Selpercatinib for previously treated RET fusion-positive advanced non-small-cell lung cancer

3.1 The patient and clinical experts explained that the symptoms of advanced non-small-cell lung cancer (NSCLC; including breathlessness, cough, and weight loss) are hard to treat. Typical treatments for RET fusion-positive advanced NSCLC in the NHS are chemotherapy (such as platinum doublet chemotherapy) and immunotherapy (such as pembrolizumab). The clinical expert and the Cancer Drugs Fund clinical lead from NHS England explained that, for previously treated RET fusion-positive NSCLC, docetaxel is the main treatment. But they also explained that some people may also be offered nintedanib with docetaxel, and that these are the only standard treatments for this indication. They explained that use of docetaxel with nintedanib is decreasing because of its limited benefit and increased side effects compared with docetaxel alone. This leaves few options for people with RET fusion-positive advanced NSCLC. Selpercatinib is the first treatment targeted at RET fusion-positive advanced NSCLC and has shown high response rates in some people with this tumour type. The committee concluded that people with RET fusion-positive NSCLC would welcome the introduction of selpercatinib as a treatment option.

3.2 In its original submission, the company provided evidence for a range of comparators based on the NICE scope for this appraisal. Through clinical advice and discussion at technical engagement, the company refined the list of comparators down to docetaxel alone and docetaxel with nintedanib. The ERG suggested that pemetrexed with carboplatin, and platinum doublet chemotherapy remained relevant comparators. The committee discussed atezolizumab as well. The company explained that advice to both itself and ERG had been clear that people would most likely have immunotherapies first. The company said it was advised that people who have immunotherapies first are not then offered them second line, meaning this class of therapy is irrelevant for this indication. The company said it was also advised that pemetrexed with carboplatin and platinum doublet chemotherapy are rarely used second line. The committee concluded that docetaxel was the main comparator and that docetaxel with nintedanib was also an appropriate comparator for people with RET fusion-positive NSCLC.

3.3 The evidence for selpercatinib comes from the LIBRETTO‑001 clinical trial. This is a single-arm, open-label, multicentre phase 1 to 2 trial including people with advanced solid tumours with RET activations. The primary outcome of the trial is objective response rate. Secondary outcomes include progression-free survival (PFS), overall survival (OS) and health-related quality of life. A total of 329 people with RET fusion-positive advanced NSCLC were enrolled, and:

3.4 The trial population included people who had had platinum chemotherapy, some people who had also had immunotherapy, and some people who had also had a multikinase inhibitor (MKI) such as cabozantinib. The ERG said it would have been more appropriate to provide data for people who had only had chemotherapy and people who had only had immunotherapy. The ERG also said people were unlikely to be offered MKIs in the NHS as part of treatment for RET fusion-positive NSCLC. This is because MKIs do not have a UK marketing authorisation for this indication specifically. The clinical expert said the trial population did reflect the NHS population for this indication. The company provided data to show that the trial groups who had and had not had MKI treatment had similar responses. The ERG acknowledged that the data for the IAS MKI‑naive group was similar to the data for the IAS overall. The committee accepted that the LIBRETTO‑001 trial population was generalisable to the NHS population of people with RET fusion-positive advanced NSCLC.

3.5 The marketing authorisation for selpercatinib does not differentiate between people with squamous and non-squamous advanced NSCLC. However, because of the rarity of RET gene fusions in squamous NSCLC, clinical advice, and the very small number of people with squamous NSCLC in the LIBRETTO‑001 trial, the company did not present any evidence on using selpercatinib to treat these tumours. The clinical expert said they might expect some difference in the effectiveness of selpercatinib in treating squamous advanced NSCLC. This is because people with squamous NSCLC may be older, have a higher chance of being smokers and be less fit. However, the clinical expert expected there would still be some level of response. The Cancer Drugs Fund clinical lead said that the NHS would expect to follow the same recommendation for people with squamous advanced NSCLC as for people with non-squamous advanced NSCLC. The committee agreed that the recommendations in this technology appraisal would apply to both squamous and non-squamous advanced NSCLC. This is because of the wording of the marketing authorisation and because the squamous population is so small.

