Complications associated with erythropoietin-stimulating agents in patients with metastatic breast cancer

A Surveillance, Epidemiology, and End Results-Medicare Study


  • Presented at the 2009 Annual Meeting of the American Society of Clinical Oncology, May 29-June 2, 2009, Orlando, Florida.

  • This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, NCI; the Office of Research, Development and Information, CMS; Information Management Services (IMS), Inc; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.



The authors evaluated the patterns of use and the risk of thromboembolic events (TEE) associated with erythropoietin-stimulating agents (ESAs) in older patients with metastatic breast cancer who were receiving chemotherapy.


The study was retrospective and used the SEER-Medicare linked database. Stage IV breast cancer patients diagnosed from 1995-2005, treated with chemotherapy, ≥66 years old, with full coverage of Medicare A and B were included. The World Health Organization's International Classification of Diseases (ICD-9) and the Healthcare Common Procedure Coding System (HCPCS) were used to identify the use of ESAs, chemotherapy, and complications of therapy. Analyses included descriptive statistics and logistic regression.


Of 2266 women, 980 (43.3%) received ESAs, and 1286 (56.7%) did not. Patients diagnosed after 1999 or who received treatment with taxanes, anthracyclines, or vinorelbine were more likely to receive ESAs. Patients receiving ESAs had higher rates of stroke (18.5% vs 15.1%, P = .031); deep-vein thrombosis (DVT; 21.3% vs 14.4%, P<.001), other/unspecified thromboembolic event (TEE; 19.8% vs 14.7%, P = .001), and any clot (31.3% vs 23.4%, P<.0001). In multivariate analysis, patients receiving ESAs had increased risk for DVT (odds ratio [OR], 1.36; 95% confidence interval [CI], 1.05-1.75), and any clot (OR, 1.26; 95% CI, 1.02-1.57). A dose-dependent effect was evident for stroke, DVT, other TEE, and any clot.


In this cohort of patients, the use of ESAs increased the risk of TEEs, with a dose-dependent effect for stroke, DVT, other TEE, and any clot. The data show that among patients treated with chemotherapy and ESAs for metastatic breast cancer, TEEs are a common event. Therefore, caution is recommended when using these agents. Cancer 2011;. © 2011 American Cancer Society.

The erythropoiesis-stimulating agents (ESAs), erythropoietin and darbepoetin, have been widely used to increase hemoglobin values and reduce transfusion requirements in cancer patients.1-3 Recent clinical trials and various meta-analyses have raised safety concerns, as results suggest that the use of ESAs is associated with adverse outcomes including thromboembolic events (TEEs) and potentially tumor progression and decreased survival.4-6

It is well known that patients with cancer have an increased risk of TEE7; however, an even higher incidence of embolic and thrombotic events has been seen in cancer patients treated with ESAs for different malignancies.8-11 A recent meta-analysis showed that patients with cancer who received ESAs had a higher risk of TEEs (hazard ratio [HR], 1.57; 95% CI, 1.31-1.87) and mortality (HR, 1.10; 95% CI, 1.01-1.20) than control patients.5 Current American Society of Clinical Oncology (ASCO) and National Comprehensive Cancer Network (NCCN) guidelines recommend awareness of the adverse outcomes associated with ESA use and advise its use exclusively in patients with chemotherapy-induced anemia who are being treated with noncurative intent.12, 13

Despite evidence, there is not a clear description of the different TEEs that are associated with ESAs use. Also, questions concerning a dose effect remain unanswered, and it is unclear whether increasing doses or longer durations of treatment are associated with worse outcomes. Although ESAs are agents that are widely used to treat cancer patients, to the best of our knowledge, there is no clear description of the patterns of ESAs use in patients with metastatic breast cancer. In this retrospective cohort study, using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database, we sought to estimate the use of ESAs and to evaluate the TEEs associated with its use in patients older than 66 years of age and who were receiving chemotherapy for metastatic breast cancer.


Data Source

We used the SEER-Medicare linked database. The SEER program, supported by the US National Cancer Institute (NCI), collects data from tumor registries; during the years included in this study, it covered 14% to 25% of the US population.14 The Medicare program is administered by the Centers for Medicare and Medicaid Services and covers 97% of the US population aged 65 years and older.15 SEER participants are matched with their Medicare records under an agreement between the NCI and Centers for Medicare and Medicaid Services. Of SEER participants who were diagnosed with cancer at age 65 years or older, 94% are matched with their Medicare enrollment records.15 Patient demographics, tumor characteristics, and treatment information were extracted from the SEER-Medicare Patient Entitlement and Diagnosis Summary File.

