Venous thromboembolism in patients with acute leukemia: incidence, risk factors, and effect on survival

Cancer-associated thrombosis is a common complication of patients with malignancies.^1-4  Although the risk associated with different malignancies has only more recently been quantified,^5,6  it is generally thought that solid tumors, such as pancreatic^5,6  and ovarian cancer,⁷  and brain cancer,^6,8  carry a much higher risk of venous thromboembolism (VTE) than hematologic malignancies such as lymphoma and leukemia.

Recent studies suggest that the incidence of VTE in patients with hematologic malignancies may be similar to the incidence observed in patients with solid tumors. In a population based case-control study of patients with a first episode of VTE, Blom et al found that the odds of developing VTE among patients with hematologic malignancies was approximately 26 compared with the general population.⁹  However, the majority of these patients had lymphoma or myeloma and not acute leukemia.

Although the association between malignancy and thrombosis has been well recognized, less is known about the risk of thrombosis in patients with acute leukemia and the impact of VTE on survival. Certainly there is abundant biochemical evidence for thrombin generation and disseminated intravascular coagulation in patients with leukemia.¹⁰  The few single-center reports of the incidence of venous thrombosis in patients with leukemia have focused primarily on children with acute lymphoblastic leukemia. These studies have suggested the cumulative incidence varies between 2% and 10.6%.^11,12 

The primary aim of this study was to better define the incidence of VTE in patients with both acute myelogenous leukemia and acute lymphoblastic leukemia in patients diagnosed in California between 1993 to 1999, when outpatient treatment of VTE was not the standard of practice. Secondary aims included the determination of risk factors associated with the development of VTE and analyzing the effect of VTE on mortality.

Since 1988, the California Cancer Registry has collected information on all residents with clinical, radiologic, or pathologic diagnosis of cancer, with the exception of nonmelanoma skin cancer and carcinoma in situ of the cervix. The registry estimates that 99% of cancer cases are captured and that approximately 95% of all cancer cases are diagnosed pathologically antemortem, with only 5% diagnosed at autopsy or without tissue diagnosis.¹³  Registry data include the date of initial diagnosis, primary anatomic site, histologic type, Surveillance, Epidemiology, and End Results (SEER) stage, and basic demographic information.

Since July 1990, the State of California Office of Statewide Health Planning and Development (OSHPD) has required that all public hospitals report the medical diagnoses (up to 25) and procedures (up to 30) performed on every patient hospitalized in the state. The medical diagnoses and procedures are coded using International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) coding. An encrypted version of the social security number permits a unique patient identifier, which allows linkage of serial hospitalizations and as well as linkage with other datasets.

To determine the incidence of VTE, the CCR was linked to the PDD as well as the state's master death registry. Residents who emigrate out of the state or who are hospitalized outside of California are not captured by these databases. Residents who die out of state are captured in the master death registry.

The acute leukemia cohort included all patients of any age who were diagnosed with either acute myelogenous or acute lymphoblastic leukemia between January 1, 1993, and December 31, 1999, exclusive of cases diagnosed in 1996. Because of the marked difference in age, race, and type and duration of treatment for acute lymphoblastic leukemia (ALL) versus acute myelogenous leukemia (AML), analyses were carried out separately for these 2 types of acute leukemia. As home treatment of VTE using low-molecular-weight heparin was not approved until January 1999, essentially all cases diagnosed with VTE were captured in the PDD. However, approximately 5% of the hospitalized cases or CCR cases lacked a social security number and because these cases could not be linked, they were excluded.

