Cost-Effectiveness of Vinorelbine Alone or Vinorelbine Plus Cisplatin for Stage IV NSCLC
Cost-Effectiveness of Vinorelbine Alone or Vinorelbine Plus Cisplatin for Stage IV NSCLC
In 1996, the Health Protection Branch of Health, Canada, approved the use of vinorelbine (Navelbine, NVB) for the treatment of advanced non-small-cell lung cancer (NSCLC). Among the data that led to this approval was a large European clinical trial that demonstrated that treatment with vinorelbine (Navelbine) in combination with cisplatin (Platinol) (NVB-P) resulted in a higher response rate and longer survival than vindesine plus cisplatin (VDS-P) (P = .04), or vinorelbine (NVB) alone (P = .01). The Ontario Practice Guideline Initiative Lung Disease Site Group (LDSG) reviewed the evidence on the efficacy of vinorelbine alone and in combination with cisplatin and concluded that this new semisynthetic vinca alkaloid was active against NSCLC and, in combination with cisplatin, was superior to standard chemotherapy (VDS-P). Inferentially, it could also be concluded that NVB-P is superior to best supportive care (BSC), as VDS-P was superior to BSC in a National Cancer Institute of Canada (NCIC) randomized controlled clinical trial. The practice guideline developed by the LDSG recommended that vinorelbine/cisplatin be offered to medically appropriate patients with advanced NSCLC after a full discussion with the patient of the potential harms and benefits of treatment. However, all new anticancer drugs are relatively expensive, and there were concerns that this new agent would significantly increase the costs of care of patients with incurable NSCLC. It was, therefore, felt that it would be useful to estimate the total burden of cost of vinorelbine assuming that all metastatic NSCLC patients in Canada were treated. Furthermore, it was felt that information on the relative cost-effectiveness of vinorelbine-based chemotherapy compared to other treatment options would be useful to clinicians who have to make decisions about which patients to treat and which chemotherapy agents to use.
A study was therefore undertaken in collaboration with Statistics Canada using their POpulation HEalth Model (POHEM).[4-6] The economic analyses were possible because detailed information on drug and other resource utilization and raw survival data were available from the trial reported by Le Chevalier et al. In addition, a cost analysis had been undertaken in the NCIC clinical trial (BR.5), which compared vindesine-cisplatin to BSC, and this provided data on hospitalization and follow-up costs in the Canadian context.
Materials and Methods
Statistics Canada is developing a POpulation HEalth Model to simulate the common illnesses of Canadians. In addition to lung and breast cancer, POHEM will eventually incorporate other malignancies, such as colon and prostate cancer, cardiovascular disease, arthritis and dementia. POHEM generates a synthetic cohort of people with the demographic characteristics, risk factor exposures, and health histories typical of Canadians. POHEM assigns synthetic patients to diseases proportional to their incidence in the Canadian population and, for lung cancer cases, with the same distribution of histology and stage as found in Canadian lung cancer cases. The model then assigns data on diagnostic methods, treatment and its outcomes, health care utilization, and direct care costs based on information extracted from provincial health care data bases and individual institutions. The perspective of the costing model is that of the government as payor in a universal health care system.
Each simulated patient in the POHEM lung cancer submodule is assigned a standard set of diagnostic tests, procedures, and visits, including an initial medical contact with a family physician, specialist consultation and diagnostic work-up appropriate for disease stage. For patients presenting with stage IV NSCLC, the number of physician assessments and associated fees were determined from the number of chemotherapy treatments given and the expected number of physician encounters during these treatments. The frequency of visits for patients receiving best supportive care was extracted from the NCIC BR.5 study. The costs of laboratory and imaging studies were estimated from their reported frequency in the methods section of the study protocols and the fees listed in the Ontario Schedule of Benefits. It was assumed that tests were not duplicated and that treatment was uncomplicated.
The European randomized trial provided careful documentation on the dosage and administration schedules of NVB, NVB-P, and VDS-P.  From this it was possible to calculate the cost of chemotherapy and its administration. In the study, patients were randomly assigned to one of three regimens: (1) vinorelbine alone at a dosage of 30 mg/m² intravenously (IV) weekly (NVB); (2) vinorelbine, 30 mg/m² IV weekly plus cisplatin, 120 mg/m² on day 1 and day 29, then every 6 weeks (NVB-P); or (3) vindesine, 3 mg/m² weekly for 7 weeks, then every 2 weeks, plus cisplatin at the same dosage and schedule as for NVB-P (VDS-P).
