Animal Model
Investigations were carried out in six skeletally mature mixed-bred ewes (five to six years old with a mean body weight [and standard deviation] of 70 ± 10 kg), which were subjected to unilateral augmentation of three lumbar vertebral bodies (L2-L4) with polymethylmethacrylate. The study was approved by the State Animal Ethics Committee and was conducted according to federal and state guidelines.
Instrumentation of Animals
Anesthesia was induced with propofol (6 mg/kg) and was maintained with isoflurane (2% to 3%) in oxygen (50%). Analgesia and muscle relaxation were achieved by the administration of buprenorphine (0.005 mg/kg) and pancuronium (0.06 mg/kg), respectively. The lungs were ventilated mechanically to maintain physiologic end-tidal carbon dioxide tension prior to cement injection. End-tidal carbon dioxide tension was measured with an infrared capnometer. Ventilation parameters were not adjusted after cement injection. Lactated Ringer solution was infused through the left cephalic vein at 4 mL/kg/h. Electrocardiographic monitoring was obtained with skin electrodes.
For cardiovascular instrumentation, animals were placed in the supine recumbent position. An angiography catheter was inserted into the left carotid artery and was advanced into the left ventricle to measure blood pressure. The right carotid artery was cannulated to measure arterial blood pressure and to obtain blood samples. A Swan-Ganz thermodilution pulmonary artery catheter was inserted into the right jugular vein and was floated into the pulmonary artery to measure central venous pressure, pulmonary arterial pressure, and cardiac output and to obtain mixed venous blood samples. The correct position of the ventricular and pulmonary catheters was confirmed by recording typical pressure waves. Catheters were connected to pressure transducers (Uniflow; Baxter, Volketswil, Switzerland) by means of pressure tubing filled with liquid (Ringer lactate solution). Catheters were flushed with heparinized Ringer lactate solution (5000 IU/L) after insertion and after taking blood samples. Injected volumes were approximately 15 mL/h and deemed not to affect hemostasis26. Heart rate was derived from the electrocardiogram. Cardiovascular pressures and the electrocardiographic recordings were digitized at 1 Hz with use of an analog-digital converter (Hellige Messturm; Marquette Hellige Medizintechnik, Freiburg, Germany) and stored on a computer for offline analysis.
Surgical Procedure and Cement Injections
Animals were placed in the right lateral recumbent position on the operating table. A retroperitoneal approach was used to expose the lateral aspect of three lumbar vertebral bodies (L2-L4). A cement injection hole (3.5 mm in diameter) was drilled into the cephalad aspect of each vertebral body to a depth of 10.0 mm. The proximal part of the injection hole was carefully widened to a diameter of approximately 4.0 mm, so that the tip of a 3-mL syringe would fit tightly into it.
Polymethylmethacrylate bone cement formulated for vertebroplasty (low viscosity, 30% barium) (Mendec; Tecres Medical, Verona, Italy) was used for injection. Powder and chilled liquid (5°C) were mixed in an open bowl for thirty seconds and then drawn into 3-mL polycarbonate syringes (Medicor, Cham, Switzerland). The cement was left to polymerize at room temperature (21° to 23°C; 27% to 34% humidity) until an appropriate viscosity for injection was reached. A volume of 6.0 mL of cement was injected over thirty seconds until filling of the vertebral body was achieved.
Experimental Protocol
Pressure transducers were zeroed at the level of the heart. Blood pressures and electrocardiographic activity were continuously recorded until sixty minutes after beginning the last (i.e., third) cement injection. Cardiac output and blood gas parameters were measured before and at least twenty minutes after completion of the surgical approach. Cardiac output was measured three times and averaged. Arterial and mixed venous blood samples were drawn for blood gas analysis, which was carried out immediately. Cement was injected into three lumbar vertebrae (L2-L4) after a twenty-minute time-interval between injections. The duration of the time-intervals was sufficient for stabilization of cardiovascular parameters to a new steady state10,11. Postinjection evaluation of cardiac output and blood gas parameters was carried out one minute and ten minutes after having started the first injection; ten minutes after having started the second injection; and ten minutes, thirty minutes, and sixty minutes after having started the third injection. At the end of the protocol, the animals were killed by intravenous injection of pentobarbital (1 g/sheep) and potassium chloride (2 mmol/kg).