Study Population

In the SEER-Medicare database, 268,198 men and women were diagnosed with breast cancer between January 1995 and December 2005. We included patients aged ≥66 years who had a diagnosis of stage IV breast cancer (according to the American Joint Committee on Cancer staging system, third edition) and were treated with chemotherapy. Patients were required to have Medicare Part A and Part B and not to be members of a health maintenance organization (HMO) for 1 year before and after their breast cancer diagnosis because Medicare claims are not complete for HMO members. Patients who had end-stage renal disease (ESRD), prior cancer, or a noncarcinoma tumor histology were excluded.

From the initial 268,198 patients with breast cancer diagnosed from 1995 to 2005, 23,885 had a history of prior or subsequent malignancies, 843 had unknown month of diagnosis, 85,266 were aged younger than 66 years at diagnosis, 151,066 had an initial stage other than IV, 2047 did not have full coverage from both Medicare A and B or were members of an HMO, 16 had noncarcinoma histology, and 45 had ESRD. From the remaining 5030 patients, 2764 did not receive chemotherapy. A total of 2266 patients were included in this study.

Data Extraction and Definitions

To identify patients who received ESAs, we looked for related codes in the Medicare claims at any time after diagnosis. The following HCPCS codes were used for defining ESAs regimens: Q0136, J0885, Q0137, J0880, J0881, and C1774. For this analysis, a single dose of erythropoietin alfa was defined as 40,000 U and was considered equivalent to a dose of darbepoetin alfa of 200 μg.12 To identify patients that received chemotherapy, we searched for the following codes: ICD-9-CM procedure code 9925 for a hospital inpatient or outpatient facility claim of chemotherapy; HCPCS codes J8510, J8520, J8521, J5530 through J8999, and J9000 through J9999, excluding J9202, J9209, J9212 through J9214, J9217, J9218 in physician, outpatient or durable medical equipment (DME) files. Revenue center codes 0331, 0332, and 0335 for an outpatient claim of chemotherapy; and the ICD-9-CM V codes V58.1, V66.2, V67.2 in inpatient, outpatient, physician, or DME files were also used.

To look for toxicities associated with ESAs, TEEs were identified by using the following ICD-09 diagnosis codes in inpatient, DME, physician, and outpatient files: Diagnosis code 410 was identified as myocardial infarction (MI) 410; 433-436, 438 as stroke; code 415.1x was identified as pulmonary embolism (PE); codes 451, 453.1, 453.2, and 453.4 were identified as deep vein thrombosis (DVT); and codes 452, 453, 453.0, 453.3, 453.5, 453.6, 453.7, 453.8, and 453.9 were identified as other/unclassified TEEs. The other/unclassified category included codes for various conditions such as portal vein thrombosis and Budd-Chiari syndrome, renal vein thrombosis, thrombosis of veins excluding the pulmonary, cerebral and coronary veins, and embolism or thrombosis of unspecified veins, among others. A comorbidity score was calculated using Klabunde's adaptation of the Charlson comorbidity index from the SAS macro provided by the National Cancer Institute (NCI).16-19 The comorbidities included in the score are myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, diabetes (with and without end-organ damage), chronic pulmonary disease, connective tissue disease, ulcer disease, liver disease, renal disease, hemiplegia, and acquired immune deficiency syndrome (AIDS).

Demographic data was obtained from the SEER-Medicare Patient Entitlement and Diagnosis Summary File. For the census tract variables of education and poverty level, quartiles were calculated in increasing order. Data from the 2000 census were supplemented with 1990 data when 2000 data were missing. Patients were followed from diagnosis date until loss of Medicare coverage, enrollment in a health maintenance organization, or death.

