Fax: (313) 874-6656
Economic burden associated with breast cancer recurrence
Findings from a retrospective analysis of health system data
Article first published online: 29 MAR 2006
Copyright © 2006 American Cancer Society
Volume 106, Issue 9, pages 1875–1882, 1 May 2006
How to Cite
Lamerato, L., Havstad, S., Gandhi, S., Jones, D. and Nathanson, D. (2006), Economic burden associated with breast cancer recurrence. Cancer, 106: 1875–1882. doi: 10.1002/cncr.21824
- Issue published online: 18 APR 2006
- Article first published online: 29 MAR 2006
- Manuscript Accepted: 21 NOV 2005
- Manuscript Revised: 21 OCT 2005
- Manuscript Received: 14 JUL 2005
- AstraZeneca (Macclesfield, United Kingdom)
- cost of care;
- early breast cancer;
- cohort study;
- economic burden
The economics of breast cancer recurrence are poorly understood. For this retrospective cohort study, the authors evaluated the economic burden of breast cancer recurrence by using data from a large Midwestern healthcare system.
Women with Stage I or II breast cancer (according to the American Joint Commission on Cancer staging criteria) were identified from the tumor registry of the Henry Ford Health System. The economic burden of breast cancer recurrence was estimated from patient charges (adjusted to 2003 U.S. dollars).
From 1996 to 2002, 1616 patients with early breast cancer were identified, including 192 patients who had recurrent tumors. Patients with recurrence had significantly greater charges in the 6-month and 12-month postrecurrence periods ($45,855 and $79,253, respectively) compared with the 6-month and 12-month prerecurrence periods ($10,715 and $12,344, respectively; both P <.001). This was evident for all recurrence types (locoregional, contralateral breast, and distant), but it was most evident for distant recurrences. In a regression analysis that was adjusted for baseline characteristics, the mean monthly charges were significantly greater for patients with recurrence versus patients without recurrence (P <.001), and this was true for each recurrence type. For women with recurrence (n = 74 patients), the mean charges during the 6-month postrecurrence period were significantly greater than mean charges during the initial 6-month period after diagnosis ($50,355 vs. $38,254; P <.01). Quarterly charges for continuing care postrecurrence were significantly greater than prerecurrence charges ($4934 vs. $1825; P <.001). The mean charges for terminal care were significantly greater (P <.01) for women with recurrence (n = 27 patients, $63,434) versus women without recurrence (n = 65 patients, $53,872).
Patients with early breast cancer who experienced recurrence required more costly care than patients who did not develop recurrent disease. Therapies that reduce the risk of recurrence may reduce costs significantly. Cancer 2006. © 2006 American Cancer Society.
Breast cancer is associated with a substantial medical and economic burden. In the U.S., for example, it is estimated that breast cancer will account for approximately 32% of all new cancer diagnoses among in women during 2005 and for >40,000 deaths.1 Consequently, the management of breast cancer accounts for a large percentage of healthcare budgets. In the U.S., such management accounts for 15% to 20% of all cancer costs and for 1% of the total healthcare budget.2
In a routine literature search, we identified several previous studies that evaluated the cost of breast cancer care,3–7 but none of those authors investigated the economic burden associated with disease recurrence among women with early breast cancer. This represents an important gap in current knowledge for physicians and payors alike, because such information may help to quantify the benefit of recommended new (and generally more expensive) treatments that significantly reduce the risk of breast cancer recurrence.8–10 Therefore, the current study was designed to examine the economic burden associated with disease recurrence retrospectively in patients with early breast cancer from the perspective of a large, integrated U.S. healthcare system.
MATERIALS AND METHODS
A retrospective cohort study design was used to investigate the economic burden of recurrence among women who were treated for early breast cancer (Stages I and II according to the American Joint Commission on Cancer staging criteria) within the Henry Ford Health System (HFHS). The HFHS is a large, integrated healthcare system located in the greater metropolitan area of Detroit, Michigan, and it provides acute, primary, specialty, and tertiary care to approximately 400,000 individuals. Annually, care provided within the HFHS includes 2.5 million patient contacts, 40,000 surgeries, and 40,000 hospital admissions. The catchment's population of the HFHS is diverse ethnically (approximately 35% African American, compared with 12% in the U.S. population) and represents a variety of payor groups, including a health maintenance organization (50%), Blue Cross (12%) and other commercial payors (6%), Medicare (10%), Medicaid (2%), and private payors (4%).
