Ingrid Pabinger-Fasching, Clinical Division of Haematology and Haemostaseology, Department of Medicine I, Waehringer Guertel 18-20, 1090 Vienna, Austria. Tel.: +43 1 40400 4448; fax: +43 1 40400 4030. E-mail: firstname.lastname@example.org
Summary. Background: In cancer patients, laboratory parameters that predict venous thromboembolism (VTE) are scarce. Increased platelet count has been found to be a risk factor for VTE in cancer patients receiving chemotherapy (CHT). We have assessed high platelet count as a risk predictor for VTE in patients with cancer undergoing discriminative anti-cancer treatments and investigated whether platelet count correlates with thrombopoietin (TPO) levels. Design and methods: The Cancer and Thrombosis Study (CATS) is an ongoing prospective observational study of patients with newly diagnosed cancer or progression of disease, which started in October 2003. Occurrence of VTE and information on the patients’ anti-cancer treatment during follow-up were recorded. Results: Between October 2003 and February 2008, 665 patients with solid tumors were included (314 female/351 male, mean age 62 years). VTE occurred in 44 patients (18 female/26 male, mean age 62 years). The cumulative probability of VTE after 1 year was 34.3% in patients with a platelet count (PC) above the 95th percentile representing 443 × 109/L compared with 5.9% in those below 443 × 109/L. High platelet count [hazard ratio (HR): 3.50, 95% confidence interval (CI): 1.52–8.06, P = 0.0032], soluble P-selectin [HR: 2.66, 95% CI: 1.42–4.96, P = 0.0021] and surgery [HR: 4.05, 95% CI: 1.74–9.46, P = 0.0012] were statistically significant risk factors for VTE in multivariable analysis along with leucocyte count, age, gender, radio- and CHT. We found no correlation between platelet count and TPO levels. Conclusions: High PC is a clinically important, independent risk predictor for VTE in cancer patients. PC was not found to be associated with TPO levels.
Patients with cancer are at an increased risk of venous thromboembolism (VTE) . The clinical presentation and the management of VTE in cancer patients differ from patients that are free from malignancy. In cancer patients , incidence rates for VTE differ remarkably, depending on tumor site, stage and treatment . Several studies focused on additional risk factors such as immobilization, co-morbidities or treatment-associated factors such as surgery, chemotherapy (CHT), hormone therapy and central venous catheters . Despite the large number of clinical risk factors, only limited data exist on laboratory parameters with a predictive value for the risk of VTE in cancer patients .
Thrombocytosis is not uncommon in cancer patients [6–8] and evidence exists that platelets play an important role in tumor angiogenesis .
Several studies have focused on the role of cytokines involvement in tumor-associated thrombocytosis [10–15]. However, it is still under discussion as to what extent thrombopoietin (TPO) secretion regulates the platelet count (PC) in cancer patients. To our knowledge, the role of TPO levels as a risk factor for VTE in cancer patients, either independently or in conjunction with high PC, has not been investigated yet.
Whether high PC has a clinically relevant impact as a risk factor for VTE is still under investigation. To date, two studies support its role in the occurrence of VTE. In a retrospective case–control study on medical patients, Zakai identified a PC of > 350 × 109/L at admission as an independent risk factor for VTE . In 2005, Khorana and colleagues analysed data from a prospective study on 3003 ambulatory cancer patients that had received at least one cycle of CHT . Over a median follow-up of 2.4 months, VTE occurred in 58 patients (1.93%). An elevated pre-CHT PC (≥ 350 × 109/L) was found to be an independent risk factor and was significantly associated with VTE. The aim of the present study was to investigate whether an elevated PC is associated with an increased risk for the development of VTE in patients with malignancy during a follow-up period of up to 2 years. Furthermore, we investigated whether a predictive value of TPO levels exists and whether PC is correlated with TPO levels, platelet volume, leukocyte count, fibrinogen levels and soluble (s)P-selectin.
