Risankizumab for previously treated moderately to severely active Crohn's disease

The standard initial treatment for Crohn's disease is conventional treatment with corticosteroids and immunomodulators such as azathioprine, mercaptopurine and methotrexate. If these treatments fail, people are offered biological treatment. The clinical experts stated that tumour necrosis factor (TNF)-alpha inhibitors (infliximab or adalimumab, including biosimilars) are usually used first. Ustekinumab or vedolizumab are used when a TNF-alpha inhibitor has failed, is contraindicated or cannot be tolerated. These treatments are recommended in NICE's technology appraisal guidance on infliximab and adalimumab, ustekinumab and vedolizumab. A patient expert noted that NICE's recommendations on infliximab, adalimumab and ustekinumab specify that when more than 1 treatment is suitable the least expensive option should be used. NICE's recommendations also state that the benefit of continuing these treatments should be assessed at 1 year. The clinical experts explained that, when there is evidence of clinical benefit, treatment continues beyond 1 year and that stopping effective treatments would be very rare. They stated that if the treatment no longer works (including after a dose increase, if relevant), another biological treatment would be considered. The clinical experts stated that although there are several clinically effective biological treatments for Crohn's disease, these do not cause long-term disease remission for everyone. Some people may have a sequence of biological treatments followed by surgery. A patient expert stated that complications of surgery and the potential effect on fertility can have a large impact on people with the condition. The clinical and patient experts agreed that it is very important to have a range of treatment options to help more people gain and regain disease remission, and delay surgery. The committee concluded that people with Crohn's disease would value the availability of a further treatment option to improve symptoms and induce disease remission.

Risankizumab is a novel treatment with a different mechanism of action to existing treatments. The company proposed it can be used either after conventional treatment (conventional care failure population) or after a biological treatment (biological treatment failure population) as an additional biological option. The clinical experts explained that treatment is not dependent on disease location. They agreed with the company's positioning of risankizumab in the clinical pathway and that it was relevant to consider both the conventional care failure and biological treatment failure populations. However, they noted that it would probably be used more after biological treatment. This is because TNF-alpha inhibitors are an effective and less expensive treatment option (because biosimilars are available), so most people have them first. A patient expert highlighted the need for people with the condition to have access to the most effective treatment first. This is because it can take years before they find an effective treatment, and this process can delay finding a treatment that induces disease remission. The committee concluded that risankizumab could be an option for both the conventional care failure and biological treatment failure populations. It further concluded that the relevant comparators for the conventional care failure population are adalimumab, infliximab and ustekinumab. The relevant comparators for the biological treatment failure population are ustekinumab and vedolizumab.

The co-primary outcomes for all 3 trials were clinical remission and endoscopic response. Clinical remission was measured by either a CDAI score below 150 or patient-reported outcomes on stool frequency and abdominal pain, but both outcomes were collected in all trials. The SES‑CD was used to measure endoscopic response alongside either measure of clinical remission. The clinical experts explained that the CDAI is used primarily in clinical trials as a measure of remission but not in clinical practice because of the time needed to complete the measurements included in this index. The clinical experts stated that the Harvey‑Bradshaw Index is broadly comparable to the CDAI and is used in clinical practice. A clinical expert stated that the SES‑CD is used increasingly in clinical practice. The committee concluded that the measures of remission used in the trials would give applicable estimates of expected remission and endoscopic response rates in clinical practice.

The results from the induction trials suggested that risankizumab is associated with higher rates of clinical remission and endoscopic response compared with placebo in the conventional care failure and biological treatment failure populations. The results from FORTIFY suggested that risankizumab is associated with higher rates of endoscopic response compared with placebo in the conventional care failure and biological treatment failure populations. The committee noted that the FORTIFY subgroup results in the conventional care failure and biological treatment failure populations were not statistically significant for clinical remission assessed by the CDAI. The committee concluded that risankizumab is associated with higher rates of clinical remission and endoscopic response compared with placebo (withdrawal) when used as a first biological treatment or after a previous biological treatment.

Because of the lack of direct comparative evidence, the company's original submission included network meta-analyses for induction and maintenance treatment in the conventional care failure and biological treatment failure populations. The outcomes assessed were clinical remission and response (defined by the CDAI). In its original submission, the company used a Bayesian risk difference fixed-effects model. For the network meta-analyses for maintenance treatment, the company split the clinical trial evidence into 2 separate networks (risankizumab and ustekinumab, and adalimumab, infliximab and vedolizumab). It stated it chose this approach because:

• of differences in drug mechanism of action, induction duration and half-life

• single network analyses lacked face validity (the estimated rates of remission were higher in people who had placebo)

• of methodological challenges in accounting for the heterogeneity.
  