Venous thromboembolism is usually defined as lower-extremity venous thrombosis or pulmonary embolism. Cases with VTE were defined using specific ICD-9-CM codes: 451.1x, 451.2; 451.81, 453.1, 453.2, 453.8, 453.9, 415.1x; and, in addition, 997.2 or 997.3 when coupled with a secondary diagnosis of VTE. Cases with a VTE code had to have a length of stay of more than 2 days, unless they died. Cases coded as having both deep vein thrombosis and pulmonary embolism were classified as having pulmonary embolism. Cases coded as having upper extremity deep vein “thrombophlebitis” (451.83, 451.84, and 451.89) were analyzed separately from the cases with deep venous thromboembolism, and cases coded as having superficial phlebitis in either upper extremity or lower extremity superficial were not included.

Placement of a central venous catheter (CVC) was defined using Current Procedure Terminology (CPT) codes 36433 and 35489. Cases that had a catheter placed as an outpatient could not be identified. To identify only incident cases of VTE, all cases that had a prior hospitalization with a VTE code or that were coded as having a prior history of VTE (V12.51 or V12.52) were excluded. This study was approved by the University of California Davis Institutional Review Board.

Presence of medical comorbidity was analyzed using a modification of the Elixhauser comorbidity index.¹⁴  The 29 chronic medical conditions that make up this index were reduced to 24 by excluding terms for cancer (tumor, metastatic cancer, lymphoma) or conditions likely associated with an acute medical condition (electrolyte disturbance, coagulation deficiency). The presence or absence of chronic comorbidity was based on coding during the index hospitalization and any other hospitalizations in the year prior to diagnosis of leukemia.¹⁵ 

Primary outcomes were the incidence of VTE and upper extremity thrombophlebitis within 3 months, 3 to 6 months, 7 to 12 months, and 13 to 24 months of acute leukemia diagnosis. Incidence rates were calculated as both cumulative incidence (CI) and person-time (events per 100 patient-years). Cox proportional hazards modeling was performed to assess the associations of age, sex, race, leukemia subtype (using the French-American-British [FAB] classification scheme), and catheter placement on the incidence of VTE. The date of catheter insertion was assumed to be constant and on the date of admission. To assess the effect of VTE on the risk of death, multivariate models were generated using VTE as a time-dependent covariate and adjusting for age, race, and FAB subtype. All analyses were performed with SAS statistical software (SAS Institute, Cary, NC), and P values less than .05 were considered statistically significant.

The demographics characteristics of the AML cohort are shown in Table 1. There were 5394 cases with AML with a male predominance and mean age of 60.4 plus or minus 21 years. Only 14.1% of AML cases were younger than 35 years and 52% were 65 years or older. The majority of patients were white (70%), with African Americans, Hispanics, and Asians contributing 5%, 15%, and 9% of the cases, respectively; in the general California adult population in 1996, 56% were white, 6.6% were African American, 25.7% were Hispanic, and 11.1% were Asian or Pacific Islanders. The majority of the AML cases were classified as “not specified” (NOS). The distribution of the remaining cases that were classified using FAB subtype had a high proportion of M4/5 (48%) and lower proportion of M2 (23%).

As shown in Table 2, the 2-year cumulative incidence of all venous thromboembolism cases in the AML cohort was 5.2%, with 3.6% coded as having deep vein thrombosis or pulmonary embolism and 1.6% coded as having thrombophlebitis of the deep veins of the upper extremity or thorax. One hundred eighty-one (3.3%) of the 282 thrombotic events occurred within the first 3 months after diagnosis of leukemia, and only 18 (0.3%) occurred 1 to 2 years after diagnosis. The corresponding rate of thrombosis, expressed as an incidence density (rate per 100 patient-years) was 19.2% in the first 3 months, and 1.4% between 1 and 2 years after diagnosis. Twenty-six (22%) of the 120 deep vein thrombosis/pulmonary embolism events that were diagnosed in the first 3 months were coded as having pulmonary embolism (PE) and 13 (17%) of the 75 later events were PE.

Among 337 patients with AML-M3, 12 were diagnosed with VTE within 6 months (3.6%) and another 9 had an upper extremity thrombophlebitis (2.7%); corresponding figures for patients with AML NOS were 3% and 1.2%, respectively.

