Administration of Anesthesia and Preparation
All animal procedures were approved and performed in accordance with the local animal care committee guidelines. Thirty New Zealand White male rabbits (Charles River Laboratories, Saint-Constant, Quebec, Canada), ranging in weight from 3.0 to 4.2 kg (mean [and standard deviation], 3.4 ± 0.4 kg), were studied. The rabbits were premedicated with an intramuscular injection (lumbar muscles) of ketamine hydrochloride (35 to 50 mg/kg) and xylazine (5 to 10 mg/kg) and maintained under anesthesia with a continuous intravenous infusion of ketamine (10 mg/kg/hr) and xylazine (2 mg/kg/hr). The outer layers of the neck were exposed, a cervical tracheotomy was performed, and a 3.5-mm uncuffed endotracheal tube was inserted and tied in place. Pancuronium bromide (0.1 mg/kg/hr) was continuously administered intravenously to obtain complete muscle relaxation. The tracheal tube was directly connected to a respiratory circuit, and the animals were ventilated with FiO2 0.5 (oxygen mixed with medical air), a tidal volume of 9 mL/kg, and a respiratory rate of 16 to 20 breaths/min to achieve partial pressure of arterial carbon dioxide (PaCO2) values between 35 and 45 mm Hg and a peak end expiratory pressure of 1 cm H2O. A constant-flow, volume-limited ventilator (model RV5; Voltek Enterprises, Toronto, Ontario, Canada) with a water column for applying peak end expiratory pressure was utilized. The animals were kept at a constant body temperature (mean, 39.1°C ± 0.5°C; range, 38.1°C to 39.8°C) with the aid of a circulating heating blanket and warmed intravenous fluids. An intravenous catheter was inserted into a marginal ear vein for fluid (10 mL/kg/hr of warmed Ringer lactate solution) and drug administration. A subcutaneous injection of buprenorphine (0.02 to 0.05 mg/kg) was administered as an additional analgesic while the animal was under anesthesia. The left carotid artery was cannulated with a 20-gauge angiocatheter for blood sampling and connected to a calibrated pressure transducer to continuously monitor mean arterial pressure. The animals were allowed to stabilize for fifteen minutes after the initiation of mechanical ventilation before baseline measurements of respiratory mechanics were performed and blood samples were taken.
Operative Procedure
The animals were randomly assigned to one of four groups: hemorrhagic shock and resuscitation followed by induction of fat embolism (HR/FE), hemorrhagic shock and resuscitation only (HR), fat embolism only (FE), and control (Fig. 1). Randomization of the animals was accomplished with use of sealed envelopes chosen at the beginning of the experiment each day.
In the nine animals in the HR/FE group, hypovolemic shock was induced by means of constant carotid bleeding into a polypropylene syringe over a twenty-minute period. Approximately one-third of the rabbit's blood volume (60 to 80 mL of blood in total), calculated on the basis of the animal's weight, was removed. Sodium citrate (a 3.8% solution) was added to the shed blood at a ratio of 1:9. The mean arterial pressure was maintained at 30 to 40 mm Hg for one hour. The animals were then resuscitated (over a forty-minute period) with the reinfusion of the entire volume of shed blood titrated with an additional volume of saline solution necessary to restore the baseline mean arterial pressure. Following a one-hour stabilization period, both condyles of the right femur were exposed through a medial parapatellar approach to the knee in preparation for induction of fat embolism. After drilling into the medullary cavity in a retrograde fashion, the canal was successively reamed with use of 3.5, 4, and 4.5-mm-diameter T-handle reamers (Biomet Orthopaedics, Warsaw, Indiana). The medullary canal was pressurized with use of a standardized injection (five to ten seconds in duration) of 1 to 1.5 mL of low-viscosity methylmethacrylate bone cement (Simplex P; Howmedica, Rutherford, New Jersey) in its liquid state. After cement curing, the patella was reduced and the incision was closed.
The experimental procedure used to induce shock in the HR/FE group was used for the six animals in the HR group as well. After a one-hour stabilization period following the resuscitation from the shock (as described above), a sham knee incision was made, exposing both condyles of the right femur. The incision was closed without drilling, reaming, or pressurization.
