Cost utility analysis of tisagenlecleucel vs salvage chemotherapy in the treatment of relapsed/refractory diffuse large B-cell lymphoma from Singapore’s healthcare system perspective

Abstract

Background

Patients with relapsed or refractory diffuse large B-cell lymphoma (r/r DLBCL) have limited treatment options and poor prognoses. Tisagenlecleucel, a chimeric antigen receptor (CAR) T-cell therapy has shown early promise in improving survival outcomes, but at a high upfront cost. This study evaluated the cost-effectiveness of tisagenlecleucel versus salvage chemotherapy for treating patients with r/r DLBCL who have failed at least 2 lines of systemic therapies.

Methods

A hybrid decision tree and three-state partitioned survival model (progression-free (PF), progressive disease and death) was developed from the Singapore healthcare payer perspective. Survival curves from JULIET trial and CORAL-1 extension study were extrapolated beyond trial period over a 15-year time horizon to estimate the underlying progression-free survival and overall survival parametric distributions for both arms. Health state utilities were retrieved from the literature, and direct costs were sourced from public healthcare institutions in Singapore. One-way probabilistic sensitivity analyses and scenario analyses were conducted to explore the impact of uncertainties and assumptions on cost-effectiveness results.

Results

Compared with salvage chemotherapy, tisagenlecleucel was associated with a base-case incremental cost-effectiveness ratio (ICER) US$508,530 (S$686,516) per quality adjusted life year (QALY) gained and US$320,200 (S$432,269) per life year (LY) gained. One-way sensitivity analysis showed the ICER was most sensitive to time horizon, PF utility and cost of tisagenlecleucel. Scenario analyses confirmed that the ICERs remained high under favorable assumptions and substantial price reduction was required to reduce the ICER.

Conclusions

Our analysis showed tisagenlecleucel use in r/r DLBCL patients who failed at least 2 prior lines of systemic therapies was associated with exceedingly high ICER, which is unlikely to represent good use of healthcare resources. Comparative clinical evidence from the ongoing trials might provide more insight into future evaluations.

Keywords:

• Tisagenlecleucel
• CAR-T
• relapsed/refractory diffuse large B-cell lymphoma
• cost-effectiveness
• partitioned-survival

JEL classification codes:

• H51
• I18

Background

Lymphoma is one of the most frequently occurring cancers in Singapore. The age-standardized incidence rates of newly diagnosed lymphoid neoplasm in men and women were 17.6 and 11.2 per 100,000 Singapore residents, respectively¹. Nearly 90% of all lymphoma cases in Singapore were Non-Hodgkin’s lymphoma (NHL)². Diffuse large B-cell lymphoma (DLBCL) is aggressive and constitutes approximately 25–30% of all NHL cases^2,3. DLBCL can occur at any age, but people aged over 50 years old are more commonly affected⁴.

A complete remission (CR) is achievable in patients with DLBCL who receive first-line rituximab, cyclophosphamide, doxorubicin hydrochloride, vincristine sulphate, prednisone (R-CHOP) immune-chemotherapy. However, 30% to 50% of patients experience relapse while 10% have refractory disease^5,6. These patients can be given salvage chemotherapy followed by stem cell transplantation (SCT), if eligible, with the aim to achieve sustained remissions. For patients who failed 2 or more prior lines of systemic therapies, there is no standard treatment and salvage chemotherapy is currently the most common treatment option to control disease. Traditionally, prognosis is poor with a median overall survival rate of less than a year⁷. Chimeric antigen receptor T-cells (CAR-T) therapy, a new class of cancer immunotherapy, have recently emerged as a clinically significant advancement in the management of relapsed/refractory (r/r) DLBCL. It involves reprogramming patients’ own T-cells with CAR specific for a tumor antigen thereby targeting and eliminating malignant cells that express these proteins. With T cells’ capacity for memory and surveillance, some patients may not need for subsequent therapy such as SCT.

