CAR T-cell Therapy for Diffuse Large B-cell Lymphoma in Canada: A Cost-Utility Analysis

Abstract

Background

Chimeric antigen receptor (CAR) T-cell therapy is a novel cell therapy for treating non-Hodgkin lymphoma. The development of CAR T-cell therapy has transformed oncology treatment by offering a potential cure. However, due to the high cost of these therapies, and the large number of eligible patients, decision makers are faced with difficult funding decisions. Our objective was to assess the cost-effectiveness of tisagenlecleucel for adults with relapsed/refractory diffuse large B-cell lymphoma in Canada using updated survival data from the recent JULIET trial.

Methods

We developed an individual-simulated discrete event simulation model to assess the costs and quality-adjusted life-years (QALY) of tisagenlecleucel compared with salvage chemotherapy. Survival estimates were obtained from a published clinical trial and retrospective analysis. If patients remained progression free for 5 y, they were assumed to be in long-term remission. Costing and utility data were obtained from reports and published sources. A Canadian health care payer perspective was used, and outcomes were modeled over a lifetime horizon. Costs and outcomes were discounted at 1.5% annually, with costs reported in 2021 Canadian dollars. A probabilistic analysis was used, and model parameters were varied in 1-way sensitivity analyses and scenario analyses.

Results

After we incorporated the latest clinical evidence, tisagenlecleucel led to an additional cost of $503,417 and additional effectiveness of 2.48 QALYs, with an incremental cost-effectiveness ratio of $202,991 compared with salvage chemotherapy. At a willingness-to-pay threshold of $100,000/QALY, tisagenlecleucel had a 0% likelihood of being cost-effective.

Conclusions

At the current drug price, tisagenlecleucel was not found to be a cost-effective option. These results heavily depend on assumptions regarding long-term survival and the price of CAR T. Real-world evidence is needed to reduce uncertainty.

Highlights

• For patients with diffuse large B-cell lymphoma who failed 2 or more lines of systemic therapy, CAR T was not found to be a cost-effective treatment option at a willingness-to-pay threshold of $100,000.

• These results heavily depend on the expected long-term survival. The uncertainty in the model may be improved using real-world evidence reported in the future.

In 2019, it was estimated that there were 10,000 new cases of non-Hodgkin lymphoma (NHL) in Canada.^1,2 Diffuse large B-cell lymphoma (DLBCL) is the most common subtype, accounting for 30% to 40% of newly diagnosed lymphoma cases.^1,2 Of these patients, 60% will receive first line chemoimmunotherapy and be successfully treated.^3–5 The remaining will experience relapsed or refractory (r/r) disease and require subsequent therapy. Standard therapy for r/r DLBCL is salvage chemotherapy followed by high-dose therapy and autologous stem-cell transplantation (SCT). ⁶ Following this treatment, only 30% will achieve long-term remission, and more than 50% will relapse. ⁷ Few treatment options remain after autologous SCT, and treatment often results in poor outcomes. ⁸ Chimeric antigen receptor (CAR) T-cell therapy has emerged as an innovative and promising therapy in the management of r/r DLBCL. CAR T-cell therapy is a personalized therapy in which a patient’s own T-cells are harvested, modified to specifically target cancer cells, and readministered to the patient, where they bind to and kill cancer cells.^9,10

Tisagenlecleucel (tisa-cel) is a type of CAR T-cell therapy approved for use in select adult lymphoma populations. The treatment targets a CD-19 antigen that is expressed in more than 95% of B-cell malignancies. ⁷ Tisa-cel was approved by the United States Food and Drug Administration in May 2018 and by Health Canada in May 2019.^11,12 The indication for this CAR T-cell product is adult patients with r/r DLBCL after 2 or more lines of systemic therapy.^11,12 In 2019, Canada’s national health technology assessment agency, the Canadian Agency for Drugs and Technologies in Health (CADTH), recommended funding tisa-cel conditional on a substantial price reduction. ¹³ This decision was based on an analysis that used short-term survival data with a median follow-up of 14 mo.^14,15

