Differentiation at Autopsy Between in Vivo Gas Embolism and Putrefaction Using Gas Composition Analysis

Int J Legal Med DOI 10.1007/s00414-012-0783-6

ORIGINAL ARTICLE

Differentiation at autopsy between in vivo gas embolism and putrefaction using gas composition analysis

Yara Bernaldo de Quirós & Oscar González-Díaz & Andreas Møllerløkken & Alf O. Brubakk & Astrid Hjelde & Pedro Saavedra & Antonio Fernández

Received: 18 January 2012 /Accepted: 4 October 2012 # Springer-Verlag Berlin Heidelberg 2012

Abstract Gas embolism can arise from different causes New Zealand White Rabbits models: control or putrefaction, (iatrogenic accidents, criminal interventions, or diving relat- infused air embolism, and compression/decompression. The ed accidents). Gas analyses have been shown to be a valid purpose of this study was to look for qualitative and quan- technique to differentiate between putrefaction gases and titative differences among groups and to observe how pu- gas embolism. In this study, we performed systematic nec- trefaction gases mask in vivo gas embolism. We found that ropsies at different postmortem times in three experimental the infused air embolism and compression/decompression models had a similar gas composition prior to 27- – Electronic supplementary material The online version of this article h postmortem, being typically composed of around 70 – (doi:10.1007/s00414-012-0783-6) contains supplementary material, 80 % of N2 and 20 30 % of CO2, although unexpected which is available to authorized users. higher CO2 concentrations were found in some decom- : Y. Bernaldo de Quirós A. Fernández (*) pressed animals, putting in question the role of CO2 in Veterinary Histology and Pathology, Department of Morphology, decompression. All these samples were statistically and Institute of Animal Health, Veterinary School, significantly different from more decomposed samples. University of Las Palmas de Gran Canaria (ULPGC), Gas composition of samples from more decomposed ani- Trasmontaña s/n. 35416, Arucas, Las Palmas, Spain mals and from the putrefaction model presented hydrogen, e-mail: [email protected] which was therefore considered as a putrefaction marker.

O. González-Díaz Keywords Putrefaction . Gas embolism . Decompression . Physical and Chemical Instrumental Center for the Development . of Applied Research Technology and Scientific estate (CIDIA), Gas composition Nitrogen Edificio Polivalente 1, University of Las Palmas de Gran Canaria (ULPGC), Campus de Tafira s/n, 35017, Introduction Las Palmas, Spain

