3 The manufacturer's submission
3.1 The manufacturer based its evidence submission on the assessment report and model from 'Etanercept and efalizumab for the treatment of adults with psoriasis', NICE technology appraisal guidance 103 (TA103). In this document these are referred to as the York report and the York model, respectively. The manufacturer stated that the population of interest should be people with a PASI score of 10 or more and a DLQI score of more than 10, in line with the recommendations in TA103. The PASI is a measure of severity of disease in terms of 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. The manufacturer compared infliximab with etanercept, efalizumab and supportive care. In TA103, etanercept is recommended for intermittent use, in which treatment stops when remission is achieved. However, the manufacturer argued that in current clinical practice etanercept is used continuously, in which treatment is continued to maintain response, and therefore continuous etanercept was a more appropriate comparator than intermittent etanercept.
3.2 The manufacturer identified four randomised controlled trials (RCTs) that compared infliximab with placebo: Chaudhari et al. (n = 33, 10-week duration), a phase III, randomised, double-blind, placebo-controlled trial based in the USA; SPIRIT (n = 249, 10-week duration, 30-week follow‑up), a phase II induction safety and efficacy study based in the USA; and EXPRESS (n = 378, 10-week duration, 50-week follow-up) and EXPRESS II (n = 835, 10-week duration, 36-week follow-up), which were both phase III multicentre, multinational, randomised, double-blind, placebo-controlled, parallel trials. The participants in the SPIRIT, EXPRESS and EXPRESS II trials had a PASI of at least 12. No trials were identified that compared infliximab with etanercept or efalizumab.
3.3 The results of all four RCTs showed statistically significant improvements in the percentage of people with a PASI 75 (a 75% improvement in the PASI score) after 10 weeks of infliximab compared with placebo (relative risk [RR] 4.5, 14.9, 31 and 39.2, respectively). In the SPIRIT and EXPRESS and EXPRESS II studies, statistically significant improvements were also observed in the percentage of people with PASI 50 and PASI 90. In the EXPRESS and EXPRESS II studies, after week 24 of follow-up the differences were no longer statistically significant, but the manufacturer attributed this to crossover.
3.4 In the absence of any direct trials comparing infliximab with etanercept or efalizumab, the manufacturer carried out an indirect comparison using a meta-analysis and Bayesian hierarchical model. The manufacturer used data from four infliximab trials, four efalizumab trials and three etanercept trials. The random-effects model used for combining 10-week data for infliximab resulted in an RR of 20.49 (95% confidence interval [CI] 16.28 to 25.37). The pooled RR calculated from the four trials of efalizumab was 7.41 (95% CI 5.96 to 9.09) and for 25-mg intermittent etanercept, the RR calculated using pooled data from the three trials was 9.06 (95% CI 7.03 to 11.53).
3.5 The manufacturer based its cost-effectiveness analysis on the York model. This was a two-state Markov model (the two states were on-treatment and off-treatment); alterations were made to include the new data from the infliximab studies. The rates of transitions between states in the model were informed by response and withdrawal rates from the RCTs. The economic analysis included comparisons with etanercept 25 mg, both intermittent and continuous, efalizumab and supportive care. There were no trials identified for continuous etanercept so the manufacturer used the RR for intermittent etanercept in subsequent analyses. The model had a 10-year time horizon and included a trial period after which treatment could be switched to efalizumab or supportive care if the patient's condition had not responded to initial therapy (defined as achieving PASI 75). For infliximab this trial period was 10 weeks (on the basis of RCT evidence), whereas for etanercept and efalizumab it was 12 weeks (corresponding to TA103). The cost and resource use data were obtained from the York report (inflated to present values), NHS reference costs and BNF 53, and were also supported by data that the manufacturer had on file and by clinical opinion.
3.6 The utility data were obtained from the York report. These utilities were based on values from etanercept trials that linked the DLQI with the PASI. A linear transformation was then used to calculate Euro Quality of Life Questionnaire (EQ‑5D) scores. In the York report two populations were defined: the all-patient group and a 4th-quartile group, which corresponded to a subgroup with more severe psoriasis, defined as the 25% of people with the highest scores on the DLQI. In both groups the participants had a PASI greater than 10. The manufacturer used the utilities from the 4th-quartile DLQI group for its base-case analysis to represent those with the worst quality of life at baseline.