No hemorrhage or shock was induced in the FE group. The eight animals were ventilated for three hours, and then pressurization of the femoral medullary canal was carried out to create fat embolism. This three-hour ventilation time corresponded to the time it took to establish the shock and resuscitate and stabilize the animals in the HR and HR/FE groups. The femoral canal was then exposed, drilled, and reamed as described above, and a methylmethacrylate bone cement injection was performed. Although the pressure was not measured during cement injection, the tip of the 10-mL syringe fit perfectly inside the canal to prevent cement leakage and thereby allowed consistency during injection.
In the control group, the seven animals were ventilated for three hours and a sham knee incision was made (as in the HR group), exposing both femoral condyles, but it was immediately closed without drilling, reaming, or pressurization.
Blood Sampling
In all four groups, the animals were mechanically ventilated for an additional monitoring period of four hours after surgical closure. Arterial blood samples were collected from the carotid arterial cannula preoperatively, prior to the knee incision, and two and four hours after knee closure. For a complete blood-cell count, blood was collected into ethylenediamine tetra-acetic acid BD Vacutainers (Becton Dickinson, Franklin Lakes, New Jersey) and processed within two hours. For plasma cytokine measurements, blood was collected into 4.5-mL EDTA BD Vacutainers, spun at 2000 rpm for fifteen minutes, aliquoted, and stored at -70°C. For flow cytometric evaluation of neutrophil activation, 0.5 mL of blood was mixed with CTAD (citrate, theophylline, adenosine, and dipyridamole) anticoagulant (Diatube-H; Becton Dickinson) at a final concentration of 1:10 CTAD:blood. Samples were processed within six hours after blood collection. The amount of blood removed for assays was replaced with saline solution at each of the measurements.
Blood gas analysis of arterial blood samples was performed with a 178P8 Blood Gas Analyzer (Corning, Medfield, Massachusetts) at baseline, prior to and five minutes following closure of the incision, and at one-hour intervals thereafter. The partial pressure of arterial oxygen (PaO2) and carbon dioxide (PaCO2) were measured. The mean arterial pressure was recorded at the same standardized time intervals.
Serum lactate levels were measured in the two groups that underwent the shock and resuscitation procedures (the HR and HR/FE groups). Blood was sampled into a 3-mL potassium oxalate sodium fluoride BD Vacutainer, centrifuged at 3700 rpm for eight minutes, and analyzed according to the protocol used in our institution's clinical diagnostic laboratory.
Postmortem Lung Sampling
Following the four-hour monitoring period, the rabbits were killed with an intravenous overdose of pentobarbital. A postmortem thoracotomy was performed, and the pulmonary vessels were ligated. The lungs were removed en bloc, and the left lung was fixed in inflation with 10% buffered formalin at a pressure of 25 cm of fixative for at least forty-eight hours. Following fixation, the samples were sectioned sagittally and three stratified random blocks of known size were taken from the midsagittal slice of each lung. Specifically, three blocks were taken from the left lung of all of the rabbits: one from the upper lobe, one from the middle lobe, and one from the lingula. Approximately 75% of the left lung was utilized for histological analysis. The right lung was used for bronchoalveolar lavage only. The specimens were embedded in paraffin, processed for histological examination, and cut at a 5-µm thickness.
Postmortem bronchoalveolar lavage with 30 mL of saline solution was performed through the right mainstem bronchus on the right lung (repeated three times to better sample alveolar space) for measurement of cytokine levels. Heparin (0.5 mL) was added to the salvaged fluid (approximately 15 mL). The fluid was then centrifuged at 2000 rpm at 4°C for ten minutes to remove the cells. The cell-free supernatant was divided into several aliquots and stored at —70°C until it was assayed.
Cytokine Assays
Plasma and bronchoalveolar lavage fluid were analyzed for monocyte chemotactic peptide-1 (MCP-1) and interleukin-8 (IL-8) concentrations in triplicate and in a blinded fashion with use of the ELISA (enzyme-linked immunosorbent assay) technique and previously described methodology31. MCP-1 and IL-8 were used as markers because they are early-response cytokines that induce leukocyte activation and are associated with the development of acute lung injury, such as acute respiratory distress syndrome. More specifically, MCP-1 is detectable in bronchoalveolar lavage fluid at the onset of acute respiratory distress syndrome, while IL-8 is an early biochemical marker predicting the onset of multiple organ failure.