A. Møllerløkken : A. O. Brubakk : A. Hjelde Air embolism (AE) is the entry of atmospheric or alveolar air Department of Circulation and Medical Imaging, into the vascular system and is mainly an iatrogenic problem Norwegian University of Science and Technology, [1]. In many cases, it arises as a complication of numerous Trondheim, Norway invasive medical procedures [1], criminal intervention, or P. Saavedra barotraumas [2]. However, gas embolism of other origins Department of Mathematics, can occur as a consequence of significantly supersaturated University of Las Palmas de Gran Canaria (ULPGC), gas tissues after decompression [3], when the sum of the Campus de Tafira s/n, 35017, Las Palmas, Spain dissolved gas tensions (oxygen, CO2,nitrogen,andhelium) and water vapor exceeds the local absolute pressure [4]. Present Address: The diagnosis of gas embolism at autopsies is based on Y. Bernaldo de Quirós morphological findings and on the chemical analysis of the Biology Department, Woods Hole Oceanographic Institution, MS#50, gas collected from the heart [5]. To avoid atmospheric air Woods Hole, MA 02543, USA entrance as a result of dissection, Richter’s technique is Int J Legal Med typically used [6]. Gas chromatography can differentiate Experimental putrefaction model between gas embolism and putrefaction gases [2, 5, 7, 8]. Putrefaction is a continual process based on tissue break- Anesthetized NZWR were euthanized with an intraperitone- down by microorganisms such as bacteria, fungi, and pro- al injection of 200 mgkg−1 diluted pentobarbital. Euthanized tozoa (from the intestine and the environment) following animals were kept in hermetically sealed plastic boxes for autolysis [2] and results in the production of gases, liquids, biological material at room temperature (22.9–24.6 °C) for and simple molecules. Therefore, differentiation between 1, 3, 6, 12, 27, 42, 47, 53 (n01 for each time) and 67 h (n0 gas embolism and putrefaction gases might be postmortem 2) PM prior to necropsy. (PM) time and decomposition status-dependent. PM time is unknown on many occasions, and putrefaction processes are Experimental infused AE model dependent on the weather, especially temperature and hu- midity. Decomposition status can be coded based on the Anesthetized NZWR received atmospheric air infusion morphological status of the external and internal organs of through a catheter (0.36 mm I.D.) placed in the central vein the animal [9]. Since environmental conditions are expected of the ear with the use of a pump at 2.2 mLmin−1 until the to vary with different crime scenarios, comparison of gas rabbit expired. The total volume of air infused varied be- composition within the same decomposition code seems tween 4.5 and 13 mL. The dead animals were placed in more reliable. There are few systematic studies, using labo- hermetically sealed plastic boxes for biological material at ratory animals, that compare gas composition with PM time room temperature (23–25.3 °C) for 0, 20, and 40 min, and 1, or decomposition status [10, 11]. Such studies have tradi- 3, 6, 12, 27, 42, 53, and 67-hPM (n01 for each time). tionally focused on iatrogenic air embolism, while other gas embolism causes, such as those involving supersaturated Compression/decompression model tissues, after decompression have not been studied in much detail: In these cases, high nitrogen content is expected [12]. In this case, anesthetized NZWR were compressed in pairs in Although for diving-related accidents gas analysis is also a dry, hyperbaric chamber (Animal Chamber System, NUT, sometimes performed [13], there are no studies on how Haugesund, Norway) to eight absolute atmospheres during a these gases evolve with PM time or how they interact with bottom time of 45 min, followed by a fast decompression putrefaction gases. (0.33 ms−1). Diving profile was selected for explosive decom- In the present study, we analyzed gas composition in pression induction. Animals that died were placed in hermet- three rabbit experimental models (putrefaction, infused air ically sealed plastic boxes for biological material at room embolism, and gases produced by decompression) and its temperature (23–25 °C) for 0, 20, and 40 min, and 1, 3, 6, evolution with PM time and decomposition status. 12, 27, 42, 53, and 67-hPM (n02 for each time). Animals that survived for 1 h after decompression were euthanized with an intraperitoneal injection of 200 mgkg−1 diluted pentobarbital. Material and methods Only those animals that died because of decompression (n0 11) (without euthanasia) were considered for this study. Experimental procedures PM procedures A total of 41 New Zealand White Rabbits (NZWR, Animal Supply Center of the Negrín Hospital, Spain and the Norwe- Complete necropsies were carried out for each animal at its gian University of Science and Technology, Norway) of 2.5– scheduled PM time. In addition to PM time, a decomposition 3.8 kg were used. The animals were divided into three exper- code from 1 (very fresh) to 5 (advanced putrefaction) based on imental groups: (1) control, experimental putrefaction model the conservation state of the body was given to each animal (n010); (2) infused AE model (n011); and (3) compression/ [9]. Necropsy was performed with the rabbit in dorsal decu- decompression model (n020). All experiments were per- bitus position. Dissection was done carefully to avoid cutting formed in accordance with the European Union regulations of large vessels. First, the skin was removed and the abdom- for laboratory animals and were conducted under surgical inal cavity opened to expose the vena cava. Secondly, the anesthesia (Medetomidine of 0.5 mgkg−1 and Ketamine of thoracic cavity was opened, allowing access to the heart. 25 mgkg−1, subcutaneously). Experimental protocols for the Finally, the animals were completely submerged in water to infused AE model and for the putrefaction model were ap- sample the gas and to avoid atmospheric air pollution. Gas proved by the Ethical Committee for Animal Experiments of sampling, storage, and analysis were performed in accordance the University of Las Palmas de Gran Canaria (Spain). The to Bernaldo de Quirós et al. [14]. Gases were sampled from Norwegian Committee for Animal Experiments approved the the intestine, right heart, left heart, vena cava, and interstitial protocol for the compression/decompression model. emphysema where present. A total of 158 samples were Int J Legal Med

