Tamoxifen therapy is reported to increase the risk of deep venous thrombosis and pulmonary embolism (DVT/PE). To the authors' knowledge, it is not yet known whether the risk changes with the amount of time elapsed since the initial tamoxifen prescription. This information would be valuable in identifying patients at high risk for DVT/PE.
The relation between timing of tamoxifen use and venous thromboembolism risk was examined. The study population of 16,289 women was identified from the clinical database of the Danish Breast Cancer Cooperative Group. It included women diagnosed with International Union Against Cancer (UICC) stage I or stage II estrogen receptor-positive breast cancer between 1990 and 2004 at ages 45 to 69 years. Risks, risk ratios (RRs), and crude and adjusted hazards ratios were calculated for each of the first 5 years after breast cancer surgery and then cumulatively over the next 5 years.
The 5-year risk of DVT/PE was 1.2% for women receiving tamoxifen and 0.50% for women not receiving tamoxifen. Women treated with tamoxifen were at a higher risk for DVT/PE during the first 2 years after exposure (RR, 3.5; 95% confidence interval [95% CI], 2.1-6.0). Subsequently, their risk was not found to be substantially increased (RR, 1.5; 95% CI, 0.88-2.5). Older women taking tamoxifen appeared to be at higher risk than younger women during the first 2 years of exposure.
The association between venous thromboembolism (VTE) and cancer has been known for nearly 150 years. At autopsy, VTE is found in >50% of cancer patients.1 Tamoxifen, a selective estrogen receptor (ER) modulator, effectively prevents breast cancer recurrence among women with estrogen-positive tumors.2 Although the pathogenesis is complex, there is strong and consistent evidence from clinical trials that tamoxifen use is a risk factor for deep venous thrombosis and pulmonary embolism (DVT/PE) in breast cancer patients. Estimates of the risk ratio (RR) of thromboembolic events range from 1.3 to 7.0.2-6 Chemotherapy, radiotherapy, and comorbid diseases such as hypertension and diabetes increase the risk of DVT/PE among tamoxifen users.7, 8
Because many trials that established the association between tamoxifen and DVT/PE were conducted in selected populations with short-term follow-up, it is important to examine the association in a clinical setting over the longer term. In addition, because tamoxifen has been shown recently to have a role in preventing breast cancer,9 understanding its risk profile is essential to weighing its benefits against associated risks.
The objective of the current study was to expand current knowledge by using a large population-based cohort of breast cancer patients with nearly complete case ascertainment and long follow-up to estimate the risks and hazards ratios (HRs) of developing DVT/PE. The study also examines the potential of other risk factors for thromboembolic events to modify tamoxifen's effect on the risk of DVT/PE.
MATERIALS AND METHODS
Women eligible for the study were diagnosed with International Union Against Cancer (UICC) stage I or stage II estrogen receptor-positive breast cancer between 1990 and 2004 at ages 45 to 69 years, as reported to the Danish Breast Cancer Cooperative Group (DBCG) clinical database. We linked the study population to the Danish National Registry of Patients (NRP) covering all Danish hospitals to obtain information regarding venous thromboembolism occurrence and comorbidities, using each patient's central personal registry (CPR) number. The CPR number is a unique identification number assigned to all Danish residents alive on or born after April 1, 1968 or at the time of immigration.10 Since 1977, the NRP has recorded the CPR number of the patient, dates of hospital admission and discharge, surgical procedure(s) performed, and up to 20 diagnoses, classified according to International Classification of Disease-8 (ICD-8) codes for events occurring before 1994 and ICD-10 codes for all subsequent events. Outpatient and emergency room data have been available since 1994. Study participants were followed until December 31, 2005.
For each patient, the following demographic information was collected from the DBCG: age, year of breast cancer surgery, height and weight (to calculate body mass index), county of residence, and hospital of diagnosis, as well as tumor characteristics (stage, size, morphology, and laterality). The prescribed treatment protocol was also collected, as well as the start and stop dates for tamoxifen, if available. Data regarding comorbidities were obtained from the NRP.
