Cost-effectiveness analysis of KTE-X19 CAR T therapy versus real-world standard of care in patients with relapsed/refractory mantle cell lymphoma post BTKi in England

Abstract

Aims

The objective of this study is to estimate the cost-effectiveness of KTE-X19 versus standard of care (SoC) in the treatment of patients with relapsed/refractory (R/R) mantle cell lymphoma (MCL) post-Bruton tyrosine kinase inhibitor (BTKi) treatment from a UK healthcare perspective.

Materials and Methods

A three-state partitioned survival model (pre-progression, post-progression and death) with a cycle length of one month was used to extrapolate progression-free and overall survival over a lifetime horizon. Population inputs along with KTE-X19 (brexucabtagene autoleucel) efficacy and safety data were derived from the single-arm trial ZUMA-2 (NCT02601313). The composition of SoC was informed by a literature-based meta-analysis, SoC efficacy data were obtained from the SCHOLAR-2 real-world study. Survival was modelled using standard parametric curves for SoC and a mixture-cure methodology for KTE-X19. It was assumed that patients whose disease had not progressed after five years experienced long-term remission. Costs, resource use and utility, and adverse event disutility inputs were obtained from published literature and publicly available data sources. An annual discount rate of 3.5% was applied to costs and health outcomes. Modelled outcomes for KTE-X19 and SoC included expected life years (LY), quality-adjusted life years (QALY) and total costs. Deterministic and probabilistic sensitivity analyses and scenario analyses were performed.

Results

Estimated median survival was 5.96 years for KTE-X19 and 1.38 for SoC. Discounted LYs, QALYs and lifetime costs were 8.27, 5.99 and £385,765 for KTE-X19 versus 1.98, 1.48 and £79,742 for SoC, respectively. The KTE-X19 versus SoC cost per QALY was £67,713 and the cost per LY was £48,645. Influential scenario analyses use alternative KTE-X19 survival curves and discount rates, and shorter time horizons.

Conclusion

Considering the survival and quality of life benefits compared to SoC, KTE-X19 for R/R MCL appears as a cost-effective treatment in the real-world UK setting.

Keywords:

• Chimeric antigen receptor T-cell
• KTE-X19
• mantle cell lymphoma
• relapsed refractory mantle cell lymphoma
• CD19 antigens
• T-cell therapy
• cost-effectiveness
• brexucabtagene autoleucel

JEL Classification Codes:

• C51
• C5
• C
• I10
• I1
• I

Introduction

As a rare form of non-Hodgkin lymphoma (NHL), mantle cell lymphoma (MCL) originates from the accumulation of malignant B-cells in the mantle zone of lymph nodes. In the United Kingdom (UK), approximately 560 people are diagnosed with MCL annually, representing approximately 5% of all people diagnosed with NHL¹. With a median age at diagnosis of 72.9 and a male-to-female ratio of 2.6:1, MCL most often affects older men¹.

Current treatment options for MCL in the UK are defined by the British Society for Haematology and National Institute for Health and Care Excellence (NICE) which have established a clear treatment pathway for patients receiving first and second-line treatment for MCL^2,3. In the first line, patients are recommended to receive a high-dose cytarabine regimen followed when possible by an autologous stem cell transplant (auto-SCT) with or without rituximab maintenance^2,4. Patients who are not eligible for auto-SCT will instead receive immunochemotherapy with or without rituximab^2,3. Patients who are refractory or relapse are eligible to receive treatment with Bruton tyrosine kinase inhibitor (BTKi) ibrutinib^3,4. Current treatment options for higher subsequent relapses are not well established. The use of alternative immunochemotherapy to that adopted in the first line, but also treatment with venetoclax and lenalidomide, commonly results in inferior response and rapid progression^4–7.

With current treatment, MCL is generally incurable⁸, and patients typically relapse, obtaining a worse prognosis with each subsequent treatment line^8–11. A recent UK real-world analysis of patients receiving ibrutinib at first relapse, for instance, estimated median survival of 23.9 months post-BKTi initiation¹². However, the prognosis for patients who have relapsed or are refractory to ibrutinib is extremely poor, with the median second line and subsequent survival across treatments in the UK being estimated at 9.6 and 7.5 months, respectively^9,13.