Analysis of Cardiovascular Data
In order to obtain preinjection values of continuously recorded parameters (i.e., blood pressures and heart rate), data were averaged over five minutes. For postinjection values, data were averaged over twenty seconds. Cardiac index, pulmonary vascular resistance index, systemic vascular resistance index, physiologic dead space, and intrapulmonary shunt were calculated with use of standard formulas.
Measurements of Coagulation Parameters
Citrated blood samples were taken prior to induction of anesthesia (jugular vein) and prior to the first cement injection (i.e., after instrumentation and surgical approach) (carotid artery). Further samples (carotid artery) were taken one minute and ten minutes after having started the first injection; ten minutes after having started the second injection; and ten minutes, thirty minutes, and sixty minutes after having started the third injection. Blood samples were centrifuged immediately after sampling, and the plasma was stored at —70°C for batch analysis. The following parameters were measured: thrombocyte count, prothrombin time, partial thromboplastin time, fibrinogen, D-dimer, and thrombin-antithrombin complexes. Analyses were performed in two certified laboratories with use of standard protocols. A given parameter was always measured in the same laboratory.
Histopathology
After the animals were killed, two lung tissue samples were taken from two predetermined areas of each of the five lung lobes (right cranial, middle, and caudal lobes and left cranial and caudal lobes) and were fixed in 10% neutral buffered formalin. Specimens were stained with hematoxylin and eosin and oil red O (fat stain). Two microscopic views (at five times magnification) were analyzed from each sample for the presence of intravascular fat and bone marrow cells. Semiquantitative analysis of intravascular fat was performed by counting the number of emboli in each photomicrograph. Counts were averaged and presented as a histopathologic score.
Statistical Analysis
According to data from previous studies10,20, the chosen sample size was sufficient to detect clinically relevant changes of blood pressure and blood gas variables from preinjection and postinjection values with the power ranging from 88% to 97%. Data were calculated and given as the mean and the standard deviation of the mean. One-way analysis of variance for repeated measures was used to test for significant differences. Post hoc analyses were performed with use of the Bonferroni test. The paired Student t test was used to test for differences in histopathologic scores between the different lung lobes. A p value of =0.05 was considered significant for all statistical analyses. Statistical analyses were performed with use of Statistica 7 software (StatSoft, Hamburg, Germany).
Cement Injections
On the average, 4.7 ± 0.7 mL of polymethylmethacrylate was injected into each vertebral body over an average of 31 ± 11 sec. One injection was judged inadequate because insufficient cement volume had been injected (i.e., <3 mL). In that case, the cement viscosity had been too high to inject more volume. Cardiovascular data from this injection were excluded from analysis. At the postmortem examination, cement leakage into epidural and paravertebral veins was discovered in two animals.
Cardiovascular Parameters
There were no significant (p = 0.71) differences among the cardiovascular changes after the three cement injections and among the three preinjection values (p = 0.98). Injections of polymethylmethacrylate elicited a consistent cardiovascular response after each injection. Thus, the data of the continuously recorded parameters (i.e., invasive pressures and heart rate) from the three embolization events were pooled, and statistical analysis was performed.
Injection of polymethylmethacrylate elicited a significant (p < 0.0001) increase in the mean pulmonary vascular resistance index (150% ± 53%) and the mean pulmonary arterial pressure (108% ± 32%) (Tables I and II). Consequently, there was a significant (p < 0.004) decrease in the mean arterial blood pressure (36% ± 16%) and a significant (p < 0.004) increase in the mean central venous blood pressure (54% ± 29%). Values of mean arterial blood pressure and central venous blood pressure were no longer significantly (p > 0.9) different from preinjection values two minutes after inducing fat embolization. The mean pulmonary arterial pressure remained elevated until nine minutes after embolization (Fig. 1).
