Tumor-Derived Tissue FactorBearing Microparticles Are Associated With Venous Thromboembolic Events in Malignancy

Thromboembolic disease is a recognized complication of cancer (1,8). Patients with cancer experience venous thromboembolic events (VTE) at a rate four to seven times higher than that of the general population (9). Although all tumor types are associated with thrombosis, glioblastoma and carcinomas of the ovary and pancreas are consistently associated with the highest incidence of VTE (10, 11). Despite the strong association between malignant disease and thromboembolic disorders, the molecular and cellular basis of this relationship remains uncertain. Hypotheses regarding the mechanism have included the activation of blood coagulation by tissue factor in tumors (12), a factor Xactivating cysteine protease (13), mucinous glycoproteins (8), and MET oncogene activation (14). Laboratory markers of coagulation activation are often elevated in patients with cancer (15,21) but are of little clinical utility in assessing individual risk for thrombosis.

Tissue factor, a receptor protein constitutively expressed on the plasma membrane of most nonvascular cells and inducibly expressed on monocytes and endothelial cells, initiates blood coagulation in vivo (22). In addition to its expression in the vessel wall, tissue factor is associated with cell-derived vesicles in blood that bind to the developing thrombus and play a role in fibrin propagation (23, 24). These vesicles, known as microparticles based on their submicron diameter, express protein surface components from the parent cell from which they are derived. Monocyte-derived microparticles that express tissue factor accumulate in the developing thrombus through the interaction of P-selectin and PSGL-1, thus delivering tissue factor to the site of vascular injury and facilitating fibrin propagation (24). Microparticles derived from blood and vascular cells circulate in the blood of normal subjects, and these populations are altered qualitatively and quantitatively in specific disease states (25,28). Tissue factorbearing microparticles have been observed in the plasma of cancer patients, and a possible role for these microparticles in cancer-associated thrombosis has been proposed (29, 30). To evaluate whether tissue factorbearing microparticles are a risk factor for and a possible cause of cancer-associated thrombosis, we quantified tissue factorbearing microparticles in cancer patients using impedance-based flow cytometry and analyzed their association with acute venous thromboembolic disease. We show that impedance-based flow cytometry allows identification and quantitation of tissue factorbearing microparticles, that most of these tissue factorbearing microparticles are tumor-derived, and that the presence of detectable tissue factorbearing microparticles is associated with an increased risk of thromboembolic disease accompanying malignant disease.

Humanized monoclonal antibody cH36 against human tissue factor was the generous gift of Dr. Hing Wong (Altor Biosciences, Miramar FL). Antibodies against MUC1 were generously provided by Dr. Donald Kufe (Dana Farber Cancer Institute, Boston MA); these antibodies were labeled with Alexa 488 using standard methods. ASPC-1, a human pancreatic carcinoma cell line, was obtained from the American Type Culture Collection. Goat anti-human Ig coupled to Alexa 488 was obtained from Invitrogen, anti-CD41a-FITC antibody and phycoerythrin-conjugated mouse anti-human monoclonal TF antibody from BD Biosciences, and all other immunochemical reagents from Sigma-Aldrich. Fluorescently labeled polystyrene microspheres were purchased from Bangs Laboratories. These microspheres are homogeneous with regard to size by 10.

ASPC-1 cells were incubated at 37C in RPMI media for 72 h. The cells were removed from conditioned media by centrifugation twice at 2,100 g. Aliquots (10 L) of supernatant were labeled with (0.25 g) humanized antitissue factor monoclonal antibody cH36 and labeled with a goat anti-human IgG-Alexa 488 antibody.

Platelet-rich plasma was centrifuged for 2 min at 2,100 g. A 40 L aliquot of plasma, containing both platelets and platelet microparticles, was incubated with 0.125 g of anti-CD41a-FITC antibody for 30 min. The plasma samples were diluted 1:50 in NPE Iso-Diluent with 0.78-m polystyrene microspheres at a final concentration of 6,000/L. Samples were analyzed on a Beckman Coulter Quanta flow cytometer equipped with a 40-m flow cell, a Becton Dickinson LSRII flow cytometer, and a Becton Dickinson FACSCalibur flow cytometer.

