Vedolizumab for treating moderately to severely active Crohn's disease after prior therapy

Clinical effectiveness

4.3 The Committee discussed the generalisability of the GEMINI II and III trial populations. The Committee noted that, although a patient's Crohn's Disease Activity Index (CDAI) score could potentially exceed 600, the maximum CDAI score permitted at trial entry was 450. It heard from the clinical experts that only a few patients with a CDAI score greater than 450 were seen in routine clinical practice and consequently considered that the spectrum of disease severity in patients in GEMINI II and III (CDAI score 220–450) was broadly comparable to that seen in clinical practice. The Committee concluded that the clinical characteristics of the populations in GEMINI II and III were generalisable to the population likely to have vedolizumab in clinical practice in England.

4.4 The Committee discussed which patient groups in the GEMINI II and III trials most closely matched the population likely to have vedolizumab in clinical practice in England. It was aware that the ERG had suggested that the clinical efficacy of vedolizumab may be different in those who had received previous TNF‑alpha inhibitor treatment compared with those who had not, making it difficult to interpret the results from the intention‑to‑treat mixed populations, which included both patients who had not previously had TNF‑alpha inhibitor treatment and those in whom a TNF‑alpha inhibitor had failed. The Committee noted that 58% of patients in GEMINI II and 76% of patients in GEMINI III had experienced failure of a TNF‑alpha inhibitor. The Committee concluded that for the purposes of its decision‑making, it would be appropriate to evaluate vedolizumab in 2 distinct populations: those who had not had a TNF‑alpha inhibitor before and those in whom a TNF‑alpha inhibitor had failed.

4.5 The Committee discussed the induction regimens used in GEMINI II and III, including the timing of response assessment. It heard from the clinical experts that clinical trials used stringent definitions of response and remission, including the timing of assessment, whereas this was more flexible in clinical practice. The Committee noted that in the marketing authorisation for vedolizumab, the recommended induction dosing regimen is 300 mg at weeks 0, 2 and 6, corresponding with the induction regimen in GEMINI III, whereas patients in GEMINI II had vedolizumab at weeks 0 and 2. The Committee also noted that, although response was assessed at 6 weeks in GEMINI II and III, the marketing authorisation allowed for an additional dose to be given at week 10 (with the first dose of maintenance treatment administered at week 14). The Committee further noted that the company's new evidence submitted in response to the appraisal consultation document stated that, in patients in whom a TNF‑alpha inhibitor had failed, a numerically higher percentage of patients experienced remission at 10 weeks than at 6 weeks. The Committee heard from the clinical experts that they considered 6 weeks to be too early to discontinue treatment if a response had not been observed, and that induction response would generally be assessed later than this in clinical practice. It further heard that assessing remission rates at 6 weeks, may therefore not have reflected vedolizumab's true clinical efficacy, and that in clinical practice a patient was likely to have 4 doses before a decision was made to discontinue treatment because of a lack of response. The Committee concluded that assessing response at 6 weeks, as in GEMINI II and III, would not detect all patients whose disease would respond to induction treatment, and that using data from later time points in the trials could potentially increase the efficacy estimates for vedolizumab.

4.6 The Committee considered the clinical effectiveness of vedolizumab compared with placebo during induction treatment in GEMINI II and III. It noted that vedolizumab was more effective in inducing remission at 6 weeks than placebo in the intention‑to‑treat mixed population in GEMINI II and III. It further noted that the remission rate in GEMINI II was numerically higher with vedolizumab than placebo in both the population who had not had a TNF‑alpha inhibitor before and the population in whom a TNF‑alpha inhibitor had failed. The Committee was aware that in GEMINI III the primary outcome was remission at 6 weeks in the population in whom a TNF‑alpha inhibitor had failed (76% of the trial population). It noted that although the primary outcome in GEMINI III had not been met, an exploratory secondary analysis at 10 weeks did show a statistically significant benefit with vedolizumab in this group. The Committee concluded that for induction, vedolizumab improved clinical remission rates compared with placebo in the whole population, people who had never had TNF‑alpha inhibitors and people in whom TNF‑alpha inhibitor treatment had failed.

