Ulcerative colitis (subacute manifestations) - infliximab

3. The manufacturer’s submission

The Appraisal Committee (appendix A) considered evidence submitted by the manufacturer of infliximab and a review of this submission by the Evidence Review Group (ERG; appendix B).

3.1. The manufacturer’s submission used a major systematic review conducted by the Cochrane Collaboration as a starting point for the literature search submitted. The search strategy then identified seven randomised controlled trials (RCTs), of which two were excluded from the analysis because they were not placebo controlled. The remaining five trials compared infliximab in ulcerative colitis with placebo. The trials had different settings and, despite having similar endpoints, used different scales to measure disease activity. In general, clinical remission was defined as a clearance of symptoms, and clinical response was defined as a marked reduction in symptoms as measured on disease activity scales. The ERG reviewed the search strategies used by the manufacturer and concluded that it was unlikely that major sources of relevant evidence had been missed. However, the ERG noted that people with ulcerative colitis requiring management in the inpatient setting would have different characteristics from the outpatient-managed population, and would also have a higher probability of requiring urgent surgery. Therefore, the ERG considered that the evidence from the registration trials in the manufacturer’s meta-analysis – ACT I (n = 364) and ACT II (n = 364) – was applicable only to the management of subacute manifestations of ulcerative colitis that did not require urgent hospitalisation.

3.2. In the manufacturer’s submission, ACT I and ACT II were considered to be the studies most relevant to this appraisal because the setting and dosing schedules were in line with the marketing authorisation. The inclusion criteria excluded patients expected to undergo surgery within 1 year. Pooled results for the major endpoints of the ACT trials at weeks 8 and 30 and results of ACT I at week 54 showed that infliximab was clinically effective, because it increased the rate of clinical response (relative risk [RR] 2.0, 95% confidence interval [CI] 1.6 to 2.4 at week 8; RR 1.8, 95% CI 1.4 to 2.3 at week 30; and RR 2.3 at week 54), remission (RR 3.7, 95% CI 1.7 to 8.5 at week 8; RR 2.3, 95% CI 1.6 to 3.3 at week 30; and RR 2.1 at week 54) and mucosal healing (RR 1.9, 95% CI 1.5 to 2.3 at week 8; RR 1.8, 95% CI 1.3 to 2.3 at week 30; and RR 2.5 at week 54). It also improved health-related quality of life (HRQoL) over the 54-week follow-up.

3.3. Safety data derived from the ACT I and ACT II trials showed no difference in total adverse-event rate between the infliximab and placebo groups. A published meta-analysis of trials for all TNF-α inhibitors in indications other than ulcerative colitis reported an increased elevation in the incidence of lymphoma in patients treated with TNF-α inhibitors when compared with patients who had not received them. However, the same meta-analysis showed that the rates of other types of malignancy in patients receiving TNF-α inhibitors were similar to the background level.

3.4. A state-transition cost-effectiveness model was developed by the manufacturer to evaluate the cost effectiveness of infliximab plus standard care when compared with standard care alone in subacute manifestations of ulcerative colitis. The model follows a cohort of hypothetical patients with moderate or severe ulcerative colitis, from entry through to 10 years, with patients being tracked as they move between the nine states of the model in each of the 8-week cycles. The three pre-surgery disease states in the model are defined using the Mayo score (remission: Mayo 0–2; mild: Mayo 3–5; and moderate/severe: Mayo 6–12). Two separate treatment strategies were evaluated. Strategy A modelled the continuation of infliximab treatment in patients whose condition showed a clinical response to the induction regimen of infliximab and maintained mild disease or remission. Strategy B modelled infliximab treatment continuation only in patients whose disease was in remission after the induction regimen and maintained remission after the subsequent doses.

3.5. The probabilities of moving between states are taken from the placebo and infliximab arms of the ACT trials, with the baseline risk of disease progression assumed to be that observed in the placebo arms of those trials. The primary source for the HRQoL estimates is a separate unpublished study based on the Health Outcomes Data Repository (HODaR), which was conducted using the EQ-5D questionnaire in patients with ulcerative colitis. Resource use was estimated based on the opinion of a panel of UK gastroenterologists and using baseline data from the ACT trials. Unit costs were extracted from UK-specific reference sources.

