Lisocabtagene maraleucel for treating relapsed or refractory large B-cell lymphomas after first-line chemotherapy when a stem cell transplant is suitable

Large B‑cell lymphoma is an aggressive type of non-Hodgkin lymphoma. There are different subtypes of large B‑cell lymphoma, including those considered within this evaluation:

• diffuse large B‑cell lymphoma (DLBCL)

• high-grade B‑cell lymphoma (HGBCL)

• primary mediastinal B‑cell lymphoma (PMBCL)

• follicular lymphoma grade 3B (FL3B).
  
  DLBCL is the most common type. The disease characteristics and treatment pathways of each of these subtypes are considered similar at second line. People with large B‑cell lymphoma can have swollen lymph nodes, night sweats, fever, weight loss and itching. The patient expert explained that large B‑cell lymphoma has a large impact on daily life. Also, people may need the support of a carer because of physical weakness and fatigue. They also described the significant mental health challenges that people may have from:

• worry about the effects of the condition

• the impact it has on friends and family

• worry about not being able to tolerate the substantial side effects of current treatment options.
  
  The committee recognised that relapsed or refractory large B‑cell lymphoma after first-line chemotherapy has a large disease burden.

DLBCL, PMBCL, HGBCL and FL3B are generally managed using the same clinical pathway in NHS clinical practice. But some treatments are only reimbursed for specific large B‑cell lymphoma types. People with untreated large B‑cell lymphoma may be offered rituximab, cyclophosphamide, doxorubicin, vincristine and prednisolone (R‑CHOP). In 2023, NICE recommended polatuzumab vedotin with rituximab, cyclophosphamide, doxorubicin and prednisolone (R-CHP) for untreated DLBCL. For large B‑cell lymphoma that is relapsed or refractory to initial treatment, clinicians may offer salvage chemotherapy. If the condition responds after salvage chemotherapy, high-dose chemotherapy and, for people who are able to have one, a stem cell transplant can be offered. Transplant suitability is based on the person's tolerance of intensive treatment and is usually only offered to people under 70 years. The clinical experts said that high-dose chemotherapy is associated with high toxicity and can cause substantial side effects for the people who have it. The patient expert also explained that some people are unable to tolerate the side effects of intensive chemotherapy. The clinical experts noted that people with large B‑cell lymphoma that relapses within 12 months or is refractory to initial treatment and who can have an autologous stem cell transplant may be offered axicabtagene ciloleucel (axi‑cel; see NICE's technology appraisal guidance on axi-cel for treating relapsed or refractory diffuse large B-cell lymphoma after first-line chemoimmunotherapy, from now TA895). But axi‑cel is only available for use at second line through the Cancer Drugs Fund, so this does not represent routine clinical practice. The committee concluded that people with relapsed or refractory large B‑cell lymphoma and clinicians would welcome a new treatment option.

The company proposed lisocabtagene maraleucel (liso‑cel) for a narrower population than its marketing authorisation. It focused on DLBCL, PMBCL, HGBCL and FL3B that were refractory to, or had relapsed within 12 months of, first-line chemoimmunotherapy in adults who could have a stem cell transplant. This was to align with the key clinical trial, TRANSFORM (see section 3.5). The committee agreed that the company's positioning of liso‑cel was appropriate.

The committee recalled that relapsed or refractory large B‑cell lymphoma after first-line chemoimmunotherapy is usually treated with salvage chemotherapy, high-dose chemotherapy and an autologous stem cell transplant (from now, called standard care). The clinical expert submission said that axi‑cel was expected to be the main alternative for liso‑cel in clinical practice. Both liso‑cel and axi‑cel are chimeric antigen receptor (CAR) T‑cell therapies (also called CAR‑T therapies). The clinical experts at the committee meeting also noted that they key difference between liso‑cel and axi‑cel was the safety profile. There are expected to be substantially lower grade 3 and 4 adverse events for people having treatment with liso‑cel. They said that this would be important for the quality of life of people having treatment. They also expected that it will reduce resource use, including length of hospital stay and intensive care use. The committee recalled that axi‑cel had not been recommended for routine commissioning at second line, so was not an appropriate comparator in this evaluation. The committee concluded that standard care was the relevant comparator.