Tisagenlecleucel, a type of anti-CD19 CAR-T therapy, has been approved by the US Food and Drug Administration and the European Medicines Agency for adult patients ≥18 years with r/r DLBCL after at least 2 lines of systemic therapies, and is expected to be approved in Singapore. The pivotal international, single-arm, open-label JULIET trial demonstrated median overall survival (OS) of 12.0 months (95% confidence interval [CI], 7.0 months to not reached) for 93 patients who received tisagenlecleucel infusion, over a median follow-up of 14 months⁸. In the updated analysis of JULIET trial with a follow-up duration up to 19 months, the median OS was 11.1 months (95% CI, 6.0 months to not reached) among 115 patients⁹. The clinical benefits of tisagenlecleucel are further corroborated by median OS results of 22.2 months (95% CI not available) reported by a case-series study from Schuster et al.¹⁰ However, tisagenlecleucel is also associated with known serious adverse events such as cytokine release syndrome and B-cell aplasia, which require monitoring by trained specialists and potential lifelong treatment with intravenous immunoglobulin (IVIG).

While tisagenlecleucel could purportedly result in lifelong clinical benefit, much concerns arose regarding its high upfront cost in the absence of direct comparative effectiveness with other treatment options and long-term safety data. The list price of tisagenlecleucel for lymphoma was US$373K (∼S$500K) in United States¹¹. Correspondingly, there is heightened interest on the economic value of tisagenlecleucel to inform policy-makers’ funding decisions alongside other relevant considerations. Therefore, the present study aimed to evaluate the cost effectiveness of tisagenlecleucel compared with salvage chemotherapy for the treatment of r/r DLBCL, from a healthcare system perspective in Singapore.

Methods

Model structure

A hybrid decision tree and partitioned survival model (PSM) (also known as the area under the curve (AUC) model) was constructed to assess the cost effectiveness of tisagenlecleucel versus salvage chemotherapy in a hypothetical cohort of patients with a median age of 56 years who have failed two or more lines of systemic therapies, consistent with the trial population reported in the JULIET study. A 15-year time horizon was chosen to capture all important differences in costs or outcomes between the treatments being compared. It reflected the timeframe by which majority of patients were expected to have died; this was considered clinically plausible based on average life expectancy in Singapore of 83.2 years in 2019, and patients suffered up to 10 years’ loss compared with the general population due to the condition and cumulative toxicity from several rounds of invasive chemotherapy treatments^12,13.

Decision tree model

Patients were assumed to receive either tisagenlecleucel or salvage chemotherapy. Treatment with tisagenlecleucel starts with obtaining a patient’s T cells through leukapheresis. The harvested cells would be sent to an overseas manufacturing facility for enrichment and activation. The entire process takes at least 3–4 weeks before infusion of final product could take place. During this period, patients require bridging chemotherapy to stabilize the disease. Prior to the infusion, patients would also receive lymphodepleting therapy to create a favorable immune environment for CAR-T persistence. However, a small proportion of patients could experience deaths (10%) or other events (23%) such as manufacturing failures and adverse events during the manufacturing time lapse. We assumed that these patients would have undergone leukapheresis and some of them would have received bridging chemotherapy (50%) or lymphodepleting therapy (50%) but were no longer eligible for tisagenlecleucel infusion in the model to reflect real-life clinical practice. Patients who remained alive following infusion failure would accrue outcomes and costs similar to those who were treated with chemotherapy instead. About 6.3% of patients who were successfully infused with tisagenlecleucel did not have a response and underwent SCT in JULIET trial; this was also accounted for in the decision tree. As their OS and progression free survival (PFS) results were not reported separately, this group of patients were assumed to achieve the same outcomes as the overall trial population.

For the comparator arm, patients received two cycles of chemotherapy before they were assessed for response and eligibility to proceed with autologous or allogenic SCT. In line with the CORAL extension study, ∼31% of patients underwent SCT, while the rest of patients who were ineligible continued with chemotherapy treatment. Given all patients followed specific clinical pathways with very few outcomes or recurrences of any outcomes, a decision tree was deemed appropriate to capture the initial clinical pathway (Figure 1) in the progression-free health state.

Figure 1. Decision tree structure. Abbreviations. r/r DLBCL, relapsed or refractory diffuse large B-cell lymphoma; SCT, stem cell transplant; PSM, partitioned survival model.

Partitioned survival model

To model the clinical pathway over the long-term time horizon, a PSM was constructed with a cycle length of one month (with a half-cycle correction). The PSM model comprised 3 health states: alive with no progression [“progression free” (PF)], alive with progression [“progressed disease” (PD)] and dead (Figure 2). All patients entered into the model in the PF health state. At the beginning of each cycle, patients would either remain in the PF health state, transit to the PD or Death state. Patients who progressed to PD state could stay within the same health state or die; but not revert to the PF state. The proportion of patients in each health state over time was derived directly from the observed cumulative survival distributions for OS and PFS and extrapolated using parametric functions after the observed periods. The proportion of patients in the PD health state was derived from the difference between OS and PFS. The model was implemented in Microsoft Excel 2013 (Microsoft Corp, Redmond, WA).