In an updated survival analysis, at a median follow-up of 40.3 mo, tisa-cel resulted in an overall response rate of 53% and complete response of 45%. ¹⁶ While the single-arm study showed improved survival in select patients, ¹⁶ the high cost of CAR T-cell therapy ($513,655 per patient), ¹⁷ together with the substantial use of health care resources, make this therapy a challenging drug reimbursement problem. ¹⁸ The cost-effectiveness of tisa-cel, using the updated survival analysis figures, is currently unknown. The objective of this study was to use the updated survival estimates to evaluate the cost-utility of tisa-cel in adult patients with r/r DLBCL after 2 or more lines of systemic therapy.

Methods

Model Overview

We developed an individual-simulated discrete event simulation (DES) model using R (version 4.2.1) to assess the long-term costs and outcomes of patients with r/r DLBCL who received CAR T-cell therapy compared with a similar population who received a salvage chemotherapy regimen (rituximab, gemcitabine, dexamethasone, and cisplatin [R-GDP]). DES models can simulate individual patients as they transit through a health care system, with transitions dependent on both patient characteristics and the availability of resources. These types of models can determine if queues will develop due to resource constraints and can also evaluate the impact of wait-time strategies. ¹⁹ In Canada, wait times are on average 1 mo after apheresis, as cells need to be shipped to a manufacturer in the United States. The DES model allows us to capture the wait times that patients may experience. A schematic of the treatment pathway and structure of the model is shown in Supplementary Figure 1. The model simulates individual patients, their health states related to treatment, and clinical states to reflect the natural history of DLBCL. A simulated cohort of 2,000 patients was followed starting at a mean age of 58 y. The start ages were randomly sampled based on a normal distribution with a mean of 58 y for all simulated patients. This start age aligns with the median age reported in the clinical trial. ²⁰ The model consists of 3 main health states: 1) progression free, 2) progressed, and 3) long-term remission. A patient enters the model in the progression-free health state and can remain progression free or move to the progressed, long-term remission or death health states over time. A person can only enter the long-term remission health state after 5 y of being progression free. ²¹ Individuals who progress after treatment have the opportunity to receive autologous SCT after salvage chemotherapy and allogeneic SCT after CAR T-cell therapy.^14,22

Events included in the model are grade 3/4 adverse events reported in the clinical trials (e.g., cytokine release syndrome, neurologic events) and death. Wait times were also modeled using a fixed queue (first in, first out) in our simulation and represented the amount of time that a patient must wait to receive CAR T-cell therapy (base case = 1 mo). This was to reflect delays, such as extra manufacturing time, traveling and delivery time, waiting for specialists, and reimbursement approval by public health care payers.

All patients in the model eventually died from DLBCL or background mortality. A lifetime time horizon was taken, and future costs and benefits were discounted at 1.5% annually based on published Canadian guidelines. ²³ An economic analysis plan was developed for this study and is available online through the Canadian Institutes of Health Research.

Treatment Strategies

We considered 2 treatment strategies: 1) salvage chemotherapy (R-GDP) and 2) CAR T-cell therapy (tisa-cel). There are several other chemotherapy combinations for salvage chemotherapy available. These include rituximab, dexamethasone, cytarabine, and cisplatin (R-DHAP) and less commonly rituximab, ifosfamide, carboplatin, and etoposide (R-ICE) ³ ; however, the most common treatment in the Canadian setting for adult patients with DLBCL is R-GDP. ²⁴ Therefore, we assumed that R-GDP was used for all patients in the salvage chemotherapy arm.