The clinical-effectiveness evidence for liso‑cel compared with standard care came from TRANSFORM. This was a phase 3 randomised open-label trial. It included adults with primary refractory or early relapsed (within 12 months of first-line treatment) DLBCL, HGBCL, PMBCL, T‑cell histiocyte rich B-cell lymphoma (THRBCL) or FL3B eligible for a stem cell transplant. Standard care consisted of 3 cycles of re-induction therapy followed by high-dose chemotherapy and an autologous stem cell transplantation if the condition responded. People in the standard-care arm could cross over to have liso‑cel if their condition:

• did not completely or partially respond by 9 weeks after randomisation

• progressed at any time, or

• needed to start a new antineoplastic therapy because of efficacy concerns (absence of complete response) 18 weeks after randomisation.
  
  The primary end point was event-free survival (EFS) defined as:

• the time from randomisation to progressive disease

• failure to have a complete response or partial response by 9 weeks after randomisation, or

• start of a new antineoplastic therapy because of efficacy concerns or death from any cause, whichever happens first.
  
  At the final data cut-off in October 2023, there was a statistically significant benefit for liso‑cel compared with standard care for EFS (hazard ratio [HR] 0.38, 95% confidence interval [CI] 0.26 to 0.54). The difference in overall survival (OS) was not statistically significant (HR 0.76, 95% CI 0.48 to 1.19). But the result was confounded by the high proportion (66.3%) of people in the standard-care arm who crossed over to have liso‑cel as a subsequent treatment. Median OS could not be estimated for liso‑cel or standard care at the final data cut-off. The committee concluded that the results of the trial showed a statistically significant EFS benefit for liso‑cel compared with standard care.

The company noted that TRANSFORM was done specifically in the population of interest (see section 3.3). It allowed people to cross over from the standard-care arm to have subsequent liso-cel, and chemotherapy-based bridging therapy regimens were used. This was in contrast with the ZUMA‑7 trial used to inform TA895. ZUMA‑7 was a phase 3 randomised trial of axi‑cel used after chemoimmunotherapy in adults with primary refractory or early relapse DLBCL who were due to have a stem cell transplant. Crossovers between treatment arms and chemotherapy bridging were not included in ZUMA‑7. So, the company considered that the design of TRANSFORM better reflected NHS clinical practice than that of ZUMA‑7. The company did acknowledge that TRANSFORM differed from NHS clinical practice in some respects. Firstly, TRANSFORM was done before several treatments for subsequent use in the pathway were available in routine practice. See NICE's technology appraisal guidance on:

• glofitamab for treating relapsed or refractory diffuse large B-cell lymphoma after 2 or more systemic treatments

• loncastuximab tesirine for treating relapsed or refractory diffuse large B-cell lymphoma and high-grade B-cell lymphoma after 2 or more systemic treatments

• epcoritamab for treating relapsed or refractory diffuse large B-cell lymphoma after 2 or more systemic treatments).
  
  This meant that few people in TRANSFORM had these subsequent treatments; most had subsequent chemotherapy. So, the company thought that OS in the liso‑cel arm was potentially underestimated relative to NHS clinical practice because these subsequent treatments are more effective than chemotherapy. Secondly, people in TRANSFORM had leukapheresis before being randomised to either liso‑cel or standard care. Also, liso‑cel manufacturing was done for people in both arms to enable rapid liso‑cel infusion after crossover (see section 3.5). The clinical experts explained that, in NHS clinical practice, people cannot have apheresis at second line in anticipation of needing a subsequent CAR‑T therapy. So, there is a greater delay between progression on standard care at second line and the subsequent CAR‑T therapy in NHS clinical practice compared with in TRANSFORM. The clinical experts said that the design of TRANSFORM to allow people to cross over to liso‑cel quickly was beneficial for the people in the trial. It also favoured the standard-care arm. The company also noted that, by having apheresis before randomisation in TRANSFORM, people may have had improved T‑cell fitness compared with people who have apheresis after progression on standard care in clinical practice. So, the company thought that OS in the standard-care arm was overestimated relative to NHS clinical practice. The clinical experts estimated that outcomes may improve by about 10% for people who have had apheresis before needing subsequent CAR‑T therapy compared with having apheresis at third line, as in clinical practice. In addition to the generalisability issues noted by the company, the EAG was also concerned that:

• drop out between leukapheresis and infusion in the liso‑cel arm of TRANSFORM was lower than expected in clinical practice

• the proportion of people having bridging therapy in TRANSFORM was lower than in NHS practice

• more people were expected to have had polatuzumab vedotin with R‑CHP at first line in clinical practice than did in TRANSFORM.
  
  The clinical experts noted that drop out between leukapheresis and infusion had improved in clinical practice over the last 5 years. But they were still concerned that some people will not live long enough between T‑cell collection and reinfusion. They also commented that the availability of polatuzumab vedotin with R‑CHP was expected to reduce the population size at second line because of its higher efficacy than R‑CHOP. But they did not expect any biological differences or impact on efficacy at second line for people who had had polatuzumab vedotin with R‑CHP compared with R‑CHOP. The committee acknowledged the issues of generalisability to NHS practice, and that this increased uncertainty in the clinical- and cost-effectiveness results. But it concluded that TRANSFORM provided the best available evidence for liso‑cel compared with standard care.

The company provided a partitioned survival model to estimate the cost effectiveness of liso‑cel compared with standard care. The model had 3 health states: event-free, post-event and death. The company justified using EFS to inform the model health states because it was:

• the primary end point in TRANSFORM

• consistent with the model health states used in the economic model to support TA895.
  
  The clinical experts at the committee meeting agreed that EFS was a relevant outcome. They explained that it was standard practice to collect it in clinical trials (such as ZUMA‑7) for relapsed or refractory large B‑cell lymphomas after first-line chemotherapy. The EAG was concerned that progression from the event-free state to a post-event health state did not reflect an objective change in health status. It said that the modelled cohort with large B‑cell lymphoma that is cured at subsequent treatment lines would not be assigned the health benefits associated with cure. This was because they would remain in the same post-event health state. It also noted that the post-event health state included people who were cured (for example, after subsequent CAR‑T therapy) and not cured, so was not a homogenous population. The EAG acknowledged that an economic model based on EFS had been accepted by the committee as part of the axi‑cel evaluation. But it said that suitable alternatives may not have been available for consideration then. So, the EAG preferred to use progression-free survival on subsequent treatment (PFS2) to partition the model health states instead of EFS. The company noted that the model based on PFS2 was limited because of discrepancies in follow up between death and disease progression. It explained that, after 36 months, people in TRANSFORM were only followed up for OS. So, it thought that the PFS2 endpoint could have been underestimated because people were censored from this dataset but known to be alive after the 36‑month timepoint. The company was also concerned that the PFS2 model structure assumed that there is no health-related quality of life detriment for people who move from second line to subsequent treatment for any reason. The committee noted the EAG's concerns. But it thought that the pre‑PFS2 health state in the EAG's preferred model structure included people having second and third lines of treatment, so was also not homogenous. The committee concluded that the company's model with health states based on EFS was appropriate for decision making.

The company said that plateaus were seen in the OS data from TRANSFORM, suggesting that some people had long-term remission and survival. So, it fitted mixture cure models to each treatment arm to model the long-term OS outcomes. The company used the log-normal curve in its base-case analysis for liso‑cel OS because it had:

• the best statistical fit

• a good visual fit to the observed data

• a cure fraction (60.8%) that aligned with the estimate of one of its clinical experts.
  
