3.1 The manufacturer's submission presented clinical-effectiveness data derived from one phase III trial (CONFIRM), supported by results from two dose-ranging phase II trials (FINDER-1 and FINDER-2). Women were eligible for these three studies if they were postmenopausal and had oestrogen-receptor-positive breast cancer. Their cancer could have relapsed during or within 12 months of completing adjuvant hormone therapy (with an anti-oestrogen or an aromatase inhibitor) for early breast cancer; or it could have progressed on anti-oestrogen or aromatase inhibitor therapy for advanced breast cancer provided that this hormone therapy was started more than 12 months after completion of adjuvant hormone therapy (anti-oestrogen or aromatase inhibitor); or it could have progressed while they were on first-line hormone therapy (anti-oestrogen or aromatase inhibitor) for advanced breast cancer. All three trials excluded patients who had received two or more lines of previous hormone therapy for locally advanced or metastatic breast cancer.
3.2 The CONFIRM trial was an international multicentre double-blind parallel-group randomised controlled trial (RCT) that included 736 patients who had previously received an anti-oestrogen or an aromatase inhibitor for the adjuvant treatment of early breast cancer or as palliative therapy for advanced breast cancer. Patients were randomised on a 1:1 basis to receive either fulvestrant 500 mg or fulvestrant 250 mg. The mean age of the patients was 61 years. The baseline characteristics of the groups in the two arms of the trial were generally comparable, although more patients in the fulvestrant 250 mg arm (102 compared with 69) had received radiotherapy as treatment for advanced disease.
3.3 The primary outcome measure in the CONFIRM study was median time to progression (TTP). Median TTP was statistically significantly longer in the overall mixed population (that is, including both patients who had previously received an anti-oestrogen and patients who had previously received an aromatase inhibitor) for the fulvestrant 500 mg arm compared with the fulvestrant 250 mg arm (6.5 months compared with 5.5 months; hazard ratio [HR] 0.80; 95% confidence interval [CI] 0.68 to 0.94; p = 0.006). A pre-planned analysis was done for the subgroups of patients last treated with an anti-oestrogen (58%) or an aromatase inhibitor (42%). The median TTPs for the fulvestrant 500 mg and fulvestrant 250 mg arms were 8.6 months and 5.8 months respectively (HR 0.76; 95% CI 0.62 to 0.94; p = 0.013) for the population last treated with an anti-oestrogen, and 5.4 months and 4.1 months respectively for the population last treated with an aromatase inhibitor (HR 0.85; 95% CI 0.67 to 1.08; p = 0.195).
3.4 Secondary outcomes reported in the CONFIRM study included objective response rate, clinical benefit rate and overall survival. The results suggested no statistically significant differences between the fulvestrant 500 mg and 250 mg arms for these outcomes, although the median overall survival was greater in the fulvestrant 500 mg group (25.1 months compared with 22.8 months). Log-rank tests suggested a trend for improved overall survival in the fulvestrant 500 mg group (HR 0.84; 95% CI 0.69 to 1.03; p = 0.091). Overall survival data from the CONFIRM trial were not mature: 51% of patients had died at the time of primary data cut-off for TTP. The manufacturer stated that it plans to re-analyse the overall survival data when 75% of patients have died.
3.5 A total of 2443 adverse events were reported by 483 (66%) of the 735 patients in the safety analysis in the CONFIRM trial. A serious adverse event was reported for 54 patients (7%), including 11 patients (1%) who died. Seventeen patients (2%) discontinued fulvestrant treatment because of an adverse event. There were no notable differences in the incidence of adverse events between treatment groups. The most common adverse events were injection-site pain (11.6%), nausea (9.7%) and bone pain (9.4%).
3.6 The manufacturer also provided health-related quality of life data taken from the CONFIRM study for a total of 145 women who completed the Functional Assessment of Cancer Therapy-Breast (FACT-B) questionnaire at baseline. No significant differences were detected between the fulvestrant 500 mg and 250 mg study arms.