Statistical Analysis

Demographic and tumor characteristics between patients receiving or not receiving ESAs were compared using the chi-square test, or Wilcoxon test as indicated. To determine the factors associated with ESAs use, a multivariate logistic regression model was used to calculate the odds ratio (OR) of receiving ESAs. The variables included in the multivariate logistic regression model included the type of chemotherapy that was used (taxanes, anthracyclines, vinorelbine), age, sex, race, marital status, education level, poverty level, year of diagnosis, geographical location, tumor grade, estrogen receptor status, and comorbidities. The different TEE categories evaluated were MI, stroke, PE, DVT, other/unclassified TEE, and any clot. Any clot was defined as at least 1 event in any of the other previously mentioned categories. For each TEE category, the proportion of patients experiencing the outcome was calculated according to ESAs group. To avoid bias, we excluded the patients that had the outcome under study the year before their breast cancer diagnosis. For each of these outcome variables, we performed a univariate analysis with ESAs in the model. Then for the multivariable logistic regression analysis, we adjusted ESAs by age, race, education level, poverty level, year of diagnosis, geographical region, estrogen-receptor status, comorbidities and transfusion. Results are expressed in odds ratios (OR) with 95% Wald confidence intervals (CIs).

To evaluate for a potential dose effect, a similar multivariate analysis was performed using ESAs as a categorical variable based on the number of doses received per patient (>0-5, 5-16, >16), the reference value was no use for ESAs. When a change in the estimate by dose was seen, a trend was evaluated by introducing the ESA dose category as a numerical variable and testing the significance of the model's linearity in the same multivariate model. All computer programming and statistical analyses were performed with the SAS system (SAS Institute, Cary, North Carolina), and all tests were 2 sided.

According to the NCI regulations, and to preserve confidentiality, in Table 1, thecategories that included less than 11 patients were omitted (educational level unknown, poverty level unknown); however, such variables were still included in our analysis.

Table 1. Patient and Tumor Characteristics According to Erythropoiesis Stimulating Agent Treatment Status
VariableStrataNo ESAsn=1268ESAsn=980Pa
  No. (%)No. (%) 
  • a

    Chi-square P value.

  • b

    Education and poverty levels are categorized by census tract.

Age group, y66-70359 (27.9)338 (34.5)<.0001
 71-75374 (29.2)282 (28.8) 
 76-80306 (24)233 (23.8) 
 >80243 (19)127 (12.8) 
SexFemale1272 (98.9)968 (98.8).764
 Male14 (1.1)12 (1.2) 
RaceWhite1057 (82.2)800 (81.6).928
 Black138 (10.7)110 (11.3) 
 Other91 (7.1)70 (7.1) 
EducationbHighest quartile306 (23.8)263 (26.8).327
 High quartile322 (25.0)242 (24.7) 
 Low quartile323 (25.1)244 (24.9) 
 Lowest quartile335 (26.0)231 (23.6) 
PovertybHighest quartile306 (23.8)261 (26.6).072
 High quartile307 (23.9)261 (26.6) 
 Low quartile335 (26.0)230 (23.5) 
 Lowest quartile338 (26.0)228 (23.3) 
Year of diagnosis1995-1999465 (36.2)155 (15.8)<.0001
 2000-2002422 (32.8)393 (40.1) 
 2003-2005399 (31.0)432 (44.1) 
RegionConnecticut126 (9.8)80(8.2)<.0001
 Detroit153 (11.9)93 (9.5) 
 California+Hawaii337 (26.2)286 (29.2) 
 Iowa135 (10.5)60 (6.1) 
 New Mexico34 (2.6)15 (1.5) 
 Seattle85 (6.6)53 (5.4) 
 Utah41 (3.2)18 (1.8) 
 Atlanta+Rural Georgia60 (4.7)53 (5.6) 
 Kentucky76 (5.9)49 (5.0) 
 Louisiana55 (4.3)76 (7.8) 
 New Jersey184 (14.3)195 (19.9) 
Tumor grade167 (5.2)48 (4.9).736
 2343 (26.7)253 (25.8) 
 3486 (37.8)393 (40.1) 
 unknown390 (30.3)286 (29.2) 
ER statusPositive644 (50.1)497(50.7).009
 Negative285 (22.2)259 (26.4) 
 unknown357 (27.8)224 (22.9) 
Charlson score0990 (77.0)742 (75.7).776
 1194 (15.1)155 (15.8) 
 2+103 (7.9)83 (8.5) 
AnthracyclineNo946 (73.6)592 (60.4)<.0001
 Yes340 (26.4)388 (39.9) 
TaxaneNo865 (67.3)356 (36.3)<.0001
 Yes421 (32.7)624 (63.7) 
VinorelbineNo1204 (93.6)714 (72.9)<.0001
 Yes82 (6.4)266 (27.1) 
Blood transfusionNo1087 (84.5)727 (74.2)<.0001
 Yes199 (15.5)253 (25.8) 


Our final cohort included 2266 patients, 980 (43.3%) of them received ESAs, and 1286 (56.7%) did not; median time of follow-up was 37 months. Patient demographics and tumor characteristics are listed in Table 1. The median time from diagnosis to first dose of chemotherapy was 3 months. Among patients receiving ESAs, median time to first ESAs dose was 5 months, and the median number of doses administered was 9 (standard deviation, 18.3).