  The EAG thought that the TRANSFORM OS data was less mature than the EFS and PFS2 data. This meant that it was less likely that the true cure fraction was estimated accurately. It also noted that the OS follow up from TRANSFORM was less mature and had a smaller sample size than ZUMA‑7. So, it thought that ZUMA‑7 was a more reliable source than TRANSFORM for estimating the long-term efficacy of liso‑cel, despite being for a different treatment. The EAG thought that the company's preferred log-normal curve for liso‑cel OS was too optimistic. This was because the cure fraction was higher than estimated by models fitted to PFS2 data, and because it did not expect cure to happen after the PFS2 outcome. The EAG also noted that the company's predicted long-term survival outcomes for liso‑cel were higher than the long-term survival accepted for axi‑cel in TA895. The EAG preferred to use SurvInt to model liso-cel OS. SurvInt is a freely available R Shiny tool that uses user-specified population survival at key time points to produce parametric extrapolations. Its inputs into SurvInt included:

• a survival estimate at 11.05 months from TRANSFORM

• a survival input at 4 years from ZUMA‑7

• a cure fraction of 50%, chosen for consistency with the cure fractions estimated from the TRANSFORM PFS2 data, and extrapolations from ZUMA-7

• a log-logistic model.
  
  The company said that the EAG's use of ZUMA‑7 data to inform the efficacy of liso‑cel was not appropriate. It thought that the study design of TRANSFORM better reflected NHS clinical practice (see section 3.6). It noted that differences in the survival outcomes between TRANSFORM and ZUMA‑7 could be explained by differences in the trial designs. It also noted that both trials were expected to underestimate long-term OS because of the recent availability of novel subsequent treatments in clinical practice (see section 3.6). The company commented that liso‑cel and axi‑cel are different treatments with different manufacturing processes, and that TRANSFORM provided relevant data for liso‑cel in the population of interest. The company also recalled the issue of censoring in the PFS2 data from TRANSFORM (see section 3.7). It said that this likely influenced the difference in cure rates between the OS and PFS2 models. The company was also concerned with the use of SurvInt to extrapolate survival in its preferred analyses. It said that the SurvInt approach ignored most of the observed trial data for liso‑cel, and arbitrarily used 2 survival inputs to inform extrapolations. It also noted that the cure fraction was arbitrarily chosen. But the cure fractions predicted by its mixture cure models were based on the observed data, and produced from an approach aligned to NICE's technical support document on survival analysis for economic evaluations alongside clinical trials and NICE's technical support document on flexible methods for survival analysis. The clinical experts thought that longer-term OS estimates were likely to be similar between liso‑cel and axi‑cel. They also commented that the OS estimates for liso‑cel (based on the company's mixture cure models) at 5, 10 and 15 years were reasonable. The committee commented on the usefulness of the SurvInt tool for exploring the sensitivity of extrapolated outcomes. But it was concerned that the tool did not use most of the observed data for liso-cel, and it was uncertain of the tool's reliability for use in decision making. The committee concluded that the company's mixture cure OS model was acceptable for liso‑cel, but there was remaining uncertainty on long-term survival.

The company's clinical experts thought that all the survival curves produced using mixture cure models for standard care overestimated long-term survival compared with clinical practice. The company explained that OS estimates for standard care may be higher than expected in NHS clinical practice because of the design of TRANSFORM (see section 3.6). It used the log-normal curve in its base-case analysis because it had the best statistical fit and it estimated the lowest cure fraction (50.7%). It noted that this approach was biased in favour of the standard-care arm because the curve likely overestimated survival for people having standard care. The EAG agreed that survival was likely overestimated by all the company's mixture cure models because of immaturity of the data. The EAG preferred to use a log-logistic curve from SurvInt to estimate standard-care OS in the absence of a suitable alternative. Its inputs to SurvInt included:

• survival estimates at 6.59 and 17.76 months from TRANSFORM

• a cure fraction of 35%, chosen for consistency with the cure fractions estimated from the TRANSFORM PFS2 data.
  