In a multivariate analysis, sex, older age, presence of multiple comorbidities, and presence of a central venous catheter were all significant predictors of deep venous thrombosis or pulmonary embolism in the first year after diagnosis of AML (Table 3). Female sex was associated with a 40% higher risk of VTE (HR = 1.4; CI: 1.1-1.9). Cases aged 25 to 59 years had an increased risk of being diagnosed with VTE relative to younger cases, although this finding was of borderline significance. Age of 60 years or older was not significantly associated with an increased risk of VTE compared with the patients younger than 25 years (HR = 1.5, CI: 0.7-2.8). Compared with cases that had no comorbidity, cases with 2 comorbidities had significantly higher risk of developing VTE (HR = 1.8; CI: 1.2-2.7). Asian race was associated with a lower risk of developing venous thromboembolism (P = .05) consistent with other studies.⁵  In contrast to reports from other investigators,¹⁶  AML FAB M3 was not associated with an increased risk of VTE.

Eighty-six (31%) of the 281 cases that developed deep venous thrombosis or PE within 2 years were coded as having upper extremity thrombophlebitis. The cumulative incidence of upper extremity deep vein thrombophlebitis was 1.1% at 3 months and 1.6% at 2 years. Placement of a central venous catheter for chemotherapy was strongly associated with a diagnosis of upper extremity DVT (HR = 3.4; CI: 1.8-6.5) (Table 4). No other potential risk factor was a significant predictor of upper extremity deep vein thrombosis.

There were 2482 cases of ALL (Table 1), and the average age was 25.3 plus or minus 25 years (± SD), which was significantly less than the AML cohort (P < .001). There were more males than females with ALL: 42.8% female (P < .001) versus 57.5% male, and this was most striking among Hispanics in whom 39.7% were female and 60.4% were male. Overall, 51% of the cases with ALL were younger than 15 years and only 11.6% were older than 65 years. The most common FAB subtype was L1. Immunophenotype was not available in this dataset.

The 2-year cumulative incidence of all venous thrombotic events among the ALL cohort was 113 events, 4.5% (Table 2). Similar to cases with AML, the preponderance of thrombotic events occurred in the first 3 months after diagnosis, with the incidence decreasing quickly over time (2.5% in first 3 months, and 2.0% in next 21 months). The corresponding rate expressed as events per 100 person-years was 11.1% in months 0 to 3, falling to 0.6% between 1 to 2 years after diagnosis. Five (10%) of the 48 deep vein thrombosis/pulmonary embolism events that were diagnosed in the first 3 months were coded as having PE, and 11 (25%) of the 44 later events were PE.

Table 5 shows the results of a multivariate analysis of potential risk factors associated with deep venous thrombosis or PE within 1 year of diagnosis of ALL. As was noted in the AML cohort, placement of a central venous catheter was strongly associated with VTE. Age 25 to 59 years and age older than 60 years were strongly associated with a higher risk of developing thrombosis compared with cases younger than 25 years. Presence of 3 or more chronic comorbidities also significantly increased the risk of developing VTE within 1 year of the diagnosis of ALL (HR = 2.7; 95% CI: 1.3-5.4). Asian race was not associated with a lower hazard ratio for VTE, although there were only a total of 218 Asian cases.

There were 21 ALL cases diagnosed with upper extremity thrombophlebitis, with a corresponding cumulative incidence of 0.6% between 0 to 3 months and 0.9% after 2 years of follow-up. The placement of a central venous catheter was strongly associated with development of upper extremity deep vein thrombophlebitis, but because of the small number of cases this was only of borderline statistically significance (HR = 4.1; CI: 0.9-18.2; Table 6). Presence of one or more chronic comorbid conditions was a strong risk factor for upper extremity thrombophlebitis, and presence of 3 or more comorbidities was associated with almost a 7-fold higher risk of thrombosis (HR = 6.7; 95% CI: 1.6-29). None of the patients younger than 16 years of age developed upper extremity thrombophlebitis, and in the multivariate model, age older than 25 years was associated with about a 3-fold higher risk.

