Respironics :: Activa, UMFF :: ResMed

[Velbor]

This posting is a continuation of my previous post,
“AutoPAP, Activa, UMFF and Mouth Taping” found at:
viewtopic.php?f=1&t=37954
There I described my methods of data collection and analysis, using a ResMed S8 AutoSet Vantage, presenting and comparing the performance of a ResMed Activa mask and a ResMed Ultra-Mirage Full Face (UMFF) mask at a variety of machine settings.

Reviewing the above post will provide a context for what I now begin to describe here. This post will describe the same sort of “experiments”, but this time using a Respironics REMstar M-Series Auto machine. Since I am using similar protocols with respect to the Respironics data, it will also be possible to very carefully compare this new information to the previous ResMed data.

Efficacy Statistics

This posting, beginning an anticipated series, will look at the efficacy and pressure patterns of the Respironics REMstar M-Series Auto (with A-Flex and C-Flex, but not using these functions; also no ramp) set in AutoPAP mode, with pressures from 6.0 to 20.0 cmH2O. (For those of you who do not “like” such a “wide” pressure range, please see my previously cited ResMed posting, which demonstrated no statistically significant difference between “wide” and “narrow” AutoPAP pressure ranges. In addition, I intentionally now use the same settings on the Respironics as I did with the ResMed, to facilitate comparisons.) The following table presents the average of nightly values, the range of nightly values, and the standard deviation. Percent time in apnea is calculated nightly from the EncoreViewer report “Therapy Data Summary – All Data” (last page) “Average Time in Apnea Per Day” divided by “Total Blower Time”.



(My one night using mouth taping reflects a triumph of curiosity over wisdom. Based on my ResMed findings, I simply had to determine whether the previously-noted degraded efficacy was or was not machine specific. I intended to spend 5 nights using tape; after reviewing the data from a single night I abandoned this plan. Clearly, all efficacy parameters are markedly worse when tape is used, particularly those related to apnea. No explanation for this phenomenon is available.)

There is virtually no difference between the masks as regards AHI. The Activa is slightly superior for reducing apneas (AI and percent-time-in-apnea), while the UMFF is slightly superior for reducing hypopneas (HI), though none of these differences appear to be statistically significant.

Still, viewing a scatter plot of the efficacy parameters is informative:



On four of the twenty nights (20%), Activa values fall within the arbitrary target box (AHI ≤ 10, AI ≤ 1), while the UMFF never achieves this goal. While the values for each mask overlap, and despite the almost identical AHI averages, visually the “clusters” for each mask appear distinct. This is due to the higher AI values for the UMFF. The near-parallelism of the two linear regression lines emphasize the similarities between the two masks; their separation emphasizes the differences. The temptation to remove the “outlier” values for the Activa had to be strongly resisted. Were the single entry for the Activa WITH mouth taping to be included on this chart (AHI=14.1, AI=8.8), the AI axis would need to be extended.

I would also call attention to results for two of the three “process” indices (parameters which drive the Respironics AutoPAP algorithm but which are not conventional “therapeutic outcome” factors):



Non-Responsive Apneas appear only sporadically in isolated occurrences with both masks, and no association with higher pressure is noted. The likelihood that these reflect true central apneas, and the relevance of the reported occurrences, are highly doubtful. Whether the slightly higher frequency during the night using the Activa WITH mouth taping is of any relevance is also doubtful.

Similarly, the small difference in Flow Limitation is of no clinical nor statistical relevance.

Vibratory Snore, however, is massively higher for the UMFF than for the Activa, and the difference is statistically significant. That this factor should differ so markedly with the selection of mask is a VERY surprising finding. Exactly what the Respironics sensors are detecting and calling snore is unclear; whether the mask itself is vibrating at critical frequencies, or whether the mask is affecting underlying physiology (as mouth-taping arguably does) and “creating” actual snoring, cannot be resolved here.