th obtained (41 from the putrefaction model, 56 from the infused where Ni,H denotes the nitrogen observed in the i animal at AE model, and 61 from the compression/decompression mod- the Hhours postmortem, ai is the effect of the animal, and el). Gas composition of samples was calculated in micromoles finally, ei,H is the variability within each animal. Temporal and normalized in micromole percentage. Samples with an variation for gases of C/D was not detected in either cluster atmospheric air-like composition were considered as polluted during the considered period (β100), but it was detected in and were excluded from the study (n05). the AE model (β1<0). In order to find out if embolic gases could be discriminated Statistical analyses from putrefaction gases, we performed an analysis of the receiver operating characteristics (ROC) with observations made Wilcoxon’s test was applied to examine the differences in the until 27-hPM. A ROC curve was calculated for those samples content of nitrogen, CO2, and hydrogen in the AE and com- where differences in nitrogen content compared to putrefaction pression/decompression (C/D) model during the first 27-hPM samples were found: AE and C/D samples of cluster 1. Because compared to gases produced by putrefaction alone. During both groups of samples showed trends in the period considered, this time, the gas composition of nitrogen and CO2 samples their values were adjusted at 27-hPM in the form Ni,H−β1⋅(H− corresponding to the C/D model showed a high variability in 27) to perform the ROC analysis. In all ROC analyses, optimal comparison with samples of the AE model (Fig. 1). In antic- cut-off was considered as that which minimizes the distance ipation of these data having two distinct patterns, cluster between the curve and upper left corner. All data were analyzed analysis was performed using the K-means cluster procedure using the “R” data analysis software, version 2.11.1 [17]. (Fig. 4)[15]. In this way, two clusters were found in the C/D model: cluster 1 (n031) and cluster 2 (n010). To analyze the evolution of nitrogen (in percentage of Results micromoles) with time until 27-hPM, we considered for each group (AE, C/D cluster 1, and C/D cluster 2) the next Experimental putrefaction model mixed effects model [16]: Intestinal gas composition varied greatly between animals. Ni;H ¼ b þ b H þ ai þ ei;H ð Þ 0 1 1 However, CO2 was usually the main compound. The highest

Fig. 1 Evolution of oxygen, a) nitrogen, CO2, and hydrogen content along PM time in the < 27 hours PM 27 hours PM > 27 hours PM N2 AE model a and b in the C/D CO2 model H2 O2 mol % gas µ 0 20406080100

036 12 27 42 53 67 PM time (h) b)

< 27 hours PM 27 hours PM > 27 hours PM N2 CO2

H2 O2 mol % gas µ 0 20 40 60 80 100

036 12 27 427 6 PM time (h) Int J Legal Med

values for nitrogen and CO2 were of 61.55 and 81.13 μmol%, respectively. Presence of methane and hydrogen in the intestine was random and did not seem to have any relationship to PM time or decomposition code (Supplemental Fig. 1). Gas produced by putrefaction alone was recovered from the heart only after 42-hPM (decomposition code 4). Gas in the left heart was exclusively found in most decomposed rabbit with the code 5 (at 67-hPM). At this PM time,

abundant subcapsular emphysema was also observed. Gas 2 composition recovered from the vena cava, heart, and sub- H 100 2 capsularly in different organs was very similar. It consisted N μ 80 of a mixture of CO2 (28.3±12.8 mol%), nitrogen (35.8± 60 12.3 μmol%), and hydrogen (29.0±6.5 μmol%) in similar 40 proportions. Hydrogen was always present (Fig. 2). Thus, 20

0 102030405060 0 when representing these data in a three-dimensional scatter 0 20 40 60 80 100 plot, where y-axis is hydrogen, all values lay in the vertical CO2 plane (Fig. 3). Oxygen and methane were only found in low Fig. 3 Three-dimensional scatter plot showing the content of nitrogen μ concentrations (5.36±4.8 and 1.52±3.5 mol%) if present. (x-axis), hydrogen (y-axis) and CO2 (z-axis), according to treatment (AE model is represented in squares, C/D model in circles and putre- rhombus Air embolism model faction model in ). Data set was further divided into samples recovered up to 27 hours PM (open symbols: in blue for AE samples and green for C/D samples) and after 27 hours PM (solid symbols in Intestinal gas was always composed of high CO2 levels, reach- red color for all treatments) ing values as high as 100 %, except for three cases. These exceptions presented nitrogen levels of around 70 %, while all other samples presented nitrogen levels lower than 25 %. Once with time according to Eq. 1 (P<0.001). A slight decrease of more, hydrogen and methane seemed to have a random appear- nitrogen content and increase in CO2 content could be observed ance, but this time they appeared more frequently and in higher at 27-hPM (decomposition code 3) (Figs. 1 and 2). After 42-h concentration in the most decomposed animals. After 12-hPM, PM, at decomposition code 4 or higher, hydrogen started to the summation of hydrogen and methane reached values close appear in some samples, contributing to an increase in the to the 50 % of the sample composition (Supplemental Fig. 2). variability of the gas composition of the samples (Fig. 1). Gas composition of intravascular bubbles of animals studied within 27-hPM was very similar, regardless of sample location. Compression/decompression model Before 12-hPM (decomposition code 2), gas composition was consistently composed of nitrogen (77±5 μmol%) and CO2 In this model, intestinal gas had a more stable composition (22±6 μmol%) (Figs. 1 and 2). However, nitrogen decreased between animals and PM time. The major compound in all