Tamoxifen use was defined as a dichotomous variable, ascertained from DBCG records as a component of a patient's prescribed breast cancer treatment protocol. The tamoxifen group was comprised of women receiving any treatment that included tamoxifen, and the unexposed group was defined as women undergoing all treatments excluding tamoxifen. The outcomes of interest were DVT/PE (ICD-8 and -10 codes 45,099; 45,100; DI260; DI269; DI269A; DI801; DI802; DI802B; DI803; and DI803E). Age, surgical procedures (other than breast cancer surgery), metastatic tumors other than breast cancer, radiotherapy, chemotherapy, diabetes, stroke, chronic obstructive pulmonary disease, and heart failure were assessed at baseline as potential confounders.
Analyses were restricted to women with no existing cardiovascular disease (defined using ICD-8 and ICD-10 codes) as of the date of breast cancer surgery. Descriptive analyses compared demographics, breast cancer characteristics, and comorbidities among tamoxifen-treated and untreated women. Follow-up was initiated 3 months after the surgery date (the average time between the date of surgery and the start of tamoxifen therapy). Follow-up ended on December 31, 2005. Risks of events were analyzed individually by year for the first 5 years of follow-up, and then cumulatively for Years 1 to 5. RRs and 95% confidence intervals (95% CIs) were calculated as estimates of the association between tamoxifen therapy and incident thromboembolic events. Numbers needed to treat were calculated as the inverse of the risk difference. Cox proportional hazards models were used to estimate crude HRs and adjusted HRs controlling for confounding, for Years 1 to 5 individually, and for Years 5 to 10 taken together. We tested the proportional hazards assumption by adding a covariate to the model to represent the interaction between exposure and the log of survival time. The assumption of proportional hazards was satisfied for all models.
Interactions between tamoxifen therapy and chemotherapy were examined using proportional hazards models, with women unexposed to tamoxifen serving as the reference group. Crude and adjusted HRs were estimated for those patients receiving tamoxifen and chemotherapy, and those receiving tamoxifen only. The assumption of proportional hazards was satisfied for all interaction models. We also examined departure from the additive effects of tamoxifen therapy and age. Women were categorized as aged ≤50 years or aged >50 years. Risks, risk differences, and the interaction contrast were calculated.11
All analyses were performed using SAS statistical software (version 9; SAS Institute, Inc, Cary, NC).
After excluding 515 (3.0%) women with pre-existing cardiovascular disease, 16 (0.09%) women who died before the initiation of follow-up, and 262 (1.5%) women with missing information on assigned treatment protocol, the study population was comprised of 16,289 women. Tamoxifen exposure was distributed almost equally among subgroups, with 8232 tamoxifen-treated women and 8057 untreated women in the study.
The median follow-up time was 4.0 years, with a minimum of 0 years and a maximum of 14.5 years. The distributions of demographic variables between the exposed and unexposed groups of women were similar, including baseline comorbidity levels (Table 1). However, exposed women were younger and more likely to have their surgery performed in earlier years than unexposed women. As expected, women receiving tamoxifen were more likely to have positive lymph node status or a grade 2 or grade 3 tumor. Nearly all women who were excluded from the analysis due to a missing treatment protocol assignment variable were aged <55 years.
Table 1. Distribution of Demographic and Health-Related Characteristics Among Breast Cancer Patients by Exposure to Tamoxifen
BMI indicates body mass index; UICC, International Union Against Cancer; BCS, breast cancer surgery.
