AHI - what's the magic number?

[split_city]

Ok, I have asked our sleep lab manager why we have classified 'normal' RDI as <15. I must make an apology as I had assumed this was the criteria for all labs in Australia. However, this is the criteria set for our lab only. Other labs have not adopted it at this stage. The reason why we classify normal as being <15 is due to a study our lab ran which looked at what normal was in a group of asymptomatic normal subjects. This is also at the bottom of our sleep reports

Normative data: In an Australian study1 of randomly-selected asymptomatic normal subjects aged 30 - 70 years respiratory disturbance index (apneas + hypopnoeas/ hr of sleep) was 6.1 + 7.2 (Mean + 2SD). i.e. RDI upper limit of "normal" in this population was approximately 13. Clinicians should, however, note that RDI was averaged for the whole study population and RDI is known to increase with age. Further, epidemiological studies suggest that increased risk of hypertension, cardiovascular disease and neurobehavioural disturbance may be present at RDIs as low as 5-10/ hr. Thus, when advising an individual patient, judgement is required in assessing what constitutes a "normal" RDI
Current versus historical RDI values.

This along with the use of pressure transducers (information below) resulted in our lab's current criteria for assessing OSA severity

In accordance with recent internationally agreed recommendations2, the AISH laboratories transferred in 2002 from using oronasal thermistors to using nasal pressure transducers to measure airflow and at the same time adopted a new standard for scoring apneas and hypopnoeas2. Nasal pressure is a more sensitive measure of airflow and as a consequence, average AISH reported RDI values are now higher for the same severity sleep disordered breathing. The best estimate is that RDI has increased by approximately 40.%. This is relevant when interpreting current patient results with respect to our earlier studies of normal subjects1 and other previous epidemiological studies, which relied on thermistors to measure airflow.

Severity of sleep disordered breathing. Bearing in mind the above, AISH consultants have adopted the following consensus classification of severity of sleep disordered breathing

RDI: <15 Normal (see above caveats)
>15 - <30 Mild
>30 - <45 Moderate
>45 Severe

Hmmm, now SAG is greatly confused. Regardless of the diagnostic airflow methodology used, during intervention the waveform from the PAP machine is used to score events, and that is a pressure transducer. Are you then saying that RDI during titration and APAP is normal as long as it is <15?

No, we try to abolish respiratory events. But this isn't achievable in all cases.

Also, what are your components of RDI? Are they apneas, hypopneas and RERAs? RERAs and hypopneas need to have an associated event, be it an arousal and/or a desaturation. And >5 events per hour of those events are abnormal.

Yes, I'm pretty sure what you have in bold are the components of RDI

Or put another way, the diagnostic device selected is really academic. Either device (thermistor or pressure transducer) will identify an apnea, but beyond that, a respiratory event without an associated desaturation or at least an arousal is nothing.
SAG

Or put another, another way, let's say a patient has 14 respiratory events, all of which cause desaturations. Under the thermistor rule, that would be a problem. Are you saying that under pressure transducer, that's no longer a problem?

OK, let's say instead, there were 14 arousals that were associated with thermistor-defined events. Are you saying that with pressure transducer technology, they are no longer considered to be respiratory-associated events? Especially if the waveforms have a flattened appearance? And if not, then what do you call them?

The whole idea behind pressure transducer is that it picks up events that thermistor may miss. And it's not the respiratory event per se that's the issue, it's the associated desaturation, arousal and/or increased effort (i.e., the flattened waveform seen in UARS). And that number is almost universally accepted as being 5.
SAG

I think the confusion to why our RDI is so high is due to our classification of what a hypopnea is. Our hypopnea rule is "a discernable decrease in amplitude (flow) accompanied by a 3% or greater desaturation." I.e. a 50% decrease in flow is not used

_________________

CPAPopedia Keywords Contained In This Post (Click For Definition): Titration, Arousal, RDI, Hypopnea, APAP

[Nev]

The reason why we classify normal as being <15 is due to a study our lab ran which looked at what normal was in a group of asymptomatic normal subjects. This is also at the bottom of our sleep reports