As mentioned previously, it appears that the occurrence of whatever the machine is reporting as “Vibratory Snore” may be producing “runaway” pressure increases. An example follows:



Pressure Profiles

The Respironics AutoPAP algorithm runs at higher pressures with the UMFF than it does with the Activa. Although these differences may not achieve statistical significance, they are arguably clinically relevant. The 90th centile “Titration pressure” for the UMFF mask is almost 3.5 cmH2O higher than that for the Activa mask, a surprisingly large value. In addition, the maximum pressure encountered with the Activa was 16cm, while the maximum pressure with the UMFF was 20 (although only for one minute on one night). Further, with the UMFF, the maximum pressure reached 18cm or higher on five of the twenty nights (25%), while with the Activa the maximum pressure exceeded 13cm on only one night. In addition, there is evidence of “runaway” pressure increases with the UMFF, which will be further reviewed in the discussion of “Vibratory Snore” below. Whether such high-pressure interventions were “necessary,” and/or whether setting a lower “maximum pressure” when using the UMFF mask might be desirable, are unclear.

As suggested by the pressure statistics presented above, the pressure profiles differ for the two masks. The amount of time spent at each pressure range for each of the masks appears below:



As is demonstrated above, with the Activa the machine spends 90% of its time at pressures of 9 and below. With the UMFF, the machine spends 29% of its time at pressures of 10 and higher.

The pressures at which apneas are reported can be similarly displayed:



While the modal pressure for apneas is at 8 cmH2O for both masks, with the Activa 35% of apneas occur above that pressure, while with the UMFF 56% of apneas occur above that pressure. With the Activa the highest pressure at which apnea occurred was 15cm (2 apneas on one night), while with the UMFF the highest pressure at apnea was 19cm (only once; there were also 2 apneas on one night at 18cm).

The correspondence of these two attributes, the distribution of pressure over time, and the distribution of apnea at various pressures, can also be displayed for each mask:





With both masks, there is a close correspondence of the two curves. The significance or desirability of this situation is unclear: does it indicate that the machine is successfully working toward its intended goal, or not? Are pressures rising to follow the apneas, or are apneas following rising pressures?

This question is of particular interest because the one night use of the Activa WITH mouth taping produces the following graph, with unknown statistical reliability, but with a clearly different pattern than displayed above for each curve, and for the relationship between the two curves:



Leak Information

Leak patterns for the masks were examined by doing with Excel spreadsheets what ResMed machines do within their firmware: the pressure-specific venting rates (as provided in manufacturer data) are subtracted from each leak datum based on the pressure at each time point. Respironics machine literature does not quantify its leak specifications, so the categories I have adopted are based on ResMed specifications. With this adjustment and classification, the results are as follows:



It is seen that the Activa mask, with mouth leaks eliminated by taping, exhibits very few higher-level leaks. Differences between the Activa without tape, and the UMFF, are minimal, and leaks are not excessive. This similarity is somewhat surprising, since mouth leaks are expected with the Activa without tape, while only skin-seal leaks are expected with the UMFF. At no time was any “Large Leak” reported by the Respironics machine. Issues regarding the methodology used in this analysis will be revisited below.

Conclusion

Different performance of the two masks (Activa and UMFF) have been described. Under identical conditions with a Respironics AutoPAP, there are differences in efficacy, pressure patterns and leak characteristics. Overall AHI is virtually identical for both masks, though control of apneas is better with the Activa. The Activa also runs at lower overall pressures than does the UMFF. Leak rates are acceptable for both masks.

Most surprising is the detection of significant occurrence of Vibratory Snore with the UMFF, but virtually none with the Activa. Vibratory Snore, as a driver of the Respironics AutoPAP algorithm, appears to contribute to the higher pressures encountered with the UMFF mask, and possibly contributes to “runaway” pressure increases.

Velbor

[Velbor]

Respironics in comparison with ResMed

We proceed now to the more delicate task of comparing the above information from the Respironics REMstar M-Series Auto, to similar information regarding the ResMed S8 AutoSet Vantage, which has previously been posted.

We cannot compare the “efficacy numbers” for different machines (AHI, AI, HI, % time in apnea) and utilize them to make value judgments as to which machine “works better.” The numbers mean different things; the words behind them describe different conditions and situations.
(See: viewtopic/t40350/viewtopic.php?f=1&t=40350&p=353046)
Value judgments based on such comparisons are faulty in their assumptions and reflect misunderstanding of underlying realities.