Design and methods
A cohort of patients with newly diagnosed cancer or progression of disease was analyzed. All patients are participants of the Cancer and Thrombosis Study (CATS), an ongoing prospective cohort study that is conducted at the Medical University of Vienna. For this study, ethical approval was received and the patients signed informed consent after they had been personally informed about the details of the study. Patients were recruited at the Departments of Medicine I and IV, Surgery, Radiotherapy and Radiobiology, Gynecology, Urology and Neurosurgery at the Medical University of Vienna between October 2003 and February 2008. Inclusion criteria for this analysis were as follows: (i) patients with newly diagnosed cancer of the breast, lung, upper and lower gastrointestinal tract, pancreas, kidney, prostate or gynaecological system and sarcoma or progression of disease after complete or partial remission, (ii) histological confirmation of diagnosis, (iii) age over 18 years, (iv) willingness to participate and (v) written informed consent. Exclusion criteria for all participants were: (i) overt bacterial or viral infection, (ii) venous or arterial thromboembolism within the last 3 months and (iii) continuous anticoagulation with vitamin K-antagonists or low-molecular-weight heparin (LMWH). Additional exclusion criteria for patients with progression of disease were (iv) surgery or radiotherapy (RT) within the past 2 weeks and (v) CHT within the past 3 months. Patients were allowed to take aspirin, ticlopidine or clopidogrel and immobilized patients were allowed to be treated with LMWH as thrombosis prophylaxis during their hospital stay.
Patients underwent a structured interview on their medical history. Data on the site of tumor, histology and tumour stage were documented. Additionally, they received detailed written information on the symptoms of VTE and were asked to report immediately to our department in case of developing symptoms. At the time of inclusion a blood sample was drawn.
The observation period started at the time of blood sampling. Patients were contacted every 3 months regarding occurrence of VTE and anti-cancer-treatment and clinical records were examined with respect to the occurrence of thrombotic events. In case patients did not respond, their family doctors or relatives were contacted. Once a year, the Austrian Mortality Registry was searched for entries concerning study participants. Each patient was followed over a 2-year period or until death.
Symptomatic or fatal VTE were recorded as an event. There was no routine screening for VTE. When a patient developed symptoms of VTE, medical imaging was performed. Diagnosis of VTE was always confirmed by objective methods, for example duplex sonography or phlebography for deep vein thrombosis and computerized tomography for pulmonary embolism (PE). In case of a fatal event, diagnosis was confirmed by autopsy. Once a year all events were presented to an adjudication committee formed by independent experts in the fields of angiology, radiodiagnostics and nuclear medicine, who were completely unaware of the laboratory results of the patients. These experts confirmed diagnosis and examined and proved the clinical significance of the events. Accidentally detected thrombotic events (e.g. PE detected during a routine computerized tomography) were accepted as events, when the adjudication committee decided that this was a clinically significant event (mentioned separately in the results section). Patients that had developed VTE were treated with long-term anticoagulation with therapeutic doses of LMWH or with vitamin K antagonists. Patient who received LMWH or vitamin K antagonist because of other reasons than VTE within the observation period (e.g. a trial fibrillation) were excluded from analysis. In case a patient died death certificate protocols and, when an autopsy was performed, the written protocols were reviewed with regard to the presence of VTE.
Blood collection and laboratory tests
Venous blood samples were collected in Vacutainer K3-EDTA tubes (Vacuette®; Greiner-Bio One) and Vacutainer citrate tubes (Vacuette®; Greiner-Bio One) containing 1/10 volume sodium citrate stock solution at 0.129 mmol/L by sterile venipuncture drawn when the patients entered the study. PCs were evaluated from EDTA-blood within 2 h after collection by the automated hematology analyzer XE-2100 (Sysmex, Kobe, Japan). To obtain platelet-poor plasma, the citrated blood was centrifuged (ROTANTA/TRC®; Hettich) at 1500 g for 15 min. Plasma aliquots were stored at −80 °C until they were assayed for determination of TPO levels in series. Samples were coded prior to laboratory analysis. The technicians were unaware of the patients’ characteristics all the time. TPO levels were quantified by a commercially available ELISA kit (Quantikine; R&D Systems, Minneapolis, MN, USA). The minimum level of detection of the assay was 30 pg/mL TPO.
