Deep Stops Increases DCS

[David Doolette]

I have been reading this thread with some interest, watching Ross reiterate, unchanged, points that were argued against 3 years ago in the Rebreatherworld deeps stops thread and subsequent threads, including, if memory serves me two years ago in the present thread. At the moderator’s suggestion, I will re-post some of my earlier material. Here is a slightly edited (split to be under the 10000 character limit) version of my first post to the Rebreatherworld thread:

The Navy Experimental Diving Unit (NEDU) “deep stops” study (Doolette DJ, Gerth WA, Gault KA. Redistribution of decompression stop time from shallow to deep stops increases incidence of decompression sickness in air decompression dives. Technical Report. Panama City (FL): Navy Experimental Diving Unit; 2011 Jul. Report No.: NEDU TR 11-06) was undertaken to determine if deep stops decompression schedules, such as those prescribed by bubble decompression models, are more efficient that the traditional shallow stops schedules prescribed by “Haldanian” models. More efficient in this context means a decompression schedule of the same or shorter total decompression time has less risk of decompression sickness (DCS) than an alternative schedule. Theoretical analysis at NEDU and by others had suggested this might be the case, and bubble models were being considered for calculating air decompression tables to replace the Standard Air Decompression Table that had been in the U. S. Navy Diving Manual since 1959, but this big change required a test.

To be clear about the purpose, methods, and outcome of the study, we need to be clear what is meant by decompression efficiency. The purpose of a decompression schedule is to reduce the risk of DCS to some acceptably low level. The cost of a low risk of DCS is time spent decompressing; efficiency relates to this cost/benefit trade off. In comparing two decompression schedules, if one could achieve the same target level of DCS risk with a shorter total decompression time than the other, the shorter schedule is more efficient.

With this definition in mind, one way to test if a deep stops schedule is more efficient than a shallow stops schedule would be to show that a deep stops schedule has the same (or less risk) than a longer shallow stops schedule. However, this is not a good scientific design because you are varying two things, stop depth distribution and total decompression time , and you will not know which was responsible if the result does not show deep stops to have lower risk. A better scientific design is to compare a deep stops schedule and a shallow stop schedule that have the same total decompression time and see which is riskier - only one thing is varied, the stop depth distribution, and any difference can be attributed to that alone. This latter is the method we used.

Remembering that the purpose of a decompression schedule is to reduce the risk of DCS, the definitive way to evaluate a schedule is to conduct many man-dives, following the schedule exactly, and count the incidence of DCS; the incidence of DCS is an estimate of the risk and the more man-dives the more confidence there is in this estimate. To compare two schedules, dive both, and count which results in more DCS. It is meaningless to compare the decompression efficiency of two schedules that are very unlikely to result in DCS – imagine conducting a thousand man-dives on each of two schedules with no DCS occurring, all you have learnt is both schedules are very low risk, and probably quite inefficient.

 

[rossh]

removed to allow continuity in surrounding posts

 

[David Doolette]

Part 2

So this is the experiment. In the wet pot of the NEDU Ocean Simulation Facility, where we can precisely control depth, time, water temperature, divers’ workload (all things that influence DCS risk), divers undertook two different profiles. Both profiles were to 170 fsw for 30 minutes during which time the divers exercised on cycle ergometers, followed by 174 minutes of decompression stops during which divers were at rest. The water temperature was 86 °F (30 °C) and dives wore only swimsuits and t-shirts and became cold during decompression. Divers were submerged and breathed surface supplied air throughout. More on all these conditions later. The only difference between the two profiles was the distribution of the total stop time among stop depths. The shallow stops schedule had stops of (fsw/minutes): 40/9; 30/20; 20/52; 10/93. The deep stops schedule had stops of (fsw/minutes): 70/12; 60/17; 50/15; 40/18; 30/23; 20/17; 10/72.

