|Cost effectiveness of paclitaxel/cisplatin compared with cyclophosphamide/cisplatin in the treatment of advanced ovarian cancer in Belgium
|Neymark N, Gorlia T, Adriaenssen I, Baron B, Piccart M
This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.
Two treatments for women with advanced ovarian cancer were examined. These were paclitaxel-cisplatin (TP) and cyclophosphamide-cisplatin (CP). The dose of cisplatin was 75 mg/m2 and was identical in both treatments. The initial dose of cyclophosphamide was 750 mg/m2, while the dose of paclitaxel was 175 mg/m2. Both treatments were administered for six cycles, repeated every 3 weeks.
Economic study type
The study population comprised patients with histologically verified advanced epithelial ovarian cancer and International Federation of Gynaecology and Obstetrics (FIGO) stage IIB, IIC, III or IV disease, with or without successful debulking surgery. A detailed list of clinical exclusion criteria was provided in the primary trial.
The setting was secondary care. The economic study was carried out in Belgium.
Dates to which data relate
The collection of data on both effectiveness and resource use started in 1995. No formal closing date for the trial was reported. The price year was 1998.
Source of effectiveness data
The effectiveness evidence came from a single study, the main details of which were published in a separate paper (see Other Publications of Related Interest).
Link between effectiveness and cost data
The costing was performed prospectively on a subgroup of patients used in the effectiveness analysis.
Preliminary power calculations were conducted. These indicated that a total of 600 patients would be required to detect a 33% improvement in the main outcome measure in the standard arm of the trial, with an 80% power at a 5% significance level (two-sided test). The method of sample selection was not reported. Overall, an initial study sample of 680 patients was enrolled into the analysis. There were 338 patients with a mean age of 58 years (range: 22 - 85) in the CP group, of which 72.5% were FIGO stage III. There were 342 patients with a mean age of 58 years (range: 23 - 79) in the TP group, of which 74.9% were FIGO stage III. However, in the CP group, 336 started the treatment and 330 were fully eligible, while in the TP group, 339 started the treatment and 338 were fully eligible.
This was an open randomised controlled trial, which was carried out in 73 institutions located in Europe and Canada. The method used for randomisation was not reported, but the authors stated that stratification factors included the treating institution, the FIGO stage, the amount of residual disease, the World Health Organisation (WHO) performance status, and the tumour grade. The median follow-up was 38.5 months for the whole group. A total of 79% of the patients in the CP group and 85% of those in the TP group completed the protocol therapy.
Analysis of effectiveness
The basis of the analysis of the clinical study was intention to treat. This means that all randomly allocated patients were analysed according to the arm to which they were assigned. The primary health outcome used in the effectiveness analysis was the progression-free survival time (PFS). A Cox regression analysis was performed to adjust the treatment comparison for known prognostic factors. The authors stated that groups were well balanced for age, WHO performance status, FIGO stage, amount of residual disease following laparotomy, presence of measurable disease, cell type and tumour grade.
The median PFS was 15.5 months in the TP group and 11.5 months in the CP group, (p=0.0005).
Median overall survival was 35.6 months in the TP group and 25.8 months in the CP group, (p=0.016).
The Cox regression analysis showed a significant 27% reduction in instantaneous death rate with TP over CP (hazard ratio 0.73, 95% confidence interval: 0.60 - 0.89).
The effectiveness analysis showed that the TP treatment was effective in improving overall survival and PFS in patients with advanced ovarian cancer.
Measure of benefits used in the economic analysis
The benefit measure used in the economic analysis was mean survival time. It was derived from the effectiveness analysis, where it represented a secondary outcome measure (the primary end point being PFS). It was selected to make the results of the present analysis more easily comparable with those from other studies. Overall survival was discounted at a rate of 3%. The mean survival time was estimated using the criterion of the restricted means method suggested by Karrison (1997).
A 3% discount rate was used since the costs were incurred over more than two years (see data on the length of follow-up). The unit costs and the quantities of resources were reported separately. The health service and patient costs included in the analysis were for hospitalisation (both inpatient and outpatient fees), the laboratory, radiology, patient co-payments for hospitalisation, ambulatory visits, laboratory sample, X-ray, computerised tomography scan, electrocardiogram, laparotomy, laparoscopy, chemotherapy drugs (both inpatient and outpatient costs) and anti-emetics. The cost/resource boundary adopted in the study was that of the Belgian health insurance and financing system.
The unit costs were estimated on the basis of nomenclature of scheduled tariffs in Belgium, and each fee covered the costs of equipment used in performing the procedure or providing the service. The quantities were estimated on the basis of a sub-sample of 231 patients recruited by the European Organization for Research and Treatment of Cancer (EORTC) centres (participating in the trial recruitment). A smaller sub-sample comprising 64 of the 231 patients in the EORTC centres was then used for a more detailed cost data collection. The dates during which the resource use data were gathered were unclear. The price year was 1998. The method proposed by Lin et al. (1997) was used to correct for censoring cost data, and subintervals of one day each were used to minimise the bias of the proposed approach.
Statistical analysis of costs
The costs were treated deterministically in the base-case.
The indirect costs were not included in the analysis.
A bootstrap analysis was conducted to assess the impact of simultaneous variations in costs and survival on the estimated incremental cost-effectiveness ratios (ICERs). An acceptability curve was also derived from the plot of bootstrap replicates.
Estimated benefits used in the economic analysis
The time point of restriction, which was estimated using the criterion of the restricted means method, was 44.8 months or 3.73 years. Thus, the mean survival was 30.6 months in the TP group and 26.6 months in the CP group. Consequently, the survival gain of TP over CP was 4.0 months.
