioned in this article.
Medical disclaimer
AirwayLab is an educational tool, not a medical device. The analysis provided is based on published research methodologies applied to your PAP device's flow data, but it is not a substitute for polysomnography or clinical evaluation. Always discuss therapy changes with your sleep physician. The metrics described here are for educational purposes and to support informed conversations with your clinician.
See What AHI Is Hiding
AirwayLab runs four research-grade analysis engines on your ResMed SD card data -- Glasgow Index, FL Score, NED, RERA estimation -- and shows you the breathing patterns your device's summary screen discards. Free, open-source, and 100% private. Your data never leaves your browser.