Deep Stops Increases DCS

[igor p]

Hi all,

here is a comment regarding the discussion here Bruce Wienke asked me to post. I can add I agree with him.

" Long time ago, I stopped worrying too much about the
ins and outs of deep stops vs shallow stops as ranted and raved
on blogs. The proof is in the pudding -- thousands of divers
employ deep stops and have used them safely without anything
but nominal DCS in the process for many years now. No probs.
No matter what model (USN, ZHL, VPM, RGBM) somebody
will always get hit, but the concern is how many and how often.
To date that number is very small by DCS incidences. As C&C
Dive Team Ldr, that is the bottom line for our operations. And
me personally.
You can argue deep stops vs shallow stops til you are
blue in the face. They both work, but for different reasons
is the simple fact. Here at LANL, we have dived both without
mishaps. Deep stops are faster and get us out of the water quicker
so we use.them. Period.
Just a couple of thoughts to pass on:
1 -- deep stop meters, tables and software have now been
around for 10s of years without reported problems with
DCS; and judging from the licenses we issue will enjoy
even more usage;
2 -- data is data no matter what generates it and for whatever
reason and LANL computer downloaded data is headed to
DAN as I have time to write translation software for 3000+
profiles;
3-- both VPM and RGBM work well and have no reported
DCS spikes (unless misused) across meters, tables and
software renderings;
4-- Balestra of DAN did a study of DCS rates in ZHL and RGBM
computers and found DCS incidence rates almost exactly
equal and small;
5 -- shallow stops vs deep stops are now for us mostly "religion"
questions;
6 -- the NEDU experiment suggested to us as operational divers
that if you don't do shallow stops and/or deep stops carefully
you can get hurt. Hats off though to all involved in that experiment
because it showed us what not to do. That is important information
but doesn't discredit deep nor shallow stops when looking at the
track record in the field -- where it counts. Both can be done safely
and that's good. End of story for us, training agencies, 1000s of
tec and rec divers, and growing numbers of same.

When I get to my office, will send you published versions of papers on the above.

Thanks,
BW"

 

[leadduck]

I think RSS only contributes to risk if positive, else 0. Once SS above threshold is below zero no further risk is added.

OK, that's fine.

If the dives are both to the same depth and vary only in bottom time, then ISS will be larger for the 200min RT. So the simply ISS would not be misleading. But we all know that dive has more deco stress.

I meant two dives with equal depth and bottom time, but comparing one fast deco profile with early exit, the other one slowly ascending. I see no reason why not to use ISS to compare their DCS risk, just because their runtime is different.

I think it's just safe to restrict "raw ISS" to comparisons of dives that vary only by distribution of stop time. If used that way, then it can be a reasonable measure of decompression stress. If you use raw ISS beyond that I think you could get false signals, and the more you diverge from comparing similar dives the more chance that will occur.

I think the reason why studies compare profiles of equal runtime is not that ISS wouldn't work when comparing profiles of different runtime, but that they care about efficiency. The profile with 200min runtime may have less ISS and less PDCS than the one with 100min runtime, but it keeps you twice as long in the water. So in order to compare two models regarding deco efficiency, you set their conservatism parameters so that they create schedules of equal runtime for a given depth+bottomtime, then compare PDCS. So I think setting the critical radius in VPM-B within Erik Baker's recommended range so that it matches runtime of NEDU profiles is the right thing to do when comparing VPM-B and NEDU in the way how they distribute supersaturation across compartments.

Alternatively you could plot the ISS of each compartment relative to the fastest one; this would allow to compare VPM-B+1, +2, +3, +4, GF70/70, GF40/70, GF30/85, NEDU-A1, NEDU-A2, ... in one chart although their runtimes are different. I expect you will see that all levels of VPM-B+x distribute the supersaturation similar to NEDU-A2. The higher conservatism levels of VPM-B buy safety by extending the runtime of an otherwise sub-optimal deco curve shape, hence inefficient.