KTE-X19 (brexucabtagene autoleucel), a chimeric antigen receptor (CAR) T-cell therapy directed against CD19, is proposed as a treatment option for relapsed/refractory (R/R) MCL post two or more lines of systemic therapy including a BKTi. The efficacy and safety of KTE-X19 have been studied in phase 2, multicenter, single-arm open-label study (ZUMA-2, NCT02601313) in patients with R/R MCL and ≤5 prior therapies, including a BTKi^14,15. Patients underwent leukapheresis and conditioning chemotherapy followed by a single infusion of KTE-X19. At a median follow-up of 17.5 months, KTE-X19 was associated with durable responses, with an objective response rate of 92% in the safety population¹⁶. In a more recent data set with a median follow-up of 25.5 months, KTE-X19 continued to be associated with durable responses with an objective response of 91% and a complete response rate of 68% in the safety population (Supplementary Table 6). In this recent dataset, the median duration of response was 24.8 months, progression-free survival (PFS) was 25.3 months and median overall survival (OS) had not yet been reached (Supplementary Table 6).

This paper investigates the comparative effectiveness and cost-effectiveness of KTE-X19 versus standard of care (SoC) containing cytotoxic chemotherapy, proteasome inhibitors, immunomodulatory drugs and Bcl-2 protein inhibitors for the treatment of patients with R/R MCL in the post-BTKi setting in England. While the analyses were conducted from the perspective of the National Health Service (NHS), personal and social services perspective, the modelled results may allow generalization to other settings.

Materials and methods

Overview

Based on the health-economic model submitted to the National Institute for Health and Care Excellence (NICE) in the UK¹⁷, a partitioned survival model was built to reflect the clinical pathway, health outcomes and costs of patients with R/R MCL post-BTKi (such as ibrutinib), driven by their PFS and OS. Partitioned survival models have previously been used for the modelling of MCL and other oncology indications, as rather than defining transition probabilities, the model utilizes parametric survival models for OS and PFS to model health state membership¹⁸. The modelled population reflected the ZUMA-2 safety population in terms of patient characteristics at baseline (see Supplementary Table 1 for patient characteristics). The intervention KTE-X19 was compared to SoC, which contained multiple agents reflecting clinical practice (see Table 1). In contrast to the model submitted to NICE, SoC efficacy was informed by the real-world study SCHOLAR-2, which was conducted in several European countries¹⁹ (see Supplementary Table 1 for patient characteristics) rather than by a literature-based meta-analysis. SoC composition in the SCHOLAR-2 study was not initially reported¹⁹, therefore, the SoC composition was informed by a literature-based meta-analysis of SoC used for the treatment of R/R MCL post BTKi²⁰, which agreed with the composition observed in the latest analysis of SCHOLAR-2 (see Supplementary Table 2). In the model, the SoC basket comparator contained cytotoxic chemotherapy (bendamustine, cytarabine and doxorubicin-hydrochloride), proteasome inhibitors (bortezomib), immunomodulatory drugs (lenalidomide, rituximab) and Bcl-2 protein inhibitors (venetoclax)²⁰. Although allogeneic stem cell transplantation was used as a consolidation strategy for patients with chemosensitive disease³, it did not contribute to SoC as only a fraction of patients would be eligible due to its substantial toxicity and feasibility limitations including inadequate response to conditioning chemotherapy prior to consolidation with stem cell transplantation, inability to find a suitable stem cell donor and graft vs host disease^3,21. KTE-X19 efficacy was informed by data analysis of the ZUMA-2 trial done at a median of 25.5 months of follow-up (Supplementary Figure 1), thus making use of longer follow-up data of the ZUMA-2 trial compared to the model reviewed by NICE. Mixture-cure modelling (MCM) methodology provided survival curves for KTE-X19 and standard parametric models provided survival curves for SoC to inform a three-state partitioned survival model. In line with UK guidelines, the cost-effectiveness model took a lifetime horizon and applied a 3.5% discount annually to both costs (2020 UK pound sterling, £) and health outcomes²². The cycle length used was one month and a half-cycle correction was applied. The primary outcomes of the model were the incremental cost-effectiveness and cost-utility ratios (ICER and ICUR) of KTE-X19 versus SoC, calculated based on the cost per life year (LY) and cost per quality-adjusted life year (QALY), respectively.

Table 1. Model input data and sources.

