Certolizumab pegol for treating rheumatoid arthritis after inadequate response to a TNF-alpha inhibitor

The committee was aware of its conclusion on the efficacy of certolizumab pegol and other bDMARDs (see section 4.6), It queried the base-case incremental cost-effective ratios (ICERs) in the company's submission for the populations of people for whom either rituximab or methotrexate are contraindicated or not tolerated. It would have expected to see similar quality-adjusted life year (QALY) gains to other bDMARDs, but the incremental QALY gain for certolizumab pegol plus methotrexate and certolizumab pegol as monotherapy, were 0.260 for both populations. The committee noted that the company stated there was a lack of comparative evidence in the population who have had TNF‑alpha inhibitors before and therefore had to place assumptions on comparative effectiveness for the comparator bDMARDs. Therefore, the model assumed that the efficacy of adalimumab, etanercept and infliximab were equivalent to golimumab. The efficacies of adalimumab monotherapy and etanercept monotherapy were modelled using the effect size estimates for golimumab compared with certolizumab pegol (both in combination with methotrexate) from the network meta-analysis. The committee noted that these assumptions were not applied to certolizumab pegol.

The committee then considered the ERG's scenario analysis in which it assumed that certolizumab pegol had equal efficacy to etanercept, adalimumab and infliximab (all plus methotrexate) for people for whom rituximab is contraindicated or not tolerated. The ERG also assumed that certolizumab pegol monotherapy had equal efficacy to etanercept and adalimumab monotherapies for people for whom methotrexate was contraindicated or not tolerated. The committee was aware that the etanercept biosimilar had been included in this sequence and agreed that this was appropriate. The committee noted for these equal length sequence analyses, that the ICERs for certolizumab pegol with methotrexate and as monotherapy were dominated; that is, certolizumab pegol plus methotrexate was more expensive but just as effective as the comparator bDMARDs. When the committee looked at the incremental increase in total costs between certolizumab pegol and the etanercept biosimilar it noted that there was very little difference so equivalence among the bDMARDs could be accepted. The committee considered the ICERs that incorporated confidential patient access schemes for abatacept and tocilizumab, the results of which cannot be shown here. Even when these schemes were taken into account, the committee noted that there were similarities in effects and costs and so concluded that certolizumab pegol plus methotrexate, or as monotherapy, can be considered a cost-effective use of NHS resources for people for whom rituximab or methotrexate are contraindicated or not tolerated.

The committee had concerns about the company's approach to evaluating the cost effectiveness of certolizumab pegol plus methotrexate for this population. In particular, it was not persuaded that an intervention treatment sequence containing certolizumab pegol and 6 other treatments should be compared with the same sequence without certolizumab pegol (see section 4.4). The committee was aware from past technology appraisals that using different sequence lengths can increase modelling uncertainties. It heard from the ERG that the company's model may not be appropriate for comparing sequences of different lengths and this point was highlighted in the ERG's exploratory analysis in which the use of the same model type resulted in some counterintuitive results; the clinical benefit (shown by the QALY gain) appeared to be greater if a person had received rituximab plus methotrexate than if a person had received both certolizumab pegol plus methotrexate and rituximab plus methotrexate. In addition the committee also understood that not all possible treatment sequences for this population had been included in the company's analysis. It noted that, to address this, the ERG had included 2 additional sequences in its exploratory analyses, in which certolizumab pegol plus methotrexate was placed after, and instead of, rituximab plus methotrexate. The committee noted that, after consultation, the company had accepted the relevance of the replacement sequence (that is, instead of rituximab plus methotrexate), but did not consider the sequence of certolizumab pegol after rituximab to be within the scope of the appraisal. The committee agreed with this but commented that placing certolizumab pegol plus methotrexate before rituximab plus methotrexate was also unsatisfactory (see section 4.4). It concluded that treatment sequences of the same length are preferable because they are subject to less uncertainty and that its focus should be on the sequence in which certolizumab pegol plus methotrexate replaces rituximab plus methotrexate.

In the revised base-case analysis submitted by the company after consultation, the committee understood that the company had accepted most of the ERG's preferred assumptions, except treatment duration for biological therapies, and the retreatment interval for rituximab. The committee noted that these were key drivers of cost effectiveness. It concluded that each of these should be examined before considering the ICERs for its preferred treatment sequence.

The company provided evidence from 2 studies to support an assumption of equal treatment duration for all biological therapies. A study by Ramiro et al. (2015) provided the evidence for a longer treatment duration with TNF‑alpha inhibitors compared with non‑TNF‑alpha inhibitors, whereas a study by Du Pan et al. (2012) provided evidence for a shorter treatment duration with TNF‑alpha inhibitors compared with non‑TNF‑alpha inhibitors. The committee was not persuaded that this opposing evidence should be interpreted as a basis for equal treatment duration. Also, it was not persuaded that these sources of evidence were methodologically stronger than the source preferred by the ERG (the REFLEX extension trial). In the Ramiro et al. (2015) trial, more people received a TNF‑alpha inhibitor than a non‑TNF‑alpha inhibitor. Also, this study was done in the USA where prescription patterns, reimbursement decisions and patients' comorbidities differ from England. The committee had fewer concerns with the Du Pan et al. (2012) study because it had enrolled more comparable numbers of people on TNF‑alpha and non‑TNF‑alpha inhibitors. Although the committee acknowledged the company's concerns that trial conditions may not represent clinical practice, it regarded the evidence for rituximab, the comparator of interest, to be superior to that for a collection of non‑TNF‑alpha inhibitor technologies. The committee concluded that the data from the extension phase of the REFLEX trial provided the most appropriate source of evidence for treatment duration.

The committee considered the most plausible assumption for the retreatment interval of rituximab in the model. It noted that the summary of product characteristics for rituximab states that the 'need for further courses should be evaluated 24 weeks after the previous course', but did not consider that this was the same as specifying a 6‑month retreatment interval. It also noted that the committee had previously discussed this assumption in NICE's technology appraisal guidance on adalimumab, etanercept, infliximab, rituximab and abatacept for rheumatoid arthritis, and concluded that treatment was unlikely to be as frequent as every 6 months for every person receiving rituximab. It therefore preferred the ERG's value of 10.09 months, which was sourced from the REFLEX trial. The committee considered that it was appropriate to use available trial evidence for rituximab to inform this assumption, and concluded that it was appropriate to use a retreatment interval for rituximab of 10.09 months.

In line with its conclusion about treatment sequences (see section 4.4), the committee considered the ICERs when certolizumab pegol plus methotrexate was placed in a sequence instead of rituximab plus methotrexate. The company's base-case estimate for this comparison was in excess of £130,000 per QALY gained. However, the committee recognised that its preferred assumptions for the treatment duration for bDMARDs and the rituximab retreatment interval were not incorporated in this estimate. When these preferred assumptions were included, certolizumab pegol plus methotrexate was dominated by rituximab plus methotrexate. This analysis did not take into account the confidential patient access scheme discount for tocilizumab, a treatment included in the treatment sequence after rituximab. When the confidential discount for tocilizumab was included, certolizumab pegol plus methotrexate was still dominated. In summary, the committee concluded that certolizumab pegol plus methotrexate could not be considered a cost-effective use of NHS resources when rituximab plus methotrexate is a treatment option. For completeness, the committee looked at the elongated sequence, in which certolizumab pegol plus methotrexate was placed before rituximab plus methotrexate, which the committee had rejected earlier (see section 4.4 and section 4.10). The committee concluded that, with its preferred assumptions this sequence was still dominated and therefore was not a cost-effective use of NHS resources.