Cost-Effectiveness of brexucabtagene autoleucel for relapsed/refractory mantle cell lymphoma

Abstract Brexucabtagene autoleucel is a chimeric anti CD19 antigen receptor T-cell therapy that allows durable responses in relapsed/refractory (R/R) mantle cell lymphoma (MCL). The present study compared the clinical and economic outcomes of R/R MCL patients (pre-exposed to ibrutinib and chemoimmunotherapy) treated with brexucabtagene autoleucel versus Rituximab, bendamustine, cytarabine (R-BAC) in the Italian Healthcare System. A partitioned-survival model extrapolated survival and healthcare costs of R/R MCL patients over a lifetime horizon. Discounted and quality-adjusted life expectancy (QALY) was 6.40 versus 1.20 for brexucabtagene autoleucel versus R-BAC and lifetime costs were €411,403 versus €74,415, respectively, which corresponds to a cost of €64,798 per QALY gained. The results were highly sensitive to brexucabtagene autoleucel acquisition cost and to assumptions on long-term survival, therefore the cost-effectiveness of brexucabtagene autoleucel for patients with R/R MCL requires validation with longer follow-up data and in specific risk subgroups.


Introduction
A dismal prognosis is forecasted for patients with mantle cell lymphoma (MCL) who experience treatment failure (R/R) after chemo-immunotherapy and ibrutinib. No standard treatment is available for such patients, who are variably assigned to chemoimmunotherapy, lenalidomide, venetoclax or bortezomib. Several retrospective studies involving patients with R/R MCL pre-exposed to BTK inhibitors, have consistently reported dismal overall survival (OS) [1][2][3][4][5].
KTE-X19 is a chimeric antigen receptor (CAR) T-cell therapy based on autologous anti-CD19-transduced CD3+ cells, which harness the body's own immune system to target neoplastic cells that can be administered to patients with R/R MCL. Autologous patient lymphocytes are collected by apheresis and manipulated at a remote site. Candidate patients receive lymphodepletive therapy before CAR-T reinfusion and are hospitalized for a median of about 3 weeks in order to monitor and treat the occurrence of severe (grade > =3) cytokine-release syndrome and neurotoxicity which occur in 15% and 31% of patients, respectively. Despite such adverse events, ZUMA-2 phase-2 trial reported very low treatment-related mortality (2 out of 74 patients), a very high response rate (93% overall objective response rate and a complete response in 67% of the treated patients), with median progression-free survival (PFS) of 25.8 months [6,7]. An indirect comparison of OS from ZUMA-2 (higher than 60% at 36 months) with patients receiving standard-of-care (SOC) treatments reported propensity-adjusted hazard ratios (HR) ranging from 0.32 to 0.49 [1]. Based on such data, international transplant scientific societies (NCCN, ESMO) recommended CAR-T for R/R MCL patients, and marketing authorization for KTE-X19 was granted by FDA, EMA, and the Italian Agency for Drugs [8,9].
Despite MCL is a rare disease with a reported incidence rate of 0.72/100,000/year, representing about 476 newly diagnosed cases in Italy each year [10], the opportunity cost of KTE-X19 needs to be balanced against all the other novel treatments available. Therefore, we used a partitioned-survival model to extrapolate PFS, OS, and healthcare costs of R/R MCL patients in the perspective of the Italian HealthCare System. The simulated patient cohort was assigned either to KTE-X19 or to rituximab-bendamustine-cytarabine (R-BAC), which is the most commonly used treatment in Italy for R/R MCL patients failing ibrutinib [3].
Here we report the results of the above pharmacoeconomic assessment of KTE-X19 in the Italian healthcare setting.

Methods
A partitioned survival mixture cure model was developed to estimate monthly patient transition through three health states, namely pre-progression, post-progression, and death. A latent "cured fraction" of patients was estimated [11]: patients who survived and whose disease has not progressed after 5 years, i.e. "cure time-point", were assumed to experience long-term remission [12]. The population entering the model reflected the ZUMA-2 population and model inputs included age (63.2 years), sex (84% males), weight (81.8 Kg), and body surface area (1.98 sqm). Italian age-and sex-specific mortality rates were driven from 2018 national life tables.
