ID3863 lifileucel - To PM for consultation (online commenting) [noCON].docx

Melanoma is a cancer of the skin. Symptoms of unresectable or metastatic melanoma will depend on where the cancer has spread to. But general symptoms may include weight loss, muscle pain, confusion, memory problems, mood or personality changes, feeling sick and extreme fatigue. The patient experts explained that unresectable or metastatic melanoma has a significant effect on quality of life. They also described how people with melanoma live with the fear of progression, which has a major impact on mental wellbeing. The clinical experts explained that, in many people, the condition progresses after all available systemic treatments have been used. But these people are still fit enough to have further treatment. Without another treatment option, these people will eventually experience disease progression and death. The committee recognised the need for an effective treatment for previously treated unresectable or metastatic melanoma. It concluded that unresectable or metastatic melanoma has a significant impact on the quality of life for people with the condition.

First-line treatment options for unresectable or metastatic melanoma include PD‑1 blocking antibodies (nivolumab, alone, with relatlimab or with ipilimumab, or, pembrolizumab alone). For people whose cancer is BRAF V600 mutation positive, a BRAF inhibitor, with or without a MEK inhibitor, is also a treatment option, either first or second line. The clinical experts explained that treatment depends on tumour characteristics and patient and healthcare professional choice. But they added that most people would have a PD‑1 blocking antibody, either alone or in combination, as first-line treatment. The committee agreed that it was in line with the treatment pathway for unresectable or metastatic melanoma in NHS clinical practice to position lifileucel:

• at second line, after a PD‑1 blocking antibody

• at third line after a BRAF inhibitor, with or without a MEK inhibitor, if the cancer is BRAF V600 mutation positive.

Lifileucel is a one-off single-dose tumour-infiltrating lymphocyte (TIL) cell therapy. TILs are immune cells that naturally target and destroy tumour cells. The company explained that lifileucel is made using cells from the tumour of the person that will have the treatment. After the cells have been collected, they are sent to a manufacturing site where they are multiplied into billions of cells. The product is then sent back to the hospital as a cryopreserved product. The company explained that lifileucel is administered in a 3‑step process for which people need to be admitted to hospital. Before lifileucel is infused, people have lymphodepleting chemotherapy to prepare the body for the TIL cell therapy and to eliminate any suppressive cells. After lifileucel is infused, people have high-dose interleukin‑2 to improve the chances of the TIL cells surviving in the body.

The company submission compared lifileucel with ipilimumab, chemotherapy and best supportive care (BSC), which aligned with the NICE final scope. The clinical experts were supportive of the choice of comparators. They explained that, at second line (or third line if the cancer is BRAF V600 mutation positive), most people will have BSC. But a smaller proportion of people will not have had ipilimumab at earlier lines of treatment (with a PD‑1 blocking antibody), so would instead have ipilimumab. The clinical experts explained that chemotherapy is not often offered because, compared with BSC, it is not thought to provide any significant advantages. But they explained that some people choose to have chemotherapy. The committee agreed that ipilimumab, chemotherapy and BSC are appropriate comparators for people with previously treated unresectable or metastatic melanoma.

The clinical evidence for lifileucel came from C‑144‑01, an ongoing phase 2 open-label single-arm trial. It recruited people with unresectable or metastatic melanoma who had had 1 or more systemic treatments, including a PD‑1-blocking antibody and, if BRAF V600 mutation positive, a BRAF inhibitor with or without a MEK inhibitor. The company provided data from the June 2023 data cut, in which the median follow up for key efficacy outcomes was 27.6 months. The company explained that a later data cut was available, which was not presented in its submission. The study had 4 cohorts:

• cohort 1: people who had non-cryopreserved lifileucel

• cohort 2 and 4: people who had cryopreserved lifileucel

• cohort 3: people who had retreatment with lifileucel.
  
  The company said that only data from cohorts 2 and 4 should be considered because non-cryopreserved lifileucel is not in clinical use and lifileucel's license will not include retreatment. It explained that cohorts 2 and 4 had the same eligibility criteria, and used the same lifileucel manufacturing process and treatment regimen, but enrolled people over different timeframes. The company used data from cohorts 2 and 4 to generate the pooled analysis set, which was used in its economic model. The EAG noted that there were differences in the baseline characteristics between cohorts 2 and 4. Cohort 4 included a higher proportion of people with stage 4 melanoma, more than 3 lesions, elevated lactate dehydrogenase (LDH) levels, and liver or brain metastasis. It also said that people in cohort 4 had nearly twice the cumulative duration of prior anti‑PD‑1 treatment. The EAG explained that treating the melanoma in cohort 4 might have been more challenging than in cohort 2. But, despite the differences between the cohorts, the EAG thought that pooling the data was reasonable. The company said that the population in C‑144‑01 may have had more treatment before lifileucel than the population who would have lifileucel in NHS clinical practice. The clinical experts thought that, despite the possible differences in the level of pretreatment, the pooled population in C‑144‑01 would be generalisable to the population that would have lifileucel in the NHS.
  