However, with care, valid comparisons CAN be made. We can DESCRIBE, with value-neutral language, the “numbers” which different machines produce, within statistical limitations, under the same conditions. We can directly compare the PRESSURE patterns produced by machines with different algorithms (since the meaning of cmH20 does not vary between manufacturers), though we need to be extremely careful in assessing whether one is “better” than another.

The Respironics pressure and efficacy parameters presented above are repeated here, this time with corresponding data from the ResMed machine used at the same pressure settings and with the same masks:



It needs to be reinforced that we are looking at different data from each machine. This is clarified for the pressures, where we can emphasize the distinction by employing different column titles, but it must be constantly remembered that definitions of “apnea” and “hypopnea” differ between the machines.

Pressure Profiles

When looking at the “high-end” (9x centile) pressures, the difference between the two machines is smaller for the UMFF (0.8cm, with the Respironics higher), but more pronounced for the Activa (1.7cm, with the ResMed higher). (The ResMed would be expected to be higher, due to use of the higher centile figure.) The practical implication of these differences is that each machine would “recommend” a different “titration pressure” for fixed-pressure CPAP: for the UMFF, 12.0cm by the ResMed and 12.5 – 13.0cm by the Respironics, and for the Activa 9.5cm by the Respironics as compared to 11.0cm with the ResMed, a considerable difference.

When looking at the “central” (mean or median) operating pressures, difference between the machines is also considerable for both the UMFF (1.2cm) and for the Activa (2.0 cm), with the ResMed being consistently higher.

Graphs of the pressure profiles for each mask, when used with each machine, can be compared:



The shapes of the curves above are notably different. When using the UMFF mask, the Respironics machine algorithm runs at lower pressures more of the time than does the ResMed machine algorithm. Nonetheless, the Respironics algorithm has utilized higher pressures (to 20.0 cm) than have been employed by ResMed (to 16.8cm) with this mask.



To an even greater extent with the Activa than with the UMFF, the Respironics machine algorithm runs at considerably lower pressures much more of the time than does the ResMed machine algorithm. Both machines have relatively sharp cut-off’s at the12 or 13 cm ranges for the Activa, as opposed to the much more gradual drop-off at higher pressures with the UMFF.

It may also be useful to observe that on typical nightly reports, regardless of mask, the ResMed algorithm tends to “jump” fairly rapidly upon my falling asleep, from its minimum pressure of 6.0cm, to about 10.0cm, and to remain about that level with little variation:



This particular ResMed night also demonstrates the typical cyclic pattern associated with Activa “lip-flutter” or “puffing” exhalation leaks.

The Respironics algorithm, on the other hand, tends typically to keep pressure near, and frequently return to, its minimum value of 6.0cm:



Efficacy Statistics

We need to be far more careful comparing the efficacy numbers, since they measure different things. ResMed tends to produce “better numbers” for apneas (numerical differences are statistically significant), while Respironics tends to have lower numbers for hypopneas (differences for the UMFF appear to be statistically significant, but not for the Activa).



We can conclude NOTHING about relative efficacy between the two machines from this data. All that can be said is that this data does support the “conventional wisdom” which often holds that ResMed hypopnea indices “tend to run higher” than do Respironics numbers.

The percent-time-in-apnea requires even more careful consideration. These numbers for either mask are significantly higher for the Respironics machine, though this is not a datum which the Respironics software provides directly. Again, it must be remembered that the definitions of apnea different. [While ResMed statistics are truncated rather than rounded after the first decimal digit, this would not account for the extent of the difference between the machines. Other recent posts have also called into question the accuracy of Respironics time-in-apnea information.]

Leak Information



It appears that the methodology distinguishing between “zero leak” and “mild leak” renders differentiation between these two categories as meaningless; for both masks and for both machines these two states occupy well over 85% of the night, and there is no reason for the Respironics machine, running at lower average pressures, to generate a higher leak profile. At the higher leak levels, it is not surprising that for each machine the Activa has higher leak figures than does the UMFF: it is known that I experience “lip flutter” or “puffing” exhalation leak with the Activa. It is surprising that the Respironics numbers for the Activa are higher than those for the ResMed, since the Respironics is operating at lower pressures, which should reduce the exhalation mouth leak. There is no reason for the UMFF higher-level leak to be higher with the Respironics machine; the procedural methodology I have used for “normalizing” and categorizing leak is clearly imperfect.