Overall, the consumption of resources was similar in the two treatment groups. The total costs during the whole trial period were EUR 23,792 (+/- 18,558) in the TP group and EUR 16,775 (+/- 15,869) in the CP group.
The main cost drivers were chemotherapy drug acquisition (higher in the TP group) and second-line treatment (higher in the CP group).
In the restricted observation period (3.73 years), the average total costs were EUR 23,324 in the TP group and EUR 16,529 in the CP group.
The drug acquisition costs explained most of the difference in the total costs of the two treatments.
Synthesis of costs and benefits
An incremental cost-effectiveness analysis was conducted to combine the costs and survival of the two treatments.
The incremental cost per life-year gained with TP over CP was EUR 20,385 (point estimation).
A bootstrap analysis showed that TP improved survival in 96.7% of the replicates and dominated CP in 3.3% of the replicates.
At a threshold value of EUR 12,400 for the decision-maker's willingness to pay for the treatment, only about 20% of the ICERs would be acceptable;
at a threshold value of EUR 24,88, the proportion of acceptable ICERs would be 60%; and
at a threshold value of EUR 50,000, the proportion of acceptable ICERs would be 87%.
Compared with cyclophosphamide-cisplatin (CP), the paclitaxel-cisplatin (TP) treatment was effective in improving survival in patients with advanced ovarian cancer at an average extra cost of about EUR 20,000 per life-year gained from the perspective of the Belgian health insurance system.
CRD COMMENTARY - Selection of comparators
The rationale for the choice of the comparators was clear. CP represented the standard treatment for women with advanced ovarian cancer, while TP was a new combination therapy introduced in the second half of the 1990s. However, after the encouraging results of a US study published in 1996, TP became the standard care in the USA. You should decide whether they are widely used treatments in your own setting.
Validity of estimate of measure of effectiveness
The effectiveness analysis used a multicentre randomised controlled trial, which was appropriate for the study question. Overall, the internal validity of the study was quite good. Appropriate power calculations were performed initially, so the likelihood that the results were due to chance is low. Selection bias should be low since the patients were randomised. However, the methods of randomisation and sample selection were not reported. There was some loss to follow-up, but it was similar for the two groups and an intention to treat analysis was performed, so attrition bias should also be low. Measurement bias is not an issue for survival outcomes. The study groups were comparable at baseline so the likelihood of confounding is low.
Validity of estimate of measure of benefit
The life-years gained were used as the benefit measure in the economic analysis. This permits the results of the present study to be compared with the benefits of other interventions. Appropriate discounting was conducted in view of the long time horizon of the analysis.
Validity of estimate of costs
The cost analysis was satisfactorily conducted. The perspective adopted in the study was reported, and it appears that all the relevant categories of costs have been included in the analysis. Actual costs (not charges) were used and discounting was performed, which was appropriate since the costs were incurred over more than 3 years. As with the benefits, the authors discussed the problem of using censored cost data and explicitly reported the approach used to deal with such an issue. The costing was conducted prospectively on a subgroup of patients used in the effectiveness analysis. However, the subgroup was considered to be representative of the whole sample of patients. All the above indicate high internal validity of the cost estimates. The unit costs and the quantities of resources used were reported separately and the price year was given. These factors enhance the reproducibility of the study in other settings. In terms of uncertainty in the results, the costs were treated deterministically in the base-case, but statistical analyses were conducted to examine the impact of variations in the unit costs.
A number of factors enhanced the external validity of the economic study. First, the authors compared their findings with those from published studies and stated that the present analysis offered several methodological improvements over prior analyses. Second, the authors discussed the issue of the generalisability of the study results to other settings and stressed the point that the unit costs may vary across different contexts, thus the comparison was focused on effectiveness and resource consumption.
Some limitations were also acknowledged. First, the authors stated that the specific approach used for dealing with censored data might not be free of bias. It was chosen because it represented a conservative method, and thus estimates of survival may have been underestimated. Second, the cost data were collected from very different health care settings and were then simply aggregated as if they had been obtained from a single centre. Thus, this pooling might have resulted in the cost estimates not being representative of any of the study centres, although the statistical power of the cost analysis was enhanced. Third, the authors justified the adoption of a limited perspective (that of the public insurance system) as the relevant one for the actual decision-making.
Implications of the study
The study confirms results showing that TP represents the treatment of choice for women with advanced ovarian cancer. The actual adoption of such treatment in clinical settings depends on the decision-maker's willingness to pay. The authors pointed out that the introduction of generic competitors for paclitaxel might further reduce the ICER of the TP treatment.
Source of funding
The clinical study was funded by Bristol-Myers Squibb.
Neymark N, Gorlia T, Adriaenssen I, Baron B, Piccart M. Cost effectiveness of paclitaxel/cisplatin compared with cyclophosphamide/cisplatin in the treatment of advanced ovarian cancer in Belgium. PharmacoEconomics 2002; 20(7): 485-497
Other publications of related interest
Piccard MJ, et al. Randomized intergroup trial of cisplatin-paclitaxel versus cisplatin-cyclophosphamide in women with advanced epithelial ovarian cancer: three year results. Journal of the National Cancer Institute 2000;92:699-708.
Karrison T. Use of Irwin's restricted mean as an index for comparing survival in different treatment groups: interpretation and power consideration. Controlled Clinical Trials 1997;18:151-67.
Lin DY, et al. Estimating medical costs from incomplete follow-up data. Biometrics 1997;53:419-34.
Subject indexing assigned by NLM
Antineoplastic Combined Chemotherapy Protocols /economics /therapeutic use; Cisplatin /administration & Cost-Benefit Analysis; Cyclophosphamide /administration & Female; Health Care Costs; Humans; Ovarian Neoplasms /drug therapy /mortality; Paclitaxel /administration & dosage; dosage; dosage
Date bibliographic record published
Date abstract record published