INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

MODELING OF DECOMPRESSION SICKNESS (DCS) risk has been impeded by the inability to identify correlated physiological variables. Some studies have tried to find a correlation between DCS risk and variables such as body temperature, body weight, exercise, gender, adiposity, age, serum cholesterol, sensitivity to complement activation, Doppler bubble grades, and patent foramen ovale (4, 12, 16, 23, 28). However, where some studies have found a correlation, others refute those results (5, 8, 16). The only physiological variable that has been undisputedly correlated with DCS risk in rats is body weight (20). Because reliable physiological correlates are lacking, researchers have used a variety of models based solely on the physical history of the compression and decompression sequence to find variables that can predict the probability of DCS (26, 31-34).

The DCS risk assessment used in this study builds on previously published models used in DCS research (26, 31, 33, 34). The goal is to estimate the beneficial effects on DCS risk of the active removal of tissue H₂ by injecting H₂-metabolizing microbes into the intestines of pigs during simulated H₂ dives and to suggest a physiological mechanism for the process called H₂ biochemical decompression (19). The metabolism of H₂ in the intestine is readily followed by measuring the release of CH₄, the metabolic end product of the microbial metabolism (21)

(1)

The model presented here differs from earlier models of DCS (26, 31, 33, 34) in that a parameter for the microbial metabolism of H₂ is included. In constructing this model, the microbial metabolism of H₂ was considered to have a direct physiological effect by influencing the gas kinetics. The measure of H₂ metabolism was based on either the total microbial activity injected into the animals (Inj), or as the CH₄ release rate from individual animals, assuming that there is a direct correlation between the H₂ metabolized inside the intestines and the release of CH₄ from the intact animal (Eq. 1). This model was used to predict the advantage of biochemical decompression for hyperbaric H₂ exposures in general and thus offers a predictive advantage over the descriptive logistic regression model of biochemical decompression that was derived earlier (10, 18).

As a result, this study used a measured physiological variable, namely the rate of removal of H₂ by the metabolism of intestinal microbes, together with the physical history of the hyperbaric exposure to predict the DCS incidence after a hyperbaric exposure to H₂.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Data Set and Case Descriptions

Pigs (Sus scrofa, 17-23 kg) were used for all experiments. The pigs were housed before experiments in an accredited animal care facility and had ad libitum access to water. The pigs were fed once daily with laboratory animal chow (Harlan Teklad, Madison, WI; 2% by body wt). All procedures were approved by an Animal Care and Use Committee. The experiments reported here were conducted according to the principles presented in the Guide for the Care and Use of Laboratory Animals (National Research Council, 1996).

The data set contains 109 well-documented hyperbaric exposures, with 53 severe DCS cases (Table 1). All animals were juvenile male Yorkshire pigs, either castrated (n = 98) or noncastrated (n = 11). There was no difference in DCS incidence between the castrated and noncastrated animals (logistic regression analysis, P > 0.3). Consequently, these two groups were pooled for all subsequent analyses.

The animals were divided into three groups: untreated control (UC, n = 69, mean body wt 19.6 ± 1.4 kg), surgical control (SC, n = 11, mean body wt 19.6 ± 1.6 kg), and treated (n = 29, mean body wt 19.5 ± 1.3 kg). Treated animals had H₂-metabolizing microbes injected into the large intestine (mean 0.92 ± 0.49 mmol CH₄/min, range 0.20-2.20 mmol CH₄/min), a procedure requiring major surgery. This surgical procedure and the culturing of Methanobrevibacter smithii are similar to prior work in rats (19) and have been described elsewhere (18). In the SC group, animals received intestinal injections of saline bubbled with CO₂ to deoxygenate it. The SC group has been described in detail along with the surgical procedure (18). Because DCS incidence was indistinguishable between the SC and UC groups (P > 0.29, Fisher exact test), the two groups were pooled and will be referred to as control animals (C = UC + SC, n = 80).