  The EAG acknowledged that its SurvInt model underestimated the tail of the Kaplan–Meier curve from TRANSFORM. But it thought that this was appropriate given that TRANSFORM was expected to overestimate survival compared with NHS clinical practice (see section 3.6). The company was concerned with the EAG's use of the SurvInt approach (see section 3.8). The committee recalled its concerns with the SurvInt approach and concluded that the company's mixture cure OS model for standard care was the most appropriate.

Time to next treatment was defined as the time from randomisation to death from any cause, or to the start of new antineoplastic therapy, whichever happened first. The company extrapolated data for time to next treatment from TRANSFORM using mixture cure models to inform the modelling of subsequent treatments. It noted that all the extrapolations for the liso‑cel arm had similar estimates of long-term survival. This meant that there was low uncertainty associated with the choice of curve for time to next treatment. The EAG was concerned that the company's extrapolations for time to next treatment were more optimistic than the EFS extrapolations, given they had similar definitions. It thought that EFS was the more mature outcome, and that it was likely to give a more reliable long-term extrapolation. It also noted that the extrapolations of time to next treatment from TRANSFORM were more optimistic than the extrapolations from ZUMA‑7 in TA895. So, the EAG preferred to use the EFS extrapolations from TRANSFORM to model time to next treatment. The committee thought that the dataset for time to next treatment provided the best available evidence for the outcome for time to next treatment. So, it preferred to use the company's extrapolations for time to next treatment in the model.

The company used the mean age of people in TRANSFORM to inform the starting age at model entry. The company considered the mean age to be confidential, so it cannot be reported here. The EAG preferred to align the model starting age with data provided by NHS England. This suggested that the mean age of people who have had second-line axi‑cel since it entered the Cancer Drugs Fund is 59 years. The committee noted that the model starting age had a minimal impact on the cost-effectiveness estimate. It concluded that company's use of the mean age of people in TRANSFORM was acceptable for the model starting age.

The company applied a weekly cycle length for the first 5 years in its economic model, followed by an annual cycle length. It discounted costs and benefits at a rate of 3.5% per annum in its base-case analyses. The EAG disagreed with the annual application of the discount rate during the weekly cycle period and preferred to use a per cycle discount rate for the first 5 years. The committee noted that application of the discount rate in the first 5 years of the model had a minimal impact on the cost-effectiveness estimate. It concluded that EAG's application of a per cycle discount was acceptable.

Health-state utility values in the company's base-case analyses were estimated using EQ‑5D data from TRANSFORM. A value of 0.852 was estimated for the event-free health state. This value was also used to inform the pre‑PFS2 health state in the EAG's preferred model structure (see section 3.7). The EAG thought that the utility value of 0.852 was too optimistic. This was because it was higher than the event-free utility value of 0.785 used in TA895 and similar to the general population utility estimate of 0.853. The EAG preferred to use the utility value of 0.785 from the axi‑cel evaluation for the event-free and pre‑PFS2 health states. The committee noted that there was a low completion rate for EQ‑5D data in TRANSFORM, and that data was not collected after treatment switching. The company explained that there had been challenges completing the data during the COVID‑19 pandemic. The committee noted the uncertainty in the EQ‑5D data from TRANSFORM. But it thought that TRANSFORM provided the most relevant EQ‑5D data for liso‑cel in the population of interest. The committee also commented that the total incremental quality-adjusted life years (QALYs) were more conservative when using the TRANSFORM data to inform health-state utility values, than when using data from TA895. It concluded that TRANSFORM was the most appropriate source for the event-free health-state utility value.

