Between these two auto cpap machines, which one has a better algorithm?
They are both fine machines with auto algorithms that have been thoroughly tested. And both machines can provide top-notch therapy to a patient. But, while the algorithms are not the same, it's impossible to say one is somehow better
than the other. It's rather like debating whether Coke or Pepsi tastes better: It all depends on who you ask. And rather like Coke vs. Pepsi, some folks have very, very strong preferences, but for the most part, most people can't really tell enough of a difference to for it to matter.
Still there are some real differences in the algorithms and for a small minority of patients, the subjective way the two machines feel will make a difference in their comfort while using the machines.
Both machines will respond to snoring, flow limitations, and clusters of OAs and Hs by increasing the pressure. Both machines will attempt to determine whether a given apnea is a clear airway apnea or an obstructive apnea. Both machines will lower the pressure back down after the algorithm is satisfied that your breathing is now stable.
The S9 algorithm tends to respond to events by rapidly increasing pressure and then, once it is happy with the shape of the wave flow, it immediately starts to slowly decrease the pressure back down. And it will keep decreasing the pressure until the machine detects snoring, flow limitations, OAs, or Hs. If more events occur, the machine once again will rapidly increase the pressure. This gives the S9's pressure curve a characteristic "wave" appearance where the fronts of the waves are steep and the back sides of the waves are much more gently sloped.
The System One algorithm is slower to respond to events and is slower to start decreasing the pressure back down once it is happy with your breathing. The System One also uses a "hunt-and-peck" algorithm for determining the optimal pressure: About every ten minutes or so, the System One will increase the pressure by 2cm over something like a two minute period while checking for improvements in the wave flow pattern. If improvements are found, the baseline pressure is reset to the pressure that gave the optimal wave flow. If no improvements are found, the pressure is then dropped back to the baseline setting over the course of a minute or so. And the machine waits for about 5-6 more minutes before starting the new hunt-and-peck cycle. To decrease the pressure, the System One does a reverse "hunt-and-peck": It temporarily decreases the pressure slightly looking for any deterioration in the shape of the wave flow. If it sees any deterioration, no matter how minor, it bumps the pressure back up to the current setting. But if no deterioration is seen, then the baseline pressure setting is reset at the lower level. And the machine then waits for about 5-6 minutes before testing whether the pressure can be decreased again. The "hunt-and-peck" algorithm used on the Respironics machines gives the pressure curve a characteristic saw tooth appearance.
The net result of both algorithms is that the machine can gracefully respond to a need for a modest increase in pressure and hence improve patient comfort by reducing the overall pressure needed to splint the airway open. Both machines have some problems providing high quality therapy if the minimum pressure setting is way, way below what the person needs. (Hence running the machine wide open with a pressure range of 4--20cm usually does not improve patient comfort.)
But each has its own strengths and weaknesses in terms of therapy:
The S9 can respond very quickly to a nasty cluster of events (in an effort to "bust the cluster up" as it's often put around here). And that quick response to events is a real strength of the S9 algorithm. But it can be very slow to lower the pressure back down. And if the min pressure is set far too low, it can lower the pressure back down to a pressure level that is clearly not high enough, which can then trigger another cluster of events. The fact that the S9 increases pressure only in response to snoring, flow limitations, OAs, and Hs also means that it always
waits until something happens to increase the pressure and it always
tries to lower the pressure all the way back down the minimum pressure setting.
The System One is much slower to respond to clusters of events. Judging from the pressure curves I've seen it appears that the System One will raise the pressure in 1-2 cm increments once a minute in response to snoring, flow limitations, OAs, and Hs. And then it waits to see if things clear up before it increases the pressure again by a 1--2 cm increase. So that can make it less effective at "busting up" clusters of events. But the System One is less likely to lower the pressure too much after the events are over. And theoretically that "hunt-and-peck" algorithm proactively raises the pressure before
the flow wave deteriorates to the point of where snoring, flow limitations, OAs, and Hs can take place. In other words, the "hunt-and-peck" algorithm is supposed to help the machine figure out the appropriate pressure level even in the absence of event clusters
As for comfort? Again there are strengths and weaknesses of both algorithms:
The S9's rapid pressure increases disturb some folks and wakes them up. Those rapid pressure increases can also trigger leaks. But except for when the machine is actively increasing the pressure in response to clear events (including snores and flow limitations as well as OAs and Hs), the machine is running at a stable or slowly decreasing pressure. And if the min pressure is pretty much set to what you need for at least 50%-60% of the night, you'll have nice long periods where the pressure is not changing at all.
The System One's pressure increases are not dramatically fast. And that means they are less likely to wake you up. And they're less likely to trigger a leak. But that hunt-and-peck algorithm is constantly kicking in. And so every 10 minutes or so there is a pressure increase. And that's enough to bother some people. But that hunt-and-peck algorithm is also going to help keep the machine from decreasing the pressure below a therapeutic level when the min pressure setting is set too low: Once the pressure is suboptimal, the subtle differences in the wave flow combined with the hunt-and-peck algorithm will prevent the machine from further decreasing the pressure much of the time.
Overall, there is some evidence that the System One's algorithm provides less overall variation in pressure and a quicker return to the baseline pressure after a series of "events" and attempts to locate a meaningful baseline pressure through the hunt-and-peck algorithm. The S9's algorithm provides a quicker response to clusters of events, a gentler decrease in pressure going back down to baseline, and a stronger effort to use the min pressure setting as the target baseline pressure.
Finally, Robert McCoy and Ryan Diesem have written a scholarly paper that discusses the Resmed and Respironics auto algorithms, along with auto algorithms from several other manufacturers. The paper is a bit dated: The machines being looked at are the Resmed S8 and the Repironics M-Series. This generation of machine was not capable of distinguishing between CAs and OAs. But the two companies overall choices in the design of their auto algorithm have not changed and so the stuff in the paper about how and when pressure is increased and decreased pretty much applies to the current S9 AutoSet and the PR System One Auto. The link to this paper is given below.
Link to A Bench Comparison of Five Auto-Adjusting Positive Airway Pressure Devices, Response to Apnea, Hypopnea and Flow Limitation, by Robert McCoy BS RRT FAARC, Ryan Diesem BA, Valley Inspired Products Inc. (December 1, 2008)