Figure 2. Partitioned survival model with three health states.

Treatment strategies

For each scenario, tisagenlecleucel was compared with current standard of care for patients with DLBCL who have failed at least 2 lines of systemic therapies. In Singapore, the most common treatment option given was salvage chemotherapy with or without subsequent SCT. This was used as the main comparator in our study (Figure 3). While there are no specific salvage chemotherapy regimens predominantly prescribed locally, the following platinum-based regimens were identified following consultations with local oncologists: R-ICE (rituximab, ifosfamide, carboplatin, etoposide, mesna and peg-filgrastim), R-DHAP (rituximab, dexamethasone, cytarabine, cisplatin and peg-filgrastim) and R-GDP (rituximab, gemcitabine, dexamethasone and cisplatin). Patients were assumed to receive either R-ICE (50%) or R-DHAP (50%) at third-line setting, followed by the less intensive R-GDP regimen following response failure to either therapies.

Figure 3. Modelled treatment strategies. Abbreviations. R-CHOP, Rituximab, cyclophosamide, doxorubicin, vincristine, prednisone; RT, radiotherapy; SCT, stem cell transplant; R-ICE, rituximab, ifosfamide, carboplatin, etoposide, mesna and peg-filgrastim; R-DHAP, rituximab, dexamethasone, cytarabine, cisplatin and peg-filgrastim; R-GDP, rituximab, gemcitabine, dexamethasone and cisplatin; BEAM, carmustine, etoposide, cytarabine, melphalan; CR, complete response; PR, partial response.

Model parameters

Clinical efficacy data

In the absence of head-to-head comparative efficacy and safety of tisagenlecleucel and salvage chemotherapy, PFS and OS data were derived from two separate sources: JULIET trial for tisagenlecleucel (primary analysis with a median follow-up of 14-month) and CORAL extension study for salvage chemotherapy¹⁴. Updated median OS results of JULIET trial published in Mar 2019 with an additional 5-month of follow-up was generally in line with primary analysis (11.1 months vs 12 months) but was not included in our cost-effectiveness analysis as PFS data was not reported⁹.

In the CORAL extension study, outcomes of 203 patients aged 18–65 years with previously treated DLBCL who could not proceed to scheduled autologous SCT (ASCT) in the main CORAL study and received third-line salvage therapy were reported. Patient characteristics in this study were similar to JULIET trial in terms of median age, prognosis status and prior treatment with rituximab, and considered largely representative of r/r DLBCL patients who would otherwise be eligible for tisagenlecleucel if it was available in Singapore. The study also provided two separate Kaplan-Meier (KM) OS curves for patients having subsequent SCT and for those who did not proceed with SCT that allowed us to model the patients separately in “SCT” and “no SCT” arms. The median OS for the SCT and no SCT arm were 11.1 months and 3.3 months, respectively. As the PFS curves were not reported in CORAL extension study, PFS data was estimated from the OS based on the assumption of proportional hazards, with a hazard ratio (HR) of 0.65, the mean cumulative HR calculated from the main CORAL study¹⁵.

The published KM OS and PFS curves were digitalized using a validated graphical digitizer (WebPlotDigitizer version 4.2; Ankit Rohatgi, CA, USA)¹⁶. Then, an algorithm developed by Guyot et al. was applied to reconstruct the individual patient data (IPD) underlying the OS and PFS curves¹⁷. Following that, a regression analysis was carried out using the R statistical software to fit parametric functions onto the OS and PFS curves, enabling extrapolation beyond the trial follow-up period. Standard single parametric models (exponential, Weibull, lognormal, loglogistic, Gompertz and generalized gamma distributions) were fitted to the data. To also account for the potential curative nature of tisagenlecleucel and the emerging plateau at the tail-end of the survival curves, both mixture-cure models (MCM) and spline models were fitted to the tisagenlecleucel data, respectively^18,19. MCM models were also fitted to the chemotherapy with SCT arm. The final model was selected based on Akaike Information Criterion (AIC) value and visual inspection of the curves against the actual data. The AIC values and extrapolated curves are reported in the Supplementary materials.

