Tocilizumab for the treatment of rheumatoid arthritis

The Committee first discussed tocilizumab given as monotherapy. It noted that the only clinical evidence for tocilizumab monotherapy came from a trial that included people who had not been previously treated with methotrexate and that tocilizumab monotherapy treatment for this population was outside the licensed indication of tocilizumab. The Committee also noted that no cost-effectiveness estimates of tocilizumab given as monotherapy had been presented by the manufacturer. It concluded that no evidence for tocilizumab monotherapy within its licensed indication was available, and therefore no recommendations for tocilizumab as a monotherapy could be made.

The Committee considered the evidence on the clinical effectiveness of tocilizumab plus DMARDs compared with placebo plus DMARDs. The Committee concluded that tocilizumab plus methotrexate was clinically effective compared with placebo plus DMARDs when given before TNF‑alpha inhibitors and when given before rituximab.

The Committee then considered the evidence for the relative efficacy of tocilizumab compared with etanercept and compared with rituximab when all treatment strategies were in combination with methotrexate. It understood that tocilizumab had not been compared head-to-head with etanercept (or any other TNF‑alpha inhibitor) or rituximab, and that indirect evidence had been combined in a mixed treatment comparison for this purpose. It noted the concerns raised by the ERG and clinical specialists regarding the mixed treatment comparison. The mixed treatment comparison assumed that the TNF‑alpha results could be regarded as a class; however, when merged, the overall results reduced the efficacy of etanercept. The Committee noted that the manufacturers had responded to its requests to remove the Klareskog trial of etanercept from the analysis because this was a large RCT with unusually high control-arm response rates and did not correspond with the inclusion criteria of the mixed treatment comparison. With this trial removed, the Committee noted that etanercept appeared at least equal to, and possibly had higher efficacy than, tocilizumab.

The Committee further noted the concerns of the DSU in its 2010 report regarding the adjusted ACR response rates from the mixed treatment comparison compared with the 'unadjusted' point estimates from the individual trials. It understood that the proportions of people achieving ACR20, ACR50 and ACR70 response rates for etanercept and rituximab resulting from the mixed treatment comparison were lower than the corresponding unadjusted trial ACR response rates. Conversely, the proportions of people achieving ACR20, ACR50 and ACR70 response rates were higher for tocilizumab in the adjusted mixed treatment comparison analysis than the unadjusted trial rates. The 2010 DSU report clarified that the counterintuitive results of the mixed treatment comparison had possibly arisen when the comparator response rates from all of the trials had been pooled. The Committee considered that the mixed treatment comparison included a set of heterogeneous trials, which meant that the results were subject to considerable uncertainty, and that limited confidence could be placed in the adjusted ACR response rates in the manufacturer's revised base case. The Committee concluded that using the unadjusted trial estimates in the analyses was more appropriate.

The Committee considered the relative efficacy of tocilizumab compared with etanercept and also with rituximab using the unadjusted trial estimates of ACR rates. It considered that the evidence was not conclusive of a benefit of any 1 drug over another. The Committee concluded that no convincing evidence had been presented to demonstrate the superiority of tocilizumab over etanercept or rituximab, but that the estimates were in a similar range to etanercept and rituximab.

The Committee considered the clinical evidence for tocilizumab after treatment with rituximab. Based on previous discussions it recognised that tocilizumab plus methotrexate is clinically effective compared with placebo plus methotrexate (see section 4.7). It noted the evidence from the RADIATE trial in which a subgroup of people had rheumatoid arthritis that had responded inadequately to 2 TNF‑alpha inhibitors. It understood that this was the only available evidence to consider the effectiveness of tocilizumab after rituximab. The Committee considered that it indicated a benefit of tocilizumab after 2 biological treatments. In view of this evidence and considering the comments from patient experts and clinical specialists, the Committee, on balance, agreed that tocilizumab was likely to benefit people whose rheumatoid arthritis has responded inadequately to rituximab.

The Committee discussed the appropriate approach for determining the cost effectiveness of tocilizumab. It understood that before the 2010 DSU report the manufacturer's ICERs were based on adjusted trial response rates from the mixed treatment comparison. It also understood that the 2010 DSU report presented analyses using 4 different approaches to evidence synthesis (see section 3.35). The Committee considered, on the basis of previous discussions (see section 4.8), that approach 1, in which the ACR response rates came from the mixed treatment comparison, was not appropriate. The remaining 3 approaches to evidence synthesis used the unadjusted trial response rates for all treatments and incorporated degradation rates. The Committee understood that approaches 2 and 3 only used the unadjusted ACR response rate from a single trial for etanercept, rather than from the 2 available trials. The Committee had a strong preference for approach 4, which used data from both of the etanercept trials. Approach 4 also corrected the counterintuitive ACR70 response rate for tocilizumab used as a third biological treatment in the treatment sequence noted by the DSU in the 2010 report. The Committee concluded that approach 4 to evidence synthesis (see section 3.35) was the most appropriate for consideration.