Continuous variables are given as mean ± standard deviation, in case of an abnormal distribution median and interquartile range. Categorical variables are given as absolute values and percentages. Observation endpoint was fatal or non-fatal VTE. Data were censored at death, end of the observational period after 2 years or loss to follow-up. The reverse Kaplan–Meier method was used to calculate the median follow-up time. Two Kaplan–Meier plots were used to visualize the differences in cumulative probability of VTE between patients with higher and lower PCs: the first using a cut-off of 443 × 109/L (representing the 95% percentile) and the second using a cut-off of 350 × 109/L. Univariate and multivariable Cox regression analyses were used to calculate the risk of VTE. The variable of main interest was PC; either as continuous variable per increase of 50 × 109/L or as binary variable with 443 × 109/L and 350 × 109/L as cut-offs. The multivariable Cox regression analysis was further adjusted for tumor stage (localized or disseminated disease), age at study inclusion, gender, leucocyte count (log-2 transformed because of skewed distribution), sP-selectin (log-2 transformed because of skewed distribution) and three time-dependent variables that were surgery, CHT and RT. We assumed that these treatments would influence the risk of VTE not only at the exact time-point of the procedure but for a longer time-period. Therefore these time-dependent binary variables were modeled to indicate a possible influence on the risk of VTE in the following way: surgery from the day of surgery plus 6 consecutive weeks, CHT from the first day of the first treatment cycle until the last day of the last cycle plus 4 weeks and RT from the first day of a treatment until the last day plus 4 weeks. TPO levels were available in 514 patients including 31 patients who developed VTE. For these patients, a further Cox regression analysis was conducted with TPO levels (logarithmically transformed because of skewed distribution). We tested for all pairwise interactions and interactions with log(time) by means of candidate variables within the multivariable statistical models. As no significant interaction was found (P-value smaller than 0.01, because of the multiple testing problem), no interaction was added to the multivariable Cox regression models.
Correlations among TPO levels and PC were calculated using Spearman’s rank correlation. Furthermore, we correlated PC with platelet volume, PC and sP-selectin using Spearman’s rank correlation, PC and leucocyte count, PC and fibrinogen levels using Pearson’s correlation, respectively.
A P-value smaller than 0.05 was used to indicate statistical significance. Statistical analysis was performed with sas Version 9.2 (SAS Institute Inc., Cary, NC, USA).
Six hundred and sixty-five patients with solid tumors including breast cancer (n = 139), lung cancer (n = 127), malignancies of the upper (n = 41) and lower (n = 116) gastrointestinal tract, pancreatic cancer (n = 55), renal cancer (n = 31) and prostate cancer (n = 107) were studied. Forty-nine patients suffered from other solid tumors than those mentioned above, most of them gynecologic malignancies and sarcomas. Another 49 patients were enrolled but excluded for various reasons: total tumor resection before inclusion (n = 16), anticoagulation in a therapeutic dose (n = 7), benign histology (n = 10), thrombosis within the last 3 months or at recruitment (n = 8), CHT within the last 3 months (n = 5), pregnancy (n = 1) and withdrawal of consent (n = 2). The median observation time was 443 days (1st–3rd quartile 214–731).
During follow-up, 312 (46.9%) patients underwent a surgical procedure, 413 (62.1%) patients received CHT and 300 (45.1%) RT; 288 (43.3%) patients were treated by surgery, CHT or RT alone, 82 (12.3%) by surgery and CHT, 72 (10.8%) patients by surgery and RT and 90 (13.5%) patients by CHT and RT. In 83 (12.5%) patients, surgery, CHT and RT were performed. Two hundred and sixty-six patients died within the follow-up period without clear evidence of VTE. Main characteristics of the study population according to their PC are given in Table 1.
