(04-24-2016 09:29 PM)DeepBreathing Wrote: There's not really a lot of detail in this paper. However the prime criterion was the ability to achieve "breathing normalisation" which appears to be the ability of the machine to "blow through" an obstructive apnea, with only the Resmed and F&P reaching a high enough pressure to do so.
I think you are misinterpreting what the authors mean by "breathing normalization". I think the authors are talking about how quickly additional events are "eliminated" and the breathing returns to normal.
Of course, the authors are also running all the machines wide open---i.e. the machines are all running in APAP 4-20cm. And the authors have rigged the test so that the test SDB breathing pattern won't clear up until a pressure level of 12cm has been reached; in other words they keep throwing SDB breathing patterns at the machine every few minutes as long as the pressure is still below 12cm. And the test is run for only 30 minutes. So the test clearly favors machines that quickly raise the pressure at the first sign of trouble and to raise pressure rapidly until things improve. Once the pressure is above 16, only a few isolated events or flow limitations get through to cause further increases.
In their (highly artificial) test, the A10 running in both AutoSet modes breaks up the apnea cluster within the first 5 minutes by aggressively raising the pressure from 4cm to 16cm during that time. Think about that, however: The machine has just raised the pressure at an average rate of 2.4cm per minute for 5 minutes
. That's enough to wake a lot of us up. Or trigger aerophagia. Or both.
In other words, the way the authors set the test up means that APAPs that are less aggressive are at a serious disadvantage in this particular test. The machines that don't "normalize breathing" all probably do a fine job in treating OSA while running in APAP mode if they are used with a reasonable min pressure setting.
Quote:I wonder how useful this type of study really is? There's possibly a lot of unreported raw data which may be useful to machine designers and engineers. But I don't think this type of trial can in any way be taken as a guide to how the machines will perform on real patients in their own beds.
Bench tests that are funded by company money usually are not designed to "prove" anything about how the machines will work in practice. They are, however, designed to show the strengths of the algorithm being used. In this case, the strength that is being demonstrated is the ability of the A10 to respond quickly to a dense cluster of events and potentially prevent the cluster from growing into a 30 minute long cluster. This can be a real advantage if a person needs a lot more pressure (as in 8-10cm more pressure) when they hit REM than when they're in non-REM sleep. There are people like that and if they can sleep through the wild pressure swings at the start a nasty cluster, they may be able to use a Resmed A10 AutoSet with a significantly lower min pressure than they would need for a PR DreamStation.
But another important thing about bench tests is that it allows you to see how different machines algorithms will kick in. The real takeaway from this bench test is this: The Resmed and Icon are far more aggressive in terms of pressure increases than the other machines in the study. In particular, they're far more aggressive than the PR machines.
There are (older) studies funded by PR where the criteria was set to favor machines that would "control" a given kind of test SDB breathing pattern with the least
average pressure. In these tests, the test SDB breathing pattern "clears up" quickly when the fake airway is given time to stabilize between pressure increases and/or responds well to the PR Search algorithm. And they couch the results in language about how other machines (particular Resmed) over titrate because they keep responding too aggressively to the breathing pattern.
Finally it is possible to do a bench test that doesn't clearly favor one approach over another. But it is more expensive to do such a bench test. You have to have multiple kinds of breathing patterns involved, including ones that are known to cause problems for any given approach and ones that play to the strengths of any given approach. Running machines wide open is considered part of the challenge: Too many manufacturers suggest in their advertising that running them wide open is an acceptable thing to do, or at least acceptable as part of a home titration. So you want to know what machines are going to do when run wide open.
The study published on the NIH site
was NOT funded by any manufacturer. They also ran the machines wide open, but with a variety of different kinds of test breathing, including a really interesting "full night" test breathing complete with fake REM cycles. The Resmed S9 and PR System One came out about about the same in the overall test, but did perform quite differently. And neither was successful in bringing the residual obstructive part of AHI down to below 5. But the machines that did bring the residual obstructive part of the AHI down below 5 were also noted as having real problems with detecting CAs (even when they claimed they could) and raising pressure in the presence of CAs. The authors made no conclusions about "machine A" being better or worse than "machine B". Rather they conclude that there is a wide variance in how APAPs react to the same breathing pattern and that variance may have significant clinical implications, particularly if an APAP is being used a home titration device in lieu of a titration PSG. They also point out that if APAP is going to be used as a full time therapy, then narrowing the range by increasing the min pressure up to just below the 90-95% pressure level should improve the efficacy of the APAP therapy since the machine will not need to increase the pressure as much if/when a cluster of events occurs.
The much older 2009 bench study published in Sleep Review
was also funded without manufacturer money. It was designed to investigate the varying responses of the APAPs to different kinds of SDB patterns and came away with a conclusion that manufacturer choices could lead some APAPs to deliver far more pressure than the patient actually needed. The aggressiveness of the F&P SleepStation and the Resmed S8 in attacking flow limitations with pressure was noted. The authors did invite the manufacturers to respond to the paper, and their responses are integrated into each section in a way that makes it clear that the manufacturer is talking. (The manufacturer comments are illuminating on the kinds of decisions that they each make in designing their APAP algorithms. Many of those choices are still evident in today's APAP algorithms.