Fig. 2 Gas composition (expressed as the percentage of the average mole fraction) of samples obtained from the three models. AE and C/D models are divided in different PM observational periods according to Fig. 1. Samples from the C/D model recovered before 12-h PM were further divided in clusters 1 and 2 Int J Legal Med the samples (except for one sample taken 20-minPM) was the samples from the putrefaction model (represented in the

CO2, reaching values as high as 92 %. Most samples com- solid rhombus) were in the vertical plane, while samples prised 45–70 μmol% CO2,10–30 μmol% N2,and8– recovered within the first 27-hPM (represented with open 15 μmol% of CH4 (Supplemental Fig. 3). symbols) from the AE or C/D model laid in the horizontal Gas composition of samples obtained from the circulatory plane (except for the two samples previously mentioned) system was very similar to that of infused AE but with higher (Fig. 3). Samples from the AE and C/D model were mostly variability within the first 12-hPM (Fig. 1). Although most of found in the vertical plane together with the putrefaction the samples contained nitrogen levels of around 70–80 μmol% samples, although some were in the horizontal plane as well. and CO2 content of around 20–30 μmol%, there were some Hydrogen median for the putrefaction group was 27.4 % samples with much higher CO2 levels(upto98μmol% al- (IQR018.8, 35.6). Thus, we found statistically significant though most of the values of these samples were between 30– differences (P<0.001) between “in vivo” gas embolism 60 μmol%) (Fig. 1). When performing the K-means cluster sampled up to 27-hPM (animals with decomposition codes procedure for samples obtained during the first 27-hPM, two 2 and 3) and putrefaction gases (obtained from animals with clusters were obtained (Fig. 4). Cluster 1 (n031) had nitrogen decomposition codes 4 and 5). Statistically significant dif- levels of 60 μmol% or higher and CO2 levels of 40 μmol% or ferences were also found for nitrogen (P<0.001) when lower, while cluster 2 (n010) had nitrogen concentrations comparing samples obtained within 27-hPM of the AE lower than 60 μmol% and CO2 concentrations higher than model and cluster 1 from the C/D model to putrefaction 40 μmol% (Figs. 2 and 4). After 42-hPM (decomposition code gases. Nitrogen was not significantly different (P00.971) 4), hydrogen started to appear in some samples, contributing to for cluster 2 from the C/D model vs. putrefaction gases an increase in the variability on the gas composition of the (Table 1). On the other hand, differences in CO2 content samples as it did in the AE model (Fig. 1). were nonsignificant for the AE and cluster 1 compared to putrefaction gases (P00.308 and P00.127, respectively). In Comparison of the three experimental models accordance with these results, the ROC curve was only calculated based on nitrogen content for the AE model and Hydrogen gas was not detected in either the AE or C/D for cluster 1 from the C/D model. Optimal cut-off for AE at animals with 27-hPM or less (decomposition codes 2 and 27-hPM was 59 % with a sensitivity of 95 % and specificity 3), except for 2/60 samples. But it was detected in 22/33 of 100 %. Optimal cut-off for D/C (cluster 1) at the same samples recovered from both models after 27-hPM (decom- observation time was 45 %, with a sensitivity of 100 % and position codes 4 and 5). On the other hand, hydrogen was specificity of 80 % (Fig. 5 and Table 1). Therefore nitrogen present in all the samples from the putrefaction model. content is a better indicator to distinguish between in vivo

Indeed, when gases from the three models were represented gas embolisms and putrefaction gases than CO2. by scatter plots based on their hydrogen (y-axis), nitrogen Gas composition of samples of cluster 1 from the C/D

(x-axis), and CO2 (z-axis) content, we could observe how all and from the AE model was very similar, but gas composition of cluster 2 was clearly different, with higher CO2 content and lower nitrogen concentrations (Fig. 4). Al-

A Air Embolism though this cluster was different and could not be distin- A Cluster 1 (Compression/Decompression) guished from the putrefaction gases based on nitrogen or A Cluster 2 (Compression /Decompression) AA AAA CO2 content, it could be distinguished based on the absence AAA A of hydrogen (P<0.001) (Figs. 2 and 3). A AA 2 N Discussion

Gas composition from in vivo gas embolism was similar up to 12-hPM regardless the cause of embolism (infused air embolism vs. decompression) and different from putrefac-

0 20 40 60 80 100 tion and intestinal gases. It was mainly composed of high

0 20 40 60 80 100 values of nitrogen (70–80 %) and around 20–30 % of CO2.