4.7 The Committee considered the clinical effectiveness of vedolizumab compared with placebo for maintenance treatment. It noted that the evidence was based only on the results of GEMINI II because GEMINI III did not include a maintenance phase. The Committee heard from the clinical experts that, in Crohn's disease, long‑term maintenance of remission was the primary goal of treatment. The Committee noted, however, that the duration of GEMINI II was 52 weeks, that only limited data up to 104 weeks were available from GEMINI LTS. At 52 weeks, vedolizumab showed higher remission rates than placebo in the intention‑to‑treat population, in the population who had not had a TNF‑alpha inhibitor before and in the population in whom a TNF‑alpha inhibitor had failed. The Committee noted that the absolute remission rate was lower with both vedolizumab and placebo in the population in whom a TNF‑alpha inhibitor had failed than in the population who had not had a TNF‑alpha inhibitor before (see section 3.14). However, it observed that despite this lower absolute rate, there was a similar relative treatment effect and the remission rate was approximately twice as high with vedolizumab than with placebo in both of these populations. The Committee heard from the clinical experts that this reflected the fact that these patients had more established and difficult to treat disease, and that even a reduced absolute treatment effect would be perceived as highly beneficial in these patients because of the lack of other treatment options. Although it was uncertain about longer‑term effects, the Committee concluded that, for maintaining remission up to 52 weeks, vedolizumab was significantly better than placebo in the whole population, in the population who had not had a TNF‑alpha inhibitor before and in the population in whom a TNF‑alpha inhibitor had failed.

4.8 The Committee discussed the impact of vedolizumab on health‑related quality of life, and was encouraged that the company had included self‑reported quality of life in its clinical trials. It focused on the EQ‑5D results because these were used to generate utility values for the company's economic model. It noted that the company's submission stated that a decrease in EQ‑5D score of at least 0.3 points was considered a clinically meaningful improvement, and that all results had been presented as decreases in score compared with baseline values. However, the Committee was aware that an improvement in quality of life would be reflected by an increase in EQ‑5D score. The company was unable to explain why it considered that vedolizumab improved quality of life, as assessed by EQ‑5D, yet the scores decreased. The Committee concluded that, although the results using other assessment methods in GEMINI II and III suggested that vedolizumab could improve quality of life, it was unable to conclude what its effect was using EQ‑5D scores.

4.9 The Committee discussed the safety of vedolizumab. It heard from the clinical experts that vedolizumab was thought to have a more favourable adverse‑event profile than other biological treatments (such as TNF‑alpha inhibitors) because of its gut‑specific effect and fewer systemic side effects. The Committee understood that this could be a consequence of its selective mechanism of action, because the α4β7 integrin is expressed only on gut‑selective T‑helper lymphocytes. The selective effect might be a particular advantage in some people for whom a TNF‑alpha inhibitor is contraindicated. The Committee was aware that progressive multifocal leukoencephalopathy, a fatal condition affecting the brain has been seen with natalizumab, an antibody that inhibits α4‑integrin. It was aware that because vedolizumab also inhibits a α4‑integrin, the incidence of progressive multifocal leukoencephalopathy in people who have treatment with vedolizumab is being closely monitored, but noted that there have been no reports to date. It concluded that fewer systemic effects may be an advantage of this therapy over existing biological treatments for some patients, particularly if TNF‑alpha inhibitor treatment was contraindicated.

4.10 The Committee considered the validity and usefulness of the company's network meta‑analyses that compared vedolizumab with adalimumab and infliximab. It agreed with the ERG and the company that it was more meaningful to evaluate the clinical effectiveness of vedolizumab according to TNF‑alpha inhibitor status rather than the overall population, because the results for this population were difficult to interpret and generalise (see sections 3.27, 3.36 and 3.37). It also agreed with the company and the ERG that the network meta‑analyses of the results for the population in whom a TNF‑alpha inhibitor had failed had a number of serious flaws. The Committee accepted that the trial populations in the induction phase in the vedolizumab and adalimumab studies were not comparable, and that it had not been possible to present an analysis for the maintenance phase. The Committee concluded that the network meta‑analyses for the population in whom a TNF‑alpha inhibitor had failed would not inform its decision‑making, but that it should further consider the network meta‑analysis in the population who had not had a TNF‑alpha inhibitor before.

