- © 2002 by American Society of Clinical Oncology
Economic Analysis of the TAX 317 Trial: Docetaxel Versus Best Supportive Care as Second-Line Therapy of Advanced Non–Small-Cell Lung Cancer
- Natasha B. Leighl,
- Frances A. Shepherd,
- Rita Kwong,
- Ronald L. Burkes,
- Ronald Feld and
- Pamela J. Goodwin
- From the Department of Medical Oncology, Princess Margaret Hospital/University Health Network, and Mount Sinai Hospital, University of Toronto, Toronto, Ontario, Canada.
- Address reprint requests to Natasha Leighl, MD, Princess Margaret Hospital, 5-222, 610 University Ave, Toronto, Ontario, Canada M5G 2M9; email: Natasha.Leighl{at}uhn.on.ca
Abstract
PURPOSE: To determine the cost-effectiveness (CE) of second-line docetaxel compared with best supportive care (BSC) in the TAX 317 trial, a randomized clinical trial of second-line chemotherapy in non–small-cell lung cancer.
METHODS: A retrospective CE analysis of the TAX 317 trial was undertaken, evaluating direct medical costs of therapy from the viewpoint of Canada’s public health care system. Costs were derived in 1999 Canadian dollars, and resource use was determined through prospective trial data.
RESULTS: The incremental survival benefit in the docetaxel arm over BSC was 2 months (P = .047). The CE of docetaxel was $57,749 per year of life gained. For patients treated with docetaxel 75 mg/m2, the CE was $31,776 per year of life gained. In univariate sensitivity analyses, CE estimates were most sensitive to changes in survival, ranging from $18,374 to $117,434 with 20% variation in survival at the recommended dose. The largest cost center in both arms was hospitalization, followed by the cost of drugs, investigations, radiotherapy, and community care. BSC patients had fewer hospitalizations than patients in the chemotherapy arm and were more often palliated at home.
CONCLUSION: Although the decision to treat should not be based on economic considerations alone, our CE estimate of $31,776 per year of life gained (at the currently recommended dose of docetaxel) is within an acceptable range of health care expenditures, and the total costs of therapy are similar to those of second-line palliative chemotherapy for other solid tumors.
LUNG CANCER IS NOW the leading cause of cancer death for both men and women in North America. Approximately 200,000 individuals will be diagnosed with lung cancer in the year 2000, and more than two thirds will present with advanced disease.1-3 The ability of first-line palliative chemotherapy to prolong survival and improve quality of life (QOL) in patients with advanced lung cancer has been clearly established.4-7 However, until recently, the usefulness of second-line chemotherapy for non–small-cell lung cancer (NSCLC) had not been documented.8,9 As the use of palliative chemotherapy increases, the costs of cancer chemotherapy will escalate, and it is important to understand how these costs relate to benefits. It is estimated that the annual direct medical care costs of lung cancer in North America are in the range of more than $5 billion United States dollars.10 In an ideal health care system, treatment decisions would not be based on economic considerations. However, in most countries, health care resources are not unlimited, and economic constraints do factor into resource allocation. With the emergence of effective second-line or even third-line therapy for breast, gastrointestinal, and lung cancer, these costs may be expected to increase even further.
Several economic evaluations have been performed comparing costs and benefits of palliative treatment in lung cancer.11 These have evaluated first-line treatment, which in most studies has demonstrated a modest yet clinically significant improvement in survival and QOL.4-7 The TAX 317 trial was the first trial of second-line chemotherapy for advanced NSCLC to compare treatment, in this case with docetaxel, with best supportive care (BSC).8 A significant survival benefit was demonstrated that was similar in magnitude to that achieved with first-line chemotherapy for advanced lung cancer.6 Improvements in symptoms and QOL were also demonstrated for patients on the chemotherapy arm. Although only a select group of patients with advanced lung cancer are suitable candidates for second-line lung cancer chemotherapy, the high prevalence of this disease compels us to examine the costs of this new strategy, which we anticipate will become widely accepted in the developed world.
The economic analysis described here was prompted by the magnitude of the survival benefit observed in the treatment arm. This evaluation is of the direct medical costs of therapy and takes the viewpoint of the Canadian public health care system.
