Venous thromboembolism (VTE) contributes to morbidity and mortality in cancer patients and is a frequent complication of anticancer therapy. In the current study, the frequency, risk factors, and trends associated with VTE were examined among hospitalized cancer patients.
A retrospective cohort study was conducted using the discharge database of the University HealthSystem Consortium. This included 1,824,316 hospitalizations between 1995 and 2003 at 133 U.S. medical centers.
Among 1,015,598 cancer patients, 34,357 (3.4%) were diagnosed with deep venous thrombosis and 11,515 with pulmonary embolism (PE) (1.1%) for an overall VTE rate of 4.1%. Subgroups of cancer patients with the highest rates included black ethnicity (5.1% per hospitalization) and those receiving chemotherapy (4.9%). Sites of cancer with the highest rates of VTE included pancreas (8.1%), kidney (5.6%), ovary (5.6%), lung (5.1%), and stomach (4.9%). Among hematologic malignancies, myeloma (5%), non-Hodgkin lymphoma (4.8%), and Hodgkin disease (4.6%) had the highest rates of VTE. The rate of VTE increased by 28%, secondary to a near-doubling of PE rates from 0.8% to 1.5% (P < .0001). Among patients receiving chemotherapy, the rates of VTE rose from 3.9% to 5.7%, an increase of 47% (P < .0001). In multivariate analysis, risk factors associated with VTE included age ≥65 years, female sex, black ethnicity, use of chemotherapy, primary site of cancer, presence of comorbidities, and year of admission.
The association between cancer and thrombosis has been known since at least the 19th century.1, 2 Cancer-associated venous thromboembolism (VTE) has significant clinical consequences for patients. Thromboembolism is a leading cause of death in cancer patients3 and cancer patients who develop VTE have a significantly worse survival.4, 5 Cancer patients with VTE also suffer a higher rate of both bleeding complications and recurrent VTE.6 Finally, VTE consumes healthcare resources; the mean length of deep venous thrombosis (DVT)-attributable hospitalization in a retrospective analysis was 11 days and the average cost of hospitalization for the index DVT episode was $20,065 in 2002 U.S. dollars.7
The risk of VTE is not uniform across cancer subgroups. Certain sites of cancer, such as the pancreas, stomach, brain, and lung, and the presence of metastatic disease are associated with higher rates of VTE in multiple studies.8–12 Cancer patients receiving active therapy are also at a greater risk for the development of VTE. In a population-based study, cancer was associated with a 4.1-fold greater risk of thrombosis, whereas the use of chemotherapy increased the risk 6.5-fold.2, 13 In women with stage II breast cancer, the risk of VTE has been shown to decline after chemotherapy is completed.14, 15 Although newer anticancer therapies have decreased toxicity compared with traditional chemotherapy, some, in particular antiangiogenic agents, are associated with very high rates of VTE.16, 17
The published literature contains inconsistent findings regarding the frequency of VTE in hospitalized cancer patients. In an analysis of Medicare claims data for hospital discharges from 1988–1990, the rate of VTE was 0.6%.8 A more recent but smaller study that included both ambulatory and hospitalized patients identified a VTE incidence of 7.8% over 26 months.11 There is an increased perception in the oncology community that VTE is being diagnosed more frequently, possibly due to an increased awareness and an increased usage of diagnostic procedures. A large study of patients with 19 selected malignancies hospitalized between 1979 and 1999 from the National Hospital Discharge Survey reported a DVT rate of 2% and a pulmonary embolism (PE) rate of 1%.12 In that study, the incidence of VTE began to increase in the late 1980s and this trend continued into the late 1990s. In our analysis of neutropenic cancer patients hospitalized between 1995 and 2002, 5.4% of patients developed VTE per hospitalization and rates increased over the period of study as well.10
It is unclear from available studies whether the risk of VTE is increasing for all cancers or only for specific subgroups of cancer patients. We hypothesized that the rate of VTE was increasing primarily in patients on active chemotherapy and was not because of increased diagnostic testing. In the current study, we analyzed data from hospital discharge summaries of all cancer patients admitted to U.S. academic medical centers between 1995 and 2003 to identify the frequency, risk factors, and trends in cancer-associated VTE during hospitalization.