4.11 The Committee considered the validity and usefulness of the company's network meta‑analyses that were provided at the clarification stage for the population who had not had a TNF‑alpha inhibitor before. It considered the population in the network meta‑analyses to be broadly generalisable to the population presenting after conventional non‑biological therapy has failed in clinical practice in England. However, it had several concerns:

Cost effectiveness

4.12 The Committee took into account the ERG's and the company's view that it was more meaningful to evaluate vedolizumab according to TNF‑alpha inhibitor status rather than the mixed population, because the results of the mixed analysis were difficult to interpret and generalise to the NHS (see sections 4.3 and 4.10).

4.13 The Committee considered the structure of the company's economic model. It noted the ERG's concerns about several structural assumptions, including that the relapsing and remitting nature of the disease had not been captured, the simplistic approach to modelling surgery, the assumption that all patients whose condition did not respond to treatment had moderate to severe disease for the full duration of the model and that there would be no difference in outcomes between patients whose disease responded and patients whose disease did not respond in the moderate to severe health state. The Committee heard from the clinical experts that, for patients in whom multiple lines of therapy failed, the assumption of long‑term continuation in the moderate to severe state as in the company's model was not unreasonable. The Committee concluded that the number of concerns raised by the ERG meant that it was uncertain if the model was structurally sound but that, overall, it was acceptable to inform its decision‑making.

4.14 The Committee discussed the dosing assumptions used during induction in the company's economic models. It noted that in the first 2 models the same treatment duration (6 weeks) was assumed for all therapies for the induction phase and that these were aligned with the clinical trials, rather than the marketing authorisations or clinical practice. The Committee appreciated the difficulty in aligning the model with clinical practice when this did not necessarily correspond with the trial data, but considered the assumption of 6 weeks for all therapies was a weakness because it did not give an accurate estimate of costs and clinical outcomes in clinical practice. However, the Committee was satisfied with the dosing assumptions used in the company's third model, and concluded that evaluating response at 10 weeks was appropriate.

4.15 The Committee discussed the assumptions related to treatment continuation in the company's economic models. It addressed the assumption that biological treatment would stop in all patients at 1 year. The Committee heard from the clinical experts that they would try to withdraw biological therapy after 1 year where remission had been achieved, but that treatment would be continued if there was a high risk of relapse or surgery, and that treatment would be likely to continue if the patient had not experienced remission but was gaining a benefit from treatment (that is, an improvement in symptoms). In general, it heard that they would not wish people to continue on a treatment if it was not needed. The Committee concluded that the company's approach to discontinuing biological therapy after 1 year of maintenance treatment was not unreasonable, but that treatment duration in clinical practice could be longer.

4.16 The Committee considered how the company had modelled long‑term clinical effectiveness of treatment. It recalled that the NICE reference case specifies that the time horizon should be long enough to capture all associated costs and benefits. Because Crohn's disease is a chronic condition, the Committee took the view that a lifetime horizon (as in the third model) was more appropriate than a 10‑year time horizon (as in the first 2 models). However, it had concerns about the relationship between the constant transition probabilities and the time horizon adopted. The Committee considered that the long‑term benefit was difficult to predict, particularly when treatment was stopped at 1 year, and if overestimated, this would be amplified with a longer time horizon. The Committee concluded that there were uncertainties in the long‑term modelling of clinical effectiveness of vedolizumab and its comparators, but that the company's approach in the third model (which adopted a lifetime horizon as recommended in the NICE reference case) was reasonable given the evidence available.

4.17 The Committee discussed the modelling of long‑term adverse effects of conventional non‑biological treatment and surgery. It was aware that repeated use of high‑dose oral corticosteroids (which may be the mainstay of treatment when other medical options, including TNF‑alpha inhibitors, have failed) was associated with a range of long‑term adverse effects such as diabetes and osteoporosis. The Committee considered that these could have a substantial impact on health‑related quality of life and could be associated with significant costs (for example, treating hip fractures and complications of diabetes). It considered that these would potentially have a greater impact when adopting a lifetime time horizon rather than 10 years. It noted that disutilities and costs for treatment of long‑term adverse effects of corticosteroids had not been included in the company models. The Committee also noted consultation comments about costs of treating any unwanted consequences of surgery, such as infertility. The Committee concluded that the company's model did not capture all the costs and disutilities associated with existing treatments for Crohn's disease, and that incorporating them would be likely to reduce the ICER in favour of vedolizumab.

