Does leak rate affect OAI and CAI? My S9 data shows no.

Quite a few comments have been made in a number of threads about the effect of leak rate on the number of obstructive and central (clear airway) apneas. I decided to see if a relationship between leak rate and OAs and CAs could be seen in the summary data from my S9. I extracted the data for the 95% leak rate, OAs, and CAs from my summary data and put it into a spreadsheet. The graphs below are the result.

I looked at data from November to now, just over 5 months worth. Looking at the graphs below you can see there is no clear correlation between OAs and leak rate, and between CAs and leak rate. It looks like there might be a weak correlation between leak rate and OAs showing the number of OAs decreases as the leak rate increases. However, this really isn't the case. I think what we see on the graph is the result of having a greater number of data points at the lower leak rates, thus the greater likelihood of more OAs per unit of leak rate. A regression analysis within the spreadsheet calculates a slope of the regression line at -0.01 for the OA graph and 0.01 for the CA graph. In other words the line that best shows how the number of events changes as the leak rate changes is almost flat, meaning that the number of events is virtually independent of the leak rate.

Now, you'll see my apnea index is less than 2 and the leak rate is mostly less than 30. In fact 3/4 of the leak rates in my data are less than 17 lpm. I haven't counted, and don't plan to, how many data points are in the graphs. There are 154 points in the data list, but if there are 2 or more exact duplicates for any one point, only one shows on the graph.

So how applicable my results are to others is open to question. Nevertheless, I think this is kind of interesting as in my case at least, leak rate per se seems to have no impact on my OA and CA indexes.

As long as the leak rate is within the particular machine's ability to compensate, I wouldn't expect it to have much, if any, impact.

Although I expected my data to show much the same (cumulative 95% leak is 1.2), my export shows 0 for all but one day for leaks. I wonder if resscan.xml has something out of whack on my machine...

.

Thanks for sharing your work, idamtnboy.

BernieRay wrote:As long as the leak rate is within the particular machine's ability to compensate, I wouldn't expect it to have much, if any, impact.

I think idamtnboy presented data that his own leaks and apneas above don't correlate rather than presenting conclusive population or epidemiology type data. More importantly, his leak data provided are actually gross S9 leak estimates----and exclude measurement of potentially-disruptive dynamic/brief leaks. That means any correlation between highly transient leaks and apnea data will likely not reflect above----as a result of that leak data being imprecise PAP estimates rather than accurate measurement.

We have had multiple posters share charts in which their leaks and apneas seem to correlate more strongly than idamtnboy's. And we have collectively scratched our heads why some subset of posters seem to present a stronger leak/apnea correlation than other posters.

Rhetorically: can leaks serve as disruptive stimuli in at least some patients for: 1) defensive upper airway closures, or 2) central-controller loop-gain perturbations leading to central apneas?

-SWS wrote:...
We have had multiple posters share charts in which their leaks and apneas seem to correlate more strongly than idamtnboy's. And we have collectively scratched our heads why some subset of posters seem to present a stronger leak/apnea correlation than other posters.

Even within the machine's range to compensate for leaks? I could see things being somewhat mixed at fixed pressures and no EPR, but I would have thought that APAP, either w/ or w/o EPR, wouldn't have shown a stronger correlation unless it was near or above the machine's ability to compensate. Just goes to show that no matter how well a machine is designed to handle things, humans don't fit into nice, cleanly defined boxes.

It's too bad ResScan doesn't export all summary data rather than just 95% values. Having median and max to go along with it would have the potential to shed a bit more light on this.

-SWS wrote:...
Rhetorically: can leaks serve as disruptive stimuli in at least some patients for: 1) defensive upper airway closures, or 2) central-controller loop-gain perturbations leading to central apneas?

I would think so, but I haven't seen any evidence of either occurring in my data. I've never had any trouble using a nasal mask. In fact, if I can't breathe through my nose, I tend to wake up. I don't have many centrals either, so I can't shed any light in that regard.

Given how many people do have trouble adjusting to any xPAP, the answer to the head-scratching may lie in the realm of how individual CNS react to 1) forced air via nose and/or mouth and 2) changes in pressures due to leaks, if there aren't other underlying and unidentified processes in play. Heck, I'm still wondering why I have no correlation between flow limitations and obstructive events.