Patients were assigned to pre-progression state based on PFS curves from ZUMA-2 and phase 2 studies for R-BAC [3,13]. PFS of patients assigned to brexucabtagene autoleucel was based on estimates from the second data cut (31 December 2020) for the safety population (68 patients), that is the population who received the infusion of CAR-T cells. Standard parametric models were fitted to data and, for the base case, the exponential model was selected and extrapolated over a lifetime horizon since it reported a similar model fit and cure fraction (52.14%) as other parametric models but requires the smallest number of parameters. Log-normal distribution for PFS was the best-fitting model for R-BAC survival curves and was adopted without a fraction of cure. A subset of patients achieving a response to R-BAC was allowed to receive allogeneic SCT: such a proportion was 31% in the retrospective study reported by McCullogh, therefore the same proportion was used at sensitivity analysis [3].
Patients were assigned to the death state based on OS curves corrected for background mortality, in order to keep OS lower than the age-and sex-adjusted general population OS. Lognormal distribution for OS was the best fitting model for R-BAC, while exponential parametric curves were fitted to ZUMA-2 18-mo follow-up data. After the cure point, patients were assigned background mortality [14,15], however, at sensitivity analysis it was adjusted with a hazard ratio of 1.09 [16].
Finally, the post-progression state was estimated as the difference between PFS and OS.
Healthcare resource consumption for patients assigned to R-BAC included the charge for outpatient parenteral administration, based on national charges, and ex-factory drug costs. However, 10% of the patients assigned to RBAC were assumed to achieve a good response to RBAC and to undergo allogeneic stem cell transplant (SCT) [17]. The cost of allogeneic SCT was not estimated uniquely based on national charges for hospital stay (DRG 481) since the overall treatment pathway included initial workup and post-transplant management, which increases the overall costs to threefold higher than the hospital charge for the transplant itself [18].
Healthcare consumptions of patients receiving brexucabtagene autoleucel included leukapheresis, conditioning chemotherapy (outpatient setting) and a single infusion of 2*10^6 anti-CD19 CAR T cells per kilogram body weight at a cost of €360,000 to which administration and hospitalization costs were added (Table 1). Bridging therapy with R-BAC (1 cycle administered in the outpatient setting) was also included for 36.8% of patients requiring a bridging therapy. Hospitalization length was modeled according to ZUMA-2 data, that is 21.2 days of hospital stay of which 23% were in the ICU. Retreatment was reported in 2.9% of ZUMA-2 trial but was not considered at the base case analysis. Daily cost for non-ICU stay was estimated based on daily charges for DRG481, which is assigned to a hospital stay for cellular therapies, namely autologous and allogeneic transplants. Rather, daily cost of ICU stay was based on a European study including Italy [19]. Additional healthcare resource consumption related to CRS grade 2-4 and selected grade 3-4 adverse events were considered, namely encephalopathy, pneumonia, acute renal failure and sepsis: costs were assigned to the above events base on published studies [20,21]. Healthcare consumption related to another grade 3-4 adverse events was assumed to occur during the hospital stay and was not costed separately.
Costs related to ambulatory care for MCL patients were modeled including periodical visits and imaging or lab tests: their frequency was established by national experts (Table 1). Charges were assumed to be appropriate proxies for the costs related to tests and visits.
The source of utilities for pre-progression was ZUMA-2 trial, which collected EQ-5D-5L data. Due to a lack of Italian EQ-5D tariffs, the UK tariffs were used: as a result, pre-progression utility was 0.824. Post-progression utility was estimated by reducing pre-progression utility by 0.20, as reported in the NICE ibrutinib submission of Ibrutinib for MCL. Utility decrements for adverse events (CRS grade 2-4 and grade 3-4 adverse events occurring in at least 5% of the patients) were based on ZUMA-1 patient level data analysis: utility decrements were multiplied for the event frequency and duration [22]. RBAC arm was charged a disutility of 0.03 QALYs accounting for the portion of patients undergoing allogeneic SCT, based on the post-transplant quality of life reported by a prospective study [23].
The cost-effectiveness analysis was conducted from the Italian Healthcare System perspective over a lifetime horizon and a 3% annual discount was applied to costs and health outcomes. The primary outcome of the model was the incremental cost-utility ratio (ICUR), expressed as cost per quality-adjusted life year gained (QALY).
Both univariate and probabilistic sensitivity analyses were conducted. Inputs were varied using the standard deviation, when available, or a 20% range. Beta distributions were used for probabilities and gamma distributions for costs. The acceptability curve plotted the chance brexucabtagene autoleucel reported of achieving 1 QALY at a cost lower than the willingness to pay.