It is possible the UE DVT was miscoded as LE DVT because to be abstracted as a 451 code a physician would have to specifically mention the term phlebitis or thrombophlebitis. If the diagnosis was “deep venous thrombosis” of an upper extremity or thoracic vein, then the code would be 453.8 and would have been considered a LE DVT in this analysis. Based on previous studies that approximately 1 of 3 VTE events should be pulmonary embolism,^7,8,17,18  a sensitivity analysis indicated that 77 of the AML cases classified as lower extremity DVT may have actually had an upper extremity DVT. Thus, in the AML cohort there may have been as many as 163 cases of upper extremity deep vein thrombosis, representing 58% of the 281 total thrombotic events. In the ALL cohort the sensitivity analysis revealed that 46 of the cases classified as lower extremity venous thrombosis may have actually had an upper extremity deep vein thrombosis. If this were the case, as many as 67 (59%) of the 113 of total thrombotic events in the ALL cohort might have had an upper extremity DVT.

In a risk-adjusted multivariate model of death within 1 year, both advancing age and the presence of chronic of comorbidities were significant predictors in the AML cohort (Table 7). Respectively, compared with FAB M3, cases with M1/2 or M4/5 had an approximately 40% and 50% higher risk of dying in the first year. Asian race was associated with a 20% decreased risk of early death (HR = 0.8; CI: 0.7-0.9). Insertion of a central venous catheter was associated with a 50% lower risk of dying in the first year in the AML cohort. Diagnosis of venous thrombosis or pulmonary embolism was not a significant predictor of death within 1 year of diagnosis of AML.

In the ALL cohort, increasing age and the presence of one or more chronic comorbidities were significant predictors of death within 1 year (Table 8). FAB subtype L3 (Burkitt) was also associated with reduced survival. Race/ethnicity had no significant influence on survival. Presence of a code indicating insertion of a central catheter was also associated with a better 1-year survival. After adjustment for other risk factors, development of deep vein thrombosis or pulmonary embolism was associated with a significant increase in the risk of dying within 1 year (HR = 1.4; CI: 1.0-2.0).

Although the association of cancer and thrombosis has been well described,¹⁹  there is a paucity of data that pertain to patients with hematologic malignancies. The results of this large population-based study indicate that lower extremity deep vein thrombosis, upper extremity deep vein thrombosis, or pulmonary embolism occurred frequently patients with either AML or ALL, particularly in the first 3 months after diagnosis, when these patients receive intense treatment. The rate of all venous thromboembolic events in the first 3 months of follow-up was very high, 19.2 events per 100 patient-years, in patients with AML and was only modestly lower in patients with ALL, 11.1 events per 100 patient-years. The corresponding 3-month cumulative incidence of any venous thrombotic event was 3.3% in AML patients and 2.5% in patients with ALL. The overall 2-year cumulative incidence of any thrombosis, 5.2% in patients with AML and 4.5% in patients with ALL, is similar to the incidence of thrombosis observed in patients with colon, esophageal, and renal cancer in some analyses.¹⁹ 

The size of this study allowed a robust multivariate analysis of potential risk factors. In a risk-adjusted model, women with AML had a 40% greater risk of developing deep vein thrombosis or pulmonary embolism. To our knowledge, this has not been previously reported and is unexplained. The only other significant risk factors for VTE in patients with AML were presence of chronic comorbidity and placement of a central venous catheter (CVC). The association of presence of a catheter and VTE may be due to inclusion of a significant number of cases with upper extremity deep vein thrombosis in the venous thromboembolism cohort. The only risk factor associated with upper extremity deep vein thrombophlebitis was presence of a central venous catheter.

