The occurrence of deep venous thrombosis (DVT) or pulmonary embolism (PE), together described as venous thromboembolism (VTE), leads to significant morbidity and is potentially fatal. Among the 500,000 new cases of VTE diagnosed annually in the United States, ≥20% are associated with a malignancy.1, 2 VTE is at times the presenting sign of an underlying cancer.3 Several factors, including procoagulant agents secreted by tumor cells, immobilization, surgery, indwelling catheters, and systemic treatment (including chemotherapy), contribute to an increased risk of VTE in cancer patients.4-14 Treatment of VTE in patients with cancer can be challenging, as there is a higher risk of both recurrent VTE and bleeding complications, although this can vary according to the stage and type of the malignancy.15, 16 As VTE is common in cancer patients and is associated with significant morbidity and an increased risk of death,17-19 there is increasing interest in instituting primary prophylaxis in high-risk patients to prevent incident (first-time) VTE events. The identification of patients at sufficiently high risk for VTE to warrant primary thromboprophylaxis is essential. Certainly perioperative thromboprophylaxis in patients with active cancer is necessary and appropriate, based on current American College of Chest Physicians, American Society of Clinical Oncology, and National Comprehensive Cancer Network guidelines.1, 20, 21 Data for hospitalized medical cancer patients are more limited, but most guidelines recommend thromboprophylaxis during hospitalization for essentially all patients judged to be at more than minimal risk.1, 20 Whether these cancer patients require extended prophylaxis after hospital discharge merges with the question of whether patients with active cancer should receive primary prophylaxis. In this concise review, we will discuss risk stratification models that have been specifically developed to identify cancer patients at high risk for VTE and therefore might be useful in future studies designed to determine the potential benefit of primary thromboprophylaxis.
Venous thromboembolism (VTE) is common in cancer patients, and is associated with significant morbidity and mortality. Several factors, including procoagulant agents secreted by tumor cells, immobilization, surgery, indwelling catheters, and systemic treatment (including chemotherapy), contribute to an increased risk of VTE in cancer patients. There is growing interest in instituting primary prophylaxis in high-risk patients to prevent incident (first-time) VTE events. The identification of patients at sufficiently high risk of VTE to warrant primary thromboprophylaxis is essential, as anticoagulation may be associated with a higher risk of bleeding. Current guidelines recommend the use of pharmacological thromboprophylaxis in postoperative and hospitalized cancer patients, as well as ambulatory cancer patients receiving thalidomide or lenalidomide in combination with high-dose dexamethasone or chemotherapy, in the absence of contraindications to anticoagulation. However, the majority of cancer patients are ambulatory, and currently primary thromboprophylaxis is not recommended for these patients, even those considered at very high risk. In this concise review, the authors discuss risk stratification models that have been specifically developed to identify cancer patients at high risk for VTE, and thus might be useful in future studies designed to determine the potential benefit of primary thromboprophylaxis. Cancer 2012;3468–3476. © 2011 American Cancer Society.
Risk Factors for VTE in Cancer Patients
Although patients diagnosed with cancer have a 6-fold to 7-fold increased risk of developing thrombosis,7, 22 not all cancers pose the same risk.5, 6, 17, 23, 24 For example, pancreatic cancer is associated with a significantly higher risk as compared with head and neck cancer.11 Several risk factors for acute VTE have been identified in cancer patients that can be categorized as patient (or host) related, cancer related, and treatment related (Table 1). More recently, biomarkers have been identified that are associated with an increased risk of VTE and are being incorporated into risk assessment models.25, 26, 80.