Table 1. Characteristics of the cohort (665 patients) based on platelet count categories (tertiles)
PC ≤ 217 × 109/L
PC > 217 and ≤ 278 × 109/L
PC > 278 × 109/L
PC, platelet count; BMI, body mass index. *Avaliable in 652 patients (not including tumor associated symptoms or depression).
Age (years, mean/standard deviation)
Female gender [n (%)]
BMI (mean/standard deviation)
Three or more comorbidities [n (%)]*
Disseminated stage [n (%)]
Surgery within the observation period [n (%)]
Chemotherapy within the observation period [n (%)]
Radiotherapy within the observation period [n (%)]
Forty-four patients (6.6%) developed VTE during the observation period: thirty-three patients symptomatic and 11 patients asymptomatic VTE. Asymptomatic events were detected incidentally on CT scan, they were, however, considered to be clinically significant by the adjudication committee and therefore classified as events. Nineteen patients developed deep venous thrombosis (DVT) of the leg, two DVT of the arm, 17 developed PE, including one fatal PE, one DVT of the leg and PE, one DVT of the arm and PE, two portal vein thrombosis, one sinus vein thrombosis and one DVT of the leg and portal vein thrombosis. One renal vein thrombosis was detected incidentally, but considered to be not clinically significant by the committee and was therefore not classified as an event. Overall, the cumulative probability of VTE was 3.4% after 3 months and 7.1% after 1 year (Fig. 1).
An increased PC was shown to be a significant risk factor for VTE in univariate (HR of an increase of 50 × 109/L: 1.28, 95% CI: 1.12–1.46. P = 0.0003) and multivariable analysis with age, gender, stage, surgery, CHT, RT, sP-Selectin and leucocyte count (HR: 1.18, 95% CI: 1.01–1.38, P = 0.0406). A high PC (higher than 443 × 109/L representing the 95% percentile of our patients′ cohort) increased the risk for VTE more than 5-fold in univariate analysis (HR: 5.07, 95% CI: 2.35–10.95, P < 0.0001) and more than 3-fold in multivariable analysis (3.50, 95% confidence interval (CI): 1.52–8.06, P = 0.0032] compared with patients with a PC below the 95th percentile (Fig. 1). Other significant risk factors in multivariable analysis were sP-selectin (HR: 2.66, 95% CI: 1.42–4.96, P = 0.0021) and surgery (HR: 4.05, 95% CI: 1.74–9.46, P = 0.0012) (Table 2).
Table 2. Hazard ratio of venous thrombosis by platelet count and other risk factors in a multivariable model
*logarithmically transformed because of skewed distribution. CI, confidence interval; HR: hazard ratio; ns: not significant.
Platelet count ≥ 443 × 109/L
sP-Selectin* (per 2-fold increment)
Leucocyte count* (per 2-fold increment)
Age (per increase of 1 year)
A PC higher than 350 × 109/l increased the risk for VTE more than 2-fold in univariate analysis (HR: 2.42, 95% CI: 1.22–4.80, P < 0.011), in multivariable analysis, statistical significance was lost (HR: 1.63, 95% CI: 0.79–3.37, ns).
Cumulative probabilities of developing VTE in patients with a PC above and below the 95th percentile were 13.2% and 2.9% after 3 months and 34.3% and 5.9% after 12 months in Kaplan–Meier analysis, respectively (Fig. 1).
TPO levels were not shown to be associated with the occurrence of VTE in a univariate Cox regression model (HR 0.99, 95% CI: 0.72–1.35, P = 0.93). No correlation between PC and TPO levels was found (P = 0.39). Statistically significant but weak correlations were observed between PC and platelet volume (r = −0.35; P = 0.0001), between PC and leukocyte count (r = 0.32; P < 0.0001), between PC and fibrinogen (r = 0.36: P < 0.0001) and between PC count and sP-selectin (r = 0.15: P = 0.0001).
We demonstrated that a high PC is strongly associated with the occurrence of VTE, exceeding an at least 3-fold increased risk. The risk was independent of other established risk factors for thrombosis in cancer patients, such as sP-Selectin or surgery.