Among the case with ALL the only risk factors associated with VTE were presence of comorbidity, age older than 25 years, and presence of a CVC. Again, a substantial fraction of the cases with VTE likely had an upper extremity DVT. Interestingly, risk factors for upper extremity deep vein thrombophlebitis included not only presence of a CVC, but presence of multiple chronic comorbidities and age older than 60 years.

In our cohorts, the 6-month and 24-month incidence of thrombosis in AML-M3 was not significantly higher compared with cases with subtypes or unspecified AML. In addition, in multivariate models AML-M3 was not a significant independent predictor of deep vein thrombosis/pulmonary embolism or upper extremity thrombophlebitis. It is possible that the previous studies might have overestimated the incidence of thrombosis in AML-M3 due to case selection bias. However, because 73% our AML cases were classified with AML-NOS (not otherwise specified), many cases with AML-M3 many have been included in the group, preventing accurate quantification of the relative risk of thrombosis.

The results of the present study extend the findings of smaller cohort studies from single institutions. Ziegler et al retrospectively determined the incidence of VTE in 719 consecutive patients with acute leukemia.¹²  They reported that only 15 patients (2%) developed acute venous thrombosis at time of or in the 4-month period prior to the diagnosis of leukemia, but before the initiation of chemotherapy. These authors reported rates of venous thrombosis were similar in patients with AML (2%) and ALL (2.1%), with the highest incidence noted in patients with AML-M3 (6.5%). An Italian observational cohort study reported that the 6-month cumulative incidence of VTE in a cohort of 379 adult patients with acute leukemia was 6.3%.¹¹  The corresponding 6-month incidence in the present study was 4.2% in the AML cohort and 3.5% in the ALL cohort. In a study by Khorana et al that included only hospitalized neutropenic leukemia patients, the incidence of VTE was 4.4%,²⁰  and this included patients with either upper or lower extremity venous thrombosis.

There were some limitations to this study. The outcome measure of venous thrombosis or pulmonary embolism was based on the presence or absence of specific ICD-9-CM codes, which are used to identify medical conditions in hospital discharge administrative datasets. Previous validation studies have shown these codes to have high predictive value in the range of 70% to 95% compared with abstracted chart documentation of objectively confirmed VTE.^21,22  However, a major source of false-positive coding for acute VTE is inclusion of patients with a prior VTE, particularly patients requiring chronic warfarin treatment. This potential source of error was minimized by excluding all patients who had a history of VTE. To the extent that asymptomatic thrombotic events were included that were detected by a screening test or simply noted on a computerized tomographic study, the incidence of thrombosis may have been overestimated. A small number of catheter placements might have been missed if placed in the outpatient setting; however, these patients had acute leukemia and the majority were treated, at least initially, as inpatients. Regardless, there might have been an underestimate of the effect of CVC on VTE. Changes in routine VTE prophylaxis, rates of CVC-associated thrombosis, leukemia therapy, and increased overall rates of VTE in hospitalized patients in general^20,23  may limit applicability of results to present-day practice.

Use of administrative data does not allow precise determination of the location of venous thrombotic events; instead the location must be surmised based on the specific ICD-9-CM code selected. Patients with upper extremity thrombosis are coded as having an upper extremity “thrombophlebitis” only if the physician specifically states that the patient has a deep vein “thrombophlebitis.” If this term is not used, upper extremity deep vein thrombosis events are assigned a nonspecific venous thrombosis code. In a sensitivity analysis based on the assumption that the number of patients with lower extremity deep vein thrombosis should equal approximately twice the number of patients with pulmonary embolism, as many as 58% of the AML cases and 59% of the ALL thrombosis cases might have had an upper extremity deep vein thrombosis. The fact that a central venous catheter was a strong predictor of deep venous thrombosis or pulmonary embolism suggests that many of these patients did have an upper extremity deep vein thrombosis. VTE classified as DVT NOS might have also skewed this analysis.