Model parameter	Value	Reference
KTE-X19 treatment costs ($)
KTE-X19	£316,118	Kite pharma list price
Leukapheresis	£1,521	NHS reference cost^34 (SA34Z and SA18Z)
Fludarabine (conditioning and bridging chemotherapy, £/admin)	£166.40	BNF^33
Cyclophosphamide (conditioning and bridging chemotherapy, £/admin)	£11.44	BNF^33
Standard of care treatment distributions^
Rituximab^a	56.7%	Based on the studies used in the literature-based meta-analysis^19
Venetoclax^b	12.7%
Bendamustine^c	35.0%
Bortezomib^d	6.4%
Lenalidomide^e	15.9%
Cytarabine^f	32.5%
Doxorubicin (Anthracycline)^g	3.2%
Standard of care acquisition costs (£/admin)
Rituximab	£705.49	BNF^33
Venetoclax	£8.55
Bendamustine	£11.73
Bortezomib	£39.15
Lenalidomide	£33.11
Cytarabine	£22.39
Doxorubicin (Anthracycline)	£5.13
Standard of care administration costs (£/admin)
Intravenous	£241.06	NHS reference cost^34 SB12Z
Hospitalization
KTE-X19: Intensive care unit (%)	22.7%	NICE axi-cel^25
KTE-X19: Non-intensive care unit (%)	77.3%	NICE axi-cel^25
KTE-X19: hospitalization duration (days)	21.2	ZUMA-2 trial
Hospitalization costs (£)
Intensive care unit	£1,428.09	NHS reference cost^34 XC01Z-XC07Z
Non-intensive care unit	£316.81	NHS reference cost^34 SA31A-F
Health state costs (£ per cycle)
Pre-progression	£349.76	Model calculations*
Pre-progression (long-term remission, ≥60 months progression-free)	£28.58	Model calculations*
Post-progression	£588.51	Model calculations*
End-of-life
Palliative care days	37	Bennett et al.^37
Palliative care cost per day	£447.00	Curtis and Burns^38
Adverse event costs (£)
KTE-X19	£12,691	Model calculations*
Standard of care	£0.00	Model calculations*
Health state utility values
Pre-progression	0.780	NICE ibrutinib, 2016^36
Pre-progression (long-term remission, ≥60 months progression-free)	0.805	Calculated from Ara and Brazier, 2010^39
Post-progression	0.680	NICE ibrutinib, 2016^36

^The treatment distributions for SoC sum to more than 100% as patients can be given combination treatment; ^aOne dose every 4-week cycle until progression or maximum 6 cycles; ^bDaily dose of venetoclax until progression, starting at 20 mg daily and ramping up weekly to 1,200 mg by week 7⁵; ^cTwo doses every 4-week cycle until progression or maximum 6 cycles; ^dFour doses every 3-week cycle for nine cycles or until progression; ^eDaily dose for 3 weeks every 4-week cycle until progression; ^fThree doses every 4-week cycle until progression or maximum 6 cycles; ^gOne dose every 3-week cycle until progression.

* Model calculations are outlined in Supplementary Table 3.

Model structure

A partitioned survival model with health states for ‘pre-progression’, ‘post-progression’ and ‘death’ was developed. The model structure is presented in Figure 1 and further described in Simons, Malone²³ where it was used in R/R MCL patients. Based on the 25.5 months ZUMA-2 trial data on PFS and OS (Supplementary Figure 1), the model estimated the transition of KTE-X19 patients through the health states by using MCM.

Figure 1. Model schematic.

Survival

KTE-X19

Data from the 25.5 months follow-up of the safety population (n = 68) in the ZUMA-2 trial were used to model the efficacy of KTE-X19 (brexucabtagene autoleucel, Supplementary Figure 1), this population excluded patients who received conditioning chemotherapy but did not receive KTE-X19. MCM was fitted based on maximum likelihood and selected using the goodness of fit criteria including the Akaike information criterion (AIC) ²⁴. The standard parametric models fitted demonstrated similar goodness of fit with the data. The MCM methodology was chosen to reflect the response dichotomy seen with the use of CAR T therapies, with a proportion of patients in the clinical trial setting appearing to show a deep and durable response with follow-up of 5-years^23,25–29. Such models had been used in previous CAR T publications and aligned with recently published data suggesting that a proportion of R/R MCL patients may experience long-term remission after treatment with KTE-X19^{16,23,25,27,28}. The MCM methodology modelled survival conditional on this cure assumption by deriving a proportion of patients that are cured and at risk for dying only from background mortality (further described in Simons, Malone²³) Survival in the non-cured proportion of patients who remain at risk of disease progression and death from R/R MCL was modelled by fitting standard parametric survival curves^30,31. The estimated percentages of patients experiencing long-term remission, as well as estimated median PFS and OS values, were consistent across the models. The exponential MCM was selected to extrapolate PFS and OS over the model’s lifetime horizon as it minimized AIC indicating the best fit with the trial data, and required the least number of parameters (see Supplementary Figure 1 and Supplementary Figure 2). To further reflect the presence of these latent groups, the model assumed that patients in the ‘pre-progression’ health state experience long-term remission after 5 years, which affected their utilities and healthcare resource use. This assumption has been applied in previous CAR T publications in other indications^28,29. In a scenario analysis, the sex- and age-matched background mortality used in the OS curve was increased by a factor of 1.09 as described by Maurer et al.³², to account for the possibility that prior treatment with chemotherapy and other treatments (such as SCT) may negatively impact survival in patients in R/R MCL post-BTKi long-term remission.