Flow Cytometry
Flow cytometric methods were used to characterize the activation of neutrophils. Whole blood was stained with conjugated monoclonal antibodies against CD45 (FITC [fluorescein isothiocyanate-labeled bovine serum albumin]) and CD11b (biotin; Research Diagnostics, Flanders, New Jersey) and then a PE (phycoerythrin)-conjugated streptavidin antibody (BD Biosciences). Ten microliters of whole blood was incubated with 50 µL of binding buffer, 5 µL of CD45-FITC (1:2 dilution), and 8 µL of CD11b-biotin (1:2 dilution) for ten minutes. Red blood cells were lysed with the addition of 450 µL of ACK lysing buffer (150 mM NH4Cl, 1 mM KHCO3, and 0.1 mM Na2 ethylenediamine tetra-acetic acid; pH 7.2-7.4) for ten minutes; then, the samples were spun at 200 × g for five minutes and the supernatant was discarded. The pellet was stained with 10 µL of streptavidin-PE (1:50 dilution) for twenty minutes and diluted with 300 µL of phosphate-buffered saline solution. All incubations were performed at room temperature in the dark. Samples were acquired immediately with use of a 15-mW FACSCalibur flow cytometer (Becton Dickinson) equipped with a 488-nm argon ion laser. Twenty thousand white-blood-cell events were collected with a trigger threshold on CD45. Neutrophils were identified and differentiated from other white blood cells by characteristic CD45 fluorescence and light scatter properties. CD11b expression was used because it is an early indicator of the acute inflammatory reaction preceding lung injury. Moreover, CD11b expression on circulating neutrophils increases several hours after major trauma and CD11b expression on monocytes increases as a result of shock following primary hip arthroplasty.
Morphology
Sections of lung tissue from all specimens were stained with hematoxylin and eosin and examined and scored by an experienced pathologist in a blinded fashion32,33. Ten random fields were examined under a 40× objective lens for scoring of all of the parameters described below.
Alveolar hemorrhage was scored as 0 if no evidence of erythrocytes was seen in the alveolar spaces, as 1 if fewer than five erythrocytes were seen in one or two alveolar spaces per high-power field, as 2 if five or fewer erythrocytes were seen in more than two alveolar spaces per high-power field, and as 3 if more than five erythrocytes were seen in more than two alveolar spaces per high-power field.
For scoring of infiltration by polymorphonuclear leukocytes, the numbers of polymorphonuclear leukocytes with in the alveolar walls and within the alveolar spaces were averaged to determine the most severely affected alveoli within each of the ten high-power fields. The average rounded number was considered to be the score for each specimen.
The score for alveolar hyaline membrane formation was 0 if no membranes were found, 1 if one membrane was found, 2 if two membranes were found, and 3 if more than two membranes were found.
The score for alveolar edema was 0 if there was no evidence of edema—i.e., no amorphous proteinaceous coagulum or eosinophilic-staining fluid within the alveolar spaces; 1 if only two alveolar spaces contained this coagulum, within all ten fields; 2 if more than two but less than six alveolar spaces contained evidence of edema; and as 3 if six or more alveolar spaces contained evidence of edema.
Statistical Analysis
We present a new experimental model that has several important outcomes, each of which might be considered a primary outcome. Consequently, there is a risk of associations occurring by chance as a result of multiple statistical comparisons. Therefore, the following approach was used to analyze the data. Data were reported as the mean and standard error of the mean and analyzed with use of the SPSS software package (SPSS, Chicago, Illinois). For sequential continuous-variable measurements, repeated-measures two-way analysis of variance with one within-subjects factor (time) and one between-subjects factor (group) was conducted with a significance level of 0.05. When time was identified as a significant factor, a repeated-measures one-way analysis of variance was performed for each of the four groups with the default level of significance being 0.05. If one-way analysis of variance showed significance, then a series of paired-samples t tests was performed to determine which time measurement was different from baseline. For multiple comparisons, a Bonferroni-adjusted level of significance was used at level 0.05/number of t tests. When a group was identified as a significant factor, a series of one-way analyses of variance was conducted with a Fisher least-significant-difference post hoc test for each time point to determine which of the three experimental groups was different from the control group. A default level of significance of 0.05 was used. The data derived with histological analysis are reported as the mean score and the range. A Kruskal-Wallis test was used to compare differences between groups. When p was <0.05, the Mann-Whitney U test was used to compare differences between individual groups. A one-way analysis of variance was used to measure differences in cytokine levels between groups.
