- © 2000 by American Society of Clinical Oncology
Incidence of Recurrent Thromboembolic and Bleeding Complications Among Patients With Venous Thromboembolism in Relation to Both Malignancy and Achieved International Normalized Ratio: A Retrospective Analysis
- Barbara A. Hutten,
- Martin H. Prins,
- Michael Gent,
- Jeff Ginsberg,
- Jan G. P. Tijssen and
- Harry R. Büller
- From the Department of Clinical Epidemiology and Biostatistics and Department of Vascular Medicine, Academic Medical Center, University of Amsterdam, the Netherlands; Hamilton Civic Hospitals Research Centre; and McMaster University Medical Centre, Hamilton, Ontario, Canada.
- Address reprint requests to Barbara A. Hutten, MSc, Department of Clinical Epidemiology and Biostatistics, Room J2–204, Academic Medical Center, University of Amsterdam, PO Box 22700, 1100 DE Amsterdam, the Netherlands; email B.A.Hutten{at}amc.uva.nl
Abstract
PURPOSE: Initial heparinization followed by vitamin K antagonists is the treatment of choice for patients with venous thromboembolism. There is controversy whether known malignancy is a risk factor for recurrences and bleeding complications during this treatment. Furthermore, the incidence of such events in these patients is dependent on the achieved International Normalized Ratio (INR). The aim of this study was to assess the incidence of venous thromboembolic recurrence and major bleeding among patients with venous thromboembolism in relation to both malignancy and the achieved INR.
PATIENTS AND METHODS: In a retrospective analysis, the INR-specific incidence of venous thromboembolic and major bleeding events during oral anticoagulant therapy was calculated separately for patients with and without malignancy. Eligible patients participated in two multicenter, randomized clinical trials on the initial treatment of venous thromboembolism. Patients were initially treated with heparin (standard or low-molecular weight). Treatment with vitamin K antagonists was started within 1 day and continued for 3 months, with a target INR of 2.0 to 3.0.
RESULTS: In 1,303 eligible patients (264 with malignancy), 35 recurrences and 12 bleeds occurred. Patients with malignancy, compared with nonmalignant patients, had a clinically and statistically significantly increased overall incidence of recurrence (27.1 v 9.0, respectively, per 100 patient-years) as well as bleeding (13.3 v 2.1, respectively, per 100 patient-years). In both groups of patients, the incidence of recurrence was lower when the INR was above 2.0 compared with below 2.0.
CONCLUSION: Although adequately dosed vitamin K antagonists are effective in patients with malignant disease, the incidence of thrombotic and bleeding complications remains higher than in patients without malignancy.
VENOUS THROMBOEMBOLISM is a well-known complication of malignant disease that is partly attributed to a perturbed hemostatic system.1,2 In addition, patients with cancer are prone to prolonged immobilization, frequently undergo surgical interventions, or receive chemotherapy with or without the insertion of an indwelling central venous catheter. All these conditions contribute to the susceptibility of patients with cancer for thrombotic complications.3
Initial heparinization followed by vitamin K antagonists is the treatment of choice for an episode of venous thromboembolism, both in patients with and without cancer. There is, however, controversy about whether patients with cancer are at increased risk of recurrences and bleeding complications during anticoagulant therapy. Several investigators have observed an increased risk of both hemorrhagic and thrombotic complications in cancer patients,4,5 whereas others could not, or only partially, confirm this.3,6
It has now been documented in several studies that the incidence of both major bleeding and recurrence of thrombosis in patients with venous thromboembolism is highly dependent on the achieved International Normalized Ratio (INR) level.7,8 Therefore, we have carried out a study designed to assess the incidence of major bleeding and recurrence among patients with venous thromboembolism in relation to both the presence of malignancy and the achieved INR.
