The results of randomized clinical trials have suggested that after receiving radiotherapy and/or chemotherapy, patients with primary breast carcinoma have an increased risk of developing leukemia. In the current study, the authors set out to assess the reported association between breast carcinoma treatment and leukemia risk.
A registry of all patients with breast carcinoma who were treated at a community-based institution since 1989 (updated annually for recurrence and/or vital status) was linked to the National Cancer Institute Surveillance, Epidemiology, and End Results database to confirm complete ascertainment of leukemia cases occurring within this registry population. Incidence rates were calculated for women who were treated for primary Stage 0–III breast carcinoma and had a follow-up duration of ≥ 24 months (n = 2866). Patients who did not undergo surgery (n = 5), patients for whom chemotherapy records were incomplete or who received nonstandard chemotherapy regimens (n = 69), patients who underwent stem cell transplantation (n = 83), and patients who were lost to follow-up or who had unknown disease status at follow-up (n = 81) were excluded from the analysis (total, n = 238).
Among patients diagnosed with breast carcinoma between 1992 and 1999, the crude overall leukemia incidence rate was 0.28%, and the acute myelogenous leukemia (AML)/myelodysplastic syndrome (MDS) incidence rate was 0.11%. The average follow-up duration was 5.46 years (minimum, 2 years). Eight incident cases of leukemia were documented (2 cases of AML, 1 case of acute lymphoblastic leukemia, 1 case of MDS/refractory anemia with excess blasts, 2 cases of chronic myelogenous leukemia, and 2 cases of chronic lymphocytic leukemia). National age-adjusted overall leukemia incidence rates for the period 1996–1998 predict the occurrence of 9 cases (incidence rate, 0.31%) in the current cohort of women ages 21–94 years. The incidence of leukemia by treatment category was as follows: no surgery/no chemotherapy/no radiotherapy, 2 of 154 patients (1.30%); surgery/no chemotherapy/radiotherapy, 4 of 1403 patients (0.29%); surgery/chemotherapy/no radiotherapy, 0 of 352 patients (0%); and surgery/chemotherapy/radiotherapy, 2 of 957 patients (0.21%).
Potential associations between leukemia risk and various adjuvant treatment regimens (chemotherapy and/or radiotherapy) have been evaluated in a number of different cohorts of patients with breast carcinoma. Although a clear relation between the use of older alkylating agents (e.g., melphalan) and acute myelogenous leukemia (AML) has been established, debate regarding the possible role of cyclophosphamide and anthracyclines in the development of leukemia is ongoing.1–3 In particular, Smith et al.4 recently found that leukemia risk increases with increasing cyclophosphamide dose, and Diamandidou et al.5 reported a 1.5% probability of developing leukemia (AML, myelodysplastic syndrome [MDS], or chronic myelogenous leukemia [CML]) for patients receiving fluorouracil plus doxorubicin plus cyclophosphamide (FAC)-style regimens. Findings such as these have generated concern regarding long-term patient health, given the increased efficacy of current chemotherapeutic agents in the treatment of breast carcinoma.6 In fact, a 2000 National Cancer Institute (NCI) consensus statement regarding the adjuvant treatment of breast carcinoma acknowledges the association between anthracycline-based therapy and elevated leukemia risk.7
In view of this debate regarding anthracycline-containing regimens and chemotherapy in general, we examined the incidence of all types of leukemia, including acute leukemia, in a community-based cohort of patients treated for breast carcinoma at the Swedish Cancer Institute (Seattle, WA) between 1989 and 2001.
MATERIALS AND METHODS
A registry of all patients treated for breast carcinoma at our community-based institution was created in 1989 and contains detailed information on diagnosis, staging, surgery, chemotherapy use, and radiotherapy use. Only patients diagnosed on or before December 31, 1999, were included in the study cohort; thus, all included patients had a follow-up duration of at least 2 years. Registry follow-up data were updated annually, with data on recurrence, subsequent treatment, and vital status being current through 2001. Vital and disease status data were obtained via chart review if the patient was still attending our institution or via a physician-directed follow-up letter if follow-up care was being provided elsewhere. Patients who were not under the care of a managing physician were contacted by mail with an institutional review board (IRB)-approved letter from the diagnosing physician requesting annual follow-up information. If no response was received, our institution's cancer registry and the NCI Surveillance, Epidemiology, and End Results (SEER) database were reviewed for information on vital and disease status. Our IRB-approved breast carcinoma registry was maintained in a manner that was compliant with the Health Insurance Portability and Accountability Act and was protected by password. All patient identifiers were removed and replaced with a nonlinked assigned number before any type of analysis was performed.