When evaluating the factors associated with ESAs use, we observed that the patients who received anthracyclines (OR, 1.85; 95% CI, 1.48-2.31), taxanes (OR, 2.74; 95% CI, 2.25-3.35), and vinorelbine (OR, 4.51; 95% CI, 3.37-6.04) were more likely to receive ESAs than patients who were not treated with such antineoplastic agents. Other factors associated with the use of ESAs were year of diagnosis (2000-2002 vs 1995-1999; OR, 3.06; 95% CI, 2.32-4.04; and 2003-2005 vs 1995-1999; OR, 3.76; 95% CI, 2.83-5.00) and some geographical locations. The complete multivariate analysis is shown in Table 2.

Table 2. Multivariate Analysis of Factors Predicting Erythropoiesis Stimulating Agents Use
EffectOR95% CI
  1. OR indicates odds ratio; CI, confidence interval.

Age, y, 66-70 (reference)
Race, black vs white1.050.74-1.49
Race, other vs white1.00.67-1.48
Year of cancer diagnosis 1995-1999 (reference)
Region California+Hawaii (reference)
 Atlanta+Rural Georgia1.370.87-2.18
 New Jersey0.860.63-1.17
 New Mexico0.410.20-0.83
Tumor grade 1 (reference)
 Tumor grade 20.960.61-1.50
 Tumor grade 30.960.62-1.49
 Tumor grade unknown1.150.74-1.80
Estrogen receptor negative (reference)
 Estrogen receptor positive0.880.68-1.12
 Estrogen receptor unknown0.720.56-0.91
Charlson comorbidity score 0 (reference)
 Charlson comorbidity score 11.040.80-1.36
 Charlson comorbidity score ≥21.210.85-1.71
Anthracycline treatment1.851.48-2.31
Taxane treatment2.742.25-3.35
Vinorelbine treatment4.513.37-6.04
Blood transfusion1.931.52-2.44

The rates of TEEs among patients who received and did not receive ESAs are shown in Figure 1. Higher crude rates of stroke (18.5% vs 15.1%, P = .031), DVT (21.3% vs 14.4%, P < .0001), other/unspecified TEE (19.8% vs 14.7, P = .001), and any clot (31.3% vs 23.4%, P<.0001) were seen in patients who received ESAs. No difference in the rate of MI (P = .989) or PE (P = .523) was observed. The crude and adjusted estimates for the analysis of ESAs use on TEEs are shown in Table 3. After adjusting for possible confounding variables, we observed that patients receiving ESAs had an increased risk of developing a DVT (OR, 1.36; 95% CI, 1.05-1.75) and any clot (OR, 1.26; 95% CI, 1.02-1.57) compared with patients who did not received ESAs. A trend for increased risk was seen for stroke (OR, 1.24; 95% CI, 0.96-1.60) and other/unspecified TEE (OR, 1.11; 95% CI, 0.86-1.44), but statistical significance was not achieved.

Figure 1.

Thromboembolic events (TEEs) among patients who received and did not receive ESAs are shown. The numbers above each histogram bar are percentages. *The chi-square test for the differences in percentage between ESA-treated and non–ESA-treated patients are statistically significant (P < .05).

Table 3. Univariate and Multivariate Analysesa of Different Thromboembolic Events According to Erythropoiesis Stimulating Agents Use
EventESAs TreatmentUnivariateMultivariateESAs DoseEventUnivariateMultivariatePtrend
 Yes (%)No (%)OR (95% CI)OR (95% CI) YesNoOR (95% CI)OR (95% CI) 
  • MI indicates myocardial infarction; PE, pulmonary embolism; DVT, deep-vein thrombosis.

  • a

    Multivariate analysis was adjusted for age, race, education, poverty level, year of cancer diagnosis, region, tumor grade, estrogen receptor status, Charlson comorbidity, treatment of anthracycline, taxane, vinorelbine and blood transfusion.