The Committee also discussed the 2 sensitivity analyses presented by the DSU within approach 4 in the 2010 report. The first concerned evidence supplied by the manufacturer for a long-term HAQ improvement. It understood that the data for a HAQ improvement with tocilizumab treatment came from open-label extension studies in which only the HAQ scores for people who remained on treatment were available. It noted that, for the open-label extension trial assessing the benefits of tocilizumab after the failure of conventional DMARDs (that is, before etanercept), approximately 30% of people had stopped treatment. It further noted that the confidence intervals around the mean HAQ scores at each point in time were wide. The Committee therefore considered that the manufacturer's evidence was not a robust estimate of the long-term HAQ improvement on tocilizumab and was subject to uncertainty. Furthermore, the manufacturer had not provided any comparable investigation into long-term HAQ trends for the comparator biological treatments other than rituximab. The manufacturer presented a graph of a stable HAQ trend for people on rituximab from the REFLEX trial. However, no data had been supplied by the manufacturer to support the graph. The Committee questioned the comparability of the rituximab and tocilizumab HAQ trend lines, and considered that single-arm extension trial data did not provide a direct comparison of the relative benefits between the 2 treatments. In addition, the Committee heard from patient experts and clinical specialists that it was unlikely that tocilizumab would provide a long-term HAQ benefit over and above that of any other biological treatment. Overall, the Committee could not support the assumption that there is a long-term HAQ gain with tocilizumab (that is, a HAQ improvement with tocilizumab) compared with no HAQ improvement with other biological treatments. It concluded, on the basis of the evidence presented, that the long-term HAQ improvement on tocilizumab treatment had not been demonstrated. The Committee agreed that the analyses that assumed no long-term HAQ improvement with tocilizumab were therefore the most appropriate for consideration.

The second sensitivity analysis that the Committee considered concerned the exclusion of negative utilities (health states worse than death) from the incremental analysis. The Committee noted that the manufacturer's mapping of HAQ scores to EQ-5D utility values resulted in negative utility values. It discussed that excluding negative utility values could be considered counterintuitive and did not allow for a worsening of quality of life when a person had rheumatoid arthritis. The Committee heard from the manufacturer that it was possible that there were some people with rheumatoid arthritis who may experience negative utility values. The Committee noted that the impact of removing the negative utilities from the incremental analysis was minimal. The Committee agreed that although the exclusion of negative utility values was subject to some debate, it was not a key issue in determining the cost effectiveness of tocilizumab. The Committee therefore accepted that the calculation of some ICERs would include negative utility values but concluded that this was acceptable because of the low impact on the ICERs.

The Committee considered the administration costs of tocilizumab. It noted comments received during consultation in 2010 that, although the infusion took 1 hour, the total time taken to administer tocilizumab in an organised unit would be at least 2 hours. The Committee then discussed the 2010 DSU analysis using approach 4 with no long-term HAQ improvement and the administration costs doubled. It heard from the DSU that the decision to double the cost was not based on a robust estimate of the time taken to administer tocilizumab, but was intended to illustrate the sensitivity of the ICERs to this assumption. Although the Committee agreed that a cost based on an administration time of 1 hour represented the minimum cost to the NHS, it did not agree that the true cost would be as much as double. The Committee therefore considered that it was not appropriate to double the administration cost of tocilizumab and concluded that the manufacturer's revised estimate of £154 was acceptable.

The Committee noted that some modelling assumptions in the manufacturer's submission had not been investigated by the DSU in the 2010 report. These included, first, any difference in the adverse events that may occur on biological treatment compared with those that might occur in palliative care. Second, that despite previous requests to the manufacturer to use directly observed EQ‑5D data, the revised base-case ICERs from the manufacturer were still subject to a HAQ mapping algorithm. The Committee highlighted its concern with this, but acknowledged that the data had not been available to investigate these assumptions.

In summary the Committee concluded that the best estimate of cost effectiveness of tocilizumab in any position in the treatment pathway should be based on approach 4 to evidence synthesis in which the ACR response rates came from the trials rather than the mixed treatment comparison and used a corrected degradation factor for tocilizumab (see section 3.35). In addition, it concluded that no long-term HAQ improvements with tocilizumab should be assumed.

