3 The manufacturer's submission
3.1 The decision problem in the manufacturer's submission compared ustekinumab with adalimumab, efalizumab, etanercept, infliximab and supportive care. Three doses of etanercept were considered: 25 mg twice weekly given intermittently as recommended in NICE's technology appraisal guidance on etanercept and efalizumab for the treatment of adults with psoriasis, 25 mg twice weekly given continuously, and 50 mg twice weekly given for the first 12 weeks followed by a reduction in dose to 25 mg twice weekly. Clinical outcomes in the manufacturer's submission included improvements in PASI and DLQI scores. PASI is a measure of disease severity based on body surface area affected and the extent, scaliness, thickness and redness of plaques, with scores ranging from 0 to 72. The DLQI is a disease-specific quality-of-life measure with scores ranging from 0 to 30. Moderate to severe psoriasis was defined as a PASI score of 10 or more and a DLQI score of more than 10.
3.2 The manufacturer's submission included evidence from three randomised controlled trials (RCTs):
PHOENIX-1 (n = 766, 5 years' duration), a phase III, multicentre, parallel, randomised, double-blind, placebo-controlled trial based in the USA, Canada and Belgium.
PHOENIX-2 (n = 1230, 5 years' duration), a phase III, multicentre, parallel, randomised, double-blind, placebo-controlled trial based in Europe and North America.
ACCEPT trial (n = 903, 64 weeks' duration), a phase III, multicentre, parallel RCT based in Europe and North America, which compared ustekinumab with etanercept (50 mg twice weekly for the first 12 weeks).
3.3 In each of the RCTs, two doses (45 mg and 90 mg) of ustekinumab were investigated and patients were randomised to groups regardless of their body weight. To reflect the licensed dosing of ustekinumab, the manufacturer presented two analyses in their submission. The first analysis used data from all the patients enrolled in the clinical trials and included dosing outside the marketing authorisation (that is, patients weighing 100 kg or less who received ustekinumab 90 mg and patients weighing over 100 kg who received ustekinumab 45 mg). The second was a subgroup analysis that included data only for patients who received ustekinumab according to the marketing authorisation (weight-based dosing; that is, 45 mg for people weighing 100 kg or less and 90 mg for people weighing over 100 kg).
3.4 The results of the three RCTs using data for all patients demonstrated statistically significant differences in the percentage of patients treated with ustekinumab who achieved a 75% or greater reduction in PASI score (PASI 75; the primary endpoint in the trials) compared with those who received placebo. The percentages of patients with at least a PASI 75 response at week 12 in the ustekinumab 45 mg, ustekinumab 90 mg and placebo groups were 67%, 66% and 3% respectively in the PHOENIX-1 trial (p < 0.001 for both ustekinumab doses compared with placebo) and 67%, 76% and 4% respectively in the PHOENIX-2 trial (p < 0.001). In the ACCEPT trial, the percentages of patients with at least a PASI 75 response at week 12 in the ustekinumab 45 mg, ustekinumab 90 mg and etanercept groups were 68%, 74% and 57% respectively (p = 0.012 for ustekinumab 45 mg and p < 0.001 for ustekinumab 90 mg compared with etanercept).
3.5 For secondary outcomes recorded in the RCTs, such as the physician's global assessment (PGA) score, the DLQI score and other health-related quality-of-life scores, the ustekinumab groups showed statistically significant improvements compared with the placebo groups. In the PHOENIX-1 trial, the mean change in DLQI score at week 12 was −8.0 for ustekinumab 45 mg, −8.7 for ustekinumab 90 mg and −0.6 for placebo (p < 0.001 versus placebo for both ustekinumab doses). In the PHOENIX-2 trial, the values were −9.3, −10.0 and −0.5 respectively (p < 0.001 versus placebo for both ustekinumab doses). DLQI data were not collected in the ACCEPT trial.
3.6 Data from the clinical trials suggested that 90 mg is a more effective dose of ustekinumab than 45 mg for patients who weigh more than 100 kg. For example, in the PHOENIX-1 trial, 69% of patients weighing more than 100 kg who received ustekinumab 90 mg achieved a PASI 75 response at 12 weeks, compared with 54% of those who received ustekinumab 45 mg. In the PHOENIX-2 trial, the values were 71% and 49% respectively.
3.7 The manufacturer included longer-term data from the PHOENIX trials for the weight-based dosing subgroup analysis. These data suggested that the PASI response rates observed during the double-blind, randomised phases of the studies were maintained in the longer term. In the PHOENIX-1 trial, the percentages of patients achieving a PASI 75 response at week 24 were 83% and 80% for ustekinumab 45 mg and 90 mg respectively. In the PHOENIX-2 trial, the respective percentages were each 80%.
3.8 In the PHOENIX-1 trial, the percentages of patients having one or more adverse events were 57.3%, 51.4% and 47.8% in the ustekinumab 45 mg, ustekinumab 90 mg and placebo groups respectively. The percentages of patients having a serious adverse event were 0.8%, 1.6% and 0.8% respectively. Similar rates of adverse events were reported in the PHOENIX-2 trial. In the ACCEPT trial, the percentages of patients having one or more adverse events were 66.0%, 68.3% and 69.5% in the ustekinumab 45 mg, ustekinumab 90 mg and etanercept groups respectively. The percentages of patients having a serious adverse event were 1.9%, 1.2% and 1.2% respectively.
