ResMed vs. Respironics: AHI and Variable Breathing

Muffy wrote: if VB and A/H/FL never appear concurrently, would that not suggest that A/H/FL are NOT reported when VB Mode is active?

Always less successful in choosing my wording than I would hope. Sorry.

In terms of the Respironics 30-second data reporting windows, A / H / FL / VB appear to be mutually exclusive. (I have no formal proof of this - just LOTS of data and no counter-example.)

In terms of my data for the night of 23 August, there are just over 45 minutes (over 90 windows) of ALMOST continuous VB entries - with the exception of five discrete (non-contiguous) windows in which the "state" of VB is NOT indicated but other "events" (1 A, 4 H) are reported. Two possible explanations immediately arise in my mind: 1) either we are in the "state" of Variable Breathing for the entire 45 minutes, and despite that "state", these five events were detected and "allowed" to be reported (in preference to reporting continuing VB), or 2) during that 45 minutes the VB state "ended" five times, at which moments there "happened" to be A or H events which were duly recorded, such events also triggering resumption of the VB state. Again, I can't prove it, but the first explanation seems a lot cleaner.

Muffy wrote:Also, how would you rate the quality of your sleep, including periods of wake, relative to events on the graph?

I have no way of knowing when I was "awake," but I have indicated on the graph when I was awake AND turned off the blower: I have inserted "zero pressure" entries when I arose to the bathroom (at "relative times" 5:00:41 and 7:47:21) and when I finally got up in the morning (at 9:30:50). Prior to the 5:00 arising, the last recorded "event" was the end of a "run" of VB about 7 minutes prior; I don't consider this temporally related. Prior to the 7:47 arising, a "run" of VB had ended just 30 seconds earlier; I had mentioned in a prior post that this seems to occur with some frequency. The same situation, with a VB "run" ending 30 seconds earlier, preceded my final arising at 9:30. Again, no other respiratory events, and no changes in pressure or leak patterns, precede my arisings. (I typically average 3 spontantous arisings nightly.) Beyond that, I have no useful information.

Velbor

Velbor, curious that your VB pattern seems to be that it ends and then you typically either have an event or shortly thereafter wake up...makes me wonder if this is just your pattern or is characteristics for all with VB. If characteristic of VB in general, could it be related to the increases in hormones prior to getting up, which then begs the question (at least in my mind) could VB be related to disturbances in the output of these hormones.

Muffy wrote:A better question would have been "How many users have Autos (where the AHI should be 0.0, having spent all that money on a machine to do exactly that)"?

Or "How many users have Autos with a too low starting pressure (that could get you a bunch of events)"?

It would seem that the aggessiveness of the Resmed A10 algorithm would have stomped out all the hypopneas, as suggested by the bench test by McCoy, Eiken and Diesem:



Maybe it's the "5 breath moving average" that adds in a few events during periods of artifact.

However, I'm sticking with Muffpothesis as the reason for the Respironics HI < ResMed HI (Respironics filters out poor sleep) until I see a comparison of ResMed vs Respironics with someone who has a Variable Breathing% of 0.0.

Muffy

This post by Muffy was made in another thread: viewtopic.php?f=1&t=44743
and I quote it in full here (hopefully with no objection by Muffy) because of its applicability to this discussion.

In particular, I am impressed by a comment in the original cited article:

"Common in the Respironics APAP models is a "hunt and peck" search pattern, as part of the device's algorithm is to find optimum therapy pressures in the face of airway resistance. The unit will increase pressure by 1.5 cm H2O over the course of 3 minutes and monitor the flow to see if there is an improvement. If there is none, the unit returns to the original pressure within 1 minute. If an improvement is recorded during the 1.5 cm H2O pressure increase phase, the device will lower the pressure by only 0.5 cm H2O. As there is no physiological component in this bench test, the unit consistently returned to the original pressure setting during periods of normal breathing. These are seen in the resulting device pressure profiles as triangular pressure increases/decreases."

If these Respironics "hunt and peck" pressure changes have been discussed here in the past, I have missed it. It appears from my own data (though I have never looked closely at the phenomenon before), that each such event typically takes 3 minutes: rising 1cm above the minimum pressure baseline and holding there for 2 minutes, then another 1cm rise for 30 seconds, followed by a return to baseline. Then there is typically 6 minutes of unchanged pressure (in the absence of a trigger-event), for a total cycle time of about 9 minutes. This appears consistent with the display in the article cited by Muffy.