A Quanta (NPE Systems) or Beckman Coulter SC Quanta flow cytometer (Beckman Coulter) were used to measure particle size by impedance. Both flow cytometry instruments required postmarket installation of smaller flow cells (25 m or 40 m, respectively) to analyze particles between 0.3 and 1.0 m in diameter. A 0.5 g aliquot of humanized monoclonal antibody cH36 against human tissue factor or a purified human IgG antibody control was added to platelet-poor plasma, and the sample was incubated for 30 min at room temperature. Subsequently, a secondary goat anti-human IgG coupled to Alexa 488 was added, and the sample was incubated for 30 min at room temperature. An aliquot of the antibody-treated platelet-poor plasma (10 L) was diluted 1:100 into filtered NPE Iso-Diluent (Beckman-Coulter) and assayed in triplicate in the Quanta flow cytometer. Microparticle concentration, particle size, and particle fluorescence were recorded. Fluorescent 0.78-m microspheres were used for instrument calibration. The cell-free supernatant of ASPC-1 pancreatic cells containing tissue factorbearing microparticles served as positive controls. The lower limit of detection of tissue factorbearing microparticles at these dilutions was estimated to be 1 10⁴ microparticles per L of plasma. Dual label experiments were done using 5 L of platelet-poor plasma and 0.05 g phycoerythrin-conjugated mouse anti-human monoclonal tissue factor antibody (BD Biosciences) and 0.5 g mouse anti-human MUC-1 monoclonal antibody labeled with Alexa 488 (Invitrogen). Isotype-matched mouse IgG controls were similarly labeled.

The plasma of cancer-free subjects and subjects with surgically unresectable or metastatic cancer were assayed for tissue factorbearing microparticles. Malignancy diagnosis required histologic documentation as well as radiographic evidence of disease. Cancer diagnoses included pancreatic carcinoma, nonsmall cell lung cancer, ovarian carcinoma, colorectal cancer, and breast cancer. Cancer-free controls were cancer-free by history, without a history of VTE, and could not be taking warfarin for any indication.

Patients with an active malignancy who presented with symptomatic proximal deep venous thrombosis or pulmonary embolism were enrolled. A qualifying VTE required radiographic evidence of proximal deep vein thrombosis by compression ultrasound and/or a pulmonary embolism by high probability ventilation/perfusion lung scan or computed tomography angiogram of the pulmonary arteries. Patients who experienced a thrombotic event within 30 d of a surgical procedure or following trauma were excluded.

The control population was a cohort of cancer patients without evidence of venous thromboembolic disease in the previous 5 y, matched for cancer diagnosis, cancer stage at time of blood draw (equal stage or higher), age (within 10 y), and sex, with each cancer patient in the acute VTE cohort. Two control patients were enrolled for each cancer patient with acute venous thromboembolic disease. This provided a control group with a similar distribution of diagnoses, age, and sex. Radiotherapy or chemotherapy administered within 30 d prior to enrollment was documented. For patients in the cancerno VTE group, a retrospective analysis was done to evaluate the incidence of VTE; all radiographic reports for the cancerno VTE group were analyzed by a reviewer blinded to microparticle status. Only documented evidence of a new proximal extremity deep vein thrombosis or pulmonary embolism was included in the analysis. A third comparator group included patients who presented with an acute idiopathic symptomatic VTE. An idiopathic event was defined as not occurring within 30 d of surgery or trauma and no history of malignancy within the last 5 y. Patients with hereditary thrombophilias were not specifically excluded.

All venipunctures were done within 72 h of initial diagnosis of acute VTE. Blood anticoagulated in citrate buffer was centrifuged at 2,100 g for 20 min twice to obtain platelet-poor plasma and stored at 80C. All study protocols were approved by the Human Investigation Review Board at Beth Israel Deaconess Medical Center and the University of Southern California prior to the enrollment of study subjects. All patients signed informed consent prior to laboratory testing and personal health information was protected in accordance with Health Insurance Portability and Accountability Act guidelines.