PATIENTS AND METHODS
Patients
Patients who took part in the Tasman or Columbus trials comprised the study population.9,10 These two open, multicenter, randomized clinical trials have been described in detail elsewhere.9,10 Briefly, the Tasman study evaluated the efficacy, safety, and cost-effectiveness of two treatment strategies (ie, intravenous unfractionated heparin versus subcutaneous low molecular-weight heparin) in the initial treatment of patients with symptomatic proximal deep vein thrombosis. The Columbus study also compared the efficacy and safety of subcutaneous low molecular-weight heparin and intravenous unfractionated heparin in the initial treatment of patients with objectively documented venous thrombosis and/or pulmonary embolism. In both studies, patients were randomly assigned to receive study drug for at least 5 days. Treatment with vitamin K antagonists was started either directly or the next day and continued for 3 months, with a target INR range between 2.0 and 3.0. An INR measurement was made every 2 or 3 weeks or more often, if indicated. For the present study, we used the data from the period between cessation of initial heparin treatment and discontinuation of oral anticoagulant therapy.
Patients who experienced a complication were often treated with concomitant medication and/or targeted at another INR level. Therefore, patients were not included when they had a major bleeding event or a recurrence of thrombosis before cessation of initial (low molecular-weight) heparin treatment.
Outcome Events
The principal study outcome events were recurrent venous thromboembolism and major bleeding. All outcomes were assessed by an independent central adjudication committee that used predefined criteria and whose members were unaware of treatment allocation. Bleeding was defined as major if it was clinically overt and met one of the following criteria: it was associated with a decrease in the hemoglobin level of at least 2 g/dL or a need for the transfusion of 2 or more units of red cells; it was retroperitoneal or intracranial; or it warranted permanent discontinuation of treatment. All other overt bleeds were classified as nonmajor. Recurrent venous thromboembolism was defined as symptomatic recurrence of venous thromboembolism, confirmed by objective diagnostic tests.9,10 If a patient experienced more than one major bleed or more than one recurrence, only the first event was considered.
Estimation of Time Spent in Different INR Categories
Four different INR categories were defined as follows: (1) ≤ 2.0; (2) more than 2.0 and ≤ 3.0; (3) more than 3.0; and (4) a missing INR category. Linear interpolation was used to estimate the number of days spent in these INR categories.11 This method assumes that the INR value between two consecutive measurements varies linearly from the value of the first to the second measurement. If the number of days between two consecutive measurements exceeded 28 days, the INR was considered not predictable for a middle part of this interval. For example, if the time between two consecutive INR measurements was 32 days, the INR level in the 4 days in the middle of this interval was allocated to the missing INR category.
If the number of days spent in different INR ranges could not be calculated at all, patients were excluded from analysis. The mean percentage of time spent in the target range (INR, 2.0–3.0) for patients with and without cancer was compared by means of a Student’s t test.
Analysis
The number of outcome events in a particular INR range was determined by taking the INR at the time of the event and, if this was not available, by taking the last INR measurement before the event occurred. If the time span between date of event and the last INR measurement before this event exceeded 14 days, the INR was considered unknown. The complication rate for both major bleeding and recurrent venous thromboembolism was calculated by determining the total number of events in a particular INR range and divided by the number of patient-years spent in that INR range. All calculations were performed separately for patients with and without known malignant disease.
Rate ratios were calculated for recurrence and major bleeding among patients with malignancy compared with those without. For calculation of the 95% confidence intervals (CI) of the incidence rates and rate ratios, the Poisson distribution was assumed. The effects of differences in baseline characteristics on the risk of venous thromboembolism and major bleeding were adjusted by using logistic regression analysis.
RESULTS
Patients
A total of 1,421 patients had been included in the two treatment studies. Of these, 118 patients (38 with malignant disease) were not included in the analysis for several reasons, which are detailed in Table 1. Of the 80 patients who had to be excluded because of missing dates or INR results, six patients (one patient with cancer) suffered a recurrence, and one patient without cancer had a major bleeding. Thus, 1,303 patients (92%) were available for analysis, of which, 264 (20%) were known to have malignant disease. Baseline characteristics for patients with and without malignancy are listed in Table 2. Patients with known malignant disease were older on average. Malignancies were commonly located in the genitourinary tract (29%), the gastrointestinal tract (19%), and in the breast (15%).
Reasons for Noninclusion in Present Analysis
Baseline Characteristics of Study Patients
Percentage of Time in Different INR Categories
The percentages of time spent in the different INR ranges (≤ 2.0; > 2.0 and ≤ 3.0; and > 3.0) are shown in Fig 1. Although the difference in percentage of time spent in the therapeutic range is small between patients with and without cancer (50% v 54%, respectively), it is statistically significant (P = .005).
