Tocilizumab for the treatment of systemic juvenile idiopathic arthritis

The Committee considered the evidence for the effectiveness of tocilizumab and noted that the manufacturer derived data from the TENDER trial, part one of which was a 12-week randomised controlled trial that compared the efficacy of tocilizumab with placebo. The Committee heard from the clinical specialists that the population was largely generalisable to the UK, but that the mean age of approximately 10 years in the TENDER trial was older than the population they would treat with tocilizumab in routine clinical practice. The Committee concluded that the trial generally reflected the UK population of children with systemic JIA but agreed that the mean age in the trial was older than the population treated in the NHS.

The Committee then considered the evidence for patients whose systemic JIA had not responded to NSAIDs, systemic corticosteroids and methotrexate. The Committee noted that the manufacturer had indicated that in the TENDER trial an inadequate response to methotrexate was defined as patients still showing symptoms of active systemic JIA at baseline despite being on a standard dose of methotrexate for a period of 3 months. The Committee agreed that the 95% of patients in the TENDER trial who were either being treated with methotrexate or had previously been treated with methotrexate could be considered to have disease that had not adequately responded to methotrexate. The Committee therefore concluded that data for these patients should be considered in any comparison of tocilizumab with TNF-alpha inhibitors or anakinra.

The Committee considered the evidence presented for the two populations defined in the scope and the different views of the population definitions from the manufacturer and the ERG. For the population of patients whose systemic JIA had failed to respond to NSAIDs and systemic corticosteroids, the Committee noted that only 5% of the TENDER trial population were methotrexate naive. The Committee also noted that the manufacturer had used a post-hoc analysis to compare patients receiving tocilizumab with those patients in the placebo group receiving methotrexate and that this was not methodologically acceptable. The Committee noted that in practice some patients' systemic JIA would still be responding adequately to methotrexate but no data for these patients had been presented. The Committee therefore concluded that there was no evidence to allow them to further consider the clinical or cost effectiveness of tocilizumab compared with methotrexate.

The Committee considered the evidence on the clinical effectiveness of tocilizumab in the population who had experienced treatment failure with NSAIDs, systemic corticosteroids and methotrexate. It noted that there were statistically significant improvements in the primary efficacy endpoint (ACR30 response and no fever) and all secondary endpoints at 12 weeks. The Committee was satisfied with the results from the TENDER trial and concluded that tocilizumab was efficacious for the treatment of patients whose systemic JIA had not responded to NSAIDs, corticosteroids and methotrexate.

After requests made at the first Appraisal Committee meeting the Committee noted that radiographic data from the TENDER trial was still unavailable. It also noted that the radiographic progression data from clinical trials presented by the manufacturer were conflicting and the sample sizes of the trials were small, and therefore there was little certainty about the findings for radiographic progression. The Committee concluded that it was possible that patients receiving tocilizumab would experience a delay in the progression of joint damage.

The Committee next considered the manufacturer's indirect comparison of tocilizumab with TNF-alpha inhibitors and anakinra. The Committee noted that the manufacturer used data from the TENDER trial and the NCT00036374 trial to perform an indirect comparison between tocilizumab and infliximab, and used infliximab to represent the class effect of the TNF-alpha inhibitors. The Committee was aware that the NCT00036374 trial was not specifically for patients with systemic JIA and included patients with other subtypes of JIA. However, the Committee noted that patients on tocilizumab were significantly more likely to reach an ACR30 response than patients on infliximab. The Committee also noted the manufacturer presented evidence from the ANAJIS trial that compared anakinra with placebo. The primary outcome measured in the ANAJIS trial was a modified ACR30 response without fever, measured after 4 weeks. The Committee noted that the TENDER and ANAJIS trials were used for an indirect comparison analysis of tocilizumab with anakinra. The Committee also noted that the manufacturer used the whole population from the TENDER trial to represent tocilizumab, whereas the ERG used only the 95% of patients whose condition had not responded to methotrexate. The Committee noted that patients on tocilizumab were significantly more likely to reach an ACR30 response than patients on anakinra. The Committee also noted that there was no significant difference between tocilizumab and anakinra in terms of ACR30 response plus absence of fever. The Committee concluded from the indirect comparison data that tocilizumab was clinically effective compared with anakinra and infliximab.

The Committee considered the revised economic model submitted by the manufacturer that included the patient access scheme for cost-effectiveness analysis. The Committee heard how homogenous health states had been defined by CHAQ score and had been categorised as 'controlled', 'mild', 'moderate', and 'severe'. The Committee noted that health states were based on CHAQ scores and regression models had been used to predict expected CHAQ categories using ACR responses. The Committee also noted that there was the possibility that individuals could have the same ACR response and still be in different baseline CHAQ categories, The Committee heard from the ERG that using ACR to define response transitions adhered to common modelling practice. It further noted that the application of the regression model from the 12 week data from the TENDER trial to the baseline CHAQ score to predict 12 week CHAQ values had not been adequately explained. The Committee concluded that although the manufacturer had submitted a revised economic model that represented the natural history of systemic JIA and its response to treatment better than the original model, there were still concerns with some aspects of the model.