Based on our preliminary data that tissue factorbearing microparticles were measurable in patients with prothrombotic malignancies (31), the study was powered to detect a difference of 40 percentage points between the percentage of individuals in the cancer-VTE group versus the cancerno VTE group who had measurable tissue factorbearing microparticles. We anticipated enrollment of a minimum of 21 patients with canceracute VTE with a 2:1 matched enrollment of no VTEcancer controls (n = 42) to achieve a statistical power of 80 with a two-sided level of 0.05. Matching was implemented to assure similar age, sex, and cancer diagnosis distribution in the VTE and no-VTE groups; there was no plan to do a matched analysis. We also anticipated comparing the cancer-VTE group with an idiopathic VTE group, with planned enrollment of 21 patients with acute idiopathic VTE. To account for this second comparison involving the cancer-VTE group, significance by Fisher's exact testing was defined a priori as P 0.025.

The determination of significance between groups based on the presence or absence of detectable tissue factorbearing microparticles was done using Fisher's exact analysis. The odds ratio with 95 confidence intervals for thrombotic risk attributable to tissue factorbearing microparticles was calculated by logistic regression; stepwise logistic regression was used to identify factors that significantly contributed to thromboembolic risk in addition to tissue factorbearing microparticles. Absolute differences in tissue factorbearing microparticles were analyzed by the Wilcoxon rank-sum test as the median values for the cancerno VTE control and idiopathic VTE groups fell within the undetectable range. The incidence of VTE among cancer patients without VTE at the time of blood sampling was estimated by the method of cumulative incidence, identifying death without VTE as a competing risk; only patients alive without VTE were censored. Differences in the time to VTE using the cumulative incidence method were assessed by the Gray test.

To investigate whether tissue factorbearing microparticles are associated with VTE in cancer patients, we first evaluated impedance-based flow cytometry as an accurate method for microparticle detection and subsequently carried out a series of clinical investigations to establish a link between elevated tissue factorbearing microparticles and cancer-associated thrombosis as outlined in Supplementary Fig. S1 (supplementary data).

We previously showed in a mouse model of thrombosis that tissue factorbearing microparticles accumulate in the developing thrombus in an interaction that is mediated by platelet P-selectin and microparticle PSGL-1 (24). However, our efforts to detect, quantitate, and size these tissue factorbearing microparticles by light scattering using flow cytometry were limited by conventional methods. When particles are of the same order of magnitude as the incident wavelength, the angular distribution of forward light scatter is not dependent on particle size (32, 33). Moreover, the sizing of particles by light scatter is influenced by the refractive index, internal structure, and presence of absorptive material on the particle surface. The absolute sizing of cells or particles by flow cytometry is typically done by referencing the light scatter characteristics of a uniform population of sized beads, ignoring the fact that the refractive index of polystyrene microspheres is much greater than cellular membranes (33). These issues have been recognized by others, but for lack of better alternatives, light scatteringbased flow cytometry has remained a standard method for microparticle detection and measurement.

For these reasons, we have explored a novel method for identifying, sizing, and quantitating microparticles utilizing impedance-based flow cytometry. This system, based on the Coulter principle, determines the electronic volume of a particle based on the fact that the electronic volume is proportional to the change in the impedance associated with the displacement of electrolyte in a flow cell by the particle of interest. The system is calibrated using fluorescent polystyrene microspheres of uniform size. In the example shown, polystyrene microspheres 520 nm in diameter were monitored by impedance-based flow cytometry. A dominant population with a diameter of 520 nm is observed in the histogram (Fig. 1A) as are smaller populations of microsphere aggregates. A plot of fluorescence against particle size illustrates the size distribution of fluorescently labeled microspheres (Fig. 1B). Using polystyrene microspheres of known concentration, analyses of solutions containing these microspheres determined concentration within 5 of the theoretical concentration value. The mean coefficient of variation for microspheres with a diameter of 0.78 m included an electronic volume and fluorescence of 25.2 and 9.2, respectively (n = 15). Populations of polystyrene microspheres of two sizes can be resolved from each other based on their diameter, and independently quantitated (Fig. 1C).

To compare microsphere and microparticle sizing, we measured platelet microparticles, calibrated microspheres, and platelets in human plasma both by impedance- and light scatterbased flow cytometry. When characterized by forward light scatter, there is poor size discrimination between platelets and platelet microparticles (Fig. 2A and B). Furthermore, the platelet population and the platelet microparticle population overlap with the 0.78-m microspheres. Using impedance-based flow cytometry, the dot plot of the diameter versus fluorescence showed a platelet microparticle population measuring less than the 0.78-m microspheres that is completely resolved from the population of platelets (Fig. 2C).