4.18 The Committee discussed how surgery had been modelled. It was concerned that, in the first 2 models, around one‑third of patients who had surgery would remain in the surgical health state in the next cycle, and noted that the associated health state costs were considerable (around £10,000 per cycle). The Committee considered that this was an unreasonably high proportion, and the modification made in the company's third model, which reduced the repeat surgery rate to 1%, was more appropriate. The Committee concluded that it was likely that the company's first 2 models overestimated the proportion of patients having repeated surgery, which would have the effect of overestimating the total costs for these patients, but that the proportion and costs used in the third model were acceptable.

4.19 The Committee discussed the clinical parameters used in the company's economic models. For the population in whom a TNF‑alpha inhibitor had failed, it accepted that the network meta‑analyses did not permit a robust comparison with the other biological therapies (see section 4.10). For the population who had not had a TNF‑alpha inhibitor before, it noted that the company's submission was unclear about which results from the network meta‑analyses had informed the model. The Committee was aware that the ERG considered that the results of the company's network meta‑analysis could underestimate the uncertainty in treatment effects, because fixed‑effects models had been used despite evidence of heterogeneity among the clinical trials in the network. The Committee noted the ERG's concerns about the lack of information about how the GEMINI II and III results had been pooled, and the estimated rates of discontinuation because of adverse events. It noted further concerns that the company had used data from the single‑arm ACCENT‑I trial but excluded the placebo‑controlled Targan trial when estimating the relative treatment effect of vedolizumab compared with infliximab during induction, and that the company had not discussed the limitations associated with ACCENT‑I. It recalled that the ERG was not satisfied with the company's approach to modelling maintenance data for vedolizumab compared with adalimumab because CLASSIC II had been excluded (see section 3.38). Lastly, the Committee agreed that, in the absence of data suggesting otherwise, it should be assumed that the mortality rate is the same for all treatments and that the same risk should be applied to all Crohn's disease health states (as in the third model). The Committee concluded that, for the population in whom a TNF‑alpha inhibitor had failed, it was appropriate to derive the clinical parameters from the GEMINI II and III clinical trial results for the comparison of vedolizumab with conventional non‑biological therapy. The Committee further concluded that the clinical parameters derived from the network meta‑analyses for the population who had not had a TNF‑alpha inhibitor before were subject to considerable uncertainty.

4.20 The Committee discussed how the health‑state costs had been modelled by the company. It noted that the costs had been based on those reported in Bodger et al. (2009) in the first 2 models, but that they had been updated in the third model so that they were derived from a survey of 8 clinical experts based in England. It noted that amending the health‑state costs had significantly reduced the ICER for vedolizumab compared with conventional non‑biological therapy in patients in whom a TNF‑alpha inhibitor treatment had failed (see section 3.78). The Committee heard from the company that many of the costs described by Bodger et al. were out of date because of advances in diagnostic tests and monitoring in Crohn's disease. The clinical experts at the Committee meeting agreed with the company and stated that the heath‑state costs derived from the survey seemed reasonable because managing Crohn's disease in patients in whom TNF‑alpha inhibitors have failed uses very significant NHS resources. The Committee concluded that the health‑state costs used in the third model were more likely to reflect current NHS practice in England.

4.21 The Committee considered the cost‑effectiveness results for vedolizumab compared with conventional non‑biological treatment and TNF‑alpha inhibitors in people with moderately to severely active Crohn's disease. It noted that the ERG did not consider the company's ICERs to be robust because of the model's structural issues, and that consequently the ERG had not conducted exploratory analyses. The Committee noted the ERG's concerns but, in the absence of alternative estimates, decided to consider the company's analyses in further detail.

4.22 The Committee discussed the cost‑effectiveness results for vedolizumab compared with conventional non‑biological treatment and TNF‑alpha inhibitors in the population who had not had a TNF‑alpha inhibitor before. It noted that vedolizumab was subject to extended dominance and that it had been excluded from the ERG's fully incremental analysis (see section 3.70). It noted that the probability of vedolizumab being the most cost‑effective treatment option at £20,000 per QALY gained was less than 1%. The Committee concluded that vedolizumab was not cost effective compared with TNF‑alpha inhibitors for treating moderately to severely active Crohn's disease in the population who had not had a TNF‑alpha inhibitor before.