Model Inputs

Survival data

Overall survival (OS) and progression-free survival (PFS) data were collected from the published findings of trials investigating CAR T-cell therapies for patients with r/r DLBCL. The JULIET trial was a multicenter, single-arm, phase 2 study that used tisa-cel to treat patients with r/r DLBCL and followed patients up to a median of 14 mo. ¹⁵ A long-term follow-up analysis was published with a median follow-up of 40.3 mo. ¹⁶ For the salvage chemotherapy arm, data from the SCHOLAR-1 trial were used for OS. SCHOLAR-1 is a retrospective study that consisted of 636 relevant patients undergoing mixed salvage chemotherapy regimens across 4 cohorts. A subset of patients who had a similar Eastern Cooperative Oncology Group performance status (score 0–1) to patients in 2 CAR T-cell therapy trials (ZUMA-1 and JULIET) were examined. ⁸ Since SCHOLAR-1 reported only OS, we used a proportional relationship based on published PFS and OS curves for patients using rituximab, dexamethasone, cytarabine, and cisplatin (R-DHAP) to derive the PFS curve.^25–27

To extrapolate survival data beyond the trial period, the published OS and PFS curves from the clinical trials were digitized. An algorithm developed by Guyot et al. ²⁸ was applied to reconstruct the individual patient data underlying the OS and PFS curves. A number of parametric functions were fitted onto the OS and PFS curves. Standard parametric models (exponential, Weibull, lognormal, loglogistic, and gamma distributions) were fitted to the data, and the one with the lowest Bayesian information criterion was chosen. Since the plateau of the survival curve is difficult to accommodate with a single parametric function, we used piecewise functions that connected 2 parametric functions based on overall goodness of fit. ²⁹ This approach allowed us to capture the structural changes of the survival curves. Using the best-fitted parametric models, the time to disease progression and death were sampled from the quantile function for each patient. ¹⁹ Supplementary Table 1 lists the details of the distributions used for the survival curves needed by the model to compute the monthly transition probabilities between the disease progression and survival health states.

Patients who survived beyond 5 y were considered to be in long-term remission. Patients in this health state could die of background mortality. This assumption was based on 2 studies that assessed long-term outcomes in patients with DLBCL and the low probability of progression past this point.^21,26 The survival estimates of the general population were derived from the 2017 to 2019 Statistics Canada Life Table. ³⁰

Adverse events

The administration of CAR T-cells is associated with significant short-term and long-term side effects. During the first month, common adverse events include cytokine release syndrome (CRS), cytopenia, infection, and febrile neutropenia. Grade 3 or higher CRS is seen in up to 23% of patients who received tisa-cel. ¹⁶ Neurologic events (i.e., encephalopathy), grade 3 or higher have been reported in up to 4% with tisa-cel. ¹⁶ Grade 3 or higher CRS commonly requires an intensive care unit (ICU) admission. Thus, patients must be closely monitored for severe side effects such as CRS after infusion. In addition to these adverse events, we considered other grade 3/4 events reported for tisa-cel if they affected more than 10% of patients in the trial. Although less than 10% of patients who received CAR T-cell therapy experienced grade 3/4 nausea, fatigue, and vomiting, we added these adverse events to the model to keep the reporting consistent with the salvage chemotherapy arm.

For salvage chemotherapy, adverse events were not reported in the clinical trial used for the efficacy data. It was assumed that the adverse events that occurred with this treatment matched a clinical trial in which patients with r/r B-cell lymphoma were randomized to receive GDP or DHAP. ³¹ Table 1 lists the adverse events considered in our model along with their probabilities.

Table 1.