The experiments were carried out over a period of 30 mo (May 1997-Dec 1999) and include a variety of different pressurization and depressurization sequences (Table 1). The hyperbaric H₂ exposures were performed in a dry chamber (WSF Industries, Buffalo, NY) of 5,600-liter volume at 1 atm. Each animal was subjected to only one hyperbaric exposure. The chamber was controlled by a computer that stored the ambient pressure (Pamb, atm), chamber temperature (°C), O₂ concentration (%), and elapsed time (min). A gas chromatograph (Hewlett-Packard 5890A, Series II, Wilmington, DE) measured the chamber gases O₂, H₂, He, N₂, and CH₄ every 12 min.

Each hyperbaric exposure commenced with a pressurization of the chamber with He to 11 atm, followed by a flush of the chamber with H₂ until the H₂ concentration was over 60%. The initial pressurization with He was necessary as a safety measure to prevent an explosive mixture of H₂ and O₂ in the chamber, as described in detail elsewhere (18-20). Pressurization then continued with H₂, with addition of O₂ as needed to maintain normoxia (0.2-0.4 atm O₂). Maximum pressure of the exposures ranged from 22 to 26 atm (absolute pressure; 2.23-2.63 MPa; 700-825 feet of seawater pressure equivalent; Table 1). Final gas composition in the chamber at maximal pressure was roughly 88% H₂, 9% He, 2% O₂, and 1% N₂ for exposures lasting 3 h, with more H₂ and less He and N₂ present in the chamber over time. Chamber concentrations of CH₄ ranged from <1 ppm to 8 ppm after 3 h, and up to 40 ppm after 24 h.

The start of the decompression (time 0) was set as the time that animals were last definitely free of DCS signs (T₁). The chamber was decompressed at 0.45-1.8 atm/min to 11 atm while each animal was observed closely. Decompression could not continue past 11 atm because of the need for a normoxic environment and safe handling of H₂ and O₂ mixtures as described earlier (18-20). On arrival at 11 atm, the animal was made to walk on a treadmill inside the chamber at 5-min intervals for up to 1 h (67-90 min, Tend) or until the animal was declared to have DCS (T₂) (18).

Most animals displayed violaceous macular lesions with or without pruritus. On the basis of previous descriptions in the literature, these lesions resembled skin DCS (6, 9). However, skin DCS alone was not considered a DCS case for the purpose of this study. Severe DCS symptoms were of two types: cardiopulmonary or neurological (3, 6, 9). Severe symptoms of DCS included walking difficulties, fore- and/or hindlimb paralysis, falling, seizures, labored breathing, and convulsions. Every diagnosis of DCS was determined by a consensus of at least three observers who discussed their diagnosis as the events occurred. The observers were not blind to the treatments or dive profiles, resulting in the possibility of biased diagnosis. However, the rigorous compression and decompression sequences used throughout this study caused signs of DCS that were in most cases severe enough to leave no ambiguity. In the rare event that the observers did not agree, a detailed description or a viewing of a video recording of the animal was given to a neurologist or diving medical officer to evaluate and pass final judgment.

DCS Risk Assessment Modeling

The probabilistic models used previously, unlike most logistic regressions, were constructed to be well behaved in the sense that the P(DCS) was predicted to be 0 when no pressure reduction occurred and increased with increasing pressure reduction (34).

The probability of having DCS symptoms at a time t, after a hyperbaric exposure, is defined as

(2)

whereas freedom of symptoms until time t is defined as

(3)

where r is the instantaneous risk (7), and r  0. The r depends on the theory used to describe the mechanism of DCS and can be one of several measures integrated over the dive and postdive period.

The probability of developing DCS for a particular hyperbaric exposure is defined as the integrated r over the exposure period through the end of the postdecompression observation period (Tend). Consequently, if DCS was observed, the following formula was used

(4)

This is the product of the probability of DCS not occurring in the interval [0-T₁] and the probability of DCS occurring in the interval [T₁-T₂] (26, 34). T₁ is defined as the last time when the diver was definitely free of DCS symptoms, and T₂ the time the diver was declared to have definite signs of DCS (26, 34). For an animal with no DCS, Eq. 2 is used to estimate the P(DCS) with the upper limit for integration at the end of the postdecompression observation period (Tend). For an animal displaying DCS, Eq. 4 is used and the integration is performed until the time the animal definitely shows signs of DCS (T₂).