|Patient-Related Factors||Cancer-Related Factors||Treatment-Related Factors|
|Demographics: age, sex, race||Primary site of cancer||Chemotherapy, hormonal, and biological therapy|
|Prior history of VTE||Stage||Indwelling catheters|
|Obesity||Time interval from cancer diagnosis||Supportive measures such as erythropoietin-stimulating agents, other growth factors|
|Number of chronic medical comorbidities||Hospitalization|
In retrospective studies of hospitalized cancer patients, age ≥65 years was shown in multivariate analysis to be an independent risk factor for VTE.23, 27 Age ≥60 years was implicated in a prospective study of postsurgical cancer patients.4 However, advanced age was not found to be a strong risk factor in other retrospective studies.17, 28
Asians and Pacific Islanders have a lower risk of VTE compared with other ethnicities as shown by several large population-based studies.30-32 This also appears to be true in patients with cancer.17, 23 In addition, concordant with data from the general population, African American cancer patients appear to have a higher risk of VTE compared with Caucasians,17, 23 but this could differ with the site of cancer.17
Genetic thrombophilias, such as factor V Leiden and prothrombin gene mutations, are highly prevalent in the Caucasian population,33 and they confer a relatively modest increased risk for VTE. As cancer itself is associated with increased risk of developing VTE,7, 22 it is not definitely known whether the presence of underlying thrombophilia increases the risk of VTE in either an additive or multiplicative fashion. A recent report suggested that factor V Leiden does increase the risk of VTE in patients with breast cancer who are receiving adjuvant therapy, although the absolute risk is quite low.34 Therefore, the utility of identifying these genetic thrombophilic conditions and using them in the assessment of overall risk is uncertain.22, 35
Performance status is used in oncology to quantify a patient's level of function and capacity for self-care, and to determine appropriate treatment regimens, dose adjustments, and prognosis. Several scoring systems can be used; the most popular and widely used are Karnofsky's index of performance status and the Eastern Cooperative Oncology Group (ECOG) performance status.36, 37 Poor performance status is associated with higher risk of VTE in patients with cancer.38 This is likely because of higher tumor burden and impaired mobility. Performance status may be useful to incorporate into a risk-assessment model for cancer-related VTE.
In the general population, a history of VTE is a significant risk factor for recurrence, increasing the risk approximately 8-fold.39 Similarly, a prior history of VTE in cancer patients, even before cancer diagnosis, is also associated with increased risk of acute recurrent VTE.4, 9, 40
A body mass index (BMI) of ≥35 kg/m2 has been found to be an independent risk factor for VTE in the general population, for cancer patients in general41 and in patients with ovarian cancer.40, 42 BMI was found to be an independent risk factor for VTE in ambulatory cancer patients receiving chemotherapy, in a risk model developed by Khorana and colleagues.25
The number of chronic medical conditions has been consistently shown to be an independent risk factor in several of our retrospective population-based studies of patients with VTE, specifically in patients with breast, colorectal, lung, ovarian, and bladder cancer, malignant gliomas, and acute myeloid leukemia.28, 43-47 As the risk of VTE increases with the number of chronic comorbidities, we have shown that it is the overall burden of chronic comorbid conditions that increases the risk, not the presence or absence of specific comorbidities. In a retrospective study of hospitalized cancer patients, patients with pulmonary disease, infection, or kidney disease had higher risks compared with patients without these.23
Primary site or type of cancer.
It is a classical concept that solid tumors, especially mucin-secreting carcinomas of the gastrointestinal tract, are associated with the highest risk of VTE.48 In large population-based analyses using the California Cancer Registry, we reported that the incidence of VTE was highest in patients with metastatic stage cancer of the pancreas (20%), stomach (11%), bladder (8%), uterus (6%), kidney (6%), and lung (5%),17 corroborating this classic notion that these types of cancer are highly prothrombotic. However, recent studies have also shown a relatively high incidence of VTE in acute leukemia (5%)44 and in other hematological malignancies, including multiple myeloma and lymphomas.49 The incidence is relatively low in patients with cancers of the breast, prostate, and head and neck cancers, particularly among patients who are not diagnosed with metastatic disease.6, 11
Histological subtypes affect the risk of acute VTE as well.29, 45, 50 The incidence of VTE was higher in patients with adenocarcinoma of the lung (VTE rate of 9.9 per 100 patient years in the first 6 months after diagnosis) compared with squamous cell cancer (VTE rate of 4.8 per 100 patient years in the first 6 months; hazard ratio, 1.9; 95% confidence interval [CI], 1.7-2.1).29
Advanced or metastatic stage cancer has been strongly associated with a higher risk of VTE in several large studies,11, 17 but in a study of ambulatory cancer patients receiving chemotherapy, the cancer stage was not found to be a significant risk factor.8 This absence of association could be a selection bias, whereby patients with the most aggressive (advanced) disease were not treated with chemotherapy.51
Time after initial diagnosis.