  • Ptrend is testing for upward or downward trend in the odds ratios as the total doses of ESAs increases.

No925 (94.4)1214 (94.4)ReferenceReference072 (56.7)1214 (56.8)ReferenceReference.905
Yes55 (5.6)72 (5.6)1.00 (0.70,1.44)0.99 (0.65,1.52)1-520 (15.7)335 (15.7)1.01 (0.60,1.68)1.01 (0.59,1.75) 
     5-1615 (11.8)308 (14.4)0.82 (0.46,1.45)0.87 (0.47,1.61) 
     >1620 (15.7)282 (13.2)1.20 (0.72,2.00)1.11 (0.61,2.02) 
No799 (81.5)1092 (84.9)ReferenceReference0194 (51.7)1092 (57.7)ReferenceReference.014
Yes181 (18.5)194 (15.1)1.28 (1.02,1.59)1.24 (0.96,1.60)1-555 (14.7)300 (15.9)1.03 (0.75,1.43)1.04 (0.74,1.46) 
     5-1660 (16.0)263 (13.9)1.28 (0.93,1.77)1.29 (0.91,1.83) 
     >1666 (17.6)236 (12.5)1.57 (1.15,2.15)1.54 (1.08,2.21) 
No891 (90.9)1179 (91.7)ReferenceReference0107 (54.6)1179 (57.0)ReferenceReference.659
Yes89 (9.1)107 (8.3)1.10 (0.82,1.48)1.00 (0.71,1.40)1-529 (14.8)326 (15.7)0.98 (0.64,1.50)0.91 (0.58,1.42) 
     5-1629 (14.8)294 (14.2)1.09 (0.71,1.67)0.99 (0.62,1.58) 
     >1631 (15.8)271 (13.1)1.26 (0.83,1.92)1.15 (0.71,1.86) 
No771 (78.7)1101 (85.6)ReferenceReference0185 (47.0)1101 (58.8)ReferenceReference.000
Yes209 (21.3)185 (14.4)1.61 (1.30,2.01)1.36 (1.05,1.75)1-551 (12.9)304 (16.2)1.00 (0.71,1.40)0.90 (0.63,1.29) 
     5-1672 (18.3)251 (13.4)1.71 (1.26,2.32)1.51 (1.08,2.12) 
     >1686 (21.8)216 (11.5)2.37 (1.77,3.18)2.01 (1.43,2.84) 
Other clot
No786 (80.2)1097 (85.3)ReferenceReference0189 (49.3)1097 (58.3)ReferenceReference.010
Yes194 (19.8)189 (14.7)1.43 (1.15,1.79)1.11 (0.86,1.44)1-546 (12.0)309 (16.4)0.86 (0.61,1.22)0.75 (0.52,1.08) 
     5-1668 (17.8)255 (13.5)1.55 (1.14,2.11)1.26 (0.89,1.77) 
     >1680 (20.9)222 (11.8)2.09 (1.55,2.82)1.56 (1.10,2.21) 
Any clot
No673 (68.7)985 (76.6)ReferenceReference0301 (49.5)985 (59.4)ReferenceReference.000
Yes307 (31.3)301 (23.4)1.49 (1.24,1.80)1.26 (1.02,1.57)1-580 (13.2)275 (16.6)0.95 (0.72,1.26)0.86 (0.64,1.16) 
     5-16107 (17.6)216 (13.0)1.62 (1.24,2.11)1.44 (1.07,1.92) 
     >16120 (19.7)182 (11.0)2.16 (1.66,2.81)1.82 (1.34,2.47) 

To explore whether a dose effect was associated with the development of TEEs, among patients who received ESAS, we categorized the number of doses as follows: >0-5, ≥5-16, and ≥16. When such categories were entered in the multivariate model, we observed an association between higher number of doses and an increased risk of stroke (Ptrend = .014), DVT (Ptrend < .0001), other/unclassified TEE (P = .010), and any clot (Ptrend < .0001). In Figure 2, we show the plot of the estimated odds ratios (OR) to develop a TEE according to different ESA doses.

Figure 2.

Plot of estimated odds ratios for developing thromboembolic events is shown according to different ESA dose distribution. The x axis is the distribution of accumulated ESA units. The y axis is the estimated odds ratio from the univariate logistic regression for different toxicity events. The 3 horizontal lines are odds ratios = 1, 2, 3; the 2 vertical lines are the cutoff ESA units at 5 and 16.