The Committee considered the cost-effectiveness analyses submitted by the manufacturer in 2011 that were based on the preferred approach (see section 4.17) and that incorporated tocilizumab at the discount agreed as part of the patient access scheme (see section 2.4). It also considered the DSU 2011 report when reviewing the manufacturer's submission. It discussed the manufacturer's analyses, which the DSU replicated including fully incremental calculations (see sections 3.45 to 3.47) for all 3 patient subgroups: people whose rheumatoid arthritis has responded inadequately to 1 or more conventional DMARDs (DMARD-IR analysis); people who are intolerant to rituximab, or for whom rituximab is contraindicated (DMARD-IR rituximab intolerant); people whose rheumatoid arthritis has responded inadequately to TNF‑alpha inhibitors (TNF-IR analysis). The Committee accepted the DSU separate exploratory incremental analyses. It noted the DSU's comment from the 2011 report that the manufacturer's analysis had not taken into account extended dominance (when 1 or more drug sequences are less effective than and at least as costly as another sequence) and that this had an impact on the ICERs. The Committee concluded that the DSU's 2011 exploratory analyses should be used as the basis for determining the cost effectiveness of tocilizumab.

The Committee also considered the straightforward inferences that could be made from its separate clinical effectiveness and costing conclusions. These were that for the DMARD-IR population (who had not received a TNF‑alpha inhibitor or any other biological treatment) tocilizumab was similar in clinical effectiveness (see section 4.10) to the TNF‑alpha inhibitors and could be considered a plausible alternative. In the case of the TNF‑IR population (whose condition had failed to respond to a TNF‑alpha inhibitor but who had not yet tried rituximab), the position was different. Although tocilizumab might be as clinically effective as rituximab, it was also more expensive and so the Committee concluded tocilizumab could not be considered an option unless rituximab was contraindicated, not tolerated or had failed.

The Committee considered the DMARD-IR ICERs in the DSU's 2011 exploratory analysis. It noted from the total costs and QALYs for the sequences that when tocilizumab was the first biological treatment rather than etanercept, it was associated with fewer QALYs and less cost. It understood that this was because of the percentage of non-responders on tocilizumab (approximately 40%) when taken as a first-line biological treatment, which resulted in reduced time on tocilizumab treatment and therefore lower cost of the sequence. The Committee noted that this improved the cost effectiveness of tocilizumab. However, on the basis of previous discussions (see section 4.10) the Committee was not convinced that the clinical effectiveness of tocilizumab would be superior to that of etanercept. The Committee concluded that the improved cost effectiveness of tocilizumab as the first biological treatment compared with etanercept was due to the cost of time on treatment, rather than any substantial differences in clinical or cost effectiveness between tocilizumab and etanercept.

The Committee further considered the DMARD-IR ICERs from the DSU's 2011 exploratory analysis. It noted that although tocilizumab appeared cost effective as the first biological treatment (£5,700 per QALY gained), this sequence had rituximab as the third biological treatment in the sequence, rather than the second. The Committee raised concerns that this was counterintuitive because the total drug treatment cost of rituximab is approximately half that of either tocilizumab or etanercept. On the basis of previous discussions (see section 4.11) the Committee was not convinced that the clinical effectiveness of etanercept or tocilizumab would be sufficiently superior to rituximab such that a sequence in which rituximab was third would be more cost effective than 1 in which rituximab was second. The Committee noted that a sequence in which tocilizumab was the first biological treatment, followed by rituximab, followed by etanercept, had not been included in either the manufacturer's or the DSU's 2011 analyses. It was aware that in clinical practice this sequence would involve off-licence use of rituximab because the marketing authorisation restricts rituximab to use after an inadequate response or intolerance to other DMARDs including 1 or more TNF‑alpha inhibitors. However, the Committee considered that to understand the impact on the cost effectiveness of placing tocilizumab first in the sequence, it was important to consider all possible treatment sequences. It noted that in their exploratory incremental analysis from 2011, the DSU had incorporated an alternative baseline sequence of tocilizumab followed by rituximab. The Committee accepted this sequence as a proxy for tocilizumab, followed by rituximab, followed by etanercept. It noted that when this alternative baseline sequence was included in the exploratory analysis, 3 sequences were extendedly dominated (see section 3.49) leaving the baseline sequence of tocilizumab followed by rituximab, and the sequence of etanercept, followed by rituximab, followed by tocilizumab. Comparing these 2 sequences, tocilizumab as the third biological in the sequence had an ICER of £28,400 per QALY gained, compared with tocilizumab as the first biological treatment in the sequence. It accepted that some uncertainty around the point estimates of the ICERs was likely. However, the conclusion to this analysis was consistent with the reasoning in section 4.18. The Committee concluded that tocilizumab should be recommended as an option when used in the same way as the TNF‑alpha inhibitors etanercept, adalimumab, infliximab, golimumab and certolizumab pegol recommended in NICE's technology appraisal guidance on adalimumab, etanercept and infliximab; certolizumab pegol; and golimumab. The Committee understood that its recommendation would apply to people whose rheumatoid arthritis has a DAS28 score of greater than 5.1. It also understood that treatment should normally be initiated with the least expensive drug (taking into account administration costs, required dose and product price per dose) and this may need to be varied in individual cases because of differences in the mode of administration and treatment schedules.

