3 Committee discussion
The appraisal committee considered evidence submitted by Celgene, a review of this submission by the evidence review group (ERG), and the technical report developed through engagement with stakeholders. See the committee papers for full details of the evidence.
After technical engagement, there were remaining areas of uncertainty associated with the analyses presented. The committee took these into account in its decision making. It discussed the following issues (issues 1 to 6), which were outstanding after the technical engagement stage.
3.1 The clinical and patient experts noted that people with previously treated follicular lymphoma have limited treatment options. The choice of treatment for previously treated follicular lymphoma depends on individual circumstances and takes into account previous chemotherapy, the age and fitness of the patient, plus clinician and patient preferences. The patient experts explained that chemotherapy has unpleasant side effects and any treatment that avoided chemotherapy would be welcomed. The committee concluded that lenalidomide with rituximab would be welcomed as a new treatment option for people with previously treated follicular lymphoma.
3.2 The clinical experts noted that rituximab monotherapy, which is the comparator used in the AUGMENT trial used in this appraisal (see section 3.3), would rarely be given to people whose disease had relapsed. The company included obinutuzumab with bendamustine as a comparator, but this combination is only recommended for use in the Cancer Drugs Fund. The committee did not consider obinutuzumab with bendamustine to be used in routine commissioning and did not consider it a relevant comparator. The committee understood that rituximab with either CHOP (cyclophosphamide, doxorubicin, vincristine and prednisolone) or CVP (cyclophosphamide, vincristine and prednisolone) is used to treat follicular lymphoma in people who have had at least 1 previous treatment. The clinical experts noted that the company's proposed treatment pathway originally separated treatments for previously treated follicular lymphoma depending on whether the disease was refractory to rituximab or not. The clinical experts considered that being refractory to rituximab or not was not a clinically appropriate way to separate treatment choices, and that time until relapse after initial chemo-immunotherapy may be more relevant for determining treatment choices in clinical practice. The committee agreed that the most appropriate comparators for this appraisal were rituximab with chemotherapy (CHOP or CVP).
Clinical evidence for lenalidomide with rituximab and rituximab with chemotherapy is compared using a matching-adjusted indirect comparison
3.3 The evidence for lenalidomide with rituximab came from AUGMENT, a phase 3 multicentre randomised controlled trial that used rituximab monotherapy as the comparator. In the absence of direct comparative evidence of lenalidomide plus rituximab compared with rituximab plus CHOP (R‑CHOP) or rituximab plus CVP (R‑CVP), the company provided a matching-adjusted indirect comparison to compare both treatments. Data for R‑CHOP and R‑CVP are from either Van Oers et al. (2006) or the Haematological Malignancy Research Network (HMRN) registry. The committee agreed that using the Van Oers data was not appropriate, because patients were not from the UK and had not had previous treatment with rituximab. The committee concluded that the HMRN registry data were most suitable for R‑CHOP and R‑CVP because the data were from the UK and include people who had had previous treatment with rituximab.
3.4 The company combined the R‑CHOP and R‑CVP populations for the matching-adjusted indirect comparison to increase the sample size of the comparator. The matching-adjusted indirect comparison showed an improvement in progression-free survival for lenalidomide with rituximab compared with R‑CHOP and R‑CVP (exact data are confidential and cannot be reported here). It also showed an improvement in overall survival for lenalidomide with rituximab compared with R‑CHOP and R‑CVP (exact data are confidential and cannot be reported here). Median progression-free survival and overall survival could not be estimated because the follow-up data were immature (not yet complete). When assessing the validity of combining R‑CHOP and R‑CVP, the committee understood that observed data for R‑CHOP in the HMRN registry for overall survival, progression-free survival and time to next anti-lymphoma treatment appeared similar to R‑CVP. The committee also understood that Cox proportional hazards model analyses of overall survival, progression-free survival and time to next anti-lymphoma treatment showed no statistically significant difference in outcomes between R‑CHOP and R‑CVP. The clinical experts noted that, in untreated follicular lymphoma, R‑CHOP and R‑CVP are not clinically equivalent. They considered that R‑CHOP has a longer time to treatment failure than R‑CVP (despite similar response rates) and progression-free survival is longer with R‑CHOP. They also noted that R‑CHOP is given to younger, fitter patients who can tolerate the additional chemotherapy component (doxorubicin) in CHOP, while R‑CVP is given to older, less fit patients. The clinical experts acknowledged that there was no evidence for R‑CHOP and R‑CVP in previously treated follicular lymphoma. They accepted that, without any other data sources, it may be appropriate to assume R‑CHOP is clinically equivalent to R‑CVP. The committee concluded that it was appropriate to assume R‑CHOP is clinically equivalent to R‑CVP in the economic model.