Our study shows that, in patients aged older than 66 years with metastatic breast cancer receiving chemotherapy, the use of ESAs increases the risk of TEEs, specifically DVTs and any clot. The magnitude of the increase was notable. Patients treated with ESAs had an absolute increase in the rates of clots of almost 10% compared with patients not treated with ESAs. A dose-dependent effect for stroke, DVT, other/unspecified TEE, and any clot was seen, such that the risk was increased for patients who received more than 5 doses of ESAs, and the risk was even higher in those receiving more than 16 doses. There was no apparent increase in risk for patients who received only 1-4 doses.

Our results are consistent with previously published data. In a recent meta-analysis that included 51 phase 3 clinical trials (n = 8172),5 the risk of TEE (HR, 1.57; 95% CI, 1.31-1.8) was higher among patients treated with ESAs. This estimate is higher than what we report, but it is important to mention that in this meta-analysis, patients with different malignancies were included and also treatment and target hemoglobin varied among trials, as also did the TEE definitions. Aapro et al20 performed a meta-analysis that included 12 trials and a total of 2297 patients with nonmyeloid haematological malignancies and different solid tumors. The analysis was stratified for a target hemoglobin of ≤11 g/dL, corresponding to current European Organization for Research and Treatment of Cancer guidelines.21 An increased frequency of TEEs was seen with epoetin-beta versus control (7% vs 4%); however, TEE-related mortality was similar in both groups.20 In our analysis, the absolute increase in risk of TEEs was more pronounced, possibly because our population was older and our study cohort was not limited to patients on clinical trials, who presumably would be healthier.

Hershman et al22 very recently reported data on an observational study using the SEER-Medicare database. They included patients receiving chemotherapy who were diagnosed with diffuse B-cell lymphoma, colon, or breast and lung cancer in different stages. They observed that the use of ESAs was associated with more recent diagnosis, younger age, comorbidities, female sex, and metastatic or recurrent cancer. From the 14,024 patients with breast cancer, 27.4% received ESAs, and 12% had an episode of thrombosis. In the analysis of the complete cohort, they observed higher rates of thrombosis in the patients who received ESAs (HR, 1.93; 95% CI, 1.79-2.07).22 The risk estimates that we found are in same direction but of less magnitude, perhaps associated with our more homogeneous patient population.

We did not observe an increased risk of MI in patients receiving ESAs. This association has not been consistently reported in cancer patients; however, it has been reported in the renal literature.23, 24 In our study, the absolute number of patients with a stroke was higher among patients receiving ESAs. In the multivariate model, there was a nonsignificant trend for increased risk of stroke in patients treated with ESAs, and we observed a significant trend in the number of doses and the risk of developing stroke. This observation has not been reported in the oncological literature. However, higher rates of fatal and nonfatal stroke have been seen in diabetic patients with ESRD who were receiving darbepoetin (HR, 1.92; 95% CI, 1.38-2.68).25 In approximately 30% of patients with ESRD treated with ESAs, hypertension develops or worsens,23 and this phenomenon has been linked to a higher risk of vascular events.23, 26 Further studies are warranted to define the cardiovascular and cerebrovascular events associated with ESA use in cancer patients.