For tisagenlecleucel, the base case model selected were one-knot spline and three-knot spline for OS and PFS, respectively. MCM (log-logistic) was not used in the base case given the lack of an apparent plateau in the observed KM curve combined with fewer patients towards the end of the tail signaling the extrapolated curve might be an overestimation of the true clinical effect. For PFS, the three-knot spline model had the lowest AIC and fitted the digitalized curve better than MCM based on visual inspection. Our choice of model was further validated against the latest JULIET trial results; at 18 months, the probability of OS reported was 43%, which was similar to the predicted figure by spline model (44%); whereas MCM (log-logistic) represented a modest overestimation at 47%. The standard Gompertz model was chosen for both OS and PFS in the chemotherapy without SCT arm while MCM-gompertz was chosen for both OS and PFS in the chemotherapy with SCT arm.

In accordance with published literature, local experts agreed that patients with DLBCL are deemed to be “cured” at a 5-year point^3,20. Therefore, we further assumed that all patients who remained alive after five years in both arms were long-term survivors and followed mortality rates of the general population (that is, standard mortality rate of 1.0)¹².

Costs

In line with the Singapore healthcare system perspective, direct healthcare costs were considered in the analysis^21,22. These included costs of treatment, consultation visits, monitoring and managing adverse events (AEs). Specific to tisagenlecleucel, we assumed that the cost of infusion was only incurred by patients who had successfully received the infusion, in line with terms specified in risk share agreements between the manufacturer and other payers²³. The frequency and types of relevant outpatient consultation visits, monitoring tests and scans were based on local expert opinions. All costs were sourced from public healthcare institutions and adjusted to 2018 Singapore dollars using the healthcare consumer price index (CPI)²⁴. All reported costs were in local currency and in US dollars (where US$1 = S$1.35 for the year 2018). Table S1 in the Supplementary shows the clinical and economic inputs that informed the model.

Utility values

Health-related quality of life (HRQoL) outcomes were assessed in pivotal JULIET trial, as reported by Maziarz et al.²⁵ Aggregated HRQoL results collected using FACT-Lym and SF-36 questionnaires were reported at baseline, 3, 6, 12- and 18-months post-infusion. However, no relevant values were published for progressed state. Therefore, utility values for the various health states and disutilities associated with treatments and adverse events were derived from other literature instead.

Utilities of 0.7 and 0.59 were used for the PF and PD states, respectively, based on a utility study by Wang et al. conducted in DLBCL patients²⁶. In the study, EQ-5D-5L data from 319 newly diagnosed (2004–2015) patients in the UK’s population based Haematological Malignancy Research Network (HMRN) were used to calculate utilities. Disutilities due to treatment, SCT and adverse events were applied to both treatment arms. These values were assumed to apply for the entire duration of the treatments and for one episode for each AE type. Patients were assumed to have additional disutility if they have grade 3 or above cytokine release syndrome (CRS) that resulted in intensive care unit (ICU) stay (Supplementary table S1). All costs and utilities were discounted at 3% per annum.

Sensitivity analyses

One-way sensitivity analyses (OWSA) were conducted to evaluate the impact of uncertain model parameters on the ICER. Each parameter was varied across the range of their 95% CI; otherwise, a ± 20% variation was applied to the base-case value.

Probabilistic sensitivity analysis (PSA) was also conducted to assess the variations of multiple model inputs. Model inputs that were varied in the PSA were related to the proportions of the patients who underwent leukapheresis, bridging chemotherapy, lymph depleting chemotherapy, tisagenlecleucel infusion, or SCT and survival functions for the chemotherapy arm. Probability distributions for these parameters were beta distribution for parameters with two variables or Dirichlet distributions if they have more than two variables. The cost of drugs and routine clinical care were assumed to be certain and were not varied in the PSA. The survival functions for the tisagenlecleucel arm were also not varied as the preponderance of the uncertainty in extrapolation is due to the choice of fitting approach rather than the uncertainties in the survival distribution parameters. The probability of cost-effectiveness was assessed using the cost-effectiveness acceptability curve (CEAC).

Additional analyses

Additional supplementary analyses were done to investigate the impact of alternative survival curve extrapolation approach and success rate of tisagenlecleucel infusion on the base-case ICER. A threshold analysis was also conducted to determine the price of tisagenlecleucel required to achieve thresholds ranging from US$37K (S$50K) to US$148K (S$200K)/QALY gained.