3.9 The manufacturer compared ustekinumab with other biological therapies (that is, adalimumab, efalizumab, infliximab and etanercept) using a mixed treatment comparison. This included data from studies that compared different biological therapies directly, as well as indirect comparisons using data from studies that compared biological therapies with placebo using the placebo group as the common factor. The manufacturer included data from the three ustekinumab RCTs, as well as from three RCTs comparing adalimumab with placebo, five comparing efalizumab with placebo, five comparing etanercept with placebo and four comparing infliximab with placebo. The results from the mixed treatment comparison using the ustekinumab data for all patients suggested that the mean probabilities of achieving a PASI 75 response were 69% for ustekinumab 45 mg (95% confidence interval [CI] 62% to 75%), 74% for ustekinumab 90 mg (95% CI 68% to 80%), 58% for adalimumab (95% CI 49% to 68%), 80% for infliximab (95% CI 70% to 87%), 39% for etanercept 25 mg (95% CI 30% to 48%), 52% for etanercept 50 mg (95% CI 45% to 59%), 26% for efalizumab (95% CI 21% to 32%) and 4% for supportive care (95% CI 3% to 4%). The manufacturer also included a mixed treatment comparison for the weight-based dosing subgroup analysis. However, the ustekinumab data from this comparison were provided as academic in confidence.
3.10 The manufacturer based its cost-effectiveness analysis on the economic model used in TA103 and subsequently in NICE's guidance on infliximab for the treatment of adults with psoriasis and adalimumab for the treatment of adults with psoriasis. The model was adapted by the manufacturer of ustekinumab to incorporate additional evidence, including the results of the mixed treatment comparison described in section 3.9.
3.11 In the model, each person had an initial period of treatment after which response was assessed (this was referred to as the trial period). Continuation of treatment into the next phase (referred to as the treatment period) occurred only if a PASI 75 response was achieved in the trial period. The time at which the response was assessed varied for the different drugs, depending on their dosing regimen. The assessment points were at 12 weeks (etanercept), 10 weeks (infliximab) and 16 weeks (adalimumab and ustekinumab). It was assumed that for people whose psoriasis responded to treatment, 20% stopped treatment each subsequent year. The mean time on treatment using this assumption was calculated to be 3.65 years. The same assumption was used for all biological therapies.
3.12 The utility data used in the model were based on an estimate of the relationship between PASI response rates and changes in DLQI score from the PHOENIX-1 and PHOENIX-2 trials mapped to EQ-5D scores. First, the mean change in the DLQI score between baseline and week 12 was estimated for groups of patients with different levels of PASI response. Secondly, the manufacturer estimated an algorithm to map DLQI scores to EQ-5D scores from a scatter plot published in the assessment report of TA103. The changes in mean EQ-5D score for PASI responses of less than 50%, between 50% and 74%, between 75% and 89%, and 90% or more were estimated to be 0.04, 0.17, 0.22 and 0.25 respectively.
3.13 The costs in the economic model included drug costs, administration costs and monitoring costs, and were taken from the model in TA103, NHS Reference Costs and the BNF (edition 56). The Personal Social Services Research Unit (PSSRU) inflation index was used to update costs from 2006 values if current costs were not available. The model assumed that people whose psoriasis had not responded adequately to treatment would have an inpatient admission of 21 days' duration once a year.
3.14 The manufacturer's base-case analysis assumed a weighted average of weight-based dosing whereby 80% of people received ustekinumab 45 mg and 20% of people received ustekinumab 90 mg. The manufacturer also provided analyses using the data from all patients in the clinical trials and the data from the weight-based dosing approach with separate estimates for ustekinumab 45 mg and 90 mg. All the analyses in the submission assumed that the patient access scheme (see section 2.4) was in place. Under the original patient access scheme the company provided 2x45 mg pre-filled syringes, for patients who needed the higher dose of 90 mg, at the same total cost to the NHS as for a single 45‑mg pre-filled syringe. The patient access scheme was withdrawn in January 2017 because the company now provides a 90‑mg vial at the same cost as the 45‑mg vial.
3.15 The base-case analysis showed that when ustekinumab was compared with supportive care, the QALY gain was 0.156 at an incremental cost of £4615, giving an incremental cost-effectiveness ratio (ICER) of £29,587 per QALY gained. The ICER for ustekinumab compared with etanercept 25 mg given intermittently (assuming 88% of the cost of continuous etanercept) was £27,105 per QALY gained. The ICER for infliximab compared with ustekinumab was £304,566 per QALY gained. Adalimumab and etanercept given continuously rather than intermittently were dominated by ustekinumab (that is, ustekinumab had both greater effectiveness and lower costs). Probabilistic sensitivity analyses suggested that the probabilities of ustekinumab being cost effective at £20,000 and £30,000 per QALY gained were 7.4% and 48.5% respectively. The manufacturer's analyses suggested that ustekinumab was the only biological therapy that was likely to be cost effective at £20,000 and £30,000 per QALY gained.