The clinical experts explained that bridging therapy is treatment offered to control large B‑cell lymphomas and symptoms between T‑cell collection and reinfusion. In TRANSFORM, 63% of people had bridging therapy. The company modelled bridging therapy costs (proportion of people having bridging therapy, and the distribution of the bridging therapy regimens) based on TRANSFORM. Clinical experts consulted by the EAG suggested that the proportion of people having bridging therapy and the distributions would differ from those modelled in the company's base case. So, the EAG preferred to use UK‑specific data based on a study by Boyle et al. (2023) to estimate the proportion of people having bridging therapy and the distribution of the bridging therapy regimens. The clinical experts at the committee meeting noted that bridging therapy is commonly used in NHS clinical practice. The NHS England Cancer Drugs Fund clinical lead said that, of the 255 people who had axi‑cel at second line, 96% had had bridging therapy. The committee thought that it was important to align modelled costs and benefits. It preferred to use the TRANSFORM data for costing bridging therapy in its decision making, but it noted the generalisability concerns of this to NHS clinical practice.

Subsequent treatment costs were applied as a one-off cost based on data for time to next treatment from TRANSFORM. The company calculated that the proportion of events for time to next treatment that were the start of a new treatment was 69.6% in the liso‑cel arm and 94.2% in the standard-care arm. These percentages were applied to the relevant extrapolation for time to next treatment (see section 3.10) to calculate the total proportion of people who had at least 1 subsequent treatment. The EAG's clinical experts thought that the proportion of events for time to next treatment that were the start of a new treatment was higher than expected in clinical practice for the standard-care arm. They said that a third of people would have palliative care after an unsuccessful stem cell transplant at second line. So, the EAG assumed that 66% of events for time to next treatment were the start of a new treatment for standard care. The company modelled the distribution of subsequent treatments from TRANSFORM in its base case but noted that they did not fully reflect current NHS clinical practice (see section 3.6). The EAG preferred to use the estimates from the company's clinical experts, which it said were similar to estimates from the EAG's clinical experts. The company said that the EAG's base case substantially underestimated subsequent treatment costs in the standard-care arm. It also noted that the EAG's approach changed the costs to reflect NHS clinical practice but did not also adjust the efficacy. The EAG explained that its preferred efficacy estimates already deviated from the trial data (see section 3.8 and section 3.9). So, it did not agree that it had not considered an adjustment to the clinical outcomes as well as the costs of subsequent treatment. The company noted that it had presented a scenario analysis that used estimates from UK clinical experts to inform the distribution of subsequent treatments. In this scenario analysis, a more optimistic Weibull curve was used for liso‑cel OS to model the increase in survival expected from having more effective subsequent treatments in clinical practice. At the same time, a weighted average OS curve for standard care was applied to lower survival to a range expected in NHS clinical practice. The EAG noted that the weighted OS curve partly used data from CORAL. This was unlikely to have included subsequent treatment with bispecific antibodies, so it was also not reflective of current clinical practice. The clinical experts and NHS England Cancer Drugs Fund clinical lead explained that the treatment pathway for large B‑cell lymphomas was rapidly changing. The clinical experts commented that it was unusual for a person not to have subsequent treatment at third line if they were able to. They explained that the absolute number of people who go on to have a subsequent treatment after liso‑cel was expected to be lower than after standard care because of the reduced risk of relapse. But, of the people that did relapse, they expected a similar proportion of people (up to 80.0%) to go on to have subsequent treatment in both treatment arms. The clinical experts said that most people would be given a bispecific antibody as subsequent treatment after liso‑cel in clinical practice. They also noted that, generally, the preference is to use CAR‑T therapy after standard care in clinical practice if the person is fit enough. But they explained that use was unlikely to be as high as the 94% of people as reported in TRANSFORM. The committee agreed with the clinical experts' expectations that liso‑cel would lower the risk of relapse compared with standard care but that, after relapse, a similar proportion of people would have subsequent treatment in both treatment arms. So, it preferred to set the proportion of events for time to next treatment that were the start of a new subsequent treatment in the model to be equal for liso‑cel and standard care. The committee concluded that the clinical experts' estimate of up to 80.0% was acceptable to use in the model. This was because it was also between the 69.6% value in the liso-cel arm and the 94.2% value in the standard-care arm from TRANSFORM. It agreed with the company that it was important to align modelled costs and benefits. It recalled its preference for modelling OS based on mixture cure models fitted to the TRANSFORM data (see section 3.8 and section 3.9). But it also noted that the subsequent treatments modelled did not reflect NHS clinical practice. In the absence of a method to reliably adjust the treatment effectiveness, the committee concluded that it preferred to model the proportion in each arm as equal. But it agreed that it would accept the distribution of subsequent treatments based on the data from TRANSFORM. The committee noted that it had remaining concerns for the generalisability of this trial data to NHS clinical practice. It also noted that the resulting impact on the clinical- and cost-effectiveness estimates was uncertain and would be considered in its decision making.