We planned to conduct 350 man-dives on each schedule, but to protect the diver-subjects from unnecessary risk, we also had several rules by which the experiment would stop early. We had stopping rules if both schedules had unexpectedly high or low risk, which were likely to result in severe DCS or an inconclusive result, respectively. We never came close to these. We were also to stop if, at an interim analysis, we saw a statitiaclly significant (at one-sided alpha=0.05) higher incidence of DCS on the deep stops schedule than the shallow stops schedule, and this is what happened. At approximately the mid-point of the experiment we had 10 DCS out of 198 man-dives on the deep stops schedule and 3 DCS out of 192 man-dives on the shallow stop schedules. Incidentally, we also measured venous gas emboli (VGE) and these were higher on the deep stops than the shallow stops schedule. As has been pointed out, a one-sided alpha=0.05 is, in statistical terms, only moderately strong evidence that the deep stop schedule was riskier than the shallow stops schedule. So why did we stop? Because it is very strong evidence that the deep stops schedule is not better and, because deep stops better was the only result of any consequence to the U. S. Navy (shallow stops are the status quo).

I want to talk about the schedules we tested in some detail, because these have been the source of a lot of confusion and misdirection in various forums. Clearly they do not look like technical diving schedules - they are not, they are deep air decompression schedules. In selecting the test pair of schedules, there were two principal criteria. First, they had to result in some DCS so there was something to compare. Second, they had to be long, so that they could have substantially different stop depth distribution, i.e. the deep stops schedule should require a substantial amount of time at deep stops, so any deep stops effect (good or bad) can manifest. There is no point in testing, for instance, two 90-minute decompression schedules where one has five or ten minutes of time spent at deeper stops – I would happily move five or ten minutes around in a 90-minute schedule and not expect it to make a any detectable change in my risk of DCS. Remember that the purpose of a decompression stop (deep or shallow): we stop to limit gas supersaturation and thereby limit bubble growth, and we stay to washout inert prior to moving to the next stop. The staying is important, the amount of gas washout that occurs in the course of one, two, or five minutes is relatively inconsequential. In other words, a one, two, or five minutes of deep stops probably does no harm, nor provides any benefit.

The final test pair of schedules was the result of hundreds of hours of analysis and the final design was peer-reviewed in a workshop attended by many people working in the field of decompression (acknowledged in NEDU TR 11-06). The shallow stops schedule was calculated using the VVal-18 Thalmann Algorithm. This algorithm was developed at NEDU for air and constant PO2-in-nitrogen rebreather diving and about 1500 man-dives were conducted during its development (NEDU TR 11-80, NEDU TR 1-84, NEDU TR 8-85). VVAl-18 Thalmann Algorithm is still very much in use, it runs in the U. S. Navy Dive Computers, desktop decompression software, and was used to calculate the MK 16 MOD 0 and MK 16 MOD 1 N2-O2 decompression tables in the U. S. Navy Diving Manual. For a 170 fsw / 30-minute bottom time air decompression dive VVal-18 requires the 174 minutes of decompression stops given above for the shallow stops schedule. Although this particular schedule was not tested during the development of Val-18, many deep, long air dives were, and lengthy air decompression was required. Many U. S. Navy dives are conducted with the diver working on the bottom and, because wet suits are often used, cold during decompression. This combination makes for a lot of required decompression because blood flow is increased, and therefore inert gas uptake is relatively fast, during the working bottom time, and blood flow is decreased, and therefore inert gas washout is relatively slow, when divers are at rest and cold during decompression. U. S. Navy decompression algorithms are designed to account for this worst case situation and tested under these conditions. Just to clarify some comments in this thread about the effects of cold, cold can increase the required decompression time, but cold does not cause DCS.

It has been suggested in this thread that 174 minutes decompression is 100 minutes too long for a 170 fsw / 30-minute dive – well, of course, you can do 74 minutes of decompression if you want, if you accept a high risk of DCS. In fact, 74 minutes is close to the time required in the new Air Decompression Table in the U. S. Navy Diving Manual Revision 6 (2008): 170 fsw / 30 minutes requires 88 minutes of air decompression stops, which has an estimated risk of DCS of about 6% (NEDU TR 09-05), but this exceptional exposure schedule is for emergency use only (this dive is required to be planned using the lower risk oxygen decompression schedule). Why didn’t we test this schedule or something like it? Two reasons: 1) 88 minutes is not long enough to allow meaningful redistribution of time to deep stops; and 2) even 174 minutes of decompression resulted in 13 cases of DCS.