Standard of care

Survival outcomes with SoC were estimated based on data from a real-world study in patients with R/R MCL (SCHOLAR-2, N = 59) in France, the UK, Germany, Italy, Spain, Sweden and Denmark¹⁹. SCHOLAR-2 was a retrospective chart review study designed to collect data on the outcomes of adult patients with R/R MCL who received active treatment in the routine care setting after failing or being intolerant to BTKi therapy¹⁹. The primary objective of SCHOLAR-2 was to estimate OS among real-world patients in the post-BTKi setting. OS was measured from the cohort index date to the date of death from any cause; patients who were known to be alive or had been lost to follow-up were censored at the last date they were known to be alive.

Standard parametric models, adjusting for background mortality, were fitted to the OS data. All curves except for the gamma curve converged. The exponential curve was chosen for the extrapolation of OS over time based on the shape of the distribution which fitted well with the entire trial data, and considering the goodness of fit statistics indicated the exponential curve provided the second-best fit after the Weibull curve which showed less good fit in the tail of the curve (see Supplementary Figure 3). PFS data was not collected in SCHOLAR-2, so to estimate PFS with SoC, a hazard ratio adjustment was applied to the OS survival estimate. This ratio of 0.727 was informed based on the OS and PFS results of a previously conducted literature-based meta-analysis in R/R MCL²⁰. The ratio was assumed to be constant over time (see Supplementary Figure 4).

A propensity score-matched analysis was conducted to reduce differences in patient characteristics between the ZUMA-2 and SCHOLAR-2 studies¹⁹. SCHOLAR-2 patients were matched to the ZUMA-2 safety cohort by means of a logistic regression model predicting treatment allocation. The model included patient gender, age, number of prior treatments, prior auto-SCT, prior BTKi duration and response to BTKi, and stage four R/R MCL as covariates, with the choice of covariates being determined based on clinical relevance, availability, and statistical considerations such as the overlap of distributions of propensity scores and sufficient variability¹⁹ (see Supplementary Table 1 for matched patient characteristics). This reduced the effective sample size to 40.1 and resulted in only minor changes to the OS curve as the matched OS curve shifted slightly downwards during the first 20 months, aligned with the unmatched curve between 20 and 32 months of follow-up and then shifted slightly downwards again during the last months of follow-up (see Supplementary Figure 5). This implies that the outcomes of the matched SCHOLAR-2 population were similar or somewhat inferior to those of the unmatched SCHOLAR-2 population. Likewise, the hazard ratio between SCHOLAR-2 and ZUMA-2 before matching (0.37, confidence interval [CI] 0.20; 0.66) was higher compared to the hazard ratio after matching (0.33, CI 0.18; 0.59). As a conservative approach, the unmatched SCHOLAR-2 data was used in the economic model¹⁹.

Cost and resource use

The model considered costs of treatment with KTE-X19, costs of SoC treatment, adverse event (AE) costs, and costs related to disease management and end of life.

KTE-X19 and related costs and resource use

Costs of treatment with KTE-X19 included drug acquisition cost based on the manufacturer’s list price of £316,118 for a single infusion of 2 × 10⁶ anti-CD19 CAR T cells/kg¹⁴, along with leukapheresis, conditioning and bridging therapy and hospitalization costs, leading to total costs per KTE-X19 administration of £341,003. Conditioning chemotherapy was given prior to the administration of KTE-X19 to induce lymphodepletion and consisted of fludarabine (30 mg/m² intravenous [IV]) and cyclophosphamide (500 mg/m² IV)¹⁴. Conditioning chemotherapy was assumed to occur in the inpatient setting. Based on ZUMA-2 trial data, hospitalization lasted 21.2 days (see Supplementary Table 3) and consisted of 22.7% of days in the intensive care unit (ICU), and 77.3% of days in non-ICU care²⁵. A proportion of patients received bridging therapy for disease control during CAR T manufacturing, the proportion of patients was obtained from the ZUMA-2 trial (36.8%)¹⁴. Costs of bridging therapy with a cycle of rituximab, bendamustine and cytarabine between leukapheresis and KTE-X19 infusion were included to reflect the most common clinical practice in the UK¹⁷.

A retreatment cost item was included to account for retreatment with conditioning chemotherapy in 7.4% of patients (see Supplementary Table 3), it was assumed that the second round of leukapheresis was not costed. All model input values and sources relating to costs and resource use with KTE-X19 treatment are summarized in Table 1.