Data on the overall registry population, including patients who eventually were excluded, were matched with SEER registry data to confirm complete ascertainment of leukemia cases. All cases of leukemia found following a diagnosis of primary breast carcinoma were consistent with SEER data regarding disease status, and no additional cases were found in the SEER database. Patients who did not undergo surgery (n = 5), patients for whom chemotherapy records were incomplete or who received nonstandard chemotherapy regimens (n = 69), patients who underwent stem cell transplantation (n = 83), and patients who were lost to follow-up or who had unknown disease status at follow-up (n = 81) were excluded from the analysis (total, n = 238). The analytic data set included women who were treated for AJCC (1998) Stage 0–III primary breast carcinoma and who had a minimum follow-up duration of 24 months (n = 2866).
Crude cumulative incidence rates were calculated for the population as a whole and also according to disease stage and according to type of therapy received. Observation began on the date of diagnosis. Censoring occurred on the date of leukemia diagnosis, the date of death, or the date of last follow-up. Expected numbers of leukemia cases were calculated using rates (published by the American Cancer Society) that were based on National Cancer Institute incidence data.8 Incidence rates for leukemia (any type) and for specific subtypes of leukemia were applied to the study population and corrected for age if applicable. Chemotherapy regimens were classified as doxorubicin-containing or non-doxorubicin-containing regimens for the purposes of the analysis.9–12
Cumulative 10-year incidence rates were calculated with adjustment for survival using Kaplan–Meier plots.13 Expected cumulative incidence rates at 5 and 10 years were estimated using the product-limit (Kaplan–Meier) method.14 Ninety-five percent confidence intervals (CIs) for cumulative incidence were calculated using the bootstrap method15; use of the arcsine and logarithmic methods yielded highly similar results. Incidence per person-year was calculated as the number of events divided by the total number of person-years of follow-up, and exact 95% CIs were obtained using the Poisson distribution.
The average age of patients in the current all-female cohort was 57 years (range, 21-94 years). Disease stages observed at diagnosis included Stages 0–IIIB, with approximately half of all patients being diagnosed with Stage I primary breast carcinoma (Table 1). Approximately three-quarters of all tumors had positive estrogen receptor status, and 60% had positive progesterone receptor status. Eighty-two percent of patients received radiation therapy, 46% received chemotherapy, and 57% received tamoxifen therapy. Of the 423 patients who did not receive doxorubicin, 413 received standard cyclophosphamide, methotrexate, and fluorouracil (CMF) according to the method described by Bonadonna et al.11 or the method described by the Southwest Oncology Group12; of the remaining 10 patients, 1 received paclitaxel, 3 received paclitaxel plus carboplatin, and 6 received cyclophosphamide plus docetaxel. The average duration of follow-up was 5.4 years (maximum, 12.3 years). The majority of patients (86%) were alive with no evidence of disease (NED) in 2001. The 10-year recurrence-free survival rate was 90% for patients with Stage I disease, 75% for patients with Stage IIa or IIb disease, and 34% for patients with Stage IIIa or IIIb disease (Fig. 1). The number of cases diagnosed per year increased steadily over time, with 40% of the cohort being accrued between 1997 and 1999 (n = 1160) (Table 2). The total number of person-years analyzed was 15,665, with patients diagnosed in 1991 accounting for the greatest number of person-years (2049).
Table 1. Descriptive Statistics (n = 2866)
No. of patients (%)
NED: no evidence of disease.
