3 The manufacturer's submission
3.1 The main evidence for the clinical effectiveness of vinflunine was from 1 open-label, phase III, randomised controlled trial (study 302, the registration trial) that compared vinflunine plus best supportive care with best supportive care alone in patients with advanced or metastatic transitional cell carcinoma of the urothelial tract whose disease had progressed after platinum-based chemotherapy. Results from 2 open-label, single-arm, phase II studies (study 202 and CA001) were also provided. The manufacturer's submission highlighted issues around using best supportive care alone as the control arm in study 302. In particular, patients had to be fit enough to receive chemotherapy but willing to accept randomisation to best supportive care. The manufacturer stated that patients in the trial had a poorer prognosis, as indicated by the high percentage (approximately 74%) of patients with visceral involvement in both groups in the trial. The manufacturer also highlighted that there is currently no standard chemotherapy regimen for patients with advanced transitional cell carcinoma of the urothelial tract whose disease has progressed after a prior platinum-containing chemotherapy, and there is a lack of trial evidence of survival advantage from chemotherapy in this clinical situation. Therefore no standard active treatments were available to use as a control, and best supportive care was considered the most appropriate comparator for vinflunine.
3.2 Patients were included in study 302 if they had progressive disease after at least 2 cycles of platinum-based first-line chemotherapy (or after 1 cycle if there was clear evidence of disease progression at this point), an ECOG performance status of 0 or 1, and an estimated life expectancy of at least 12 weeks. Previous systemic chemotherapy must have been stopped at least 30 days before randomisation. Patients were excluded if they had received more than 1 previous systemic chemotherapy for advanced or metastatic disease, or if they had been treated with neoadjuvant or adjuvant chemotherapy. Patients were randomised on a 2:1 basis to vinflunine plus best supportive care (hereafter called the vinflunine arm) or best supportive care alone (hereafter called the best supportive care arm). Patients in the vinflunine arm initially received 320 mg/m2 every 21 days via infusion, but the protocol was subsequently amended to allow a lower starting dose (280 mg/m2) in patients at greater risk of haematological toxicity. Best supportive care included palliative radiotherapy, antibiotics, analgesics, corticosteroids and blood transfusions.
3.3 A total of 370 patients were enrolled into the study (253 in the vinflunine arm and 117 in the best supportive care arm). The median age of study participants was 64 years, and 79% were male. Most baseline characteristics were similar across the 2 treatment arms. However, a greater proportion of patients in the vinflunine arm had an ECOG performance status of 1 compared with the best supportive care arm (72% and 62% respectively) although this difference was not statistically significant. Cisplatin was the most common first-line platinum treatment and had been received by more patients in the best supportive care arm than in the vinflunine arm (73% and 65% respectively), although this difference was not statistically significant. More patients in the vinflunine arm than in the best supportive care arm had received carboplatin as first-line platinum treatment (30% and 20% respectively; p=0.044).
3.4 Study results were provided for 4 study populations, only 2 of which are presented here: the intention-to-treat (ITT) population, which included all randomised patients, and the 'eligible ITT population'. The latter excluded 13 patients who were found, upon retrospective review of the patient inclusion criteria, not to have progressive disease at the time of entry into the study, and who therefore should not have been randomised (4 patients in the vinflunine arm and 9 in the best supportive care arm; 3 of the 4 excluded patients in the vinflunine arm and 6 of the 9 excluded patients in the best supportive care arm were also ineligible because they had received neoadjuvant or adjuvant chemotherapy). The primary outcome of study 302 was median overall survival. For the ITT population, this was 6.9 months in the vinflunine arm compared with 4.6 months in the best supportive care arm (hazard ratio [HR] 0.88, 95% confidence interval [CI] 0.69 to 1.12, p=0.2868). A pre-planned multivariate analysis, adjusting for a number of prognostic factors (performance status, visceral invasion, alkaline phosphatase, haemoglobin and prior pelvic irradiation), showed a statistically significant overall survival benefit for vinflunine (HR 0.77, 95% CI 0.61 to 0.98, p=0.036). For the eligible ITT population, median overall survival was 6.9 months in the vinflunine arm and 4.3 months in the best supportive care arm (HR 0.78, 95% CI 0.61 to 0.99, p=0.0403). An extended multivariate analysis was also done, adjusting for the same prognostic factors outlined above plus additional baseline characteristics such as age, sex and disease stage at diagnosis. This analysis also showed a statistically significant overall survival benefit for vinflunine (HR 0.68, 95% CI 0.52 to 0.88, p=0.0035).