Conclusion

In summary, real distinctions between the machines, operated under almost identical conditions, have been noted. However, there is no objective criterion for assigning relative value labels (“better” or “worse”) to these differences at this time.

Always remembering that while my anatomy and physiology are not identical to yours, and my data may have more or less relevance to your particular situation, it is highly likely that my OSA is similar in some respects to that of someone else, and my findings should not be dismissed until you’ve proven that you are not that someone.

This effort will continue, several weeks from now, with use of the Respironics machine in fixed-pressure CPAP mode at 11.0 cmH2O, the same pressure utilized with the ResMed machine. This will remove algorithm-related differences, and hopefully leave us able to look at the effects of differences in the definitions of “apnea” and “hypopnea”.

Velbor

[klockemy]

Wow! Now that is a very analytical look at managing your therapy. I will have to print this out and read it at home a couple times with a glass of brandy (or scotch) to fully understand it. Excellent job, Velbor!

[cinco777]

You have provided an awesome level of common sense and statistical analyses applied to many different CPAP therapy variables: machines, masks, pressure, algorithms, etc. You've given me more food for thought re the analyses that I am doing on my sleep therapy "numbers" and equipment usage/settings. You've set a high bar for others to follow. Thanks.

[Velbor]

11cm H2O: Respironics M Series: Activa, UMFF and F&P 431 masks

Following completion of the previously posted data collection with the Respironics REMstar M-Series Auto set at 6.0 – 20.0 cmH2O, I began collecting data with the machine set in constant-pressure CPAP mode, at 11.0cm. This pressure choice was consistent with an average of the Auto data (90th centile pressures for the Activa mask at 9.5cm, and for the UMFF mask at 13.0cm), but selected primarily to maintain consistency with previously posted data using the ResMed S-8 AutoSet Vantage machine at 11.0cm.

In addition to the ResMed Mirage Activa nasal mask and the ResMed UltraMirage Full Face mask which have been included in past reports, during this testing period I acquired a Fisher & Paykel FlexiFit 431 Full Face Mask, which is now newly introduced into the comparative data. In addition, three one-night-stands under dubious circumstances were also undertaken and here reported upon.

Efficacy Statistics

The following table (which repeats the previously presented 6-20cm data – grayed out – for ease of comparison) presents the average of nightly values, the range of nightly values, and the standard deviation. Percent time in apnea is calculated nightly from the EncoreViewer report “Therapy Data Summary – All Data” (last page) “Average Time in Apnea Per Day” divided by “Total Blower Time”. Thus, three sets of old data, and six sets of new data (presented in order of increasing efficacy) are included:



With curiosity again triumphing over wisdom, and I spent a single night at a sub-therapeutic 4.0 cmH2O (the lowest pressure possible with the machine still recording data). This provides, as close as can be simulated at home, for reference purposes, a picture of what my breathing would be like WITHOUT CPAP. The numbers are, as would be expected, terrible, though it is curious that the AI was equal to that obtained with CPAP therapy which SHOULD have been effective.

While I already knew that mouth-taping has the paradoxical effect of WORSENING my breathing (although improving leak statistics), I found it hard to talk myself into using taping with a full-face mask. The presence of tape should be irrelevant, and I fully expected the experiment to be a waste of time. I was quite startled to find that taping, with the UMFF mask, resulted in markedly worsened indices and percent-time-in-apnea. In fact, the AI was fully as bad as my experiment at 4.0cm, and the %-time-in-apnea was even worse, than the “no-CPAP” simulation. With only a single night under each of these conditions, the strength of the statistics is indeterminate. Nonetheless, the numbers are as they are, inexplicable as they may be.

The single night using the Activa nasal mask, with mouth taping, at least was not surprising, producing results comparable to previously presented data. As in the past, it remains a puzzle why this paradoxical worsening of breathing should appear. And again, despite my inability to offer any etiologic explanation, and the indeterminate statistical strength of these figures (though considerable statistical power for this finding has previously been presented), the numbers were not unexpected.