4.23 The Committee considered the cost‑effectiveness results for vedolizumab compared with conventional non‑biological treatment and TNF‑alpha inhibitors in the population in whom a TNF‑alpha inhibitor had failed. It acknowledged the opinion of the clinical experts that these were the people for whom access to a new agent would be of most value because of the very limited treatment options available to them (see section 4.2). It considered therefore that the most appropriate comparator in this population was conventional non‑biological therapy. It noted that the ICER for vedolizumab compared with conventional non‑biological therapy generated using the company's third model (that is, the model that incorporated many of the Committee's preferences) was £21,600 per QALY gained. The Committee was aware that in clinical practice, vedolizumab treatment could be longer than the 1‑year duration assumed in the company's base case (see section 4.15) and noted that extending vedolizumab's treatment duration to 2 or 3 years increased the ICER to £24,700 and £26,200 per QALY gained respectively. The Committee considered that the ICERs for vedolizumab compared with conventional non‑biological treatment were within the range that would normally be considered a cost‑effective use of NHS resources (that is, £20,000–30,000 per QALY gained) but it was concerned about the uncertainty in the modelling of long‑term treatment effects of vedolizumab (see section 4.16) and some of the structural assumptions (see section 4.13). However, it recalled that disutilities and costs for treatment of long‑term adverse effects of corticosteroids and surgical complications had not been included in the company models (see section 4.17), which would reduce the ICER, and noted the substantial unmet need in this population of patients. It concluded, on balance, that vedolizumab could be considered a cost‑effective use of NHS resources and should be recommended for people in whom TNF‑alpha inhibitor treatment has failed. It also concluded that when treatment duration is greater than 1 year, it would be important to identify the people who would continue to derive ongoing clinical benefit. Therefore it concluded that, in the absence of treatment failure, people receiving vedolizumab should be reassessed at 12 months and then at least annually to see if continued treatment is justified.

4.24 The Committee gave further consideration to whether there was a subgroup of patients who cannot take TNF‑alpha inhibitors because they are contraindicated or not tolerated, and in whom vedolizumab would provide the only medical alternative to conventional non‑biological therapy. The Committee appreciated the high unmet clinical need in this group (see section 4.2) for whom there would otherwise be no biological treatment options, and concluded that when TNF‑alpha inhibitors were contraindicated or not tolerated, it was reasonable for vedolizumab to be prescribed.

4.25 The Committee contemplated whether vedolizumab could be considered an innovative technology. It noted that vedolizumab has a different mechanism of action to other drug treatments for Crohn's disease, and the clinical experts' opinion that the systemic side effects of treatment were lower with vedolizumab compared with other treatments (including TNF‑alpha inhibitors). The Committee found it plausible that vedolizumab's gut‑selective mechanism of action could result in a more favourable side‑effect profile than other treatments (including other types of biological treatment) which have more systemic effects. It recalled that there were no head‑to‑head trials comparing vedolizumab and TNF‑alpha inhibitors and that the company had not presented adverse‑event data in the network meta‑analyses in its submission. The Committee concluded that vedolizumab has a different mechanism of action to other drug treatments for Crohn's disease and, in this regard, was innovative. As discussed in section 4.17, it also considered that the costs and disutilities associated with infertility and the long‑term adverse effects of oral corticosteroid use had not been captured in the company's cost‑effectiveness calculations.

4.26 The Committee discussed the potential equality issues raised during consultation. It was aware that surgery can have an impact on fertility and that surgery has particular issues for people following certain religious practices. The Committee heard from the clinical and patients experts that, in general, people with Crohn's disease preferred to manage their condition medically rather than surgically, and therefore considered that this did not represent an equality issue. The Committee considered that there were no equality issues associated with benefit according to age because vedolizumab brought treatment benefits to adults of all ages. The Committee concluded that there was no need to alter its recommendations because of any equality issues.

Summary of Appraisal Committee's key conclusions