Source of Funding
No authors received financial benefit from this study. No external funding was obtained.
We used a newly developed rabbit model of fat embolism34 to study the effects of fat-induced lung injury in the setting of acute hemorrhagic shock and resuscitation. Respiratory failure is a frequent complication after polytrauma and may occur as a result of either direct or indirect injury to the lung. Cofactors such as proinflammatory and coagulation mediators and the presence of other injuries and hemodynamic instability, in addition to long-bone fractures, have been shown to potentiate the development of lung injury1,6-10,19-23,35,36.
The animals in the FE group displayed a familiar physiological response to fat embolism37 (Figs. 4, 5, and 6). They exhibited a significant decrease in mean arterial pressure immediately following induction of fat embolization, and the pressures remained below baseline values for the remainder of the experimental period. The PaO2 level in the FE group decreased significantly as well. Although the animals in the HR/FE group showed a similar trend of physiological response to fat embolism, the changes were more transient. Following the one-hour shock period, the shed blood was reinfused and supplemented with an additional volume of saline solution in order to reach baseline values of mean arterial pressure. It is possible that the added volume of saline solution compensated to maintain cardiac output following the induction of the fat embolism. Conversely, in the HR/FE group, three animals died following canal pressurization as a result of severe systemic hypotension and eventual cardiac arrest, and this is the only group in which animals died. It is postulated that preexisting damage caused by the shock and resuscitation compromised the ability of the animals to respond to the fat embolism. The volume of bone marrow fat embolus released from the medullary canal of the long bone is uncontrolled, and the animal's cardiopulmonary reserve or ability to respond varies; this could also account for differences between individual responses. Since shock alone has the potential to produce greater injury to lung tissue than fat embolism alone (Figs. 7, 8, and 9), we postulate that a period of initial shock may "prime" the lung for subsequent fat embolism damage, which may be greater than the lung injury incurred from an isolated episode of fat embolism. The duration and intensity of shock required to produce such a priming effect may be the subject of future work.
The present histological findings support previous observations that fat embolism by itself does not cause lung injury38. However, the score for infiltration of neutrophils into the alveolar spaces was higher in the HR/FE group than it was in the controls. Although the difference was significant, the number of counted neutrophils was still small in the HR/FE group; this could be attributed to the relatively short duration of the experiment. This finding may be clinically important as the extent of neutrophil influx and the presence of neutrophil products in alveolar lavage fluid have been correlated with the severity of lung injury39. The results of our white-blood-cell flow cytometric studies complement the histological findings, with the CD11b mean channel fluorescence measurements being significantly higher than the baseline and control values only in the HR/FE group at two and four hours after canal pressurization.
Neutrophils have been implicated as having a pivotal role in the development of acute lung injury. Although the neutrophil percentage increased in all groups with time in this study, CD11b mean channel fluorescence was significantly elevated only in the HR/FE group after knee closure. The binding of neutrophils to endothelial cells and their subsequent transmigration to the site of inflammation are essential in a multistep process. Transmigration of leukocytes is dependent on the expression of cell-surface glycoprotein adhesion molecules. Molecular markers, such as CD11b expression, are used as early indicators of the acute inflammatory reaction preceding lung injury. CD11b expression on circulating neutrophils has been shown to increase six and twelve hours after major trauma40, and CD11b expression on monocytes has been shown to increase after surgical trauma caused by primary hip arthroplasty41.
Cytokines have an important role to play as a marker for the inflammatory response. Consistent with the CD11b-expression results, the cytokine levels in the bronchoalveolar lavage fluid in the present study were significantly higher than the control values only in the HR/FE group (Figs. 8 and 9). Early-response cytokines such as tumor necrosis factor-a(TNF-a)42 and chemokines such as IL-8 and MCP-1, which induce leukocyte activation, are associated with the development of acute lung injury such as acute respiratory distress syndrome43. MCP-1, which is known to regulate monocyte recruitment, is detectable in bronchoalveolar lavage fluid at the onset of acute respiratory distress syndrome and persists in the lungs of patients with sustained acute respiratory distress syndrome43. IL-8 may also be an early biochemical marker predicting the onset of multiple organ failure44. Therefore, since levels of chemokines, neutrophil activation, and alveolar infiltration were significantly higher only in the HR/FE group, an amplified response has been detected. The above findings are consistent with those in several other studies45,46.