Standard of care and related costs and resource use

The treatment regimens included in SoC are outlined in Table 1 based on the definition of SoC in the literature²⁰, which had been validated by UK clinical experts. The literature-based definition of SoC aligned with the treatments used in the SCHOLAR-2 study, which are presented in Supplementary Table 2. Costs of SoC consisted of acquisition costs based on British National Formulary (BNF, May 2021) costs³³. Administration cost applied to IV treatments and was informed by 2018–2019 NHS reference costs for IV administration of simple parenteral chemotherapy at first attendance³⁴. Drug dosage and treatment duration for each treatment were taken from the respective guidance, except for venetoclax, for which there was no indicated posology for the treatment of MCL, whereby the average maximum dose in Eyre, Follows⁵ was used. Dose intensity of 100% with no vial sharing was generally assumed, a dose intensity of 70% was explored in a scenario analysis³⁵. All model input values and sources relating to costs and resource use with SoC treatment are summarized in Table 1.

Disease management costs and resource use

Evidence on health state-related health care resource utilization was not collected in the ZUMA-2 trial and was assumed to be equal to the health state-related health care resource utilization reported in the NICE technology appraisal for ibrutinib for R/R MCL [TA502] ³⁶. Evidence used in the appraisal stemmed from a survey amongst active practicing hematologists and oncologists in the NHS and was validated by UK hematologic experts. Table 1 outlines the cost in the pre-progression and post-progression health states. Health state-related costs were assumed independent of treatment and applied until the patient died or reached long-term remission. Long-term remission was assumed to commence after 60 months in pre-progression. The resource use of patients in long-term remission was assumed to include two doctor’s visits per year at a unit price of £171.46 per visit, see Supplementary Table 3 for details.

Adverse events

AEs of grade 3 or 4 with an incidence of ≥5% based on the ZUMA-2 trial were modelled for KTE-X19 (see Supplementary Table 4, additional details on AEs in Supplementary Table 5). Additionally, cytokine release syndrome (CRS) of grade ≥2 (61.8% of patients) was modelled by assuming treatment with tocilizumab, CRS-related hospitalization in the ICU was assumed to be included in the ICU care component of KTE-X19 administration. AEs were costed in line with NHS reference costs³⁴, hypogammaglobulinemia (20.6% of patients) was assumed to be treated with 0.5 g/kg of IV immunoglobulin every 4 weeks for 12 months, and BNF prices were applied³³. Although AEs occurred with the individual SoC treatment components, limited safety data were identified from the literature and consequently, no AEs were modelled for SoC as a conservative assumption.

End of life costs

Costs relating to intensified care towards the end of life are applied to all patients irrespective of their cause of death and prior treatment. A fixed cost of £16,539 was applied reflective of 37 days of palliative care at a cost of £447 per day (see Table 1)^37,38.

Utilities

In absence of utility values for R/R MCL patients post BTKi, health state utilities from the ibrutinib NICE submission in R/R MCL overall are used and applied irrespective of treatment. The utility of patients experiencing long-term remission was assumed to be equal to an age-matched UK general population and was informed by published estimates (see Table 1)³⁹. Pre-progression and post-progression utilities were adjusted not to exceed the utility of age- and gender-matched UK population. QALYs were derived from the multiplication of each health state’s utility and the time spent in the health state³⁶.

Utility decrements were used to reflect the impact of AEs following the treatment with KTE-X19. These were calculated based on the incidence of the AE, the applicable utility decrement, and the duration of the AE. Decrements and duration were informed by patient-level data of the ZUMA-1 trial used in the NICE assessment of axicabtagene ciloleucel for the treatment of diffuse large B-cell lymphoma and primary mediastinal B-cell lymphoma after 2 or more systemic therapies²⁵. The incidence of the AEs was informed by the safety population of the ZUMA-2 trial (see Supplementary Figure). In case no disutility was identified for an AE, the highest disutility of the remaining AEs (excluding CRS) was used as a proxy. Decrements were applied during the first model cycle, and in parallel to KTE-X19 administration.

Outcomes

The outcomes of treatment with KTE-X19 vs. SoC in terms of OS expected LYs, QALYs and total costs were calculated and presented alongside incremental costs and effects, and the ICER and ICUR. All presented results are discounted and considered a lifetime horizon.

Uncertainty analyses

In addition to deterministic results, results of probabilistic sensitivity analysis are presented where the values of model parameters were varied simultaneously within their individual uncertainty distributions. In case distributional information was not available from the original source, a standard error of 20% was used as a proxy.

A univariate sensitivity analysis was conducted where model parameters were varied one at a time to the lower and upper 95% confidence interval. Results of the analysis are presented in tornado diagrams including the ten most influential parameters and depicting their impact on the incremental costs, effects and the ICUR. The probabilistic sensitivity analysis is presented on a cost-effectiveness plane and the probability that each treatment is cost-effective at different levels of willingness-to-pay per QALY is presented using a cost-effectiveness acceptability curve (CEAC).