Mean age in yrs (range)
Positive estrogen receptor status
Positive progesterone receptor status
Mean follow-up in yrs (range)
Alive with original malignancy
Alive with other malignancy
Died with original malignancy
Died with other malignancy
Table 2. Person-Years and Mean Years to Censoring or Leukemia Diagnosis (n = 2866)
Yr of diagnosis
Person-yrs of follow-up
Mean yrs of follow-up
No. of patients (%)
Between 1992 and 1999, 8 incident cases of leukemia were documented (2 cases of AML, 1 case of acute lymphoblastic leukemia [ALL], 1 case of MDS/refractory anemia with excess blasts [RAEB], 2 cases of CML, and 2 cases of chronic lymphocytic leukemia), for a crude incidence rate of 0.28%. The crude MDS/AML incidence rate was 0.11% (3 of 2866 patients). Using Kaplan–Meier one-minus-survival curve estimates, the 10-year cumulative incidence rate was calculated to be 0.51% (95% CI, 0.001–1.44%) for MDS/AML (Fig. 2) and 0.81% (95% CI, 0.19–1.84%) for leukemia overall (Fig. 3). The 10-year cumulative incidence rate for patients receiving chemotherapy (n = 1309) was 0.19% (95% CI, 0.018–0.55%). It should be noted that the two chemotherapy patients who developed AML posttreatment had the second cancer occur during the first 5 years after treatment, and thus the observed 5-year and 10-year incidence rates were identical. The rate of incidence per 1000 person-years was 0.19% (95% CI, 0.04–0.56%) for MDS/AML and 0.51% (95% CI, 0.22–1.01%) for leukemia overall. The rate of AML incidence per 1000 person-years for patients receiving chemotherapy was 0.29% (95% CI, 0.035–1.04%).
The time from primary breast carcinoma diagnosis to leukemia diagnosis (any type) ranged from 3 months to 10 years (median, 37 months) (Fig. 3). Two of the 8 patients who developed leukemia had been treated for recurrent breast carcinoma—one (who developed CML) had been treated with radiation alone, and the other (who developed MDS/RAEB) had been treated with chemotherapy (Abeloff chemotherapy and 2 cycles of cyclophosphamide, etoposide, and cisplatin [CEP]), radiotherapy, and tamoxifen.9, 10 Calculation of the time to development of leukemia from the date of recurrence instead of the date of initial diagnosis changed the time to diagnosis of leukemia from 33 months to 8 months for one of these two patients and from 120 months to 96 months for the other. The remaining six patients had no evidence of breast carcinoma at the time of leukemia diagnosis. Cytogenetic data were available for all cases of AML/MDS and are presented in Table 3. Abnormalities at chromosome 5 and/or chromosome 7 were present in one of the two AMLs (patient 1) and in the lone case of MDS (patient 4). These two patients who had chromosomal abnormalities had short durations of survival from the time of leukemia diagnosis. It is noteworthy that the patient who had AML with no abnormality at chromosome 5 or chromosome 7 (patient 2;Table 3) remains alive with disease in first complete remission at the time of the current report (survival duration, 76 months as of June 2003).
Patient 4 had a breast carcinoma recurrence 25 months after her initial diagnosis and was subsequently treated with Abeloff chemotherapy and cyclophosphamide, etoposide, and cisplatin (2 cycles) plus radiotherapy and tamoxifen.
Time from secondary treatment.
Patient 5 had a breast carcinoma recurrence 24 months after her initial diagnosis and was subsequently treated with localized radiotherapy.
CAF × 6
Pt. 1: Cytogenetic findings: inv(3)(q11.2-q26) and missing chromosome 7
CMF × 6
Alive as of 12/2001
Pt. 2: Cytogenetic finding: t(16;16)(p13.1;q22) abnormality
The distribution of leukemia cases by treatment category was as follows: surgery/chemotherapy/no radiotherapy, 0 of 352 patients (0%); surgery/chemotherapy/radiotherapy, 2 of 957 patients (0.21%); surgery/no chemotherapy/no radiotherapy, 2 of 154 patients (1.30%); and surgery/no chemotherapy/radiotherapy, 4 of 1403 patients (0.29%) (Table 4). When only Stage II–III breast carcinoma was considered, there were 3 cases of leukemia documented in 1259 patients (0.24%).