  The EAG disagreed with the company on:

• Splitting networks: The EAG noted that connections in network meta-analyses should be based on comparator connections, not drug characteristics. It also noted ustekinumab and vedolizumab, used in the different networks, have a similar half-life and are comparable treatment options. The EAG presented results using a single network. The committee agreed that a single network was more appropriate.

• Using a fixed-effects model rather than a random-effects model: The EAG agreed with the company that there were several differences between the trials that made doing network meta-analyses more challenging. This included differences in baseline risks, stratification by the conventional care failure and biological treatment failure populations, and an observed temporal effect in which remission rates in placebo groups appeared higher in more recent trials. Given these differences, using a random-effects model is more appropriate. The committee noted the company's concerns that its exploration of a random-effects model produced results with wide confidence intervals and included values which favoured placebo over biological treatments. However, the committee agreed with the EAG that a random-effects model was preferable.

• Lack of adjustment for baseline risks or temporal effect: The EAG noted that the company's risk difference approach was not an adjustment for heterogeneity. The committee agreed that there was no evidence that the company's approach minimised differences between placebo group results between trials in the network. The EAG preferred to include an adjustment for the temporal effect observed in placebo remission rates. The committee agreed an adjustment for temporal effect was needed.
  
  Overall, the committee concluded that it preferred the EAG's approach because it was more methodologically appropriate, but that the relative clinical effectiveness of risankizumab compared with other biological treatments was highly uncertain with either approach. The committee further noted that models using risk ratios rather than risk differences to compare risankizumab with the other biological treatments may be more informative, given the heterogeneity of studies in the network. This is because relative effect tends to be more stable across risk groups than absolute risk. It would also allow further, and more straightforward, exploration of data to improve the precision of the modelled comparative effectiveness estimates (by using an informative prior). At the first committee meeting, the committee concluded that updated analyses were needed:

• using a single maintenance network with an adjustment for temporal effect

• using risk ratios rather than risk difference and presenting the credible intervals around the estimates

• exploring both random-effects models and fixed-effects models.

In response to draft guidance consultation, the company used the committee's preferred methods for its network meta-analyses. The company stated that the results suggested risankizumab and its comparators had similar clinical effectiveness, consistent with its previous analyses. The company noted that no statistically significant differences were observed. The committee noted that a lack of statistically significant differences between treatments did not mean that they were equally clinically effective. The exact point estimates for the comparisons and credible intervals are confidential and cannot be reported here. However, the committee noted that in all comparisons the point estimates of risk ratios for CDAI-measured clinical remission were below 1 for adalimumab, infliximab and ustekinumab compared with risankizumab, and around 1 for risankizumab compared with vedolizumab. In all comparisons, the 95% credible intervals were wide and included values above and below 1. In these analyses, a value below 1 suggests that clinical remission is more likely with the comparator than risankizumab, and a value above 1 suggests that clinical remission is more likely with risankizumab than the comparator. The EAG did not consider that the network meta-analyses supported equivalent clinical effectiveness across treatments. A clinical expert noted the difficulty in assessing clinical equivalence without clinical trials directly comparing the treatments. Based on experience in clinical practice and clinical trial evidence on risankizumab, the clinical expert considered that risankizumab is likely similarly effective to TNF-alpha inhibitors and ustekinumab, probably more effective than vedolizumab, and likely at least as effective as ustekinumab. They also cited a published meta-analysis suggesting that risankizumab's effectiveness is similar to other treatments (Barberio et al. 2022). The committee noted difficulties in doing network meta-analyses with the available trial data (for example, because of differences in treatments, trial designs or populations) and agreed that the results are uncertain. It stated that clinical equivalence or inferiority is usually assessed in specifically designed large-scale trials. The clinical expert also said that treatment response has been seen in people having risankizumab after previous biological treatment. The committee agreed that, based on the available evidence, there was only enough certainty that risankizumab had at least equivalent benefits to vedolizumab.

