ASV users: the everything ASV thread.

[-SWS]

What you say would make perfect sense in the world of, say, mathematical proofs. In the world of statistical associations, however, it is incorrect. The fact that A != B does not mean you cannot infer anything about B from A. For example, let us suppose that half of obstructive events in plain OSA are the same as obstructive events in ComplexSAS, while the other half are unique to plain OSA and have a different pathogenesis. You would say "Hey, they aren't the same! Stop trying to make claims about one population from results on the other!"

I can agree with this.

But statistically, this means that any evidence we have on, say, the performance of ASV vs. CPAP on one population applies 50% to the other. If ASV is twice as good as CPAP at reducing the OAI in ComplexSAS, we would expect it to be 1.5x as good on OSA. That is basic statistics. Of course, if we tested it, it probably wouldn't turn out to be 1.5x as good, because the two treatments likely wouldn't be exactly equal at the 50% of events that are unique to plain OSA. But the most accurate guess we can make using that information is 1.5x - that ASV will do twice as well on the half of events that are the same, and that the two will perform equally well on the events that are different (that's our best guess when we have no other information about relative performance on this sample).

You never presented a shred of evidence for commonality. Rather you worked on a vague and under-researched supposition of commonality as established fact. Sleep researchers have yet to agree on commonality, if any, in the pathogenesis for these two populations. And yet you claim to know:

After all, the ComplexSAS population is not some totally different group of people with a totally different condition. It is simply the set of OSA sufferers who cannot splint their airway with enough pressure without inducing central apneas.

In order for ASV vs. CPAP performance on obstructive apneas in ComplexSAS to tell us *absolutely nothing* about ASV vs. CPAP performance on obstructive apneas in OSA, it would have to be the case, not merely that they are somewhat different, not merely that the populations aren't identical, but that there was absolutely no similarity. If there is any overlap in pathogenesis between the events in the two populations, then we can statistically infer that a treatment that is better on one is, before testing, more likely than not to be better on the other.

Your scope of statistical analysis also fails to factor iatrogenic effects of unnecessary PS. I linked that above and you ignored that as well. You offer a lot of number-cranking for a guy who has thus far presented nothing more than vague, under-researched suppositions.

Sorry for the long rant, I'm frustrated that what seems like such a simple statistical argument is not clear, and I'd like to hash this one out in detail, both because it is important in and of itself (ASV vs. CPAP for plain OSA seems to be a topic of much interest on this board), and so that when I make similar statistical arguments in the future, y'all accept and understand them.

I really think you're over-invested in your own powers of analysis.

[patrissimo]

Patrissimo, first comment: IMO, you are putting too much stock in the data from these machines. Think about this well know piece of information: The previous version of the Resmed S9 scored tons of hypopneas, way more than the S9 does. If someone upgraded from the S8 to an S9, did their SDB suddenly worsen overnight? No, the algorithm changed. From one machine to the next, the data is different. I've been on Respironics CPAP and APAP, Resmed CPAP and APAP, Resmed bilevel and Auto-bilevel, and now Resmed ASV. They don't even score FLs the same, even within the same brand. This isn't a sleep study (which would also yield different results from night to night and lab to lab). .

I'm a bit confused. The paper I was referring to ("Adaptive Pressure Support Servoventilation: a Novel Treatment for Sleep Apnea Associated with Use of Opioids") used PSGs to get the result that ASV decreased OAs more than CPAP did. I think all of the papers were based on PSGs, not on data from the machines. Where was I referring to data from the machines?

Quite honestly, I'm not reading carefully this big debate about statistics and I'm not interested enough in all this to wade through those articles, checking out sample sizes, power analyses, etc. If you want to try ASV, try ASV. I sincerely doubt that -SWS would try to talk you out of it, particularly if you've exhausted your other options. So you probably don't need to work so hard to try to convince him that ASV is a modality worth trying for someone who has tried everything else...

Oh, definitely. I'm going to try ASV based on my own experiments and opinion. I've never been one to care much about other people approving my ideas. But I'd love to have more data to guide my decision, and I was hoping to get some here. I would also like others to have good data to make their own personal decisions. It doesn't serve anyone to have erroneously pro-ASV opinions or erroneously anti-ASV opinions. I believe strongly that people are best served by the truth.