It has been suggested that hemoglobin level may relate to the risk of TEEs, with greater risk associated with higher hemoglobin levels.27 Some of the trials where an excess of TEEs was reported, targeted hemoglobin levels were higher than those in enrolled patients who were not anemic at baseline or targeted hemoglobin levels higher than those recommended by current ESA labeling. In the Breast Cancer Erythropoietin Survival Trial (BEST) study,2 patients with metastatic breast cancer treated with chemotherapy were randomized to receive erythropoietin or not. Similar rates of TEE were seen between groups (16% vs 14%). The trial was stopped because of increased mortality in the group treated with erythropoietin; a subsequent chart review revealed an increased number of TEE-associated deaths in those who received ESAs (14 vs 4). Patients who were not anemic were targeted to hemoglobin levels of 12 g/dL to 14 g/dL. The target was maintained for 59% of patient weeks in the treatment arm, compared with 45% of patient weeks in the placebo group (P < .001).2 In a similar group of patients, The Breast Cancer—Anemia and the Value of Erythropoietin (BRAVE) trial8 evaluated whether erythropoietin use improved survival in patients with metastatic breast cancer. A total of 463 patients treated with anthracyline- and/or taxane-based chemotherapy were randomized to receive 30,000 U of erythropoietin beta or control once weekly for 24 weeks. After 18 months, no differences in overall survival (P = .522) or progression-free survival (P = .448) were seen; however, patients treated with ESAs experienced more TEEs than controls (13% vs 6%, P = .012), with no difference in the rate of serious TEEs (4% vs 3%).8 In a different patient population, Henke et al28 attempted to sensitize head and neck tumors to radiation therapy by randomizing patients to receive erythropoietin or placebo and targeted a hemoglobin level of 12 g/dL to 14 g/dL in women and 13 g/dL to 15 g/dL in men. Vascular events defined as hypertension, hemorrhage, DVT, and PE were seen in 11% of patients in the treatment arm and in only 5% of the patients randomized to placebo. The number of patients who died from cardiac disorders was higher in the treatment arm (n = 10 vs 5). Unfortunately, in the SEER-Medicare dataset, no information was available on patient hemoglobin, so it was not possible to evaluate the impact of hemoglobin levels.

In our study, we found that patients who received more than 5 doses of ESAs had a higher risk of developing a stroke, a DVT, other/unspecified TEE, and any clot. Contrary to our results, Rosenzweig et al10 found no relation with the number of doses or the mean number of weeks a patient received erythropoietin. This trial was stopped early because 28.5% of the patients in the treatment arm developed a TEE, compared with no event at all in the control arm.

Recently, Khorana et al29 identified an association between red blood cell transfusions and an increased risk of TEEs. In a large, retrospective, cohort study in patients with cancer, they observed that transfusions increased the risk of venous thromboembolism (OR, 1.53; 95% CI, 1.46-1.61). To avoid having our risk estimates confounded by this association, we included transfusion in our regression model.

To the best of our knowledge, this is the largest study, including exclusively patients with metastatic breast cancer receiving chemotherapy, in which an assessment of the patterns of ESA use and the relation between ESAs and different TEEs is made. One of the strengths of this study is that it involves a large, population-based cohort of older patients; therefore, it includes patients who may be excluded from randomized clinical trials, possibly reflecting real clinical practice. It is important to mention that in our cohort of patients, we observed a high incidence of TEEs. Independent of whether patients received ESAs or not, we believe that this study includes a high-risk group. Old age, comorbidities, and that all of the patients had metastatic disease and were receiving chemotherapy likely contributed to the high incidence of TEEs observed. In addition, we defined our outcomes by the presence of at least 1 diagnosis code specifically for TEE, which could have resulted in an overdiagnosis of the outcomes. However, if misclassification occurred, it would not have differed between our study groups; thus, misclassification is not likely to have caused any change in our risk estimates.

A limitation of our study is that the SEER-Medicare data do not allow assessment of the extent of the disease, the severity of outcomes, or for an analysis that takes into account the patient's hemoglobin level, which, as previously mentioned, has been related with adverse outcomes. It is possible that factors such as a large volume of disease or genetic predisposition may impact TEE incidence, but we were not able to adjust for such factors. One will expect, however, that these potential confounders were present in similar proportion in patients who received and who did not receive ESAs. It is possible that patients receiving ESAs were not captured in Medicare claims; however, because of the expense of the medication and the exclusion of patients with secondary insurance, we believe is unlikely this occurred in a significant number of patients. A minor limitation in our study is that our cohort exclusively included patients aged 66 years and older. The results, thus, may not be applicable to a population of younger, and in general, healthier patients. Additional studies are needed to prospectively confirm these findings and to assess the complications and toxicities associated with the use of ESAs in younger women with metastatic breast cancer. It is important to mention that observational studies cannot establish causality. There are data from randomized trials proving the association between the use of ESAs and TEEs. However, our study cannot rule out that the reason for ESA administration also placed patients at higher risk of developing a TEE.

In summary, we have shown that in this cohort of patients, the use of ESAs increased the risk of certain TEEs. Our data show that among patients treated with chemotherapy and ESAs for metastatic breast cancer, TEEs are a common event. Therefore, ESAs should be used for the minimum necessary time to reduce the risk of complications.


This work was supported by National Institutes of Health grant NIH K07-CA109064 (Giordano).