Also not unexpected is the superiority of the Activa over the UMFF mask. Although the index numbers are consistently better with the Activa, their statistical power is borderline at best. (Although %-time-in-apnea exhibits a stronger statistical difference, I have expressed grave concerns regarding the validity of this datum as derived from Respironics machine/software data.)

To a degree, the Activa superiority under these fixed-pressure conditions is not surprising, since the CPAP setting of 11.0 cmH20 is 1.7cm higher than the 90th centile Auto (6-12) pressure for the Activa (9.3cm), while it is 1.7cm lower than the 90th centile Auto pressure for the UMFF (12.7cm). Thus, if the previous 6-12cm Auto experiment were viewed as a “titration” session, at 11.0cm I have been using is “too high” a pressure for the Activa and “too low” a pressure for the UMFF. Still, it is puzzling to me that for BOTH masks, the fixed-pressure CPAP numbers are better than those provided by the AutoPAP algorithm. Acknowledging (at least for the present) that the Auto algorithm might have performed better with a higher minimum pressure, it is still puzzling why the Auto algorithm DID NOT PERFORM BETTER. The Auto algorithm is “supposed” to actively adjust pressure to achieve “optimal” results. Clearly, and particularly with the Activa, this goal was not achieved.

My recent introduction to the Fisher & Paykel FlexiFit 431 Full Face Mask, although with only half the number of nights of data collection, shows it to be relatively comparable to (even if the numbers show performance slightly worse than) the UMFF mask. This supports my assertion that different masks display different efficacy characteristics, and also supports the hypothesis that the TYPE of mask (nasal vs. FF) may be the primary factor in producing different therapeutic results.

These findings once again reinforce my previously stated conclusion that a single “titration pressure” from a one-night sleep study (given the high night-to-night variability exhibited), without reference to a specific interface device (given the significant mask-related performance differences), is overly simplistic. While it may be unwieldy to write different CPAP pressure prescriptions for each interface device which might be used, and while it may also arguably be an overreaction to change pressures each night if different interfaces are frequently alternated, it nonetheless does seem reasonable, if a single interface is used regularly by an individual, to determine the optimal pressure(s) for use with that specific device.

As I have suggested in the past, viewing a scatter-plot of the efficacy parameters is informative:



Items which may be noted include:

 All 20 of the nights using the Activa mask fall within the arbitrary target box (AHI ≤ 10, AI ≤ 1). Only 3 of the nights using the UMFF mask (15%) fall within the target box. Only 1 of the nights (10%) using the FP431 mask falls within the target box.
 The slopes of the linear regression lines for all three of the masks are remarkably similar. The slope of the UMFF mask is about 1.6; that is, when the difference in AHI between two nights is 1.6, this variability will be associated with a change of about 1.0 in AI. Correspondingly, the same AHI variability would be associated with a change of about 0.6 in HI. In other words, night-to-night variability, independent of mask, has a 5:3 AI:HI ratio, or alternatively, my night-to-night variability is about 62.5% attributable to AI and 37.5% attributable to HI.
 The range of night-to-night variability should be appreciated. The notion that meaningful data upon which a user might make supposedly “informed” changes in settings, can be obtained in 3 nights, or even 7 nights, is sheer folly. I find that only as I approach 20 nights do measures of central tendency “settle down,” and even then standard deviations are higher than would be hoped.
 While my “experiment” with mouth-taping with the Activa mask would just barely be included at the far left of the chart, inclusion of the results of mouth-taping with the UMFF, and the results of 4.0cm pressure, would require that the chart be considerably extended.

For ease of comparison, I once again display the similar (less effective) data associated 6-12cm AutoPAP therapy:



I would again also call attention to average (range & standard deviation) results for the “process” indices (these parameters drive the Respironics AutoPAP algorithm but are not conventional “therapeutic outcome” factors) reported by the Encore software. At fixed pressure, only Vibratory Snore is relevant:



As with the Auto settings, so also with fixed pressure, Vibratory Snore was present every night, often significantly, with the UMFF, while it is virtually never reported (an index of 0.1 on two nights, otherwise zero) with Activa. The F&P 431 mask exhibited less VS than did the UMFF, though only one night of the ten had no such events. Once again, that this factor should differ so markedly with the selection of mask is a VERY surprising finding. Exactly what the Respironics sensors are detecting and calling snore is unclear; whether the mask itself is vibrating at critical frequencies, or whether the mask is affecting underlying physiology (as mouth-taping arguably does) and “creating” actual snoring, cannot be resolved with data available to me. Further, while my own computation of the respiratory indices from the raw data generally agrees well with the index values provided by the Encore software, for Vibratory Snore the software puzzlingly tends to give significantly higher values than are supported by the raw data.

