Cost-Effectiveness of Docetaxel-Cyclophosphamide Versus Doxorubicin-Cyclophosphamide for Breast Cancer

In the United States, adjuvant chemotherapy with an anthracycline-based regimen is a widely used treatment for operable, invasive breast cancer. However, improved survival times in early-stage breast cancer patients are accompanied by concerns regarding the long-term toxicity of chemotherapeutic drugs, particularly anthracycline-associated cardiomyopathy.^1-4

Taxanes (eg, docetaxel) are among the most widely used therapies in the treatment of metastatic breast cancer.⁵ Recently, the effi cacy of these drugs has been demonstrated in the adjuvant setting. Results of the US Oncology Research trial 9735 (USOR 9735) demonstrated significant improvements in 7-year disease-free survival (DFS) (81% vs 75%; hazard ratio [HR] 0.74, P = .033) and overall survival (OS) (87% vs 82%; HR 0.69, P = .032) for docetaxel and cyclophosphamide (TC) versus doxorubicin and cyclophosphamide (AC) in women with resected node-positive or high-risk nodenegative breast cancer (

Figure 1

).⁶

The results of the USOR 9735, along with concerns about anthracycline-associated cardiac toxicity, may provide support for the selection of TC over AC for women with nodenegative and lower-risk, node-positive breast cancer. In the past, the evaluation of pharmaceuticals by managed care organizations for inclusion on the formulary had a primary focus on safety and effi cacy. With rising drug expenditures in the United States over the last decade, cost considerations have now become essential.⁷ Accordingly, the objective of this analysis was to assess the cost-effectiveness of adjuvant TC therapy compared with standard AC therapy from the perspective of US managed care organizations.

METHODSStudy Design

The analysis was conducted using a lifetime decision model, developed in Microsoft Excel (Microsoft, Redmond, WA), simulating disease progression in early-stage breast cancer (

eAppendix A

, available at www.ajpblive.com). The model followed a cohort of patients for their lifetimes.

One arm of the model considered adjuvant chemotherapy with TC, while the other considered adjuvant chemotherapy with AC. Over each 1-month cycle, patients either remained in the “alive” state or were transitioned to the “dead” state. These transitions were based on survival data reported in the USOR 9735.6 Costs and outcomes beyond 1 year were discounted at 3% annually.^8,9

Disease recurrence was modeled as an event rather than a health state, and it was assumed that patients who did experience recurrence could do so only once during the trial time period. This simplifying assumption was deemed reasonable given the relatively low recurrence rate during the trial period. Those experiencing a recurrence were attributed a utility decrement, representing quality of life (QoL) lost as well as the costs associated with that recurrence.

Patient Population

The base case estimates applied to a “typical” patient enrolled in the USOR 9735.¹⁰ This prospective, phase III, randomized study evaluated TC or AC treatment in 1016 women with operable stage I to III breast cancer. These patients had undergone complete surgical excision of the primary tumor, had no evidence of metastatic disease, and were eligible for adjuvant chemotherapy. The median age of patients was 51 years; 84% were younger than age 65 years. The majority of patients (71%) had breast cancer that was estrogen receptor positive and/or progesterone receptor positive and 52% of patients were node positive.

Treatment Arms

Patient treatment in the USOR 9735 consisted of four 3-week cycles of either TC (75 mg/m² of docetaxel and 600 mg/m² of cyclophosphamide) or AC (60 mg/m² of doxorubicin and 600 mg/m2 of cyclophosphamide). Accordingly, these were the therapy regimens evaluated in this analysis.

Effectiveness Data

The long-term effectiveness for TC and AC was derived from USOR 9735 and extrapolated the potential clinical benefits and costs over the lifetime of a patient cohort. Survival over the first 7 years in our analysis followed that of patients in USOR 9735 and survival beyond 7 years matched that of the general population of US women aged 58 years (25.7 years at age 58). Given that the cycle duration of the model was 30 days to allow for the attribution of costs and utility, the OS curve was digitized using TechDig software (TechDig, Toronto, Ontario) to derive monthly transition probabilities. It was assumed that all patients alive and disease-free at 7 years, regardless of treatment, remained disease-free for the rest of their lives. This simplifying assumption was deemed reasonable given that 8-year data reported by Jones et al suggested a low risk of death or recurrence for TC patients after year 7, while the risk of death or recurrence for AC patients persisted beyond 7 years; thus, this simplifying assumption would bias the results against TC.⁶