MATERIALS AND METHODS
All discharge summaries of adult cancer patients admitted between 1995 and 2003 to 1 of 133 academic medical centers in the U.S. were reviewed using the discharge database of the University HealthSystem Consortium. Patients were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes that contained at least 1 diagnosis of malignant disease (ICD-9-CM codes 140–208). Patients with thromboembolism were identified using codes for venous thrombosis (451, 452, and 453) and pulmonary embolism (415.1–415.19). Patients receiving active therapy were identified using codes for chemotherapy (99.25, V58.1, and V67.2), high-dose interleukin-2 (00.15), biologic therapy (99.28), adverse events from chemotherapy (E930.7 and E933.1), and neutropenia (288.0). Major categories of cancer that were studied included lung (162–163), breast (174,175), colon (153), rectum (154), stomach (151), pancreas (157), other abdominal cancers (152, 155, 156, 158, and 159), ovary (183), endometrium and cervix (179–182), head and neck (140–149 and 160–161), esophageal (150), Hodgkin disease (201), non-Hodgkin lymphoma (200 and 202), brain (191–192), prostate (185), bladder (188), renal (189), leukemia (204–208), testicular (186), myeloma (203), and sarcoma (170–171). Comorbidities and risk factors included infection (001–139.8, 480–486, and 9966.2), pulmonary disease (487–519), hypertension (401), renal disease (580–593), diabetes mellitus (250–250), congestive heart failure (428–428), hepatic disease (570–576), anemia (280–285), obesity (278.0), and the use of transfusions (99.0–99.09). Arterial thromboembolism was identified using codes for arterial embolism (444), acute cerebrovascular disease (433–434 and 436), and acute coronary arterial disease (410 and 411.1–411.8). Selected surgical oncologic procedures included mastectomy or lumpectomy (85.21–85.23, 85.34, 85.36, and 85.4), unilateral or bilateral radical cervical lymph node dissection (40.4), partial or total pancreatectomy (52.5–52.7), and spinal procedures (03.0, 03.09, 03.3, 03.32, 03.39, and 03.4). Diagnostic procedures for VTE that were studied included vascular ultrasound (88.77), ventilation perfusion pulmonary scanning (92.15), and computed tomography (CT) scan of the chest (87.41).
Trends in the proportion of patients with thromboembolism were studied for the time period 1995 through 2003 and assessed using the Cochran- Armitage test for trend. For patients with multiple hospitalizations, only a single randomly chosen hospitalization was included for analysis. The increase in events during the study period was estimated by comparing the proportion of the events during the first 2 years (1995–1996) to the proportion during the last 2 years (2002–2003), as a ratio of 2 ratios. Proportion of the increase, and P value testing of the hypothesis of no increase were reported. Logarithms of the proportion of events were assumed to follow a normal distribution and the difference in logarithms were compared with zero, similar to the methodology used for the relative risk estimation.18
Multivariate logistic regression modeling was performed to identify factors associated with a greater risk of thromboembolism. The model was fixed with medically relevant covariates chosen before analysis. The cancer type was included in the model with all cancer categories first. After adjusting for the additional covariates in the model, cancer types associated with an increased risk of VTE were kept as separate categories and the rest were grouped into the reference category. The full cohort of patients was used in the model. The data regarding age and years of study were complete. For ethnicity, the group ‘other/unknown’ was created. Binary clinical covariates were created based on the presence or absence of the relevant diagnostic code. The chi-square test was used to compare dichotomous outcomes for categoric variables. To address large sample size and multiple testing, only P values < .001 were considered significant. Statistical analysis was performed using SAS software (version 9.1.3; SAS Institute Inc, Cary, NC).