CO2 Differences were found in some samples from the decompression model. These samples contained higher concentra- A Fig. 4 Nitrogen and CO2 content of samples from the AE model ( ), tions of CO but no hydrogen. Differentiation of in vivo gas and from the two clusters of the C/D (cluster 1 in open circles, cluster 2 2 in solid circles) obtained by K-means clusters procedure. All these embolism from putrefaction gases was possible until 27-h samples were obtained within the first 27-hPM PM or decomposition code 3, where hydrogen was absent. Int J Legal Med

Table 1 Statistical descriptors for nitrogen content of the three models: putrefaction, AE, and C/D (clusters 1 and 2)

Putrefaction N010 AE N019 C/D (cluster 1) N031 C/D (cluster 2) N010

Median (IQR) 36.2 (21.8; 43.4) 76.0 (74.1; 79.5) 69.0 (63.1; 77.9) 37.9 (30.3; 43.9) p valuea – <0.001 <0.001 0.971 Prediction 27 h (SE) – 66.2 (3.7) 59.8 (5.3) – AUC (95 % CI) – 0.989 (0.965; 1.00) 0.923 (0.837; 1.00) – Optimal cut-off – 59.48 44.87 – Sensitivity b – 0.95 (0.75; 0.99) 1.00 (0.89; 1.00) – Specificityb – 1.00 (0.72; 1.00) 0.80 (0.49; 0.94) –

In the AE and C/D models, samples over 27-hPM were excluded a Comparison of the median of each group of fresh gas with the median of putrefaction gas receiver operator curve descriptors for those groups with significant differences in nitrogen content compared to putrefaction gases (AE and cluster 1 from the C/D model) b Sensitivities and specificities were obtained for the optimal cut-off

From our results, we can conclude that hydrogen is a The composition of air embolism and its evolution with putrefaction marker in accordance with Pierucci and Gherson PM time is more or less clear since very similar results to [10]. It was present in all gas samples produced by putrefaction ours have previously been reported [10, 11]: starting values alone, while it was absent in 58 out of 60 samples taken within of nitrogen are around 70–80 %. There are some discrep- the first 27-hPM from animals of the AE and C/D models ancies between our results and previous observations re- (decomposition codes 2 and 3). This difference was statistical- garding how the remaining 20 % is distributed between ly significant and allowed us to distinguish between “fresh” oxygen and CO2. Pierucci and Gherson [11] report low samples (<27-hPM) and samples affected by putrefaction gas- CO2 concentrations on fresh animals compared to ours. es. Therefore, we would like to highlight the importance of Perhaps this difference follows non-use of anesthetics by analyzing hydrogen content in suspicious gas embolism cases. Pierucci and Gherson [11]. We used medetomidine/ket- ROC calculated at 27-hPM for nitrogen in the AE and amine. This anesthetic is reported to reduce heart and respi- cluster 1 from the C/D model confirm the good sensitivity ration rate, a significant drop in arterial pO2 and an increase and specificity of nitrogen for distinguishing in vivo gas in arterial pCO2 in NZWR [18]. embolism from putrefaction gases, although better results Both studies agree that around 24-hPM, which corre- were obtained for AE than for C/D. It is obvious that if sponds to decomposition code 3 in our study, a decrease of differences can be found at 27-hPM (decomposition code nitrogen content could be observed at the same time as CO2 3), they will be found more easily at shorter PM intervals content increases proportionally. Indeed, we have mathe- (decomposition code 2). matically confirmed this tendency with Eq. 1. Previous

Fig. 5 Receiver operating characteristics to distinguish a) b) between AE a and C/D cluster 1 b from putrefaction gases Sensitivity Sensitivity

cut off: 59.48 cut off: 44.87 Sensitivity: 0.9474 Sensitivity: 1.0000 Specificity: 1.0000 Specificity: 0.8000 AUC: 0.9895 AUC: 0.9323 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 1−Specificity 1−Specificity Int J Legal Med

studies and ours also agree in the appearance of hydrogen sickness risk when rats breathed elevated levels of CO2 in after 42-hPM (decomposition code 4) when a mix with either He–O2 or N2–O2 mixtures during the hyperbaric putrefaction gases is produced. Thus, the presence of putre- exposure. faction gases does not rule out the possibility of an in vivo Physical models have shown that CO2 diffuses rapidly gas embolism. Simply, it cannot be confirmed or denied into existing bubbles. One example is the empirical obser- through gas analysis. vation made by Harvey [26] where bubbles moved verti- All this process is less clear in the C/D model. Many cally through different water layers alternately saturated samples from the fresh animals belonging to the C/D model with air or CO2. The bubbles increased in size in the CO2- presented a similar gas composition to those of the infused saturated water and decreased in the air-saturated layer. AE with high values of nitrogen (70–80 %) and around 20– These changes in volume were very fast, indicating an