3.7 The manufacturer's base-case analysis (using 4th-quartile DLQI utilities) against continuous etanercept resulted in a cost of £26,095 per quality-adjusted life year (QALY) gained. The incremental cost-effectiveness ratio (ICER) for infliximab compared with supportive care was £22,240 per QALY gained. The manufacturer carried out one-way sensitivity analyses. These demonstrated that changes in response rates and patients' weight (the dose of infliximab is dependent on a patient's weight, see section 2.3) had the greatest impact on the ICER. The probabilistic sensitivity analysis gives probabilities of being cost effective at £20,000 and £30,000 thresholds of 10% and 73%, respectively.
3.8 The manufacturer presented, in an appendix, an ICER for infliximab compared with supportive care, using the all-patient utilities, of £41,351 per QALY gained. The probabilistic sensitivity analysis gives a probability of being cost effective at the £30,000 threshold of 0%.
3.9 The ERG had three main areas of concern over the modelling.
The ERG expressed concern regarding the reasoning behind the exclusive use of the 4th-quartile DLQI utility values. This does not correspond to the total population in the decision problem (that is those with a PASI score of at least 10 and a DLQI score greater than 10) or the data used for the indirect clinical effectiveness estimation. The ERG was unclear what severity of psoriasis this analysis would apply to.
The assumed annual drop-out rate in the model was considered by the ERG to be an underestimate because it was based on 6-month rather than annual data. The ERG postulated that the drop-out rate might be as high as 50%. This would result in the ICER against continuous etanercept increasing to approximately £37,000 per QALY gained.
The ERG considered that the cost of an inpatient stay might have been overestimated because it was based on an elective inpatient code rather than elective and non-elective codes with excess bed days incorporated. The cost of an inpatient day would be reduced from £6189 to £5091 using elective codes and to £5488 using a combination of codes for elective and non-elective admissions. Using a cost of £5091 would increase the ICER against continuous etanercept to approximately £30,000 per QALY gained.
3.10 The ERG produced a cumulative scenario analysis in which all of the changes arising from the assumptions described in section 3.9 were combined. This increased the ICER of infliximab compared with continuous etanercept from £26,095 to approximately £41,000 per QALY gained when the alternative drop-out rate and inpatient costs assumptions were combined. When the all-patient utility was included, the ICER increased to approximately £77,000.
3.11 The ERG also extended the manufacturer's probabilistic sensitivity analysis to include the extra variables of annual drop-out rate, cost of infliximab, length of inpatient stay and number of outpatient visits. The combined result of these changes gave an ICER of £33,200 using 4th quartile utilities and a 38% probability of being cost effective at a £30,000 threshold.
3.12 At the request of the Committee the manufacturer undertook additional analyses, which are described in sections 3.13 to 3.15.
3.13 The manufacturer presented an analysis using utilities derived from the EXPRESS trial. It converted SF-36 (36-item Short Form Health Survey) quality of life data into EQ-5D utilities by using an algorithm that was based on unpublished research. However, the manufacturer argued that the utilities from the York report are more appropriate on the grounds that: a) they are more generalisable to the wider patient population because they are based on data that reflect clinical practice; b) SF-36 displays a floor effect and as such can underestimate the impact of some chronic conditions on health-related quality of life. The manufacturer produced utilities for the whole trial population corresponding to the all-patient group defined in the York report, and a 4th-quartile group defined as those with a PASI greater than 12 and a DLQI greater than 18. In addition, the manufacturer combined the utilities from the York report and those from EXPRESS to obtain a pooled mean estimate.
3.14 The manufacturer presented analyses using three different values for the cost of infusions. It used its base case of £65.02 per infusion from national reference costs for dermatology outpatient visits, £78.20 derived from TA103 and £124 described as the cost per administration in the assessment report for rheumatoid arthritis (a systematic review of the effectiveness of adalimumab, etanercept and infliximab for the treatment of rheumatoid arthritis in adults and an economic evaluation of their cost effectiveness, West Midlands Health Technology Assessment Collaboration, October 2005).
3.15 For the all-patient population the ICER against etanercept varied from £44,000 to £49,000 per QALY gained, and in comparison with efalizumab the ICER varied from £42,000 to £47,000 per QALY gained depending on the utilities and costs used. For the population defined as being in the 4th quartile of DLQI values, the ICER against continuous etanercept varied from £26,000 in the base case to a maximum of £35,000 and when compared with efalizumab the ICER varied from £25,000 in the base case to a maximum of £34,000. Combining the use of EXPRESS utilities and a cost of £124 per infusion resulted in the highest ICERs.
3.16 Full details of all the evidence are in the manufacturer's submission, the ERG report and responses to clarification requests.