I find this phenomenon of particular relevance to this Variable Breathing discussion in light of the graph I had posted earlier:



Note that during periods of Variable Breathing, these "hunt and peck" pressure variations appear to be shut down. This strikes me as the strongest evidence I've seen that VB is not simply a "failed" data-collection system whose "remnants" have simply been left within the Encore data system. Rather, VB appears to have an active effect on what the Respironics Auto algorithm does or does not do, and when it does or does not do it.

Velbor

Velbor wrote:If these Respironics "hunt and peck" pressure changes have been discussed here in the past, I have missed it.

It pops up from time to time. It was discussed by RG and -SWS in the Bev opus:

viewtopic/p399814/viewtopic.php?f=1&t=3 ... rs#p303441

Speaking of



get a load of this



where they say:

19. The system of claim 17, wherein the controlling means determines whether such a patient is experience a breathing instability based on the parameter, and delays searching for Pcrit or delays searching for Popt responsive to detection of the breathing instability.

Muffy

Here's another hopefully good example, this time of an atypically TERRIBLE night.

I could stay in bed for only about 6.5 hours, rather than my typical over 8 hours sleep.
I had more arisings than usual.
My periods of sleep were shorter than usual, none reaching 2 hours.



Things to notice:

Variable breathing occupied 61.9% of the night (equivalent VB Index: 70.9 events per hour!), rather than my usual ~33%.
This would be considerably higher if only the first 5 hours were included.
Hypopneas were the principal drivers for pressure increases.
Note especially the VB run from about 4:10 to 4:50 -- it is clearly "punctured" by five hypopneas.
Since VB and H are mutually exclusive, it appears that the H is detected and reported preferentially over the VB, even if VB conditions are continuing.
The "chair" pressure patterns of the Respironics algorithm "hunt and peck" are virtually obliterated most of the night by the VB.
The overall respiratory events (AHI = 4.0, AI = 0.5, HI = 3.5) aren't too shabby - the AI is about normal, though the HI is considerably higher than typical - despite the very high VB.

Leaks may look erratic, but really aren't all that bad: there's very little time spent above 24 lpm (adjusted for venting).
I believe that my mouth leak is associated with deeper / REM sleep.
The lack of significant mouth leak may well be due to how LITTLE time I spent in deep sleep, if we use VB as a surrogate for "waking" breathing.

Subjectively, I don't feel any "different" this morning than any other morning, though I know that I had a poor night's sleep.

Hope this adds to the conversation. Velbor

Muffy wrote: get a load of this



where they say:

19. The system of claim 17, wherein the controlling means determines whether such a patient is experience a breathing instability based on the parameter, and delays searching for Pcrit or delays searching for Popt responsive to detection of the breathing instability.

That perfectly matches up with the Pcrit and Popt "idle time" searches occurring in the lower-priority autoPAP control layer. That autoPAP control layer is the bottom-most control layer regarding priority, and the VB control layer is second from bottom.

But Respironics consistently describes the control layer that manages apneas and hypopneas as having higher algorithmic or treatment priority than either the VB or autoPAP control layers.

Velbor wrote: Since VB and H are mutually exclusive, it appears that the H is detected and reported preferentially over the VB, even if VB conditions are continuing.

I agree that VB and H are mutually exclusive regarding algorithmic response. And I also agree that VB and H are not mutually exclusive in physiology, since they can occur simultaneously.

Velbor wrote: The "chair" pressure patterns of the Respironics algorithm "hunt and peck" are virtually obliterated most of the night by the VB.

That perfectly corroborates that the VB control layer has higher algorithmic priority than the autoPAP control layer (that latter being the control layer that "hunts and pecks" for Pcrit and Popt during either "idle time" or even during certain subtle flow-signal variations that hint as early signs of FL).

Velbor wrote: it appears that the H is detected and reported preferentially over the VB, even if VB conditions are continuing.

And that corroborates all the Respironics patent descriptions that claim the apnea/hypopnea control layer has higher algorithmic priority over the VB control layer.