A piece of advice from someone who's been around here a lot longer than you have (and whose FLs are worse than yours, if that impresses you ):

You're obviously very bright and trying hard to catch up reading all the stuff that I would have expected a guy as bright as you to have read BEFORE subjecting himself to invasive surgery. Whatever it is you know, however, is a drop in the bucket compared to what some of the people here know, especially -SWS and Mollete. So you can spend your time in a big debate about statistical significance and try to teach them something about the null hypothesis (and show off your statistical prowess) OR you can let them teach you and maybe, just maybe, make a dent in your EDS.

Yes, I should have read more (and tried xPAP more) before surgery. I had an irrational desire to find a permanent solution that wouldn't require all the tinkering and annoyance of CPAP. And I believed Dr. Li's 90% claim for MMA.

I'm not trying to show off my statistical prowess - I'm not even a statistician, I've only taken one stats class, I've just worked with statisticians enough to know the basics. I don't think I'm any kind of a stats whiz, I'm a first-grader compared to lots of people I know. But I don't understand how I can learn from someone who says "A is false because of B" when B does not actually disprove A at all. I'm here trying to learn, and trying to listen, but when we're arguing whether the guy who batted 0.320 last year is more likely to bat higher than the guy who batted 0.275, or more likely to bat lower, and someone says "Well, they are going to face some different pitchers this year, in a different order, with different weather" I don't how to learn from that, or do anything other than say "Huh?"

I would love to read anything that anyone has to say about addressing my EDS. That's why I posted my health history & sleepyhead data threads. I would especially love to hear from anyone who has successfully treated the situation I'm in - Mild AHI, significant FLs according to the machine, feel slightly better on CPAP, but still wake up 3-8x/night w/ nocturia (strong sign of untreated SDB), and feel unrefreshed and awful, have had tests on hormones and iron and all that stuff to look for other causes, and suspect that they are unusually sensitive to even mild SDB. The only doctor who claims to have a cure for this population is Dr. Krakow, and the only model that fits it is the "UARS as hypersensitive response to mild flow limitations" model. I would love to have other models and other doctors to read, and especially other patients to talk to who have been in the same situation and fixed it.

[patrissimo]

But statistically, this means that any evidence we have on, say, the performance of ASV vs. CPAP on one population applies 50% to the other. If ASV is twice as good as CPAP at reducing the OAI in ComplexSAS, we would expect it to be 1.5x as good on OSA. That is basic statistics. Of course, if we tested it, it probably wouldn't turn out to be 1.5x as good, because the two treatments likely wouldn't be exactly equal at the 50% of events that are unique to plain OSA. But the most accurate guess we can make using that information is 1.5x - that ASV will do twice as well on the half of events that are the same, and that the two will perform equally well on the events that are different (that's our best guess when we have no other information about relative performance on this sample).

You never presented a shred of evidence for commonality. Rather you worked on a vague and under-researched supposition of commonality as established fact. Sleep researchers have yet to agree on commonality, if any, in the pathogenesis for these two populations.[/quote]

I don't see why I need to prove commonality between the two populations when they are so similar. Sure, the opioid-induced SDB population may be totally different. But the ComplexSAS population? They started out the same as the OSA population, and the two split based on whether their obstructive apneas could be treated by a level of pressure that didn't introduce centrals. I am baffled that you could think that makes the two populations 100% different. Or even 50% different!

We start out with 100 people with OSA, and we crank up their CPAP pressure. 80 of them find a pressure that brings their AHI < 1, and causes no centrals. The other 20 of them develop CPAP-induced central sleep apnea, and are moved to ASV machines, which bring their AHI < 1. It seems you are saying that the obstructive apneas that come up during the night for the first 80 people have a completely different pathogenesis than the obstructive apneas that come up for the second 20 people.

That just seems really implausible to me. I'm not saying that the pathogenesis is 100% identical, I totally buy that there are differences between the populations, and differences between the obstructive events that they experience, but how can you claim it is 100% different and has no similarity? There are only so many mechanisms for pathogenesis, it seems really implausible that the separation procedure that I just outlined would neatly divide the 100 people into 80 people with only pathogenesis types A, B, and C, and 20 people with only types X, Y, and Z. Doesn't it?

In order for ASV vs. CPAP performance on obstructive apneas in ComplexSAS to tell us *absolutely nothing* about ASV vs. CPAP performance on obstructive apneas in OSA, it would have to be the case, not merely that they are somewhat different, not merely that the populations aren't identical, but that there was absolutely no similarity. If there is any overlap in pathogenesis between the events in the two populations, then we can statistically infer that a treatment that is better on one is, before testing, more likely than not to be better on the other.