If figuring out all the pieces that come into play were easy, then our machines would probably handle most events for most people more consistently and there wouldn't be such unexplained differences.

BernieRay wrote:It's too bad ResScan doesn't export all summary data rather than just 95% values. Having median and max to go along with it would have the potential to shed a bit more light on this.

Those two are in the str.edf file, so maybe in a day or two I'll plot them just to see what kind of difference there is. I just this would be interesting to post mostly to show that maybe we, the forum members, sometimes get caught up in trying to capture clouds.

Yeah, well. I recently tried going back to a nasal mask after using a full face mask for almost a year. Man! Some of the leaks would raise your hair. Obviously, a full face mask has allowed me to open my mouth at will so I no longer even try to keep it closed and would have to "re-train" myself before I could go back to a nasal cushion mask.

But I digress: when I had a HIGH leak rate and a LOT of HIGH leak rate my AHI data went DOWN. Not up. Go figure. I think the ole PAP just gave up trying to compensate or respond or collect any data until the leak comes down to a somewhat decent level.

Conversely, if I have a bad night w/my full face and some high but not unreasonable (over 40 but not outta sight) leaks my AHI will be higher than if I have a really low leak rate (like less than 15-20).

Slinky, I was just the opposite. After 10+ years on a nasal mask, I tried a full face mask and had all kinds of troubles with it. At the time, that was a bummer because it meant nightly Afrin (Flonase and Astelin had quit working) until I was able to have sinus surgery a few months later. It's a good thing we have options.

idamtnboy wrote:

BernieRay wrote:It's too bad ResScan doesn't export all summary data rather than just 95% values. Having median and max to go along with it would have the potential to shed a bit more light on this.

Those two are in the str.edf file, so maybe in a day or two I'll plot them just to see what kind of difference there is. I just this would be interesting to post mostly to show that maybe we, the forum members, sometimes get caught up in trying to capture clouds.

Ah! Serves me right for assuming you used a ResScan export instead of EDFBrowser.

Man, I am so glad that I'm not the only one who sometimes thinks we're grasping at clouds, all potential value as a source of humor to Heisenberg not withstanding. I look at my data every morning, though. I guess it's a combination of being a data junkie and those 3 years of engineering courses that I tossed aside for a degree in computer science. I just can't not look!

Re-edit

From Resmed's engineers about the S9 AutoSet:

The CSAD algorithm determines resistance of the upper
airway. To do this accurately the impedance of the circuit
needs to be known. CSAD works optimally when recommended
circuit configurations are used, but it does not rely
on correct menu setup1. The ability to determine resistance
is also affected by large leaks. Small leaks are tolerated but
the accuracy falls rapidly with leaks greater than 30 L/min.

Apneas are classified as being central (open upper airway),
obstructive (closed upper airway) or unknown. Central apneas
are scored when the resistance is low, and obstructive
apneas when the resistance is high.

The classifier
labels apneas ‘unknown’ when the inadvertent leak exceeds
30 L/min or the resistance is indeterminate2. Central
and unknown apneas do not cause an increase in delivered
pressure.

.

lars4life wrote:The vertical axis is only labeled Avg Leak.
Does anyone know what this means?

It's just an overall average leak number for the hours involved. Depending on what your default software is set at (Total leak or Unintentional Leak) it is just an average. High leak and low leak all averaged out.

Just to add to Pugsy's reply, the values are most likely liters/minute.

To add a little more, it's probably the median leak rate rather than the mean.

If we had enough data, we might see a correlation between leak rate and OAI for APAP users simply because the machine will run at a higher average pressure on nights when the user's airway is more obstructed. For some people, there would also be a correlation because apnea causes them to move around more and hence be more likely to dislodge the mask. Also leaks make FOT less accurate, which could compromise both data and therapy. Add in that large leaks wake people up and so both prevent OAs and score false CAs, and I think there are just too many variables for a ragtag band of hoseheads to properly analyze. Even for people who do see a correlation between leak rate and OAI or CAI, it's near impossible to tell what that means.