A total of 1,824,316 hospitalizations occurred in 1,015,598 cancer patients between 1995 and 2003 at 133 medical centers, including 338,552 patients (33.3%) with multiple hospitalizations (Table 1). Greater than one-third of patients (43.5%) were aged ≥65 years. Greater than two-thirds (71.4%) of the population was white, with blacks representing 12.8% and Hispanics 3.7%. Lung cancer was the most common cancer diagnosis (10.6%), followed by prostate (9.1%) and breast (7%) cancer (Table 2). Non-Hodgkin lymphoma (5.6%) and leukemia (4.6%) were the most common hematologic malignancies. A small minority of patients (1.8%) had multiple cancer diagnoses. Approximately 14% of the study population was identified as receiving active anticancer therapy. Of these, 74.6% received chemotherapy during the hospitalization being analyzed; 12.8% were admitted with a diagnosis of neutropenia, suggesting use of outpatient chemotherapy; and 12.6% both received chemotherapy and had a diagnosis of neutropenia.
Table 1. Characteristics of the Study Population and Associated Rates of Venous Thromboembolism
All patients 1995–2003
VTE indicates venous thromboembolism.
Cervical lymph node dissection
Congestive heart failure
Table 2. Site of Cancer and Associated Rate of Venous Thromboembolism
VTE indicates venous thromboembolism.
Sites of cancer
Endometrium and cervix
Head and neck
VTE events were reported in 41,666 patients (4.1%) during hospitalization (Table 1). DVT occurred in 34,357 patients (3.4%) and pulmonary embolism PE occurred in 11,515 patients (1.1%). Black patients had the highest rates of VTE (5.1%), followed by whites and Hispanic patients (4%); Asians/Pacific Islanders had the lowest rates (3.3%) (P < .0001). Sites of cancer with the highest rates of VTE included the pancreas (8.1%), other noncolorectal abdominal sites (6.6%), kidney (5.6%), ovary (5.6%), lung (5.1%), and stomach (4.9%) (Table 2). Among hematologic malignancies, myeloma (5%), non-Hodgkin lymphoma (4.8%), and Hodgkin disease (4.6%) had the highest rates of VTE. Patients receiving active therapy had a VTE rate of 4.9%, which is significantly greater than the rate of 4% observed in the rest of the study population (P < .0001). Other variables found to be significantly associated with VTE on univariate analysis included age, female sex, the presence of comorbidities (particularly renal disease, infection, congestive heart failure, hepatic disease, arterial thromboembolism, and anemia), and the use of transfusions (P < .001 for each).
Trends in VTE
Overall, the proportion of patients with VTE increased from 3.6% per hospitalization in 1995 through 1996 to 4.6% in 2002 through 2003, an increase of 28% (P < .0001) (Fig. 1). Particularly striking was the near-doubling in the rate of PE from 0.8% in 1995 through 1996 to 1.5% in 2002 through 2003 (P < .0001). The increase in VTE rates was not uniform across all subgroups of cancer patients. Among various ethnicities, the rise in VTE was highest in black patients, an increase of 36.7% compared with 26.8% for other ethnicities (P = .07). The increase in VTE also differed by site of cancer. Rates of VTE increased disproportionately in patients with cancers of the lung, breast, head and neck, esophagus, ovary, and sarcoma. In contrast, patients with pancreatic and renal cancer had among the highest rates of VTE, but the incidence did not change significantly over the period of study (P = .67 for cancer of the pancreas and .99 for renal cancer). Among patients with hematologic malignancies, patients with lymphoma had the greatest increase (28.7%; P < .0001).
Trends in VTE were also not uniform across the spectrum of cancer therapy. The increase in VTE was disproportionately greater in patients receiving chemotherapy, from 3.9% per hospitalization in 1995 through 1996 to 5.7% in 2002 through 2003—an increase of 47% (P < .0001). In contrast, cancer patients undergoing selected surgical procedures did not demonstrate similar trends. Patients undergoing surgery for breast, head and neck, pancreatic, or spinal cancers experienced no significant change in the rate of VTE during this period.
We examined whether the increase in reported rates of VTE was associated with a similar increase in the use of diagnostic procedures for VTE. Lower extremity vascular ultrasound examinations were performed in 1.7% of all hospitalizations in 1995 and 1996, declining to 0.8% in 2002 through 2003 (P < .0001 for trend). Similarly, the use of ventilation-perfusion scanning declined from 0.7% to 0.16% over the same period (P < .0001 for trend). The use of CT scan of the chest also declined, although not as precipitously (2.6% to 2.1%; P < .0001 for trend). Of note, however, the proportion of patients who underwent a CT scan of the chest and also carried a diagnosis of PE increased significantly, from 1.7% in 1995 to 7.2% in 2003 (P < .0001 for trend).