The choice of T₁ can be made in various ways, as detailed elsewhere (34). Because of subjective aspects inherent in diagnosing DCS, T₁ is set to the start of the decompression for this study. This approach is conservative and may result in some loss of temporal information. However, because T₁ is difficult to determine in some cases, this approach seems warranted.

Model Variations

View larger version (18K): [in this window] [in a new window]   Fig. 1.   Model behavior showing the ambient pressure (Pamb), estimated tissue tension for H2 and He (Ptis), the probability of decompression sickness [DCS; P(DCS)], the instantaneous risk (r1, Eq. 5) for the Immediate (A) model, and the relative supersaturation (r2, Eq. 6) for the Delayed (B) model using a time constant () and a scale factor (G) for a sample dive to 24 atm. The Pamb, the r1, and the r2 for both models (C) are displayed to show the delayed maximum for the r2 compared with r1. Constant pressure was maintained for 3 h followed by decompression at 0.9 atm/min to 11 atm. The Immediate model (A) used  = 0.80 and G = 2.23 whereas the Delayed model (B) used  = 0.75 and G = 0.14. The units for r are inverse of time; the scale for r is arbitrary for convenience.

It is widely accepted that the presence of gas bubbles in tissues leads to DCS (14). If one believes that the presence of bubbles triggers the symptoms, it may be most appropriate to use a risk function in which the greatest r occurs immediately after a decompression step (34). Alternatively, if one prefers the theory that DCS develops after the bubbles have grown to a certain size (28) or that they trigger an immune or hematological response (17, 24, 35), r may rise slowly to a maximum and finally decrease slowly. Unrelated to the theory used, when r is zero there should be no occurrence of DCS, and as r increases so should the DCS incidence (34).

Immediate model. The instantaneous risk (r₁), is defined as the relative difference between the tissue tension (Ptis, atm) and the absolute Pamb (atm) above a threshold (Thr) (26, 33, 34)

(5)

where Ptis refers to the sum of the tissue tensions for H₂ (Ptis_H2, atm) and He (Ptis_He, atm) and G₁ is a scaling factor (min¹) to be determined from the fitting procedure. The inclusion of a threshold parameter has been shown to improve the fit for some human data sets (33), whereas this parameter has not been very successful in describing others (26). The contributions of O₂, CO₂, and water vapor to Pamb (1) were ignored, as in some other DCS modeling efforts (29). In this model, r₁ is constrained to be 0, and, accordingly, r₁ will be set to 0 at any time Pamb > Ptis (33, 34). The maximum value of r₁ is reached at arrival at 11 atm, after which it decreases quickly as Ptis approaches Pamb (Fig. 1A).

Delayed model. The second model has the same structure as the Immediate model, but here the relative supersaturation is integrated over the exposure time (34)

(6)

For this model, the relative supersaturation is integrated from the first occurrence that Ptis > Pamb, followed by integration of both positive and negative values of supersaturation. However, the risk cannot be negative, so r₂ is constrained to be 0. The value of r₂ increases more slowly compared with r₁ and reaches its maximum value at the time when Pamb > Ptis (Fig. 1B). As can be seen in Fig. 1C, the r₂ maximum value occurs slightly later in the decompression compared with r₁. This leads to a delayed effect on the accumulation of P(DCS) for the Delayed model (Fig. 1B) compared with the Immediate model (Fig. 1A).