METHODS
General Method of Economic Evaluation
A retrospective economic evaluation of the TAX 317 randomized trial of second-line docetaxel for metastatic NSCLC was performed by using the technique of cost-effectiveness analysis (CEA). The trial compared chemotherapy with single-agent docetaxel versus BSC in patients previously treated with cisplatin-based chemotherapy. Costs of therapy were taken from the time of randomization until death, and treatment benefit was described as the mean survival after randomization onto the study.
Costing was performed at one participating center, the Princess Margaret Hospital (PMH), a tertiary care cancer center. Patterns of resource use were determined through prospectively collected information taken directly from trial data of patients enrolled onto each arm of the study. Any missing data on direct medical costs were collected retrospectively for patients from two participating centers in Toronto—PMH and the Mount Sinai Hospital, a general teaching hospital. Valuation of resources was determined retrospectively in 1999 Canadian dollars. Nonmedical direct and indirect costs were not included in the analysis.
TAX 317 Study Design
The design and results of the TAX 317 study have been published previously.8 To summarize, patients with advanced NSCLC who had received one or more cisplatin-based chemotherapy regimens were randomized either to second-line chemotherapy treatment, with docetaxel 100 mg/m2 or 75 mg/m2 intravenously over 1 hour every 21 days, or to BSC. The higher dose of docetaxel was used in the first half of the study. However, this dose was reduced to 75 mg/m2 for the second half of the trial, because the interim safety-data monitoring identified a higher toxic death rate on the docetaxel 100 mg/m2 arm. Patients were evaluated at baseline with physical examination; standard bloodwork, including hematology and biochemistry; chest radiograph; and clinically indicated imaging studies, including computed tomographic scans of brain, chest, and abdomen, ultrasonography, and radionuclide bone scans. All patients completed a QOL questionnaire. Patients in both arms were evaluated with history and physical examination with bloodwork, chest radiograph, and QOL assessment every 3 weeks. After 18 weeks on the study, patients in the BSC arm were evaluated every 6 weeks until disease progression. Patients in the chemotherapy arm were treated until disease progression or until the development of unacceptable toxicity. All patients were offered supportive care interventions as deemed appropriate by the treating physician, including palliative radiotherapy, analgesics, transfusions, home oxygen therapy, and antibiotics. No chemotherapy was offered to patients on the BSC arm while they were on the study protocol. However, patients were discontinued from the study protocol on disease progression, after which investigators could offer chemotherapy at their discretion. Patients were observed until death, with determination of overall survival, toxicity, and clinical benefit, including QOL, tumor-related medication use (including analgesics), and use of palliative radiotherapy.
Patient Population
Two hundred four good performance (Eastern Cooperative Oncology Group performance status ≤ 2) status patients with a diagnosis of stage IIIB or IV NSCLC, previously treated with cisplatin-based chemotherapy, were randomized onto the study, 100 to the BSC arm and 104 to the docetaxel arm (49 docetaxel 100 mg/m2 and 55 docetaxel 75 mg/m2). Although 36 centers in Europe and North America participated in the study, 68 patients (33%) were treated at the PMH and Mount Sinai Hospital in Toronto, and 77 (38%) were from the province of Ontario, where the costing was performed.
Treatment Outcomes
The measures of treatment outcome were obtained directly from the therapeutic results of the TAX 317 trial. The primary treatment outcome was mean survival (in months), measured from the date of randomization to the date of death or, if still alive, to the date of last contact. Survival data were analyzed with the log-rank test. The incremental therapeutic effectiveness was determined by subtracting the estimated mean survival time on the BSC arm from the estimated mean survival time on the chemotherapy arm.
Determination of Costs
All costs were identified retrospectively at the PMH. This analysis includes only those costs related to treatment of patients with NSCLC; costs relating to research were excluded. Discounting was not used in this analysis, because the median duration of survival in both arms was less than 1 year.
TAX 317 patients generated costs relating to outpatient assessment, chemotherapy administration, hospitalization, radiation therapy, community-based nursing and supportive care, and miscellaneous items. The methodology for determining these costs is described below. Trial data and medical records from treating institutions, as well as other facilities at which care was received, were examined in detail for all 68 patients entered from the two participating Toronto centers.