Monthly Probabilities Related to Treatment and Adverse Events

Parameter	Value	Range	Distribution	Reference
Proportion of patients receiving bridging chemotherapy
Tisa-cel	0.90	0.68–1.00	Beta	Schuster et al. 16
Proportion of patients receiving SCT
Tisa-cel	0.05	0.04–0.06	Beta	CADTH 14
Salvage chemotherapy	0.30	0.22–0.37	Beta	Corbett et al. 22
Adverse events (grade 3–4) with tisa-cel
Cytokine release syndrome	0.23	0.17–0.28	Beta	Schuster et al. 16
B-cell aplasia	0.30	0.23–0.38	Beta	Schuster et al. 16
Febrile neutropenia	0.17	0.12–0.21	Beta	Schuster et al. 16
Anemia	0.39	0.29–0.49	Beta	Schuster et al. 16
Encephalopathy	0.04	0.03–0.05	Beta	Schuster et al. 16
Hypotension	0.09	0.07–0.11	Beta	Schuster et al. 16
Neutropenia	0.20	0.15–0.25	Beta	Schuster et al. 16
Pyrexia	0.05	0.04–0.07	Beta	Schuster et al. 16
Thrombocytopenia	0.12	0.09–0.15	Beta	Schuster et al. 16
White blood cell count decreased	0.32	0.24–0.40	Beta	Schuster et al. 16
Infections	0.19	0.14–0.24	Beta	Schuster et al. 16
Nausea	0.01	0.01–0.01	Beta	Schuster et al. 16
Fatigue	0.06	0.05–0.08	Beta	Schuster et al. 16
Adverse events with salvage chemotherapy
Febrile neutropenia	0.23	0.17–0.29	Beta	Crump et al. 31
Infections	0.09	0.07–0.11	Beta	Crump et al. 31
Nausea	0.08	0.06–0.10	Beta	Crump et al. 31
Vomiting	0.07	0.05–0.09	Beta	Crump et al. 31
Fatigue	0.09	0.07–0.11	Beta	Crump et al. 31
Wait time for CAR T	1 mo	3–6 mo		Clinician input 37

CADTH, Canadian Agency for Drugs and Technologies in Health; CAR, chimeric antigen receptor; SCT, stem-cell transplantation; tisa-cel, tisagenlecleucel.

Costs

This analysis adopted a Canadian health care payer perspective, and as such, direct health care costs were considered (Table 2). These costs included pretreatment costs, cost of treatment, monitoring, and adverse events. The pretreatment costs for CAR T-cell therapy included leukapheresis and lymphodepleting chemotherapy with fludarabine and cyclophosphamide. A proportion of patients also incurred the cost of bridging chemotherapy before receiving CAR T-cell infusion. The primary sources of data for costs included the Ontario Case Costing Initiative, the Canadian Institute for Health Information, the Ontario Schedule of Benefits and Fees for Physician Services, and the Ontario Drug Benefit.^32–35

Table 2.