Death during hospitalization occurred in 67,748 patients (6.7%). Mortality was significantly and consistently greater among patients who developed VTE compared with patients who did not over the duration of study (16.3% vs 6.3%; P < .0001) (Fig. 2). In-hospital mortality rates were significantly higher among black (18.9%), Hispanic (19.1%), and Asian (20.5%) patients with VTE compared with white patients with VTE (15.4%; P < .0001). Overall, however, the risk of in-hospital mortality declined from 1995 to 2003 both for patients who did and did not develop VTE (P < .0001 for trend for each). The mortality rate was higher in patients who developed PE compared with those who did not (24.8% vs 6.5%; P < .0001). The mortality rate associated with PE was higher among black (27.8%) and Hispanic (32.4%) patients than among white patients (23.8%; P < .0001).
In a multivariate logistic regression analysis, variables associated with VTE included age ≥65 years, female sex, black ethnicity, use of chemotherapy, sites of cancer (including pancreas, other abdominal sites, kidney, brain, ovary, and lung), the presence of comorbidities (including arterial thromboembolism, pulmonary disease, renal disease, infection, and anemia), and the use of red cell or platelet transfusions (Table 3). The year of admission was also found to be significantly associated with the development of VTE, with the risk increasing over the period of study.
Table 3. Predictors of Venous Thromboembolism by Multivariate Logistic Regression Analysis
OR (95% CI)
OR indicates odds ratio; 95% CI, 95% confidence intervals.
ORs are in comparison with white ethnicity.
ORs are in comparison with all other cancers.
ORs are in comparison with baseline years 1995–1996.
We examined the frequency, risk factors, and trends for VTE in hospitalized cancer patients over a period of 8 years. In our analysis, 4.1% of all cancer patients were diagnosed with VTE during hospitalization. This is higher than the rates reported by 2 studies from earlier time periods.8, 12 The first, by Levitan et al.,8 studied patients between 1988 and 1990 and reported a rate of 0.6%. The second study, by Stein et al.12 and conducted between 1979 and 1999, reported an overall rate of 2%. In the current analysis as well as in prior reports, VTE rates have risen over time, which could account for the higher rates observed in our more contemporary study. Rates of VTE in the study by Stein et al.12 approached 4% in the late 1990s, which is consistent with our findings.
A notable finding of this analysis was the significant association between ethnicity and VTE. Black patients had a 5.1% rate of VTE, the highest of any ethnicity. Furthermore, the rate of VTE increased at a higher rate in blacks than in other ethnicities, and the in-hospital mortality rate for PE was significantly higher in blacks than in whites. Black ethnicity continued to be significantly associated with VTE on multivariate analysis, after adjusting for known risk factors (odds ratio [OR] of 1.18; 95% confidence interval [95% CI], 1.15–1.22). The association of black ethnicity with a higher rate of VTE, particularly in nonidiopathic cases, has been previously reported in the general population.19 However, to our knowledge, the current study is the first such report in the cancer population. A possible explanation may be that black patients are more likely to have elevated Factor VIII levels,20 but to our knowledge this has not been studied specifically in cancer patients. The use of thromboprophylaxis has also been reported to be lower in black patients.21 Our observations are deserving of further study.