Costs of Clinic Visits
The costs of clinic visits were determined using the hotel approximation method. Hotel costs included the portion of hotel costs for the entire institution that could be attributed to the dedicated outpatient lung cancer clinic at the PMH, and these costs were averaged over outpatient visits to the clinic. These included such costs as overhead, administration, plant costs (including maintenance, housekeeping, and porter services), equipment, central supply, and medical records. For medical costs, the cost of nursing care and other supportive staff was derived from salary, including benefits and educational leave, and hours attributed to the lung cancer clinic and ensuing workload. These total costs were averaged over the number of patient visits to the lung clinic. Costs for physicians’ services and investigations performed (radiology and laboratory) varied according to what actually occurred at the clinic visit for each patient, as derived from the trial data and hospital records. The frequency of clinic visits per patient was similarly derived. All physician service costs were based on the 1999 Ontario Health Insurance Plan fee schedule,12 as were the costs of laboratory and imaging investigations. The medical costs were added to hotel costs to arrive at the average cost per clinic visit for the patients on each arm of the trial.
Chemotherapy Costs
Costs relating to chemotherapy administration included the costs of the chemotherapeutic agents and supportive medications (antinauseants and hypersensitivity prophylaxis), with the actual doses and number of courses obtained directly from the study and hospital records. Drug preparation costs were included (averaged over the total number of drugs prepared). The cost of visiting the chemotherapy daycare clinic was derived by adding the attributed hotel cost (including intravenous solutions, medical and surgical supply, and support staff) to the medical costs of physician supervision and direct nursing care. After review of chemotherapy nursing practice at the PMH, it was estimated that a chemotherapy nurse could share time between multiple patients, averaging 50 minutes per patient receiving docetaxel (ie, supervising two patients simultaneously). Double counting of costs was avoided.
Costs of Acute Care Hospitalization
The costs of acute care hospitalization were determined for the dedicated inpatient solid tumor ward at the PMH. Hotel costs were derived with the hotel-approximation method, with attributed costs averaged over the total number of inpatient days on the ward, resulting in a per diem hotel cost. Medical care costs varied according to the admission. Medications and investigations (laboratory and radiology) were detailed for each patient through extensive review of trial and hospital records, and thus costs varied among patients. The nursing support staff cost was determined by using the annual expenditure of salaries for the solid tumor ward, averaged over the total number of inpatient days. Costs of physicians’ services were included here and varied among patients. These were summed for a per diem average acute care inpatient cost for each arm of the study. Costs for admission through the emergency department, derived similarly, were counted where appropriate, again with avoidance of double counting.
Costs of Palliative Care Institutionalization
The costs of palliative care unit admissions (averaged at $409.70 per day) were obtained from the Toronto Grace Hospital, a hospital with a large palliative care inpatient ward (S. Bradshaw, personal communication, March 2000). This includes an average medication cost, other direct medical costs, and hotel cost per patient day. Where possible, individual patient data on medication use or investigations were calculated and used instead.
Costs of Radiation Therapy
The number and dose of radiation fractions were obtained from medical records. The cost per fraction, as well as visit costs for radiation review and follow-up clinic, were updated from the data of Warde and Murphy,13 from an earlier costing study. These costs similarly reflect attributed hotel costs plus direct medical costs of radiation therapy at the PMH. In 1996 Canadian dollars, the estimated cost per fraction with a high-energy linear accelerator was $104.72. With the Consumer Price Index for Health Services, the cost per radiotherapy fraction was $109.18 in 1999 Canadian dollars.14 Given that the TAX 317 patients were treated with palliative intent, the standard practice in Ontario was to not use computed tomographic planning, and this is reflected in the costing method (P. Warde, personal communication, February 2000).
Costs of Toxicity
Major toxicities are listed in Table 1. The differences in neutropenic sepsis, pneumonia, and toxic deaths on the higher docetaxel dose arm reflect the increased days of hospitalization for the chemotherapy arm. These costs have already been counted in the hospitalization costs.