Costs and Utilities

Parameter	Value	Range ($)	Distribution	Reference
Pretreatment
Leukapheresis	1,441	1,081–1,801	Gamma	OCCI 30
Pretreatment tisa-cel	152	114–190	Gamma	Ellis 24
Bridging chemotherapy	20,625	15,469–25,781	Gamma	CADTH 14
Treatment
Tisa-cel	513,655	385,241–642,068	Gamma	RedBook 17
Administration of CAR T	105	79–131	Gamma	SoB 31
Hospitalization (regular ward)	1,950	1,462–2,437	Gamma	CIHI 33
Salvage chemotherapy (R-GDP, 6 cycles)	22,344	16,758–27,936	Gamma	ODB 32
Administration of chemotherapy	75	56–94	Gamma	SoB 31
Stem cell transplant	170,201	127,651–212,751	Gamma	Ontario Health 35
Posttreatment follow-up and monitoring
Office consultation, hematology	169	127–211	Gamma	SoB 31
PET scan	1,706	1,279–2,132	Gamma	Ellis 24
Adverse events
Febrile neutropenia	7,137	5,352–8,921	Gamma	OCCI 30
Infections	470	352–587	Gamma	OCCI 30
Pyrexia	447	335–558	Gamma	OCCI 30
Neutropenia	534	400–667	Gamma	OCCI 30
Anemia	8,584	6,438–10,730	Gamma	OCCI 30
Thrombocytopenia	459	344–574	Gamma	OCCI 30
Hypotension	590	442–737	Gamma	OCCI 30
White blood cell count decreased	456	342–570	Gamma	OCCI 30
Encephalopathy	4,271	3,203–5,338	Gamma	OCCI 30
Nausea	266	200–333	Gamma	Ellis 24
Vomiting	342	257–428	Gamma	Ellis 24
Fatigue	334	251–418	Gamma	Ellis 24
CRS hospitalization (intensive care unit)	3,958	2,968–4,947	Gamma	CIHI 33
Tocilizumab (2 doses)	2,553	1,915–3,191	Gamma	ODB, 32 Ellis 24
IVIG	2,556	1,917–3,195	Gamma	ODB, 32 Ellis 24
Predeath costs
Care in the last year before lymphoma-related death	58,709	58,687–58,731	Gamma	De Oliveria et al., 39 Ellis 24
Health care costs in the last year of life	60,147	45,111–75,184	Gamma	Tanuseputo et al., 40 Ellis 24
Utilities
Progression-free health state	0.80	0.53–1.00	Beta	Lin et al. 43
Progressed health state	0.72	0.39–1.00	Beta	Lin et al. 42
Disutility with tisa-cel	0.05	0–0.13	Beta	Lin et al. 42
Disutility of salvage chemotherapy	0.15	0.11–0.19	Beta	CADTH 14
Disutility of stem cell transplant	0.57	0.31–0.87	Beta	Sung et al. 45
Additional utility starting from 18 mo after treatment	0.06	0.05–0.07	Beta	Patrick et al. 46

CADTH, Canadian Agency for Drugs and Technologies in Health; CAR, chimeric antigen receptor; CIHI, Canadian Institute for Health Information; CRS, cytokine release syndrome; IVIG, intravenous immunoglobulin; OCCI, Ontario Case Costing Initiative; ODB, Ontario Drug Benefit; PET, positron emission tomography; SoB, schedule of benefits; tisa-cel, tisagenlecleucel.

The cost of CAR T-cell therapy included the cost of the CAR T-cell product, physician costs for monitoring during the administration of CAR T-cell therapy, and an inpatient hospital stay. The cost of CAR T-cell therapy was based on the wholesale acquisition cost of $513,655 CAD for tisa-cel. ³² The cost of salvage chemotherapy (gemcitabine-dexamethasone-cisplatin-rituximab) was based on 6 cycles within 21 d. ²⁴ For the cost of drugs in which the dose is weight based, an average body surface index of 1.7 m² was used. ²⁴ For patients who received an SCT, the cost of an adult allogeneic matched donor was applied. ³⁶

Posttreatment monitoring included the cost of physician consultations (hematology) twice a week in the first month after infusion, weekly in the second month, monthly thereafter up to 6 mo, every 3 mo for up to 2 y, and every 6 mo until 5 y.^37,38 The cost of a positron emission tomography scan was applied at 3 mo, 6 mo, 1 y, and yearly up to 5 y. ²⁴

We included the cost of the top grade 3 or 4 adverse events reported in each CAR T-cell therapy clinical study. The most significant adverse events associated with CAR T-cell therapy are CRS and B-cell aplasia. The cost of CRS was calculated by multiplying the cost of an ICU stay and 2 doses of tocilizumab by the time to CRS resolution reported in the clinical trials. The average cost of an ICU stay was obtained from the Canadian Institute for Health Information. ³⁵ For those who experienced B-cell aplasia, the cost of intravenous immune globulin was applied monthly up until 1 y. ¹³ The costs of all other adverse events were based on hospital costs obtained from the Ontario Case Costing Initiative, which reports the costs of inpatient hospital visits in Ontario. ³²