  The committee agreed that cohorts 2 and 4 were the relevant cohorts from C‑144‑01 to consider, so pooling the data from these may be reasonable. It also noted that pooling the cohorts resulted in a larger sample size, which is a benefit, especially because the data is used in indirect treatment comparisons (ITCs; see section 3.6). But the committee raised concerns that the median overall survival (OS) was substantially better in cohort 2 than cohort 4. This suggests that the presence of negative prognostic factors in cohort 4 had a substantial impact on the effectiveness of lifileucel. It was also aware that cohort 4 was the primary evidence used as the basis for lifileucel's Food and Drug Administration marketing authorisation application. The committee noted that the differences between cohorts resulted in uncertainty when interpreting the pooled analysis set. So, it requested the following further analyses:

• an explanation for why the baseline characteristics were different between the 2 cohorts when the eligibility criteria were the same

• a scenario analysis using only the latest data cut from cohort 4 in the ITCs and economic model, to help understand the impact of uncertainty when using the pooled analysis set

• further evidence on the generalisability of the pooled analysis set to the population who would have lifileucel in NHS clinical practice, considering the difference in baseline characteristics between the 2 cohorts

• outcomes for the pooled analysis set using the latest data cut.

Because there were no trials directly comparing lifileucel with comparators, ITCs were needed for this comparison. To develop comparative clinical-effectiveness data, the company did:

• a simulated treatment comparison (STC) to compare lifileucel with ipilimumab (see section 3.7)

• a naive comparison to compare lifileucel with chemotherapy (see section 3.8).

To estimate the cost effectiveness of lifileucel, the company presented a partitioned survival model with 3 health states: progression free, progressed disease and death. The model used a lifetime time horizon. It also included a decision tree for the lifileucel arm. This governed whether, after tumour harvest, people:

• stopped treatment before the infusion

• died before the infusion

• had an infusion that was out-of-specification

• had an infusion that met lifileucel's summary of product characteristics specifications.
  
  Only people in the last group were assigned the costs and outcomes associated with lifileucel. People who stopped before an infusion or had an infusion out-of-specification were assigned part of the lifileucel treatment costs. But outcomes were derived from an estimated market-share-based weighting of ipilimumab, chemotherapy and BSC outcomes. The committee concluded that the company's model structure, and the assumptions for people who stopped treatment before an infusion or had an out-of-specification infusion, were suitable for decision making.

The company used a mixture cure model (MCM) to model progression-free survival (PFS) and OS for lifileucel. The MCM assumed that there were 2 groups:

• people whose melanoma is cured immediately after having lifileucel and who will have a long-term survival similar to people without melanoma of the same age and sex

• people whose melanoma is not cured, and who have a continued risk of progression and death because of it.
  
  The MCMs estimated a cure fraction that represented the proportion of people who were in the first group. The company said that using an MCM was supported by an observed plateau in C‑144‑01's PFS and OS Kaplan–Meier curves. The company's clinical experts said that people whose melanoma has not progressed after 3 years are assumed to have a long-term survival similar to the general population. The company selected an exponential MCM for OS. It said that the exponential MCM provided the best statistical fit and was clinically intuitive because it assumed that people whose melanoma is uncured die at a constant rate. The company selected a log-normal MCM for PFS, informed by clinical expert opinion.
  