Patient Characteristics and Disease Status
The HFHS cancer registry was used to identify all women who were diagnosed with early breast cancer between 1996 and 2002. This data source provided baseline characteristics on patient age and race as well as baseline cancer characteristics, including disease stage, initial treatment (including type of surgery and medical therapy, such as chemotherapy, hormone therapy, or radiation therapy), and hormone (i.e., estrogen, progesterone) and other receptor status. Physician diagnoses for patient encounters 1 month prior to and 12 months after breast cancer was diagnosed were obtained from health system administrative databases and were used to calculate a comorbidity score with the Charlson comorbidity index.11 Originally developed to assess survival probabilities based on inpatient medical record review, this methodology has been modified for use with administrative data bases as a means of measuring underlying illness burden.12
Recurrence of breast cancer was determined by medical record abstraction using the electronic medical records for the entire study population. For the purpose of the current analysis, new contralateral breast primary tumors also were defined as a recurrence of breast cancer (in accordance with the definition of recurrence adopted by the Early Breast Cancer Trialists' Collaborative Group13 and more recent clinical trials in this setting [e.g., the Arimidex, Tamoxifen, Alone or in Combination Trialists' Group9]). Records were reviewed from the date of diagnosis until the end of the study (March 31, 2004), unless the patient died or was lost to follow-up at an earlier date. Accredited cancer registry professionals performed abstraction. Recurrence events included locoregional, contralateral (breast), and distant (i.e., beyond the breast, regional lymph nodes, or regional tissue) recurrences. The first recurrence event for individual patients was used to assign patients to the corresponding recurrence groups. Locoregional or contralateral breast recurrences were confirmed by tissue sampling and histology reports. Distant recurrences were confirmed by results of imaging studies, including scans, X-rays, and/or histology. In some cases of distant recurrences, the diagnosis was confirmed by recorded information regarding tumor growth noted clinically and/or by imaging studies without histology. Recurrence-free probabilities were plotted subsequently by using the Kaplan–Meier estimation method.
Calculation of Economic Burden and Statistical Analysis
The economic burden of breast cancer recurrence was estimated by using data on patient charges that were incurred during the prerecurrence and postrecurrence periods. Charge data were derived from the integrated healthcare system's internal records, which were used for billing purposes. The analysis on charges of care included both total charges of care and potential noncancer-care charges. All inpatient and outpatient charges for care provided within HFHS were included in the analysis. In addition, for the patients who were enrolled in the healthcare system's health maintenance organization (n = 1031 patients [63%]), charges delivered outside of HFHS were captured fully through external claims. Charges for chemotherapy and radiation therapy provided with the healthcare system and external to the healthcare system for the health maintenance patients were included. Charges for prescription drugs were not included in the analysis because of the lack of a consistent and reliable data source. Patient charges were adjusted to 2003 U.S. dollars using the U.S. Consumer Price Index.14
Patients were censored at the time of death or separation from the healthcare system. Among 1013 patients with health maintenance insurance, 80% were enrolled at the time of diagnosis to the end of follow-up. The electronic medical record was reviewed to determine last date of contact with the healthcare system (for censoring) and to determine the first date of disease recurrence.
Charge data were log transformed prior to statistical testing to adjust for skewness in the data. Paired t tests were used to compare charges for care prerecurrence and postrecurrence using 6-month and 12-month comparisons of mean charges. Two-sample t tests were used to compare charges for initial, continuing, and terminal care between patients with and without recurrence. Linear regression analysis was used to compare monthly charge rates between patients with recurrences and patients without recurrences during the period of the study while adjusting (when possible) for the baseline variables of age at diagnosis, race, comorbidity score, hormone receptor status, stage at diagnosis, type of surgery, hormone therapy, and chemotherapy. The time to recurrence curve was constructed using the Kaplan–Meier method. Among the patients who developed recurrences, differences in the time to recurrence were tested using a log-rank test.