3.5 Progression-free survival for the ITT population was 3.0 months in the vinflunine arm compared with 1.5 months in the best supportive care arm (HR 0.68, 95% CI 0.54 to 0.86, p=0.0012). In the vinflunine arm, 46.5% of patients had stable disease after second-line treatment, 44.9% had progressive disease, and 8.6% had a partial or complete response. In the best supportive care arm, 27% of patients had stable disease, 73% had progressive disease, and none had a partial or complete response. These outcomes were not reported for the eligible ITT population. After disease progression, 29% of patients in the vinflunine arm and 34% of patients in the best supportive care arm received palliative chemotherapy; 60% of these re-treated patients received multi-agent chemotherapy.
3.6 Quality of life was assessed using the cancer-specific European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30 questionnaire. This was done at study entry and at the end of cycles 1, 2, 4 and 6 for both arms. There were no statistically significant differences in overall EORTC QLQ-C30 global health status score between the 2 arms (p=0.658).
3.7 In the 2 phase II, single-arm trials (study 202 and CA001), vinflunine was given every 21 days at a dose of 320 or 280 mg/m2. In study 202 (n=58), the overall response rate (partial or complete response) was 18%, median progression-free survival was 3 months (95% CI 2.4 to 3.8 months) and median overall survival was 6.6 months (95% CI 4.8 to 7.6 months). In CA001 (n=151), the overall response rate was 14.6% (95% CI 9.4% to 21.2%), median progression-free survival was 2.8 months (95% CI 2.6 to 3.8 months) and median overall survival was 7.9 months (95% CI 6.7 to 9.7 months).
3.8 The most common adverse events (any grade) associated with vinflunine across the 3 phase II and phase III studies (n=450) were constipation (55%), nausea (41%), infusion-site reactions (28%), stomatitis/mucositis (27%) and vomiting (27%). Overall, there were 6 deaths related to treatment (1.3%), of which 4 were a result of myelotoxicity. Four treatment-related deaths occurred in the vinflunine arm of study 302. Grade 3 or 4 toxicities relating to neutropenia, anaemia and constipation occurred in 50%, 19% and 16% respectively of patients in the vinflunine arm of study 302, compared with 1%, 8% and 1% of patients respectively in the best supportive care arm. Febrile neutropenia occurred in 6% of patients receiving vinflunine (none in the best supportive care arm).
3.9 The manufacturer submitted an economic analysis comparing vinflunine plus best supportive care with best supportive care alone. The manufacturer's model was similar to a Markov cohort model in that it included 3 health states: pre-progression, post-progression and dead. The model calculated the proportion of patients expected to be in each health state, based on the estimated survival curves for the eligible ITT population from study 302. The model assumed that treatment is administered in cycles of 21 days until disease progression, major toxicity or other reason for treatment discontinuation, or death (if occurring before progression). All patients are assumed to be in a pre-progression health state at model entry (baseline). Patients who experience disease progression are assumed to stop treatment with vinflunine and remain in the post-progression state until death. The cycle length of the model was 1 day and the time horizon was 5 years.
3.10 Costs of vinflunine were based on the mean dose (287 mg/m2), the mean body surface area (1.85 m2) and the mean number of treatment cycles (4.2) in study 302. Other treatment costs included administration for intravenous infusion every 21 days in an outpatient setting, complete blood count before drug administration and constipation prophylaxis. Drug wastage was assumed to be zero in the base-case analysis. The total per-patient cost of treatment with vinflunine included in the model was £10,207. Costs for 3 common adverse events were included in the model: constipation (£39; based on 1 GP consultation and 1 pack of laxatives), febrile neutropenia resulting in hospitalisation (£3538; NHS HRG [healthcare resource group] costs) and abdominal pain resulting in hospitalisation (£557; NHS HRG costs).