The incidence of VTE has been shown to be the highest within the first few months after diagnosis of cancer.5, 17, 22 This may be related to the principal risk factor for VTE being related to the biological aggressiveness of the cancer, which for most cancers is a characteristic present at the time of diagnosis. The early biological properties of the cancer may be more important than how quickly the cancer spread.43 In addition, the time period immediately after diagnosis is when treatment begins, including major surgery, radiation, and chemotherapy.17, 22
Treatment with chemotherapy significantly increases the risk of VTE by 2-fold to 6-fold, with doxorubicin-containing regimens being especially implicated.5, 7 The use of newer immunomodulatory drugs, such as thalidomide and lenalidomide for multiple myeloma, leads to particularly high risk, especially in combination with high-dose dexamethasone and other chemotherapies. The incidence of VTE has been reported to be as high as 27% in treatment-naive patients receiving thalidomide.52 The use of thalidomide with high-dose steroids and/or chemotherapy is thought to be associated with sufficient risk as to warrant primary thromboprophylaxis.20 The antiangiogenic agent bevacizumab is associated with an increased risk of VTE, with an incidence of 11.9% in a large meta-analysis.10 Somewhat paradoxically, the risk of hemorrhage is also increased with this drug. A recent meta-analysis of 12,617 patients with a variety of solid tumors illustrated that the risk of high-grade bleeding was significantly increased with bevacizumab treatment (relative risk, 1.91; 95% CI, 1.36-2.68).53
Certain hormonal therapies may increase the risk of VTE. Tamoxifen has been associated with increased risk of VTE.54 Studies of aromatase inhibitors are limited with respect to risk for VTE, but do illustrate that the incidence of VTE is reduced with aromatase inhibitors compared with tamoxifen.55, 56 Androgen blockers have been shown to slightly increase the risk of VTE.5, 57 Use of erythropoietin-stimulating agents has been recognized to be associated with an increased risk of VTE and was 1 reason for new restrictions on use in cancer patients; a large Cochrane meta-analysis showed that the relative risk was 1.67 with the use of erythropoietin-stimulating agents.58 One study has shown that blood transfusions are associated with increased risk of both venous and arterial thrombosis.59
Immobilization and subsequent venous stasis in nonsurgical hospitalized patients, as well as the comorbidities associated with hospitalization or institutionalization, lead to an increased risk of VTE in this setting.9 Primary thromboprophylaxis is to be considered for all hospitalized cancer patients in the absence of bleeding or other contraindications.20
Indwelling central venous catheters carry an increased risk of thrombosis. The incidence of upper limb DVT (detected by venography) is reported to range from 27% to 66%, and in autopsy reports incidences of PE of up to 50% have been reported.13 Unfortunately primary pharmacological prophylaxis to decrease catheter-related thrombosis in cancer patients has been unsuccessful.62
Laboratory or radiological tests that could predict for increased risk of VTE in patients with cancer would be of great utility for targeting primary prophylactic strategies. Readily available laboratory values include the white blood cell count and the platelet count.