3.7 The FINDER-1 study was a multicentre parallel-group double-blind phase II RCT conducted in Japan. A total of 143 patients recruited from 40 centres were randomised on a 1:1:1 basis to receive fulvestrant 500 mg, fulvestrant 250 mg or fulvestrant 250 mg with a loading dose. The FINDER-2 study was a multicentre international double-blind phase II RCT conducted in seven European countries and Canada. A total of 144 patients were recruited from 34 centres and randomised on a 1:1:1 basis to receive fulvestrant 500 mg, fulvestrant 250 mg or fulvestrant 250 mg with a loading dose. The primary outcome in the FINDER-1 and FINDER-2 trials was objective response rate, with secondary outcomes including clinical benefit rate and TTP. The findings from these trials were broadly in favour of fulvestrant 500 mg compared with fulvestrant 250 mg.
3.8 The manufacturer conducted a network meta-analysis to compare overall survival and TTP for fulvestrant 500 mg with the comparators listed in the scope. Five RCTs that included three of the other comparators (anastrozole, letrozole and fulvestrant 250 mg) listed in the scope were identified in the systematic literature review, resulting in eight trials being included in the network meta-analysis. Data from the total population in the fulvestrant trials were included, with the FINDER-1 and FINDER-2 trials contributing only to the TTP network meta-analysis. The manufacturer stated that inclusion of the group from the CONFIRM trial who had received an aromatase inhibitor as their last treatment did not alter the results in favour of fulvestrant. The manufacturer did not include exemestane as a comparator in the base-case network meta-analysis because of a lack of any relevant trials in which 70% or more patients had documented hormone-receptor-positive advanced breast cancer in a population who had received an anti-oestrogen. Therefore a secondary scenario analysis, as part of the cost-effectiveness analysis comparing fulvestrant 500 mg with exemestane, was carried out by the manufacturer.
3.9 For the base-case network meta-analysis, data on two outcomes were collected: overall survival and TTP. Data from the eight included trials were pooled and extrapolated. Based on patient-level data from the CONFIRM trial, the Weibull distribution was identified as the best-fitting distribution to estimate overall survival. Because hazard ratios in the CONFIRM trial were constant over time (the shape parameters were very similar for both treatment groups), the relative treatment effects of the alternative treatments were applied to the baseline treatment (fulvestrant 250 mg) using a pooled hazard ratio for overall survival estimated from the network meta-analysis. For TTP, the log-normal distribution was identified by the manufacturer as the best-fitting distribution for data from the CONFIRM trial because it was inappropriate to assume that hazard ratios were constant over time. A simultaneous extrapolation and network meta-analysis of TTP curves for all comparator treatments were derived from the available RCTs. This was done by relating the TTP Kaplan-Meier curves of each of the comparators directly to the parameters of the log-normal survival curves. A fixed-effects model was used to simultaneously extrapolate Kaplan-Meier curves over time by means of log-normal curves, to synthesise and to indirectly compare the different treatments. The shape and scale parameters for the baseline treatment (fulvestrant 250 mg) were estimated and used as the anchor to obtain estimates for the shape and scale parameters of the other comparators. Pooled TTP curves for each treatment were produced and the corresponding area under the curve was calculated to obtain the mean TTP estimates for each treatment.
3.10 The results of the network meta-analysis presented by the manufacturer suggested that fulvestrant 500 mg was associated with longer overall survival compared with fulvestrant 250 mg, anastrozole and letrozole, but this finding was not statistically significant. The results of the TTP network meta-analysis suggested that fulvestrant 500 mg was associated with a statistically significantly longer TTP than fulvestrant 250 mg, whereas anastrozole was associated with a statistically significantly shorter TTP than fulvestrant 250 mg. There were no statistically significant differences in TTP between letrozole 2.5 mg and fulvestrant 250 mg.
3.11 The manufacturer developed an Excel-based cost–utility model, based on a time-in-state model structure. The model structure is similar to that of a Markov cohort model, with three possible health states: pre-progression, post-progression and death. However, instead of using transition probabilities to determine movement between health states, the model calculates the proportion of patients in each health state according to the estimated survival functions for TTP and overall survival. All patients are assumed to be in the pre-progression health state at model entry (baseline). The duration of second-line hormonal therapy is assumed to be the same as the amount of time spent in the pre-progression health state. The post-progression health state captures a series of subsequent therapies, including third-line hormonal therapy, up to three sequential lines of chemotherapy, and supportive palliative care. Patients can move to the state of death from either the pre-progression or the post-progression health state, which captures death from any cause. The model uses monthly cycles with a lifetime (13-year) time horizon.