 

[leadduck]

When I first read the NEDU article, it struck me as odd that they used a one-sided Fisher's exact test, I probably would have chosen a two-sided test. I ran a 2x2 contingency table and got a p of 0.087 using a two-sided test and a p of 0.049 using a one-sided test. The NEDU study states a p of 0.47. This is not rocket science, the results are not significant using a two-tailed test and are right on the border using a one-tailed test. Statistical significance was nearly lost with exclusion of one episode of DCS in the deep stops group.

They write in the report why they use a one-sided test, see second paragraph on p.4.

The reason why the p-Value is right on the border is simply that they stopped the experiment as soon as the border was reached because then they were sure that A2 profile was significantly worse than A1 and didn't have to bend more test subjects. Had they continued with 200 more dives as originally planned, they might have reached a much smaller p-value.

 

[rossh]

For DCS risk, the area below the curves ("integral supersaturaion") is relevant to compare, not their maximum. And the x-axis better be time, not depth for that comparison. You can see in your own plots that the area below the NEDU-A2 profile is smaller than the one below the VPM-B+0 profile, which means that NEDU-A2 protects the fast tissues even more strongly than VPM-B+0, while ZHL-16C puts the most stress on them.

For the rest of your plots I think you miss the difference between compartments and show only the fast ones. Deep stops shift the supersaturation stress from the fast to the medium and slow compartments, so you're completely missing the point if you look only on a few faster compartments or just sum up all of them. UWSojourner's plot of integral supersaturation vs all compartment indices show this. There you can also see how similar NEDU-A2 and VPM prioritize fast vs slow compartments in terms of supersaturation, although their stop times and depth are different.



All models including VPM and RGBM, and GF addition to ZHL, demand scrutiny. These in particular, because
(1) Unlike many other models and tables, their parameters were not calibrated with controlled experiments to minimize DCS risk, but were just set so that their runtime is similar to older methods and the curves look somehow familiar. If you use such an uncalibrated model with no experimental link to DCS risk, then you might as well just guess a runtime, take a paint program and just draw a curve that you feel looks right, and dive it.
(2) It's conspicuous how these uncalibrated and untested models were and still are pushed commercially with dive planning software and licensing to dive computer vendors. They promote an unhealthy habit of deep stopping among technical divers.

I think you have missed some relevant points in there.

The plots are all scaled for time - you will notice three basic x axis plot lengths for the dive portion - all scaled properly, but time number are not shown - I can add em if you really need to see that.

The plots show the maximum (lead / controlling) cell at every point. Fast ones start at the beginning (left side), slower ones later at the surfacing end. I have already addressed how the A2 and real VPM-B are different in all the appropriate ways.

These are the actual data points that the decompression model uses - no room for dispute here.

This method of displaying stress and supersaturation, is independent of model. It can be used across many different planning types. These tools are inbuilt to MultiDeco - you can make all that yourself right now, and check it all you want. It was these discussions that made me add this feature, so anyone can do these.

Integral supersaturation, is the area between 0 (pAmb) and the peak supersaturation shown. If you add up the volume of that area (pinky bits), you have integral supersaturation. You can also visualize it pretty easy from this display format.

.Supersaturation is the industry accepted risk measure. David Doolette describes this in the papers introduction, and its exactly what is shown in these graphs.

Your assumption that different profiles make different types of off gas and supersaturation patterns is correct. BUT, how much does it really change things? You can see all that for yourself in the graphs, or make your own in MultiDeco.

.

 

[UWSojourner]

These are the actual data points that the decompression model uses - no room for dispute here.

Actually, I'll dispute that --- at least if you claim the plots represent supersaturation.

Integral supersaturation, is the area between 0 (pAmb) and the peak supersaturation shown. If you add up the volume of that area (pinky bits), you have integral supersaturation.

This is just simply NOT TRUE. There are 16 compartments in Buhlmann's tissue model all contributing (potentially) to decompression stress. You can't just look at the area under the most heavily supersaturated compartment and claim it's integral supersaturation.

Supersaturation is the industry accepted risk measure. David Doolette describes this in the papers introduction, and its exactly what is shown in these graphs.