For all patients who died, the predeath health care costs in the last year before the lymphoma-related death were applied.^39,40 These predeath costs were obtained from 2 retrospective cohort studies and included any health care visits, hospitalizations, medications, and home care visits paid for by Ontario Health. All costs were inflated to 2021 Canadian dollars using the Consumer Price Index for health care services in Ontario. ⁴¹ The United States prices of CAR T-cell therapy were converted to Canadian dollars using the purchasing power parity exchange rate. ⁴²

Utilities

Utility values (Table 2) for the progression-free and progressed health states were obtained from published quality-of-life data, based on a safety management cohort from the ZUMA-1 trial, which were generated using the EQ-5D-5L instrument. ⁴³ The ZUMA-1 trial focused on patients treated with axicabtagene ciloleucel (axi-cel), which is a similar product to tisa-cel. A total of 34 patients with r/r DLBCL treated with axi-cel were initially part of the quality-of-life study, with a median follow-up of 5.1 mo. The mean EQ-5D-5L index was 0.80 (SD = 0.14) for the progression-free health state and 0.72 (SD = 0.17) for progressed disease. ⁴³ To capture short-term adverse events associated with treatment, we applied a disutility while patients were receiving treatment. For CAR T-cell therapy, the disutility associated with treatment was 0.05 (SE = 0.04) and applied over a 1-y period at the time of the infusion. ⁴³ For those who experienced grade 3/4 CRS, health utility was reduced to 0 for the first 2 mo after the CRS-related event. ⁴³ For salvage chemotherapy, a disutility of 0.15 was applied for the treatment (6 cycles of chemotherapy),^14,44 based on a randomized multicenter study comparing CHOP with CHOP plus G-CSF.^14,44 The disutility associated with SCT (0.57) was obtained from a study examining the quality of life of young adults with acute myeloid leukemia who had an allogeneic bone marrow transplantation. ⁴⁵ This disutility was applied at the time of therapy. An additional utility of 0.06 starting 18 mo after treatment was applied, as a study examining the quality of life of patients who received CAR T-cell therapy indicated that the utilities increase at this time point. ⁴⁶

Analysis

The DES model captured patient-level variability and parameter-level uncertainty. For the base case, a probabilistic sensitivity analysis was conducted to account for the uncertainty of all the parameters. Parameters were sampled from appropriate distributions (e.g., beta distribution for probabilities) informed by the corresponding ranges. The ranges were based on the published 95% confidence intervals or, in the absence of literature, ±25% from their base case value. We simulated 2,000 individuals and performed 1,000 simulations for each treatment strategy to estimate the mean outcomes. We estimated the likelihood of each strategy being more favorable across a range of willingness-to-pay (WTP) thresholds using cost-effectiveness acceptability curves (CEACs). An upper limit of $100,000 per QALY was used as a cost-effectiveness threshold, as this threshold is commonly used by Canada’s national health technology assessment agency. ¹⁴

Deterministic sensitivity analysis

To assess the impact of each parameter on the results, we conducted deterministic 1-way sensitivity analyses, varying the input parameters one at a time according to their 95% confidence intervals, or ±25% from their base case value.

Scenario analysis

Scenario analyses were conducted on the time to achieve long-term remission, mortality in the long-term remission health state (standardized mortality ratio of 1.09), ⁴⁷ alternative parametric distribution (i.e., second best fit) for the OS and PFS extrapolation, and alternative wait time to receive CAR T-cell therapy. We also conducted a scenario analysis using survival and cost estimates for axi-cel, which is another CAR T-cell therapy approved in Canada.^20,48,49 The survival curves for axi-cel can be found in Supplementary Figure 4 and the parameter estimates in Supplementary Table 2.

Validation

To validate our model, the Kaplan-Meier survival curves for the OS and PFS from the trials were plotted alongside our model outcomes. We visually inspected the overlay by comparing the OS and PFS curves from the model and the original trials. Supplementary Figures 2 and 3 demonstrate a comparison between the trial data and the output of our simulation when comparing the JULIET and SCHOLAR-1 trials for OS. The face validity of the model structure, assumptions, data, and results were validated with clinical experts.