Coagulation Parameters
There was a significant (p < 0.05) decrease in thrombocyte count and a significant increase in the concentration of thrombin-antithrombin complexes from the preanesthetic to preinjection values (Table III). The mean plasma concentrations of thrombin-antithrombin complexes increased from the value before the injection (3.01 ± 0.75 µg/L) to one minute after the injection (4.26 ± 1.29 µg/L), and the mean prothrombin time decreased from the value before the injection (74% ± 9%) to sixty minutes after the injection (68% ± 7%). However, these changes were not significant (p > 0.3). There were no other changes in coagulation parameters from preinjection to postinjection values.
Histopathology
Intravascular fat and bone marrow cells were present in all lung lobes (Fig. 2). The mean histopathologic score for all five lobes (thirty specimens) was 4.9 ± 3.7. There were no significant (p = 0.16) differences in the histopathologic scores among the different lung lobes. Traces of intravascular bone cement (i.e., barium particles) were detected in two lung lobes. Cardiovascular responses recorded in these animals were not significantly (p = 0.9) different from responses in animals without pulmonary cement embolism.
The cardiovascular response to bone marrow fat embolism after intravertebral injection of polymethylmethacrylate was characterized by a sudden (one minute postinjection) and dramatic (>100%) increase in mean pulmonary arterial pressure and a decrease in mean arterial blood pressure (36%). There were no significant changes in any coagulation parameter from preinjection to postinjection values.
In the present study, fat embolization during vertebroplasty in sheep with use of polymethylmethacrylate did not affect any of the measured coagulation parameters. In contrast, Barie and Malik27 observed a decrease in fibrinogen and an increase in fibrin degradation products after an intravenous injection of allogenic bone marrow suspension in sheep. Defibrinogenation completely prevented any cardiovascular changes. Fibrinogen depletion prevented thrombogenesis and therefore the blockage of the pulmonary vasculature. The process of harvesting bone marrow may have activated the coagulation cascade in their study. However, the present model mimics the clinical situation more closely.
Thrombogenesis has also been observed after arthroplasty and long-bone intramedullary reaming. These interventions cause more vascular and tissue damage compared with the injection of polymethylmethacrylate into vertebral bodies. An increase in thrombin-antithrombin complex concentration and changes in antithrombin and fibrinogen values were observed after reaming of the femoral medullary cavity for hip arthroplasty24 and intramedullary nailing28, respectively. Modig et al.22 investigated the role of coagulation, fat embolism, and methylmethacrylate in the development of arterial hypotension and hypoxemia during total hip arthroplasty with cement. Activation of the coagulation cascade was observed immediately after the insertion of the femoral component. The recorded cardiovascular changes were correlated with the degree of coagulation activation, but not with the severity of fat embolization or the plasma concentrations of methylmethacrylate. More recently, coagulation activation during total hip arthroplasty with and without cement was demonstrated by an increase in thrombin-antithrombin complex and D-dimer values23.
It has been suggested that coagulation may be activated by cell fragments released from the bone marrow cavity22 or thromboplastin (protease, which converts prothrombin to thrombin)22,29. Thromboplastin may be released from adipose bone marrow tissue29 or as a result of subendothelial tissue damage in the bone marrow cavity28 or in the lungs27. Furthermore, polymethylmethacrylate may have inherent thrombogenic properties24,25. Serious concerns have been raised with regard to the use of calcium phosphate cement for vertebroplasty as it may aggravate cardiovascular complications by stimulating coagulation30. Calcium phosphate cement may provide a scaffold for clot formation or may initiate the coagulation cascade by surface contact activation and providing calcium ions, an important cofactor for coagulation. However, more recent studies have found no coagulation activation after both polymethylmethacrylate and calcium phosphate cement came in contact with circulating blood in vitro31 and in vivo32,33.