Our results are in agreement with the retrospective study of Zakai  and the prospective study of Khorana . Additionally, Khorana et al. presented a multivariate risk model for VTE in cancer patients in whom a new CHT regimen was initiated and reported on the risk of VTE during an observation period of 2.5 months . Next to PC (> 350 × 109/L), site of cancer, decreased hemoglobin or use of erythropoesis stimulating factors, increased leucocyte count and body mass index were included in the analysis. With this model, the authors were able to classify patients into three categories, according to their risk for VTE. The main differences between the aforementioned studies and our study are the cut-off that was used for definition of the PC, the co variables which were used in multivariate analysis, the duration of the observation period and the treatment procedures. Whereas Khorana and Zakai used a cut off of 350 × 109/L, the one we used represented the 95th percentile of our patient cohort (above 443 × 109/L), leading to a better distinction of the high-risk patient group. However, for a better comparability with the aforementioned studies we also used a cut off of 350 × 109/L, but using this cut-off statistical significance was lost in multivariate analysis (web only supplemental Fig. S1).
The incidence of thrombosis in our study is comparable to other studies, which investigated the risk of VTE in cancer patients [19,20]. We were able to follow our patients for a median time period of 13 months. Additionally, we also included patients with types of treatment other than CHT and we used time-dependent variables to find out whether treatments such as surgery, CHT and RT increase the risk for VTE within the period of and during the weeks after their application.
The underlying mechanism that leads to the high PC in a subgroup of cancer patients is yet unclear. We were not able to find a statistically significant correlation between PC and TPO levels. Furthermore, levels of TPO were not shown to be significant risk factors for VTE in Cox regression analysis. To the best of our knowledge this has not been investigated yet. Whether increased TPO levels play a role in tumor-associated thrombocytosis is still under discussion [21,22]. A correlation between TPO and PC is not supported by the data from our study. Contrary to the investigations performed by Kaser et al.  we did not investigate whether PC was interleukin-6 (IL-6) mediated, but we investigated whether there was a correlation between PC and leucocytes as well as between PC and fibrinogen – markers that would represent an inflammatory process or an acute phase reaction. Correlations were statistically significant but weak. To clarify the mechanisms of thrombocytosis in cancer patients, a longitudinal observation with repeated blood sampling would probably be more appropriate to elucidate the mechanisms behind the increased PC.
In the present study, other risk factors for VTE were surgery and sP-selectin, which turned out to be significant in multivariable analysis. Surgery is a well-known risk factor for the occurrence of cancer-associated VTE [23,24]. Recently, our group described sP-selectin as a predictive parameter for VTE in patients with malignancy . sP-selectin levels did not correlate with PCs and increased the risk independently of the PC.
The HR for CHT was around 1.7; however, statistical significance was not demonstrated in our study, although CHT is a well-known risk factor for VTE . In contrast to other studies, we calculated the risk of CHT as a time-dependent variable by considering the time span of CHT plus a 4-weeks lasting post-chemotherapeutic period in the analysis. Moreover, the majority of our patients (almost two-thirds) received CHT during the observation period. These circumstances might explain the fact that CHT was not a significant risk factor in our analysis.
Some limitations of our study need to be mentioned: our data can only be applied to patients similar to our study cohort with regard to tumor sites, stage, patients’ age and ethnicity. Our patients were not recruited consecutively, because of organizational reasons and involvement of various departments of the Medical University of Vienna; however, there was no specific selection in order of recruitment. More than 50% of our patients suffered from disseminated stage of cancer which means that distant metastases were existent at time of inclusion. A limitation might be that there are only 44 events. Compared with larger studies, which are mainly based on registry data, in our study multivariable analysis may be underpowered because of the smaller number of events and the number of variables assessed. In our study, data were collected carefully and each event was proofed by a panel of experts. Incidence rates of VTE differ remarkably in different studies, depending on the type and stage of cancer and not at least by different methods of inclusion criteria and detection of VTE (routine screening vs. detecting symptomatic VTE) [3,26,27]. In a minority of patients, PE was not symptomatic, but was described in reports of computerized tomographic imaging during routine follow-up investigation. We included certain events in these patients, when our panel of experts assessed these events as clinically significant. Interestingly, O′Connell et al.  described 70 patients with incidentally detected PE and found out, that these events had a significant adverse impact on survival.
