To give you a more technical answer:
Your CPAP machine detects apneas and hypopneas by analyzing the airflow into/out of your lungs. It tracks each breath to establish a running "baseline" for your normal breathing pattern during the last five or ten minutes or so. Whenever the airflow drops by 50% from the running baseline for at least 10 seconds, the machine will flag a hypopnea. Whenever the airflow drops by 80% from the running baseline for at least 10 seconds the machine will score an apnea.
Since you are using a Resmed S9, if the machine detects very little or no airflow (ie airflow has dropped by 80% from baseline) for about 6 seconds or so, the machine will use a proprietary algorithm called FOT to determine the patency of your upper airway. In other words, the machine will start oscillating the pressure at a set frequency and analyze what happens to the air flow. The air flow signal from the FOT is different when your airway is clear
(ie NOT obstructed) instead of when your airway is obstructed
(ie your airway has collapsed). If the pause in airflow is sufficiently long enough to score an apnea and the FOT detects an obstructed
airway, the machine will score an OA. If the pause in airflow is sufficiently long enough to score an apnea and the FOT detects that the airway is clear
, the machine will score a CA. Because the machine has detected no evidence of an obstruction, CAs are often thought to be central apneas
, but the definition of a central apnea on a PSG does not involve testing the patency of the airway---Central apneas on PSGs are scored when there's no air moving into/out of your lungs for 10 seconds or more and the belts around your chest and abdomen indicate you are making NO effort to breathe.
How accurate is a CPAP's AHI? Well it's accurate enough for trending
purposes. But it's not 100% on the spot accurate for any given night. To understand this, you need to know the following two definitions:
- In-lab PSG AHI = (number of events scored while you are asleep)/(total time you are asleep) where "you are asleep" is detected by the EEG readings.
- CPAP machine AHI = (number of events recorded during the night)/(total time the machine was running during the night)
Where do the errors in the machine AHI occur?
As others have noted, a CPAP machine cannot tell when you are asleep and when you are awake. Since the CPAP computes the AHI by simply dividing the number of detected events by the total time the machine was running for the night, that introduces two kinds of obvious errors as compared to an in lab PSG where the EEG is used to determine when you are asleep.
- False events are "apneas" and "hypopneas" that the machine scores when you are actually awake. On a PSG, any "event" that occurs during a time period when the EEG indicates that you are actually awake is NOT counted towards the AHI. The reason is simple: Sleep apnea is a breathing problem that occurs when you are asleep. The control of our sleep breathing is different than the control of our wake breathing. And our wakeful breathing is much less regular than our sleep breathing. We sigh. We hold our breath momentarily when concentrating on something as simple (or complex) as turning over in bed while managing a six foot hose attached to our nose. We consciously take several deep breaths in a row to relax ourselves and then breath much more shallowly for a few more breaths. And some of us have a bit of a tendency to not breathe for a bit when we're transitioning from Wake to Sleep as our body resets the control of breathing. On an in-lab PSG, none of these kinds of "events" count because we're awake when they start. But a CPAP doesn't know that and scores them anyway. The net result of these "false" events is to make the AHI increase a bit because there can be a few more events scored than should/would have been scored on a PSG. And so the numerator of the AHI fraction is higher than it should be, which makes the computed AHI go up.
- Inaccurate estimate of sleep time affects the denominator of the AHI fraction. The presumption by the CPAP makers is that your sleep time and the machine run time will be close enough to each other that the error introduced by using machine run time will be insignificant. And if you are one of those people who falls asleep within 5-15 minutes of going to be and doesn't spend much time awake during the night, that's actually a pretty valid assumption. But if you've got major insomnia and you're lying in bed with the mask on and the machine running for several hours each night, using machine run time instead of total sleep time to compute the AHI can increase the size of the denominator enough to reduce the AHI, perhaps significantly. (If two fractions have the same numerator, the fraction with the larger denominator will be the smaller of the two fractions---i.e. 29/8 is smaller than 29/6.)
people on most
nights, the minor errors caused by the machine recording a few false events and using the run time instead of the sleep time more or less cancel each other out. And so the nightly machine AHI is, for most
people a decent approximation
of the true AHI on most
nights, and so the machine numbers are useful for looking at trends.
In addition to the two obvious kinds of errors discussed above, there are also algorithm errors. Sometimes the machine will miss what looks like an obvious event---perhaps because the machine thinks it's not quite 10 seconds long. Or perhaps the drop from the current baseline is not quite enough. Sometimes the machine will score an event and you'll look at the data and not see anything that even vaguely resembles an apnea/hypopnea. These errors pop up often enough that if you zoom in on your wave flow data every single night, you're sure to see a few of them each week, but you probably wont see them every night. These random missed events and random extra events tend to cancel each other out in the long run. And so they don't tend to affect the overall trend of the AHI data over the course of several days or weeks.
And so when we consider the machine AHI data as trending data
, it is considered accurate enough to be significant, even if any individual night might be off.
For example, if you have one night with an AHI that is up around 9, but your typical AHI is usually between 1.0 and 3.0, that one night is likely just an outlier: It may have been simply been a bad night for your OSA. Or it could be that the data for that night is particularly inaccurate---perhaps because you were very, very restless all night long. But that one night with an AHI of around 9 doesn't really indicate that there's something wrong with your overall therapy because the trend of data is that your machine AHI is consistently below 3.0, which clearly indicates that your real AHI is consistently well below the desired 5.0.
But on the other hand, if you are seeing machine AHI's that are consistently above a 5.0 night after night for several weeks, then there's a good chance that something is not yet "optimized" about your therapy---particularly if you continue to experience daytime sleep apnea symptoms. It could be that your pressure is set too low and too many events are still occurring. It could be that you are in the unlucky 10% of new CPAPers who wind up developing problems with pressure induced central apneas It could mean that your leaks are not under control and you are losing too much pressure for your therapy to be effective. In any case, an AHI that is consistently above 5.0 for a period of several weeks should be brought to your sleep doc's attention.
Now in your case, you say the machine is recording a fair number of CAs, while your in-lab titration study only record one CA. If those CAs are clustered around times that you vaguely remember being awake/restless or occur just as you are falling asleep or occur just as you are waking up, then as Bons says, there's a pretty good chance they are false events---i.e. events that would not have been scored in the lab.
And yes, the restless legs might trigger inaccurate readings. And finally, if you are doing a lot of mouth breathing, that too can affect the accuracy of the data. You need to look at the leak data as well as the AHI data if you're doing a lot of mouth breathing.
I'd suggest that you have a printout of some typical night's detailed data with you when you have your follow up visit with your sleep doc. And get his/her take on what might be going on.