Several scenario analyses were conducted in order to assess the impact of baseline assumptions, including the choice of comparator therapy and other input parameters. In particular, we examined the assumption of adopting a mix of therapies including R-BAC, acalabrutinib, venetoclax, lenalidomide in the comparator arm of the model. The mix was adapted to a meta-analysis of 4 published studies addressing post-ibrutinib MCL patients [24]. OS was derived from the meta-analysis [2][3][4][5] and PFS from two of the above studies [2,3]. A Gompertz shape for OS and lognormal shape for PFS were assumed for estimating long-term outcomes, based on fitting [25]. The mix was also adapted according to the OS reported in a real-world retrospective study, named SCHOLAR-2, including patient-level data of 59 R/R MCL patients matched to ZUMA-2 enrolled ones. PFS for SCHOLAR-2 patients was estimated based on the PFS to OS ratio reported in the meta-analysis, namely 0.76 [1]. A Weibull distribution was used to model long-term OS in this simulated SOC population. Table 2 outlines discounted base case results. The model estimated that OS at 5 and 10 years was 57% and 49% for brexucabtagene autoleucel and 30% and 20% for R-BAC, respectively ( Figure 1). Median survival was 9.7 years for brexucabtagene autoleucel and 2.1 for R-BAC and discounted progression-free life years increased from 1.46 to 7.09 (+5.63 life years). Quality-adjusted life expectancy significantly improved from 1.20 of patients treated with R-BAC to 6.40 QALYs (+ 5.20) of those treated with brexucabtagene autoleucel: the majority of QALYs were accrued in the pre-progression health state. Also, post-progression lifespan was estimated to be increased by 1.06 QALYs. Total discounted costs were quite low for R-BAC, reflecting the outpatient setting of administration and the low drug cost. Costs were higher for brexucabtagene autoleucel: treatment-related costs were the main cost drivers, contributing to 68% and 86% of brexucabtagene autoleucel and R-BAC total costs, respectively. These values translated to a cost per QALY of €64,798 (€46,498 undiscounted).

Results
As compared to mixed-treatment SOC from real-life SCHOLAR-2, brexucabtagene autoleucel gained 4.72 life years and 6.55 QALYs at a cost of €57,915 per QALY gained ( Table 3). As compared to mixed treatment SOC from meta-analysis of published studies, similarly brexucabtagene autoleucel gained 7.20 life years and 5.18 QALYs at a cost of €56,010 per QALY gained ( Table 3).
The univariate sensitivity analysis indicated that the incremental effectiveness results were most sensitive to the utilities (Figure 2(A)) and to assumptions on long-term survival. In particular, if the projected median OS in patients treated with brexucabtagene autoleucel declined to 46 months, the incremental gain of brexucabtagene autoluecel versus BAT would also decline to 1.2 QALYs. Moreover, by increasing mean age of patients by 5 years, to 68 years, reduced the incremental gain to 4.33 QALYs.
Incremental costs were sensitive to allogeneic SCT costs, ICU daily costs, and resource consumption for hypogammaglobulinemia (Figure 2(B)). Similarly, as the proportion of patients receiving allogeneic SCT after BAT therapy decreased from baseline 31% to 21%, the incremental cost-effectiveness ratio of brexucabtagene autoleucel versus BAT increased to €68,248/QALY. Incremental costs were also sensitive to the acquisition cost of brexucabtagene autoleucel and the rate of re-treatment: a 10% retreatment and re-payment rate would increase the cost-utility ratio to €71,721, while a 20% reduction of brexucabtagene autoleucel acquisition cost would reduce the cost per QALY gained to €50,954. On the converse, an increase of patients requiring bridging therapy from 36.8% to 100% would modestly increase the cost per QALY gained by brexucabtagene autoleucel to €65,157.
Also, some time-related assumptions of the analysis determined its outputs. In particular, the cost-effectiveness ratio minimally increased to €65,651  unexpected since they are typical in pharmacoeconomic assessments of potentially curative therapies. Probabilistic sensitivity analysis (Figure 3) showed that for a willingness-to-pay above the ICUR (i.e. €64,798 per QALY gained), brexucabtagene autoleucel was cost-effective. At a willingness-to-pay threshold of €87,330 per QALY [25], the probability of brexucabtagene autoleucel was cost-effective was 88%.