Tissue Inert Gas Tension

The differential equation used to describe the inert gas tissue tension was as follows

(7)

where Ptis is the tissue tension of the inert gas (atm), Pblood is the arterial blood tension of the inert gas (atm),  is the time constant (min) to be determined from the data, and t is time (min). The time constant determines the flux of gases in and out of the tissues. The change in Pblood during compression and decompression is assumed to be instantaneous and therefore equal to the alveolar partial pressure for that inert gas at all times. The alveolar partial pressure in turn is assumed to be equal to the Pamb for that gas.

The effect of a change in Pblood on the Ptis is computationally extensive and has been described in detail elsewhere (26, 33). The inert gas flux during the hyperbaric experiment was calculated by dividing the compression and decompression sequence into pressure-time ramps (33, 34). For the present model, it is assumed that  is the same for He and H₂. Hence, three parameters need to be determined: , G, and Thr.

Effect of Microbial H₂ Metabolism on Ptis_H2

(8)

CH₄

1

CH₄

View larger version (34K): [in this window] [in a new window]   Fig. 2.   Proposed physiological mechanism for H2 biochemical decompression. For each figure, the number in parentheses represents the net result of the process of gas uptake (compression) or elimination (decompression) by the body at the end of a circulatory pass from the gas exchange surface throughout the body and returning to the gas exchange surface. Units in these figures are hypothetical and are used only to describe the current working hypothesis.   During compression (A, top) in an untreated animal exposed to a hyperbaric environment of 100 units of H2 pressure (PH2), arterial blood is loaded with H2 at the gas exchange surface. Alveolar PH2 is somewhat less than the ambient PH2, until the tissues have reached saturation. The PH2 of alveolar gas is determined by a balance of 2 processes: removal of H2 by capillary blood to the tissues and replenishment by alveolar ventilation. By the time blood leaves the lungs, alveolar PH2 is in complete equilibrium with the blood (Pblood). H2 is delivered to all regions of the body via the arterial blood supply. H2 in the blood diffuses into tissues until the blood and tissues are in equilibrium. Saturation is reached when Pblood, Ptis, and Pamb are all equal.   In the treated animal (A, bottom), the process is slightly different. Microbial metabolism of H2 in the intestines works as a sink, resulting in a mixed venous return that will have a lower PH2 compared with the untreated animal. As venous blood reaches the lung, alveolar PH2 will be somewhat lower than in the untreated animal, assuming that the ventilation is the same in the treated and untreated animal. The result is a slightly lower arterial PH2 compared with the untreated animal. Overall magnitude is dependent on the efficacy of the metabolism of H2. The result is a prolonged time to equilibrium, a somewhat lower PtisH2 at equilibrium, and a chronically subsaturated state.   During decompression (B, top), fluxes of inert gases are reversed. In the case of the untreated animal, H2 is transported from tissues to blood and via the blood to the gas exchange surface. This leads to a continuous decrease in the PtisH2. In a treated animal, there is a dual removal of H2: one from the intestines and one from the gas exchange surface. This enhances the reduction of the PtisH2 in the treated animal. The lower overall PtisH2 of the mixed venous blood in the treated animal (B, bottom) may result in a lower risk of bubble formation compared with the untreated animal, thereby reducing the probability of DCS.

The parameters were determined from the data by fitting the estimated P(DCS) to the actual outcome for each hyperbaric exposure, using Eq. 2 or 4. If the animal did not suffer from DCS by the end of the 1-h postdecompression observation period (Tend), the hyperbaric exposure was considered to be safe throughout the experiment.

Null Models

(9A)

(9B)

Model Analysis

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The cumulative distribution of the 53 cases of DCS vs. time is shown in Fig. 3. The data show that the occurrence of DCS was tightly clustered within the first half hour after reaching the observation pressure (11 atm, Fig. 3). No cases of DCS were observed during the transition to lower chamber pressure (Fig. 3).

View larger version (15K): [in this window] [in a new window]   Fig. 3.   Cumulative number of DCS cases with time after observation pressure (11 atm) is reached for the 53 observed cases of DCS out of 109 hyperbaric H2 exposures.