Leukocytosis was previously found to be associated with cerebrovascular disease and ischemic heart disease.63, 64 It has been shown by several investigators to be independently associated with cancer-related thrombosis as well.25, 65 Leukocytosis may reflect the degree of inflammation known to promote VTE. Leukocytes themselves may play a role in VTE pathogenesis, especially monocytes and neutrophils.66
Thrombocytosis has also been associated with increased risk of cancer-related thrombosis.25, 67 A prechemotherapy platelet count of >350,000/mm3 was associated with a 3-fold increased risk of VTE in a prospective study of patients undergoing chemotherapy.8
The role of other biomarkers, including circulating cell or microparticle associated tissue factor, D-dimer, soluble P-selectin, C-reactive protein, and factor VIII levels, are under further investigation and may prove to be predictive of VTE risk.68-75
Risk Stratification and Development of a Risk Assessment Model
As above, there are several factors that contribute to increased risk of VTE in cancer patients. Incorporating these into a risk assessment model that would effectively stratify VTE risk among all cancer patients would be very useful. Patients identified as at high risk of VTE could be studied to determine whether primary thromboprophylactic strategies would lead to a decrease in the incidence of VTE with an acceptable incidence of bleeding. Not only might this decrease VTE-related morbidity, but because VTE has been associated with reduced survival, such prevention may in theory lead to improved survival in patients with cancer. This is an area of ongoing investigation. A useful model would use easily identifiable risk factors that would optimize the stratification of the risk of VTE and be valid across different practice environments.
Various risk models have been developed for hospitalized surgical and medical patients. An example used to estimate the risk of VTE for hospitalized medical patients is shown in Table 2.76 A score of ≥4 would prompt the use of thromboprophylaxis via an alert in the electronic health record. By using this risk tool, most patients with cancer fall in the higher-risk category; hence, prophylaxis is recommended to essentially all hospitalized cancer patients who do not have a contraindication.20
|Recent major surgeryd||2|
Although thromboprophylaxis is underused,77-79 there is little controversy regarding the need to institute prophylaxis in hospitalized cancer patients. Guidelines however are based on studies on the general medical inpatient population and not specifically those with cancer. The issue of the duration of primary thromboprophylaxis after medical hospitalization is unclear, but the indication for posthospital thromboprophylaxis and the indication for primary prophylaxis could be very similar depending on whether active tumor-directed therapy is being administered.
The majority of patients with cancer are treated in an ambulatory or outpatient setting. Identification of high-risk patients for thromboprophylaxis in this common situation has been more problematic. A risk stratification tool for ambulatory cancer patients receiving chemotherapy has been developed by Khorana and colleagues (Table 3).25 By using readily available clinical and laboratory parameters, this model has been used to identify ambulatory cancer patients who are at high risk for developing VTE during the time they receive active antineoplastic chemotherapy.
|Patient Characteristics||Risk Score|
|Site of cancer|
|Very high risk (stomach, pancreas)||2|
|High risk (lung, lymphoma, gynecologic, bladder, testicular)||1|
|Prechemotherapy platelet count ≥350,000/mm3||1|
|Hemoglobin <10 g/dL or use of red cell growth factors||1|
|Prechemotherapy leukocyte count >11,000/mm3||1|
|Body mass index ≥35 kg/m2||1|
The model was developed using data from a multicenter prospective observational study of patients with different cancer diagnoses who were initiating outpatient chemotherapy. A total of 4066 patients were followed prospectively for a maximum of 4 cycles of chemotherapy. Model derivation and validation were based on a split-sample method. Two thirds of patients (n = 2701) were randomly assigned to a derivation cohort, and ⅓ (n = 1365) were tested in an independent validation cohort. Sixty patients (2.2%) in the derivation cohort and 28 (2.1%) in the validation cohort developed VTE. Multivariate analysis revealed the site of cancer (divided as very high, high, or “other” risk), prechemotherapy leukocytosis and thrombocytosis, anemia and/or use of erythropoietin-stimulating agents, and BMI of ≥35 kg/m2 to be the most predictive factors for the development of symptomatic VTE. Anemia and the use of erythropoietin-stimulating agents were found to be strongly associated, with the majority of patients receiving erythropoietin-stimulating agents for hemoglobin levels of <10 g/dL; hence, these were taken as a single factor. Multiple other factors, including age, stage, comorbid conditions, use of myeloid growth factors, and performance status, were not found to be statistically significant.
