Only O2 Sat matters, or OA False positives - Part Deux

[rbookman]

[This got a bit long.....]

Earlier, as I was grappling with the data from my DreamStation bPAP (DSX700), I wondered aloud on the forum about exactly what algorithm was used by SleepyHead or DreamStation software to declare an OA or obstructive apnea event. After an overnight sleep test, I was diagnosed with severe obstructive sleep disorder and had reported AHI values > 40, with OA events counts usually in excess of 200 per night and CA counts that were either really low or, sometimes, around 100.

We have been struggling to get the bPAP settings right and various forum members have been incredibly helpful. I am using supplemental O2 (4 lpm) through the mask at least until we get this thing better understood and treated. Trying to make the DreamStation/SleepyHead AHI/OA/CA/HA numbers better has been a real struggle. And, I could not make subjectivesense of the OA counts when I tried to map what was happening in the 2-20 minutes while I was falling asleep to what the record showed the next morning when I downloaded the data.

I do want to quickly add that despite current AHI and apnea counts, I am sleeping much better and feeling more energetic. Whether that is the cervical collar, the supplemental O2, the bPAP machine....we don't know yet.

Meanwhile....

I've been recording O2 saturation and pulse rate every night on my Masimo MightySat along with my bPAP data...and over the weekend, I wrote some code to dig into this a bit more. The Masimo bluetooth stream the data to my iPhone in real-time...and I can email myself the 25,000+ time points in a csv file when I wake up.

Here's the hypothesis for your consideration:
All the OAs & CAs & HAs in the world really don't matter as long as you are not desaturating. If your O2 saturation wobbles around between 90-99 all night long....you are not in trouble. It's "severe" desaturation that is the concern...and this is usually defined as at or below 88%.

So now, let me share data from last night.

First, here's the entire night's PulseOx and heart rate record from the Masimo MightySat. The little green "+"'s mark desaturation events which were defined as periods with O2_Sat<=88%. The time resolution of the MightySat is 1 second and I defined as a desat event 1s or more of PO2<=88. The straight black line indicates the user-set threshold of 88% saturation. There were 19 desaturation events which tended to occur in bursts. Over 8.6 hours of sleep, I was at or below desaturation threshold for about 10 minutes, total. More time details of those desat events are in the figure. 


The Dreamstation numbers seen in the figure (upper left) are from the DreamMapper app, not SleepyHead. PulseOx data are longer than bPAP data....because I start the PulseOx before and stop after starting and stopping the bPAP.


Now, here's a magnification of the time axis to focus in on one of the DeSat bursts, composed of 4 deSat events, that occured around 2AM.



So....what do we learn from SleepyHead and the bPAP data?
49 CAs, 20 HAs, 187 OAs, and AHI = 29.96.

The next figure takes the detailed pulseOx graph and lines it up on top of SleepyHead graphs, zoomed in to have the exact same time period. At the top of the SleepyHead record is, as usuaal, the full timeline with all the events marked...the CAs, the OAs, etc etc. The zoomed graph show 4 of the 19 total desat events I recorded last night.

For the 3 longer deSat events, they were associated with BND events and two of them were associated with CA events. But note that there are other CA and OA events called by the software that are not associated with any significant desaturation. Are they false positives?

For me, I don't think the SleepyHead Event record is particularly helpful....in my case. Your case may be different. But, if you can see some value in my records, please share the insights.

I think it makes most sense for me to focus on O2 Sat/deSat and not worry about making machine or mask or any other adjustments to change any other measurement. Just focus on O2 Sat.

What do you think? 
Thanks in advance for your feedback, suggestions, and comments.

[Walla Walla]

Even though your O2 may be fine if your sleep is disturbed over and over it still can effect your health.

[srlevine1]

A few observations ...

(1) Your finger-tip pulse-oximeter is not a reliable or accurate instrument when it comes to fine resolution data. Movements and/or holding your breath while repositioning yourself can produce data anomalies. Heart arrhythmias can also affect pulse-ox events. Therefore, one should look at the trend rather than individual data points. I have tested and compared my CMS-50F, Dynamed Professional Pulse-Ox, and the Nonin XPOD 3012 with ResMed Adapter and the generally compare with some small differences in results for movements. Even with O2 sensors on adjoining fingers are not absolutely in-sync with each other when moving around. I often see those exaggerated downward 02 spikes when I move around. You might want to test this hypothesis while awake.

(2) To concentrate on the results of your pulse-ox over trying to manage your other numbers may be a major error in judgment as a number of OSA events over time can have a major cumulative effect on your physiology regardless of some desaturation events.

(3) The detection algorithms live in the machine and the software simply reports the result, it does not, to my knowledge, attempt to re-compute the numbers from the actual data. Everything is a function of flow, pressure, and time. The definitions of events are well known and standardized over almost all machines; what the algorithm does is to determine how the machine responds to the sensed data.

(4) You treat the patient, not the machine. Attempting to respond to anything but trends is bound to drive you crazy in the short-term. I suggest you discuss your issues with your pulmonologist, cardiologist, and sleep doctor.

Best of luck. Try for a good night's sleep for a better tomorrow.

[rbookman]

Thanks, @srlevine1!!

In response to your speicifc points:

1) Because much of my sleep apnea can likely be attrributed to position, I not only sleep on my back (only), but also sleep with a soft cervical collar. Just to be sure, I tried recording while moving more than I do during sleep and while small (~2%) variations were noted, I never could get below 88% saturation, so I don't think movement artifacts are a significant issue....and generally treat the pulseOx data as reliable. I see no reason not to. DeSat events lasting 20-60 seconds are not likely to be artifacts, as far as I can tell. If you have evidence otherwise, that would be very helpful.