Table 4. Comparison of Studies Investigating the Incidence of Leukemia Following Breast Carcinoma
There was 1 case of AML documented in the doxorubicin-containing chemotherapy (cyclophosphamide, doxorubicin, and fluorouracil [CAF])/radiotherapy subgroup (1 of 886 patients [0.11%]) and another case documented in the non-doxorubicin-containing chemotherapy (CMF)/radiotherapy subgroup (1 of 423 patients [0.24%]). MDS/RAEB was found in one patient (Patient 4) who was treated with surgery alone at diagnosis and later received doxorubicin-containing chemotherapy and a high-dose cyclophosphamide regimen plus radiotherapy for a breast carcinoma recurrence (Table 3).
An increased risk of leukemia (any type) was not observed among patients treated with surgery, chemotherapy, and radiotherapy compared with patients treated with surgery and radiotherapy; however, the three documented cases of MDS/AML were found exclusively in patients who were treated with both chemotherapy and radiotherapy either initially or for recurrent disease. Two of these three patients were treated with doxorubicin-containing chemotherapy.
National age-adjusted leukemia incidence rates (any type) for women during the period 1996–1998 have been collected and are published annually.8 Using these national age-specific rates, we calculated a total of 8.8 expected cases of leukemia in the current cohort of patients ages 21–94 years who had varying durations of follow-up; the actual number of cases observed was 8 (Table 5).
Table 5. Crude Age-Specific Leukemia Rates (All Cases through 2001)
Leukemia rate (%)
No. of patients (%)
The objective of the current prospective cohort study was to evaluate leukemia incidence in a community-based population of patients with primary breast carcinoma treated with surgery alone; surgery and chemotherapy; surgery and radiotherapy; or surgery, chemotherapy, and radiotherapy and compare these rates to those expected in the general population (adjusted for age and gender) using currently available national statistics. In addition, we evaluated the difference between doxorubicin-containing and non-doxorubicin-containing chemotherapy regimens. The results of the study suggest that there is little, if any, added risk of leukemia in association with the use of either doxorubicin-containing or non-doxorubicin-containing regimens, regardless of whether radiotherapy is also received. Nonetheless, AML and MDS/RAEB were observed exclusively in patients who received both chemotherapy and radiotherapy.
The risk of developing leukemia after treatment for breast carcinoma is not well characterized. Major U.S. studies of leukemia risk have used SEER data or follow-up data from randomized clinical trials (Table 4).2, 4, 5, 16, 17 Collectively, these studies contain data on patients treated between 1971 and 2001. Three of these five studies (Fisher et al.,2 Curtis et al.,16 and Smith et al.4) reported a statistically significant association between chemotherapy (alone or in combination with radiotherapy) and leukemia risk. Diamandidou et al.5 reported that leukemia risk was increased specifically in patients receiving doxorubicin and radiotherapy. Furthermore, 4-epidoxirubicin was found to be leukemogenic in a prospective cohort study (1987–1990) of 360 patients with breast carcinoma.18 The Manitoba Cancer Registry study (1970–1985) also reported an increased risk of developing leukemia among patients receiving chemotherapy (CMF or single-agent melphalan).3 In addition, a high risk of leukemia was found to be associated with the use of radiotherapy plus mitoxantrone in patients diagnosed with and treated for breast carcinoma between 1982 and 1996, with a much smaller increase observed in the non-mitoxantrone-containing chemotherapy/radiotherapy group.19