Fig 1. Percentage of time spent in different INR ranges for patients with and without malignancy.
Incidence of Recurrences
During treatment with vitamin K antagonists, a total of 35 recurrent episodes of venous thromboembolism (31 deep leg vein thrombosis and four pulmonary embolisms) were observed. Of these, 14 were among patients with malignancy, and 21 were among patients without malignancy.
The overall incidence of recurrent venous thromboembolism was increased among patients with malignancy (27.1 per 100 patient-years; 95% CI, 14.8 to 45.4) compared with patients without malignancy (9.0 per 100 patient-years; 95% CI, 5.6 to 13.8). This difference in incidence is statistically significant (rate ratio 3.0; 95% CI, 1.5 to 5.9; P = .003).
The incidence of recurrent venous thromboembolism in relation to INR is listed in Table 3. In both groups, the highest incidence was seen for the lowest INR range. The excess risk for recurrence for patients with malignancy was 35.1 (54.0 minus 18.9) per 100 patient-years, whereas the excess risk was 8.7 (15.9 minus 7.2) per 100 patient-years for patients without malignancy during periods with an INR below 2.0. The effect of malignancy on the risk of recurrent venous thromboembolism was not influenced by the percentage of time spent in the target range and age (odds ratio: unadjusted, 2.7; adjusted, 3.2).
Incidence of Recurrent Venous Thromboembolism
Incidence of Major Bleeding
During treatment with vitamin K antagonists, a total of 12 major bleeding events occurred. Of these, seven bleeding episodes occurred among patients with malignancy, and five were observed in patients without malignancy.
The overall incidence of major bleeding was 13.3 per 100 patient-years (95% CI, 5.4 to 27.5) for patients with malignancy and 2.1 per 100 patient-years (95% CI, 0.7 to 5.0) for patients without malignancy. This difference is statistically significant (rate ratio, 6.2; 95% CI, 2.0 to 19.7; P = .002).
The incidence of major bleeding in relation to INR is listed in Table 4. As expected, in patients without malignancy the incidence of bleeding complication was highest in the INR range over 3.0, whereas this pattern was not clear among patients with a malignancy. The sites of bleeding complications are listed in Table 5. The effect of malignancy on the risk of major bleeding was not influenced by the percentage of time spent in the target range and age (odds ratio: unadjusted, 5.6; adjusted, 5.9).
Incidence of Major Bleeding
Sites of Major Bleeding
DISCUSSION
In this study, we observed that during treatment with vitamin K antagonists for documented venous thromboembolism, patients who were also known to have malignant disease had a three- to six-fold higher risk for both recurrence and major bleeding compared with patients without malignancy. In patients without malignancy, the incidence of both major bleeding and thrombotic recurrence followed the expected pattern across the INR ranges, ie, the highest incidence for bleeding was found in the INR category greater than 3.0 and the highest incidence for recurrence was found in the INR range below 2.0. For patients with malignant disease, the incidence of recurrent venous thromboembolism more or less followed the expected pattern. On the other hand, for major bleeding, an opposite pattern was seen; the lowest INR range showed the highest incidence.
When interpreting the results, some methodologic aspects should be considered. Although our analysis was conducted retrospectively, the data collection was performed prospectively in a large group of patients participating in two randomized clinical trials. However, because the underlying hypothesis was conceived in retrospect, data collection especially for the long-term treatment details might have received less attention than would have been possible if the underlying hypothesis had been conceived in advance. Consequently, in 80 patients (5.6%) the incompleteness of data led to unavailability for analysis. Still this seems an acceptable proportion. More importantly, all clinically suspected outcome events, including all bleeding and suspected recurrent thrombotic events and death had been assessed by an independent, blinded central adjudication committee on the basis of prespecified criteria, which minimizes the potential for bias. Although a relatively large number of recurrences were observed, the number of major bleeding events was quite small. Therefore, the estimation of the incidence for major bleeding in the different INR categories is less precise.