2) I grant your point here.....but not if the events are false positives. What are your thoughts about that?

3) I've wondered about where the algorithms live. Are you sure they live in the machine? For example, I have 3 sources of reports on AHI: The DSX700 built-in screen, DreamMapper software, and SleepyHead software. For last night, the reported AHI values are: 34.6, 30.9, and 28.26, respectively. If only the DSX700 machine calculated them, I'd expect all three values to be the same and equal to 34.6. Do you have any insight into why they are all different?

4) Great advice! Couldn't agree more. The question is: follow what trends? Trends in AHI? For me, I'm not so sure that is helpful. I'm suggesting that it's trends in DeSat events, total DeSat time, and max DeSat event duration that are the key variables to follow. For me. 

Thanks for your help and good wishes!

[Walla Walla]

rbookman,
I have the same machine and inputs for AHI. It except for some rounding off of less than a point mine match up fine.

It may be that the clocks are off from each other. Also there's settings in sleepyhead that could be excluding some of your recorded sleep time. Check the settings.

[mltam]

Just a few comments from a total newbie.

Medicine is a complicated "science", because it is hard to go from the statistical patterns we see to health effects.
What usually happens is that we try to go through UNDERSTANDING where a certain pattern comes from.
So, there have been some studies that have shown health affects of "sleep apnea". Well, what is sleep apnea? No red
light lights up saying "this is an event" or this person has sleep apnea. Instead, for each particular study there was a certain
definition of what apnea events are. Apnea was seen as stopping to breathe or waking up, so EEG and air flow were measured
to define apnea events. Now we see that there is a correlation between people who have these events and health effects.

But why? We don't exactly know. Since we also do not have an exact picture of what sleep is good for, it is somewhat hard to
see which of the things that happen during sleep are disrupted.

So, there is definitely a theory that says repeated low oxidation events during sleep is bad for you. Now you'd need to do a study looking at low oxidation events as opposed to low airflow events, to try to convince people (doctors) to look not just at airflow, but also at low oxygen events.
There have been a few studies like that. But again - how can you tell which is worse? How can you find people who have one and not the other?

There are theories that talk about other effect of apnea - low level waking, stress on the system from lack of breathing, effects on the heart associated with low rates of breathing, etc. etc. All these are very hard to prove/define/measure.
I also think that it is unclear what exactly is bad for you? Is it bad to have a certain number of events divided by the total time in bed/total time asleep? Or maybe a large number of events over 1 or two hours in the night is bad? Is it worse to go down to 85% saturation 15 times during the night, or 2 times to 80%? For each of these one needs a quite complicated study. So medicine proceeds just with what it knows, and that is "apnea events as defined by studies done are bad for you".


Now back to your question. I'm actually in a similar boat... My AHI aren't very high, when measured over the whole night. My main problem were 2-3 episodes where SpO2 would go down to 80% or so. Now I'm also following O2 and other events. Since I started using the CPAP, O2 has been really good, but I am tracking CA and other events. Partially to try to evaluate with an objective measure my sleep quality - I know I wake up during the night, or am restless, and I'm trying to evaluate how much.
A second thing is I'd like to keep track of CA events, to see that they're not increasing over time, that my body doesn't "forget" how to breathe...
BTW: I noticed that sometimes CA events are actually associated with an increase in O2, not a decrease. I take a couple of deep breaths, and then stop breathing for some time.

Sorry, I seem to have been rambling. What I wanted to say is O2 desaturation events are bad, and there is some reason to also track other events.

[sonicboom]

I doubt the events are false positives although there may be a few.  But overall, I would be concerned with an AHI as high as yours.  Studies have been done and published (and they were recently discussed on this forum) on the accuracy of the most prevalent CPAP machines compared to PSG testing.  The result was that the CPAP machines are fairly accurate in assessing events.  So, if it were me, I would track both O2 and AHI levels and strive towards results that reduce desats (which you have seemed to accomplish) and reduce AHIs to below 5 an hour which the medical community considers "treated" by definition.  I would also consult my doctor regarding making any changes to my therapy.

[srlevine1]

A few observations ...


(2) To concentrate on the results of your pulse-ox over trying to manage your other numbers may be a major error in judgment as a number of OSA events over time can have a major cumulative effect on your physiology regardless of some desaturation events.

(3) The detection algorithms live in the machine and the software simply reports the result, it does not, to my knowledge, attempt to re-compute the numbers from the actual data. Everything is a function of flow, pressure, and time. The definitions of events are well known and standardized over almost all machines; what the algorithm does is to determine how the machine responds to the sensed data.

About point (2) ... because the AHI calculation is time-dependent (events/time), I find it is better for me to look at reducing the number of events (CAs first, OAs second, HYPOPs third). Of course, if I really wanted to reduce the AHI all I have to do is wear the mask for a longer period of time in a semi-wakeful state or sleep longer. I do not overlook the issue of desats when looking at my progress -- but most are acceptable and the trend is positive.

About point (3) ... In a discussion with ResMed engineering, I was told that events are triggered by standard flow criteria and counted with the results stored in the machine as summary data and written to the SD card as detail data. Here are the definitions: https://www.resmed.com/us/en/healthcare-...aries.html. There will be differences even between two identical machines because they way the machine may algorithmically respond to stop the next event. I know Sleepyhead does some computations with the data to prepare for graphing but will need to look at the freely available code base to see if it re-computes the AHI.

Here is the code repository: https://sourceforge.net/p/sleepyhead/cod...leepyhead/