Scenario analyses

Scenario analyses were conducted to explore the main assumptions of the model, which included:

• Removing the long-term remission assumption regarding utilities and costs, and adjusting the long-term remission point to 2 years and 10 years,

• Use of SCHOLAR-2 survival data matched to the ZUMA-2 safety set,

• Use of ZUMA-2 survival data based on the intention-to-treat set i.e. including patients who received conditioning chemotherapy but did not receive KTE-X19 (see Supplementary Figures 6 and 7, and Supplementary Table 6),

• Use of ZUMA-2 survival data based on the intention-to-treat set and use of SCHOLAR-2 survival data matched to the ZUMA-2 ITT set, (see Supplementary Figures 6 and 7)

• Use of standard parametric survival models for KTE-X19,

• Inclusion of an adjustment of the background mortality,

• Lowering discount rates to 1.5% for health outcomes and costs,

• Assuming KTE-X19 acquisition and leukapheresis costs apply upon retreatment with KTE-X19,

• Assuming a 70% dose intensity for SoC treatments,

• Increasing the costs of neurotoxic AEs encephalopathy and confusional state,

• Reducing the time horizon to 10, 20 and 30 years.

Results

Base case results

Table 2 outlines the deterministic base case results. Survival at 24 months was 66.8% with KTE-X19 and 39.2% with SoC, and survival at 60 months was 52.0% and 9.3%, respectively. The total discounted LYs were larger for KTE-X19 than for SoC, with 8.27 and 1.98 LYs, respectively. Similarly, the total QALYs were 5.99 and 1.48, respectively, with an incremental QALY of 4.52. Survival and QALYs, as well as incremental benefits with KTE-X19, were largely accrued in the pre-progression health state. Total costs were higher for KTE-X19 (£385,765) than for SoC (£79,742), driven by drug acquisition costs of KTE-X19. Acquisition costs contributed 82% and 64% of KTE-X19 and SoC total costs and represented 87% of the incremental costs. The resulting ICUR was £67,713/QALY gained and the resulting ICER was £48,645/LY gained.

Table 2. Base case results (discounted).

	KTE-X19	Standard of care	Incremental difference
Median overall survival	5.96	1.38	4.58
Total life years	8.27	1.98	6.29
Pre-progression	6.62	1.46	5.15
Post-progression	1.65	0.51	1.14
Total quality-adjusted life years	5.99	1.48	4.52
Pre-progression	4.99	1.13	3.85
Post-progression	1.06	0.34	0.72
Adverse events	−0.05	0.00	−0.05
Total costs	£385,765	£79,742	£306,023
Treatment-related costs
Drug acquisition	£316,118	£50,986	£265,132
Apheresis	£1,521	£0	£1,521
Drug administration	£4,652	£3,500	£1,152
Conditioning chemotherapy	£6,275	£0	£6,275
Bridging therapy	£373	£0	£373
Hospitalization	£12,064	£0	£12,064
Disease management costs
Pre-progression	£11,857	£5,973	£5,884
Post-progression	£11,652	£3,624	£8,028
Other costs
End of life care	£12,092	£15,659	-£3,567
Adverse events	£9,161	£0	£9,161
Cost per life years	£67,713	–	–
Cost per quality-adjusted life years	£48,645	–	–

Uncertainty analyses

Results of the probabilistic sensitivity analysis showed costs and effects comparable to the deterministic results, see Figure 2. The resulting ICUR was £68,733. The CEAC shows the probability of KTE-X19 being cost-effective at different willingness-to-pay thresholds; see Figure 3.

Figure 2. Probabilistic cost-effectiveness plane.

Figure 3. Cost-effectiveness acceptability curves.

The univariate sensitivity analysis indicated that the incremental cost results were most sensitive toward the ICU day costs, the proportion of ICU days, the costs of conditioning chemotherapy and the cost of treatment for hypogammaglobulinemia (Figure 4). The main driver of the incremental utilities was the pre-progression utility for patients with long-term remission, followed by the utility for post-progression and pre-progression prior to long-term remission (Figure 5). The most influential parameters on the ICUR were the pre-progression utility for patients in long-term remission and the utility post-progression (Figure 6).

Figure 4. Univariate sensitivity analysis – influential parameters on incremental costs.

Figure 5. Univariate sensitivity analysis – influential parameters on incremental QALYs.

Figure 6. Univariate sensitivity analysis – influential parameters on the ICUR.

Scenario analyses

Detailed results of all scenario analyses can be found in Supplementary Table 7. The most impactful scenario analysis on the ICUR included the use of standard parametric survival curves for the modelling of KTE-X19 effectiveness, which increased the ICUR by £13,566. Reducing the time horizon to 10 years increased the ICUR to £131,804 and reducing discount rates decreased the ICUR by £12,804. The remaining scenarios impacted the ICUR by less than 10%.