Results

Base-case analysis

In the base-case with a time horizon of 15 years, tisagenlecleucel led to an overall incremental mean gain of undiscounted 0.81 LYs and discounted 0.51 QALYs, respectively, compared with chemotherapy (Table 1). The benefits were attained at an incremental cost of US$258,375 (S$348,807). These translate to incremental cost effectiveness ratios (ICERs) of US$320,200 ($432,269) per LY gained and US$508,531 (S$686,516) per QALY gained.

Table 1. Summary of cost and benefits of tisagenlecleucel versus chemotherapy in base-case and additional scenario analyses.

Base-case	Tisagenlecleucel	Chemotherapy	Incremental
PFLYs	2.534	1.259	1.275
LYs	3.616	2.809	0.807
QALYs	2.064	1.556	0.508
Total Costs (US$)	297,911	39,536	258,375
ICER (US$ per LY gained)			320,200
ICER (US$ per QALY gained)			508,531
Scenario analyses
Survival modelling approach: Log-logistic mixture cure model + standardized mortality ratio of 1.0 after 5-year cure-point^a
PFLYs	3.086	1.259	1.827
LYs	5.233	2.809	2.424
QALYs	2.944	1.556	1.388
Total Costs (US$)	300,865	39,536	261,329
ICER (US$ per LY gained)			107,821
ICER (US$ per QALY gained)			188,292
Success rate of tisagenlecleucel infusion: 100%
PFLYs	3.336	1.259	2.077
LYs	4.414	2.809	1.604
QALYs	2.536	1.556	0.980
Total Costs (US$)	428,623	39,536	389,087
ICER (US$ per LY gained)			242,506
ICER (US$ per QALY gained)			397,188

Abbreviations. PFLY, progression-free life year; LY, life year; QALY, quality-adjusted life year; ICER, incremental cost-effectiveness ratio.

^a Under this scenario, 44% of patients were cured after 5-year and their standardized mortality ratio (SMR) was adjusted to match general population at 1.0. A higher SMR due to toxic effects from previous chemotherapy therapies was not applied as this was already factored in by a shorter lifetime horizon.

Sensitivity analyses

OWSA showed that the ICER was most sensitive to time horizon, utility in PF, proportion of autologous SCT in salvage chemotherapy arm and cost of tisagenlecleucel (Figure 4). The ICER increased substantially to over US$800K (S$1.1M)/QALY when time horizon was shortened to 5 years. This was not unexpected given there is a shorter time frame to accrue clinical benefits of tisagenlecleucel after committing a high cost upfront. Using the lower bound of the utility value in PF health state increased the ICER to US$716,219 (S$966,896)/QALY, while the higher bound of the utility value lowered the ICER to US$394,216 (S$532,192)/QALY. The ICER spread was similar when the proportion of patients who underwent autologous SCT in salvage chemotherapy arm and cost of tisagenlecleucel were varied by ±20%

Figure 4. OWSA tornado diagram. Abbreviations. QALY, quality-adjusted life year; PF, progression-free; SMR, standardized mortality ratio; LT-survival, long-term survival; SCT, stem-cell transplantation; PD, progressed disease.

The PSA result was largely congruent with base-case analysis where the mean probabilistic ICER of US$452,352 (S$610,676)/QALY gained was marginally higher than the base-case ICER. This illustrated that tisagenlecleucel was consistently more effective but remained more expensive than chemotherapies. The CEAC demonstrated that compared with salvage chemotherapy, tisagenlecleucel had a 0% probability of being cost-effective at a willingness-to-pay threshold of ∼ US$280K (S$378K)/QALY or less, and a 50% probability of being cost-effective at ∼ US$423K (S$571K)/QALY (Figure 5).

Figure 5. CEAC showing the likelihood of tisagenlecleucel being cost-effective compared to salvage chemotherapy across different WTP thresholds. Abbreviations. CEAC, cost-effectiveness acceptability curve; WTP, willingness to pay.

Scenario analyses

The results from additional scenario analyses performed were shown in Table 2. All the scenario analyses showed exceedingly high ICER. To explore the likelihood of a cure fraction among patients treated with tisagenlecleucel, MCM-loglogistic was used to fit the survival data of the patients in JULIET trial. Our results showed that the ICER remained high (US$188,292 (S$254,194)/QALY) and uncertain, when 45% of patients were assumed to be cured from DLBCL and remained alive over the time horizon of 15 years. As a fraction of patients (23%) in JULIET trial failed to receive the infusion successfully due to disease progression or manufacturing failure, we also modelled a scenario whereby all patients were successfully infused with tisagenlecleucel and the ICER decreased to US$S$397,188 (S$536,204)/QALY.