Normative data: In an Australian study1 of randomly-selected asymptomatic normal subjects aged 30 - 70 years respiratory disturbance index (apneas + hypopnoeas/ hr of sleep) was 6.1 + 7.2 (Mean + 2SD). i.e. RDI upper limit of "normal" in this population was approximately 13. Clinicians should, however, note that RDI was averaged for the whole study population and RDI is known to increase with age. Further, epidemiological studies suggest that increased risk of hypertension, cardiovascular disease and neurobehavioural disturbance may be present at RDIs as low as 5-10/ hr. Thus, when advising an individual patient, judgement is required in assessing what constitutes a "normal" RDI

Can you point me to the original study referred to?

How were the "asymptomatic" "normal" subjects identified? If they had RDI of 6 or up to 13 they were hardly asymptomatic??

If we define "normal" as up to 15 and then study a "normal" population so defined it is hardly surprising that we find they have RDI of up to 15?

[split_city]

The reason why we classify normal as being <15 is due to a study our lab ran which looked at what normal was in a group of asymptomatic normal subjects. This is also at the bottom of our sleep reports

Normative data: In an Australian study1 of randomly-selected asymptomatic normal subjects aged 30 - 70 years respiratory disturbance index (apneas + hypopnoeas/ hr of sleep) was 6.1 + 7.2 (Mean + 2SD). i.e. RDI upper limit of "normal" in this population was approximately 13. Clinicians should, however, note that RDI was averaged for the whole study population and RDI is known to increase with age. Further, epidemiological studies suggest that increased risk of hypertension, cardiovascular disease and neurobehavioural disturbance may be present at RDIs as low as 5-10/ hr. Thus, when advising an individual patient, judgement is required in assessing what constitutes a "normal" RDI

Can you point me to the original study referred to?

How were the "asymptomatic" "normal" subjects identified? If they had RDI of 6 or up to 13 they were hardly asymptomatic??

If we define "normal" as up to 15 and then study a "normal" population so defined it is hardly surprising that we find they have RDI of up to 15?

Here's the study:

SIOBHAN BANKS1 , 2 , MAREE BARNES3 , NATALIE TARQUINIO3 ,
ROBERT J . PIERCE3 , LEON C. LACK1 , 4 and R . DOUG McEVOY

Factors associated with maintenance of wakefulness test mean
sleep latency in patients with mild to moderate obstructive
sleep apnoea and normal subjects

J. Sleep Res. (2004) 13, 71–78

This is how 'normal' patients were recruited:

Forty-one normal subjects also participated in this study. They were randomly selected from the telephone book within a 30-km radius of both the hospitals and were invited to participate via a telephone call. They answered a general health questionnaire, which screened for health and sleep difficulties. Normal volunteers were excluded from the study if they were <30 or >70 years of age, if English was not their first language, if they were heard to snore more than one night a week, or if they were on any medications for respiratory, cardiovascular or psychiatric disorders, or if they had a history of drug or alcohol abuse.

Participants undertook a PSG and also a MWT. MWT was 36.7mins while ESS score for these volunteers was 5.44. Mean AHI was 6.1+/-3.6. Therefore, 2SD = 7.2, giving us our 'upper limit' of 13 for 'normal' AHI. Our lab decided that 15 would be the cut off for normal

_________________

CPAPopedia Keywords Contained In This Post (Click For Definition): RDI

[StillAnotherGuest]

Forty-one normal subjects also participated in this study. They were randomly selected from the telephone book within a 30-km radius of both the hospitals and were invited to participate via a telephone call. They answered a general health questionnaire, which screened for health and sleep difficulties.

41 guys, huh? Not exactly the Sleep Heart Health Study.

Regardless, the rationale is a little backwards. If your subjective analysis makes them "normal", but then your objective data (PSG) says they're not, I would contend that you can't redefine the rule book so that your objective data now coincides with the subjective data. And extending out 2 SD simply allows you to make all your "normals" normal.

Normal volunteers were excluded from the study...if they were heard to snore more than one night a week...

Who was listening? Did they set their alarm clock to be sure they were awake to track REM periods? Oh wait, that won't work...