Your scope of statistical analysis also fails to factor iatrogenic effects of unnecessary PS.[/quote]

Yes, definitely. I was only directly addressing the question of whether we can extrapolate from the performance of ASV vs. CPAP on obstructive apneas in the ComplexSAS population to an expected performance difference on the OSA population. My reason for doing that was to extrapolate an expected performance difference on not just apneas, but hypnopneas, and not just hypnopneas, but flow limitations of all sorts. And my reason for doing that was to make guesses about which would best relieve EDS in a person with SDB. As the inferences get increasingly remote from the original data, more and more other factors come into play. Perhaps one lowers the AHI through a treatment that also narrows the glottis by adduction of the vocal cords, decreasing the effective ventilation that reaches the lungs, and worsening sleep fragmentation and EDS. I am open to other metrics and other treatments.

But wouldn't an ASV tend to have less unnecessary PS, due to the split-second response, then something like fixed CPAP or BiPAP which you have to set high enough to splint the airway during all sleep stages?

[old64mb]

Patrissimo,

First, I'm sorry you've had so many problems before coming here and that they're still bothering you so severely. As you have noted, you might have been a lot better off had you found resources like this before taking radical steps like surgery.

Second, the very short answer to your question about ASV and unresolved hypopneas is...maybe. Several of the senior people on the board who you've not run across have reported pretty good experiences with it doing so.

Third, as you've been directly diagnosed with UARS by Dr. G himself, I'd suggest a reasonable approach would be to exhaust the known modalities of treatment for UARS prior to experimenting with other stuff. I've never heard of an ASV being prescribed for UARS without comorbidities, but I have heard of bilevels working quite effectively. You've only been back on straight CPAP for, what, a month? There's a good clinical reason why step therapy is standard medical practice (before it was warped by the insurance companies to create massive headaches for physicians so they could save a few bucks), because part of it is to see how you react to different stimuli.

You may do really well on a bilevel. You may find a bilevel horrible, as some UARS patients have awful experiences with bilevel pressures. I've talked to several UARS patients who actually did best by switching between different brands of straight CPAPs and eventually finding one with an algorithm that best suited their condition. I think that's something the Stanford people could help you out with a bit, since this gets into really small sample sizes seen in clinical practice and given your surgeries, the relevant comps are even smaller than that. They've got more experience dealing with it than anyone else.

But to jump directly from CPAP to ASV is not something I'd suggest for anyone. Some of that is because when you start dealing with the central-predominant SDB conditions that ASV gets prescribed for, the pathogenesis is radically different than OSA or UARS. Practitioners have got to do some pretty careful calculus about what's the most important condition to treat; sometimes, chasing a lower AHI can come second.

However, in your case, even without centrals, you have to get the IPAP and EPAP settings right (which you can do with a bilevel) before seeing what levels of additional pressure you might need to deal with those unresolved hypopneas. And more importantly, the experiences that people have had in bringing AHI from say 2 to near 0 don't necessarily resolve EDS. It may be worth a shot to try down the road once you've exhausted everything else, but you'll also never get an insurance company to pay for an SV and will have to go into the grey market or buy it yourself, as I doubt a DME would rent one to you either. On the other hand, they probably will approve you for a bilevel if you request one at this point.

Finally, just a thought. Almost everyone who suffers from non-treated SDB tends to be a little bit grumpy and argumentative as part of the symptoms of the condition. It's worth thinking about that before you hit the submit button, since most of us are trying to help even if we disagree with you. Don't take this the wrong way, but arguing back and forth over trivial stuff tends not to get a lot of people willing to spend time working with you here.

[-SWS]

I don't see why I need to prove commonality between the two populations when they are so similar.

Actually, they might not be. One prominent theory has OA in the OSA population as passive relaxation of muscles related to airway maintenance. Another has OA in the CompSAS population as a neurologically active albeit maladaptive stimulus/response type closure. OA pathogenesis in the two populations might be largely dissimilar. I'm guessing there's a very limited variety of pathogenic types and gradients in each population. The truth is none of us knows whether the OA pathogenesis in those two populations are largely similar or largely dissimilar---or somewhere in between. Given that uncertainty, I wouldn't perform a basic statistical analysis. I'm pretty sure any epidemiologist worth her or his weight in salt would not accept your statistical analysis as "evidence" either.

Your scope of statistical analysis also fails to factor iatrogenic effects of unnecessary PS.

Yes, definitely.

You endeavor to use your own fuzzy associations as the basis for statistics----but there's a fine line between parsimonious analysis and under-factorized analysis. Rather than spending time to validate those fuzzy associations, you forged ahead with statistics as if your cursory premise was just as good as fact. We will believe that you bring us "truths" with your statistical analyses when you have earned that credibility. Not because you insist that your analysis is right and that it's important for us all to understand and accept what you have to say.