Results

Base Case Analysis

The results of the base case analysis are presented in Table 3. In the probabilistic analysis, tisa-cel compared with salvage chemotherapy resulted in an additional cost of $503,417 and additional effectiveness of 2.48 QALYs. This led to an incremental cost-effectiveness ratio (ICER) of $202,991 per QALY gained. Tisa-cel was not cost-effective based on a WTP threshold of $100,000 per QALY. Our model estimated 14.8% of cancer-related deaths were averted due to the treatment at 5 y. The results demonstrated that for tisa-cel, 100% of the 5,000 second-order simulated ICERs were in the northwest quadrant, meaning that tisa-cel was more costly and more effective compared with salvage chemotherapy (Figure 1). The CEAC demonstrated that compared with salvage chemotherapy, tisa-cel had a 0% probability of being cost-effective at $100,000/QALY (Figure 2).

Table 3.

Base Case Analysis Results

	Salvage Chemotherapy	Tisa-cel
Total expected cost, 2021 CAD	$143,115	$637,532
Incremental cost	—	$503,417
Total expected QALYs	3.55	6.03
Incremental QALYs	—	2.48
Total expected life-years	5.64	9.15
Incremental life-years	—	3.51
ICER (cost per QALY)	—	$202,991

Figure 1.

Cost-effectiveness plane for tisagenlecleucel (tisa-cel) versus salvage chemotherapy.

Figure 2.

Cost-effectiveness acceptability curve for tisagenlecleucel (tisa-cel) versus salvage chemotherapy.

Deterministic Sensitivity Analysis

Figure 3 demonstrates the top 10 variables that affected the ICERs. The ICER ranged from $152,632 per QALY gained to $444,535 per QALY gained. The model was most sensitive to the time horizon, the starting age of the cohort, discount rate, cost of the product, and the utility of the progression-free health state. Tisa-cel was not cost-effective in all of the one-way sensitivity analyses; however, it came closer to the cost-effectiveness threshold when the cost of the product was reduced to $385,241 and the starting age of the cohort was reduced to 43.5 y. In our analysis, the price of tisa-cel would need to be reduced by 51% ($251,602) to meet the $100,000 WTP threshold.

Figure 3.

Deterministic sensitivity analysis results.

Scenario Analysis

The results of the scenario analysis are presented in Supplementary Table 3. A change in the time required to achieve long-term remission, increasing the mortality for those in long-term remission, and the wait time to receive CAR T-cell therapy did not affect our assessment of the economic value of tisa-cel. The results were also robust to changes in the OS and PFS curves (i.e., using the second-best fit). Using the survival and cost estimates associated with axi-cel resulted in an ICER of $105,649/QALY compared with salvage chemotherapy, which is closer to our WTP threshold. If the cost of CAR T-cell therapy was reduced to $251,602, tisa-cel would be cost-effective compared with salvage chemotherapy at a WTP threshold of $100,000/QALY.

Discussion

CAR T-cell therapy has the potential to improve patient lives by inducing a long-term durable remission in patients who have run out of treatment options; however, this therapy comes at a high cost. After using longer-term survival data, our results suggest that tisa-cel would not be cost-effective using a WTP threshold of $100,000 per QALY. The ICERs were driven by the time horizon, starting age of the cohort, discount rate, price of the product, and assumptions regarding long-term survival.