3.5 The ERG noted that the matching-adjusted indirect comparison did not account for all potentially relevant matching criteria. Potentially relevant matching criteria included the diameter of the largest node, haemoglobin levels, duration since last treatment, high lactate dehydrogenase levels, and Follicular Lymphoma International Prognostic Index (FLIPI) risk group status. The committee understood that the most important criteria identified by clinical experts were included in the matching-adjusted indirect comparison, and that other potentially relevant matching criteria were not collected by HMRN. The committee agreed that, given the data limitations, the matching-adjusted indirect comparison could not be improved further.
3.6 The company used a partitioned survival model to determine the difference in overall survival and progression-free survival for lenalidomide with rituximab and R‑CHOP and R‑CVP. The ERG said that a state transition model would have helped to assess the validity of the extrapolations in overall survival and progression-free survival in the partitioned survival model. Following technical engagement, the company provided a state transition model for lenalidomide with rituximab compared with rituximab monotherapy (not R‑CHOP and R‑CVP). The committee did not consider this appropriate because rituximab monotherapy is not a relevant treatment for patients in the NHS. The committee also noted that a state transition model for lenalidomide with rituximab compared with R‑CHOP and R‑CVP would have its own uncertainty because of the limitations in the data. The AUGMENT data were immature, with a maximum follow up of 3.9 years, and the R‑CHOP and R‑CVP data were from a small effective sample. The committee agreed that accurately modelling transitions between intermediate health states would be difficult, and providing a state transition model would not help validate the partitioned survival model. The committee concluded that a partitioned survival model was appropriate, and that a state transition model was not needed on this occasion.
3.8 The clinical experts said that overall-survival estimates for lenalidomide with rituximab will be better than overall-survival estimates for R‑CHOP and R‑CVP for several years once treatment begins. However, these estimates will gradually decline over time and eventually become the same as overall-survival estimates for R‑CHOP and R‑CVP. The clinical experts noted that it is uncertain how long the overall-survival benefit for lenalidomide with rituximab will last, but said that it will likely start to reduce 5 to 10 years after treatment starts. The company and the ERG's estimates of treatment effect duration was 5 years in its base cases. The committee noted that this was the most conservative estimate, given the clinical expert opinion of 5 to 10 years. The committee concluded that a treatment effect duration of 5 years was appropriate, with no evidence to the contrary.
3.9 NICE's Decision Support Unit technical support document 14 states that, if parametric models are fitted separately to individual treatment arms, it is advisable to use the same type of distribution for extrapolating both arms. For overall survival, the company selected the exponential distribution for both arms to best reflect the average 20‑year prognosis for follicular lymphoma when treated with R‑CHOP and R‑CVP. The clinical experts noted that the Weibull distribution may have also been appropriate for both arms, and the committee noted that it was a slightly better statistical fit than the exponential distribution. The committee also noted that the Weibull distribution produced a higher cost-effectiveness estimate in the probabilistic sensitivity analysis results than it did in the deterministic sensitivity analysis results. The committee was aware that this was caused by lenalidomide with rituximab having lower quality-adjusted life years (QALYs) than R‑CHOP and R‑CVP in the probabilistic sensitivity analysis results compared with the deterministic sensitivity analysis results. More specifically, lenalidomide with rituximab had lower quality of life or extension of life than R‑CHOP and R‑CVP in 267 of the 1,000 iterations in the probabilistic sensitivity analysis. The clinical experts said that this was not clinically plausible. The committee agreed that the reduction in QALYs for lenalidomide with rituximab in the probabilistic sensitivity analysis results (when using a Weibull distribution for overall survival) was unusual. The committee noted that this might have been due to an error in the model or an unstable covariance matrix. Because of this, the committee did not consider the output of the probabilistic sensitivity analysis to be robust, and did not think it appropriate to use the Weibull distribution. The committee considered that the exponential distribution had good statistical fit for the observed Kaplan–Meier data and concluded that the exponential distribution was appropriate for extrapolating overall survival.