But wouldn't an ASV tend to have less unnecessary PS, due to the split-second response, then something like fixed CPAP or BiPAP which you have to set high enough to splint the airway during all sleep stages?

CPAP doesn't offer PS (okay, arguably a little when exhalation relief is enabled). BiLevel tends to offer much less PS than that large PS frequently delivered by ASV. The problem with ASV's split-second response is that it is essentially myopic: 1) no temporal analysis---just short-window baselines and flow targets, and 2) inadequate context analysis. As a result, ASV will sometimes target and perpetuate too-large volumes and at times too-small volumes.

[johnthomasmacdonald]

I've already gone this route with self treating myself with an asv unit (a resmed s8 type) and so far it has been a big improvement ( however what's weird about my particular variety of apnea ( it even perplexes Pugsy) is that it is very erratic - I can go from an ahi of 1 to an ahi of 21 from one day to the next with no change in anything that i am aware of). I was originally diagnosed by a home test with an ahi of 21 and sold an PR sys 1 apap machine - i hated it, could barely tolerate it and my ahi was all over the place - tried it as a cpap, no improvement. I then got a resmed s9 vpap unit on the "grey" market and it was a HUGE improvement in comfort ( i could tolerate it all night from the first night) and some improvement in consistency - now my ahi ranges from 2 - 12 rather than 2-21 ( i have severe nocturia when my ahi goes much above 5 - like 2-3 liters so i have a good secondary feedback as to how well the treatment is working). I sent the sd card back to the company that sold me the apap and they let me trade it in on an apap/bipap since they could see it wasn't working, sadly it was no improvement over the straight bilevel, bye bye money ( actually not as good really, but it did a good job of providing me with settings for asv ( which i again purchased on the grey market). So far, so good - it's only been a week but the worst the ahi has been is 5.9 and my nocturia is cured, i sleep through the night without having to get up. I didn't have any problem adjusting to the asv actions. I don't know if i could get further improvement with tweaking the unit but while it seems to be working i don't want to mess with it in any way

[4betterO2]

The problem with ASV's split-second response is that it is essentially myopic: 1) no temporal analysis---just short-window baselines and flow targets, and 2) inadequate context analysis. As a result, ASV will sometimes target and perpetuate too-large volumes and at times too-small volumes.

Sorry I've not properly read this interesting topic throughout; but the sentence quoted above got my attention.
Are you talking here about a ResMed or a PR machine? Do they apply ASV differently?

AFAIK, the PR ASV monitors each breath for flow volume, and calculates and applies a response based on 4 breaths.
So that is a quick but not split-second response, and it effectuates a short but not absent temporal analysis.
what window size would prevent the offshoots in the high and low end of pressure?
Of course, setting desired limits instead of leaving everything on auto, can protect the patient against these offshoot pressure values.

By "inadequate context analysis", I'm not sure what you mean, it does not include other parameters than flow volume?

Are you talking here about a ResMed or a PR machine? Do they apply ASV differently?

[4betterO2]

btw do you now a link to technical/clinical training material for the ASV?

This clinical document is a PDF file: BiPap autoSV Advanced-- System One clinical applications guide. That's the only clinical training material for your machine I know of. But I bet there's plenty more kicking around out there for professionals. You might ask your DME for a copy of the clinician's manual or request one here: http://www.apneaboard.com/adjust-cpap-p ... tup-manual.

Thanks so much for these links ! It will take me a while to absorb the material.

[-SWS]

The problem with ASV's split-second response is that it is essentially myopic: 1) no temporal analysis---just short-window baselines and flow targets, and 2) inadequate context analysis. As a result, ASV will sometimes target and perpetuate too-large volumes and at times too-small volumes.

Sorry I've not properly read this interesting topic throughout; but the sentence quoted above got my attention.
Are you talking here about a ResMed or a PR machine? Do they apply ASV differently?

Regarding running baseline, Respironics ASV targets 95% of recent-averaged peak flow while Resmed ASV targets 90% of recent-averaged minute volume.

AFAIK, the PR ASV monitors each breath for flow volume, and calculates and applies a response based on 4 breaths.
So that is a quick but not split-second response, and it effectuates a short but not absent temporal analysis. what window size would prevent the offshoots in the high and low end of pressure?