In case a patient died at home or died in a hospital and no autopsy was performed, we tried to find out whether symptoms of VTE occurred and had led to medical imagining. It is possible, that especially in patients who died at home, thrombotic events might be missed. Furthermore, in the Austrian Mortality Register the cause of death is classified as ‘cancer’, independently of the ultimate cause of death.
Evidence from the last decades resulted in recommendations on the use of anticoagulants in prophylaxis and treatment of VTE in patients with cancer . The advantages and disadvantages of anticoagulation have to be balanced in this very specific patient group, as an increased risk of bleeding might be present . In addition to clinical risk factors, predictive laboratory parameters indicating an increased risk for VTE in cancer patients could help to treat patients according to their individual risk profile. The parameter PC can easily and routinely be measured in cancer patients and might be a valuable parameter for risk stratification of cancer patients for VTE.
R. Simanek: Administrative support, provision of study material and recruitment of patients, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. R. Vormittag: Concept and design, administrative support, provision of study material and recruitment of patients, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. C. Ay: Administrative support, provision of study material and recruitment of patients, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. G. Alguel: Administrative support, provision of study material and recruitment of patients, collection and assembly of data, final approval of manuscript. D. Dunkler: Concept and design, administrative support, data analysis and interpretation, manuscript writing, final approval of manuscript. I. Schwar Zinger: Provision of study material and patients, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. G. Steger: Concept and design, provision of study material and patients, collection and assembly of data, final approval of manuscript. U. Jaeger: Concept and design, administrative support, final approval of manuscript. C. Zielinski: Concept and design, administrative support, data analysis and interpretation, final approval of manuscript. I. Pabinger: Concept and design, administrative support, data analysis and interpretation, manuscript writing, final approval of manuscript.
Cats study group
We thank all persons that supported us in patient recruitment for the Vienna Cancer and Thrombosis Study (CATS): T. Brodowicz, J. Drach, H. Gisslinger, M. Hejna, G. Kornek, M. Krainer, C. Marosi, L. Öhler, R. Pirker, M. Raderer, W. Scheithauer, M. Schmidinger, P. Valent, H. Watzke, C. Wiltschke, S. Zöchbauer-Müller, K. Elandt, M. Hassler, S. Koder, I. Kührer, A. Eisenhut, M. Gnant, B. Teleky, G. Goldner, U. Dieckmann, G. Hohenberg, P. Munda, E. Kubista, M. Schwarz, O. Wagner and many colleagues more (all from the Medical University Vienna).
We also thank R. Rataj, H. Dude and J. Raglhofer (Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Vienna) for the management of blood samples and technical assistance.
We are grateful to the members of the adjudication committee: R. Koppensteiner, M. Haumer, A. Willfort-Ehringer (all from the Department of Angiology, Medical University Vienna), S. Metz-Schimmerl (Department of Diagnostic Radiology, Medical University of Vienna) and R. Dudczak (Department of Nuclear Medicine, Medical University of Vienna).
This work was supported by a grant from the Jubiläumsfonds of the Austrian National Bank (project numbers 10935 and 12739) and by an unrestricted grant from Pfizer Austria. The sponsors had no involvement in the design and conduct of the study, in the collection, management, analysis and interpretation of the data, or in preparation of the manuscript. We thank T. Altreiter (Clinical Division of Haematology and Haemostaseology, Department of Medicine I, Medical University of Vienna) for proof-reading the manuscript and M. Pabinger (Clinical Division of Haematology and Haemostaseology, Department of Medicine I, Medical University of Vienna) who was responsible for administration of the follow-up data and who supported us in data base administration.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.