The increased risk of VTE with systemic chemotherapy has been well documented in prior population-based studies.2, 13 In a recent large cohort study, chemotherapy was associated with a 2.2-fold increase in VTE compared with cancer patients not receiving chemotherapy.22 However, to our knowledge, prior studies of hospitalized patients have not separately evaluated the risk in patients receiving chemotherapy.8, 12 In our current analysis, chemotherapy was indeed a significant risk factor for developing VTE (adjusted OR of 1.15; 95% CI, 1.12–1.18). Other variables associated with VTE in our analysis such as older age, particular sites of cancer, and the presence of comorbidities are consistent with prior reports.8, 10, 12
Overall, the rate of VTE increased over the period of analysis, and the year of admission was found to be a significant risk factor for the development of VTE after adjusting for other risk factors in multivariate analysis. This confirms popular perception regarding increasingly frequent diagnoses of VTE in cancer patients and is also consistent with rising trends in VTE observed in the late 1980s through 1990s in the study by Stein et al.12 The use of newer chemotherapy agents and regimens may contribute to the observed increase in VTE. Rates of VTE rose faster in patients receiving chemotherapy than in those not receiving chemotherapy (47% increase vs 26%). In contrast, patients undergoing major surgical procedures for cancer did not experience a significant increase in the rate of VTE. For example, VTE rates in patients undergoing chemotherapy for breast cancer increased by 47.5%, whereas rates remained stable in patients undergoing lumpectomy or mastectomy for the same diagnosis. This suggests an effect of chemotherapy rather than changes in diagnostic testing as the cause of the increase.
Another possible cause for the increased frequency of VTE diagnosis may be related to newer diagnostic technologies. Although rates of diagnostic testing did not increase in our study, the resolution of CT technology has increased significantly in recent years. In particular, the availability of multidetector-row CT technology expanded considerably in the earlier part of this decade.23 Trends in our study population demonstrated a greater likelihood of a diagnosis of PE in patients undergoing CT scans over the course of the study. Unfortunately, the administrative code for CT scan procedures does not provide information regarding the type of scan performed, and therefore we were unable to control for this variable. It should be noted that in this analysis, the diagnosis of PE carried an in-hospital mortality rate of 24.8%, strongly suggesting that PE diagnosed with higher-resolution CT scans are clinically significant. A recent analysis of ‘incidental’ PE in cancer patients found that a subsequent workup also identified DVT in 60% of patients; 2 of 16 patients in this study had recurrent PE, 1 of which was fatal.24 Indeed, in our analysis, VTE continued to be strongly associated with increased mortality throughout the period of study.
The limitations of the current analysis include its reliance on an administrative coding database; however, codes for VTE as well as comorbidities have been validated in prior reports and are considered to be accurate.25–28 The number of patients receiving active therapy may have been underestimated because we only included patients who were receiving chemotherapy during the admission being analyzed, or those admitted with neutropenia; some patients receiving outpatient chemotherapy may not have been identified. Alternatively, infection is associated with an increased risk of VTE independent of chemotherapy, and this could have increased the apparent risk in the chemotherapy group defined in this study. The diagnostic criteria used to define VTE included the code for superficial thrombophlebitis; however, < 1% of patients fell into this category and therefore did not substantially influence the analysis.
It is important to note here that VTE is potentially preventable with the use of thromboprophylaxis in appropriate patients. For medically ill patients without contraindications to anticoagulant therapy, several regimens have been demonstrated to be effective and well-tolerated29 and are recommended by major guidelines panels, including the American College of Chest Physicians and the National Comprehensive Cancer Network.30, 31 Despite this, recent retrospective analyses and surveys show that prophylaxis continues to be underutilized in hospitalized medical patients. In a recent global survey, although 52% of respondents stated that they would routinely utilize VTE prophylaxis for surgical patients, most respondents considered prophylaxis routinely in < 5% of their medical patients.32 Even worse, in a Canadian multicenter hospital audit of the use of thromboprophylaxis, patients admitted for cancer were significantly less likely to receive prophylaxis than other medical patients (OR of 0.40; 95% CI, 0.24–0.68 [P = .0007]).33 Unfortunately, we did not have access to data regarding compliance with prophylaxis in this study and are therefore unable to evaluate its contribution, if any, to the rates of VTE observed. Efforts to improve compliance with prophylaxis in appropriate hospitalized cancer patients are, however, sorely needed.
In summary, the results of the current analysis demonstrate that VTE is an increasingly frequent complication of hospitalization in cancer patients. Particular subgroups of cancer patients such as those with black ethnicity and those receiving chemotherapy are at even greater risk. The significant recent increase in VTE diagnosis cannot be attributed to an increased use of diagnostic procedures, although it may be related to improved CT scan technology. Regardless of its cause, VTE in hospitalized cancer patients is of considerable consequence, given its strong association with in-hospital mortality. Much work needs to be done to reduce the burden of VTE among cancer patients.