Philosophical question: is AHI accurate enough?

[problemchild]

I've been on CPAP therapy for 8 years.  My AHI numbers on my ASV machine (and wearing a soft cervical collar to keep my neck straight and mouth closed) is < 1.  So you'd think I'd be happy. 

But I can't help but wonder if AHI (apnea hypopnea index) really *is* an accurate means of measuring sleep apnea, the severity of the apnea events as they're happening, when they're actually happening, and what other kinds of sleep disruptions are occurring that aren't included in AHI calculations.

I'll use my own example as a for-instance...

In my sleep studies, and monitoring my apneas from home, it's obvious that I don't have many apneas outside the time when I'm in REM sleep.  If I'm not using the cervical collar, I have a *lot* of apneas, but they aren't distributed evenly throughout the night:  they're concentrated when I'm in REM sleep, then taper toward morning as I drift out of REM.  So I find myself wondering why AHI is an average that incorporates the entire night's worth of sleep, and is diluted by the number of hours where I'm not having apneas.

Also, why use AHI as an average when, if the number can be skewed when you have a short night of sleep with less time spent in REM?  My AHI numbers can rise on the weekends when I typically sleep in and have a longer night's sleep.  On a shorter night, with less time in REM, the AHI is less, but it reflects a shorter time window to collect apnea events in REM, not a significant change in the number, density, or strength of the apnea events I have when I sleep longer periods of time.

Along with that is the narrow definition of what counts for an apnea.  Is an apnea a breathing event that last 10 seconds or does it last longer? And, as the link below asks, if one person's apnea events last longer than 10 seconds, is it accurate to count that at the same weight (for calculating AHI) as someone whose events end right at (or near) 10 seconds.  Why wouldn't someone whose apnea events are significantly longer carry more weight in the AHI calculation than someone who has the same number of events, but whose events are much shorter in nature? 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3917481/

As an example, early in my treatment, I had a series of events that, strung together, accounted for more than three minutes out of a six minute timeframe, but were counted only as three or four apnea events.  If AHI is the main measure of apnea severity, such an event would mask the severity of my events compared to someone who had three or four ten second hypopneas over the same six minute period.  Also, why are obstructive or central apneas treated the same as hypopneas in the AHI calculation?  Wouldn't more serious events be reflected in the index compared to less serious ones?

Next, what about events that are disruptive to sleep, but don't count as "apneas" in AHI?.  Even *with* my almost-absent AHI numbers, my ASV records frequent (as in "all night) instance of vibratory snoring or other disruptive, but non-apnea events.  Are they serious enough to be addressed?  Do such events interfere with the amount of rest or good REM sleep I get every night, even when my official apnea events are almost non-existent these days?

Finally, what events does a machine, like my ASV, mask by intercepting apnea events before they happen?

Before I was ever diagnosed with sleep apnea, I would catch myself not breathing as I slipped off to sleep (or in the mornings if I was dozing).  My breathing rate would slow down and be paused for some period of time, right as I began to relax and drift away.  Most of the time, I wouldn't notice it, but every now and then, I'd realize, "Hey! I'm not breathing right now."  Ironically, those events seem to stop when I fall completely asleep.

After I was diagnosed with OSA, I figured my apnea machine would fix it, but of course it didn't.  Nor did the BiPAP machine I graduated to.  My ASV will now pulse a couple breaths to snap me out of that mode...which took some tweaking of the setting so it wouldn't wake me up or make my chest muscles hurt.  But I still have those events.

I thought those events would be registered with the machine as CA's or HA's, but apparently they're not.  Because I'm not in REM, and because the ASV can correct the event before it becomes an official apnea, they don't end up counting as an actual apnea event.  So they don't make my AHI score.  But I've had nights, even on my ASV, where I counted 20 or 30 of those events, where the ASV sent a couple stronger air pulses to get me breathing again as I was relaxing to fall asleep.  None of those make the chart as apneas, although my doctor told me (after I asked her to look for them) that the ASV is recording them and sending them to her.