Dvorak et al. has previously shown that procoagulant microvesicles are released from malignant cells in culture (34). Based on this observation, we examined conditioned media from ASPC-1 pancreatic cells in culture for the presence of tissue factorbearing microparticles. Using a monoclonal antibody specific for human tissue factor and a FITC-labeled secondary antibody directed against the antitissue factor antibody, tissue factorbearing microparticles were detected, sized, and quantitated by impedance-based flow cytometry (Fig. 3A). Tissue factorbearing microparticles derived from the ASPC-1 pancreatic cells varied in size from about 800 nm to 300 nm. As a control, purified human IgG was used in place of the antitissue factor antibody (Fig. 3B).

To determine whether tissue factorbearing microparticles could be detected and quantitated in platelet-poor plasma from cancer patients, blood samples were obtained from patients with advanced disease for an initial survey. Four histograms from the analysis of samples from patients with metastatic breast carcinoma, nonsmall cell lung carcinoma, and pancreatic carcinoma are shown. Tissue factorbearing microparticles were detected in these four samples, and the microparticle size varied from 0.30 m to 0.75 m, with a median value of about 0.40 to 0.45 m (Fig. 4A).

Blood samples were randomly collected from a variety of cancer patients as well as a cohort of cancer-free controls to provide preliminary information concerning the prevalence of tissue factorbearing microparticles in these disorders. These data were subsequently used to adequately power a case-control study. All cancer patients had metastatic or locally advanced disease whereas the control population was cancer free by history. Tissue factorbearing microparticles were detected in the platelet-poor plasma of subjects in all categories (Fig. 4B). Tissue factorbearing microparticles were detected in 68 of patients with pancreatic cancer, seven of whom had previously incurred a VTE. About half of the patients with metastatic ovarian carcinoma, breast carcinoma, and colorectal carcinoma had detectable tissue factorbearing microparticles. The concentration of tissue factorbearing microparticles varied from 20,000 to 10 million per L, with the lower limits of detection at 10,000 microparticles/L. Tissue factorbearing microparticles were detectable in significantly fewer cancer-free controls and patients with nonsmall cell lung carcinoma compared with patients with advanced pancreatic cancer (P < 0.001) or colorectal cancer (P = 0.03). These data show that tissue factorbearing microparticles can be measured in a significant percentage of cancer populations using impedance-based flow cytometry.

Given that tissue factorbearing microparticles can be detected in a high percentage of patients with malignant diseases using impedance-based flow cytometry, we explored the statistical association between tissue factorbearing microparticles and thromboembolic disease in patients with cancer. A case-control study was conducted to establish whether there is a relationship between the presence of tissue factorbearing microparticles and cancer-associated thrombosis.

Tissue factorbearing microparticles were measured in 30 cancer patients with acute VTE and 60 cancer patients without VTE selected to reflect a similar distribution of age, stage, sex, and diagnosis (Table 1). The median age in this cohort was 60 years. In the cancer-VTE group, one half of qualifying events were pulmonary emboli and five patients presented with both a proximal deep vein thrombosis (DVT) and pulmonary embolus at time of enrollment. Nine patients were enrolled with an isolated deep vein thrombosis; three were of the lower extremity and one patient was simultaneously diagnosed with upper and lower extremity DVT. A single patient was enrolled with a large inferior vena cava thrombus. The cancer stages of the cancer-VTE group were as follows: two myeloma patients at stage I by International Staging System, one case of chronic lymphocytic leukemia (Rai stage II), three patients with non-Hodgkins lymphoma at stage III and one at stage II, and the patient with ovarian cancer with active recurrent disease. The patient with glioblastoma had active disease, and all of the other solid tumor malignancies were metastatic except one patient with stage I breast cancer. A greater percentage of controls were actively receiving either chemotherapy or radiation therapy. With regard to antiangiogenesis agents associated with thrombosis, one individual in the acute VTE group was receiving lenalidomide whereas two control patients were taking thalidomide. Bevacizumab was administered to one individual in the acute VTE group and four in the control group. The detection of tissue factorbearing microparticles was not associated with gender, active therapy, smoking, or diabetes by Fisher's exact test (P > 0.05), or with age, complete blood count parameters (white cell count, hematocrit, platelet count) by Wilcoxon rank sum analysis (P > 0.05).