I think Respironics employs a 4-minute sliding window for their peak-flow baseline while Resmed uses a 3-minute sliding window for their minute-volume baseline. In the course of prolonged hyperpnea or unresponsive hypoventilation those windows are fairly myopic IMO. You're essentially asking me what a better design might be. I think the answer to that might entail additional temporal variable analysis and limits rather than elongating the baseline window. I'm sure that's an extremely challenging design problem. The ASV algorithms will no doubt continue to improve.

By "inadequate context analysis", I'm not sure what you mean, it does not include other parameters than flow volume?

The current ASV algorithms will periodically target too-large breaths, too-small breaths, and just-right breaths throughout the same night---irrespective of whether those offshoot volumes are beneficial or deleterious.

[mollete]

Respironics ASV targets 95% of recent-averaged peak flow...

So it strikes me that it would only work when one's baseline peak flow is normal (and one has real "peaks"):



However, if one's baseline is flow-limited, the most one could hope for is 95% of that, which may not attain "normalcy".

Perhaps one could hope that the ASV would see a failure to respond to PS and identify the event as obstructive-- however, the PR ASV response to obstruction is raise EPAP (that's the only thing it can do to correct the underlying issue), and not simply chase it all night long.

Which is why I contend that an AutoAnythingElse algorithm is better suited to attack obstructions (i.e., look for snores and flow-limitations, positive response to pressure increases, etc.).

[mollete]

However, if one's baseline is flow-limited, the most one could hope for is 95% of that, which may not attain "normalcy".

Of course, one could argue that during a prolonged stretch of FL that theoretically one would not need a huge drop to elicit a response from ASV (because flow rate is already reduced). But you would need some sort of an acute drop, and hope that ASV would correct it before the subject goes to arousal, which is the whole goal in treating UARS. It does not matter how good one makes the waveform look, if arousal is not prevented, then no real progress can be made.

[mollete]

The ASV algorithms will no doubt continue to improve.

Well, "IMHO" the target group is getting so tiny it may not be worth the R&D money.

However, I could see where a PAV-based algorithm could come into play, where you could have two treatment arms-- a Volume Assist mode to attack central component, and the Flow Assist arm to address obstruction. But again, it's difficult to cram in a (semi-) reliable diagnostic central tool like FOT or Pressure Pulses before it's time to send in a breath, so trying to get all this done and have a custom breath ready in 6 seconds is going to be tough.

[mollete]

...a Volume Assist mode to attack central component...

Although in the strictest sense, it couldn't be "Volume Assist" during frank central apnea, when there wouldn't be anything to "Assist".

[old64mb]

It does not matter how good one makes the waveform look, if arousal is not prevented, then no real progress can be made.

Agreed.

The main reason I'm not a huge fan of auto xPAP is the latter point, where it takes a series of events to actually kick up pressure support. By then, the user has already more than likely already had arousals, been kicked out of Stage III/IV and/or REM sleep, and has to try to start their sleep cycle all over again. Interrupt a cycle more than say once or twice an hour and you're going to have someone who doesn't end up with enough full cycles by the end of the night to feel rested.

I think Respironics employs a 4-minute sliding window for their peak-flow baseline while Resmed uses a 3-minute sliding window for their minute-volume baseline.

One of the things that I think often gets overlooked about SVs in actual use is the practical effect of the backup rate. If the SV is programmed to ensure that you're supposed to be breathing every x amount of seconds and it thinks you haven't inhaled, well, a few cm more inspiratory pressure is going to find its way to you on a specific breath - regardless of whether or not the algorithm determines overall pressure needs to go up after that breath based on flow volume. I'd argue that's not insignificant in eliminating events, even if those events aren't actually the centrals backup rate is designed to address.

[-SWS]

The ASV algorithms will no doubt continue to improve.

Well, "IMHO" the target group is getting so tiny it may not be worth the R&D money.

However, I could see where a PAV-based algorithm could come into play, where you could have two treatment arms-- a Volume Assist mode to attack central component, and the Flow Assist arm to address obstruction. But again, it's difficult to cram in a (semi-) reliable diagnostic central tool like FOT or Pressure Pulses before it's time to send in a breath, so trying to get all this done and have a custom breath ready in 6 seconds is going to be tough.

I see Resmed recently added a separate obstruction-addressing treatment arm to their ASV:

...auto-adjusting EPAP in ASVAuto mode or fixed EPAP in ASV mode.

http://www.resmed.com/us/products/s9_vp ... clinicians

One of the things that I think often gets overlooked about SVs in actual use is the practical effect of the backup rate.

Good point.