3.6 Because LIBRETTO‑001 was a single-arm trial, an ITC was needed to establish the relative efficacy of selpercatinib. The ERG stated that trials used for the ITC were unlikely to have contained substantial numbers of people with RET fusion-positive advanced NSCLC. This was because the mutation is rare (1% to 2% of people with NSCLC). Also, testing was not done for RET fusion status in these trials, which the company acknowledged as a limitation of the data. The company did its ITC using NMA. This method allows for the relative effects estimated in different studies to be pooled if studies are sufficiently similar. To overcome the limitations noted by the ERG, and to ensure the selected trials were comparable, a suitable cohort of people was needed as a control arm for LIBRETTO‑001. The company simulated a control arm (that is, people having docetaxel with placebo), referred to as the pseudo-control arm, by using data from the REVEL NSCLC randomised controlled trial. The aim was to allow for the LIBRETTO‑001 data to be compared with the other trials in the ITC. The committee noted that the other trial data was not adjusted for RET status. The clinical expert said that the effect of RET fusion on treatment effectiveness for people with advanced NSCLC is unknown. However, the clinical expert thought it may become clear over time as more testing is carried out for this form of lung cancer. The committee accepted that, in the absence of a direct comparator population with RET fusion-positive NSCLC, the NMA trial populations were relevant for the ITC.

3.7 In the company's original submission, the Flatiron clinic-genomic database was used to provide a range of prognostic factors (such as RET fusion status, age, smoking history and cancer histology). This was to adjust the control arm extracted from the REVEL randomised controlled trial to match the LIBRETTO‑001 population. The company said this process had generated a relevant control arm for LIBRETTO‑001, simulating the effect of using docetaxel with placebo to treat RET fusion-positive advanced NSCLC. The ERG said the methods used by the company needed multiple statistical steps, and each step created additional uncertainty. The company changed its approach after technical engagement, and the ERG pointed out that several issues either remained or had been created by using the new propensity score-matching approach. It also pointed out that the additional data provided by the company from a later point of the LIBRETTO‑001 trial had not been used in the NMAs. Doing this would have ensured as much data as possible was informing the ITC. The ERG emphasised that there was still too much uncertainty in the NMAs to make conclusions on the relative efficacy of selpercatinib and the comparators. The committee agreed that simulating the control arm using the company's approach did generate uncertainty for the relative efficacy of selpercatinib. It agreed that, in principle, using a simulated control arm was acceptable. The committee considered that there was not enough evidence to understand the effect of RET fusion status on survival. So, it thought that the relative clinical-effectiveness estimates may have lacked validity. In response to consultation, the company reported new evidence from the scientific literature. It argued this showed that RET fusion status was not prognostic, so the simulated control arm should be generated without adjustment for RET status. The company provided updated survival results for the simulated control arm without adjustment for RET status. It used these results in its NMA. The ERG said that scientific literature identified by the company was not designed to show whether RET fusion status was prognostic, and that the results were not conclusive. However, it thought that the analysis from the company that had shown its results without adjusting the simulated comparator for RET fusion status could be informative. The ERG said that there were still several issues in addition to those with the generation of the revised simulated control arm. These included:

3.8 The company used a partitioned-survival economic model that included 3 health states: progression-free, progressed and death. The committee concluded that the model was generally appropriate and consistent with the models used in other appraisals for NSCLC, including:

3.9 In the company's original submission, extrapolation of PFS and OS for selpercatinib came from LIBRETTO‑001 and the NMA (see section 3.6). The different extrapolation distributions were ranked using statistical methods, and also considered by clinical advisers to the company. The company based its conclusions for the selpercatinib arm on the Spline/Knot1 OS extrapolation. This was because its clinical advisers believed this extrapolation fitted most closely to their expectation of clinical reality, even though it was not objectively the best fit. The committee noted that clinical expert opinions drew little on experience of the rare RET fusion-positive form of NSCLC. It also noted that there is little long-term experience of using selpercatinib in the NHS. The ERG said that selection based on clinical advice rather than statistical tests, was open to bias. The direction and magnitude of any bias was not clear from the data. The ERG did not select a preferred alternative base-case extrapolation function because it thought the data and NMAs were too uncertain to make this possible. It noted that the Gompertz alternative extrapolation would match the clinical evidence most closely, and would be just as appropriate a selection of extrapolation as Spline/Knot1. However, it noted that it resulted in substantially different cost-effectiveness results. A different approach was used for PFS with selpercatinib, in that the stratified Gompertz distribution was used to fit the data. The committee discussed the differences in the extrapolated OS estimations presented and that this was, in part, caused by the short follow up of the LIBRETTO‑001 trial. The ERG said that, based on its inspection of extrapolations fit to the LIBRETTO‑001 data, OS for selpercatinib appeared to have been overestimated by the company. The clinical expert and Cancer Drugs Fund clinical lead supported this view. The committee acknowledged the uncertainty in PFS and OS estimates, particularly in the wide range of extrapolations for selpercatinib. In response to consultation, the company provided updated survival extrapolations for selpercatinib based on the results from its updated NMA and presented the updated results. It was again possible to fit a wide range of extrapolations to the data. The company repeated the process used in its original submission for choosing extrapolation curves. It selected the stratified Gompertz distribution for PFS, and the unstratified Gompertz for OS. The ERG commented that, based on the selpercatinib Kaplan–Meier curve of the data in LIBRETTO‑001, the extrapolated OS for selpercatinib appeared to have been overestimated. The ERG reiterated that several other distributions also fitted the data. The Cancer Drugs Fund clinical lead and clinical expert commented that the 5‑year survival estimates appeared to be similar to those seen in clinical practice for other targeted lung cancer therapies. They explained that the predicted survival may have been high, for example, the 38.8% survival predicted at 5 years using the company's Gompertz extrapolation. But they thought that this was plausible based on experience with other targeted treatments. The company acknowledged that its revised PFS and OS extrapolations may still have overestimated the effect of selpercatinib. However, it commented that the uncertainty could have been reduced with more mature data from LIBRETTO‑001. The committee concluded that there was still uncertainty about long-term survival with selpercatinib and that more mature data from LIBRETTO‑001 would provide more robust long-term survival estimates. However, it agreed that, based on the opinions of the clinical expert and the Cancer Drugs Fund clinical lead, the survival benefits from selpercatinib at 5 years were not unreasonable. So, the committee concluded that it was appropriate to consider the company's survival estimates for selpercatinib in its decision making.