A CAR‑T cell tariff cost of £41,101, assumed to capture all costs of care from the decision for the person to have CAR‑T therapy to 100 days after infusion, was accepted for use in TA895. This tariff cost included the costs associated with managing adverse events happening up to 100 days after infusion (excluding any costs associated with the treatment of hypogammaglobulinemia, that is, intravenous immunoglobulin). The EAG was concerned that the company applied the CAR‑T cell tariff cost to people having subsequent CAR‑T therapy in the standard-care arm but not the costs associated with adverse events for other subsequent therapies in either treatment arm. It commented that this approach biased the cost-effectiveness results in favour of liso‑cel. The EAG preferred to exclude the costs associated with adverse events (estimated by the company as £10,611) from the CAR‑T cell tariff cost when used for subsequent CAR‑T therapy. The company agreed with the EAG's adjustment of the CAR‑T cell tariff cost at third line because it had not intended to include adverse event costs for subsequent treatment. The NHS England Cancer Drugs Fund clinical lead explained that NHS England had been working with NHS trusts to determine the tariff cost that applied in NHS practice. They said that a value of £57,080 was agreed, which applied from the start of the new financial year for 2024/25. But they also noted that inflation had uplifted this value. So, a tariff cost of £58,964 was now applicable for the rest of the 2024/25 financial year and for use in this appraisal. The committee concluded that the updated tariff cost of £58,964 should be applied in the model. It agreed that the EAG's adjustment of the tariff cost in the model was acceptable.

NICE's manual on health technology evaluations notes 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. 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 that:

• relapsed or refractory large B‑cell lymphoma after first-line chemotherapy has a large disease burden (see section 3.1)

• the only treatment option available in routine clinical practice can have substantial side effects for some people (see section 3.2).
  
  But it also noted the uncertainty in this appraisal, specifically:

• issues of generalisability of TRANSFORM to NHS practice, and the impact on the clinical- and cost-effectiveness results (see section 3.6), including:

  • the proportion of people having bridging therapy, and the distribution of bridging therapies at second and third lines (see section 3.14)

  • the proportion of people having subsequent therapy, and the distribution of subsequent therapies (see section 3.15)

• long-term OS in people having treatment with liso‑cel (see section 3.8)

• the low completion rate for EQ‑5D data in TRANSFORM (see section 3.13).
  
  So, the committee concluded that an acceptable ICER would be around the middle of the range NICE considers a cost-effective use of NHS resources (£20,000 to £30,000 per QALY gained).

The exact cost-effectiveness results cannot be reported here because of confidential discounts for liso‑cel, comparators and subsequent treatments. When the CAR‑T cell tariff cost of £41,101 was used in the model (see section 3.16), the company's base-case ICER was below £30,000 per QALY gained and the EAG's base-case ICER was above £30,000 per QALY gained. The committee considered the results of the cost-effectiveness analysis for liso‑cel using its preferred assumptions, which included:

• model health states based on EFS (see section 3.7)

• the company's mixture cure OS model for liso‑cel (see section 3.8)

• the company's mixture cure OS model for standard care (see section 3.9)

• the company's extrapolations for time to next treatment to inform time to next treatment in the model (see section 3.10)