Quantification of thrombin-antithrombin complexes allows one to detect subtle degrees of coagulation activation. The key event of coagulation activation is the conversion of prothrombin to thrombin. However, only a small amount of circulating prothrombin (<1%) is activated to thrombin, and thrombin is rapidly neutralized by antithrombin, resulting in an increase in circulating thrombin-antithrombin complexes. In the present study, there was a significant increase in the concentration of thrombin-antithrombin complexes and a significant decrease in the thrombocyte count from the preanesthetic to preinjection values. This may have been the result of the instrumentation and the surgical approach to the lumbar spine. There was no significant change in thrombin-antithrombin complex or any other coagulation parameter from preinjection to postinjection values. Recording was terminated sixty minutes after the last embolization event, and coagulopathy may occur later on as a result of lung injury or blood flow disturbances34. However, the focus of the present study was on the intraoperative cardiovascular changes after bone marrow fat embolism.
Activation of the coagulation cascade generating additional emboli did not contribute to the acute cardiovascular changes after fat embolism elicited by injections of polymethylmethacrylate into vertebral bodies. However, blockage of >50% of the pulmonary arterial vasculature would be required to elicit cardiovascular changes of the magnitude recorded in the present study35,36. Blockage of this magnitude is improbable according to the histopathologic results and data reported in the literature10,29,37, and thus mechanical blockage would not seem to be responsible for the recorded cardiovascular changes in the present study. The increase in pulmonary arterial pressure may have been caused by pulmonary vasoconstriction elicited by vasoactive mediators released from the bone marrow cavity or as a result of lung injury after embolization38,39. Alternatively, a reflex response to embolization may have elicited pulmonary vasoconstriction40,41. Methylmethacrylate has also been reported to cause pulmonary vasoconstriction42,43. However, concentrations of methylmethacrylate required to elicit cardiovascular changes are more than twice as high as concentrations measured clinically44,45.
It has been suggested that kyphoplasty may carry a lower risk of fat embolism compared with vertebroplasty. Kyphoplasty is only performed in vertebral bodies with acute (i.e., mobile) fractures, and fracture lines may offer a way of pressure release and escape for the fat emboli during inflation of the balloon. Potentially beneficial effects are therefore a result of the different indications (only acute compared with acute and old fractures) and are not inherent to the type of intervention (i.e., kyphoplasty compared with vertebroplasty). Vertebroplasty of unconsolidated or highly compressed vertebral bodies also carries a low risk of fat embolism. On the other hand, augmentation of intact osteoporotic vertebral bodies at risk of fracture may carry a high risk of fat embolism.
Limitations of the present animal model have been discussed previously10,11. Briefly, vertebral filling was higher compared with the clinical situation. However, it was crucial to inject similar volumes of cement compared with the clinical situation, displacing similar volumes of bone marrow fat. The aim of the present study was to investigate the pathophysiology of cardiovascular changes after bone marrow fat embolism, rather than replicating clinical vertebroplasty. Flushing catheters with heparinized Ringer solution may have affected measured coagulation parameters. The prothrombin time decreased by 8% from before anesthesia to before injection and again by 7% and 8% from before injection to thirty and sixty minutes after injection, respectively. There was a 15% decrease from before anesthesia to sixty minutes after injection. However, these changes were not significant, and, moreover, there were no changes in partial thromboplastin time. Therefore, injected volumes of heparin appeared to be insufficient to affect measured coagulation parameters.
In conclusion, injection of polymethylmethacrylate into vertebral bodies elicited embolization of bone marrow fat with subsequent transient cardiovascular deterioration. Thromboembolism did not contribute to the observed cardiovascular changes. Cardiovascular complications as a result of bone marrow fat embolism should be considered in patients undergoing vertebroplasty. 
Note: The authors thank Boris Leskosek, Katja Nuss, and Alush Avdyli for their help with the animal experiments. They thank Trinh Cung for performing the thrombin-antithrombin complex measurements.