Now the deep stops schedule. This was calculated using a model called BVM(3) (Gerth & Vann Undersea Hyperb Med 1997;24:275-292). BVM(3) is a Bubble Volume Model, and models in this class (Mike Gernhardt’s TBDM is another example) model the growth and dissolution of bubbles using the equations that describe exchange of gas between tissue and blood (a feature of most decompression models) and the equations that describe diffusion of gas between spherical bubbles and surrounding tissue (the characteristic of this class). BVM(3) output is the estimated risk of DCS for a dive profile, and this risk is a function of bubble volume and duration in each compartment. BVM(3) is used in conjunction with an exhaustive search algorithm to find the optimum decompression schedule (under the model). This can be done two ways. First, you can specify a total decompression stop time, and an exhaustive combinations of stop depths and times (that add to the total) are tested to find the schedule that gives the minimum estimated risk. Second, you can specify a target risk, and the first step is repeated with different total stop times, searching for the shortest schedule that just reaches the target risk. We used the first step, and specified 174 minutes total stop time (the VVal-18 total stop time) and had the model find the optimum distribution of that time, which resulted in the deep stops schedule specified above. Actually, we examined hundreds of candidate schedule pairs until we decided on the 170 fsw / 30-minute dive.

 

[David Doolette]

Part 3

To interpret our results, I have to describe some “Decompression 101” theory, so this may a bit basic for a lot of you, and for brevity I am going to confine the description to diving on a single gas (e.g. air diving, as in the experiment) although it is possible to extend to multiple gas. The purpose of a decompression stop is to limit bubble formation and allow washout of tissue inert gas. Deeper stops are generally controlled by faster exchanging (short half time) compartments and shallower stops by relatively slower exchanging (long half time) compartments. Bubbles form and grow only while tissue is supersaturated and shrink when tissue is undersaturated. In a supersaturated tissue, at a deeper stop (compared to a shallower stop) less bubbles form, they will grow less rapidly, they will dissolve more quickly, and in some circumstances inert gas washout can be faster. This is all good stuff and the motivation for deep stops. However, the NEDU results indicate that emphasizing these effects in fast tissues by doing “deep stops” is not as important as previously thought, because our shallow stops schedule, in which fast tissues had substantial supersaturation, resulted in very few cases of DCS. So why did the deep stops schedule result in more DCS? We looked at the supersaturation predicted in a range of half-time compartments. In fast compartments, the deep stops schedule resulted in less, and less prolonged supersaturation than the shallow stops schedule. However, in slow compartments, gas washed out slowly or continued to be taken up during deep stops, so that later in decompression, the deep stop schedule resulted in more, and more prolonged, supersaturation than the shallow stops schedule. The increase in supersaturation in the slow compartments was greater than the decrease in fast compartment. There is a principal, Occam’s Razor, that roughly means “the simplest answer is the preferred one”. The simplest answer here is that the greater supersaturation (and by extension greater bubble formation and growth) is responsible for the greater incidence of DCS on the deep stop schedule. In other words, the cost of doing the deep stops outweighed any benefit. And remember, “any benefit” was slim, because there was very few DCS in the shallow stops schedule.

So an important question is how relevant is this result to other deep stops schedules, or put another way, is there another deep stops schedule that would have given the reverse result. Accepting the explanation that greater supersaturation is the culprit, we modeled the gas supersaturation in a range of half-time compartments for half a million different schedules, each comprising 170 fsw / 30 minutes followed by 174 minutes of decompression stops, but with different combinations of stop depths (deepest stop 100 fsw) and times. At the level of granularity we chose (5-minute blocks of time was the shortest we moved) we looked at all reasonable ‘shapes’ of decompression schedule. As it turned out, the VVal-18 shallow stops schedule resulted in near the least combined (adding together the fast and slow compartments) supersaturation. Moving a small amount of time to deeper stops resulted in no improvement, and moving any substantial amount of time to deeper stops resulted in more combined supersaturation. This would suggest that there are some schedules with a little bit of time at deep stops that are no worse than the shallow stops schedule, but most deep stops schedules will be worse. Clearly, this theoretical analysis is not proof, but it is a compelling hypothesis, and I am very confident we would not have gotten the reverse result (deep stops better) if we had tested another schedule.