Factors that potentially lead to an inflammatory response and fat embolism syndrome have also been examined in several previous studies. Giannoudis et al.12 explored the effect of femoral reaming and nailing on the inflammatory progression in a prospective clinical study. This may, in part, contribute to the pathogenesis of lung and multiple organ injury revealed by increased levels of serum IL-6 (a proinflammatory cytokine) and plasma elastase (neutrophil protease) in response to intramedullary nailing. They found CD11b levels to be elevated on admission. Only one patient died of acute respiratory distress syndrome; interestingly, the inflammatory response in that patient, as indicated by IL-6 and elastase production, was hyperstimulated after the traumatic injury and again after nailing, suggesting that the excessive inflammatory reaction was caused by femoral pressurization in the pathogenesis of posttraumatic respiratory distress.
The role of leukocyte activation was emphasized by Aoki et al.47. Fifteen patients with a long-bone fracture who underwent intramedullary nailing with no complications were compared with five patients diagnosed with fat embolism syndrome. The patients with a fracture in whom echogenic material in the right ventricle was noted with transesophageal echocardiography were compared with the patients with fat embolism syndrome. Although the percentage of lipid-laden cells in the bronchoalveolar lavage fluid did not differ significantly between the groups, the bronchoalveolar lavage fluid from the patients with fat embolism syndrome had a higher leukocytic count and albumin concentration. The authors concluded that a factor other than mechanical pulmonary obstruction by fat globules may be necessary for the development of fat embolism syndrome.
We used a rabbit model rather than a canine model for several reasons. First, although a canine model is ideal for examining mechanical and hemodynamic aspects of fat embolism, it is constrained by the inability to further assess molecular and cellular causes of inflammation because of limited accessibility to markers. The rabbit model permitted an understanding of the basic molecular and cellular pathogenic processes of lung injury to improve therapeutic strategies for fat embolism syndrome. Second, the rabbit model allowed consistent creation of sufficient emboli. Third, the rabbit model allowed for future exploration of new research directions—namely, a focus on the various molecular signaling pathways of the inflammatory response, which are currently difficult to study in larger animals.
There are several limitations to the present study. First, the results of an assessment of the inflammatory response to hemorrhage, resuscitation, and fat embolism in animals may not be completely representative of the situation in human subjects. Second, the limited number of specimens in each test group due to current budgetary constraints and logistics warrants a future study involving a larger number of samples to eliminate any possible Type-II error. Third, the short period of follow-up may not have been adequate for us to detect the full effects of fat embolism or the progression to fat embolism syndrome, which may take some time to develop48,49. It is possible that more changes would have occurred with a prolonged follow-up time. Fourth, in a clinical trauma situation, the pressurization with cement would not be done in an intact femur. A future study should perhaps be done with use of a fracture model in which fat embolism is created by, for example, stabilizing the femur with reaming and a Kirschner wire. Although the induction of fat embolism in this manner is probably less reproducible from specimen to specimen, it may be more representative of the clinical situation because fat embolism develops after fracture nailing in some patients and not in others. However, the present study is an initial step in understanding and assessing the effects of pressurization of the medullary canal. Finally, although the current model does not directly simulate the events surrounding polytrauma, such as femoral nailing of a vented fracture or hip arthroplasty with a non-vented femoral canal, the model was previously found to have generated pathophysiological and histomorphometric findings consistent with fat embolism as would occur in these clinical situations34.
In conclusion, we have proposed that the "two-hit" combination of shock and fat embolism leads to greater pulmonary dysfunction. Fat embolism due to medullary canal pressurization alone caused acute pulmonary hypertension and decreases in mean arterial pressure and systemic arterial oxygen tension but did not activate an inflammatory response in the lung. In the HR/FE group, most physiological parameters measured displayed changes consistent with those shown in the FE group alone. Hemorrhage and resuscitation did not exacerbate the immediate pulmonary gas-exchange abnormalities. The HR/FE group did demonstrate several significantly amplified responses—i.e., greater cytokine expression, neutrophil activation, and alveolar neutrophilic infiltration. Future prospective studies may determine if these findings can be used as early indicators of an augmented inflammatory response, which may play a role in the development of fat embolism syndrome associated with trauma.