Breast cancer characteristics
Lymph node status
Year of surgery
Diabetes type I or II
Moderate/severe renal disease
Congestive heart failure
Any surgery after BCS
Any other tumor
Chronic obstructive pulmonary disease
As shown in Table 2, tamoxifen therapy increased the risk of DVT/PE. The 5-year risk was 1.2% for women receiving tamoxifen and 0.50% for women not receiving tamoxifen, yielding a RR of 2.4 (95% CI, 1.6-3.4). However, there was a decreasing trend in additional risk conferred by tamoxifen therapy from Year 1 to Year 5 of follow-up. During the first 2 years, women taking tamoxifen had approximately 3.5 times the risk of DVT/PE as women not receiving tamoxifen. During the next 3 years of follow-up (Years 3-5), the risk was 1.5 times higher for women taking tamoxifen. Figure 1 shows that women treated with tamoxifen had a sharper rise in incident DVT/PE episodes (steeper slope) during the first 2 years of follow-up compared with unexposed women. The cumulative incidence curves then become nearly parallel, indicating no substantially increased risk in later years. Therefore, the number needed to treat to yield 1 thromboembolic event increased from Year 1 to Year 5. In Year 1, 319 women needed to be treated to produce 1 additional case of DVT/PE, but by Year 5, 4303 women needed to be treated to result in 1 additional case.
Table 2. Risk of DVT/PE in Breast Cancer Patients by Year After Surgery and Numbers Needed to Treat (Risks Reported in Percent)
Tamoxifen No. (Risk)
No Tamoxifen No. (Risk)
Risk Ratio (95% CI)
No. Needed to Treat 1/|RD|
DVT/PE indicates deep venous thrombosis and pulmonary embolism; 95% CI, 95% confidence interval; RD, risk difference.
The same decreasing trend was revealed in both the crude and adjusted HRs (Table 3) when proportional hazards regression was used to model each of the first 5 years of follow-up individually and then Years 5 to 10 taken together. During the overall initial 5-year period, 97 events occurred in the tamoxifen group and 40 events occurred in the unexposed group, yielding a crude HR of 2.8 (95% CI, 1.9-4.0) and an adjusted HR of 2.5 (95% CI, 1.7-3.6). However, the rate was substantially elevated only in Years 1 and 2, with adjusted HRs of 3.5 (95% CI, 1.6-7.5) and 3.4 (95% CI, 1.7-7.0), respectively. There was no increased risk of DVT/PE noted during Years 5 to 10 (HR, 1.1; 95% CI, 0.69-1.9), conditional on survival until Year 5.
Table 3. Crude and Adjusted HRs for DVT/PE Among Breast Cancer Patients by Tamoxifen Exposure and Year After Surgery
HRs indicates hazards ratios; DVT/PE, deep venous thrombosis and pulmonary embolism; 95% CI, 95% confidence interval.
Adjusted for age group, surgery, any tumor, radiation therapy, chemotherapy, diabetes, stroke, chronic obstructive pulmonary disease, and coronary heart failure.
Duration of Tamoxifen Use
Of the women treated with tamoxifen, 5.1%, 14%, 8.8%, and 72% were assigned to a 6-month, 1-year, 2-year, and 5-year treatment duration, respectively. The analysis was repeated excluding women prescribed less than a 5-year course of tamoxifen to examine whether protocol duration could explain the decreasing risk of DVT/PE over time. This analysis yielded the same declining pattern of both crude and adjusted HRs over time, with an adjusted Year 1 to 2 HR of 3.8 (95% CI, 2.2-6.4), and an adjusted HR for Years 3 to 5 of 1.8 (95% CI, 0.98-3.2).
Tamoxifen treatment and chemotherapy were examined to determine how these exposures interacted with one another to increase or decrease DVT/PE risk during Years 1 and 2. Of the 8232 tamoxifen-treated women, 1139 (14%) were also treated with chemotherapy. When these exposures were included in an adjusted proportional hazards model (Table 4), women treated with tamoxifen only (HR, 3.4; 95% CI, 2.0-5.7) and women treated with tamoxifen and chemotherapy (HR, 3.0; 95% CI, 0.92-9.6) were found to be at similar risk of DVT/PE.