In this case-control study, tissue factorbearing microparticles were identified in 60 (18 of 30) of patients with cancer-associated VTE compared with 27 (16 of 60) of patients with cancer without VTE (P = 0.01). This corresponded to a >4-fold increase risk of thrombosis associated with detectable tissue factorbearing microparticles (odds ratio, 4.13; 95 confidence interval, 1.63-10.43). Multivariable models suggest that in addition to the presence of tissue factorbearing microparticles, lower platelet count, active therapy for underlying disease, and lower hemoglobin were also significantly associated with VTE. The adjusted odds ratio for the effect of detectable tissue factorbearing microparticles on occurrence of VTE in this model is 3.72 (95 confidence interval, 1.18-11.76).

To further assess the relationship between tissue factorbearing microparticles and thrombosis, we performed a retrospective analysis of deep venous thrombosis or pulmonary emboli diagnosed in the cancerno VTE group in the two years following enrollment. Only documented evidence of a new proximal extremity deep vein thrombosis or pulmonary embolism was included in the analysis. Of the 60 patients in this group, 16 had measurable tissue factorbearing microparticles, 4 of whom subsequently developed radiographic evidence of VTE within 12 months of enrollment (Fig. 5). One thrombotic event was recorded among the 44 patients without detectable tissue factorbearing microparticles, but after 15 months following enrollment into the study. The tissue factorbearing microparticlepositive group and the tissue factorbearing microparticlenegative group did not differ significantly for age, sex, active cancer treatment, smoking status, diabetes, or the presence of metastatic disease at time of enrollment. Identifying death without VTE as a competing risk, the 1-year estimate of the rate of VTE in cancer patients with detectable tissue factorbearing microparticles was 34.8; among the same group without detectable tissue factorbearing microparticles, the 1-year rate was 0 (Gray test P = 0.002). The presence of tissue factorbearing microparticles in cancer patients initially thrombosis-free predicted a 7-fold increased risk of thrombosis over cancer patients who were negative for tissue factorbearing microparticles (odds ratio, 7.00; 95 confidence interval, 0.85-82.74; P = 0.02). Elevated levels of tissue factorbearing microparticles were associated with a sensitivity of 0.80 and specificity of 0.78 for the development of VTE with a positive predictive value of 0.25 and negative predictive value of 0.97 for a population with disease site distribution that matches that of our study.

A total of 22 patients with idiopathic venous thromboembolic disease without cancer were compared with the 30 cancer patients complicated by acute venous thromboembolic disease to determine whether the presence of tissue factorbearing microparticles was a feature of venous thromboembolic disease. The median age in the idiopathic VTE group was 55.5 years, which was younger than 59.5 years in the cancer VTE group (P = 0.01), and 54.5 subjects were female (12 of 22; P = 0.47). Nine patients were diagnosed with a pulmonary embolism; this number did not differ significantly from that of the cancer VTE group (P = 0.71). Tissue factorbearing microparticles were detectable in 23 patients with acute idiopathic VTE (5 of 22), significantly fewer than the 60 (18 of 30) in the cancer patients with acute thromboembolic disease (P = 0.01). The median number of tissue factorbearing microparticles in the cancer VTE cohort (7.1 10⁴ microparticles/L) was significantly greater than both the idiopathic VTE and cancerno VTE groups whose levels fell within the undetectable range (P = 0.002 and P = 0.03, respectively). These results confirm that the high prevalence of tissue factorbearing microparticles in patients who present with an acute VTE is associated with cancer patients and not VTE alone.

To examine whether the measured tissue factorbearing microparticles are derived from the underlying malignancy, we measured tissue factorbearing microparticle concentrations before and after definitive surgery in three patients with pancreatic carcinoma. In one patient, the preoperative concentration of tissue factorbearing microparticles was 5.40 10⁵/L. Forty-four days following distal pancreatectomy, there was a >10-fold decrease of these microparticles to 3.92 10⁴/L (Fig. 6A). In a second patient, the preoperative concentration of tissue factorbearing microparticles was 1.02 10⁶/L. These microparticles were undetectable 26 days postoperatively. In a third patient, the preoperative concentration of tissue factorbearing microparticles was 7.06 10⁵/L, and similarly these microparticles were undetectable 27 days postoperatively. These results are consistent with the concept that the tissue factorbearing microparticles in these patients are tumor-derived.