In addition to prerecurrence and postrecurrence charge analyses, patient care (log-transformed) charges were examined over the continuum of care after diagnosis by using an approach previously reported by Taplin et al.6 The periods examined for patients without recurrence were the initial treatment phase (6 months after diagnosis), continuing care (care after initial treatment; examined by quarterly periods), and terminal care (care for the 6-month period prior to death). Charges for the patients with recurrence also were examined for these periods in addition to the initial care period postrecurrence (defined as the 6-month period after the date of recurrence) and the continued care period after the initial 6 months of treatment for recurrence (examined by quarterly periods). Additional analyses (using paired t tests) for the patients with recurrence compared the initial treatment period with the initial postrecurrence treatment period and compared initial continuing care with continuing care postrecurrence. Care periods were not allowed to overlap in these analyses, in accordance with the approach adopted by Taplin et al.6 Subsequently, linear regression analysis was used to determine whether differences remained after adjusting for baseline variables. In that analysis, costs were reported as adjusted means from the linear regression model. All statistical analyses were performed using SAS software (version 9.1; SAS Institute, Inc., Cary, NC). All testing was 2-sided. and P<.05 was considered significant.
Patient Characteristics and Disease Outcome
The cancer registry search identified a total of 1808 women who were diagnosed with early breast cancer between 1996 and 2002. Of these, 192 women subsequently were excluded from further analysis. Because the presence of additional nonbreast cancers may have altered the costs of treatment and could be ascribed erroneously to breast cancer management, patients who had multiple cancers (n = 136 women) were excluded. Other patients who were excluded were women with <6 months of follow-up (n = 25 patients), women who had bilateral breast cancer at the time of presentation (n = 14 patients), women with the wrong disease stage (defined as not Stage I or II after a review of the electronic medical record; n = 11 patients), or women who received no definitive treatment (n = 6 patients).
The final study population of 1616 patients had a mean age of 60 years, 55% had Stage I disease, and 78% were hormone receptor-positive at the time of diagnosis. Nearly all patients (98.5%) underwent surgical resection of the primary breast tumor, and 65% underwent breast-conserving procedures. Adjuvant hormone therapy and chemotherapy were received by 56% and 44% of patients, respectively (Table 1). During the study period, 188 women died (11.6%), including 85 women (or 5.3% of the total study population) who died of breast cancer (based on chart review).
|Characteristic||No. of patients (%)|
|All Patients (n = 1616)||Patients without Recurrence (n = 1424)||Patients with Recurrence (n = 192)|
|Age in y|
|Mean (SD)||60 (14)||60 (14)||44 (23)*|
|Median (range)||59 (19–95)||45.5 (23–95)||39.3 (19–90)|
|White||1064 (65.8)||951 (66.8)||113 (58.8)*|
|African American||515 (31.9)||439 (30.8)||76 (39.6)|
|Hispanic||1 (0.1)||0 (0.0)||1 (0.5)|
|Asian||25 (1.5)||23 (1.6)||2 (1.0)|
|Other||11 (0.7)||11 (0.8)||0 (0.0)|
|AJCC stage at diagnosis|
|Stage I1||882 (54.6)||822 (57.7)||60 (31.3)†|
|Stage IIA||467 (28.9)||408 (28.7)||59 (30.7)|
|Stage IIB||267 (16.5)||195 (13.7)||73 (38.0)|
|Hormone receptor status‡|
|ER positive||1081/1446 (74.8)||981/1279 (76.7)||100/167 (59.9)†|
|PR positive||1007/1455 (62.2)||916/1287 (71.2)||91/168 (54.2)†|
|Mean comorbidity score (SD)§||2.3 (0.9)||2.3 (0.8)||2.5 (1.3)|
|Surgery for breast cancer|
|Biopsy only||24 (1.5)||20 (1.4)||4 (2.1)*|
|Lumpectomy||115 (7.1)||100 (7.0)||15 (7.8)|
|Lumpectomy with lymph node dissection||936 (57.9)||844 (59.3)||92 (47.9)|
|Mastectomy||541 (33.5)||460 (32.3)||81 (42.2)|
|Adjuvant hormone therapy||899 (55.6)||814 (57.2)||85 (44.3)†|
|Chemotherapy||714 (44.2)||602 (42.3)||112 (58.3)†|
|Radiation therapy||1229 (76.1)||1088 (76.4)||141 (73.4)|
By the end of follow-up (median, 44.5 mos; range, 6.1-98.3 mos), 192 women (11.9%) had experienced a breast cancer recurrence. The Kaplan–Meier estimate of the rate of recurrence at 8 years was 19.7% (95% confidence interval, 16.0-23.4%) (Fig. 1). Among patients who had disease recurrence, 112 recurrences (58%) were distant, 50 recurrences (26%) were local or regional, and 30 recurrences (16%) were in the contralateral breast. Collectively, patients who had disease recurrence were less likely to have hormone receptor-positive tumors and therefore were more likely to have received chemotherapy than patients who did not have disease recurrence (Table 1). When considering first recurrences, there was no significant difference noted in the time to recurrence between the 3 recurrence types, with median times to recurrence of 25.1 months, 17.6 months, and 25.2 months, respectively, for locoregional, contralateral breast, and distant recurrences (P = .94, log-rank test).