Previous studies of alkylating agent-based adjuvant chemotherapy regimens have revealed an increased risk of acute leukemia when such regimens contain melphalan.1–3 Curtis et al.,20 in a study in which oral cyclophosphamide generally was administered once daily, suggested that the risk of acute leukemia is not increased in patients receiving adjuvant breast carcinoma treatment when cyclophosphamide doses are < 20 g. The total cyclophosphamide dose administered in the standard CMF regimen described by Bonadonna and colleagues was 8.4 g/m2, and in the standard CMFVP regimen, the total dose was 10.8 g/m2.14, 15 Therefore, it is unlikely that a patient would receive a total cyclophosphamide dose in excess of 20 g if either of these two regimens was used. Notably, Diamandidou et al.5 reported the existence of a trend toward increasing leukemia risk in association with intravenous bolus cyclophosphamide doses in excess of 6 g/m2 in patients treated with various types of FAC regimens, which ranged in length from 6 to 24 cycles. Smith et al.4 observed an increased risk of leukemia in association with dose intensification of intravenous bolus cyclophosphamide (4.8–9.6 g/m2), while Laughlin et al.21 reported an elevated risk of leukemia in association with high-dose, bone marrow-ablative chemotherapy involving alkylating agents in patients with breast carcinoma who were undergoing autologous bone marrow transplantation. Neither Smith and colleagues nor Diamandidou and colleagues found an association between the amount of doxorubicin administered and increased leukemia risk.4, 5 In the current study, cyclophosphamide doses in patients receiving AC, CAF, and Abeloff9 chemotherapy ranged from 2.4 g/m2 (AC) to 3 g/m2 (CAF) to 5.6 g/m2 (Abeloff)9, although it should be noted that the patient who developed MDS/RAEB received 2 cycles of CEP in addition to 1 course of the Abeloff9 regimen, increasing her total cyclophosphamide dose to 13.6 g/m2. This patient is the only one in the current study who received alkylating agent doses comparable to those discussed by Smith et al.4 Primary chemotherapy regimens in the current study, as well as in the study conducted by Smith et al., did not exceed six cycles.
Cytogenetic analysis has revealed an association between abnormalities at chromosome 5 and chromosome 7 and alkylating agent-induced leukemia.22 Topoisomerase inhibitor-induced leukemia has been associated with translocations of 11q23, 21q22, and 3q23.18, 22–25 No cases of leukemia with these topoisomerase inhibitor-induced translocations were documented in the current study or in the study conducted by Diamandidou et al.5; however, in the latter series, 6 of the 10 patients with leukemia for whom chromosomal data were available had abnormalities at chromosome 5 and/or chromosome 7. Two leukemias in the current study (one AML and one MDS/RAEB) had abnormalities at chromosome 5 or chromosome 7.
Diamandidou et al.5 estimated that the 10-year rate of incidence of acute leukemia in patients receiving FAC was 1.5%. Neither our data nor the data reported by Smith et al.4 agreed with this finding when standard doses of cyclophosphamide were used. With respect to crude incidence rates, Diamandidou et al.5 reported the highest figure (0.95%), followed by Smith et al.4 (0.50%) and the current study (0.28%). Armitage et al.26 discuss the difficulties associated with comparing studies that use different methods to calculate incidence rates. With respect to leukemia following breast carcinoma treatment, we have compiled our results (Table 4) and calculated cumulative incidence rates, cases per total number of person-years, estimated rates, and rates per 1000 person-years. Unlike Curtis et al.,16 we did not calculate relative risk estimates. Nonetheless, all methods used in the current study yielded leukemia rates that were lower than those reported in other studies. Also notable is that the 95% CIs reported in the current study for chemotherapy recipients do not overlap with the corresponding CIs reported by Diamandidou et al.5
Although topoisomerase inhibitors are known to be leukemogenic, experience in Hodgkin disease suggests that these agents do not pose as serious a problem as do alkylating agents.22 In the treatment of Hodgkin disease, the mechlorethamine, vincristine, procarbazine, and prednisone (MOPP) regimen is clearly leukemogenic, with peak incidence occurring in the 5–10-year range.27 Delwail et al.28 reported that the 15-year risk of developing acute leukemia was 3.4% after MOPP chemotherapy and 1.3% after receiving doxorubicin, bleomycin, vincristine, and dacarbazine (ABVD); the corresponding risk rates for AML were 3.4% and 0.7%, respectively. Among patients receiving ABVD, there were two cases of ALL with an 11q23 translocation, an abnormality associated with topoisomerase inhibitor-induced leukemia.18, 22–24 Thus, although doxorubicin-containing chemotherapy does appear to be leukemogenic when used to treat patients with Hodgkin disease, it appears to be less so compared with alkylating agent-based chemotherapy.