To further explore this question, we examined the tissue factorbearing microparticles from these patients with pancreatic cancer for the presence of the tumor marker, MUC-1. The MUC-1 antigen is a transmembrane glycoprotein overexpressed in epithelial malignancies such as pancreatic cancer but is not found on circulating hematopoietic cells (35, 36). The preoperative samples were labeled with the antitissue factor antibody and a monoclonal antibody directed against the MUC-1 antigen. In the three patients tested, approximately 50 of the circulating tissue factorbearing microparticles in these patients were also positive for the MUC-1 antigen (Fig. 6B). These results support the premise that a population of microparticles is derived from the underlying malignancy. Whether the tissue factorpositive, MUC-1negative microparticles are derived from the underlying tumor where the MUC-1 signal may be below the level of sensitivity or from hematopoietic cells has not been established.

In this study, we have examined the hypothesis that the pathogenesis of thrombosis that complicates many forms of cancer is due to high levels of tissue factor associated with circulating tumor-derived microparticles. To test this hypothesis, we had four requirements: (a) to develop a method for reliably detecting, sizing, and quantitating tissue factor antigenbearing microparticles in platelet-poor plasma, without isolation and manipulation of microparticles; (b) to survey patients with various forms of cancer to approximate the prevalence of tissue factorbearing microparticles in their plasma; (c) to design and execute a case-control study to determine whether the detection of tissue factorbearing microparticles is more closely associated with cancer patients with acute thrombosis than with cancer patients without thrombosis; and (d) to determine whether tissue factorbearing microparticles associated with malignancy are tumor-derived.

In order to address the analytical issues of identifying and quantitating microparticles in platelet-poor plasma, we adapted a flow cytometric methodology capable of particle size measurements based upon impedance. Triggering on fluorescence, the diameter, in micrometers, of each fluorescent particle was determined. We showed that this system was capable of reproducibly determining both the size and number of fluorescent microspheres to an accuracy of 5. This method shows marked improvement over light scatteringbased flow cytometry, a nonoptimal technique for estimating microparticle size and number.

In the current work, we have shown a significant association between the presence of tissue factorbearing microparticles and acute cancer-associated thrombosis. In contrast, acute thromboembolic disease in the absence of malignant disease was not characterized by a similar increase in levels of tissue factorbearing microparticles. The cancer patients who presented with an acute VTE were similar to a cancerno VTE population for multiple factors including age, sex, cancer diagnosis, current cancer stage, diabetes, and smoking. However, the cancer patients with VTE tended to have lower platelet counts or hemoglobin and were less likely to be receiving chemotherapy. The basis for the differences in blood counts and active therapy between the cancer-VTE and cancerno VTE groups was likely due to differences in timing of study enrollment. The acute VTE group was identified at any point during the treatment cycle whereas control patients were enrolled at outpatient visits and were more likely to have recovered from chemotherapy-associated cytopenias. As both higher platelet counts and chemotherapy are considered contributing risk factors for cancer-associated VTE (9, 37, 38), the risk of thrombosis attributed to tissue factorbearing microparticles may in fact be greater than observed. In this case-control study, we established that tissue factorbearing microparticles are 4-fold more likely to be found in cancer patients with an acute thrombotic event than in cancer patients without an acute thrombotic event. These results extend previous observations that tissue factorbearing microparticles can be found in the plasma of some patients with cancer, and show that tissue factorbearing microparticles are a risk factor for the development of thrombosis.

To further explore the association of circulating tissue factorbearing microparticles and VTE, we conducted a review of all radiographic studies obtained following enrollment to assess for the development of proximal deep vein thrombosis or pulmonary emboli in cancer patients who had not incurred VTE at time of enrollment. Although this retrospective review was done in a blinded manner, the patients were not systematically assessed for the development of VTE. However, the potential for systematic bias was low as this was a nonintervention trial and enrollment did not impact the subsequent management of these patients. Only those cancer patients with detectable tissue factorbearing microparticles subsequently developed a thrombotic event in the year following enrollment. This analysis corroborated the association between tissue factorbearing microparticles and cancer-associated thrombosis identified in the case-control study and suggests that these microparticles may be central to the pathogenesis of thromboembolic disorders associated with malignant disease. Due to the inherent limitations of small case-control studies and retrospective cohort analysis, larger prospective studies with systematic monitoring for the development of VTE are required to confirm these findings.