Nasal pressure is a more sensitive measure of airflow and as a consequence, average AISH reported RDI values are now higher for the same severity sleep disordered breathing. The best estimate is that RDI has increased by approximately 40.%. This is relevant when interpreting current patient results with respect to our earlier studies of normal subjects1 and other previous epidemiological studies, which relied on thermistors to measure airflow.

If your RDI includes RERAs, then RDI won't increase at all. You should have been using some sort of method to identify RERA even with thermistor (RIP, sum channel, suprasternal notch transducer, esophageal balloon, increased abdominal/chest effort belt activity, etc.).

Then what are your normals for Arousal Index (AI)? By jacking up the RDI "normal" from 5 to 15, you're now saying that an AI of 15 is now fine, whereas when you were using thermistor it was a problem. The arousals are there whether you use thermistor or not. Using pressure transducer may make some RERAs hypopneas, but the RDI will remain the same.
SAG

[ozij]

Here is the full study from the Journal of Sleep Research, Volume 13 Issue 1 Page 71-78, March 2004


I couldn't find the so called normal AHI nubmers quoted by split_city.

As a matter of fact, here is the discription of the participants, in that study:

Participants
The patients for this study were recruited as part of a randomized, placebo-controlled trial investigating the treatment effects of CPAP and mandibular advancement splint. A total of 110 patients were invited to participate in the study after a clinical PSG that identified them as having mild–moderate OSA [categorized as an apnoea/hypopnoea index (AHI) between 10 and 30 per hour] but before any treatment was commenced. The patients had no significant co-morbidities (cardiovascular disease, respiratory disease or psychiatric disorders) and were not taking any regular medications (including hypnotic or anti-depressant medication).

My emphasis.

I also found that the study considered a score of 1 sd (standard deviation) below the mean low and 1 sd above the mean high. That was for another measure, not AHI - but so what, a standard deviation is a standard deviation.

Here's some info on probabiliteis in the normal distribution:
http://en.wikipedia.org/wiki/Normal_dis ... _deviation

Assuming OSA is normally distributed - and I don't see why this should be assumed, by pure mathematical definition, the lab's "normative" (mean + 2 standard deviations) data will include less than 5% of the population in the abnormal group.

This is an arbitrary calculation, based on the statisitical norms of what is considered "a significant differnce" that justifies getting a research paper published.

Which would be like defining the proper healthy level of sugar in your blood as that found in all but less than 5% of a non-symtomatic population. Or the right level of blood pressure, etc.

The meaning of "significance" in hypothesis testing is that there is less than a 5% chance of rejecting the Zero hypothesis by mistake. In other words, the measure mentioned by the lab has less than a 5% chance of mistaking a healthy person for one who needs therapy (what a relief...). Think of that: According to this "normative" measure, people will given therapy only if there is a less than 5% chance that they are actually healthy.

I wonder how many would die before they get therapy, just because someone was being extra extra carefull about not treating a healthy person.

Thank goodness someone in that lab has a head on their shoulders:

Further, epidemiological studies suggest that increased risk of hypertension, cardiovascular disease and neurobehavioural disturbance may be present at RDIs as low as 5-10/ hr. Thus, when advising an individual patient, judgement is required in assessing what constitutes a "normal" RDI

At least somone there does know the difference between the probability needed for publishing papers and to that relevant to treating people.


O.

[split_city]

Here is the full study from the Journal of Sleep Research, Volume 13 Issue 1 Page 71-78, March 2004


I couldn't find the so called normal AHI nubmers quoted by split_city.

As a matter of fact, here is the discription of the participants, in that study:

Participants
The patients for this study were recruited as part of a randomized, placebo-controlled trial investigating the treatment effects of CPAP and mandibular advancement splint. A total of 110 patients were invited to participate in the study after a clinical PSG that identified them as having mild–moderate OSA [categorized as an apnoea/hypopnoea index (AHI) between 10 and 30 per hour] but before any treatment was commenced. The patients had no significant co-morbidities (cardiovascular disease, respiratory disease or psychiatric disorders) and were not taking any regular medications (including hypnotic or anti-depressant medication).