Disease recurrence was calculated as the inverse of the 7-year DFS curves from USOR 9735.⁶ The proportion of patients who died was removed to more accurately estimate recurrence and avoid double-counting death in the analysis.

eAppendix B

(available at www.ajpblive.com) provides the DFS curves from USOR 9735 along with the estimated recurrence curves. Recurrence was classified as either local recurrence or distant recurrence based on data from USOR 9735; 13.6% and 18.8% of TC patients and AC patients experienced local recurrence and 86.4% and 81.2% of TC and AC patients experienced distant recurrence, respectively.¹⁰ Based on expert opinion, it was assumed that 25% of local recurrences were noninvasive and 75% were invasive.

Cost Data

Drug costs included only the cost of chemotherapy, assuming no drug wastage, a body surface area of 1.6 m², and administration every 3 weeks for 4 cycles. Unit costs for the therapies were $16.732 per milligram for docetaxel, $0.466 per milligram for doxorubicin, and $0.0186 per milligram for cyclophosphamide. These drug costs were derived from Medicare Part B (October 2008) and were based on Biologics Average Sales Price plus a 6% markup (

Table 1

).¹¹

Additional costs for the study were determined through analyses of the IMS LifeLink Health Plans Claims Database (PharMetrics)¹² and Premier Perspective Comparative Database.¹³ The IMS LifeLink Health Plan Claims Database includes longitudinal, integrated, patient-level medical and pharmaceutical claims comprising 4 billion patient observations from across the United States and more than 65 million patients from more than 95 health plans, including medical services and prescription drug information across the entire continuum of care. The Premier Perspective Comparative Database is a repository of hospital administrative data that includes approximately one-sixth of all hospitalizations in the United States. Service-level data that are recorded include charges for medications, procedures, and laboratory tests; characteristics of surgeons and hospitals are also available.

For costs associated with chemotherapy administration, the PharMetrics database was used to determine cost of administration per cycle. The sample included patients having a diagnosis of breast cancer (International Classification of Diseases, Ninth Revision [ICD-9] code 174) from January 2003 to December 2007, receiving at least 1 cycle of TC or AC following the first identified breast cancer diagnosis (assigned as the index diagnosis), and having mastectomy between index diagnosis and the first TC/AC use. The drug administration cost per cycle was determined separately for TC and AC cohorts and comprised physician visits, infusion procedure, equipment and solutions, growth factor, and other medications (ie, antinausea and antiemetic agents, agents to protect the heart).

Inpatient costs of treating grade 3 or 4 hematologic toxicities and fatal adverse events were derived from the Premier database from January 2004 to December 2007 for the events of anemia (ICD-9 code 285.9), neutropenia (ICD-9 code 288.0) and thrombocytopenia (ICD-9 code 278.4), congestive heart failure (ICD-9 code 478.0), and myelodysplastic syndrome (ICD-9 code 238.75).

Lastly, the cost of treating relapse (per episode) was estimated using the PharMetrics database. The sampled patients included those with a breast cancer diagnosis (January 2003 to December 2007) followed by a local recurrence (ie, ICD-9 codes 196, 198.2, 198.81) or distant recurrence (ie, ICD-9 codes 197, 198, except 198.2, 198.81) after the index diagnosis. The costs of chemotherapy were assumed to be the same as in the adjuvant treatment setting. Other cost components of recurrence included diagnostic and staging, physician visits, surgery, and radiation.

Total local recurrence costs were composed of the weighted noninvasive (25%) and invasive (75%) costs. Treatment algorithms for local and distant recurrence were provided by expert opinion. For patients who experienced local invasive recurrence, it was assumed that 50% of patients would receive an additional course of chemotherapy, while 50% would not receive chemotherapy. For patients who received TC as adjuvant chemotherapy, it was assumed that 100% of patients would receive AC in postadjuvant treatment of local recurrence. For patients who received AC as adjuvant chemotherapy, it was assumed that 100% of patients would receive TC in postadjuvant treatment of local recurrence. For patients who experienced local noninvasive recurrence, it was assumed that patients would be treated with surgery.