Discussion
The treatment of patients with R/R MCL is challenging, with the currently available treatments only offering transient responses, mostly granted to the small proportion of patients receiving allogeneic SCT. The ZUMA-2 trial has demonstrated that brexucabtagene autoleucel is an effective cellular therapy for this group of patients and real-world experience has confirmed a low rate of CRS [13,26]. However, the benefit-for-cost of this novel immunotherapy may change in different countries, according to commonly adopted SOC therapies and to national unit costs. Therefore, we developed a country-specific cost-effectiveness analysis for the Italian healthcare system, based on an international partitioned survival model. The model compared brexucabtagene autoleucel strategy to R-BAC and was based on the assumption that progression-free patients after 5 years do not incur any relapse, namely are assumed functionally cured. In the lifetime horizon, the model estimated that brexucabtagene autoleucel delivered 5.2 discounted QALYs over R-BAC at a cost of €64,798 per QALY gained. The survival gain of brexucabtagene autoleucel was clinically relevant and the cost-utility ratio proved inferior to €87,330/QALY, which is a common threshold for most cancer drugs reimbursed in Italy [25]. The cost-utility of brexucabtagene autoleucel also compared favorably with similar estimates of axicabtagel ciloeucel and tisagenleucel for R/R diffuse large B cell lymphoma in the Italian healthcare system perspective [26][27][28]. Furthermore, the estimated cost per QALY gained by brexucabtagene autoleucel in the UK setting was quite similar to the one reported by our analysis [29], while the lower incremental cost-utility ratio reported in the US setting was sustained by the high cost of novel BTK inhibitors modeled as SOC since the FDA label was for all RR MCL irrespectively of prior failure to BTKi [24]. Finally, the value-for-cost of this cellular therapy is inferior to other immune-modulating therapies for blood cancers assessed in the same healthcare system [30][31][32], however, the average amount of life years gained by brexucabtagene autoleucel is significantly higher than pharmacologic therapies. The survival gain of brexucabtagene autoleucel was basically stable in different scenario analyses: it was influenced only by the major assumptions, such as the time of cure and the long-term excess mortality in cured patients. On the converse, the economic outcomes of the analysis were sensitive to a certain amount of factors, such as the choice of SOC and the amount of patients undergoing allogeneic SCT, which is still recommended if CAR-T therapy has failed or is not feasible [8]. However, this result was not unexpected, since many therapies proved more cost-effective than allogeneic SCT, i.e. eculizumab for paroxysmal nocturnal hemoglobinuria [33].
Some limitations of this model-based analysis should be reckoned. First, we acknowledge that a cutoff of five years to define a patient with MCL as cured is largely bound to the purpose of the present study, and may reflect survivorship but not cure: a long time to cure, of 10 years, was explored at sensitivity analysis without any relevant impact on the estimated cost-effectiveness ratio. However, three-year clinical follow-up of ZUMA-2 has been recently published [34]: since the model is based on prior, more limited follow-up, data, a scenario analysis was conducted for accounting for the latest survival reported at 36-month follow-up. The above analysis showed that the benefit of brexucabtagene autoleucel significantly declined, but high-risk subgroups might still achieve a huge benefit, as compared with BAT. These issues are still being explored by updated models and deserve re-assessment of the cost-effectiveness.
Second, synthetic control arms have been criticized as substitutes for true control arms due to slight but significant variations in the definition of R/R status and a general exclusion of unfit patients from trials [35]. Third, the present model was not customized to analyze the cost-effectiveness of brexucabtagene autoleucel in clinical subgroups, such as patients with early progression (POD24), who were reported to achieve a shorter median PFS (11.3 vs 29.3 months), more frequent cytopenias and fewer CRS events [13]. Moreover, the model was not designed to test therapy sequencing, including pre-CART or post-CAR-T allogeneic SCT. Fifth, unit costs were mostly based on charges and no specific study ever assessed the true resource consumption patterns for MCL in Italy, therefore, costs are possibly underestimated. In addition, the present pharmaco-economic analysis of brexucabtagene autoleucel did not assess the budget impact of brexucabtagene autoleucel, which is expected to be low, due to the rarity of the disease.
New pharmacologic therapies for MCL are being developed. In particular, LOXO-305, a new-generation BTK inhibitor, amplified the therapeutic yield for MCL, however, the reported overall response rate of 51% is still suboptimal and such a therapy is not expected to compete with CAR-T in the same treatment setting [13].