• a model starting age based on the mean age of people in TRANSFORM (see section 3.11)

• a per cycle discount rate (see section 3.12)

• the event-free health-state utility value based on TRANSFORM (see section 3.13)

• use of the TRANSFORM data for costing bridging therapy (see section 3.14)

• setting the proportion of people who have subsequent therapy after a time to next treatment event to be equal for liso-cel and standard care, assuming a value of 80% (see section 3.15)

• the distribution of subsequent treatments based on the data from TRANSFORM (see section 3.15)

• the updated CAR‑T cell tariff cost of £58,964 (see section 3.16)

• adjusting the CAR‑T cell tariff cost in the third line to exclude adverse event costs (see section 3.16).
  
  With the committee's preferred assumptions, the cost-effectiveness results were above the range normally considered a cost-effective use of NHS resources. The committee also noted that the company's base-case ICER was above £30,000 per QALY gained when the updated CAR‑T cell tariff cost of £58,964 was applied in the model. So, the committee did not recommend liso‑cel for routine use in the NHS for treating relapsed or refractory large B‑cell lymphomas after first-line chemotherapy when a stem cell transplant is suitable.

Having concluded that liso‑cel could not be recommended for routine use, the committee then considered whether it could be recommended with managed access. But:

• Data from the final data cut of TRANSFORM had been used in the economic model (see section 3.5) and the company did not expect any further trial data to become available in the relevant population (see section 3.3) to support decision making.

• The committee considered the uncertainties in the evidence (see section 3.18) and the key issues raised by the EAG, and noted that:

  • The generalisability of TRANSFORM to NHS clinical practice would not be resolved by further data collection (see section 3.6). It also recalled that the treatment pathway for large B‑cell lymphomas is rapidly changing (see section 3.15). So, data collected during a period of managed access was unlikely to still be generalisable to NHS clinical practice by the time of the managed access review.

  • Long-term OS is affected by subsequent treatment use. So, the committee was not persuaded that the uncertainty associated with long-term OS in people having treatment with liso‑cel (see section 3.8) would be sufficiently resolved by further data collection in a rapidly changing treatment pathway.

• The committee thought that liso‑cel had not shown the plausible potential to be cost effective at the agreed price (see section 3.19).
  
  The committee concluded that liso‑cel did not meet the criteria to be considered for a recommendation with managed access. So, it could not recommend liso‑cel for use with managed access as an option for treating relapsed or refractory large B‑cell lymphomas after first-line chemotherapy when a stem cell transplant is suitable.

At scoping, a stakeholder noted that clinicians consider a person's fitness when deciding whether more intensive cancer treatments are suitable for them. A person's age may be used as a proxy for levels of fitness. Age is a protected characteristic under the Equality Act 2010. The committee was aware that NICE makes recommendations for technologies within their marketing authorisations. The committee recalled that the company positioned liso‑cel only for people for whom a stem cell transplant is suitable, which is usually people under 70 years. The committee considered the evidence that had been submitted. It noted that it had not seen evidence for liso‑cel for treating relapsed or refractory large B‑cell lymphomas in people for whom a stem cell transplant is not suitable, who are usually older and less well. The committee was aware of the need for new treatments in this population and was disappointed the company chose to position liso‑cel for the transplant-eligible population only. Stakeholders also commented that there is a geographic inequality because CAR T‑cell therapy is only provided at designated centres. The committee noted these concerns but concluded that its recommendation for liso‑cel would not adversely affect people protected by the equality legislation.

The committee considered whether there were any uncaptured benefits of liso-cel. It did not identify additional benefits of liso-cel not captured in the economic modelling. So, the committee concluded that all additional benefits of liso-cel had already been taken into account.

The committee concluded that the most plausible ICER based on its preferred assumptions is unlikely to represent a cost-effective use of NHS resources. So, liso‑cel is not recommended for treating relapsed or refractory large B‑cell lymphomas after first-line chemotherapy when a stem cell transplant is suitable.