Economic Analysis of Disease Recurrence
Of the 192 patients who had recurrence events, 104 patients and 62 patients, respectively, were included in the 6-month and 12-month analyses of patient care charges. Paired t tests were performed for both 6-month and 12-month prerecurrence and postrecurrence periods for the overall population of patients who had recurrences and for specific recurrence types. All comparisons demonstrated a statistically significant increase in patient charges for the postrecurrence period (Table 2). Six-month analyses demonstrated a 4-fold increase in mean charges accrued during the postrecurrence period for the entire population of patients who developed recurrences compared with the prerecurrence period ($45,855 vs. $10,715; P<.001). The differences also were evident when patients were stratified according to recurrence type. Similarly, the 12-month comparisons showed statistically significant increases in mean patient care charges after recurrence ($79,253 vs. $12,344; P<.001), and the difference was particularly marked for patients who had distant recurrences (Table 2).
|Period||No. of Patients||Mean Charges (SD), $US|
|All patients with recurrence||104||10,715 (18,592)||45,855 (35,686)*|
|Type of recurrence|
|Contralateral breast||17||4788 (6231)||35,909 (17,897)*|
|Locoregional||35||8507 (13,491)||33,174 (25,119)*|
|Distant||52||14,138 (23,186)||57,642 (42,020)*|
|All patients with recurrence||62||12,344 (15,414)||79,253 (57,103)*|
|Type of recurrence|
|Contralateral breast||12||11,162 (17,220)||50,131 (43,148)*|
|Locoregional||21||11,341 (15,192)||61,027 (49,263)*|
|Distant||29||13,560 (15,276)||104,502 (58,017)*|
For an additional measure of the burden of recurrence, the total charges for care based on recurrence status also are reported (Fig. 2). Figure 2 shows that, compared with patients who did not recur, patients who had contralateral, locoregional, and distant recurrences had higher mean total care charges by $43,803, $66,927, and $102,504, respectively. However, the differences in total charges between the patients with different types of recurrences and the patients without recurrences were not tested statistically because of differences in the length of follow-up for which we could not adjust readily.
Linear regression analysis was used to compare average monthly charges for the care of patients with recurrence compared with patients who did not develop recurrences. Compared with patients who had no recurrences, patients with contralateral breast, locoregional, and distant recurrences had higher mean monthly care charges by $612, $1495, and $2962, respectively (P <.05) (Table 3). After adjusting for baseline covariates, recurrence status remained highly related to monthly charges (P <.001) across all recurrence types.
|Variable||No. of Patients||Median Follow-Up (Months)||Mean Monthly Charges ($US)|
|Patients without recurrence||1424||45.5||2003|
|Patients with recurrence|
In the analysis of charges over the continuum of care, log-transformed charges were examined for specific periods of care (Table 4). In the initial treatment phase, the mean charges for patients without recurrence were found to be greater than for patients with recurrence ($41,345 and $38,165, respectively; P <.05). There is no obvious explanation for this increase; however, compared with the prerecurrence and postrecurrence charges shown in Table 2 and the difference in total charges shown in Table 3, the magnitude of this increase was not substantial. Conversely, a different pattern was observed for the continuing care period after initial treatment, for which charges were significantly higher among patients who had recurrences compared with patients who did not have recurrences. Again, this increase did not appear to be substantial. In the analysis of the terminal care phase for patients who died, care in the last 6 months of life was increased significantly for patients who developed recurrences compared with patients who did not develop recurrences ($63,434 and $53,872, respectively; P ≤.01).