  The EAG thought that using MCMs to model PFS and OS for the lifileucel group was appropriate. But the EAG preferred to use a log-logistic MCM for both OS and PFS. It explained that the log-logistic MCM for OS produced a cure fraction that more closely aligned with the cure fraction suggested by the company's clinical experts. The EAG noted that the company's preferred MCMs resulted in a larger difference in the cure fractions for PFS and OS compared with its preferred MCMs. The EAG explained that you would only expect the cure fraction for OS to be higher than the cure fraction for PFS if there were effective treatments after progression. It noted that this was not the case for people with previously treated unresectable or metastatic melanoma. But the clinical experts suggested that a difference in OS and PFS cure fractions was plausible. The clinical experts explained that a difference in OS and PFS cure fractions has been seen in clinical trials for other treatments. They explained that progression could include localised progression that could be treated in a way that does not affect the potential for a cure. They further explained that some melanoma treated with immunotherapies has been known to progress but then respond, which is known as pseudoprogression. The company said that data on PFS at subsequent treatment lines (PFS2) and data from a more recent C‑144‑01 data cut would support a higher cure fraction for OS than PFS. The committee agreed that further data may help resolve this uncertainty. It requested that the company provide additional analysis from C‑144‑01, outlining how this data justifies a higher cure fraction for OS than PFS, including:

• the most recent data cut

• PFS2 data

• the proportion having surgical resection after lifileucel.

For its comparison of lifileucel with ipilimumab, the company applied the ratio of adjusted and unadjusted hazard ratios from the STC (see section 3.7) to the lifileucel MCMs (see section 3.10). It did this to estimate absolute OS and PFS outcomes for people who has lifileucel in the da Silva et al. (2021) population. The company used standard parametric models fitted directly to data from da Silva et al. to estimate absolute outcomes for people who had ipilimumab. It said that the short follow-up time in da Silva et al. meant that the standard parametric modelling approach was appropriate for outcomes in the ipilimumab arm, but that an MCM approach was not. But the company's clinical experts suggested that some people who have ipilimumab survive in the long term and their melanoma is considered to be cured. The company accounted for this by assuming that people who had ipilimumab and were progression free after 3 years had long-term survival similar to people without melanoma of the same age and sex.

The EAG explained that the company's approach to estimating absolute outcomes for people who had lifileucel uplifted the OS and PFS cure fractions compared with the MCM. It noted that the implied cure fractions for lifileucel after adjustment far exceeded the cure fractions suggested by the company's clinical experts. The EAG preferred to use the unadjusted MCMs (see section 3.10) to estimate absolute outcomes for people who had lifileucel. The EAG applied the inverse of the adjusted hazard ratios from the STC to the lifileucel MCMs to estimate absolute outcomes for people who had ipilimumab. The EAG noted that the 3‑year OS cure assumption applied in the company's model resulted in a very small proportion of people surviving in the long term with ipilimumab. The exact percentage is considered confidential by the company, so cannot be reported here. The EAG explained that its preferred approach resulted in a plateau in model-predicted ipilimumab OS and PFS. This suggested long-term survival for a small, yet slightly higher, proportion of people.

The committee noted that the company used unadjusted MCMs to estimate lifileucel's treatment effect when compared with chemotherapy and BSC. So, it thought that it was more appropriate to also use unadjusted MCMs for the comparison with ipilimumab. This meant that the absolute outcomes for lifileucel were the same for all of the comparisons. It also noted that the company did not provide any evidence to support the uplifted cure fractions associated with its preferred approach. These were higher than both the cure fractions suggested by its clinical experts and those estimated from the unadjusted models. The committee noted that it was unclear whether C‑144‑01 or the da Silva et al. population more closely reflected the population who would have lifileucel in NHS practice (see section 3.5). So, it was also unclear whether using the unadjusted MCMs or adjustments from the STC based on the da Silva et al. population estimated the efficacy of lifileucel in a more relevant population. The committee concluded that using MCMs to estimate the long-term survival outcomes for lifileucel and ipilimumab was appropriate. It also concluded that the EAG's approach to modelling relative and absolute treatment effects for lifileucel compared with ipilimumab was more appropriate than the company's approach. But the committee acknowledged the most appropriate approach may change when data with more complete follow up (see section 3.5) is considered.

The company fitted standard parametric survival curves to data from the chemotherapy arm in Mangin et al. (2021; see section 3.8) to estimate long-term PFS and OS for people who had chemotherapy in the model. The EAG noted that all the standard parametric models for PFS provided a poor visual fit to the observed data. It further noted that several of the standard parametric survival models did not reflect the observed hazards. The EAG thought that more flexible parametric models, such as spline models, may better reflect the observed hazards. The company said that its systematic literature review did not identify any suitable and up-to-date data sources for BSC. So, the company chose to use data from people who had chemotherapy in Mangin et al. as a proxy. It modelled absolute OS and PFS outcomes for people who had BSC by applying a hazard ratio of 2.0 to the modelled outcomes for chemotherapy. The company explained that it chose a hazard ratio of 2.0 because its clinical experts thought that BSC survival outcomes are about 50% worse than the chemotherapy outcomes.

