Probabilistic pharmacokinetic models of decompression sickness in humans: Part 2, coupled perfusion-diffusion models.

Decompression sickness (DCS) can be experienced following a reduction in ambient pressure; such as that associated with diving or ascent to high altitudes. DCS is believed to result when supersaturated inert gas dissolved in biological tissues exits solution and forms bubbles. Models to predict the probability of DCS are typically based on nitrogen and/or helium gas uptake and washout in several theoretical tissues, each represented by a single perfusion-limited compartment. It has been previously shown that coupled perfusion-diffusion compartments are better descriptors than solely perfusion-based models of nitrogen and helium uptake and elimination kinetics observed in the brain and skeletal muscle of sheep. In this work, we examine the application of these coupled pharmacokinetic structures with at least one diffusion compartment to the prediction of the incidence of decompression sickness in humans. We compare these models to LEM-NMRI98, a well-described U.S. Navy gas content model, consisting of three uncoupled perfusion-limited compartments incorporating oxygen and linear-exponential kinetics. Pharmacokinetic gas content models with a diffusion component describe the probability of DCS in human bounce dives made with air, single non-air bounce dives, and oxygen decompression dives better than LEM-NMRI98. However, for the full data set, LEM-NMRI98 remains the best descriptor of the data.

Probabilistic pharmacokinetic models of decompression sickness in humans, part 1: Coupled perfusion-limited compartments.

Decompression sickness (DCS) is a disease caused by gas bubbles forming in body tissues following a reduction in ambient pressure, such as occurs in scuba diving. Probabilistic models for quantifying the risk of DCS are typically composed of a collection of independent, perfusion-limited theoretical tissue compartments which describe gas content or bubble volume within these compartments. It has been previously shown that 'pharmacokinetic' gas content models, with compartments coupled in series, show promise as predictors of the incidence of DCS. The mechanism of coupling can be through perfusion or diffusion. This work examines the application of five novel pharmacokinetic structures with compartments coupled by perfusion to the prediction of the probability and time of onset of DCS in humans. We optimize these models against a training set of human dive trial data consisting of 4335 exposures with 223 DCS cases. Further, we examine the extrapolation quality of the models on an additional set of human dive trial data consisting of 3140 exposures with 147 DCS cases. We find that pharmacokinetic models describe the incidence of DCS for single air bounce dives better than a single-compartment, perfusion-limited model. We further find the U.S. Navy LEM-NMRI98 is a better predictor of DCS risk for the entire training set than any of our pharmacokinetic models. However, one of the pharmacokinetic models we consider, the CS2T3 model, is a better predictor of DCS risk for single air bounce dives and oxygen decompression dives. Additionally, we find that LEM-NMRI98 outperforms CS2T3 on the extrapolation data.

The probability and severity of decompression sickness.

Decompression sickness (DCS), which is caused by inert gas bubbles in tissues, is an injury of concern for scuba divers, compressed air workers, astronauts, and aviators. Case reports for 3322 air and N2-O2 dives, resulting in 190 DCS events, were retrospectively analyzed and the outcomes were scored as (1) serious neurological, (2) cardiopulmonary, (3) mild neurological, (4) pain, (5) lymphatic or skin, and (6) constitutional or nonspecific manifestations. Following standard U.S. Navy medical definitions, the data were grouped into mild-Type I (manifestations 4-6)-and serious-Type II (manifestations 1-3). Additionally, we considered an alternative grouping of mild-Type A (manifestations 3-6)-and serious-Type B (manifestations 1 and 2). The current U.S. Navy guidance allows for a 2% probability of mild DCS and a 0.1% probability of serious DCS. We developed a hierarchical trinomial (3-state) probabilistic DCS model that simultaneously predicts the probability of mild and serious DCS given a dive exposure. Both the Type I/II and Type A/B discriminations of mild and serious DCS resulted in a highly significant (p << 0.01) improvement in trinomial model fit over the binomial (2-state) model. With the Type I/II definition, we found that the predicted probability of 'mild' DCS resulted in a longer allowable bottom time for the same 2% limit. However, for the 0.1% serious DCS limit, we found a vastly decreased allowable bottom dive time for all dive depths. If the Type A/B scoring was assigned to outcome severity, the no decompression limits (NDL) for air dives were still controlled by the acceptable serious DCS risk limit rather than the acceptable mild DCS risk limit. However, in this case, longer NDL limits were allowed than with the Type I/II scoring. The trinomial model mild and serious probabilities agree reasonably well with the current air NDL only with the Type A/B scoring and when 0.2% risk of serious DCS is allowed.