Our results are consistent with previous economic studies.^14,50,51 Three studies found tisa-cel to be cost-effective compared with salvage chemotherapy; however, there are key differences between these studies and ours.^52–54 Two of these studies relied on less mature data from the JULIET trial and selected a 3-year cure point.^53,54 In our scenario analysis, if we reduced the cure point from 5 to 3 y, the ICER of tisa-cel versus salvage chemotherapy would be reduced to $169,083/QALY, bringing the ICER closer to the WTP threshold. A study from the United States that relied on more mature data from the JULIET trial and used a 5-y cure rate found tisa-cel to be cost-effective. ⁵² However, there are a few key differences between our study and theirs, which likely had a cumulative impact on the ICER. First, Qi et al. reported the total costs associated with subsequent SCT to be $105,000 USD more with the salvage chemotherapy arm compared with the tisa-cel arm. In our model, subsequent SCT was associated with $42,000 CAD more in costs. ⁵² Our cost of SCT and rates of subsequent transplant were lower as we considered only autologous SCT for the salvage chemotherapy arm and allogeneic SCT for the tisa-cel arm. This is in line with costs and clinical practice in Canada. To be eligible for CAR T in Canada, the patient would not have previously received allogeneic SCT; however, this treatment may be offered after CAR T. As our tornado diagram demonstrates, if we increased our rate of subsequent SCT for salvage chemotherapy to 37.3%, our ICER would be $193,072 per QALY gained, which is much lower than our base case. Second, the total cost after progression for the salvage chemotherapy therapy arm and the total cost of terminal care were larger for the salvage chemotherapy arm in the study by Qi et al. ⁵² This is a function of more patients progressing and requiring SCT in the salvage chemotherapy arm. Finally, Qi et al. ⁵² reported greater incremental QALYs gained with tisa-cel versus our study, which is a result of the disutility associated with subsequent SCT and progression. All of these factors together resulted in an ICER that is more favorable than ours.

To our knowledge, this is the first study to use a DES model to estimate the cost-effectiveness of tisa-cel compared with salvage chemotherapy. This type of model further enhanced the realism of the simulations by allowing costs and effectiveness to accumulate and be discounted accurately over a continuous period of time. Another key strength of our analysis is including wait times for treatment in the model, which have been shown to affect the ICER. Our study is further enhanced by the use of more recent survival data from the JULIET trial. ¹⁶ Finally, we used a 2-phase parametric curve fitting, which greatly improved the model fit and the accuracy of our extrapolated estimates. This DES model will be used in the future to evaluate the cost-effectiveness of tisa-cel using real-world evidence so that results are more generalizable to public payers. This DES model will allow us to account for the impact of time delays to receiving treatment in the real world (e.g., survival, additional resources used).

Limitations

There are some limitations to this cost-effectiveness analysis. First, since no head-to-head trial data were available, this analysis relied on data sources from a single-arm trial. Data from a real-world study could help to inform the true estimates. Second, data from this single-arm trial included a median of 40.3 mo of follow-up, and so assumptions had to be made regarding the extrapolation of outcomes beyond this point and the percentage of patients who are deemed cured. Longer-term data can help to inform more precise estimates of cost-effectiveness over a patient’s lifetime. The collection of standardized outcomes in a registry is needed to better assess the long-term effectiveness, safety, and cost-effectiveness. Third, similar to other published cost-effectiveness analyses, we calibrated our DES model using published OS and PFS estimates. We did not have access to the individual patient-level data, which limits our ability to accurately estimate the exact timing of each event individually, which can cause us to overestimate or underestimate our results. A series of scenario analyses were run using the second best fit for the OS and PFS curves. Fourth, we assumed that after 5 y of being progression free, a patient was cured. Longer follow-up data will be needed to inform this estimate. Lastly, the utility estimates informing our study were mainly based on the quality of life of patients who received axi-cel. Although this treatment is similar to tisa-cel, future studies are necessary to validate these estimates.

Conclusion

Our study suggests that, compared to salvage chemotherapy, tisa-cel was not a cost-effective treatment option for patients with DLBCL who failed 2 or more lines of systemic therapy at a WTP threshold of $100,000/QALY. Our analysis indicated that there is still a reasonable amount of uncertainty in terms of survival and utility estimates. Given that CAR T-cell therapy has already been introduced in many jurisdictions around the world, further research based on real-world evidence can address some of the uncertainty raised in our analysis.