Summary of Toxicity Seen in the TAX 317 Trial8
Other Costs
Outpatient medication use (including dosage and number of tablets prescribed) for each patient was determined through intensive review of trial and medical records, including records from palliative and primary care physicians, where available. The number of dispensing fees per patient was derived where possible for each patient; otherwise, it was estimated that patients paid an additional fee per month per medication prescribed. Costs of outpatient medications were determined with the Ontario Drug Benefit Formulary plus a 10% markup fee, as permitted by the Ontario government.15 Nonformulary medication costs included a 15% markup, reflecting the actual practice of the PMH outpatient pharmacy (J. Kirby, personal communication, January 2000). The 1999 unit cost of processing a unit of packed RBCs was provided by Canadian Blood Services (P. Ghelani, personal communication, January 2000). Charges to the health care system for community care (J. Armstrong and G. Wilson, personal communication, January and February 2000, respectively), including nursing and homemaking services, as well as home oxygen, were used (M. Reilly, personal communication, February 2000).
Analysis of Costs
The technique of CEA was performed by using the incremental mean cost (IC) per patient receiving chemotherapy over that with BSC, divided by the incremental mean survival of the chemotherapy arm over the BSC arm. Incremental mean survival was used in the denominator of the cost-effectiveness (CE) ratio, similar to other economic evaluations in health care, including several in palliative chemotherapy.16-20 According to Smith and Hillner17 and Hillner and Smith,21 this is the preferred measure of survival in such evaluations.
Sensitivity Analyses
To take into account natural variability and potential measurement error in costs and study outcomes, the technique of sensitivity analysis has been used to generate a range of plausible values for CEA. Because hospitalization was a major cost, univariate sensitivity analysis of the number of days in the hospital was performed, varying the number of inpatient days ± 20% on each arm. Additional univariate sensitivity analyses were undertaken, verifying the cost of docetaxel chemotherapy and incremental survival ± 20%. Of note, this variation in survival extended beyond the 95% confidence interval for the different mean survival values, and, because of this, the range is broader than would be expected by chance alone. The effects of variability in costs were also examined by varying the costs of medications, clinic visits, and investigations by ± 20%. Finally, the number of chemotherapy cycles was varied from two to 12.
RESULTS
TAX 317 Treatment Outcome
Survival and clinical benefit.
Overall survival was prolonged in the chemotherapy arm, with a median survival of 7.0 months on the docetaxel arm versus 4.6 months on BSC (P = .047 by log-rank test). Patients randomized to docetaxel 75 mg/m2 in the last half of the study had a median survival of 7.5 months, compared with 4.6 months with BSC (P = .01 by log-rank test).8 Mean survival, used in calculating the CE, was 9.1 months with docetaxel versus 7.1 months with BSC for the entire cohort (P = .07), and it was 9.5 v 5.4 months, respectively, in those randomized in the last half of the study (P < .001; Table 2) (K. Yong, personal communication, December 1999). QOL scores and clinical benefit (reflected by use of tumor-related medication, including analgesics, and radiotherapy use) were also better in the docetaxel arm.
Mean Survival of Patients on the TAX 317 Study
Toxicity.
Toxicity was greater on the chemotherapy arm. Twelve patients receiving chemotherapy developed febrile neutropenia (11.5%), all but one of whom received the higher dose of docetaxel. There were three fatal episodes of febrile neutropenia, all at the higher dose. The occurrence of grade 3 or 4 anemia was similar in both arms, as was that of nausea and vomiting. Infection was seen in 34% of the chemotherapy patients, but it was also seen in 21% of the BSC patients. Further details of toxicities and benefits can be obtained from the published results of the study.8
Outcome of the subset studied in the economic analysis.
The data from the 68 patients who were studied in the economic analysis suggest that the subset is representative of the TAX 317 population with respect to survival and toxicity, with a median survival of 7.4 months in all patients treated with chemotherapy and 3.9 months in BSC patients. Radiotherapy use overall in TAX 317 patients was less than in the subset, as was use of morphine and nonmorphine analgesics for pain (Table 3). The higher use of radiotherapy and analgesics in the economic analysis subset may be a reflection of additional data collection beyond study protocol. Although there are small differences between the two groups, these suggest that our estimate of the cost of therapy may overestimate, not underestimate, the true costs.
Tumor-Related Treatment Other Than Chemotherapy: TAX 317 Population and Economic Analysis Subset
Costs
Unit costs are listed in Table 4, and mean resource utilization per patient on each treatment arm is listed in Table 5. Costs were higher on the docetaxel arm, with the IC calculated as $9,577 per patient. This represents the additional cost of the extra 2 months of survival seen on the docetaxel arm (all patients in the study). As seen in Table 6, the major cost on both arms was related to hospitalization costs, and the second major cost source was chemotherapy. Analyzing the second half of the study with docetaxel at a dose of 75 mg/m2, the costs were still higher per patient, with chemotherapy as the major cost source. An IC of $10,804 per patient for the additional 4 months of survival was seen on the lower-dose chemotherapy arm for the latter half of the study.