Radiotherapy has long been known to be leukemogenic.29 Experience in patients with Hodgkin disease has demonstrated that the addition of radiotherapy to chemotherapy dramatically increases the risk of leukemia in that setting28; radiation doses and field sizes also may influence this risk.28, 30 Delwail et al.28 recently reported a 2.4% leukemia risk in patients receiving MOPP plus limited-field radiotherapy and a 13.9% risk in patients receiving MOPP plus extended-field radiotherapy. Although the current investigation did not reveal an increased risk of leukemia of any type in any group of patients, it is noteworthy that both Diamandidou et al.5 and Smith et al.4 suggest the existence of a relation between concomitant radiotherapy and chemotherapy and leukemia risk in patients receiving adjuvant breast carcinoma treatment. Diamandidou and colleagues encountered 14 patients with leukemia, 12 of whom had received radiotherapy plus chemotherapy; overall, patients who received both types of treatment had a 5-fold-increased risk relative to patients who had received chemotherapy without radiotherapy (Table 4). In the current study, both patients who developed AML received radiotherapy in addition to chemotherapy, as did the patient who developed MDS/RAEB after treatment for recurrent breast carcinoma. In fact, no patient who received chemotherapy without radiotherapy during the study period developed any type of leukemia, and seven of the eight patients who developed any type of leukemia received radiotherapy.
The length of follow-up in studies of leukemia incidence following breast carcinoma treatment varies considerably. In the current study, the average follow-up duration was 5.4 years, with a median follow-up duration of 4.8 years for all patients and a minimum follow-up duration of 24 months. Among patients who received doxorubicin, the median follow-up duration was 4.2 years, and among those who received CMF chemotherapy, the median length of follow-up was 6.6 years. The median follow-up duration of the studies reviewed by Diamandidou et al.5 ranged from 5.5 years to 12 years; the median time from treatment to the onset of leukemia was 66 months (range, 22–113 months), although among patients with AML, CML, or MDS/RAEB who received chemotherapy, the median time to onset was 36 months (range, 31–120 months). Smith et al.4 reported a median time of 38 months to the onset of MDS/AML (range, 1 month–10 years); the minimum follow-up duration was 27 months, with the median follow-up duration across all trials reviewed ranging from 5.8 years to 11.5 years. In a review conducted by Van Leeuwen and Travis,30 the risk of alkylating agent/radiation-induced leukemia was reported to increase in the first 1–2 years following treatment and to peak at 5–10 years posttreatment. A shorter induction period was observed for topoisomerase inhibitor-induced leukemia, with a median time of 2 years to onset.29, 30 It appears unlikely that an increased risk of chemotherapy-induced leukemia, and particularly doxorubicin-induced leukemia, would have been overlooked in the current study, in which all patients had at least 2 years of follow-up, but we cannot exclude the possibility that additional years of follow-up could yield additional cases.
In conclusion, we did not observe an increased risk of any type of leukemia in patients receiving standard-dose doxorubicin and cyclophosphamide compared with patients receiving other regimens. In addition, none of the observed cases of leukemia harbored chromosomal abnormalities associated with topoisomerase inhibitors. The risk of leukemia in the current cohort was comparable to the age-adjusted risk observed among women nationally. This finding is consistent with recent data obtained by Smith et al.,4 who focused on the influence of higher doses of cyclophosphamide on leukemia rates following breast carcinoma treatment. Whether the risk increase reported by Diamandidou et al.5 can be explained by differences in the length and intensity of chemotherapy (particularly alkylating agent-based chemotherapy), the type and intensity of concomitant radiotherapy, or the length of follow-up is unknown. In light of the findings reported by Smith and colleagues, it is likely that any observed increase in risk is attributable to alkylating agent dose intensification and perhaps to concomitant radiotherapy use, rather than to the use of doxorubicin. The equivocal risk of leukemia associated with current standard treatment regimens should not serve as a deterrent to the use of these regimens to prevent breast carcinoma recurrence.