The Committee considered the utility values used in the revised economic model. The Committee noted that the manufacturer had identified limitations of the CHQ (which had been used in the TENDER trial to elicit patients' health-related quality of life) and therefore had instead used data from the CHAQ. The manufacturer had made the assumption that the CHAQ score of a child was equal to the HAQ score of an adult and that the adult EQ-5D was equal to the health-related quality of life of a child. The Committee expressed concern about the methods and assumptions that had been used by the manufacturer and considered the utility value of 0.19 assigned to the severe health state in the revised model to be implausible given that approximately two thirds of children entered the model in this state. The Committee heard from the ERG that as most children leave the severe health state after 12 weeks, this may only have a limitedeffect on the long-term model. The Committee concluded however that this would over-estimate the incremental QALY gain ascribed to tocilizumab.

The Committee considered the costs for tocilizumab used in the revised economic model. The Committee noted that the costs of treatment were a composite of cost of medication and cost of administering the medication. The Committee heard from the clinical specialists that the costs for the health states in the model were a reasonable reflection of clinical practice in the UK. The Committee concluded that the costs in the model were reasonable, however it noted that potential cost savings could result from reductions in orthopaedic surgery for future joint damage and in bone marrow transplant and stem cell procedures, and that these factors had not been taken into account in the revised model.

The Committee considered the starting age of 5 years in the manufacturer's revised economic model and noted the comment received from a consultee that this was too high. The Committee concluded that the start age in the model was a mean age of 5 years and therefore the spread would include children who were below 5 years of age.

The Committee considered the manufacturer's revised base-case results with the patient access scheme applied. The Committee noted that the anakinra treatment strategies had used the primary outcome of ACR30 response and no fever. The Committee noted that two distinct incremental analyses had been performed: an anakinra-containing strategy that evaluated the strategies of tocilizumab followed by anakinra and anakinra followed by tocilizumab, and how each compared to anakinra alone. The manufacturer had done the same analyses for tocilizumab and infliximab. The Committee heard from the ERG that the incremental analyses the manufacturer presented were not fully incremental. The ERG also noted that the manufacturer's analyses were only designed to compare two technologies with one. The Committee further heard from the ERG that a more rigorous analysis would have involved tocilizumab, infliximab and anakinra in the same sequence using tocilizumab in different places of the sequence. The Committee also noted that as well as not conducting a fully incremental analysis, the manufacturer had not conducted a revised probabilistic sensitivity analysis. The Committee therefore concluded that the manufacturer's cost-effectiveness estimates could not be considered robust.

The Committee reviewed the ERG's exploratory analyses of the revised model, which included the patient access scheme. The Committee noted the ERG's analysis for the strategy tocilizumab then infliximab, compared with infliximab alone. The Committee considered that this strategy was the most appropriate because tocilizumab is the only licensed technology for systemic JIA and because the strategy of tocilizumab then anakinra compared with anakinra was not considered to be cost effective. The Committee agreed that the ERG's deterministic sensitivity analysis results were preferable to the ERG's probabilistic sensitivity analysis because only a limited number of simulations of the data were run in the probabilistic analysis. Despite the considerable uncertainty around the ICERs, the most plausible ICERs for the strategy tocilizumab followed by infliximab compared with infliximab alone were within a range that would be considered an acceptable use of NHS resources. Given the other factors that had not been taken account of in the manufacturer's model (such as steroid sparing, a decrease in health-related quality of life of the parents or carers, reduction in future orthopaedic surgical operations, bone marrow transplantation and stem cell procedures), on balance, the Committee concluded that the resulting cost-effectiveness estimate would be at the lower end of this range.

The Committee considered the manufacturer's revised sensitivity analyses that included the patient access scheme. The Committee noted the manufacturer's results, and the uncertainty around the adjustment factor derived from the etanercept study used to take account of other types of JIA subgroups in the infliximab study. The Committee noted that an increase or decrease of the adjustment factor by 30% on the tocilizumab then infliximab strategy made only a small difference to the manufacturer's revised base-case ICER. The Committee concluded that the revised model was robust to the sensitivity analyses of the adjustment factor.

The Committee considered the manufacturer's revised scenario analyses including the patient access scheme that looked at the impact of decreasing the frequency of administration of tocilizumab from every 2 to every 4 weeks after a treatment period of 6 months and of stopping tocilizumab after 2 years of treatment. The Committee noted that both scenarios improved the cost effectiveness of tocilizumab further. The Committee had heard from the clinical specialists that in some instances tocilizumab would be stopped when patients were in complete remission or dose administrations decreased when there was a significant improvement in the patient's condition. The Committee was of the view that when these situations arise in clinical practice that clinicians could consider reducing the frequency of administration of tocilizumab or stop using tocilizumab. The Committee also highlighted the importance of registries in collecting further data on patients receiving tocilizumab so that specific information about long-term outcomes and treatment-related adverse events in systemic JIA can be collected.

In summary, because the Committee did not have any clinical evidence on the comparison of tocilizumab with methotrexate, it concluded that tocilizumab could not be recommended for the treatment of systemic JIA in children and young people aged 2 years and older whose disease continues to respond to methotrexate or who have not been treated with methotrexate. For the population of children whose systemic JIA has not responded adequately to methotrexate as well as NSAIDs and corticosteroids, the Committee accepted the manufacturer's revised model despite its reservations about some aspects of the cost-effectiveness evaluation. The Committee noted the patient access scheme proposed by the manufacturer is a simple discount and would not incur additional costs. Consequently, the Committee concluded that tocilizumab represents a cost-effective use of NHS resources and should be offered as an option for the treatment of systemic JIA in children and young people aged 2 years and older whose condition has inadequately responded to NSAIDs, corticosteroids and methotrexate.