Unit Costs in 1999 Canadian Dollars
Resource Utilization by Patient Treatment Arm
Summary of Average Costs (in Canadian dollars) per Patient Arm of the TAX 317 Study
Of particular interest, patients randomized to BSC were much less likely to be hospitalized (acute or palliative care institution) for palliation. Rather, they were palliated at home on an outpatient basis more often than those receiving chemotherapy. This difference reflects a proportion of cost savings in this group.
CE Analysis
The IC for all arms from the viewpoint of the health care system is listed in Table 7. The chemotherapy arm was more costly, and the IC of docetaxel was $57,749 per year of life gained. In analyzing the patients accrued in the last half of the study with randomization to docetaxel 75 mg/m2, the IC of docetaxel was $31,776 per year of life gained.
Measures of Economic Evaluation (in 1999 Canadian dollars)
Sensitivity Analyses
Hospitalization rates.
Varying hospital admission rates by 20% yielded a range of CE per life-year gained for all patients on the docetaxel arm of $68,580 in the setting of a 20% increase in admission rates in the docetaxel arm and a 20% decrease in the BSC arm, and it yielded $47,233 for the reverse scenario. Analyzing patients randomized in the second phase of the study to docetaxel 75 mg/m2, the corresponding CE values were $35,943 and $27,609 per year of life gained, respectively. In this analysis, the rates of admission to the hospital were clearly lower in the BSC group, in part because of a clear trend toward community-based palliation as opposed to end-of-life care in the acute setting, which was more common in the chemotherapy arm. If costs of hospitalization were varied by 20%, similar results would be seen.
Docetaxel prices.
By increasing the cost of the drug by 20%, the CE per life-year gained in the chemotherapy arm would be $58,401, and it would be $51,184 if the drug cost were lowered by 20%. By using the recommended dose of docetaxel 75 mg/m2, if the drug cost were increased by 20%, the CE per life-year gained would be $35,257, and if it were lowered by 20% it would be $28,296. Even if the drug were available at no cost, the docetaxel arm would not yield cost savings at either dose level.
Treatment outcome.
The CE figures are most sensitive to changes in survival. When we varied the duration of mean survival ± 20%, the CE per life-year gained ranged from $18,374 in the most favorable scenario for chemotherapy at the dose of 75 mg/m2 to $117,434 in the least favorable scenario. Examining the entire study population, the survival advantage of docetaxel was lost when the mean survival on treatment was decreased by 20% and that for BSC was increased by 20%. However, this scenario is outside the 95% confidence interval surrounding the survival result favoring treatment with docetaxel over BSC from the randomized trial, and it is unlikely. When the reverse was analyzed, there was a CE of $21,965 per life-year gained. Because the sample size is small and mean survival in the BSC arm for the second phase is lower than that of the entire BSC group, it is possible that the CE per life-year gained could be greater than our calculation of $31,776.
Other treatment costs.
If the cost of investigations were varied by 20%, the CE per life-year gained on docetaxel would vary from $54,480 to $61,333 (with costs decreased for chemotherapy and increased for BSC in the first figure and then reversed). By using the recommended dose of 75 mg/m2 from the trial, these figures range from $30,833 to $32,940. Although unnecessary investigations (only for trial protocol and not reasonable as part of routine care) were not included in the analysis, it may be that fewer investigations may be used in day-to-day clinical practice for both the chemotherapy group and the BSC group. This would probably have a minimal effect on CE. Varying the costs of medications and clinic visits resulted in only minor changes in the CE with chemotherapy.
Docetaxel cycles.
If two cycles of docetaxel were administered to patients randomized to the chemotherapy arm, the CE per life-year gained for the entire chemotherapy cohort would be $39,345 ($21,694 for the lower-dose arm). If six cycles were administered, the costs per life-year gained would be $77,341 and $39,368, respectively ($134,335 and $65,879 for 12 cycles). Of note, the median number of cycles received in the trial was four in the lower-dose arm and three in the entire chemotherapy group.