I suppose this is progress. Because the chart below shows the total supersaturation exposure (integral supersaturation) for a 240ft CCR dive (see this post for more info). The question still remains: If "Supersaturation is the industry accepted risk measure", then what benefit is VPM-B+3 providing for 30% more decompression stress (i.e. supersaturation exposure) at the surface? (and about 20% overall?)

Your assumption that different profiles make different types of off gas and supersaturation patterns is correct. BUT, how much does it really change things?

See this post. The distribution of stop time by VPM-B+3 vs GF60/75 caused VPM to display 30% more decompression stress when the diver surfaces. That's how much.

 

[EFX]

After reading Simon Mitchell's long post summarizing his key points regarding the NEDU study I was struck by the following comment: "Supersaturation is what drives bubble formation, and bubbles are the vectors of primary injury in DCS. More supersaturation overall is therefore likely to be bad." I thought, of course supersaturation (SS) drives bubble formation, it has to if there is going to be off-gassing. So what. Then I thought that high SS in and of itself may not be the sole problem but only one side of a two-sided approach. Bubbles below the critical size will tolerate higher levels of SS and bubbles that have grown above the critical size can cause DCS at lower levels of SS. And then I thought, "there seems to be an infatuation with SS but I don't remember reading anything recently regarding bubble mechanics and how that could be affected by the NEDU study."

So I did a search on bubble mechanics in the thread and found six entries, one of them from me (post #702). Kevrumbo posted a summary of Weinke's paper on deep stops in post #36 and after reading that I realized that bubble models, particularly VPM and RGBM, extend deep stops but shorten shallow stops allowing faster deco time than Haldane models. I was shocked and a little embarrassed by this since I assumed (incorrectly) that the bubble models would add time to shallow stops because deep stops were done. So now I understand why the NEDU A2 dive shallow stops were shortened -- it was an attempt to mimic the strategy of bubble models as well as maintaining total deco time a constant.

So, I need to rethink the NEDU study in light of bubble mechanics and see if I can come up with something that would support why BRW thinks the NEDU study's dive profiles were done wrong.

BTW UWSojourner, I went back to take a closer look at your heat maps and realized I had made some wrong assumptions regarding that to. Great work though. I really like the colored visual display format. I do have some questions though: In the profile line (profile CCR 240 ft 20 min 10/50 1.2) what does 10/50 and 1.2 mean?

 

[EFX]

BTW UWSojourner, I went back to take a closer look at your heat maps and realized I had made some wrong assumptions regarding that to. Great work though. I really like the colored visual display format. I do have some questions though: In the profile line (profile CCR 240 ft 20 min 10/50 1.2) what does 10/50 and 1.2 mean?

Wait a minute. The light bulb just went on. 10/50 is the gas mix (10% O2 and 50% helium) and 1.2 is the max ppO2. Is this correct?

 

[rossh]

.... t I realized that bubble models, particularly VPM and RGBM, extend deep stops but shorten shallow stops allowing faster deco time than Haldane models. I was shocked and a little embarrassed by this since I assumed (incorrectly) that the bubble models would add time to shallow stops because deep stops were done.

Big mistake there...

VPM uses Haldane gas tracking...... the ascent is dictated by Haldane / Schriener equations....

Bubble models DO make longer times for the deep stops...its the same formula that is in the ZHL and powers your GF as well. If it works for one, it works for the other too.

Also, VPM-B is LONGER than ZHL for most plans - yes longer, up to about 3 or 4 hours.


So what do think now?

.

 

[victorzamora]

1.2 would be the "Setpoint"....not the max ppO2, but the ppO2 throughout the dive.

 

[PfcAJ]

Big mistake there...

VPM uses Haldane gas tracking...... the ascent is dictated by Haldane / Schriener equations....

Bubble models DO make longer times for the deep stops...its the same formula that is in the ZHL and powers your GF as well. If it works for one, it works for the other too.

Also, VPM-B is LONGER than ZHL for most plans - yes longer, up to about 3 or 4 hours.


So what do think now?

.

longer than 100/100 maybe. Not for the GF values people are proposing here.