Objectives: To assess the clinical and cost-effectiveness of adjuvant carmustine wafers (BCNU-W) and also of adjuvant and concomitant temozolomide (TMZ), compared with surgery with radiotherapy.
Data sources: Electronic databases were searched up to August 2005.
Review methods: Included trials were critically appraised for key elements of internal and external validity. Relevant data were extracted and a narrative synthesis of the evidence produced. Where possible, data on absolute survival at a fixed time point were meta-analysed using a random effects model. A Markov (state transition) model was developed to assess the cost-utility of the two interventions. The model compared BCNU-W or TMZ separately with current standard treatment with surgery and radiotherapy. The simulated cohort had a mean age of 55 years and was modelled over 5 years.
Results: Two randomised controlled trials (RCTs) (n = 32, n = 240) and two observational studies of BCNU-W compared with placebo wafers as adjuvant therapy to surgery and radiotherapy for newly diagnosed high-grade glioma were identified. All the studies were in adults and provided data on 193 patients who had received BCNU-W. The RCT findings excluded under 65-year-olds and those with a Karnofsky Performance Status of less than 60. The largest multi-centre RCT suggested a possible survival advantage with BCNU-W among a cohort of patients with grade III and IV tumours, adding a median of 2.3 months [95% confidence interval (CI) -0.5 to 5.1]. However, analysis using per-protocol, unstratified methods shows this difference to be not statistically significant (HR 0.77, 95% CI 0.57 to 1.03, p = 0.08). Long-term follow-up suggests a significant survival advantage using unstratified analysis. No difference in progression-free survival (PFS) was demonstrated. Subgroup analysis of those with grade IV tumours also showed no significant survival advantage with BCNU-W [hazard ratio (HR) 0.82, 95% CI 0.55 to 1.11, p = 0.20, unstratified analysis]. It is estimated that the cost of surgery and radiotherapy, with follow-up, treatment of adverse effects and end of life care is around 17,000 pounds per patient. Treatment with BCNU-W adds an additional 6600 pounds. Across the modelled cohort of 1000 patients, use of BCNU-W costs an additional 6.6 million pounds and confers an additional 122 quality-adjusted life-years (QALYs). On average, that is 6600 pounds per patient for 0.122 QALYs (6.3 quality-adjusted life-weeks). The base-case incremental cost-effectiveness ratio (ICER) is 54,500 pounds/QALY. In probabilistic sensitivity analyses, BCNU-W was not cost-effective in 89% of the simulations assuming a willingness to pay threshold of 30,000 pounds/QALY. In 15% of simulations, BCNU-W was dominated (i.e. did more harm than good, conferring fewer QALYs at greater cost). The cost-effectiveness acceptability curve (CEAC) suggests that it is very unlikely to be the most cost-effective option at normal levels of willingness to pay (11% probability at 30,000 pounds/QALY), only becoming likely to be the most cost-effective option at much higher levels of willingness to pay (50% probability at 55,000 pounds/QALY). Two RCTs (n = 130, n = 573) and two observational studies were included, giving evidence for 429 adult patients receiving TMZ. Currently, TMZ is licensed for use in those with newly diagnosed grade IV gliomas only. The RCTs excluded those with lower performance status and, in the larger RCT, those older than 70 years. TMZ provides a small but statistically significant median survival benefit of 2.5 months (95% CI 2.0 to 3.8), giving an HR of 0.63 (95% CI 0.52 to 0.75, p < 0.001). At 2 years, 26.5% of patients treated with TMZ were alive compared with 10.4% of those in the control arm. Median PFS is also enhanced with TMZ, giving a median 1.9 months' advantage (95% CI 1.4 to 2.7, p < 0.001). No analysis of the subgroup of patients with confirmed grade IV tumours was undertaken. Subgroup analysis of patients by O6-methylguanine-DNA methyltransferase (MGMT) activity showed a significant treatment advantage for those with reduced MGMT activity but not for those with normal activity, although this analysis was based on a selected sample of patients and the test used has proved difficult to replicate. A median gain of 6.4 (95% CI 4.4 to 9.5) more life-months is seen with TMZ among those with reduced MGMT, giving an HR of 0.51 (p < 0.007). PFS is increased by a median of 4.4 months (95% CI 1.2 to 6.3), giving an HR of 0.48 (p = 0.001). The model shows a cost per patient for being treated with surgery, radiotherapy and including adverse effects of treatment and end of life care of around 17,000 pounds per patient. TMZ in the adjuvant and concomitant phase adds an additional cost of around 7800 pounds. Across the modelled cohort of 1000 patients, use of TMZ costs an additional 7.8 million pounds and confers an additional 217 QALYs. For the average patient this is 7800 pounds for an additional 0.217 QALYs (11 quality-adjusted life-weeks). The base-case ICER is 36,000 pounds/QALY. Probabilistic sensitivity analyses shows that TMZ was not cost-effective in 77% of the simulations. The CEAC suggests that there is a 23% chance that TMZ is the most cost-effective option at a willingness to pay level of 30,000 pounds/QALY, rising to be more cost-effective than no TMZ at slightly higher levels (50% probability at 35,000 pounds/QALY).
Conclusions: BCNU-W has not been proven to confer a significant advantage in survival for patients with grade III tumours when treated with the drug, compared with placebo. There does not appear to be a survival advantage for patients with grade IV tumours. No increase in PFS has been shown. Limited evidence suggests a small but significant advantage in both overall survival and PFS with TMZ among a mixed population with grade IV and grade III (7-8%) tumours. However, it remains unclear whether this is true in grade IV tumours alone. On the basis of best available evidence, the authors consider that neither BCNU-W nor TMZ is likely to be considered cost-effective by NHS decision-makers. However, data for the model were drawn from limited evidence of variable quality. Tumour type is clearly important in assessing patient prognosis with different treatments. Grade IV tumours are commonest and appear to have least chance of response. There were too few grade III tumours included to carry out a formal assessment, but they appear to respond better and drive results for both drugs. Future use of genetic and biomarkers may help identify subtypes which will respond, but current licensing indications do not specify these. Further research is suggested into the effectiveness of these drugs, and also into areas such as genetic markers, chemotherapy regimens, patient and carer quality of life, and patient views on survival advantages vs treatment disadvantages.