Table 2. Willingness-to-pay threshold analysis.

Willingness-to-pay threshold (US$)^a	Target price of tisagenlecleucel (US$)	Corresponding discount (%) from price point of ∼ US370K
∼37K/QALY	14,271	96
∼74K/QALY	42,244	89
∼111K/QALY	70,216	81
∼148K/QALY	98,189	74

^a These willingness-to-pay thresholds correspond to thresholds in local currency: S$50K, 100K, 150K and 200 K/QALY.

As shown in Table 2, our threshold analysis indicated that the price of tisagenlecleucel needed to drop by 96% to US$14,271 (S$19,266) in order to reduce the ICER to US$37,037 (S$50K)/QALY. At other thresholds between US$74,074 (S$100K) to US$148,148 (S$200K)/QALY, its price ranged from US$42,244 (S$57,029) to US$98,189 (S$132,555).

Discussion

The availability of tisagenlecleucel and other CAR-Ts provides more treatment options for patients with r/r DLBCL and closes a therapeutic gap for those who are unable to undergo curative SCT and have limited prognosis. While acknowledging the unmet clinical need and early promising outcomes, it is equally important for policymakers to consider if the benefits justify its price tag in consideration of long-term sustainability and affordability of the healthcare system. Administering CAR-Ts to all eligible patients would increase annual healthcare costs in Singapore by approximately US$22M (S$30M) to US$26M (S$35M) at steady state. To our best knowledge, this is the first study conducted in collaboration with clinical experts, to examine the cost-effectiveness of tisagenlecleucel versus salvage chemotherapy in patients with r/r DLBCL in Singapore.

Results from our study suggested that treatment with tisagenlecleucel, at a price tag of US$370K (S$500K), did not represent good value for money from a healthcare system perspective. While a large upfront cost was incurred during the decision tree stage, clinical benefits and utilities gain of tisagenlecleucel were only realized much later in the PSM over a long-term extrapolated time horizon beyond trial period. Therefore, when outcomes were discounted to present value, these benefits were not able to counteract the high incremental cost of tisagenlecleucel.

Our findings were robust across both the OWSA and the PSA. In the OWSA, the cost of tisagenlecleucel was one of the main drivers of the variation in ICER. Even at the lower limit of the ICER range, the additional benefits provided by tisagenlecleucel remained not cost-effective compared to salvage chemotherapy under commonly cited willingness-to-pay (WTP) thresholds. Based on the CEAC results, the probability of tisagenlecleucel being cost-effective was 0% at thresholds between US$100K to US$280K/QALY gained. Our scenario analysis confirmed that for tisagenlecleucel to be cost-effective at WTP thresholds between ∼ US$37K (S$50K) and ∼ US$148K (S$200K)/QALY, its price would need to be reduced by approximately 74%–96%.

Our study is associated with several limitations. Firstly, uncertainty surrounding the real clinical benefits of tisagenlecleucel inherently translates to uncertainty in its cost-effectiveness. There was no guarantee of the durability of early promising clinical benefits of tisagenlecleucel until longer term data become available, one of which includes real-world efficacy and safety data gathered from ongoing post-marketing study by the Center for International Blood & Marrow Transplant Research Cell Therapy registry. Preliminary results mirrored JULIET trial outcomes and demonstrated an overall response rate of 59.6% (28–47 patients) over a median follow-up of 5.8 months (0.9–8.9 months)²⁷. However, given the absence of head-to-head trial data to date, it remained challenging to determine the comparative benefits of tisagenlecleucel versus salvage chemotherapy. The ongoing phase 3 BELINDA randomized controlled trial, comparing tisagenlecleucel with standard of care (platinum-based immunotherapy followed by high dose chemotherapy and ASCT) will shed more insights into its comparative effectiveness and inform future analyses.