3.10 In the company's original submission, the estimates of PFS and OS for docetaxel came from the simulated control arm and NMA (see section 3.6 and section 3.7). The ERG considered that the extrapolation of survival in the control arm was likely to have been longer than expected in clinical practice. The clinical expert said they would have expected OS to be about 9 to 10 months for docetaxel, rather than the higher values seen in the survival extrapolations in the original submission. They explained that it is feasible that people with RET fusion-positive advanced NSCLC could have greater OS than people with other forms of advanced NSCLC. This is particularly because they tend to be younger and non-smokers, which might explain some of the higher-than-expected OS in the docetaxel arm. However, they noted that there was no evidence to support this. The company explained that the increase in OS from 9 months in the simulated control arm was because of the adjustment processes for RET fusion status used in its generation. The committee agreed that the survival estimates for the control arm were implausibly long, and that this would mean the conclusions based on the model were not robust. In response to consultation, the company provided revised survival extrapolations for the simulated control arm without adjustment for RET fusion status. The company considered that the updated survival extrapolations fitted more closely to the clinical expert's estimates (that is, 9 to 10 months). The ERG reiterated its opinion that there was still a lot of uncertainty. It did not think that simply removing the adjustment for RET fusion status would have accounted for all the uncertainty (see section 3.7). The ERG thought that the company had succeeded in reducing survival estimates for the simulated control arm, which had been considered to be too high. However, it pointed out that, because of limited data, the long-term survival for this group was still uncertain. The clinical expert considered the revised survival extrapolations to be more plausible. This was because of the low number of people in the stimulated control arm who were expected to be alive at 5 years and beyond. The committee agreed that the company's revised survival extrapolations for the simulated control arm were clinically plausible and appropriate for decision making. However, it also agreed that the survival estimates were still uncertain. This was because of the lack of evidence on whether RET status is a prognostic factor, so whether it should have been adjusted for in the stimulated control arm.

3.11 The original company model used PFS to calculate the cost of selpercatinib. The ERG said that using an extrapolation based on the TTD data in LIBRETTO‑001 would be more accurate. The company subsequently used an estimate for TTD in its updated model. The ERG preferred to incorporate a parametric extrapolation for TTD into the original model. The company stated that the ERG's approach overestimated TTD, and therefore costs, because the data was immature. The clinical expert said that the costs of selpercatinib would be higher if estimated using TTD rather than PFS. The reason is that it is common for a treatment to be continued even if there is disease progression because progression does not mean there is no benefit from the treatment. This could be because:

3.12 The company did not include costs for genetic testing for RET fusions into its original cost-effectiveness model. This was because it expects such testing to be done routinely within the NHS. The Cancer Drugs Fund clinical lead confirmed that testing for RET fusions is available in the NHS as a fluorescent in-situ hybridisation test. However, access to this test is not routine or part of normal screening at the NHS Genomic Medicine Service. The clinical expert said that next-generation sequencing screening panels would be adapted to include testing for RET fusions when possible. However, at the time of this appraisal for selpercatinib, this was not considered routine. Therefore, NHS England provided a suitable cost per test to the company, and the company included this in its economic model. The committee noted the response to consultation from a commentator that the cost of testing should have been shown as a percentage of the overall testing costs. The commentator said that this percentage should have represented the additional costs compared with the testing costs without testing for RET fusion status. The committee agreed that incorporating the cost of genetic testing for RET fusions was appropriate.