DISCUSSION
As was the case with the National Cancer Institute of Canada (NCIC) BR.5 study of first-line chemotherapy compared with BSC in patients with advanced NSCLC,6 this trial of second-line chemotherapy in NSCLC patients provides a unique opportunity to examine changing patterns of palliative care. Although the results of second-line lung cancer chemotherapy with respect to survival and clinical benefit seem similar in magnitude to those derived from first-line therapy, our analysis suggests that the costs of treatment are higher. In contrast to the result of the BR.5 economic analysis,16 we identified no cost savings in the chemotherapy arm of the TAX 317 study. Although a lower hospital admission rate was seen in chemotherapy-treated patients in the NCIC study of first-line chemotherapy, the converse was true in this analysis. Patients on the docetaxel arm of this analysis were more likely to be admitted to the acute care setting for palliation and were also more likely to be admitted to a palliative care facility than those on the BSC arm. In our analysis, hospitalization and chemotherapy constituted the largest costs in the chemotherapy arm. Patients on the BSC arm of this analysis tended to receive palliation at home. This probably reflects changing patterns in palliative cancer care, and palliation has previously been shown to be more cost-effective at home than in the hospital.22
The recommended dose of docetaxel from the TAX 317 trial was changed from 100 to 75 mg/m2 intravenously every 21 days because of an unacceptable toxic death rate seen at the initial dose of 100 mg/m2. A test of interaction between treatment effect on survival and dose of docetaxel (100 mg/m2 in the first phase and 75 mg/m2 in the second) indicates a trend toward statistical significance (P = .09) (K. Yong, personal communication, December 1999). Given that power to detect interactions is low, this interaction is viewed as an important one, and the lower dose has been recommended for clinical practice. As a result, we believe that our CE estimate for patients treated at the recommended lower dose of docetaxel is more clinically relevant than the CE for all patients treated on the study.
Studies of first-line chemotherapy for NSCLC suggest CE per life-year gained as favorable as a cost savings of $6,172 per life-year gained (1984 Canadian dollars) for combination chemotherapy as part of a randomized trial16 and $632 per life-year gained for single-agent therapy.18 The latter figure was predicted by the Population Health Model, developed by Statistics Canada to simulate medical care costs. However, these figures are based on the NCIC BR.5 data, which had higher admission rates for patients on the BSC arm. This was not the case in the present analysis and is inconsistent with current practice, given the shift away from inpatient palliation of cancer patients. The CE for therapy with vindesine and cisplatin chemotherapy from the NCIC trial in updated 1999 Canadian dollars would be $22,692, which is less than the CE of this study of second-line chemotherapy, which indicated similar benefits. Selected CE values published in the literature are listed in Table 8.
Cost-Effectiveness Analyses in Advanced Non–Small-Cell Lung Cancer (adapted from Goodwin and Shepherd11)
Extrapolating between different health care systems may also be misleading. For example, in the analysis of Smith et al17 of a randomized trial of vinorelbine, vinorelbine plus cisplatin, and vindesine plus cisplatin, costs in the United States system are clearly different from those in Canada. An analysis of resource utilization of patients in the Southwest Oncology Group 9509 trial, comparing vinorelbine and cisplatin with paclitaxel and carboplatin for advanced NSCLC, has recently been completed.10 After analyzing resource utilization for 24 months after randomization by using nationally standardized costs, it was estimated that first-line treatment during that period with vinorelbine and cisplatin cost $33,209 United States dollars in total, compared with $43,522 for paclitaxel and carboplatin.