30 % of CO2 (Fig. 4). These results are in accordance with immediate equilibrium with the surrounding environment. previous reports in experimental animals: cats, dogs, and In addition, bubbles formed with less mechanical agitation rats [12, 19, 20], although they present higher CO2 content and grew at a faster rate in decompressed tubes filled with than previous descriptions for rabbits or humans [13, 21]. CO2-saturated water rather than nitrogen-saturated water Bert [12] was the first to analyze the gas from decompressed [23]. This could be explained by the diffusion coefficient animals. Indeed, he suggested that the gas, which would for CO2, which is 20 times higher than for oxygen and 40 threaten life when liberated, would be exclusively that times higher than that for nitrogen [27].Basedonthis,itis which was considerably increased in the blood: nitrogen. hypothesized that the CO2 content near a newly formed However, he was very surprised to find concentrations of cavity is very high, increasing the probability that the

CO2 that varied from 15 to 20 %, in addition to nitrogen. pressure inside the bubble becomes high enough to over- Nevertheless, we found in some samples even higher come the surface tension [28], thus yielding an important

CO2 concentrations. Concentrations of 30 % of CO2 have role for CO2 in the formation of bubbles. A mathematical been described for goats [22] and even of 60 % for rats [23]. model of gas pockets in crevices suggested that, although According to the K-means clusters procedure, those samples metabolic gases might represent a small amount of the gas with CO2 concentrations of 40 % or higher constituted a inside the bubble, their presence has a significant effect on separate cluster: cluster 2 (Fig. 4). This difference was very the behavior of the bubble under decompression due to the marked. Gas composition of samples from cluster 1 (n031) high diffusivity of these gases [29]. was very similar to samples from the AE model, while Although there is no conclusive report about which tissues samples from cluster 2 (n010) had higher CO2 concentra- the vascular bubbles originate from in decompression cases, it tions. This difference in CO2 content was probably not an is generally assumed that they form on the venous end of experimental error as the same methodology was applied to capillary beds or venous sinusoids as these sites have higher both models, and those concentrations of CO2 were never inert gas tensions and lower blood pressure [30]. Bubbles have found in the AE model. One possible explanation to the been observed to appear first in the small branches, predom- variability in CO2 content in the C/D model may be related inantly in the legs and back muscles [23]. Harris et al. [23] to the origin of the formation of the bubble: the source of gas proposed that these tissues might experience elevated CO2 in the AE model is atmospheric air, while in the C/D model, tensions, reducing the magnitude of mechanical disturbance it is the gas dissolved in the tissues. There is significant necessary for creating bubbles. Thus, these sites may therefore theoretical and empirical evidence suggesting a role of CO2 facilitate bubble formation. Harris et al. [23] also suggested in bubble formation. that bubbles might contain large amounts of CO2 at their point Behnke [24] stated that empirical evidence indicates a of origin because of its availability and diffusion rate. Gas higher incidence of bends in association with a rise in the bubble composition would change as the bubbles move out

CO2 level following dives using compressed air [24]. Bull- into the larger veins and heart, where nitrogen would be frogs exposed to atmospheres of 60–70 % of CO2 and then responsible for further growth and maintenance of the bubble. rapidly decompressed (hypobaric treatment) showed mas- We have found large CO2 concentrations in samples taken sive bubbling, while only three of 18 in the control group from the vena cava, although they were also found in some