3.11 Costs for best supportive care were calculated for the pre-progression and post-progression health states. For the pre-progression health state, best supportive care included: home visits by a GP, community nurse, health home visitor and dietician, an oncologist follow-up visit (assumed to be the same for each treatment group) and palliative radiation therapy (which differed by treatment group). For the post-progression health state, best supportive care included home visits by a GP, community nurse, health home visitor and dietician, a non-consultant oncologist follow-up visit, hospice care, pain medication (assumed to be the same for each treatment group), and palliative radiation therapy and palliative chemotherapy (which differed by treatment group).
3.12 The pre-progression utility values used in the manufacturer's submission were based on responses to 1 item from the EORTC QLQ-C30 questionnaire used in study 302, which asked patients to rate their overall quality of life in the previous week. Responses were transformed to health-state utilities using a published regression model relating this measure to utility values from a time-trade-off analysis in a sample of US cancer patients and their relatives. Post-progression utility values were taken from a study reporting EQ-5D values in 1270 terminally ill cancer patients with painful bone metastases or poor-prognosis non-small-cell lung cancer. Disutility values associated with treatment-related adverse events were not included in the model.
3.13 In the manufacturer's base case, the incremental cost-effectiveness ratio (ICER) for vinflunine plus best supportive care compared with best supportive care alone was £100,144 per quality-adjusted life year (QALY) gained (incremental cost of £13,071 and incremental benefit of 0.131 QALYs). The manufacturer's deterministic sensitivity analyses showed that vial price and pre-progression utility values had the greatest impact on the base-case ICER. When a vial price of £0 was used, the ICER was £27,478 per QALY gained. When a pre-progression utility of 0.4 was used (instead of 0.65), the ICER was £133,094 per QALY gained. The ICER was also sensitive to assumptions about the number of vinflunine treatment cycles (£70,233 per QALY gained when 3 cycles were costed) and vial wastage (£121,095 per QALY gained when wastage was accounted for). The manufacturer's probabilistic sensitivity analysis suggested that vinflunine had a 6% probability of being cost effective at a threshold of £30,000 per QALY gained when compared with best supportive care alone.
3.14 The ERG considered the modelling approach and model structure used by the manufacturer to be appropriate and reasonable; however, it commented on a number of areas of uncertainty. The ERG stated that the modelled population reflected that of the pivotal trial (study 302) but may not be representative of the majority of patients whose disease progresses after first-line therapy. This was because patients who had received prior neoadjuvant or adjuvant platinum-based chemotherapy had been excluded from the trial. The ERG commented that best supportive care may not be the most appropriate comparator because alternative second-line treatments are available in UK clinical practice. However, the ERG noted that best supportive care was the comparator specified in the scope for the appraisal, and that there are no randomised controlled trials of relevant comparators for the population of interest. The ERG stated that data from the ITT population of study 302 may have been a more appropriate basis for the economic model than the data from the eligible ITT population that were used by the manufacturer. It also stated that the utility values used did not fit with the preferred NICE reference case, and that there is considerable uncertainty around these estimates because standard methods were not used. The ERG also compared the overall survival curve for vinflunine used in the manufacturer's economic model with that obtained using Kaplan–Meier estimates. It concluded that the most realistic results were those obtained using the Kaplan–Meier estimates, although it noted that the choice of survival curve did not have a significant impact on the cost effectiveness of vinflunine in the manufacturer's sensitivity analysis.
3.15 The ERG conducted an exploratory analysis using the confidence intervals around the modelled estimates of overall survival and progression-free survival. This resulted in ICERs ranging from £87,871 to £117,938 per QALY gained. In a separate exploratory analysis the ERG used estimates of progression-free survival and overall survival from the ITT population of study 302 (rather than the eligible ITT population) and corrected an error in the manufacturer's model in which the vinflunine vial cost was entered incorrectly. The resulting ICER was £99,792 per QALY gained when the manufacturer's method of estimating survival was used, and £126,422 per QALY gained when Kaplan–Meier estimates based on trial data for the ITT population were used.
3.16 Full details of all the evidence are in the manufacturer's submission and the ERG report.