It is of interest that our cost for second-line treatment of platinum-refractory NSCLC was not inconsistent with estimates of second-line therapy for platinum-refractory ovarian cancer. Although CEAs are not available, a recent study suggests that the total cost of treatment for patients receiving second-line chemotherapy for advanced ovarian carcinoma is in line with costs of therapy in our analysis of lung cancer patients.26 Another study of palliative ovarian cancer patients’ use of health care resources suggests a mean cost per patient of $53,000 (in 1994 Canadian dollars) from the start of second-line therapy.27 Costs of therapy for second-line therapy in advanced colorectal cancer are also similar.28,29
Incremental CE estimates for nonmalignant health care interventions are also in a comparable range, although society’s valuation of benefits for these interventions might differ substantially from benefits derived from palliative chemotherapy. CE estimates range from $53,000 per year of life gained (1994 United States dollars) from dialysis over supportive care for end-stage renal disease, to $251,000 for liver transplantation.11 Thus, our CE of $31,776 for second-line NSCLC chemotherapy is within an acceptable range for health care interventions.30
Our analysis is limited somewhat by its retrospective design. We believe that there may have been some degree of ascertainment bias on the BSC arm after disease progression or discontinuation of trial protocol with respect to use of health care resources, but not survival. Hence the derivation for mean community care costs may represent a significant underestimate of the cost of care in the BSC arm, even though all possible efforts were made to address this retrospectively (contact and follow-up through community services, primary and palliative care physicians, and community oncologists for all patients in the subset). However, should an underestimate of the costs of care on the BSC arm exist in our study, this serves to lessen the incremental difference in costs between the BSC and chemotherapy arms and to render treatment with chemotherapy more cost-effective. The sample size in our analysis for derivation of costs comprised one third of study patients, with potential for bias or reduced precision of CE. As discussed previously, because patients randomized to BSC in the last half of the study had shorter survival than the entire cohort on the BSC arm, the individual patient cost data used in the analysis may overestimate the cost per year of life gained of docetaxel as second-line chemotherapy at its recommended lower dose. Both treatment arms in the study subset in the analysis had higher use of medications and radiotherapy compared with the entire trial population, and although this is unlikely to significantly affect our CE estimate, if anything it would result in an overestimate of the true CE. The use of PMH, a tertiary cancer care center, as the basis for costing in this study may limit its external validity in translation to costs at a smaller center or less specialized hospital, but the magnitude of costing difference is probably minor.
Patients in the TAX 317 trial were permitted to have chemotherapy once off of the study protocol, but only 12% of patients on the BSC arm received chemotherapy at some point between entry onto the trial and death. Although this might be expected to increase the treatment costs of the BSC arm, it might also prolong survival for that arm. This would lessen the effect on CE estimates, decreasing the difference in IC between the docetaxel and BSC arms while also decreasing the incremental survival difference between the two.
Use of the perspective of the Canadian provincial health care systems may decrease the cost of care because of the country’s centralized, public administration and regulation of drug costs. Conversely, though, patients might use more resources than those in a user-pay system. We anticipate that the CE ratios in this analysis will be similar in countries with analogous health care systems. Translation of these results to the United States model, for example, may be complicated by the higher cost of care, rendering the current CE per life-year gained less attractive than in the Canadian system.
Data on utility of outcomes in this trial were not collected prospectively, and, thus, a cost-utility analysis was not undertaken. Clearly, though, when dealing with modest benefits and potential treatment toxicity, such an analysis could make an important contribution to decision-making about second-line chemotherapy treatment.
The decision to pursue second-line chemotherapy for NSCLC should not be based on economic factors alone. The benefits of prolonged survival and improved symptom control must be considered in the context of toxicity and trade-offs and the likelihood of benefit given the individual patient’s performance status and history of disease. The information from this evaluation is intended to provide insight into the cost of second-line therapy in NSCLC. It should also be recognized that the group of lung cancer patients for whom this therapy will be appropriate will be only a subset of those currently receiving first-line palliative chemotherapy and who will have been highly selected by their performance status and biology of disease, including response to initial therapy. The cost of docetaxel for second-line lung cancer chemotherapy is in keeping with the cost of second-line therapy in other tumor sites as published in the literature, and it is within an acceptable range for health care interventions. Good performance status patients previously treated with cisplatin-based therapy should be considered for treatment with docetaxel for a potential survival advantage and palliative benefit of a similar magnitude to that derived from first-line therapy.
Acknowledgments
N.B.L. was supported through the Department of Medical Oncology, Princess Margaret Hospital/University Health Network, in part through a fellowship grant from Aventis Pharma Inc, Laval, Quebec, Canada.
ACKNOWLEDGMENT
We thank P. Warde, J. Nagai, S. Boeker, A. Stewart, T. Marincic, G. Prendergast, M. Mereles-Pulcini, M. Keresteci, M.A. Brittain, J. Cowan, R. Villa, J. Armstrong, G. Wilson, and the Ontario Cancer Registry for their contributions to data collection and costing.
- Received January 23, 2001.
- Accepted October 15, 2001.