The EAG preferred to assume that outcomes for BSC and chemotherapy are equivalent because this is what its clinical experts said they would expect. It further explained that the minutes from the company's advisory board meeting stated that some of the company's clinical experts thought that chemotherapy provides no advantage over BSC. The clinical experts at the meeting explained that BSC is thought to be as effective as chemotherapy, which is only offered when someone's preference is to continue having anticancer therapy. The committee recalled that estimates based on Mangin et al. were likely to underestimate the effect of chemotherapy seen in the population having treatment in clinical practice (see section 3.8). So, using this data was likely to result in optimistic relative efficacy estimates for lifileucel compared with chemotherapy in the model. The committee also agreed that it was reasonable to assume BSC and chemotherapy are equally effective. But, because the same data was used for BSC as for chemotherapy, the same uncertainties associated with estimating the effectiveness of chemotherapy also applied to the BSC estimates. The committee concluded that it would take this into account in its decision making.

Health-related quality-of-life data was collected in C‑144‑01 using the EORTC QLQ‑C30 questionnaire. The company used the Kim et al. (2012) and the Wojciechowski et al. (2023) models to approximate EQ‑5D scores. But the company could not use these values in its model. It explained that the quality-of-life data was only an exploratory outcome, and the sample size was small. So, it was unable to determine utility value changes over time. It further explained that the missing data was not random. This could have affected the statistical reliability of mapped utilities, and the mapping models identified were not developed based on UK demographics. Instead, the company extracted utility values from the published literature and 10 previous NICE technology appraisals (progression free: 0.77, progressed disease: 0.67).

The EAG explained that the company's approach was unconventional and that it had some concerns. Firstly, clinical experts at the company's advisory board said that they would have expected a larger gap between progression-free and progressed-disease utility values. The EAG also explained that some of the utility values were not derived using the EQ‑5D measurement method as specified in the NICE reference case. But it noted that the company had presented multiple sensitivity analyses. These included analyses using the source with the largest difference between progression-free and progressed-disease utility values and excluding non-EQ‑5D values. The EAG explained that the sensitivity analysis had a small impact on the cost-effectiveness estimates. The committee emphasised that, ideally, health-related quality-of-life data should be collected for people in the relevant clinical trial using EQ‑5D. But the committee noted that NICE's health technology evaluations manual states that, if EQ‑5D data is not available from the relevant trials, it can be sourced from the literature. The committee was satisfied that EQ‑5D data was not available from the relevant clinical trial. It also thought that the company had provided sensitive analyses that shows that the choice of data set has a small impact on the cost-effectiveness estimates. The committee concluded that the company's analysis which excluded non‑EQ‑5D values should be used for decision making.

The company's base-case model included administration costs associated with lifileucel infusion, and pretreatment and post-treatment costs, specifically:

• tumour tissue procurement

• lymphodepletion chemotherapy administration

• lifileucel administration

• interleukin‑2 administration and monitoring.
  
  The EAG explained that set-up, training and logistics costs associated with introducing lifileucel into the NHS were not included in the company's base-case model. The EAG used the NHS England's CAR‑T tariff (£60,863) to account for these costs. To avoid double counting for people who had lifileucel, it removed the costs associated with lifileucel administration listed above, the adverse event management costs and the health-state costs for the first 100 days. NHS England explained that these costs are also covered by the CAR‑T tariff. It also explained that there is limited evidence for what the associated costs for lifileucel will be. It said that, to give providers certainty about income in the implementation phase, it is committed to commissioning lifileucel activity in line with the cost of the CAR‑T tariff. So, it thinks that the CAR‑T tariff should be incorporated into the evaluation. NHS England also thought that some costs in the CAR‑T tariff would not apply for lifileucel but that this would be offset by costs which would apply but are not captured in the tariff.

The committee noted there was uncertainty associated with many of the model inputs and requested the following further analyses:

• an explanation for why the baseline characteristics were different between the 2 cohorts when the eligibility criteria were the same (see section 3.5)

• a scenario analysis, using only the latest data cut from cohort 4 in the ITCs and economic model, to help understand the impact of uncertainty when using the pooled analysis set (see section 3.6)

• further evidence on the generalisability of the pooled analysis set to the population who would have lifileucel in NHS clinical practice, considering the difference in baseline characteristics between the 2 cohorts (see section 3.5)

• outcomes for the pooled analysis set using the latest data cut (see section 3.5)

• an analysis from C‑144‑01, outlining how this data justifies a higher cure fraction for OS than PFS (see section 3.10), including:

  • the most recent data cut

  • PFS2 data

  • the proportion undergoing surgical resection after lifileucel

• further details on what is included in the CAR‑T tariff that is not already covered by the company's resource use estimates (from NHS England) (see section 3.14)

• further analyses on how its resource use estimates are reflective of those that will be incurred in the NHS, including for the set-up, logistics and training associated with lifileucel (from the company; see section 3.14)

• an analysis that resolves the problems the EAG identified in the company's probabilistic analysis.
  