3.16 The analyses using data for all patients (that is, no weight-based dosing) presented separate ICERs for ustekinumab 45 mg and 90 mg. These analyses suggested that when ustekinumab 45 mg was compared with supportive care, the QALY gain was 0.1544 at an incremental cost of £4735, giving an ICER of £30,664 per QALY gained. The estimates for ustekinumab 90 mg suggested a QALY gain of 0.1563 and incremental costs of £4613, giving an ICER of £29,520 per QALY gained. The ICERs for ustekinumab in comparison with intermittent etanercept 25 mg were £36,938 per QALY gained for ustekinumab 45 mg and £28,633 per QALY gained for ustekinumab 90 mg. Etanercept 25 mg given continuously was dominated by ustekinumab. Adalimumab was dominated by ustekinumab 90 mg, but for ustekinumab 45 mg the ICER was £16,400 per QALY gained.
3.17 Sensitivity analyses were carried out to test assumptions in the economic model. When the manufacturer reduced the length of an inpatient stay for people whose psoriasis did not respond adequately to treatment from the base-case estimate of 21 days to 17.5 days, the ICER for ustekinumab in comparison with supportive care increased from £29,587 to £34,387 per QALY gained. When the length of stay was increased to 27.5 days, the ICER decreased to £20,672 per QALY gained. The manufacturer also changed the way in which estimates of utility were obtained: from EQ-5D data mapped from DLQI scores, to SF-6D data transformed from SF-36 values collected in the PHOENIX-1 trial. When SF-6D data were used to estimate utilities, the ICER for ustekinumab compared with supportive care increased from £29,302 to £49,371 per QALY gained.
3.18 The manufacturer also varied the assumptions about the cost and efficacy of intermittent etanercept. The cost of intermittent compared with continuous etanercept was changed from the base-case estimate of 88% to 74% (the figure used in TA103) and to 98%. Using an estimate of 74%, the ICER for ustekinumab compared with intermittent etanercept 25 mg increased from £27,105 to £68,339 per QALY gained. When an estimate of 98% was used, ustekinumab dominated intermittent etanercept. The relative efficacy of intermittent compared with continuous etanercept was assumed to be 81% in the base case. When this estimate was changed to 71%, the ICER for ustekinumab compared with intermittent etanercept decreased to £22,634 per QALY gained. When the estimate was changed to 91%, the ICER was £32,949 per QALY gained.
3.19 The ERG concluded that the manufacturer's submission provided an unbiased estimate of the clinical effectiveness of ustekinumab at 12 weeks based on the results of the three randomised comparisons. However, it noted that there was a lack of information about the methodology used for the weight-based dosing subgroup analysis. In addition, it could not determine whether the methods used were appropriate and whether the subgroup analysis supported the weight-based categorisation presented.
3.20 The ERG commented that there appeared to be differences between the mixed treatment comparison that had been used in the appraisal of etanercept and efalizumab (TA103) and that used in the current appraisal. The ERG also noted that the manufacturer's submission included only minimal discussion of any possible clinical heterogeneity between the trials included in the mixed treatment comparison. It further noted that in the mixed treatment comparison, data from the weight-based dosing analysis of ustekinumab were taken from a subgroup of the trial data, whereas data for all patients were used for the comparator trials. The ERG was concerned that this had affected randomisation. The ERG concluded that the clinical effectiveness of ustekinumab in comparison with the other biological therapies was uncertain.
3.21 The ERG also noted that the probabilistic sensitivity analysis in the manufacturer's submission appeared to include only variables for utilities, treatment response and the proportion of people weighing more than 100 kg. It did not include other variables to which the ICERs were sensitive, such as the number of hospital days, the effects of different inpatient costs and the relative efficacy of intermittent etanercept.
3.22 The ERG completed an exploratory analysis that amended the base-case analysis to include the price for ustekinumab 90 mg as double the list price of ustekinumab 45 mg (that is, assuming that there would be no patient access scheme in place). The results showed that the ICER for ustekinumab compared with supportive care increased from £29,587 to £40,952 per QALY gained. A further exploratory analysis assumed that the efficacy of intermittent etanercept 25 mg was the same as that of continuous etanercept 25 mg (as was assumed in the economic model for TA103). Using this assumption, the ICER for ustekinumab compared with intermittent etanercept 25 mg in the base-case analysis increased from £27,105 to £41,449 per QALY gained.
3.23 The ERG conducted an exploratory probabilistic sensitivity analysis that included a larger number of variables than were included by the manufacturer. The results of the ERG's analysis suggested greater uncertainty around the estimates of cost effectiveness, but the cost-effectiveness acceptability curves did not differ significantly from those of the manufacturer. When the ERG repeated the analysis assuming that the cost of ustekinumab 90 mg was twice that of ustekinumab 45 mg, the results showed that the probability of ustekinumab being considered cost effective at £20,000 and £30,000 per QALY gained was zero.
3.24 Full details of all the evidence are in the manufacturer's submission and the ERG report.