For patients with a distant recurrence who received adjuvant chemotherapy with TC, it was estimated that 50% of patients received a second course of TC, while 50% received a course of AC. For patients who received adjuvant chemotherapy with AC, it was estimated that 20% of patients received a second course of AC, while 80% received a course of TC. Following progression, 100% of patients, regardless of adjuvant chemotherapy, received approximately 6 cycles of chemotherapy with capecitabine. A total episode cost of distant recurrence utilized in the model was calculated by summing the cost within the first 6 months and the cost in the subsequent 6 months of recurrence.

Costs were inflated to 2008 dollars using the healthcare component of the Consumer Price Index and discounted at 3% per year.

Utility

As data specific to the US population were unavailable, the base utility value (0.79) for early breast cancer patients in remission (DFS) was obtained from a recent cost-effectiveness analysis by Wolowacz et al.¹⁴ This utility weight was calculated using the European Organisation for Research and Treatment of Cancer QoL questionnaire (QLQ-C030) on data collected in the BCIRG001 clinical trial.¹⁵

Utility decrements for local (-0.09) and distant (-0.29) recurrence were also calculated based on the analysis by Wolowacz et al¹⁴ by subtracting the respective utilities from the utility for remission. The QoL impact of recurrence was assumed to last for 1 year and the full utility decrement was applied in the first month of recurrence, in order to simplify the analysis structure. Overall survival was multiplied by health preference in order to obtain a quality-adjusted life-year (QALY). After 7 years, it was assumed that patient QoL would be comparable to that of women in the general population.¹⁶

Sensitivity Analysis

A series of 1-way sensitivity analyses were performed to explore the effect of changes in assumptions and data inputs on model outcomes (

Table 2

).

The survival advantage associated with chemotherapy regimens may extend beyond the limited follow-up of clinical trials.¹⁹ Therefore, we conducted 2 sensitivity analyses: (1) we assumed greater life expectancy in the TC arm equivalent to the mean difference in OS between the TC and AC arms within the clinical trial period (approximately 1.8 months) and (2) we assumed greater life expectancy in the TC arm equivalent to the difference in survival at the end of the 7-year trial period (approximately 22 months).

Several sensitivity analyses were conducted with respect to costs, inclusion of long-term fatal adverse events for AC, varying the utility decrements according to low¹⁷ and high¹⁸ ranges reported for local and distant recurrence, and varying the discount rate by 0% and 5%. Results for the base case and sensitivity analyses were presented as cost per life-year gained and cost per QALY gained in order to capture health-related QoL as well as OS.

RESULTSBase Case

Longer survival and fewer recurrences for patients receiving TC resulted in gains in both life-years and QALYs compared with those receiving AC (

Table 3

). Base case results of the model suggested that the average TC patient was expected to live 18.95 years, which was 10.2 months (0.85 years) longer than the average AC patient. Following adjustment for QoL, the average TC patient was estimated to experience 14.87 QALYs vs the 14.20 QALYs experienced by AC patients.

Mean total lifetime disease-related costs were higher with TC compared with AC: treatment with TC was associated with a $5897 higher lifetime cost than AC, with a total per patient cost of $37,846. This cost difference was driven primarily by the higher drug acquisition costs associated with TC. Costs per life-year and per QALY gained (TC vs AC) were $8819 and $6985, respectively (Table 3).

Sensitivity Analyses

Sensitivity analyses indicated that the results were most sensitive to changes in assumptions regarding the time horizon (date not shown). With a 7-year time horizon (ie, the time frame used in USOR 9735), the incremental cost per life-year gained remained under $50,000 ($45,037) and the incremental cost per QALY gained was very close to $50,000 ($53,983). In all remaining sensitivity analyses, cost per life-year gained remained between $3382 and $9367 and cost per QALY gained remained

between $4296 and $11,791 (

Figure 2

).

DISCUSSION

The results of USOR 9735 demonstrated that compared with AC, adjuvant treatment with TC reduced the risk of recurrence and improved OS in patients with operable, invasive breast cancer. Grade 3 and 4 toxicities were comparable between the 2 groups with the exception of anemia, almost of all which occurred in older women receiving AC (5% vs <1% in other groups) and febrile neutropenia, which was doubled in the TC group (5% vs 2.5%). There were 3 late deaths without relapse, probably related to treatment, and all occurred in the AC group.

These findings prompted increased utilization of TC in early breast cancer treatment; however, the cost-effectiveness of this therapy in the United States remained unknown. Accordingly, the present economic evaluation was undertaken to determine the cost-effectiveness of TC compared with AC in this patient population.