|Initial Care (6 Months)||Continuing Care (Average Quarterly Charges)||Recurrence Care (6 Months Postrecurrence)||Postrecurrence Continuing Care (Average Quarterly Charges)||Terminal Care (6 Months Prior to Death)|
|Patients without recurrence|
|No. of patients||1424||1317||Not applicable||Not applicable||65|
|Mean charges (SD), $US||41,345 (22,590)||1228 (1758)||53,872 (128,489)|
|Patients with recurrence|
|No. of patients||171||161||89||59||27|
|Mean charges (SD), $US||38,165 (23,329)*||2564 (2990)‡||47,581 (32,341)||4934 (5584)||63,434 (47,080)†|
An additional analysis of care charges was performed for the subgroup of patients who developed recurrences. Overall, among evaluable patients (i.e., patients with 6 months of follow-up), the mean charges for the initial care period after breast cancer diagnosis were much lower than the charges incurred after a diagnosis of recurrence ($38,254 and $50,355, respectively; P<.01). The significant impact of recurrence on costs of care also was evident for quarterly charges during the continuing care period (Table 5).
|Type of care||No. of Patients||Mean Charges (SD), $US|
|After Diagnosis||After Recurrence|
|Initial care||74||38,254 (20,738)||50,355 (33,287)†|
|Continuing care||59||1825 (1895)||4934 (5584)‡|
A linear regression analysis that was adjusted for baseline covariates was used to compare care charges for patients with and without recurrence. A model that included age and stage at diagnosis, comorbidity score, race, and hormone receptor status showed that the difference in initial care charges for patients with and without recurrence remained statistically significant. The adjusted means from the linear regression model were $29,733 and $36,316, respectively (P<.01). However, in a model that included age and stage at diagnosis, comorbidity score, race, hormone receptor status, hormone therapy, and chemotherapy, the effect of recurrence on terminal care charges was not found to be statistically significant ($27,759 for patients with recurrence vs. $15,804 for patients without recurrence; P = .29), possibly because of the decreased power of the multivariable model on our relatively small sample size.
A routine literature search suggests that the current retrospective data base study may be the first to document the economic impact of disease recurrence on the costs of care for women with early breast cancer in a ‘real-world’ setting. We examined the 6-month and 12-month prerecurrence and postrecurrence aggregate charges in such patients and observed that the overall charges for care were significantly higher for patients who developed recurrent disease compared with patients who did not develop recurrences. The greatest economic burden was observed among patients who had distant recurrences, adding to previous findings that patients with advanced breast cancer at diagnosis incur significantly more healthcare charges compared with patients who are diagnosed at an earlier stage.3–7 Furthermore, for all recurrence types (i.e., locoregional, contralateral breast, and distant recurrences), postrecurrence charges for care were significantly greater than prerecurrence charges. These results are consistent with a previous study in men with prostate cancer in which it was noted that metastatic progression and prostate-specific antigen progression were associated with significantly greater healthcare charges compared with patients who did not have disease progression.15 In addition, the estimated probability of 20% recurrence at the 8-year point indicates the medical burden of breast cancer recurrence. The use of therapy options that have the best demonstrated efficacy in reducing the rate of recurrence events may be helpful in further reducing this medical burden.
Similar to all retrospective studies that rely on administrative data, the current study has potential limitations. First, there may have been a selection bias because of the way in which patients were included in the study sample. For example, the requirement of at least 6 months of follow-up after the initial breast cancer diagnosis excluded patients who died or who left the healthcare system within a shorter time frame, and these patients may have been more ill than the study population. However, including these patients in the study probably would have increased the cost differences between patients with and without recurrence. Second, there is potential for information collection bias, because patient information was captured only within the HFHS; therefore, information regarding recurrence status or patient charges for care outside of the HFHS may be incomplete. However, an additional analysis of a subset of the study population insured under the health system-owned and operated health maintenance organization, patients for whom there are complete data on healthcare expenditures delivered outside of the HFHS, suggested that this bias (if any) is likely to be small. For example, an assessment of whether radiation therapy delivered outside of the healthcare system underestimated charges for this care indicated that <4% of women who had a record of radiation therapy in the tumor registry lacked corresponding radiation charge data within the HFHS or as external claims.
Another potential limitation is that the cut-off points for the phases of care after diagnosis may be varied and, thus, may yield different results for each phase of care. In the current study, charges were examined over the continuum of care by using an approach similar to that described by Taplin et al.6 However, if the duration of the initial phase of care was extended from 6 months to say 270 days to include radiotherapy received after 6 months, then the results for the initial and continuing care phases may change. Conversely, this would not affect the results for the overall charges for care.