Leak Performance

Leak patterns for the masks were examined by doing with Excel spreadsheets what ResMed machines do within their firmware: the pressure-specific venting rates (as provided in manufacturer data) are subtracted from each leak datum based on the pressure at each time point. Respironics machine literature does not quantify its leak specifications, so the categories I have adopted are based on ResMed specifications. With this adjustment and classification, the results are as follows:



The best leak performance was exhibited by the Fisher & Paykel 431 Full Face mask. It required considerably tighter strapping to prevent leaks (particularly around the eyes) than my other two masks, and left more (larger, redder, longer lasting) redness on the nose bridge. Nonetheless, it was surprisingly comfortable, and it was easy to grow accustomed to its under-the-chin design. Also very effective with respect to leak was the UMFF.

As expected, the Activa spends more time at higher leak rates than does the UMFF, due to the exhalation “lip flutter” mouth leaks which I am known to exhibit. In addition, as expected, the Activa has higher leak rates when used at constant 11.0cm than it did when used in Auto mode at 6-20, when the mean pressure was only 7.6cm and even the 90th centile pressure averaged only 9.3cm. Leak with the UMFF in fact DECREASED at the higher fixed pressure.

Unfortunately, while a 5% “significant leak” rate for the Activa could arguably be borderline acceptable, it is difficult to defend an almost 7% significant leak frequency. This is further worsened by the fact that 12 of the 20 nights (60%) individually had significant leak frequencies over 5%; five of the nights (25%) displayed significant leak over 10% of the time. Although there were NO “Large Leak” indications from the Respironics software for the Activa (or for any of the masks), and although there is no evidence of the leaks resulting in machine problems or in sleep disruption, this remains an area of concern.

Since, as documented by my worsened efficacy data, mouth taping is NOT an acceptable option for me, and since the Activa is documented to be the BEST performing mask in terms of efficacy, a balance needs to be found between optimal efficacy pressure, and an acceptable leak (which increases with higher pressure).

Conclusions

At fixed CPAP pressure of 11.0cm, performance of the Activa mask was superior to that of the UMFF and the F&P 431 in terms of efficacy, demonstrating that masks are not simply passive “comfort” components, but affect the quality of therapy.

In addition, therapy with both the Activa and the UMFF masks was superior at 11.0cm fixed pressure than it was with the Respironics Auto set from 6 – 20cm. While this is not overly surprising, since the average Auto 90th centile pressure was only 9.3cm for the Activa, it remains puzzling that performance in Auto mode was not better.

While the UMFF exhibited LESS leak at higher pressure, leak with the Activa (due to exhalation mouth leak) was significantly higher.

As previously demonstrated, and herein reinforced, while mouth taping very effectively eliminates leak from that source, it can have unexpected (and unexplained) deleterious effects on respiratory event indices.

As was noted in Auto mode, Vibratory Snore continues to be encountered with the UMFF in fixed pressure mode (and with the F&P 431 as well), but is virtually absent with the Activa. The appropriateness of the algorithmic application of Vibratory Snore in AutoPAP mode, if it is mask (or mask-type) dependent, is unclear.

I am unable to discern consistent subjective differences in how I feel, either between the AutoPAP and the CPAP therapies, or between the masks.

[Velbor]

Respironics in Comparison with ResMed: fixed pressure CPAP

We proceed now to comparing the above information at fixed CPAP pressure of 11.0cm from the Respironics REMstar M-Series Auto, to similar information regarding the ResMed S8 AutoSet Vantage, which has previously been posted.

The cautions about comparison between machines from manufacturers which I expressed in a previous post remain valid. The current situation, however, is a bit simpler, since it describes data from fixed-pressure CPAP use. The meaning of cmH2O is not manufacturer-dependent, and we are not now dealing with differences in algorithmic function. There still exist manufacturer differences in the definitions of “apnea” and “hypopnea”. While it remains inappropriate to make “better / worse” evaluations, it is precisely the differences DUE TO the difference in definition of events which we can not examine.