Meanwhile, growing experience with anti-CD19 CAR-Ts have demonstrated treatment resistance that leads to eventual relapse in patients who initially achieved remission. Key identified mechanisms that underlie CAR T-cell resistance include T-cell exhaustion and loss of target antigen on tumor cells²⁸. Relapse with antigen-positive disease presents a potential opportunity for re-treatment with CAR-T cells. Treatment with CAR that targets other antigens such as CD20 or combination therapy with a checkpoint inhibitor has also been observed in clinical practice. For example, combined use of CAR-T therapy and PD-1 or PD-L1 inhibitors has shown some promise²⁹. The impact of such combination or sequential use of these therapies on its cost-effectiveness was not captured, as they could not be quantified at this time due to lack of clinical data. Nonetheless, unless a significant incremental clinical benefit could be demonstrated with such management strategies, re-treatment with another CAR-T therapy or addition of another drug to the high-cost CAR-T therapy would significantly add to the costs of management and is unlikely to represent cost-effective use of healthcare resources.

Secondly, utilities values applied in our model might not be entirely representative or generalizable to the Singapore patient population as these studies were conducted in Western countries. It was not apparent from the Wang et al. study if there were any Asian representation among the DLBCL patients recruited from the UK’s population-based Haematological Malignancy Research Network²⁶. Acknowledging that non-Singapore specific utility values were used, OWSA was conducted to assess the impact of varying the utility values. While utility in PF health state appeared to be a key driver, the ICER value of US$380,545 (S$513,735)/QALY was still higher than commonly cited thresholds even with the most optimistic utility figures.

Lastly, healthcare cost could be underestimated either due to unavailability or lack of complete cost data of adverse events and long-term follow up for patients who have progressed. For example, in calculating the management cost of CRS, only the costs of tocilizumab and ICU stay were included. Costs of other monitoring tests and long-term costs associated with management of neurotoxicity were not considered due to lack of information. However, given the high cost of tisagenlecleucel, addition of these costs was unlikely to impact the ICER significantly.

Comparisons with other evaluations are complicated by methodological differences, choice of model inputs and differences in the healthcare cost structure and policies in various countries. However, findings from other studies consistently showed that the use of tisagenlecleucel in r/r DLBCL patients at current price did not provide value for money. With a 5-year PFS of 25% (which was similar to our base-case PFS at 24%), Lin et al. showed that the ICER of tisagenlecleucel remained high at US$223K/QALY (S$301K/QALY) (95% CI, US$123K to U$1,170K/QALY [S$166K to S$1,581K/QALY]) from the US healthcare payer perspective over a lifetime horizon³⁰. Meanwhile in the UK, the ICER range accepted by the National Institute for Health and Care Excellence committee was GBP43K-GBP55K/QALY (US$58K-74K/QALY) only after accounting for an undisclosed amount of discount and rebates³¹. In Canada, analyses done by the Canadian Agency for Drugs & Technologies in Health showed that the ICER was CAD212K/QALY (US$164K/QALY) when compared to salvage chemotherapy and had only a 1.8% probability of being cost-effective at CAD100K/QALY (US$77K/QALY) threshold³². In Australia, the Medical Services Advisory Committee recommended public funding of tisagenlecleucel for lymphoma with a risk-sharing agreement in place to address uncertainties in its clinical- and cost-effectiveness after a price reduction. Similarly, high ICERs were reported for axicabtagene ciloleucel, another anti-CD19 CAR-T therapy marketed for treating DLBCL^33–35.

In all, to deliver effective and cost-effective care to patients, a substantial price reduction was required by all jurisdictions to appropriately reduce the cost of tisagenlecleucel and ensure sustainability of funding by public healthcare systems.

Conclusion

Based on current evidence, our study demonstrated that tisagenlecleucel for treating r/r DLBCL in patients who have failed at least 2 lines of systemic therapies was associated exceedingly high ICER from the Singapore healthcare system perspective, which is unlikely to represent good use of healthcare resources.

Transparency

Declaration of funding

This study was not funded.

Publication fees were provided by the Government of Singapore, Ministry of Health.

Declaration of financial/other interests

WY Hwang is a clinical investigator in the BELINDA study funded by Novartis and has received research funding and honorarium from Novartis. All authors declare that they have no competing interests.

JME 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

BPC, KYG developed the economic model, performed the analyses, collected and reviewed the data, interpreted the results and drafted the manuscript. MIAZ, LL and KN reviewed the data, interpreted the result and revised the manuscript. WYKH, MLMP provided clinical input and validated model assumptions.

Acknowledgements

The authors would like to thank Dr. Lee Yee Mei and her team from the National University Hospital and Ivy Cheong Wai Cheng from the Singapore General Hospital for providing cost data for SCT.