3.12 The results of the base-case network meta-analysis of the clinical effectiveness data on TTP and overall survival were used to populate the economic model. For the base-case analysis, comparator treatments were fulvestrant 250 mg, anastrozole and letrozole. The manufacturer used the overall CONFIRM trial population (that is, a mixed population who had received either an anti-oestrogen or an aromatase inhibitor as their last treatment) in the analysis. The manufacturer reported that it was not feasible to analyse the proportion of patients with grade 3 or grade 4 adverse events because adverse events were not reported consistently across the trials included in the network meta-analysis. However, the manufacturer included serious adverse events in the model because sufficient data were available to conduct a network meta-analysis. The serious adverse event data used in the model included both treatment-related and treatment-independent events, because these were available for all relevant RCTs used to derive the estimates of TTP and overall survival in the base-case analysis.
3.13 Health-related quality of life data based on the FACT-B questionnaire were collected at baseline (pre-progression) from a subgroup of patients in the CONFIRM study. However, the model structure required utility values for the pre-progression and post-progression health states that were not collected in the CONFIRM study. Therefore the manufacturer used published pre-progression and post-progression utility values based on a systematic literature review of utility studies for metastatic or locally advanced breast cancer. The manufacturer considered that the study by Lloyd et al. (2006) provided the most appropriate utility values. In this study, utility values were taken from a relatively small sample of the general public in the UK using the standard gamble technique. The study provided utility values of 0.72 and 0.44 for the pre-progression and post-progression health states respectively. Death was assigned a utility value of zero. Disutilities associated with treatment-related adverse events were not included in the model.
3.14 Resource use and costs in the economic model included those related to each second-line hormonal treatment used during the pre-progression phase, subsequent treatments during the post-progression phase including third-line hormonal therapy, supportive palliative care and chemotherapy, and treatment-related adverse events. No treatment-related monitoring costs associated with fulvestrant 500 mg or its comparators were included in the model. An overall average cost per monthly cycle of £1084 per patient was applied to each treatment arm for the patients in the post-progression health state. For adverse events, the model assumed that each serious adverse event is associated with an average hospital stay of 5 days at a cost of £321.02 per day, which was then weighted by the proportion of serious adverse events estimated in the network meta-analysis for each hormonal treatment considered in the scope. The model assumed that one-third of patients received fulvestrant in primary care and two-thirds in hospital.
3.15 The manufacturer reported the results from the economic model for the two key clinical outcomes, TTP and overall survival. The mean TTP was 15.0 months for fulvestrant 500 mg compared with 10.8 months for fulvestrant 250 mg, 9.5 months for anastrozole and 9.9 months for letrozole. The mean overall survival was 33.4 months for fulvestrant 500 mg compared with 29.0 months for fulvestrant 250 mg, 28.5 months for anastrozole and 24.9 months for letrozole.
3.16 In the base-case incremental analysis, fulvestrant 500 mg was associated with the highest total quality-adjusted life years (QALYs) (1.487 QALYs), followed by fulvestrant 250 mg (1.256 QALYs), anastrozole (1.214 QALYs) and letrozole (1.105 QALYs). Based on an incremental analysis ranking of treatments, the base-case results demonstrated that anastrozole and fulvestrant 250 mg were extendedly dominated by (that is, were more expensive and less effective than) a combination of two other single-agent treatments, fulvestrant 500 mg and letrozole. The comparison of fulvestrant 500 mg with letrozole produced an incremental cost-effectiveness ratio (ICER) of £31,982 per QALY gained (representing incremental costs of £12,239 and incremental QALYs of 0.383). The manufacturer stated that no patients were assumed to be on an adjuvant switch hormone treatment strategy (that is, sequential treatment with an anti-oestrogen and an aromatase inhibitor).
3.17 The manufacturer conducted deterministic sensitivity analyses by varying key model input parameters. These showed that the key drivers of the cost-effectiveness results were the estimates of TTP and overall survival for all treatments and the utility values assigned to the pre-progression and post-progression health states. The widest range of ICERs was found for the comparison of fulvestrant 500 mg with letrozole, in which the ICERs ranged from £21,894 to £55,160 per QALY gained when the upper and lower 95% credibility limits for the scale and log shape of the log-normal distribution of TTP for letrozole were used.