The results for the Immediate (Eqs. 4 and 5) and Delayed (Eqs. 4 and 6) models are summarized in Table 2. Each model was tested with varying numbers and combinations of the model parameters (Imm 1-4 for the Immediate model and Del 1-6 for the Delayed model). Both models showed improvement compared with the Null model (Table 2, Null vs. Imm1 and Null vs. Del1). Therefore, incorporating a specific description of the dive history significantly improved the fit to the data. The LL values cannot be used to compare the Immediate and Delayed models with each other because neither is a subset of the other.

Addition of the threshold parameter Thr to the risk function improved the fit only for the Delayed model (Table 2, Del2, and Del5). The standard errors of the Thr parameter estimates were nearly equal to or greater than the parameter estimates themselves (Table 2). Likelihood ratio analysis of this determined that the 95% confidence limit, as based on change in the LL by 1.95 units, was asymmetric about the Thr parameter (data not shown).

Inclusion of the parameter for H₂ metabolic activity injected into the animals (A_Inj) improved the fit for both the Immediate and Delayed models (Table 2, Imm3, Del3, and Del5). The parameter for CH₄ release rate (A_CH4 ) improved the fit only for the Delayed model (Table 2, Del4). Only the Delayed model could successfully be fitted using four parameters, including both Thr and A_Inj as best fit parameters (Table 2, Del5). There was a trend for an improvement in fit when both A_CH4 and Thr were included in the Delayed model (P < 0.10, Table 2, Del6). Activity injected and CH₄ were significantly correlated with each other (r = 0.60, P < 0.001; Fig. 4); thus no model was tested that included both A_Inj and A_CH4.

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Rate of release of CH4 from animals vs. microbial activity injected into animals. Line represents least squares linear regression (y = 41.2 + 0.048x; r = 0.60, P < 0.0001). DCS outcome is indicated for each animal.

The observed vs. predicted DCS incidences for both models were tested by separating the data into groups divided by treatment or by varying compression and decompression sequences (Fig. 5). The Immediate (Imm3) and Delayed (Del4 and Del5) models satisfactorily predicted the number of DCS cases when the data set was divided into two groups (all controls vs. all treated, ² test, P > 0.4-0.8; Fig. 5A), four groups (controls at 3 different decompression rates vs. all treated animals, P > 0.3-0.7; Fig. 5B), or 14 different groups (controls vs. treated using all dive sequences listed in Table 1, P > 0.2-0.7; Fig. 5C). There was no apparent advantage to any one of these models over the other two (Fig. 5).

View larger version (10K): [in this window] [in a new window]   Fig. 5.   Observed number of cases of DCS vs. number of cases predicted by the Immediate (Imm3) and Delayed (Del4 and Del5) models. A: data divided into control and treated animal groups only. B: data divided into all treated animals and controls at either 0.45, 0.90, or 1.8 atm/min decompression rate. C: data divided into all 14 of the control and treated animal groups at all compression and decompression sequences used. Dotted line in each panel has a slope of unity, representing perfect fit of predictions to observations. All 3 models could satisfactorily predict DCS incidence values that were not different from the observed values, regardless of the manner of assigning groups (P > 0.20).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

DCS is notoriously difficult to diagnose because its manifestations are so varied and nonspecific (16). Choosing compression and decompression sequences with severe and rapid-onset outcomes as in this study helps reduce the subjectivity of the diagnosis. In this study, most DCS cases were unambiguous seizures or limb paralysis that occurred within the first 20 min after arrival at the lower chamber pressure (Fig. 3). This is similar to other studies on DCS in pigs (9), sheep (2), and rats (20). Previous experience with a pig model has indicated that virtually all manifestations of severe DCS will be evident within 1 h, with observation periods of 4-24 h adding more cases in less than 1% of tested animals (9). In a sheep model of DCS (2), a small number of cases continued to be diagnosed in the second and third hour of observation, but these may have been less severe symptoms. In DCS research in humans, symptoms may emerge as late as 20 h after return to 1 atm (2, 34). The prolonged latency observed in humans but not found in other animals may be an experimental artifact, because only obvious and early signs of DCS can be evaluated in other animals. Furthermore, other animals have only limited ability to report symptoms. Postdecompression, most pigs in this study manifested livid marks on the skin that are typical of skin DCS (6). We cataloged these marks for future reference but did not consider them sufficient to diagnose a case of DCS. This is in keeping with the practice for human divers of not offering recompression treatment for such mild symptoms alone.

