Axicabtagene ciloleucel for treating diffuse large B-cell lymphoma and primary mediastinal large B-cell lymphoma after 2 or more systemic therapies

3.1 Diffuse large B-cell lymphoma and primary mediastinal large B-cell lymphoma are aggressive subtypes of non-Hodgkin lymphoma. The condition does not respond well to chemotherapy, and outcomes for people with refractory or relapsed disease are poor and survival is limited. Axicabtagene ciloleucel is part of a group of advanced cancer treatments called chimeric antigen receptor (CAR) T‑cell therapies. CAR T‑cell therapies are personalised cancer immunotherapies which involve collecting and modifying people's own immune cells to treat their cancer. Axicabtagene ciloleucel is indicated for the treatment of relapsed or refractory diffuse large B-cell lymphoma and primary mediastinal large B-cell lymphoma after 2 or more lines of systemic therapy in adults. Clinical and patient experts explained that without CAR T‑cell therapy there is a lack of treatment options if the condition relapses after 2 therapies. The patient expert explained that chemotherapy treatments are intense, frequent and extremely difficult to tolerate. They described that chemotherapy treatments have challenging side effects and a large impact on quality of life for little clinical success. The patient and clinical experts explained that the introduction of axicabtagene ciloleucel into the treatment pathway has revolutionised treatment and provided hope of a cure for this aggressive condition. The committee concluded that axicabtagene ciloleucel is a valuable treatment option.

3.2 In the original appraisal, the company proposed that axicabtagene ciloleucel can be used in 4 possible positions in the treatment pathway, specifically for people whose disease:

3.3 The clinical-effectiveness evidence for axicabtagene ciloleucel was from ZUMA‑1, an ongoing, phase 1/2, multicentre, open-label, single-arm trial. ZUMA‑1 investigated axicabtagene ciloleucel for aggressive relapsed or refractory B-cell non-Hodgkin lymphoma (including diffuse large B-cell lymphoma, primary mediastinal large B-cell lymphoma and transformed follicular lymphoma) in adults that was refractory to treatment or had relapsed within 12 months of autologous stem cell transplant. The company presented updated overall survival results collected in ZUMA‑1 which included a minimum follow-up of 5 years. The updated ZUMA‑1 results showed that 64 patients (59%) had died at database lock (11 August 2021). Updated median overall survival was 23.5 months with an observed plateau of 45% at around 26.0 months. Updated median progression-free survival was 5.8 months with an observed plateau in progression-free survival of 40% at 9.0 months. The committee heard that ZUMA‑1 progression-free survival was only updated to 2 years, because there was no protocol mandate to collect data after 2 years. The company stated that any progression-free survival data collected after 2 years would not be consistent with progression-free survival data collected before 2 years. The committee would have preferred to see more mature progression-free survival data but accepted that additional overall survival data had reduced uncertainty. The committee concluded that additional axicabtagene ciloleucel data supported the predictions in the original submission of long survival times for a substantial proportion of patients.

3.4 Further clinical evidence was collected from the SACT dataset while axicabtagene ciloleucel was available in the Cancer Drugs Fund. The SACT dataset contains UK real-world clinical-effectiveness data from the Cancer Drugs Fund population between December 2018 and October 2021. The median age of people having axicabtagene ciloleucel was 59.5 years, similar to ZUMA‑1. Overall survival rates in the Cancer Drugs Fund population were comparable to ZUMA‑1 over the 36 months of data collection. Median overall survival of people having axicabtagene ciloleucel was 28.5 months and 45% of people were alive after 3 years. The ERG highlighted that Eastern Cooperative Oncology Group (ECOG) performance status was not collected for a proportion of people in the SACT dataset, which made comparisons with ZUMA‑1 more difficult. Clinical experts explained that they expect axicabtagene ciloleucel to perform similarly in routine NHS clinical practice to ZUMA‑1 and the SACT dataset. The committee concluded that outcomes from the SACT dataset were comparable to those from ZUMA‑1 and supported the use of axicabtagene ciloleucel in routine NHS clinical practice.