Table 4. Interaction Analysis of Tamoxifen Treatment and Chemotherapy Among Breast Cancer Patients: Crude and Adjusted HRs for DVT/PE for the First 2 Years After Surgery
HRs indicates hazards ratios; DVT/PE, deep venous thrombosis and pulmonary embolism; 95% CI, 95% confidence interval.
Adjusted for age, surgery, any tumor, diabetes, stroke, chronic obstructive pulmonary disease, and chronic heart failure.
Tamoxifen and chemotherapy
Our analyses provided evidence of an important interaction between age and tamoxifen therapy. In Years 1 to 2, the risk of DVT/PE among women aged >50 years (8.8 of 1000) was higher than that which could be accounted for by the additive effects of age and tamoxifen therapy. The latter would have predicted a risk of only 4.4 of 1000. In Years 3 to 5, the risk of DVT/PE among these older women (5.4 of 1000) was slightly lower than expected given the additive effects of age and tamoxifen therapy, which would have predicted a risk of 5.7 of 1000 (Table 5).
Table 5. Risks (Per 1000) of DVT/PE Among Breast Cancer Patients During Years 1 and 2 and Years 3 Through 5, Across Age Strata and Exposure to Tamoxifen
DVT/PE indicates deep venous thrombosis and pulmonary embolism.
Consistent with previous trials,2-6 we found that tamoxifen therapy increases the risk for DVT and PE. Women receiving this therapy were at approximately 2.5 times greater risk during the first 5 years after breast cancer surgery. However, we also found a decreasing trend in the tamoxifen effect by duration of therapy, with the first 2 years after the initiation of therapy constituting the period of greatest increased risk. Risk during the following 3 years was not found to be substantially increased. The validity of our findings depends ultimately on the quality of the discharge diagnoses. The validity is high, because the predictive value of a discharge diagnosis of PE has been reported to be 90% in administrative databases, and slightly lower for DVT.12
Our estimates of the HR for DVT/PE were adjusted for many important risk factors, including cancer (other than breast) and chemotherapy. The association between cancer and DVT/PE appears to be related to the prothrombotic responses of the host to cancer, such as inflammation and necrosis, as well as the clot-promoting mechanisms of the tumor cells themselves, including the production of procoagulant and the release of proinflammatory cytokines.13 Chemotherapy has also been shown to be related to DVT/PE, with 5-year risks in breast cancer patients treated with placebo, tamoxifen, and tamoxifen plus chemotherapy reported to be 0.2%, 0.9%, and 4.2%, respectively.7 Due to the older age of patients in the current study cohort, the confounding role of pregnancy likely did not play a substantial role in our findings. Because we were comparing a relatively homogenous group of breast cancer patients, substantial residual confounding from other covariates such as total hip and knee replacement, trauma and spinal cord injury, certain concomitant medications, and a sedentary lifestyle is unlikely, but cannot be excluded.14 An additional important factor to consider is confounding by indication. Physicians may selectively prescribe tamoxifen to women without other major risk factors for DVT/PE. This prescribing pattern would likely bias our results toward the null, because the untreated group would have a higher baseline risk of thromboembolic events.
Although existing evidence for a causal association between tamoxifen and DVT/PE is strong, to our knowledge the current study is one of the first to suggest that the RR decreases with time after initiation of therapy, with the first 2 years being the most hazardous, and with only a slightly increased risk during Years 3 to 5. A review of trials comparing 5 versus 1 to 2 years of tamoxifen use also indicated that there was little effect of longer treatment on the number of deaths attributed to thromboembolism.2 The same timing of excess risk has been shown for tamoxifen in a preventive setting. Decensi et al reported an HR of 1.63 (95% CI, 1.02-2.63) for VTE among patients taking tamoxifen for breast cancer prevention, compared with those not receiving tamoxifen. All excess VTEs due to tamoxifen occurred within the first 18 months after randomization. However, the excess stemmed entirely from superficial phlebitis of the legs, not DVT or PE.8 Because the current study had nearly 3 times as many subjects, we may have been able to detect more DVT/PE events.
