3 The manufacturer's submission
3.1 The manufacturer's decision problem specified febuxostat (80 mg or 120 mg once daily) as the intervention of interest in a population of adults with hyperuricaemia in gout. The comparator specified by the manufacturer was fixed-dose allopurinol (300 mg once daily). The manufacturer did not present comparisons with alternative comparators such as sulphinpyrazone, benzbromarone, probenecid or combinations of these treatments. The health outcomes considered included serum uric acid concentration, gout flare rates, reduction in size of tophi, and health-related quality of life. The results of the economic evaluation were expressed as incremental cost per quality-adjusted life year (QALY) over a time horizon of 2 years.
3.2 The manufacturer presented evidence on the clinical effectiveness of febuxostat from three randomised controlled trials (RCTs): the FACT trial, the APEX trial and the TMX-00-004 study. The FACT trial was a 52-week, phase III, multi-arm, randomised, double-blind, parallel-group trial that compared febuxostat 80 mg/day (n = 257), febuxostat 120 mg/day (n = 251) and allopurinol 300 mg/day (n = 254). The APEX trial was a 28-week, phase III, multi-arm, randomised, double-blind trial that compared febuxostat 80 mg/day (n = 267), febuxostat 120 mg/day (n = 269), febuxostat 240 mg/day (n = 134), allopurinol 300 or 100 mg/day (n = 268) and placebo (n = 134). The reduced dose of allopurinol (100 mg/day) in the APEX trial was used for 10 patients with renal impairment. In the APEX and FACT trials, colchicine and naproxen (a non-steroidal anti-inflammatory drug [NSAID]) were given as prophylaxis for treatment-initiated gout flares during the first 8 weeks of treatment. The TMX-00-004 study was a 4-week, phase II (dose–response), multicentre, randomised, double-blind, parallel-group trial that compared febuxostat 40 mg/day (n = 37), febuxostat 80 mg/day (n = 40), febuxostat 120 mg/day (n = 38) and placebo (n = 38).
3.3 Supplementary data in support of the clinical effectiveness of febuxostat were provided by the manufacturer from two open-label extension studies. The EXCEL trial is an ongoing open-label extension study that enrolled a subset of patients (n = 735) from the APEX and FACT trials. The EXCEL study compared febuxostat 80 mg/day (n = 299), febuxostat 120 mg/day (n = 291) and allopurinol 300 or 100 mg/day (n = 145). The FOCUS trial was a 5-year, open-label, non-controlled extension study that enrolled patients (n = 116) who completed the TMX-00-004 study. The FOCUS study evaluated febuxostat 80 mg/day, with dose titration to lower (40 mg/day) or higher (120 mg/day) doses permitted between weeks 4 and 28.
3.4 Although summaries of the clinical evidence from the respective trials were provided by the manufacturer, the main evidence presented in support of the clinical effectiveness of febuxostat was based on pooled data from the APEX and FACT trials. This evidence was not in the form of a meta-analysis of the separate studies. It involved adding together the number of events observed for each treatment group across the trials and dividing the results by the total number of patients in the combined treatment group.
3.5 The manufacturer's pooled analysis suggested that febuxostat 80 mg/day and 120 mg/day was significantly more effective (p ≤ 0.05) than fixed-dose allopurinol (300 or 100 mg/day) at lowering the serum uric acid concentration to target therapeutic levels (of below 6 mg/100 ml) at the last three visits and at the final visit. Febuxostat at either dose was also significantly more effective (p ≤ 0.05) than fixed-dose allopurinol (300 mg/day) at lowering the serum uric acid concentration from baseline levels at the final visit. No statistically significant differences were observed with febuxostat 80 mg/day compared with allopurinol (300 or 100 mg/day) in the proportion of patients requiring treatment for gout flares. In contrast, the proportion of patients requiring treatment for gout flares was statistically significantly higher (p ≤ 0.05) with febuxostat 120 mg/day than with allopurinol (300 or 100 mg/day), both during (weeks 1–8) and after (weeks 9–52) prophylaxis. The difference was more marked during the initial weeks of treatment. No statistically significant differences were found between groups in the percentage reduction in tophus area except at week 28, when statistically significantly greater reductions in primary tophus size from baseline were observed with febuxostat 120 mg/day than with allopurinol.
3.6 Post-hoc subgroup analysis of the pooled data showed that febuxostat was more effective (p ≤ 0.05) than allopurinol in lowering serum uric acid concentration to below 6 mg/100 ml in three subgroups of patients defined according to baseline serum acid concentrations of below 9 mg/100 ml, between 9 and 10 mg/100 ml and above 10 mg/100 ml. The proportion of patients whose serum uric acid concentration fell below the target level of 5 mg/100 ml was higher (p ≤ 0.05) among those receiving febuxostat than among those receiving fixed-dose allopurinol. No subgroup analyses were conducted for patients with renal impairment or those whose condition did not respond to allopurinol.