3.5 Because ZUMA-1 is a single-arm trial, the primary evidence source for the comparator was the SCHOLAR‑1 study. SCHOLAR‑1 was a retrospective study of pooled data from 4 datasets. These datasets included adults with diffuse large B-cell lymphoma (n=552), primary mediastinal large B-cell lymphoma (n=14), transformed follicular lymphoma (n=27) and 'other' (n=43). Treatment options included salvage chemotherapy, rituximab maintenance therapy and observation after autologous stem cell transplant. The company did a targeted literature review of studies on diffuse large B-cell lymphoma after 2 or more lines of systemic therapy. This was to respond to the Cancer Drugs Fund review terms of engagement's request for additional data to address uncertainties in comparator data. The company identified 3 studies, with only 1 study from the UK. Radford et al. (2019) investigated treatment patterns and outcomes for people with relapsed or refractory diffuse large B-cell lymphoma from a single UK centre. The company argued that SCHOLAR‑1 is still the most appropriate source of data for salvage chemotherapy and stressed the approach maintained consistency with the original submission. The company also outlined that SCHOLAR‑1 outcomes were similar to the third-line population in the Radford study, validating their SCHOLAR‑1 approach. Clinical experts highlighted issues with SCHOLAR‑1. They argued that the populations in SCHOLAR‑1 were not comparable to ZUMA-1. They also noted that there are large datasets that could have been used to validate SCHOLAR‑1. The ERG explained that the company's targeted literature review used only 1 database, and highlighted that important sources of comparator data may have been missed. The ERG explained that both ZUMA‑1 and SCHOLAR‑1 were not done in the UK, so the company's argument for not selecting other studies on this basis was flawed. The ERG explained that running alternative survival extrapolation scenarios with the SCHOLAR‑1 data showed that results are sensitive to changes in survival estimates for salvage chemotherapy. The committee noted that it would have liked the company to have explored further sources of comparator data. It noted the limitations with SCHOLAR‑1 but concluded that the dataset was appropriate for decision making.

3.6 The same approach used in the original appraisal to model salvage chemotherapy was applied. SCHOLAR‑1 data was adjusted to address imbalances in baseline characteristics and ensure comparability between ZUMA‑1 and SCHOLAR‑1. People with an ECOG performance status of 2 to 4, an unknown ECOG status or primary refractory disease were excluded from the SCHOLAR-1 dataset. Separate survival curves were used to generate a weighted survival estimate dependent on whether people had a stem cell transplant or not. The ERG highlighted that these adjustments substantially reduced the sample size of the study, from 562 people to 133 people, increasing uncertainty. The ERG also highlighted that this reduced sample size was similar to those in the alternative comparator studies (Radford [UK], n=89; Fuji [Japan], n=189; Nakaya [Japan], n=131). The committee asked why the company didn't apply more formal adjustment methods involving matching. The company explained that it did not update the analysis because the adjustment approaches were accepted by the committee in the original appraisal. The committee would have preferred to see matching adjustments applied in the indirect comparison. Comparing ZUMA‑1 and SCHOLAR‑1 data, axicabtagene ciloleucel is associated with longer overall survival than salvage chemotherapy and sustained progression-free survival. The committee concluded that axicabtagene ciloleucel showed clinical benefit compared with salvage chemotherapy, however limitations in the comparator data mean that the exact size of the benefit was unknown.

3.7 The company's economic model used the same approach as in the original appraisal. The model included 3 health states: pre-progression survival, post-progression survival, and death. The company used a partitioned survival modelling approach in which progression-free and overall survival estimates were modelled independently. The proportion of people whose disease progressed at each cycle was calculated as the difference between the overall survival and progression-free survival curves. The company modelled axicabtagene ciloleucel using updated data from ZUMA‑1, and salvage chemotherapy using data from SCHOLAR‑1. The company analysis was largely consistent with the original appraisal. The model time horizon was 44 years with a monthly cycle length. The company updated inputs for intravenous immunoglobulin (IVIg). In the original appraisal, the proportion of people who had IVIg was informed by ZUMA‑1 data and was assumed to be for 12 months. During technical engagement, the company and ERG agreed to use available SACT data to inform IVIg inputs, that is, 16% of people had IVIg treatment for 6.5 months. Health state utility values were applied using data from ZUMA‑1. The committee questioned the assumption that people in remission returned to general population utility. The committee thought it would be more clinically valid for people in remission to have a utility decrement applied to account for the impact of having had diffuse large B‑cell lymphoma or primary mediastinal large B-cell lymphoma and intensive treatments. The company explained that it kept the analysis consistent with the original appraisal. The committee would have preferred scenarios investigating the impact of health state utility, but concluded that, as in the original appraisal, the economic model was appropriate for decision making.

3.8 Since the original appraisal, the company collected additional 60‑month axicabtagene ciloleucel overall survival data in ZUMA‑1. The company used a mixture cure modelling approach containing a proportion of people whose disease responds to axicabtagene ciloleucel and a proportion whose does not. People whose disease responds were assumed to have a mortality rate identical to that of the general population. Survival of people whose disease does not respond to axicabtagene ciloleucel was modelled using a standard parametric survival model. The combined survival curves made up the overall population of people with diffuse large B-cell lymphoma and primary mediastinal large B-cell lymphoma after 2 or more systemic therapies. The company indicated that all mixture cure models fit well to ZUMA‑1 overall survival data and selected a log-logistic model based on statistical fit. The ERG generally agreed with the company approach and explained that the updated 60‑month ZUMA‑1 data reduced uncertainty in the estimated cure fraction. However, the ERG highlighted that all extrapolations using mixture cure and spline models were higher than ZUMA‑1 Kaplan–Meier data from 50 months onwards. It argued that this was potentially overestimating long-term survival. Clinical experts thought the company extrapolations fitted well with their experience of how CAR T‑cell therapies perform in clinical practice. They explained that CAR T‑cell therapies are expected to provide a durable response in a minority of people, and the plateau present in the extrapolations seemed reasonable. The committee concluded that the company's overall survival extrapolation modelling approach was appropriate.