The Muffpothesis has the VB control layer either suspending apnea/hypopnea scoring and thus treatment---or at the very least deciding exactly which ones to hand off to the apnea/hypopnea control layer for treatment. However, either of those scenarios would functionally place the VB control layer higher (on the algorithmic control-layering scheme) than the apnea/hypopnea control layer. And that prioritization is opposite of what all the Respironics patent descriptions say about the priority of detected VB versus H.

While the Muffpothesis is interesting and reasonable IMHO, its algorithmic prioritization scheme so far doesn't match up with the priority-order claimed by Respironics or detailed above by Velbor. Thanks for your work, Velbor, Muffy, et al.

Whoa, Velbor! That is one bitter night!

I think that this not only gives more support to Muffpothesis, but gives rise to Muffpothesis 1.1, which will be:

Since Variable Breathing Mode (or any component of an xPAP algorithm) cannot guarantee 100% efficiency in event identification, then a failure of VBM may result in the erroneous identification, reporting and response to "phantom" hypopnea.

In other words, hypopneas appearing in the midst of a VB run, can/should be considered to be VB until shown to be otherwise. In other, other words, the Respironics algorithm is behaving like A10.

BTW, Velbor, when you subtract designed leak, do you "float" that value (vary it according to the pressure)?

Muffy

Muffy wrote:Whoa, Velbor! That is one bitter night!
....
BTW, Velbor, when you subtract designed leak, do you "float" that value (vary it according to the pressure)?

Yes, Muffy, my Excel worksheet contains columns of pressure and leak values at each time-reporting-point. I created an adjoining "adjusted leak" column, using the LOOKUP function to reference a table of mask-specific pressure/vent-flow data. LOOKUP matches the pressure value for each timepoint with the mask data, and then I subtract the pressure-specific venting from the reported total leak (with a zero floor) to populate an "adjusted leak" column.

I should also clarify that when I said that VB and H are "mutually exclusive," I mean that data review strongly supports the conclusion that Respironics Encore systems REPORT, for any 30-second time-window, ONLY ONE of OA, H, FL or VB (though Vibratory Snore appears to be reported separately and can be concurrent with other events). That is not to say that MORE THAN ONE of these events/conditions may be PRESENT during any given 30-second period, but ONLY ONE will be REPORTED.

I should also have emphasized in my last post the EXTREME variability of night-to-night data which I encounter. There is no clear correlation of results with ANY variable I can identify. Such variation is inherent in biological systems (= people) and can lead to two errors: 1) throwing out discordant data without justification, or 2) going crazy because of a few bad nights. So I would take this opportunity to again emphasize: trust the machinery, trust the data, trust and rely on statistics to make sense out of the numbers, and ABOVE ALL, don't make changes based on only a few nights of information. I find that it typically takes about two weeks of data before the "central tendency" numbers start to settle down, and about three weeks of data before they're stable enough to have any meaning. Even then, standard deviations are distressingly high, and statistically meaningful conclusions are not easy to come by. (I also think that the findings of sleep studies, valuable as they are, also should always be viewed in this context.)

No comments from me on Muffpothesis 1.0 or 1.1; I'm out of my league here. Many thanks for your fascinating observations! Velbor

Velbor wrote:Leaks may look erratic, but really aren't all that bad: there's very little time spent above 24 lpm (adjusted for venting).
I believe that my mouth leak is associated with deeper / REM sleep.
The lack of significant mouth leak may well be due to how LITTLE time I spent in deep sleep, if we use VB as a surrogate for "waking" breathing.

Well, time to dump on my own theory.
It just occurred to me that I could look at the correlation, if any, between percent of time in significant leak (>= 24 lpm adjusted for venting) with my nasal (ResMed Activa) mask, and the percentage of time in Variable Breathing.

Here are the results of 15 nights of data. There are other potentially relevant variables (e.g., pressure) but I don't think they're particularly relevant.



No correlation at all. So either my mouth leak is NOT associated with deeper sleeping, and/or VB CANNOT be used as a surrogate for wakefulness.

A bit embarrassing, but negative results need to be published just as much - if not more - that positive outcomes.

Regards, Velbor