So what is the relevance of this to technical diving? For that I have to speculate a bit because I am moving away from the facts of the study. First let us deal with the issue of whether this applies to helium-based breathing mixtures. Probably. Blatteau and colleagues have done a small comparison of deep stops versus shallow stops open circuit trimix decompression profiles, using VGE as an endpoint and found more VGE with the deeps stops (Proceedings of the Decompression and the Deep Shop Workshop) and there is another, as yet unpublished, similar study from the Swedish Navy with similar results, although using algorithms used by technical divers. The more important issue is that technical divers do not do air decompression dives, they use oxygen-accelerated decompression. If decompression stops are conducted using a breathing mixture with a low inert gas fraction, then, of course, there is less gas uptake into the relatively slow compartments. The effect of this is to increase the depth at which stops become “bad” deep stops.

David Doolette

 

[rossh]

.... [ We have done this before ]....

Yes we have. And this time, the argument is NOT going to be derailed by FAKE profiles, by invalid heat maps, use worthless ISS line graphs, or railroaded by dominant players.


Let me be specific. I AGREE with the study conclusion !

It was a test for two shallow model designs that you use. It did what you needed. The procedure of moving and offsetting shallows stops, is not beneficial.

OK ?


but...

The test and its results apply to shallow stop model designs - we don't make those. No one here uses them.

There are no deep stops in your test, and there are no tech practices in your test. The connection to tech practices and real world tech deco, does NOT exist.




End of story.


Now, if only we could stop this false and invalid agenda effort to make that connection. Because as I have shown, there is no science to make that connection.

Thank you.

.

 

[rossh]

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

GF... what is GF? Oh yes,... its an add on time fudge... not a model. GF is this moving target, with no baseline, undefined datum... 40/70, no 30/50 no... 15/95... no I want to do 1111/1111.


You want to compare models, then compare models..


But all your doing here is saying... my stretched out GF plan, is longer than your stretched out GF plan. Who cares.


.

 

[Shotmaster]

Thank you Dr. Doolette!

 

[David Doolette]

Straight back to fantasy land I see.

Yes - they do show exact supersaturation values... In any pressure units you choose to select.


You want to add up 16 parallel cells, because it puts a bias in you graph for shallow stops. How surprising. But its no more relevant than reading the primary values only. Your approach of adding up 16, of which at least 10 will be insignificant, is well, just noise. How


Only one cell has control at any time. Only one cell has the limit at any one time. Tracking and summing that one primary value, is more realistic and can be used to cross compare with some validity.

30% ... of what? Noise? So there is 30% less stress in the water then. I guess its all balanced up then.

Ross

You seem to be continually confusing generating a schedule with an algorithm - which in those you are familiar with have one controlling compartment at any one time - and an algorithm-independent evaluation of a schedule by looking at supersaturation or integral supersaturation. In the latter case it is entirely appropriate to look at all compartments at all times. Since the compartments represent potential DCS-injury-sites, the probability of DCS is 1 minus the joint probability of no injury in all compartments. So to evaluate the risk of a profile you do want to add up the integral supersaturation in all compartments.

David Doolette

 

[rossh]

Ross

You seem to be continually confusing generating a schedule with an algorithm - which in those you are familiar with have one controlling compartment at any one time - and an algorithm-independent evaluation of a schedule by looking at supersaturation or integral supersaturation. In the latter case it is entirely appropriate to look at all compartments at all times. Since the compartments represent potential DCS-injury-sites, the probability of DCS is 1 minus the joint probability of no injury in all compartments. So to evaluate the risk of a profile you do want to add up the integral supersaturation in all compartments.

David Doolette

Well the problem is, adding up 16 puts a bias towards shallow stop profiles. Which then make this method invalid for comparing shallow and deep profiles.

As you fully know, if a slow tissue was to some how on gas enough, it would rise to top and become the limit. And it works this way already. if you look carefully, you can see how the deeper on gassing has given a higher surface supersaturation. The problem is solved already by Haldane / Schriener equations - it doesn't need fiddling.

.

 

[PfcAJ]

it doesn't need fiddling.

.

It clearly needs some fiddling. We already established that VPM is broken as risk of DCS goes up as deco time goes up.

ZHL is broken, too, but you can fiddle with the GFs to make it less broken.