The time spent by pharmacy and nursing personnel in the preparation and administration of the various chemotherapy regimens was determined using the Canadian Management Information System for pharmacy workload and by measuring the time taken to administer NVB and NVB-P in the chemotherapy treatment unit at the Ottawa Regional Cancer Centre (ORCC). Nursing and pharmacy personnel costs were then determined by multiplying the amount of time expended by the hourly salary rate, including benefits, at the ORCC in 1993. Chemotherapy and antiemetic (ondansetron [Zofran]/dexamethasone) drug costs were based on their acquisition costs at the ORCC in 1993. The equipment and supplies necessary to prepare and deliver the chemotherapy were quantified and costed by the pharmacy and nursing staff of the ORCC.
Because etoposide (VePesid)/cisplatin (VP-16-P) and vinblastine (Velban)/cisplatin (VLB-P) have frequently been used by Canadian oncologists who treat non-small-cell lung cancer, it was felt that it would be important to determine the costs of these treatment strategies relative to the vinorelbine-based regimens. The costs associated with the administration of etoposide/cisplatin (etoposide, 100 mg/m² IV on days 1-3; cisplatin, 25 mg/m² on days 1-3 every 3 weeks) and vinblastine/cisplatin (vinblastine, 5 mg/m² IV on days 1 and 8, and cisplatin, 100 mg/m² IV on day 1 every 4 weeks) were estimated using the same approach as described above for NVB and NVB-P.
Hospitalization and Other Costs
In both the NCIC and Le Chevalier studies, high-dose cisplatin was administered in hospital.[1,3] For the purposes of the cost estimates in this study, it was assumed that each hospitalization required 2 days. The daily costs of hospitalization for chemotherapy were based on the actual costs of giving inpatient chemotherapy at the Princess Margaret Hospital in Toronto, Canada, as determined during a previous costing study. These were adjusted to 1993 costs by multiplying the 1984 costs by the increase in the average daily cost of operating tertiary health care facilities in Canada between 1984 and 1993 (a 12.4% increase). Because there are significant barriers to the admission of patients for inpatient chemotherapy, we estimated the costs of NVB-P administration using an outpatient schedule currently used in an NCIC clinical trial (vinorelbine, 30 mg/m² weekly, and cisplatin, 50 mg/m² on days 1 and 8 every 4 weeks).
The cost of radiotherapy in the management of metastatic NSCLC, as well as clinic overhead costs (hotel costs), were extracted from the study of the costs of chemotherapy and best supportive care determined during the BR.5 study and adjusted to 1993 dollars by multiplying by the increase in the consumer price index (41.1%). The cost of best supportive care and the terminal care costs for stage IV chemotherapy-treated patients were extracted from the same study. Importantly, patients who received chemotherapy required less palliative radiotherapy and fewer inpatient bed days. This information was inserted into the model, and we assumed, based on the clinical trial data, that chemotherapy-treated patients, including those receiving NVB and NVB-P, used 17.1 hospital bed days during terminal care, as opposed to 23.6 days for those receiving best supportive care.
In order to estimate the survival gain of various chemotherapy regimens over best supportive care, the survival data for all patients treated with NVB, NVB-P and VDS-P were obtained on computer diskette from the principal investigators of the European trial. Based on these raw data, survival curves were modeled using a Weibull distribution. The survival data for stage IV patients treated in the NCIC BR.5 study were obtained from the NCIC clinical trials office and similarly modeled. Of interest was the fact that the survival curves for stage IV patients treated with VDS-P in the European and NCIC studies were superimposable. Based on other randomized trials in the literature, we assumed that the survival of patients treated with etoposide/cisplatin and vinblastine/cisplatin was the same as that of VDS-P-treated patients.[9,10]
Cost and Cost-Effectiveness
Assuming that all stage IV lung cancer patients in Canada would be treated with chemotherapy and knowing the total number of stage IV patients diagnosed in 1992, we estimated the cost of treating individual patients and the total burden of care for each of the chemotherapy treatment regimens and for best supportive care. The cost-effectiveness of these various approaches was then determined by dividing the total cost of care (including diagnostic work-up, follow-up, relapse and terminal care costs) by the estimated survival gain relative to BSC. We also undertook an analysis of cost-effectiveness relative to the standard regimens of vinblastine/cisplatin and etoposide/cisplatin.
This cost analysis was undertaken with financial support from Glaxo Wellcome Inc., Canada. No restraints were placed on the investigators in their conduct of the study and funding was provided to the Ottawa Regional Cancer Centre. The Centre in turn contracted with Statistics Canada for the use of the POHEM microsimulation software and the provision of analytical expertise for the economic analysis.