The mean age of patients with a malignancy was 66 years, whereas the mean age of the patients without malignancy was 59 years. Several studies have reported a higher risk of hemorrhagic complications with increasing age during anticoagulation, which could have partially explained the higher bleeding risk for cancer patients.12-14 However, among patients with a major bleeding event, the mean age of patients with and without cancer was similar (59 and 61 years, respectively).
Our findings are consistent with the results of Gitter et al15 who reported that malignant disease at initiation of warfarin therapy was significantly associated with an increased risk for both major bleeding and recurrent thromboembolism. Wester et al16 and Landefeld et al17 also found that malignancy was associated with a higher frequency of bleeding complications.
An interesting finding in our data was that, in the lowest INR range (≤ 2.0), the rate for bleeding complications was increased compared with higher INR levels for patients with cancer. This could be a chance phenomena because of small numbers. However, in a recent abstract, a similar higher bleeding rate was also found in association with low anticoagulation levels (< 2.0) in patients with a venous thromboembolism and cancer.18 This could indicate the importance of other causes for major bleeding in these patients.
Levine et al19,20 reported the results as a percentage of patients experiencing complications during oral anticoagulation. They found recurrent venous thromboembolism in 14% of 103 cancer patients compared with 4% of 317 noncancer patients (P = .001), which is similar to our findings.
At first sight, our results might seem inconsistent with the conclusions of Bona et al.6 They compared the complication rate (major bleeding and thrombosis) during warfarin treatment in a cohort of patients with cancer with a cohort without cancer. The investigators did not observe statistically significant differences in the rates of either major bleeding or recurrence between the two groups of patients and concluded that oral anticoagulation is safe and effective in patients with cancer. However, they reported rates of major bleeding for cancer and noncancer patients of 0.0074 per treatment-month and 0.0025 per treatment-month, respectively. If the rates of Bona et al6 are recalculated as the number of major bleeding events per 100 patient-years, they are comparable with our incidence estimates (cancer, 8.9 v 13.3, respectively; and noncancer, 3.1 v 2.1, respectively). Therefore, the conclusion of a statistically nonsignificant difference in their study is likely based on a lack of precision (only 108 patients and three major bleeding events) rather than lack of prognostic effect of the presence of malignancy.
Although Prandoni et al21 found that the presence of cancer increased the risk for recurrent venous thromboembolism (hazard ratio, 1.72; 95% CI, 1.31 to 2.25) consistent with our findings, their results for major bleeding complications were different. However, the results for major bleeding were only presented as proportions (3.4% for cancer patients compared with 3.0% for noncancer patients), not taking into account potential differences in periods at risk.
Patients with both venous thromboembolism and malignancy present a dilemma when anticoagulant treatment is indicated. Whereas they have an increased risk of major bleeding complications during anticoagulant treatment, they also have a greater risk of recurrence. Although the incidence of recurrent thrombosis in the absence of treatment is unknown in cancer patients, it is likely that treatment with vitamin K antagonists does reduce their risk of recurrent venous thromboembolism. This is supported by the observation that the incidence for the lowest INR range amounted to 54 events per 100 patient-years and was reduced to 19 per 100 patient-years in the target INR range. Because the risk ratios for bleeding and recurrence in cancer patients are both increased, the resulting risk-benefit balance for patients with and without cancer will likely be similar. Therefore, oral anticoagulants can be justifiably prescribed to patients with cancer who experienced an episode of venous thromboembolism. However, the high absolute incidence of thromboembolic events indicates that the anticoagulant treatment can be optimized. Such an optimization could be achieved by improving the current standard treatment with vitamin K antagonists in an attempt to reach a higher percentage of time spent in the INR target range. This would require more frequent monitoring in this high-risk group, even if prior INRs seem stable. Because such an approach would increase the burden of treatment associated with the use of vitamin K antagonists, cancer patients are a challenging group for which to improve long-term secondary prophylaxis after a first venous thromboembolic event.
Acknowledgments
ACKNOWLEDGMENT
We thank all the Columbus and Tasman investigators for the opportunity to evaluate the data collected on the patients included in these randomized clinical trials.
- Received February 3, 2000.
- Accepted April 26, 2000.