I always considered those falling-to-sleep events central apneas.  I know they can affect my O2 levels the times I've had surgery and have been coming out of the influence of the apnea-inducing effects of anesthetic or heavy pain meds.  I simply stop breathing for awhile.  And, though my ASV seems to be doing a good job of catching, and correcting, those events, they don't seem to be getting counted as CA's.  Or measured in my AHI.  So does that mean my ASV is making those events?  Or are they really central apneas?  If not, what *are* they?

I'm not really looking for analysis on my own sleep numbers.  I've done it in the past, but my DreamStation wasn't originally covered by Sleepyhead.  And in the end, that's not my concern or question.

What I'm pondering here is why AHI is considered to be the gold standard of measuring apnea severity, given the weakness of what AHI doesn't really cover?  Even the guideline of getting below AHI of 5 seems arbitrary and inaccurate.  I mean, if I had 50 apneas on a night where I slept 5 hours, my AHI would be 10...above the threshhold.  But if I had the same number of apneas on a night where I slept 10 hours, my AHI would be 5.  Same number of apneas, most of which would (for me) have been collected during REM.  But the score is twice as "bad" because one night I didn't sleep as long.

Why wouldn't doctors find an index based on density and severity of apnea (and disruptive) events instead of one based on a diluted average that, effectively, does a headcount of events without measuring what they're really like? 

My frustration over this is talking with doctors and folks in the industry who won't listen to you if you tell them you're still not getting rest...all because your AHi is still "okay" by the guideline standard.

Thoughts?

[Fats Drywaller]

Just a couple of little things that occur to me:

Even the guideline of getting below AHI of 5 seems arbitrary and inaccurate.  I mean, if I had 50 apneas on a night where I slept 5 hours, my AHI would be 10...above the threshhold.  But if I had the same number of apneas on a night where I slept 10 hours, my AHI would be 5.  Same number of apneas, most of which would (for me) have been collected during REM.  But the score is twice as "bad" because one night I didn't sleep as long.

It's not twice as supposedly bad; it is twice as bad, objectively, because what's being measured is the average of (counted) events per hour. Per unit of time, the quality of your sleep on the first night is worse than that of your sleep on the second night. (That's disregarding the separate point that 5 hours isn't enough.) I assume that in most cases the duration is 8 hours plus or minus whatever; that's what the body wants. So if AHIs are often assumed to represent an average over an 8-hour night, that shouldn't be surprising. An 8-hour duration is right in the middle of the bell curve.

More generally, AHI is a measure (yes, a flawed one, as you say) of how untreated a patient's sleep apnea is. Sleep apnea isn't cured by CPAP or ASV. The whole idea is to treat it, or patch it, well enough that the oxygenation and the carbon-dioxide excretion are functioning acceptably. Even an AHI of zero isn't necessarily perfection, as you can tell from the various posts here by folks who have consistently low AHIs, maybe even zero, but still feel fatigued. That's one reason the question "Yes, but how do you feel?" keeps coming up.

[Fats Drywaller]

It also seems to me that any competent sleep doctor will take all of that into account and won't rely entirely on AHI numbers to tell the whole story, but will look at the full data if available, in addition to quizzing the patient. I mean, doesn't this just get back into our standard old lament, frequently heard, that a lot of sleepydocs are incompetent? That's why there's even an emoticon for it in the AB collection.

I can't help thinking that maybe you're expecting the AHI to do too much, or to represent too much. To paraphrase Uncle Siggy, sometimes an AHI is only an AHI.

[Sleeprider]

I'm not particularly impressed with AHI as a measure of success, but it is the one metric the insurance companies and Medicare in particular use to determine your medical need and qualification for reimbursement of equipment. There are many individuals who go untreated because they don't have the number or types of events Medicare considers worthy of treatment. Doctors and sleep labs have become complicit in advancing this absurd standard. I have seen studies with obvious respiratory event related arousals, severe flow limitation and other problems that the person did not qualify for treatment because they didn't have the necessary apena or hypopnea; yet they display very poor sleep efficiency and sometimes very low oxygen saturation. We have also seen obvious cases of complex and central apnea dismissed and diagnosed as obstructive sleep apnea with a recommendation to lose weight and try CPAP.