Discussion

The analyses investigated the comparative effectiveness and cost-effectiveness of KTE-X19 (brexucabtagene autoleucel) versus SoC for the treatment of patients with R/R MCL post-BTKi in England. This cost-effectiveness model suggests that in real-world use, compared with SoC, KTE-X19 may extend median survival by 4.58 years (5.96 vs. 1.38 years) and result in 4.52 additional discounted QALYs over a lifetime horizon (5.99 vs 1.48 QALYs). The majority of these gains (85%) were made pre-progression. The incremental costs of KTE-X19 were £306,023, with drug acquisition costs of KTE-X19 being the main cost driver. The cost per QALY gained was £67,713 and the cost per LY gained was £48,645. The results provide evidence that KTE-X19 may be a valuable use of resources for the treatment of R/R MCL post-BTKi in England with the potential to provide significant QALY gains at an acceptable cost level. Sensitivity analyses showed the robustness of the results against changes in model parameters and alternative assumptions.

The utility of patients in long-term remission had the largest impact on health outcomes and on the ICUR. This can be explained by the difference in proportions of patients in long-term remission between KTE-X19 and SoC, which determines that changes in long-term remission utilities promptly translate into changes in incremental QALYs. The cost per ICU day had the largest impact on incremental costs and the third largest impact on the ICUR, explained by the fact that ICU costs only apply to the KTE-X19 arm whereby any changes in ICU costs translate into changes in incremental costs. Scenario analyses indicated that results were stable with the use of most alternative assumptions, the use of shorter time horizons increased the ICER, and the use of standard parametric survival curves for KTE-X19 and the use of lower discount rates had a larger impact on the ICUR, increasing it by £13,500 and decreasing it by £13,000, respectively.

Previous studies have assessed the cost-effectiveness of CAR T in other indications and in other settings^{23,28,29,40–42}, whereby this is the first cost-effectiveness analysis of a CAR T therapy in R/R MCL in England. A previous US-based cost-effectiveness analysis of KTE-X19 for the treatment of R/R MCL using post-BTKi data had shown survival outcomes in the form of 8.99 LYs and 7.39 QALYs, and 4.52 incremental LYs and 3.74 incremental QALYs obtained with KTE-X19 treatment compared to SoC²³. In line with the findings of the present study, a previous cost-effectiveness analysis of axicabtagene ciloleucel in adult patients with R/R large B-cell lymphoma following two or more prior therapies showed a substantial survival gain with CAR T treatment in the form of 6.90 incremental LYs and 6.54 incremental QALYs gained²⁸. ICERs were not compared due to the differences in costs and other settings as the analyses were US-based. Similar assumptions were made regarding patients in long-term remission after 5 years being subject to age- and sex-matched general population mortality. The estimated incremental costs closely align with estimates of other recent cost-effectiveness analysis of CAR T therapies⁴³.

This analysis used uniquely granular inputs which contribute to the robustness of the results. Deterministic, probabilistic and scenario analyses confirmed the results of the base case analysis, and showed the results were relatively stable with regard to changes in inputs and assumptions. The valuable insights into the comparative effectiveness and cost-effectiveness of KTE-X19 versus SoC support the advancement of treatment in this population of R/R MCL patients post-BTKi for whom SoC offers very limited hope.

The study is also subject to uncertainties and limitations. The extrapolation of short-term results over a lifetime horizon inherently introduces uncertainty into the results. Long-term survival outcomes with KTE-X19 are subject to substantial uncertainty considering the relatively low sample size (N = 68), the limited duration of follow-up of the trial data (median of 25.5 months), and the lack of long-term clinical experience with KTE-X19. The results of the analysis should be confirmed with long-term follow-up data and data from real-world application of KTE-X19.

Patients enrolled in ZUMA-2 had failed multiple prior therapies and a placebo control arm would have been deemed unethical given their poor prognosis. SCHOLAR-2 data was used for the analysis of SoC effectiveness in the absence of prospective SoC OS and PFS data in the post-BTKi setting. While the SoC survival outcomes are more mature and better validated against clinical experience than the ZUMA-2 outcomes, SoC survival outcomes should be cautiously interpreted considering the low sample size of SCHOLAR-2 (N = 59). Additionally, the use of SCHOLAR-2 data for survival with SoC treatment prevented an anchoring of the treatment comparison between KTE-X19 and SoC. The unanchored and unadjusted comparison between the treatments gives rise to potential selection bias stemming from potential differences in population characteristics, which may prevent a fair comparison between KTE-X19 and SoC and impact the relative treatment effect. A scenario analysis using propensity score-matched SCHOLAR-2 survival data showed similar results with somewhat lower SoC health outcomes and costs, and an ICUR of £66,760/QALY, indicating the adjustment of key patient characteristics between ZUMA-2 and SCHOLAR-2, including age and number of prior treatments, had little impact on the results. A further limitation and potential source of bias is that the composition of SoC treatment and thus the costs of SoC treatment acquisition was based on a literature-based meta-analysis, while the effectiveness of SoC was obtained from the SCHOLAR-2 study.