In its original submission, the company presented a cost-utility model comparing risankizumab with other biological treatments in the conventional care failure and biological treatment failure populations. It consisted of a short-term induction phase (decision tree) and a long-term maintenance phase (Markov state transition model). It assumed that people with moderately to severely active Crohn's disease have the same mortality as the general population. The maintenance phase modelled people having risankizumab or other biological treatments after a response to induction, or having conventional care if their Crohn's disease had not responded to induction treatment. After the first biological treatment, all people were modelled to have conventional care. With each maintenance treatment people were modelled to be in one of 4 health states: remission (CDAI score below 150), mild disease (CDAI score 150 to below 220), moderate to severe disease (CDAI score 220 to below 600), and surgery. It assumed that people with moderately to severely active disease could have surgery (constant rate across treatment arms based on NHS hospital episode statistics annual rates) and return to a CDAI-based health state after 8 weeks. The EAG explained that the CDAI used in the model to define clinical response and remission, and the severity of the disease, is not used in clinical practice. However, the clinical experts explained that the CDAI measure correlates with the Harvey-Bradshaw Index which is commonly used in clinical practice. The EAG further explained that the model does not reflect the lifelong relapsing–remitting nature of Crohn's disease because it does not allow people to have multiple biological treatments. Instead, it assumes all people have conventional treatment after a biological treatment. The company explained that a similar structure was used in previous NICE technology appraisals of treatments for Crohn's disease. The clinical experts agreed with the EAG that the model does not reflect the current clinical pathway. Overall, the committee concluded that although a CDAI-based model may be appropriate, the model is not suitable for decision making because it did not reflect the treatment pathway in which people can have more than 1 biological treatment.

In the model, people could either stop treatment because of a lack of efficacy or once they had reached assumed maximum treatment duration. The company assumed a 1‑year maximum treatment duration. The EAG assumed a 20‑year maximum treatment duration and applied the company's rate of discontinuing treatment because of a lack of efficacy for 20 years. The company explained that a 1‑year maximum duration is consistent with modelling in previous NICE technology appraisals of treatments for Crohn's disease. The EAG noted that most people in FORTIFY were still having treatment at 1 year and that 1 year does not reflect the lifelong nature of Crohn's disease. The clinical and patient experts agreed with the EAG that a 1‑year maximum treatment duration does not reflect clinical practice and would not be fair for people with the condition. The company confirmed that it did not intend that there would be a 1‑year stopping rule for risankizumab in clinical practice. The committee concluded that the 1‑year maximum treatment duration assumption in the model was too short. It concluded that the EAG's 20‑year maximum treatment duration was more reflective of clinical practice and appropriate for use in the cost-utility model.

The company developed a new model for cost comparison that had some different assumptions to its original cost-utility model:

• It used a 10‑year treatment duration rather than a maximum of 1 year (company's preference) or 20 years (EAG's preference) used in the cost-utility model (see section 3.11). It also did not account for some people stopping treatment before 10 years. A clinical expert stated that 10 years of treatment would be longer than expected and draft guidance consultation comments stated that 2 to 3 years is the reported median duration of treatment persistence with biologics. The EAG also provided a scenario analysis modelling treatment discontinuation. The committee considered that if using a cost-comparison analysis it was more appropriate to determine a time horizon sufficient to capture any differences in costs rather than to model treatment discontinuation. This is because treatments which are discontinued because of loss of effectiveness or tolerability may appear to be the cheapest option. It also agreed that a shorter time horizon than the company's 10 years may be appropriate.

• Infliximab and vedolizumab maintenance treatments can be delivered intravenously or subcutaneously. The company assumed that half of all people would have intravenous treatment and the other half subcutaneous. The model presented results that assumed 50% of vedolizumab and infliximab was administered subcutaneously and 50% intravenously. In the company's cost-utility model, a comparison of intravenous and subcutaneous vedolizumab and infliximab had been presented separately. A clinical expert agreed with the company, saying that the 50% split between intravenous and subcutaneous treatments, and the percentage of people having high doses assumed in the company's modelling, reflected clinical practice.

• Some maintenance treatments can be used at standard or high doses. The company assumed that 50% of people having adalimumab, 40% having infliximab, 92.5% having ustekinumab and 30% having intravenous vedolizumab start maintenance treatment with the high dose. A clinical expert stated that these assumptions were similar to what they had observed in their own centre. In the cost-utility model, an annual dose escalation using the same percentage was applied. They explained that although not everyone would start with a high dose, the dose would be increased soon after starting treatment, so the assumption reflects a simplification of clinical practice.
  
  The committee concluded that the company's modelling assumptions were appropriate, but a shorter time horizon should be considered in its decision making.

No equality or social value judgement issues were identified.