Other groups have also observed that tissue factorbearing microparticles can be measured in patients with advanced cancer (29, 30, 39). Tesselaar and colleagues showed an increase in tissue factor activity in platelet-poor plasma from patients with cancer compared with platelet-poor plasma from healthy controls (29). Tissue factor activity associated with microparticles was measured in all seven metastatic cancer patients with a recent history of venous thromboembolic disease. However, the absolute number of tissue factorbearing microparticles measured by flow cytometry was not significantly different in cancer patients compared with healthy controls (29). Tesselaar et al., using light scatteringbased flow cytometry, reported a median of 460 tissue factorbearing microparticles per L of platelet-poor plasma in pancreatic cancer patients (range, 240-1,550/L; ref. 29), or less than one tissue factorbearing microparticle per leukocyte, whereas the tissue factorbearing microparticle concentrations in our cancer patients varied from 70,000/L to 3,200,000/L. This is a discrepancy of 3 to 4 orders of magnitude, and raises issues of the sensitivity for the identification of tissue factorbearing microparticles by light scattering. Hron et al. reported that the median number of tissue factorbearing microparticles was significantly greater in 20 cancer patients with colorectal cancer compared with controls whereas the individuals' history of VTE was not specified (30). More recently, Khorana and colleagues observed that two patients with pancreatic cancer with increased levels of plasma tissue factor subsequently developed a venous thromboembolic event.(39)

We present two independent arguments that the tissue factorbearing microparticles are largely derived from the tumor itself rather than from inflammatory cells. First, in three patients with pancreatic carcinoma undergoing pancreatectomy with curative intent, the microparticle concentration before surgery was reduced to very low or unmeasurable levels about a month after surgery. Second, in three patients with pancreatic carcinoma, we showed that about 50 of the tissue factorbearing microparticles coexpress MUC-1, a tumor marker for pancreatic carcinoma. Tumor cells express tissue factor (12), the expression of tissue factor activity on tumor cells correlates with thrombotic risk in pancreatic cancer patients (40), cancer cells in culture shed procoagulant vesicular structures (34), and circulating tumor-derived tissue factorbearing microparticles can be detected in mice xenografted with human pancreatic tumors (41). These results provide compelling evidence that tissue factorbearing microparticles are derived from certain tumors. Indeed, we suspect that the variability of microparticle generation from different tumors, as well as the total tumor burden, is the basis for the variability of thrombotic events in cancer-associated thrombosis. However, the presence of circulating tissue factorbearing microparticles is not unique to cancer patients as high levels were identified in a subset of noncancer controls as well as individuals with an idiopathic VTE. The clinical significance of these elevations in noncancer populations is not known.

Venous thromboembolic disease is a leading cause of death in patients with malignant disease (42), but several recent randomized studies have failed to show the benefit of prophylactic anticoagulation in cancer patients even in the presence of indwelling central catheters (43,45). The methodologies and conclusions of this work provide a rational basis to determine whether the detection of tissue factorbearing microparticles in cancer patients and their use as a biomarker predict an increased risk of a thromboembolic event. If so, it is critical to determine whether patients identified with this biomarker will benefit from primary thromboprophylaxis.

No potential conflicts of interest were disclosed.

We thank Cynthia Kos, Koryn Johnson, and Thomas LaRocca for excellent technical assistance and Marie Mahony RN for clinical support; Michael Brochu, Sr., Rick Thomas, Michael Brochu, Jr., and Raquel Cabana at NPE Systems and Gus Alban at Beckman-Coulter for their collaboration in flow cytometer development for this application, and Vasilis Toxavidis at the BIDMC Flow Core Facility; Dr. Hing Wong at Altor Biosciences for the gift of humanized tissue factor antibody; and Dr. Donald Kufe for the monoclonal antibody to MUC1.