While in line with cure assumptions made in previous CAR T models, a universally accepted definition of long-term remission is lacking in R/R MCL, and uncertainty remains surrounding the 5-year cut-off used. Maurer, Ghesquières³² suggested that in patients with diffuse large B-cell lymphoma, event-free survival 24 months after the completion of first-line treatment should be considered a robust end point for the disease-related outcome. No definition of cure in MCL was identified, and clinical expert opinion indicated that a 24-months cut-off would likely not apply to MCL due to the occurrence of late relapses. Given the lack of evidence, assumptions were made regarding the utility and costs of long-term remission. These assumptions were explored in sensitivity analyses, which showed a limited impact of long-term remission utility and cost on the absolute outcomes. A scenario analysis removed the long-term remission assumption and increased the ICUR from £67,713 to £73,203. Clinical evidence on the timing of long-term remission post CAR T treatment for R/R MCL and its impact on costs and utility would be welcomed to address these evidence gaps.

Omitting AEs with SoC as a conservative assumption in the absence of evidence likely underestimates costs of SoC, and to a lesser degree, overestimates SoC QALYs. This likely increases the incremental costs of KTE-X19 and inflates ICER and ICUR.

Conclusions

KTE-X19 (brexucabtagene autoleucel) for the treatment of R/R MCL post-BTKi has shown significant survival benefits over SoC based on clinical and real-world survival data. This cost-effectiveness model adds to the growing evidence supporting KTE-X19 use in R/R MCL, and further research is warranted to address uncertainties regarding the long-term survival and long-term remission of R/R MCL patients post-BTKi treated with KTE-X19.

Transparency

Declaration of funding

This study was funded by Kite, a Gilead Company.

Declaration of financial/other interests

SP is employed by OPEN Health Company and is a consultant for or has consulted for Kite, a Gilead Company, Galderma, PTC therapeutics, Bristol Myers Squibb, Celgene Ltd, a Bristol Myers Squibb Company, GlaxoSmithKline, Otsuka, Merck & Co, and AbbVie in the past two years.

CLS is employed by OPEN Health Company and is a consultant for or has consulted for Kite, a Gilead Company, Galderma, PTC therapeutics, Takeda, Bristol Myers Squibb, EMD Serono, Grifols and AbbVie in the past two years.

CB is employed by OPEN Health Company and is a consultant for or has consulted for Kite, a Gilead Company, AbbVie, Amgen, TEVA, Takeda, PTC therapeutics, Galderma, Bristol Myers Squibb, Celgene, GlaxoSmithKline, Tesaro, Gilead and Otsuka in the past two years.

GC is an employee of Kite, a Gilead Company, was employed previously by Amgen, Mundipharma and Janssen Cilag.

JW is an employee of Kite, a Gilead Company, was employed previously by Amgen, Abbott, Genentech and Elan, and is a stakeholder of Gilead, Amgen, Abbott, AbbVie, Avid Bioservices, Curis, Moderna and Viracta Therapeutics.

SW is employed by Wade Outcomes Research and Consulting and is a consultant for or has consulted for Kite, a Gilead Company, AbbVie/Allergan, Johnson & Johnson, and Amgen in the past 2 years.

RS is an employee of Kite, a Gilead Company and is a stakeholder of Gilead and Amgen.

IK is an employee of Kite, a Gilead Company and is a stakeholder of Gilead.

WP is an employee of Kite, a Gilead Company.

GS has been advising Abbvie, Beigene, BMS/Celgene, Debiopharm, Epizyme, Genentech/Roche, Genmab, Incyte, Ipsen, Janssen, Kite/Gilead, Loxo, Milteniy, Morphosys, Novartis, Rapt, Regeneron, Takeda and Velosbio in the past two years.

MW has been advising Pharmacyclics, Celgene, Janssen, AstraZeneca, MoreHealth, Pulse Biosciences, Nobel Insights, Guidpoint Global, InnoCare and BioInvent over the past two years.

GH has been advising Janssen, Genmab, Gilead/Kite, AstraZeneca, Lilly, Roche, Abbvie and Morphosys/Incyte over the past two years.

Peer reviewers on this manuscript have received an honorarium from JME for their review work but have no other relevant financial relationships to disclose.

Author contributions

All named authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship of this manuscript.

Acknowledgements

The authors would like to thank Nour Chami, (employee of OPEN Health) for the collection of costing inputs used in this study. The authors would also like to thank Greg Maglinte for his contributions during the conception and design of this work.