3.18 The manufacturer also conducted six scenario analyses to assess the impact of key assumptions made in the base-case analysis. These scenarios included: expanding the patient population to allow the inclusion of exemestane in the network meta-analysis (by including trials in which at least 50% of patients had documented hormone-receptor-positive cancer and patients who had last been treated with an aromatase inhibitor [because there are no studies comparing fulvestrant with exemestane in patients treated with an anti-oestrogen]); using alternative proportions for the administration of fulvestrant in the primary care setting and in hospital; altering the cost of the post-progression health state by using an alternative mix of chemotherapies; altering the cost of the post-progression health state by eliminating treatment skipping (patients skip further hormonal treatment if the extent and duration of response to a previous hormonal treatment was insufficient); discounting costs and benefits at 0% and 6%; and altering the time horizon. In summary, exemestane, anastrozole and fulvestrant 250 mg were all extendedly dominated by a combination of fulvestrant 500 mg and letrozole. The comparison of fulvestrant 500 mg with letrozole gave a range of ICERs from £29,881 to £38,566 per QALY gained.
3.19 The results of the manufacturer's probabilistic sensitivity analysis showed that, at a threshold of £20,000 per QALY gained, there is a 2% probability of fulvestrant 500 mg being cost effective. This increased to 20% at a threshold of £30,000 per QALY gained.
3.20 The ERG commented that the manufacturer's systematic review of clinical-effectiveness studies was methodologically appropriate and that all relevant studies meeting the inclusion criteria appeared to have been identified.
3.21 The ERG commented that the CONFIRM study was well designed and that the clinical outcomes reported in this RCT and the supporting phase II trials (FINDER-1 and FINDER-2) address all the relevant outcomes outlined in the scope. However, the ERG noted that fulvestrant is currently most commonly used in clinical practice in England and Wales after aromatase inhibitors and often after an anti-oestrogen as well, and therefore it is a third- or fourth-line hormonal therapy in the treatment pathway for advanced breast cancer. In the fulvestrant trials used as the basis for direct clinical evidence and in the manufacturer's submission, fulvestrant was used in the treatment pathway in the position currently occupied by aromatase inhibitors, as second-line treatment. Therefore, the ERG commented that the generalisability of the patient population and trial results to clinical practice may be questionable, because there is a difference between the indication in the marketing authorisation for fulvestrant and its use in treatment in England and Wales.The ERG also noted that no patients were recruited to CONFIRM from the UK.
3.22 The ERG highlighted that the marketing authorisation for fulvestrant 500 mg specifies that the patient has received previous anti-oestrogen therapy, although the ERG noted that it is not clear from the wording of the marketing authorisation that eligibility for treatment depends on the last therapy received. Therefore, the ERG requested that the manufacturer divide the TTP data from CONFIRM in two main ways. First, the patients were divided into two treatment groups: patients who had received an anti-oestrogen as their last treatment (58%) and patients who had received an aromatase inhibitor as their last treatment (42%). Second, the patients were split into three treatment groups: patients who had received an anti-oestrogen but not an aromatase inhibitor; patients who had received an aromatase inhibitor but not an anti-oestrogen; and patients who had received both an anti-oestrogen and an aromatase inhibitor. The second set of data was provided 'in confidence'. The ERG noted from the patients divided into two treatment groups that 65.5% of patients who had received an anti-oestrogen as their last treatment were receiving fulvestrant as a first-line treatment for locally advanced or metastatic breast cancer, whereas 66.8% of patients who had received an aromatase inhibitor as their last treatment received fulvestrant as a second-line therapy for advanced breast cancer. The ERG also demonstrated significant differences between the demography of these two groups: the proportion of patients treated with hormone therapy for advanced disease was 34% in the anti-oestrogen group compared with 67% in the aromatase inhibitor group; and the proportion who had received two previous hormone therapies was 4% in the anti-oestrogen group compared with 27% in the aromatase inhibitor group. The ERG therefore speculated that the apparent increased benefit for fulvestrant after an anti-oestrogen rather than after an aromatase inhibitor may be influenced by where in the treatment sequence most patients received fulvestrant, rather than by whether the last treatment before fulvestrant was an anti-oestrogen or an aromatase inhibitor.