3.13 The ERG pointed out that the company's approach to utility values used in the model was inconsistent. In general, the company took its utility values from NICE's technology appraisal guidance on nivolumab for advanced non-squamous non-small-cell lung cancer after chemotherapy. However, it used the utility value for PD of 0.688 from the company's base-case analysis in the nivolumab appraisal, rather than that appraisal committee's preferred value for PD of 0.569. The ERG was concerned that 0.688 was high, preferring 0.569. For the appraisal of selpercatinib, the company collected health-related quality-of-life data in the LIBRETTO‑001 trial. However, it used the European Organisation for Research and Treatment of Cancer (EORTC) QLQ‑C30 questionnaire rather than EQ‑5D to collect this data. The company followed a method reported in the literature to map EORTC to EQ‑5D, and the calculated PD value was higher than 0.688. So, the company decided to use the midpoint between 0.569 and 0.688 in its model, which was 0.628. The ERG said this approach was arbitrary and maintained its view that the utility value of 0.569 from the nivolumab appraisal was appropriate for this population. The clinical expert stated that people with RET fusion-positive advanced NSCLC tend to be younger and have never smoked. So, they thought it was feasible they might have generally higher utility values than people with other forms of lung cancer. The committee decided that the PD value of 0.628 used by the company in the revised model was acceptable for decision making in absence of more robust data.

3.14 In response to consultation, the company presented a revised base case, which included an updated commercial arrangement for selpercatinib. The pairwise ICER was:

3.15 The committee noted that, unlike docetaxel, selpercatinib is an oral drug, and it specifically targets RET fusion-positive NSCLC. It agreed that selpercatinib would be beneficial. The committee considered that the model structure should have been able to capture the benefits and costs of selpercatinib in terms of health-related quality of life and QALYs gained. It did not think that it had not been presented with evidence of any additional benefits that were not captured in the measurement of QALYs.

3.16 The committee considered the advice about life-extending treatments for people with a short life expectancy in NICE's guide to the methods of technology appraisal. In the company's original submission, the base-case estimate for the median OS for people offered docetaxel was less than 24 months, and an estimate of the mean was not provided. However, the company explained that it thought this to be an overestimate compared with clinical expert opinion, which was 9 to 10 months. The ERG's estimates for OS for people offered docetaxel with or without nintedanib were higher than those of the company, that is, above 24 months. The company thought that the ERG's extrapolations for survival were overestimates. The committee noted the comments from the clinical expert. It considered that the expected survival of people with RET fusion-positive advanced NSCLC who were not offered selpercatinib might be much less than 24 months in practice. In the company's response to consultation, the modelled estimates for OS with docetaxel (without adjustment for RET status in the simulated control arm) were closer to the survival estimates expected by the clinical expert. The ERG reiterated that there was a lack of data to show that removing the adjustment for RET status was the correct approach, so there was still uncertainty. The committee accepted that there was uncertainty in how the simulated control arm was generated. But it agreed that the updated OS results for docetaxel were plausible and concluded that the short life expectancy criterion was met.

3.17 In its original base case, the company estimated that selpercatinib would extend life expectancy by much more than 3 months (the company's modelled estimates are confidential and cannot be presented here). The ERG thought that this was feasible according to the data, but highly uncertain because of the difference between clinical expert opinion and company estimates. The committee recalled its concerns about the uncertainty in the OS estimates generated using the company's original model. It concluded that the company's estimate of extending life expectancy was not reliable and so the life extension criterion was not met. This was because of its concerns with the original NMA and the lack of robust results from the model. In response to consultation, the company presented updated OS estimates for selpercatinib and the simulated control arm based on the generation of the revised control arm, and updated NMAs. A wide range of extrapolations could be made from the results, so the committee agreed that there was uncertainty about the extent of the additional survival gain from selpercatinib compared with the simulated control arm. However, it concluded that it was likely that people having selpercatinib would benefit from an extension to life of more than 3 months.

3.18 The committee was aware that the evidence base will necessarily be weaker for some rare indications such as RET fusion-positive advanced NSCLC because of the low number of people with the condition. The committee recalled that there are no targeted treatments currently available for RET fusion-positive advanced NSCLC, as discussed in section 3.1. It noted the clinical- and cost-effectiveness evidence was highly uncertain because of the immaturity of the data from the LIBRETTO‑001 trial. It also noted that there was still uncertainty about the ITC using NMAs based on the simulated control arm. Selpercatinib met NICE's end of life criteria. However, the committee's preferred ICER was well above the range normally considered a cost-effective use of NHS resources, even considering the end of life criteria. Therefore, it could not recommend selpercatinib for routine use for previously treated RET fusion-positive advanced NSCLC.

3.19 Having concluded that selpercatinib could not be recommended for routine use, the committee then considered whether it could be recommended for treating RET fusion-positive advanced NSCLC within the Cancer Drugs Fund. The committee discussed the arrangements for the Cancer Drugs Fund agreed by NICE and NHS England in 2016, noting NICE's Cancer Drugs Fund methods guide (addendum):