Results of the base case analysis demonstrated that treatment with TC compared with AC provided an increase of 0.85 year of life and 0.674 QALY per patient. These benefits were attained with a cost per life-year gained of $6985 and a cost per QALY gained of $8819 over a lifetime horizon. Overall, 1-way sensitivity analyses conducted on a variety of input parameters demonstrated the robustness of the base case results. Limitation of the time horizon to 7 years (the time frame of USOR 9735) resulted in an incremental cost per life-year of $45,037 and per QALY of $53,983; these values were below or close to the generally accepted threshold of $50,000 in the United States.²⁰

Strengths of the present analysis included the use of a simple, transparent analysis structure and the use of a gold standard, head-to-head, randomized controlled trial to provide the primary effectiveness inputs for the analysis. Furthermore, required assumptions were consistently biased against the new intervention, particularly assumptions regarding long-term survival. For the base case analysis, it was conservatively assumed that clinical benefits in terms of OS and recurrence did not extend beyond the end point of the clinical trial. Assuming no continued advantage for DFS and the same survival in both the TC and AC cohorts after the trial follow-up resulted in equivalent survival beyond 7 years. Since the recent Oxford overview (Early Breast Cancer Trialists’ Collaborative Group 2005) suggests that DFS and OS curves for alternative adjuvant regimens continue to diverge until at least 10 years after random assignment, it is likely that this analysis provided a conservative estimate of the longterm benefits of TC. The sensitivity analyses demonstrated that use of alternative assumptions (ie, extension of OS beyond 7 years) resulted in improved cost-effectiveness ratios, ranging from $3384 to $6574 for cost per life-year gained and $4296 to $8306 for cost per QALY gained.

Potential limitations of the present analysis included the inclusion of a single comparator, AC, without considering other regimens. However, trial-based cost-effectiveness analyses are often conducted, given that head-to-head, randomized controlled trial data provide the most direct evidence of the relative efficacy of comparator therapy. A second limitation would be the need for assumptions to be made when combining clinical data with data obtained outside of USOR 9735. For example, the availability of utility values for the calculation of QALYs in the United States was limited and varied considerably. Further, the data used for various cost inputs may not have been current, having been derived in some cases from database entries between 2003 and 2007. Overall, the analysis used the most up-to-date and accurate sources available. In addition, sensitivity analyses were conducted to address the impact of uncertainty about cost estimates on the costeffectiveness ratio, with the results being robust.

An analysis similar to the current cost-effectiveness evaluation of TC versus AC was recently conducted from the perspective of the Canadian publicly funded healthcare system.²¹ The results of this study also supported adoption of the TC regimen for the treatment of operable, invasive breast cancer, reporting comparable ratios of $6842 per life-year gained and $8251 per QALY gained over a lifetime horizon.

Our results were also consistent with a second Canadian cost-effectiveness analysis study, with recent findings presented by Younis et al at the 2008 San Antonio Breast Cancer Symposium.²² Also using data from USOR 9735, the authors found TC to be cost-effective compared with AC, with a cost per QALY of $16,753. Cost components and assumptions utilized in this Canadian analysis were similar to those used in our US analysis, with key differences being the inclusion of more than 1 opportunity for disease recurrence for each patient and use of a 10-year time horizon in the published Canadian analysis. Further economic evaluations specifically comparing these 2 chemotherapeutic regimens were not identified in the literature.

The similarity in results between the 2 published studies and this study suggests that country-specific differences in costing inputs and sources did not appear to impact the overall conclusions. Nevertheless, given the increasing emphasis on requirements for cost-effectiveness evaluations in the United States,⁷ our study provided a unique comparative economic evaluation of these 2 chemotherapy regimens, applying US-specific resource utilization data obtained from nationwide, large representative databases.

CONCLUSIONS

Previous results from USOR 9735 demonstrated that TC was superior to AC with respect to both DFS and OS for the treatment of operable, invasive breast cancer. Extrapolation of these results to a lifetime horizon in this study demonstrated that use of TC was highly cost-effective compared with AC. When the time horizon was limited to the 7-year period of the clinical trial, these incremental cost-effectiveness ratios remained at or below the established threshold for cost-effectiveness. Results were robust to the majority of the sensitivity analyses evaluated. These findings provide additional evidence to US payers and decision makers regarding the value of TC use in the management of early-stage breast cancer.