Probabilistic models for DCS have been used successfully to model human and sheep data (2, 22, 26, 31, 33, 34). The fitted values were made from the observed outcome and the physical history of the hyperbaric experiment. Physiological interpretation of the parameters from these past studies must be made with care, because no actual physiological measurements were included. In contrast, in this study the proposed mechanism of biochemical decompression has been incorporated into the models. This mechanism is described in detail in Fig. 2. We were able to include in the models a measured physiological variable that is subject to manipulation, namely an estimate of the rate at which intestinal microbes were eliminating a portion of the body burden of H₂. The inclusion of such a variable will permit an evaluation of the beneficial effects of H₂ biochemical decompression in pigs from any compression and decompression sequence in H₂ and will be one of the first demonstrations of a mathematical model using a measured physiological variable to predict DCS.

Various models were tested and not found to have better fitting ability or any additional insights compared with those reported here. Among these were models with more than one tissue, models with a separate  for H₂ and He, models with a separate  during compression and decompression, models with a T₁ set by the observers at 11 atm, and models incorporating body weight and chamber temperature.

The successful incorporation of a parameter for H₂ metabolism in the models supports our hypothesis (19) that the H₂-metabolizing microbes reduced P(DCS) by causing an overall reduction in Pblood_H2 (Eq. 8). Two terms for H₂ metabolism were tested (Table 2): one relating the removal of H₂ in the pigs' tissues to the in vitro measurement of the total activity of H₂-metabolizing microbes injected into the intestines of the animals (A_Inj) and the other relating the removal of H₂ to the measured CH₄ release rate (A_CH4 ) from the pigs. These two estimates of microbial H₂ elimination were significantly correlated with each other (Fig. 4), with CH₄ ~10% of Inj (range 4-30%).

Inclusion of the parameter A_Inj, along with  and G, improved the fit to the data for both the Immediate and Delayed models, compared with models with only two parameters ( and G, Table 2, Imm1 and Del1). The parameter A_CH4, on the other hand, improved the fit only for the Delayed model. Using CH₄ had lower predictive power than using Inj in the Delayed model (Table 2, compare LL values for Del3 vs. Del4 and Del5 vs. Del6). This probably indicates a greater accuracy in measuring the metabolic activity of the microbes in vitro before their injection compared with measuring the release of CH₄ into the chamber by the animals (18). In the latter case, multiple steps separate the evolution of a CH₄ molecule inside the animal from sampling it by the gas chromatograph, with unknown kinetics and sampling errors at each of these steps. Accurately estimating the CH₄ during the hyperbaric experiment is complicated by the large volumes of chamber gas (chamber gas volume at 24 atm is equivalent to 135,000 liters at 1 atm) and the necessarily slow exhaust rate of the chamber (18). Furthermore, the metabolism of H₂ within the intestinal ecosystem is more complex than the metabolic pathway shown in Eq. 1. Although methanogenesis is the predominant pathway for microbial H₂ metabolism (21, 25), reduction of sulfate and nitrate and formation of acetate may also consume H₂ (13, 21, 25). The potential that not all H₂ metabolism was due to methanogenesis by M. smithii may mask part of the correlation between CH₄ and DCS outcome (Table 2; Fig. 4).