(with no CO2 pretreatment but with the same decompression samples from the heart. profile) presented bubbles [23]. Additionally, when the We have also considered other factors such as the micro- authors injected a mixture of gas with 50 % CO2 into the circulation occlusion and the surface area of the bubble. blood stream of a frog, the bubbles equilibrated with the Occlusion of the microcirculation causes tissue ischemia blood in a few seconds. These authors also reported gas and also retards elimination of dissolved gas, thereby pro- analyses with CO2 contents ranging from 60 to 80 % in ducing local areas with gas tensions higher than the sur- bubbles formed in decompressed dead rats. Berghage [25] rounding tissue [31]. In this scenario, the CO2 tension is reported a statistically significant increase in decompression expected to increase. The rate of bubble growth is largely Int J Legal Med determined by the gas diffusion rate, which in turn depends Martín Barrasa for his help during the animal experiments at the Unit on the pressure difference, surface area, diffusion constants, of Research of the Negrín Hospital in Spain. This work was supported by the Spanish Ministry of Science and Innovation with two research and gas solubility [32]. Increasing local CO2 tension would projects, (AGL 2005-07947) and (CGL 2009/12663), as well as the increase the diffusion rate and thereby the rate of bubble Government of Canary Islands (DG Medio Natural). The Spanish growth. Ministry of Education contributed with personal financial support The surface area for gas exchange is reduced in bubbles (the University Professor Formation fellowship). The Central Norway Regional Health Authority and the Norwegian University of Science large enough to occupy an entire section of a vein as compared and Technology supported additionally the hyperbaric experiment. with smaller bubbles surrounded by blood. Thus, the compo- Finally, The Woods Hole Oceanographic Institution Marine Mammal sition in large bubbles would change more slowly, while at the Centre and Wick and Sloan Simmons provided funding for the latest same time occlude the circulation and alter the overall gas stage of this work. exchange between tissues and the alveoli. Furthermore, when Conflict of interest The authors declare no conflict of interest. there is massive bubbling, lungs can become overwhelmed and gas exchange become compromised. All these factors will increase gas tensions locally in tissues and peripheral veins Animal welfare The study was performed in accordance with all EU applicable laws, regulations, and standards, obtaining the corresponding and enhance bubble growth. approval from the different ethical committees. The rabbit necropsied at 12-hPM from the C/D model had the highest measured gas volume. Almost 12 mL was recovered from the right heart of this animal using a spi- rometer (although gas from the vena cava might have been References suctioned in addition). Both heart and vena cava were completely gas-filled. The blood had been pushed aside by the 1. Muth CM, Shank ES (2000) Primary care: gas embolism. N Engl J gas. Several hemorrhages were observed dispersed through- Med 342(7):476–482 2. Knight B (1996) Forensic pathology. Edward Arnold, London out the body. We consider that this animal is a good example 3. Hamilton RW, Thalmann ED (2003) Decompression practice. In: to find out how gas exchange is compromised in a lung Brubakk AO, Neuman TS (eds) Bennett and Elliott's physiology overwhelmed with gas bubbles. The reduced pulmonary and medicine of diving. Saunders, pp 455-500 surface area for gas exchange results in high local tissue 4. Vann RD, Butler FK, Mitchell SJ, Moon RE (2011) Decompression illness. Lancet 377(9760):153–164. doi:10.1016/s0140-6736 and venous gas tensions as the removal is compromised. (10)61085-9 This might explain the high concentrations of CO2 found 5. Pierucci G (1985) The post-mortem diagnosis of gas embolism. (40–60 %) in the vena cava and the right and left heart of Pathologica (Genoa) 77(1048):145–156 this animal. 6. Richter M (1905) Gerichtsärztliche Diagnostik und Technik. S. Hirzel, Leipzig Factors such as bubble size, the site of the bubble, gases 7. Keil W, Bretsxhneider K, Patzelt D, Behning I, Lignitz E, Matz J dissolved in the surrounding tissues, and the local metabolic (1980) Luftembolie oder Fäulnisgas? Zur Dianostik der cardialen rate vary greatly. Bubbles are gas phase systems in contact Luftembolie an der Leiche. Beiträge zur Gerichtlichen Medizin with a liquid phase where a continuous dynamic equilibrium 38:395–408 8. Bajanowski T, Kohler H, DuChesne A, Koops E, Brinkmann B is taking place. Since the gases dissolved in liquids and (1998) Proof of air embolism after exhumation. Int J Legal Med tissues surrounding the bubble are changing continuously, 112(1):2–7 we cannot expect the same composition for bubbles of 9. Kuiken T, García-Hartmann M Dissection techniques and tissues different size or in different tissues. Further studies are sampling. In: Newsletter (ed) 1st European cetacean society work- shop on cetacean pathology, Leiden, Netherlands, 1991 needed in order to understand the high CO2 values reported 10. Pierucci G, Gherson G (1969) Further contribution to the here for some samples in the C/D model. chemical diagnosis of gas embolism. The demonstration of In summary, gas composition from in vivo gas embolism hydrogen as an expression of “putrefactive component”. Zacchia 5(4):595–603 (70–80 % nitrogen and 20–30 % CO2) could be distinguish- 11. Pierucci G, Gherson G (1968) Experimental study on gas embo- able from putrefaction gases if gas sampling is done before lism with special reference to the differentiation between embolic 27-hPM, 12 h preferably. Higher content of CO2 in fresh gas and putrefaction gas. Zacchia 4(3):347–373 animals might be an indication of bubble formation driven 12. Bert P (1878) La Pression Barometrique: Recherches de by decompression, but further studies are needed in order to Physiologie Expérimentale. (Barometric pressure: researches in experimental physiology) (trans: Hitchcock MA, Hitchcock FA). clarify the role of CO2 in decompression. Hydrogen is a Masson, Paris putrefaction marker. 13. Lawrence C (1997) Interpretation of gas in diving autopsies. S Pac Underw Med Soc J 27(4):228–230 Acknowledgments The authors would like to thank all colleagues 14. Bernaldo de Quirós Y, González-Díaz Ó, Saavedra P, Arbelo M, from the University of Las Palmas de Gran Canaria (Spain) who Sierra E, Sacchini S, Jepson PD, Mazzariol S, Di Guardo G, contributed to this work and to the hyperbaric medicine division of Fernández A (2011) Methodology for in situ gas sampling, trans- the Norwegian University of Science and Technology (Norway) for its port and laboratory analysis of gases from stranded cetaceans. Sci scientific contribution. We would also like to thank Dr. Jose Luis Rep 1:193. doi:10.1038/srep00193 Int J Legal Med