Risk stratification was based on the score from the risk model. Patients with low risk (score 0), intermediate risk (score 1-2), and high risk (score ≥3) had rates of symptomatic VTE of 0.8%, 1.8%, and 7.1%, respectively, at a median of 2.5-month follow-up. In the validation cohort, these were found to be consistent, 0.3%, 2%, and 6.7%, respectively.
This tool has some major limitations, because the number of cases used to develop the model was relatively low, and the number of patient with specific types of cancer thought to be associated with a high risk of thrombosis, specifically gliomas, renal cell carcinoma, gastric and pancreatic cancers, lymphoma, and myeloma (recognized as being the highest risk cancers in population studies), was low.17, 49 The majority of the patients Khorana and colleagues studied also had a good performance status as they were receiving outpatient chemotherapy. The small sample size of the subgroup of patients with poor performance status precluded identifying patients with poor performance status as being at higher risk for VTE. Additionally, only patients who developed symptomatic VTE were included, as there was no routine screening for asymptomatic VTE; although the clinical significance of incidental VTE discovered in radiographic studies performed for other purposes in cancer patients has not been established. Also, as a result of the widespread use of erythropoietin-stimulating agents in this cohort, the use of erythropoietin-stimulating agents could not be separated from a hemoglobin level of <10 g/dL in this model.
Nevertheless, the Khorana model does offer a good start to developing a more robust risk model. Importantly, this risk model has been validated externally in 2 independent cohorts. The Vienna Cancer and Thrombosis study—a prospective observational cohort study of 819 patients initiating cancer treatment (including surgery and radiotherapy in addition to chemotherapy) as outpatients—found a cumulative incidence of VTE in 7.4% patients at a median follow-up of 656 days. By using the Khorana model, they found that the cumulative probability of developing VTE at 6 months was 1.5% with a score of 0, 3.8% with a score of 1, 9.6% with a score of 2, and 17.7% with a score of ≥3.80 They also expanded the risk model using 2 additional biomarkers—soluble P-selectin, and D-dimer—with 1 point added for each of these biomarkers if elevated (ie, P-selectin level ≥53.1 ng/mL and D-dimer level ≥1.44 μg/mL). In the expanded risk model, the cumulative VTE probability after 6 months was 35.0% in patients with the highest score (≥5, n = 30) and 10.3% in those with an intermediate score (score 3, n = 130), but only 1.0% in patients with score 0 (n = 200). The hazard ratio of patients with the highest compared with those with the lowest score was 25.9 (95% CI, 8.0-84.6). These 2 biomarkers definitely added value to improving the predictability of the model. This would need additional validation to be more widely accepted. Kearney and colleagues conducted a retrospective chart review of 112 patients with solid tumors or malignant lymphoma who had undergone chemotherapy within the previous 2 years, and found that the incidence of VTE was 5.0%, 15.9%, and 41.4% in the low-risk, intermediate-risk, and high-risk groups, respectively.81
Interestingly, in the Kearney study the mortality was 30.0%, 42.9%, and 69.0% in the low-risk, intermediate-risk, and high-risk categories, respectively, supporting the finding that the risk of VTE closely parallels the risk of mortality. Indeed, Kuderer also reported that this VTE risk tool was useful in stratifying risk of death.82 Death or disease progression was noted in 7%, 18%, and 28% in the low-risk, intermediate-risk, and high-risk score groups, respectively. In multivariate analysis, the risk score and VTE occurrence were both significant independent predictors for early mortality and shorter progression-free survival after adjusting for other prognostic factors, including age, stage, cancer type, and ECOG performance status.