In addition, for the current study, we used charge data instead of cost data; therefore; the results may not reflect the actual costs to the healthcare system. Conversely, an adjustment to the charges using a cost-to-charge ratio would not affect the proportional increase in costs because of disease recurrence and is unlikely to affect the corresponding level of statistical significance.
Although retrospective data base analyses have potential weaknesses, such ‘real-world’ studies have a number of potential advantages, including the ability to examine large populations for a longer time and the ability to look at specific subpopulations.16 We had the advantage of the combined data sources of the HFHS cancer registry and administrative data bases coupled with an extensive electronic medical record system, which serves as an efficient and comprehensive record of care provided within the healthcare system. The breadth and depth of such data resources are a major strength of this study.
A meaningful health economic analysis with the objective of guiding decisions on the allocation of healthcare resources should include all relevant costs in the cost-effectiveness evaluation. Several alternatives for the adjuvant treatment of women with early breast cancer are available, including aromatase inhibitors, chemotherapy, and trastuzumab17; and recent economic evaluations have reported on the cost-effectiveness of some of these approaches. Bonneterre et al.,18 for example, reported an economic evaluation of the combined fluorouracil (at a dose of 500 mg/m2), cyclophosphamide (at a dose of 500 mg/m2), and epirubicin (at a dose of 50 mg/m2) (FEC 50) protocol compared with the same protocol but with the epirubicin dose increased to 100 mg/m2 (FEC 100) and concluded that the clinical benefit of FEC 100 generated a negligible increase in cost compared with FEC 50. Similarly, an economic evaluation of the aromatase inhibitor anastrozole versus tamoxifen indicated that anastrozole is a cost-effective alternative to tamoxifen in the adjuvant treatment of women with early breast cancer,19 providing not only a superior safety profile but also significantly reducing the risk of breast cancer recurrence. These analyses accounted for all relevant costs and outcomes in their economic evaluations. The economic burden of breast cancer recurrence is an important component of the overall cost of breast cancer care. Therefore, the results of the current study will facilitate future economic evaluations that examine the cost-effectiveness of adjuvant breast cancer therapies.
The results of the current study demonstrated that recurrence poses a significant economic burden in patients with early breast cancer, with the highest costs associated with distant disease. Overall, patients who experienced a recurrence required more costly care compared with patients who did not. Furthermore, the economic burden of recurrence was apparent throughout the entire continuum of care for patients with early breast cancer. These findings highlight the potential economic benefits of breast cancer treatment alternatives that prevent disease recurrence. Treatments that are more effective in reducing the risk of recurrence may provide economic benefits in terms of avoided cost of treating breast cancer recurrence events and eventually may contribute to lowering the overall cost of breast cancer care.
The authors thank Margaret Paquette, Certified Tumor Register, Henry Ford Health System, for her expertise in cancer care abstraction and Simon Sharp for his editorial assistance with this article.
- 1American Cancer Society. Cancer facts and figures, 2005. Atlanta, GA: American Cancer Society; 2005. Available at URL: http://www.cancer.org/downloads/STT/CAFF2005f4PWSecured.pdf [accessed April 2005].
- 5Age differences in the psychosocial problems encountered by breast cancer patients. J Natl Cancer Inst. 1994; 16: 191–197., , .
- 10BIG 1-98 Collaborative Group. Letrozole vs. tamoxifen as adjuvant endocrine therapy for postmenopausal women with receptor-positive breast cancer. BIT 1-98: a prospective randomized double-blind Phase III study [abstract]. Breast. 2005; 14(Suppl 1): S4.
- 14U.S. Department of Labor. Bureau of Labor Statistics Consumer Price Index for All Urban Consumers. 2005: U.S. city average. Item: medical care. Base period: 1982-84 = 100. Washington, DC: U.S. Department of Labor, 2005. Available at URL: http://stats.bls.gov.cpihome.htm [accessed April 2005].
- 19ATAC Trialists' Group. Cost-utility analysis of anastrozole versus tamoxifen as primary adjuvant therapy in postmenopausal women with early breast cancer from a US healthcare system perspective: the 5-year completed treatment analysis of the ATAC (‘Arimidex,’ Tamoxifen Alone or in Combination) trial [abstract]. Presented at the 27th Annual San Antonio Breast Cancer Symposium, San Antonio, Texas, December 8-11, 2004., on behalf of the