A decreasing trend in HRs for VTEs has also been found for estrogen replacement therapy (ERT) and VTE risk. On the basis of data from the Women's Health Initiative, Curb et al found that the VTE risk was highest in the first 2 years of ERT (HR, 2.79; 95% CI, 1.24-6.27), compared with Years 3 to 5 (HR, 1.18; 95% CI, 0.69-2.01).15 Because tamoxifen has estrogen agonist effects in some tissues, it is not surprising that our results parallel these findings.
Our evaluation of interactions revealed that adjuvant therapy given along with tamoxifen may alter the risk of VTE during the relevant risk period of the first 2 years of treatment. However, due to small numbers of patients at risk in some of the strata and resulting wide 95% CIs, these results should be interpreted with caution. The analysis of interaction between age and tamoxifen demonstrated a departure from additivity in 2 different directions, depending on the risk period examined. During the first 2 years after treatment initiation, older women (aged >50 years) were at greater risk of DVT/PE than would be expected from the effect of age and tamoxifen alone. In Years 3 to 5, older women were found to be at a lower risk than would be expected from the effect of age and tamoxifen alone. Therefore, age and treatment appear to work together to decrease the risk of the outcome in this time period. These results further indicate an important difference between the risk periods: tamoxifen increases risk only during the first 2 years of treatment.
Because the first 2 years appear to play an important role in the risk of DVT/PE among breast cancer patients taking tamoxifen, close monitoring for thromboembolic events might best be undertaken during these years. Less frequent surveillance then may be sufficient. Surveillance for DVT/PE, including ultrasound examination and D-dimer measurement, is reported to be highly predictive and cost-efficient.14 Because the majority of patients with acute PE who receive adequate anticoagulant therapy survive, surveillance, diagnosis, and rapid therapy are crucial and effective for high-risk patients.
It is important to establish the risk period for VTEs to formulate guidelines for prescribing tamoxifen and monitoring for cardiovascular events in both the preventive and adjuvant setting. In an editorial, Goldhaber recommended that the prevention of breast cancer should take precedence over the risk of venous thromboembolism, suggesting that women with a high risk of developing DVT should take systemic anticoagulation concomitantly with tamoxifen.16 Risk factors for VTE among women receiving tamoxifen include surgery, immobilization, or fracture within the previous month.17 Atherosclerotic risk factors, including older age, higher body mass index, elevated blood pressure, high total cholesterol, smoking, and a family history of coronary heart disease, also have been shown to increase the risk of VTE among tamoxifen users.8 Therefore, in addition to anticoagulation treatment, it may be important to educate patients regarding the importance of modifying risk factors for DVT/PE, including obesity, cigarette smoking, and hypertension.
Understanding the risk associated with tamoxifen therapy is increasingly important in the era of aromatase inhibitors (AIs). Risk/benefit profiles need to be evaluated for both therapies. AIs do not appear to have the same thrombotic risks as tamoxifen. One study demonstrated that patients treated with AIs had half the risk of thrombotic events compared with those treated with tamoxifen.18 However, AIs are contraindicated for premenopausal women, as well as for postmenopausal women with poor bone health. In addition, some women may decline AIs because of cost.19-21 Tamoxifen remains the most important adjuvant therapy for premenopausal breast cancer patients and is an important adjuvant therapy for postmenopausal women, alone or in sequence with an AI.
Further understanding the cardiovascular risks and benefits associated with tamoxifen will help physicians individualize treatment for their breast cancer patients, and may even assist in decision making concerning the use of tamoxifen, AIs, or a sequential therapy. The results of the current study indicate that women may be most susceptible to DVT/PE during the first 2 years of tamoxifen therapy—the time when targeted monitoring for VTEs is most needed.