Different methods are needed to extrapolate progression-free survival for lenalidomide with rituximab and R-CHOP and R-CVP
3.10 The standard approach for extrapolating survival curves is to fit a parametric distribution to the available data, which is then used to estimate survival at any point in time. This method was used to fit the Weibull function for R‑CHOP and R‑CVP. However, the committee noted that all the standard parametric extrapolations for lenalidomide with rituximab estimated worse progression-free survival outcomes compared with R‑CHOP and R‑CVP. The clinical experts said that it was clinically implausible for the average progression-free survival of lenalidomide with rituximab to be worse than the average progression-free survival of R‑CHOP and R‑CVP. The committee understood that the relatively worse progression-free survival estimates for lenalidomide with rituximab may have been a result of the immature follow-up data from the AUGMENT trial (3.9‑year follow up), compared with the longer follow up for R‑CHOP and R‑CVP (11.6 years). To avoid the implausible crossing of progression-free survival curves, the company used the observed AUGMENT data, and then, in the absence of further trial data, the company fitted a Weibull distribution to the end of these data. The ERG noted that this approach generated a larger progression-free survival benefit for lenalidomide with rituximab than applying the standard parametric distributions. Additionally, the ERG was concerned that, by extrapolating from the end of the observed AUGMENT data when there were fewer patients to inform the extrapolation, the mean progression-free survival benefit for lenalidomide with rituximab may have been further overestimated. The committee agreed that using the observed AUGMENT data with a Weibull distribution extrapolation was the most appropriate method for estimating progression-free survival for lenalidomide with rituximab. This was because this method takes account of the clinical plausibility of a progression-free survival benefit for lenalidomide with rituximab. The committee concluded that the AUGMENT data should have been extrapolated from the midpoint, rather than from the end, because this would have reduced the uncertainty in the extrapolation by using a relatively larger sample size.
3.11 NICE's guide to the methods of technology appraisal says that, above a most plausible incremental cost-effectiveness ratio (ICER) of £20,000 per QALY gained, judgements about the acceptability of a technology as an effective use of NHS resources will take into account the degree of certainty around the ICER. Therefore, because of the uncertainty in the equivalence of R‑CHOP and R‑CVP (see section 3.4), matching-adjusted indirect comparison (see section 3.5) and treatment effect duration (see section 3.8), the committee agreed that an acceptable ICER would be around the lower end of the £20,000 to £30,000 per QALY gained range. The company's deterministic base case showed that the ICER for lenalidomide with rituximab compared with R‑CVP was £20,156 per QALY gained (including the patient access scheme for lenalidomide). The ERG presented 6 analyses, each using a different parametric distribution to extrapolate overall survival. The ERG's analyses also included the confidential commercial arrangement for obinutuzumab and biosimilar rituximab. The ICERs were within or below the range that NICE usually considers an acceptable use of NHS resources (the exact ICERs are confidential and cannot be reported here). The committee agreed that an extrapolation with an exponential distribution was appropriate for overall survival in both lenalidomide with rituximab and R‑CHOP and R‑CVP. The committee also agreed that AUGMENT data with a Weibull distribution (extrapolated from AUGMENT data midpoint) was the most appropriate for estimating progression-free survival for lenalidomide with rituximab, and that a standard parametric Weibull distribution should have been used to extrapolate R‑CHOP and R‑CVP (see sections 3.9 and 3.10). These assumptions generated a most plausible ICER that was below the range that NICE usually considers an acceptable use of NHS resources (the exact ICER is confidential and cannot be reported here). The committee therefore recommended lenalidomide with rituximab as an option for previously treated follicular lymphoma (grades 1 to 3A).