3.6. The base-case cost-effectiveness estimates presented in the manufacturer’s submission were £33,866 per additional quality-adjusted life year (QALY) gained for strategy A, and £25,044 per additional QALY gained for strategy B. For both strategies, extensive univariate sensitivity analyses were performed, which considered variations in time horizon, utility values, surgery rates, discount rates and average patient weight. Probabilistic sensitivity analyses were conducted to assess the uncertainty of the cost-effectiveness estimates by assigning distributions around some of the transition probabilities, the health-state utility estimates and some of the unit costs. These showed that: for strategy A, if willingness to pay for an additional QALY gained is £20,000 or £30,000, infliximab has an 11.4% or 33.9% probability of being cost effective, respectively; for strategy B, for a willingness to pay of £20,000 or £30,000 per additional QALY gained, infliximab has a 21.3% or 40.1% probability of being cost effective, respectively.

3.7. Following a request for clarification, the manufacturer stated that the trial data indicated no differences in efficacy between subgroups with different baseline medication when this was used as a proxy for disease severity. However, no data were available to assess whether cost effectiveness varied if the degree of disease was defined by the percentage of the colon involved. In addition, the differential cost effectiveness for the corticosteroid-refractory and corticosteroid non-refractory subpopulations could not be estimated because of a lack of data, although the trial data indicated that these subgroups had similar clinical responses.

3.8. Overall, the ERG considered that the model and the analysis presented adequately reflected the outpatient scenario, although there were still some important issues that were not fully addressed. The ERG considered that although the model structure appeared to be generally appropriate for the subacute outpatient setting, it did not include the cost and HRQoL impact of the serious adverse events reported in the trials.

3.9. The ERG considered that the small numbers used to estimate the transition probabilities could lead to a large amount of uncertainty throughout the model, especially in the case of the extrapolation of week 30–54 data to a 10-year time horizon. The ERG stated that univariate sensitivity analysis showed that the extrapolation exercise appears to have more impact on the cost-effectiveness estimates for strategy A than on those for strategy B, with infliximab becoming less cost effective with shorter time horizons. More scenarios with lifetime horizons were modelled, with different assumptions about the duration of the treatment effect, and this was again shown to have a greater effect on the strategy A results than the strategy B results. The cost-effectiveness estimates also appear to be sensitive to the assumptions made about the existence of a rebound effect.

3.10. The ERG considered that the justification provided for using utility data from studies other than ACT I and ACT II, despite these being the main sources of clinical data for the modelling work, was not reasonable. In the ERG’s view, the univariate sensitivity analyses performed highlighted the importance of the utility estimates in driving the cost-effectiveness results, because the highest incremental cost-effectiveness ratio (ICER) for both strategies is seen when the utility data from the ACT I trial are used (£60,750 and £53,504 per additional QALY gained for strategy A and strategy B, respectively). Therefore, the judgement about which set of utility data is most plausible is of critical importance when estimating cost effectiveness.

3.11. The ERG conducted further exploratory analyses to address a discrepancy between the criteria used to identify responders in the trials and the criteria used in the model to define health states: trial criteria (using both the total score and the subscores of the Mayo scale) to define response at week 8 were used to identify responders whereas the criteria used to define the health states in the model used only the total Mayo score. In these exploratory analyses, the ERG also attempted to address the fact that the transitions between the standard-care health states (to where patients whose condition does not respond to infliximab or placebo progress) are inappropriately based only on the placebo patients in the trials who had a clinical response at week 8. These exploratory analyses resulted in increased ICERs for the base case when considering the two strategies (approximately £38,000 and £33,000 per additional QALY gained for strategy A and strategy B, respectively). However, the ERG stated that these estimates would need to be considered cautiously because the full data that would be needed to follow all the patients throughout the trials were not available.

3.12. Full details of all the trials are in the manufacturer’s submission and ERG report, which are available from www.nice.org.uk/TA140