Other risk assessment models have been developed, including 1 by Kroger and colleagues.9 The risk factors found to be predictive by these investigators were inpatient treatment, prior history of VTE, family history of VTE, undergoing chemotherapy, fever (graded according to the Common Toxicity Criteria), and elevated C-reactive protein level (>5.0 mg/L). Merging these factors into 1 variable (ie, number of risk factors) demonstrated that the predicted VTE risk increased with the number of factors in both outpatients (odds ratio [OR], 1.85; 95% CI 1.18-2.88; P = .0071) and inpatients (OR, 2.34; 95% CI, 1.63-3.36; P ≤ .0001). The predicted VTE rate was 2.3% when none of these factors was present, increasing to 72% if all 6 factors were present.9 The application of this risk model is limited by the small number of patients used to develop this risk model and the need for external validation of the tool. A recent prospective, observational study83 used the Caprini risk assessment model,84 applying it to cancer patients hospitalized because of infections, chemotherapy, or palliative care. This model assigns the same risk score to all patients with cancer, not distinguishing between any features such as the primary site, stage, or active treatment. This limited the application of the model, and a modification was suggested to assign a higher risk score to patients with cancer diagnosis within the last 6 months (except skin and thyroid cancer) or active malignancy, and those receiving treatment. The authors are currently validating their suggested modification in their institution.
Conclusions and Future Directions
VTE remains 1 of the leading causes of cancer-related morbidity and mortality. Pharmacological thromboprophylaxis significantly reduces the risk of VTE in many high-risk populations, including those with cancer.85-89 However, cancer patients are also more likely to have bleeding complications relating to anticoagulation. Risk stratification is important to avoid unnecessary exposure to patients at lower risk of VTE. Risk models, therefore, are important in identifying patients with a high risk of thrombosis and consideration of primary prophylaxis.
The model developed by Khorana and colleagues25 for ambulatory patients receiving chemotherapy is simple and incorporates readily available clinical data. A prospective study to determine the efficacy of primary thromboprophylaxis in high-risk patients identified using this model is ongoing (www.clinicaltrials.gov: NCT00876915). The results of this study are eagerly awaited. A study investigating a newer anticoagulant in patients initiating chemotherapy has recently been completed; the results of this study are expected soon (www.clinicaltrials.gov: NCT00694382). Newer anticoagulant agents, including direct thrombin and factor-Xa inhibitors, may also prove to be effective and should be studied systematically in cancer patients at high risk for VTE.
In the meantime, current guidelines recommend the use of pharmacological thromboprophylaxis in postoperative and hospitalized cancer patients, and ambulatory cancer patients receiving thalidomide or lenalidomide in combination with high-dose dexamethasone or chemotherapy, in the absence of contraindications to anticoagulation.1, 20, 21 Unfortunately, worldwide surveys of clinicians have shown low rates of pharmacological thromboprophylaxis in both surgical and medical patients with cancer.90 In addition, real-world data from a US hospital discharge database indicate that approximately half (53.6%) of cancer patient discharges with an indication for thromboprophylaxis actually receive thromboprophylaxis, and only 27.0% receive thromboprophylaxis that is appropriate in terms of type, dose, and duration.77 Oncologists should be mindful of the relatively high risk of VTE in cancer patients and overcome their hesitancy to use primary prevention for VTE.
The authors received editorial support in the preparation of this article from Katherine Roberts, PhD of Excerpta Medica, funded by Sanofi-Aventis US, Inc. The authors were fully responsible for all content and editorial decisions, and received no financial support or other form of compensation related to the development of the article.
CONFLICT OF INTEREST DISCLOSURES
R.H.W.: clinical trials grant support, Bayer, Bristol-Myers Squibb, Pfizer; grant support, California Department of Health Services. T.W.: clinical trials grant support, Eli Lilly, Bayer, Glycomimetics; grant salary support, National Center for Research Resources, National Heart, Lung, and Blood Institute, and National Cancer Institute of the US National Institutes of Health.