3.23 The ERG considered that the manufacturer's base-case economic evaluation was well conducted and closely matched the NICE reference case. The main issue raised by the ERG related to the use of data from the network meta-analysis, which included patients from the CONFIRM trial who had been treated previously with an aromatase inhibitor. The ERG considered it more appropriate to base the model only on patients who had previously received anti-oestrogen therapy, particularly in view of the heterogeneity of the anti-oestrogen and aromatase inhibitor groups. The ERG considered that the advantage of this approach of reducing the heterogeneity of the compared populations outweighed the main disadvantage of reducing the statistical power of the CONFIRM trial.
3.24 In its critique of the network meta-analysis, the ERG noted that in the comparator treatment trials, none of the patients had received a prior aromatase inhibitor. In addition, the ERG noted key differences in the baseline characteristics of the populations in the trials included in the network meta-analysis. For example, the percentage of patients whose oestrogen receptor status was not known to be positive ranged from to 0% (CONFIRM, FINDER-1 and FINDER-2) to 33.1% (Buzdar 1996/98); the proportion of patients treated previously with chemotherapy ranged from 35.1% (Buzdar 1996/98) to 72.5% (FINDER-1); and the proportion of patients with visceral spread was variable, although the ERG noted that the proportion of patients with known visceral spread was high in the fulvestrant trials.
3.25 Overall, the ERG considered that the population in the CONFIRM trial was heterogeneous and that it was not meaningful to regard the group who had received an anti-oestrogen and the group who had received an aromatase inhibitor as similar. The ERG suggested that the network meta-analyses should include data only from patients who had received an anti-oestrogen as their last treatment from the CONFIRM, FINDER-1 and FINDER-2 trials. Therefore the ERG re-ran the analysis using only data from CONFIRM trial patients whose previous hormone therapy was an anti-oestrogen (n = 423). The results were comparable with those obtained when the whole population of the CONFIRM trial was included in the analysis. All hazard ratios for overall survival still favoured fulvestrant 500 mg over other treatments considered in the scope, although the results were not statistically significant.
3.26 For the TTP network meta-analysis, the ERG questioned the assumption that the CONFIRM trial results follow a log-normal distribution. A direct comparison of the Kaplan-Meier analysis of the trial results with the outputs of the manufacturer's log-normal model appeared to suggest a reasonable match between data (TTP) and model. However, the ERG noted some divergence after 18 months, which would affect the projection of survival curves beyond the observed data. Therefore, the ERG argued that because the log-normal parametric model used by the manufacturer did not adequately represent the data on which it was calibrated, it should not be used to calibrate TTP estimates for all comparators included in the network meta-analysis. The ERG observed that the results of the Kaplan-Meier analysis from the CONFIRM trial showed a higher number of progression events occurring around 90 days, followed by a 90-day period with relatively few new events. From 180 days onward, there was a clear indication of a linear relationship between time and the cumulative TTP hazard. Therefore the ERG proposed that a more accurate approach would be to split the estimation of TTP into two phases and to include only the anti-oestrogen-treated population from the CONFIRM trial. For the first part of the analysis (0–180 days), the ERG performed a network meta-analysis on the log-hazard ratios at 180 days. For the second part of the analysis (after 180 days), TTP was modelled using an exponential distribution, which has a constant hazard or linear cumulative hazard, based on a clear indication of a linear relationship between time and cumulative TTP hazard in the CONFIRM trial. The results of this analysis showed no statistically significant differences in TTP between the groups receiving fulvestrant 500 mg and those receiving other treatments for the first 180 days (a period thought to be driven by protocol activities and short-term events). However, after 180 days (the ERG stated that this period relates to long-term patient experience) fulvestrant 500 mg was associated with statistically significant improvements in TTP compared with anastrozole and letrozole.
3.27 For the overall survival network meta-analysis, the ERG commented that the parametric model used by the manufacturer to estimate overall survival in the network meta-analysis appeared to be a reasonable match with the available CONFIRM trial data. However, the ERG also noted that projections of overall survival beyond the period of observation may be substantially over- or under-estimated because of the complex changes in risk that are likely to apply at later times. Therefore, the ERG suggested that an alternative approach to projective modelling was to consider modelling post-progression patient experience directly on the basis of the trial data, and then to combine pre- and post-progression estimates to obtain the best estimate of overall survival. Examination of post-progression survival data by the ERG showed no statistically significant differences between fulvestrant 500 mg and fulvestrant 250 mg, suggesting that any overall survival gains associated with fulvestrant 500 mg were obtained only in the pre-progression phase (TTP). Therefore the ERG estimated a compatible set of survival estimates (TTP), post-progression survival and overall survival) for fulvestrant 250 mg, anastrozole and letrozole by calibrating a hazard ratio applied to the overall survival estimated for the fulvestrant 500 mg group, which generated a gain in overall survival equal to the corresponding gain in TTP. The ERG noted that although this is an approximation, it allows the timing of post-progression survival to be calculated without elaborate additional modelling. This approach appeared fully justified for anastrozole, because key clinical trials comparing anastrozole with fulvestrant 250 mg showed no statistically significant differences in TTP or overall survival. However, this approach was less clearly supported in the case of letrozole, because there are no trials that directly compare letrozole with fulvestrant.