The Respironics efficacy parameters presented above are repeated here, this time with previously-reported corresponding data from the ResMed machine (grayed-out) used at the same 11.0cm pressure setting and with the same masks:



The AHI and HI numbers reported by the ResMed machine are consistently higher than those reported by the Respironics machine for the same mask, under identical conditions; the reverse is true for the AI and percent-time-in-apnea numbers.

(Well, perhaps not identical conditions. Between the period from late-January through mid-March, 2009, when the ResMed data was collected, and the current Respironics data collection period from mid-June through early August, 2009, my average weight increased by 1.7 pounds, which is less than a 1% increase. Other unrecognized differences of possible significance may also exist.)

It is critical to understand what these numbers do and do not represent. They do not suggest that I have had FEWER apneas and MORE hypopneas on the ResMed machine than I have had on the Respironics machine. The pressures are the same; the therapy is the same; all other relevant variables are presumably the same. What is different is what ResMed DEFINES and REPORTS as an apnea: there are fewer of these than what Respironics DEFINES and REPORTS as apneas. What is different is what ResMed DEFINES and REPORTS as a hypopnea: there are more of these than what Respironics DEFINES and REPORTS as hypopneas.

It must be noted that my data in some respects conforms to what is considered “common knowledge”: that ResMed machines will REPORT higher AHI and HI values (for the same person, under the same conditions) than will be REPORTED by Respironics machines. My data complicates this “understanding” by demonstrating that difference factors are also mask-dependent.

It has been proposed by RestedGal, (particularly for use in conjunction with the notion that an “ideal” AHI “should be” 5.0 or less) that ResMed users should “mentally” recalculate AHI by using only half their reported HI, if they are using the “5.0 AHI” standard. (This has raised the objection that this thought process inappropriately makes Respironics numbers a de-facto standard; I note this but will not comment further.) Utilizing this recommendation with my data, the “comparable” results would be:

UMFF: ResMed AI 0.8 + HI 8.6 / 2 = “comparable” AHI = 5.1 rather than 9.3
Actual Respironics AHI = 3.3
Activa: ResMed AI 0.2 + HI 5.0 / 2 = “comparable” AHI = 2.7 rather than 5.2
Actual Respironics AHI = 1.2

Clearly, this approximation cannot be verified from my data; it suffers from not accounting for the LOWER AI reported by Resmed than is reported by Respironics. (On the other hand, note that my arbitrary graphic “target box” was set at a maximum AHI of 10.0 for the ResMed to obtain “reasonable” working data; the AHI could well have been limited to 5.0 for the Respironics data.)

My data CAN be used to provide a bi-factoral comparability methodology; unfortunately the solution of the relevant equations yields numbers which have no apparent “physical” meaning. To serve as a comparison for BOTH masks, each ResMed AI would have to be multiplied by 2.71 and each ResMed HI would have to be multiplied by 0.13, to produce the “comparable” Respironics AHI. In one sense, this conforms to the “common wisdom” that ResMed’s HI must be “lowered” to produce a Respironics-comparable value. Still, it gives considerably greater weight than might have been expected to the ResMed AI.



It is critical to once again emphasize that use of such factors for ResMed data is an attempt to arrive at a “Respironics-comparable” (that is, a number capable of being compared), NOT a “Respironics-equivalent” AHI number. (Understand also that this current data reflects only fixed-pressure CPAP use: the notion of number-comparability is stretched beyond meaningfulness when AutoPAP algorithmic differences are also present.)

In this regard it is also interesting to note the differences in the relative “percent contribution” of AI to the combined AHI, for the different machines and for the different masks, all at 11.0 cmH2O (these figures come from the original data, not from the averages presented above):



For Resmed data, the AI plays a relatively small role in the total AHI. For Respironics data, the AI plays a much larger role in the total AHI, though it is a minority role as compared to the HI for the Activa mask, while AI has a majority role with respect to the UMFF (and the F&P 431) mask. Who would have guessed? Who knows what it means? Or whether it is of any relevance?