The total H₂ tissue burden eliminated by the intestinal H₂-metabolizing microbes is a crucial link between the Ptis_H2 and the DCS risk. The model (Del5) provides us with a way to assess the fraction of dissolved H₂ (Ptis; that is, Ptis_H2 + Ptis_He) removed by means of microbial metabolism. For this calculation, a treated animal compressed to 24 atm for 3 h and decompressed at 0.90 atm/min will be used (Table 1). By using the parameter for A_Inj, the difference in Ptis for an animal with or without intestinal injections of M. smithii can be estimated. Let us consider an animal with an Inj of 1.50 mmol CH₄/min (activity injection in the upper range of those used in the study; Fig. 4). The estimated Ptis from this computation was compared with an animal with an Inj = 0 mmol CH₄/min. On the basis of the model computations, it can be shown that the fraction of the total gas burden removed with additional injections of M. smithii is ~5%. In a previous study of H₂ biochemical decompression in rats, there was a 50% reduction in DCS incidence associated with an elimination by the intestinal H₂-metabolizing microbes of ~5% of the estimated total gas burden (19). The calculation in rats depended on numerous assumptions and simplifications regarding H₂ solubility in tissues. The calculation performed above for pigs used fewer assumptions but arrived at a similar numerical result. Calculations based on H₂ solubility in pigs have generated results that are on the order of 2-19% reduction of total gas burden for similar decreases in DCS incidence (18).

A graphical example of the possible reduction in the Ptis using Inj and Thr with and without a high dose of H₂-metabolizing microbes is shown in Fig. 6. Because risk of DCS is the area under the curve between Ptis and Pamb, where Ptis > Pamb, the beneficial effects of biochemical decompression can be seen (Fig. 6). Even though the two curves for Ptis for control and treated animals are close, the difference in the area under the curve between Ptis and Pamb is appreciable for DCS risk [Fig. 6, compare P(DCS)_C vs. P(DCS)_T]. This illustrates how elimination of relatively small fractions of dissolved gas may have a surprisingly large impact on the DCS incidence.

View larger version (18K): [in this window] [in a new window]   Fig. 6.   Ambient pressure (Pamb) and tissue tension of H2 + He for a control animal (PtisC) and a treated animal (PtisT) injected with a total H2-metabolizing activity of 1.50 mmol CH4/min and the estimated probability of DCS for a control [P(DCS)C] and treated animal [P(DCS)T] using the 4-parameter Delayed model (Table 2, Del5). Hyperbaric exposure was to a total pressure of 24 atm for 3 h with a decompression rate of 0.9 atm/min.

The models described here can be used to predict the overall benefit from H₂ biochemical decompression for any specified compression and decompression sequence. On the basis of the parameter estimates, it is possible to predict the change in the P(DCS) with varying CH₄ or Inj for any given hyperbaric sequence. For example, the most tested compression and decompression sequence in this study was to a total pressure of 24 atm for 3 h with a decompression rate of 0.9 atm/min (n = 20 control and 18 treated animals, Table 1). The observed DCS incidence was 80% (16 cases) and 39% (7 cases) for control and treated animals, respectively (Table 1). The predicted values for the Delayed (Del5) model with inclusion of Inj and Thr were 73% (14.6 cases) and 41% (7.4 cases) for control and treated animals, respectively (Fig. 5C). The reduction in DCS risk predicted by the Delayed model with inclusion of Inj is thus close to the observed value. The Imm3 and Del4 model predictions were only 1-2 DCS cases different from the Del5 model predictions (Fig. 5C). Naturally, predictions of P(DCS) made for dive exposures or treatment dosages beyond our data set must be made with caution and tested empirically before the full benefit of this model can be evaluated.

In conclusion, a hypothesis of the mechanism supporting biochemical decompression (Fig. 2) has been postulated and incorporated into a physiologically based probabilistic model (Eq. 8). The model can be used to predict the beneficial effects of biochemical decompression from any compression and decompression sequence with pigs in hyperbaric H₂. The physiological interpretation of the model gives a foundation for further research including other physiological measurements related to gas transport kinetics and estimates of tissue inert gas content. The model supports our hypothesis that removal of H₂ by microbial metabolism reduces the body burden of gas, thereby reducing the DCS risk by as much as 50%.