If you read directly under the quote above in the paper, it explains the recruitment criteria of the 'normal' subjects. The normal AHI is in Table 1

[split_city]

Regardless, the rationale is a little backwards. If your subjective analysis makes them "normal", but then your objective data (PSG) says they're not, I would contend that you can't redefine the rule book so that your objective data now coincides with the subjective data. And extending out 2 SD simply allows you to make all your "normals" normal.

Why can't you redefine the rule book? It happens all the time..

Normal volunteers were excluded from the study...if they were heard to snore more than one night a week...

Who was listening? Did they set their alarm clock to be sure they were awake to track REM periods? Oh wait, that won't work...

I'm unsure of the exact questions they were asked in regards to "snoring." I wasn't involved in this study. I can find out though.

Nasal pressure is a more sensitive measure of airflow and as a consequence, average AISH reported RDI values are now higher for the same severity sleep disordered breathing. The best estimate is that RDI has increased by approximately 40.%. This is relevant when interpreting current patient results with respect to our earlier studies of normal subjects1 and other previous epidemiological studies, which relied on thermistors to measure airflow.

If your RDI includes RERAs, then RDI won't increase at all. You should have been using some sort of method to identify RERA even with thermistor (RIP, sum channel, suprasternal notch transducer, esophageal balloon, increased abdominal/chest effort belt activity, etc.).

Effort bands were used in the study and are used in current diagnostic studies.

[split_city]

Then what are your normals for Arousal Index (AI)? By jacking up the RDI "normal" from 5 to 15, you're now saying that an AI of 15 is now fine, whereas when you were using thermistor it was a problem. The arousals are there whether you use thermistor or not. Using pressure transducer may make some RERAs hypopneas, but the RDI will remain the same.
SAG

I think our RDI severity classification has increased due to our interpretation of what a hypopnea is. Unlike the classical hypopnea criteria, our lab looks for a at least a 3% desaturation with a discernable drop in flow (not 50% drop as stated elsewhere)

_________________

CPAPopedia Keywords Contained In This Post (Click For Definition): Arousal, RDI, Hypopnea

[ozij]

I stand corrected on those numbers, I did find them in Table 1, which is not mentioned in the participants descriptoin at all.

If I undestand you correctly, your lab is using non stadndard definitions of events. Those give higher base rates for everyone. In addition the lab prefers to state what at the upper limit of normal is on those measures - and reaches that by adding 2 standard deviations, and the some more points, to the mean score found by those supersensitive, non standard measures.

What I don't get is why create supersensitive measures, and then use them in roughest way by looking at the upper limit of whatever it is you're measuring, making absolutely sure you'll drop all but 5% of the population.

Nor do I undertand how you can use a 1 sd deviation to define a high and low score, and yet add 2 sd to a score, and still call it normal.
O.

[Nev]

Thanks for the reference to the original paper.

The 41 "normal" subjects were in fact a control group for a different study altogether. The rough and ready way they were selected may have been OK for that study which focussed on 110 mild-moderate sleep apnea patients. It's another matter to treat them as the sample for an analysis of AHI in the normal asymptomatic population. No-one actually looked for symptoms. They answered a questionnaire in which they reckoned they didn't have any synptoms. Before I was diagnosed as having sever sleep apnea, I would have said I only snored occasionally and would have given myself quite a good score on the Epworth scale. It came as a complete shock to me that I had an AHI of around 60. I think the results in the paper have been taken way out of context in using them to define the characteristics of a normal asymptomatic population.

It looks like a more specifically directed study should be undertaken before departing from the widely accepted scale regarding as normal only AHI up to 5. The accepted scale is after all based on a large number of studies of the physiological effects of sleep apnea i.e. looking at the detrimental effects and studying the level of AHI at which these effects start to become significant.

If someone can show that AHI of between 5 and 15 has no detrimental effects, then let's have another look at the criteria and not worry about treating people in that range. But I think the answer is already stated in the footnote where it is conceded: "Further, epidemiological studies suggest that increased risk of hypertension, cardiovascular disease and neurobehavioural disturbance may be present at RDIs as low as 5-10/ hr."