3.7 Results from the EXCEL extension study showed that a higher proportion of patients receiving febuxostat (80 mg/day or 120 mg/day) remained on initial treatment than among those receiving fixed-dose allopurinol (300 or 100 mg/day) after more than 24 months of follow-up. For each year of febuxostat treatment in the EXCEL trial, the number of gout flares decreased over time. However, the ERG considered that this evidence should be treated with caution, since statistical comparisons between treatment groups were not reported. In addition, no data were provided on withdrawals because of gout flares, adverse events or lack of response to treatment.
3.8 The ERG considered that the evidence presented in support of the clinical effectiveness of febuxostat in comparison with allopurinol may not be adequate. This is because guidelines for gout management from the British Society of Rheumatology and British Health Professionals in Rheumatology (BSR guidelines) and the European League Against Rheumatism (EULAR guidelines), and the SPC for allopurinol, recommend dose titration for allopurinol according to therapeutic targets. It is possible that dose-titrated allopurinol may be more effective than fixed-dose allopurinol, and that the additional clinical benefits of febuxostat may not be as great in routine practice as is suggested by the results from the RCT comparisons with fixed-dose allopurinol. However, the ERG noted that dose titration of allopurinol is rarely carried out in routine clinical practice.
3.9 The ERG expressed concerns about the analysis of clinical efficacy based on pooling data across trials, because this approach fails to preserve randomisation in the RCT evidence, which may introduce bias. The ERG carried out a corrected meta-analysis (based on both fixed- and random-effects modelling) using the RCT data and evidence presented by the manufacturer. The results of the meta-analysis showed that the probability of reaching therapeutic targets for serum uric acid concentration was statistically significantly higher for patients receiving febuxostat (80 or 120 mg/day) than for those receiving fixed-dose allopurinol (300 mg/day). The ERG's meta-analysis also showed that a higher proportion of patients receiving febuxostat needed treatment for gout flares (both during prophylaxis [weeks 1–8] and after prophylaxis [weeks 9–52]) compared with those receiving fixed-dose allopurinol. For the febuxostat 80 mg/day group this difference was not statistically significant (p > 0.18), but for the 120 mg/day febuxostat group it was (p < 0.05).
3.10 The manufacturer's submission presented an analysis of the cost effectiveness of febuxostat in comparison with fixed-dose allopurinol. A decision-tree model was provided to estimate the cost and health outcomes for patients with gout after initiation of urate-lowering therapy with febuxostat 80 mg or 120 mg daily, or allopurinol 300 mg daily. The model had a time horizon of up to 2 years. The time horizon was based on 1-year trial data and a further 1 year of extrapolation, and was extended to 5 years in sensitivity analyses. A mixed cohort of men and women with gout and a baseline serum uric acid concentration of 80 mg/100 ml or above entered the model after initiation of urate-lowering therapy. The model was split into two time periods: an initial period of 3 months, during which patients may or may not experience a treatment-initiated flare, and a treatment maintenance period from 4 to 24 months.
3.11 The results of the economic analysis were presented as incremental costs per QALY gained for febuxostat in comparison with fixed-dose allopurinol. The base-case economic analysis using pooled clinical data over a 1-year time horizon comparing febuxostat (80 mg/day and 120 mg/day) with fixed-dose allopurinol (300 mg/day) produced an incremental cost-effectiveness ratio (ICER) of £16,574 per QALY gained. An alternative base-case analysis based on a 2-year time horizon produced an ICER of £15,565 per QALY gained. The manufacturer presented the results of a probabilistic sensitivity analysis that gave a mean ICER of £16,324 per QALY gained (95% confidence interval [CI] £6281 to £239,928 per QALY). The cost-effectiveness acceptability curve reported that the probability that febuxostat 80 mg/day (titrated to 120 mg/day where appropriate) had an ICER lower than £20,000 per QALY gained compared with fixed-dose allopurinol was 63%.