The simple fact is a large portion of people the quit CPAP, do so because it is ineffective and the medical system is not interested in improving their results, instead blaming the patient for poor outcomes. Your question was a bit long, but hopefully I covered the gist of it.

[DeepBreathing]

Problemchild, I think you're quite right that AHI on its own is not a valid measure of anything, for all the reasons you and the other posters have mentioned. It is, at best, a crude approximation of your condition.

#SleepyHead displays a total time in apnea which partly addresses the question. (It's quite a while since I've used ResScan so I don't know if it does likewise). But I don't think I've seen anything which measures a cluster as such or the average duration of apneas and hypopneas - the data is all there, it just needs some software to pull it into a reasonable shape for reporting. In addition to AHI, the following would make useful parameter tools: a) AHI, b) AHI for the worst hour (or half-hour), c) mean duration of apneas, d) maximum length of apnea, e) total time in apnea.

As I said, all that data is there, but when people look at compliance and AHI and nothing else, then the intelligence needed for a thorough diagnosis and treatment goes missing.

[AshSF]

AHI serves the 80/20 rule. It’s good enough for 80% of apnea sufferers.

Then insurance cos and doctors latch on since it makes their life easier (rather than seeing graphs and individual events).

AHI doesn’t take into account sleep architecture (how much time we spend and need to spend in REM and SWS stages). I suspect that sleep architecture is the first thing to go bad many years before AHI nos starts creeping up. And conversely, treating AHI to below 5 doesn’t tell us if we fixed the underlying sleep architecture.

Unfortunately, unless we get cheap and reliable EEG sensors, we can’t answer that in a home setting.

[Fats Drywaller]

As I said, all that data is there, but when people look at compliance and AHI and nothing else, then the intelligence needed for a thorough diagnosis and treatment goes missing.

There's a natural human tendency to want things to be simple and reducible to easily digested numbers. Sometimes that happens to work, as with body temperature, blood pressure, and pulse: nice and simple, and informative. But AHI happens to be too simplistic, hiding useful information. A state vector is what's needed. However, that's great for software but not so great for people, because people don't want to have to do any arithmetic. I was thinking about a rudimentary state vector consisting of duration of sleep session along with counts of OA, CA, H, and FL. But that's not quite enough, because you're right that it would also be good to focus on the periods with the most events and the worst events. "It's always something."

[mesenteria]

I have always understood the concept of AHI to be akin to a pair of thumbs trying to do up a shirt button; they'll work, eventually, but it won't be what one would term 'efficient', or pretty.

I tend to be a late faller-asleeper.  I get to bed at a reasonable time, but some nights I lie awake for up to 40 minutes.  I won't have a single apnea/hypo event during those 40 minutes, while the machine is dutifully compiling its nightly record of my breathing.  Then, come morning, I may turn over and attempt to get another dream.  Sometimes I'm granted that giftie, others not so much.  Yet another 40 minutes while the machine records no events.  My AHI's are routinely less than 1.2, with the current 30 day average at 0.4.  Most nights I have at least three qualified events that I can click on and see how long I was in X, Y, or Z.  Mostly up to three CA's for me, none over 17 seconds.

My point is that my events are reflected, but not using a metric of density during a discrete period except for the total numbers comprising my sleep period.  So, I might have all two, three, or four events inside of a 20 minute period, which I would think would not be salutary.  However, those events, if they happen in a constrained timeframe, might be indicative of something I can modify.  In fact, we routinely look at sleepyhead data to see if those who submit their records can find ways to improve their AHI.  We see positional problems, for example.  And this is why sleepyhead, with all its indicators, is such a superior tool compared to the gross indicator that is AHI.  

Heavy reliance on AHI is probably one of the great limitations to helping with the 'face validity' of PAP in all its forms; people don't feel the improvements they hope for, and they'll abandon their therapy.

[Mosquitobait]

FWIW, a lot of sleep doctors would like to see anybody with an AHI 3 and above get a machine. I didn't qualify for a machine because my events occurred ONLY during REM. They fudged the numbers because of my very low oxygen. Otherwise, lots of paperwork would have been required for Medicare approval.