3.9 The company extrapolated progression-free survival using a standard Gompertz distribution. The company noted that this was the distribution used in the original appraisal and was based on the updated progression-free survival results at 2 years. The committee heard that the ERG had requested progression‑free survival data beyond 2 years, but the company explained this was not collected (see section 3.3). The company further explained that, because overall survival data was mature and stable, further progression-free survival data was not likely to impact cost-effectiveness estimates. The NICE technical team and ERG outlined that, because of the heavy censoring in the progression-free survival curve and low numbers at risk in the tail of the curve, any plateau in progression-free survival is still uncertain. The technical team highlighted that there were overall survival events between 2 years and 4 years and these events may well feature in the progression-free survival curve. These would cause the progression-free survival curve to drop. Clinical experts explained that people whose condition had not progressed by 2 years would likely remain progression-free. Clinical experts thought the company's axicabtagene ciloleucel progression-free survival extrapolation was plausible and stated they would be surprised to see many people progressing after 2 years. The committee noted that other plausible extrapolations increased the incremental cost-effectiveness ratio (ICER). The committee also asked why the company hadn't used mixture cure models for progression-free survival as it had for overall survival. The committee would have preferred longer-term progression data, as outlined in the terms of engagement, and flexible models considered in the base-case analysis. The committee concluded that the company analysis was acceptable but associated with uncertainty.

3.10 The company used SCHOLAR-1 data to extrapolate long-term overall and progression-free survival data for the comparator (salvage chemotherapy). For overall survival, the company used a standard generalised gamma parametric model fit to adjusted SCHOLAR-1 data in the base case. To extrapolate progression-free survival the company generated an estimate by applying the ratio between axicabtagene ciloleucel overall survival and progression-free survival to salvage chemotherapy overall survival. These were consistent with the methods used in the original appraisal. Overall and progression-free survival extrapolation curves showed a steep drop over the first 24 months before levelling off. Clinical experts explained that these extrapolations were clinically plausible for people who had the available treatment at the time. They added that the proportion of people alive in the extrapolation using SCHOLAR-1 data and people who remain progression-free using the company's methods were broadly consistent with other studies. The committee would have preferred to see consistent methods across outcomes for the 2 treatments but acknowledged limitations of available data. The committee concluded that the company's approach to modelling long-term survival for people having salvage chemotherapy was appropriate.

3.11 The company used a 'bottom-up' costing approach to calculate the cost of delivering axicabtagene ciloleucel in the NHS. The company included the costs of:

3.12 The company's base-case analysis was updated after technical engagement and aligned with the ERG's preferred analysis. Because there are confidential discounts for treatments included in best supportive care, the exact ICERs cannot be reported here. The committee concluded that the most plausible probabilistic ICER, using the £41,101 CAR T‑cell therapy delivery cost, was below £50,000 per quality-adjusted life year gained.

3.13 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 original appraisal the committee concluded that axicabtagene ciloleucel met both criteria for end of life. The company updated ZUMA‑1 survival data for axicabtagene ciloleucel for this review, but salvage chemotherapy data was not updated. The committee concluded that axicabtagene ciloleucel offers more than 3 months' extension to life for a population that has a life expectancy of less than 24 months. The committee concluded that axicabtagene ciloleucel continues to meet the end of life criteria.

3.15 New evidence was considered from the ZUMA‑1 trial and the Cancer Drugs Fund SACT data (see section 3.3 and section 3.4). The committee recognised that there was some uncertainty in the company's, progression-free survival extrapolation and limited available data on the comparator (see section 3.5 and section 3.6). The committee also acknowledged uncertainty in the true cost of providing CAR T-cell therapies in the NHS (see section 3.11). However, the cost-effectiveness estimates for axicabtagene ciloleucel compared with salvage chemotherapy were below what NICE considers a cost-effective use of NHS resources. This is in the context of the end of life criteria being met. Axicabtagene ciloleucel is therefore recommended as an option in routine commissioning for treating diffuse large B-cell lymphoma or primary mediastinal large B-cell lymphoma after 2 or more systemic therapies.