The Committee discussed the cost effectiveness of tocilizumab when a person is intolerant to rituximab or for whom rituximab is contraindicated (that is, the DMARD‑IR rituximab intolerant population). The Committee again took the view that, assuming that etanercept and tocilizumab have approximately equal effectiveness (see section 4.19) and cost, it would be reasonable for either to be an option in this position. The Committee noted that the DSU's 2011 analyses broadly corroborated these conclusions. It noted that in this population the ICER from the DSU's 2011 exploratory incremental analysis was £30,100 per QALY gained for a sequence in which etanercept was followed by tocilizumab, and £10,700 per QALY gained for a sequence in which tocilizumab was followed by etanercept (see section 3.50). The Committee concluded that tocilizumab should be recommended as an option for the DMARD-IR rituximab intolerant population. It further concluded that this recommendation should be in line with NICE's technology appraisal guidance on adalimumab, etanercept, infliximab, rituximab and abatacept, specifically the recommendations on disease activity when a second TNF‑alpha inhibitor is recommended for people in whom rituximab is contraindicated or when rituximab is withdrawn because of an adverse event.

Finally, the Committee considered the DSU's 2011 exploratory analysis for the TNF‑IR population. It understood that in this analysis, the costs and QALYs associated with earlier treatment on a TNF‑alpha inhibitor were assumed to be the same and so the analysis comprised 2 sequences containing tocilizumab (1 in which tocilizumab is followed by rituximab, and 1 in which rituximab is followed by tocilizumab) and a baseline treatment sequence of rituximab alone. The Committee noted from this analysis that the treatment strategy that placed tocilizumab before rituximab was dominated by treating with rituximab before tocilizumab (in people who had previously only had a TNF‑alpha inhibitor). The Committee accepted the ICER from this analysis as the most plausible estimate of tocilizumab following rituximab in this population (that is, £18,500 per QALY gained). The Committee also compared this ICER with the manufacturer's estimate of £22,700 per QALY gained. In view of this, the Committee concluded that tocilizumab could be considered an option after an inadequate response to treatment with rituximab but should not be recommended as an alternative to rituximab.

The Committee noted that, in clinical practice and as recommended in NICE's previous technology appraisal guidance on adalimumab, etanercept and infliximab, treatment should normally be initiated with the least expensive drug; this would not necessarily be the same drug in individual cases because of differences in the mode of administration and treatment schedules. The Committee therefore concluded that it was appropriate to recommend tocilizumab as an option following the same considerations as for the drugs recommended as options in NICE's previous technology appraisal guidance on adalimumab, etanercept and infliximab.

The Committee concluded that it was appropriate to recommend tocilizumab plus methotrexate as an option for people whose rheumatoid arthritis has a DAS28 score greater than 5.1 and has responded inadequately to 1 or more previous DMARDs if used as described for TNF inhibitor treatments in NICE's previous technology appraisal guidance on adalimumab, etanercept and infliximab, specifically the recommendations on disease activity and choice of treatment. It concluded that tocilizumab plus methotrexate could be recommended as an option for people whose rheumatoid arthritis has responded inadequately to treatment with DMARDs and a TNF inhibitor and in whom rituximab is contraindicated or who had rituximab withdrawn because of an adverse event. The Committee concluded that, for people whose rheumatoid arthritis has responded inadequately to previous TNF inhibitors, and for whom rituximab is an option, tocilizumab plus methotrexate could not be recommended because although it might be as effective as rituximab, it was more expensive and so could not be considered unless rituximab was contraindicated, not tolerated or had failed. The Committee also concluded that tocilizumab plus methotrexate could be recommended for people whose rheumatoid arthritis has responded inadequately to treatment with 1 or more previous TNF inhibitors and rituximab. It also decided that a recommendation about tocilizumab as monotherapy could not be made because there was not enough evidence of its efficacy as a monotherapy.