15. Hastie T, Tibshirani R, Friedman J (2009) The elements of 24. Behnke AR (1951) Decompression sickness following exposure to Statistical learning: data mining, inference and prediction. high pressures. In: Fulton JF (ed) Decompression sickness. Springer series in statistics, 2nd edn. Springer, New York, USA Saunders, Phyladelphia, pp 53-89 16. Laird NM, Ware JH (1982) Random-effects models for longitudi- 25. Berghage TE, Keating LJ, Wooley JM (1978) Decompression nal data. Biometrics 38(4):963–974. doi:10.2307/2529876 sickness in rats and mice rapidly decompressed after breathing 17. R-Develpment-Core-Team (2012) R: A language and environment various concentrations of carbon dioxide. In: Lambertsen CJ (ed) for statistical computing. R Foundation for Statistical Computing, Proceedings of the sixth underwater physiology symposium, vol 6. Vienna, Austria Bethesda, pp. 485–496 18. Hellebrekers LJ, deBoer EJW, vanZuylen MA, Vosmeer H (1997) A 26. Harvey EN (1945) Decompression sickness and bubble fromation comparison between medetomidine-ketamine and medetomidine- in blood and tissues. Bulletin of the New York Academy of propofol anaesthesia in rabbits. Lab Anim 31(1):58–69 Medicine 21:505–536 19. Smith-Sivertsen J The origin of intravascular bubbles produced by 27. Guyton AC (1981) Physical principles of gaseous exchange; dif- decompression of rats killed prior to hypebaric exposure. In: fusion of oxygen and carbon dioxide through the respiratory mem- Lambertsen CJ (ed) Proceedings of the fifth symposium on under- brane. In: Textbook of medical physiology. 6th edn. W.B. water physiology, Washington, DC, 1976. Bethesda MD, pp 303- Saunders, Philadelphia, pp 491-503 309 28. Dean RB (1944) The formation of bubbles. J Appl Phys 15 20. Ishiyama A, Mano Y, Shibayama M, Maeda H (1981) Analysis of (5):446–451 gas composition of dysbaric decompressed bubble. In: Miller JW 29. Chappell MA, Payne SJ (2006) A physiological model of gas pockets (ed) The sixth meeting of the panel on diving physiology and in crevices and their behavior under compression. Respir Physiol technology. The United States–Japan Cooperative Program in Neurobiol 152(1):100–114. doi:10.1016/j.resp.2005.07.010 Natural Resources, California, U.S.A., pp 201-210 30. Francis TJR, Simon JM (2003) Pathology of Decompression 21. Ishiyama A (1983) Analysis of gas composition of intra vascular Sickness. In: Brubakk AO, Neuman TS (eds) Bennett and bubbles produced by decompression. Bull Tokyo Med Dent Univ Elliott's physiology and medicine of diving. Saunders, pp 530-556 30(2):25–36 31. Davies JM (1983) Studies on bubble formation after decompres- 22. Armstrong HG (1939) Analysis of gas emboli. Engineering sion: with special reference to the development and testing of the Section Memorandum Report. Wright Field, Ohio integrating ultrasonic pulse-echo imaging system for bubble detec- 23. Harris M, Berg WE, Whitaker DM, Twitty VC, Blinks LR (1945) tion. University of Oxford, Oxford Carbon dioxide as a facilitating agent in the initiation and growth 32. Blatteau JE, Souraud JB, Gempp E, Boussuges A (2006) Gas of bubbles in animals decompressed to simulated altitudes. J Gen nuclei, their origin, and their role in bubble formation. Aviation Physiol 28(3):225–240 Space and Environmental Medicine 77:1068–1076