  The committee discussed its preferred assumptions, noting that these may change after consideration of the further analyses it has requested. But, given the data it was presented, its preferred assumptions included:

• ipilimumab, chemotherapy and BSC included as relevant comparators (see section 3.4)

• estimating the relative efficacy of chemotherapy using the naive comparison of C‑144‑01 and Mangin et al. (2021; see section 3.8)

• using a log-logistic MCM for extrapolating lifileucel PFS and OS (see section 3.10)

• estimating OS and PFS for lifileucel using unadjusted MCMs (see section 3.11)

• estimating OS and PFS for ipilimumab by applying the inverse of the STC-adjusted hazard ratios to the lifileucel MCMs (see section 3.11)

• using chemotherapy as a proxy for BSC and assuming both treatments are equally effective (see section 3.12)

• using utility values from the analysis that excluded non-EQ‑5D values (see section 3.13)

• applying a severity weight of 1.7 to the comparisons with chemotherapy and BSC (see section 3.15)

• applying a severity weight of 1.2 to the comparison with ipilimumab (see section 3.15).
  
  The committee noted that the costs associated with the set-up, logistics, training and delivery of lifileucel were uncertain (see section 3.14). The committee also noted the following additional differences between the company and the EAG's base-case analysis, which it noted had a minimal impact on the incremental cost-effectiveness ratio (ICER):

• using an assumed cure time point of 5 years

• tumour procurement costs weighted using data on the distribution of resection sites from C‑144‑01.
  
  Because of confidential discounts for lifileucel and comparators, exact ICERs are confidential, so cannot be reported here. The committee was unable to agree on its preferred ICER because further analysis is likely to affect its preferred assumptions. But the committee noted that, including the commercial discounts, all the cost-effectiveness estimates presented were substantially higher than what NICE considers a cost-effective use of NHS resources. This included both the company's and EAG's cost-effectiveness estimates.

NICE's manual on health technology evaluations notes that, above a most plausible 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. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. But it will also take into account other aspects including uncaptured health benefits. The committee noted the high level of uncertainty, specifically around the:

• generalisability of the C‑144‑01 cohorts (see section 3.5)

• STC used to compare lifileucel with ipilimumab (see section 3.7)

• absolute treatment effects for lifileucel compared with ipilimumab (see section 3.11)

• use of a naive comparison to compare lifileucel with chemotherapy (see section 3.8)

• relative treatment effects for lifileucel compared with BSC (see section 3.12)

• whether the company's approach captures all resource costs that are relevant for the evaluation (see section 3.14).
  
  The committee thought that the additional requests for analyses (see section 3.16) may mitigate some of the outstanding uncertainty. So, the committee concluded that it could not yet determine an acceptable ICER threshold.

Several equalities issues were identified by stakeholders, including that:

• some people may not be able to easily get to a treatment centre

• the suitability of lifileucel for people with a learning or physical disability will need to be considered carefully to ensure it would be offered when appropriate

• lifileucel is unsuitable for women, trans men and non-binary people who are pregnant

• lifileucel is unsuitable for some older, frailer people

• C‑144‑01 did not include people under 18 years.
  
  The committee noted that access to specialist centres is an implementation issue that cannot be addressed by a NICE evaluation recommendation. Disability, pregnancy and age are protected characteristics under the Equality Act 2010. But, because its recommendation does not restrict access to treatment for some people over others, the committee agreed the recommendation did not lead to any equalities issues.

The committee noted the uncertainties associated with many of the model inputs, including the survival estimates and costs. The committee was unable to agree upon its preferred ICER because of the uncertainties. So, lifileucel should not be used for previously treated unresectable or metastatic melanoma (Stage IIIc to Stage IV) in adults who have had:

• a PD‑1 blocking antibody, and

• a BRAF inhibitor with or without a MEK inhibitor if the cancer is BRAF V600 mutation positive.