3.28 The ERG noted several criticisms about the design of the manufacturer's economic model, which was based on separate parametric models of the time from randomisation to TTP and overall survival. The ERG commented that when different probability distributions are used to represent the two sets of data, or when the same function is used for both but does not satisfy proportional hazards criteria (that is, the risk of an event occurring on one treatment relative to another treatment is assumed not to change over time), it is possible for projected estimates of TTP to exceed the corresponding estimates of overall survival. Although the model corrected any negative post-progression survival estimates to zero, it did not compensate for any resulting overestimation of survival. Overall, the ERG concluded that the design of the manufacturer's economic model is unlikely to provide a robust basis for projecting survival beyond the observed data.
3.29 The ERG identified four issues in relation to the cost data used in the manufacturer's model. First, the manufacturer's model does not account for wastage of part-used dispensed packs at the time of disease progression. Second, the ERG questioned the use of the two expert opinions for pre-progression and post-progression health state costs, and instead proposed that such costs should be based on treatment pathways described in 'Advanced breast cancer: diagnosis and treatment' (NICE clinical guideline 81). Third, the manufacturer's model limits drug-related adverse events to serious adverse events only. Fourth, the manufacturer's approach of applying a single average cost for UK hospital admission is simplistic and inappropriate for costing adverse events associated with treatment complications in advanced breast cancer. The ERG calculated an alternative estimate of £3147 per admission, compared with the estimate in the manufacturer's model of £1605 per episode. Overall, the ERG stated that making these four modifications to the model increased the ICER in all cases but, because each change represents only a small element of the total cost, the increases were small.
3.30 Finally, the ERG noted an error in the utility values assigned to the pre-progression and post-progression health states in the manufacturer's economic model. Utility values were based on the age of the participants in the study by Lloyd et al. (2006), taken from a sample of the general UK population, and not on the age of breast cancer patients. The ERG proposed that, to ensure consistency with standard UK EQ-5D tariff scores, the mean age should be set to 47 years (the mean age of the original UK York study sample used). The ERG also accounted for the 'responder status' of patients (that is, whether or not their cancer responded to treatment) when estimating new utility values for both health states. In summary, using ERG estimated utility values of 0.7733 for the pre-progression state and 0.4964 for the post-progression state reduced the ICER for fulvestrant by £2700 per QALY gained compared with letrozole.
3.31 The ERG made eight separate modifications to explore the impact of the various issues described in the critique of the manufacturer's economic model. Seven modifications were made to the economic model logic or parameter values, and the eighth modification involved using effectiveness data from the anti-oestrogen subgroup in the CONFIRM trial instead of data from the whole trial population. The ERG presented detailed deterministic results separately for the manufacturer's base-case scenario using the whole CONFIRM population and for the anti-oestrogen subgroup.
3.32 In summary, based on the full CONFIRM trial population, the calculated deterministic cost-effectiveness results of the ERG's exploratory analyses showed that fulvestrant 250 mg was extendedly dominated by the other comparators,. The ICERs for anastrozole compared with letrozole and for fulvestrant 500 mg compared with anastrozole were both close to £30,000 per QALY gained. The ERG's preferred exploratory deterministic cost-effectiveness analysis based on the anti-oestrogen subgroup from CONFIRM and an updated network meta-analysis resulted in fulvestrant 250 mg being extendedly dominated by the other comparators. The ICER for anastrozole compared with letrozole was £1162 per QALY gained, and the ICER for fulvestrant 500 mg compared with anastrozole was £34,972 per QALY gained.
3.33 Full details of all the evidence are in the manufacturer's submission and the ERG report, which are available from www.nice.org.uk/guidance/TA239