ResMed consistently reports lower values for “percent-time-in-apnea” than does Respironics. This is to be expected, given the generally lower AI values reported by ResMed. While the accuracy of nightly data for this parameter as reported by Respironics has been questioned,

see: viewtopic.php?f=1&t=42182

the similarity for the two machines of the ratio between AI and %-time-in-apnea for the UMFF mask (ResMed: 2.67, Respironics: 2.70) is striking. (Similar figures for the Activa mask are calculable only for Respironics: 3.85).

Finally, it should be noted that both my Respironics and my ResMed data agree that the Activa mask provides consistently superior efficacy results as compared to the UMFF mask.

The risk of excessive repetition is far less than the risk of misunderstanding. I therefore repeat: We can conclude NOTHING about relative efficacy between the two machines from this data. All that can be said is that this data does support the “conventional wisdom” which often holds that ResMed hypopnea indices “tend to run higher” than do Respironics hypopnea indices.

Leak Information

As previously mentioned, leak figures for the Respironics machine have been adjusted to eliminate the expected venting, making numbers comparable to those provided by the Resmed machine.



For each machine, reported leak profile is (not unexpectedly) higher for the Activa (due to my exhalation lip flutter leak) than it is for the UMFF. For each mask, reported leak profile is higher for the Respironics machine than it is for the ResMed machine.

At fixed pressure, there is less difference between the leak profiles of the two masks than had been observed in the AutoPAP mode as previously reported. Still, there is no obvious reason for the Respironics machine to report higher leaks at the same pressure and under identical conditions.

Assuming that both machines are equally “accurate” in their measurement and data management, possible reasons for the discrepancies in leak profiles may include such factors as:

 ResMed provides raw leak data in 60-second windows, while Respironics provides leak data in 30-second windows
 ResMed provides raw leak data to a precision of 0.02 L/s (=1.2 L/m), while Respironics provides leak data to a precision of 7 L/m (=0.12 L/s)
 ResMed provides leak data already modified by the “expected” normal venting of the mask “category” identified to the machine; “exactly” how it manages the data is not known to me.
 The procedural methodology I have used for “normalizing” Respironics leak may be imperfect.
 Categorization of the lower-precision Respironics data may be affected by the use of cut-off levels which are based on ResMed data presentations.

The following graphs display the typical features of my “lip-flutter exhalation leak” with the Activa mask, on each machine. The cyclic nature of the leak, with discrete “mountains” presumably reflecting underlying deep-stage or REM sleep, is quite distinctive. From ResMed:



And from Respironics:



Conclusion

In summary, real distinctions between the machines, operated under almost identical conditions, have been noted. However, there is no objective criterion for assigning relative value labels (“better” or “worse”) to these differences at this time.

The current data has focused on fixed-pressure CPAP, in which situation fundamental differences between the two machines is largely limited to differences in definition of apnea and hypopnea. Differences in the AutoPAP algorithms would add another massive level of complexity. (Differences in the Flex/EPR comfort features are not addressed here at all.)

This effort will continue, over the coming weeks, with use of the Respironics machine in AutoPAP mode at 10.0 – 20.0 cmH2O, the same pressure utilized with the ResMed machine and previously published on this board. This will hopefully address issues related to the “conventional wisdom” that my previous 6.0 – 20.0 AutoPAP data used “too low” a minimum pressure. For the ResMed, differences resulting from changes in the minimum pressure were … minimal. We shall see, in several months, how the Respironics performs in this regard.

It must finally be repeated that this is MY data. The details of the pathophysiology underlying my OSA, and underlying my response to treatment, may be quite different from yours. I am doubtless in some respects a “special case,” but it is unlikely that I am a unique case. My data, and the conclusions which emanate from it, should neither be unquestioningly adopted, nor arbitrarily discarded, as relevant to any other particular person’s situation. To assess its value to YOU, you must compare it to YOUR data.

[WindFlyer]

You've done a very thorough job, Velbor, with all the experimenting that you've done and posted here. I wish DME's / RRT's would think more like this rather than seeming to be more about selling equipment. Most of the trade journals that are online seem focused on DME products, not cross-sectional or longitudinal studies on patients to determine what works best.