As with our previous studies for patients with solid tumors, medical comorbidities were associated with an increased risk of LE DVT and PE for both AML and ALL. As noted, curiously this was statistically significant for 2 but not 3 comorbidities in AML patients. The Elixhauser index is a measure of number, but not severity, of comorbidities. This counterintuitive finding may reflect this lack of severity component. The index is also not hierarchal with regard to thrombosis risk, and some comorbidities may be more relevant to thrombosis (congestive heart failure [CHF], poorly controlled diabetes) than others (well-controlled hypertension). Finally, with relatively small numbers this finding may have occurred merely by chance.

The incidence of thrombosis in children with ALL has been studied extensively.^24-27  In a meta-analysis of 1752 children with ALL enrolled in 17 prospective studies, the cumulative incidence of any venous thrombotic event was 5.2% (95% CI: 4.2-6.4) in a cohort with a mean age of 5.5 years.²⁷  In this analysis, the incidence of non–central nervous system venous thrombosis was only 2.3%. In the current study, the cumulative incidence of any thrombotic event after 2 years among patients younger than 25 years (mean = 8.6) was 2.6%, which is remarkably close to the value reported in the meta-analysis. As in our series, most of the thrombotic events occurred early after diagnosis, probably during induction therapy, and 59% of the non–central nervous system thrombotic events were upper extremity DVT, the same figure calculated in our sensitivity analysis.

In contrast to previous studies in solid tumors,^5,7,17,28  VTE was not associated with increased mortality in AML. There may be something intrinsically different about the mechanism or significance of VTE in acute myeloid leukemia compared with solid malignancies. For example, the metastatic potential of non–small cell lung cancer and colorectal cancer has been linked to tissue factor expression and procoagulant activation.^29,30  However, by nature, acute myeloid leukemia is disseminated and procoagulant potential may have no bearing on aggressiveness.

The association between VTE and mortality in ALL is concordant with what is seen in solid tumors. The explanation for the differential associations of VTE with mortality in ALL versus AML is uncertain. Age was more strongly associated with VTE in ALL than for AML, and increasing age is strongly associated with mortality. Therefore, there could be an interaction between age and VTE in patients with ALL. Because this was not a prespecified analysis, a formal test for interaction was not performed to prevent alpha error. There could also be interactions between treatment, VTE, and mortality that could not be determined with the information available in this dataset.

In conclusion, this study represents the largest cohort of patients with acute leukemia assembled to analyze the incidence of venous thromboembolism. The incidence of VTE in patients with acute leukemia was high, particularly during the first 3 months of treatment. Important risk factors in patients with AML include insertion of a central venous catheter, female sex, and the presence of 2 chronic comorbid medical conditions. In patients with ALL, the principal risk factors are older age, presence of a catheter, and presence of multiple medical comorbidities. In patients with ALL, development of VTE was associated with a higher risk of death within 1 year. Further research is needed to determine whether thromboprophylaxis is effective in preventing VTE in patients with acute leukemia.^31,32 

This study was supported in part by grant 1-RO3-CA99527-01 from the National Cancer Institute (Bethesda, MD; H.K.C.), UL1 RR024146 from the National Center for Research Resources (Washington, DC; T.W.), the Hibbard E. Williams Endowment (R.H.W.), and an American Society of Hematology Trainee Award (Washington, DC; G.H.K.).

Contribution: G.H.K. analyzed data, wrote the draft, and gave final approval; R.H.W. and H.K.C. developed the concept and design, acquired data, revised the draft, and gave final approval; D.J.H. analyzed data, revised the draft, and gave final approval; H.Z. analyzed data; and T.W. developed the concept and design, analyzed data, wrote and revised the draft, and gave final approval.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ted Wun, Division of Hematology and Oncology, UC Davis Cancer Center, 4501 X St, Sacramento, CA 95670; e-mail: ted.wun@ucdmc.ucdavis.edu.