3.12 The manufacturer presented a number of univariate sensitivity analyses to evaluate the impact of changing the following variables: the time horizon of the model; the protective effect of colchicine prophylaxis; discount rates; the assumed cost of febuxostat; the disutility associated with increments in serum uric acid concentration; and the proportion of patients with a serum uric acid concentration below 360 µmol/litre (6 mg/100 ml) between 4 and 24 months of treatment with febuxostat. The results of these sensitivity analyses showed that the key drivers of the economic model were: (1) the assumed cost of febuxostat; (2) the disutility associated with each increment in serum uric acid concentration; and (3) the proportion of patients with a serum uric acid concentration below 360 µmol/litre (6 mg/100 ml) between 4 and 24 months of treatment with febuxostat. The base-case ICER remained stable over time when extrapolating to a 5-year time horizon. In the base-case analysis, gout flare rates were reduced by 78% in the first 3 months by assuming that patients received 3 months of prophylaxis with colchicine. Setting the reduction in gout flares because of prophylaxis to 0% increased the ICER to £18,826. Discounting had only a marginal effect on the ICER. The manufacturer presented an additional economic analysis after consultation on the appraisal consultation document. This compared febuxostat with a no-treatment (placebo) option over a 2-year time horizon using clinical data from the placebo arm in the APEX trial. The results of the economic analysis gave an ICER of £3727 per QALY gained for a comparison of febuxostat against a no-treatment option.
3.13 The ERG noted a number of areas of uncertainty around the cost-effectiveness analyses undertaken in the manufacturer's submission. The ERG noted that the natural history of people with hyperuricaemia in gout who did not receive treatment was not modelled, and hence no inference could be made of the cost effectiveness of febuxostat compared with no treatment. The ERG requested that a sequence of strategies where patients progress to an alternative intervention (allopurinol, febuxostat or no treatment) following lack of response should be evaluated. The manufacturer declined the request, arguing that estimation of a sequential strategy was not feasible because of a lack of clinical data. In addition, the manufacturer argued that it was unethical to consider febuxostat as second-line therapy when it is cost effective as first-line therapy, and that the only appropriate comparison was that investigated in the pivotal RCTs; that is, first-line therapy. The ERG asserted that appropriate modelling assumptions could have been made to allow some exploratory analysis.
3.14 The ERG noted that no data were provided on the likely serum uric acid concentrations of patients who receive no treatment, and that the model assumes full treatment adherence and usage over the time horizon modelled. However, evidence from the FACT and APEX trials showed that treatment continuation rates were lower for febuxostat than for allopurinol. Contradictory evidence from the open-label EXCEL trial reports gave treatment continuation rates of 35% for 80 mg/day febuxostat and 10% for 120 mg/day febuxostat, but only 5% for allopurinol. Additional evidence presented by the manufacturer suggested that in a randomised subset of the EXCEL trial, 76% of patients receiving 80 mg/day febuxostat, 71% of patients receiving 120 mg/day febuxostat and 40% of patients receiving allopurinol remained on initial treatment after more than 24 months. The ERG stated that the lack of data on the likely serum uric acid concentrations of patients who receive no treatment made it difficult to accurately account for treatment discontinuation rates in the manufacturer's economic model. An assumption of full treatment adherence and usage over the time horizon could potentially bias the results of the economic analysis in favour of febuxostat.
3.15 The ERG noted that in the manufacturer's economic model, data on the number of gout flares within the initial 3 months of treatment were taken from the pooled analysis of the results from the APEX and FACT trials. The flare rates were reduced by 78% by assuming that patients received 3 months of colchicine prophylaxis. However, this reduction may be an overestimate, since in the APEX and FACT trials colchicine prophylaxis was given for only 8 weeks. After the first 3 months of treatment, the proportion of gout flares was assumed to be related to serum uric acid concentration, on the basis of a multivariate analysis of data provided in confidence by the manufacturer. The significance of the correlation between serum uric acid concentration and the number of gout flares disappears in the manufacturer's 'multivariate analysis' when other significant covariates are included. However, the manufacturer's 'multivariate analysis' appears to be the same as a bivariate analysis of the relationship between serum uric acid concentration and the odds of gout flares occurring (the p-values for the two analyses were identical). The ERG argued that a proper multivariate analysis, in which a backward stepwise analysis is carried out, should be presented. The ERG further expressed concerns about discarded data points from the datasets used in the analysis showing a relationship between serum uric acid concentration and number of gout flares. The ERG stated that a relationship, not necessarily linear, between serum uric acid concentration and number of gout flares may still be found with more appropriate analysis and larger or different datasets.
3.16 The ERG noted that the relationship between serum uric acid concentration and the expected number of gout flares (with a 'chronic utility gain' associated with a lower serum uric acid concentration and decreased utility associated with gout flares) is a key driver of the economic results presented by the manufacturer. Therefore uncertainty about this relationship translates into uncertainty about the ICER estimates presented by the manufacturer. The ERG considered that this uncertainty had not been adequately investigated. In an exploratory analysis provided by the ERG, removing the 'chronic utility gain' associated with lower serum uric acid concentration increased the base-case ICER to £81,000 per QALY gained over a 2-year time horizon. A similar analysis gave an ICER of £696,000 per QALY gained over a 1-year time horizon, and an ICER of £150,000 per QALY gained over a 5-year time horizon.
3.17 Full details of all the evidence are in the manufacturer's submission and the ERG report.