Cost-effectiveness analysis of mosunetuzumab for treatment of relapsed or refractory follicular lymphoma after two or more lines of systemic therapy in the United States

Abstract

Aims

Mosunetuzumab has received accelerated approval by the US Food and Drug Administration for adult patients with relapsed or refractory (R/R) follicular lymphoma (FL) after two or more lines of systemic therapy. We evaluated the cost-effectiveness of mosunetuzumab for the treatment of R/R FL from a US private payer perspective.

Materials and Methods

A partitioned survival model simulated lifetime costs and outcomes of mosunetuzumab against seven comparators: axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), tazemetostat (taz, EZH2 wild-type only), rituximab plus lenalidomide (R-Len) or bendamustine (R-Benda), obinutuzumab plus bendamustine (O-Benda), and a retrospective real-world cohort (RW) based on current patterns of care derived from US electronic health records (Flatiron Health). Efficacy data for mosunetuzumab were from the pivotal Phase II GO29781 trial (NCT02500407). Relative treatment efficacy was estimated from indirect treatment comparisons (ITCs). Costs included were related to treatment, adverse events, routine care, and terminal care. Except for drug costs (March 2023), all costs were inflated to 2022 US dollars. Costs and quality-adjusted life-years (QALYs) were used to calculate incremental cost-effectiveness ratios (ICERs). Net monetary benefit (NMB) was calculated using a willingness-to-pay (WTP) threshold of $150,000/QALY.

Results

Mosunetuzumab dominated taz, tisa-cel, and axi-cel with greater QALYs and lower costs. Mosunetuzumab was projected to be cost-effective against R-Benda, O-Benda, and RW with ICERs of $78,607, $42,731, and $21,434, respectively. Mosunetuzumab incurred lower costs but lower QALYs vs. R-Len. NMBs showed that mosunetuzumab was cost-effective against comparators except R-Len.

Limitations

Without head-to-head comparative data, the model had to rely on ITCs, some of which were affected by residual bias. Model inputs were obtained from multiple sources. Extensive sensitivity analyses assessed the importance of these uncertainties.

Conclusion

Mosunetuzumab is estimated to be cost-effective compared with approved regimens except R-Len for the treatment of adults with R/R FL.

Keywords:

• Cost-effectiveness
• mosunetuzumab
• lenalidomide
• bendamustine
• axi-cel
• tisa-cel
• rituximab
• obinutuzumab
• tazemetostat
• follicular lymphoma

JEL codes:

• I11
• I1
• I
• I19

Disclaimer

As a service to authors and researchers we are providing this version of an accepted manuscript (AM). Copyediting, typesetting, and review of the resulting proofs will be undertaken on this manuscript before final publication of the Version of Record (VoR). During production and pre-press, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal relate to these versions also.

Introduction

Follicular lymphoma (FL) is the most common of the indolent non-Hodgkin’s lymphomas (NHLs) and accounts for approximately 20% of all NHLs in the United States (US)[1].

Despite its indolent nature, FL is considered incurable with current routine therapies[2]. Around 20% of patients will experience early progression following first-line chemoimmunotherapy, which is associated with poor prognosis (5-year overall survival [OS] of 50%)[3]. Current treatments, such as cytotoxic chemotherapy, carry a risk of toxicities that can seriously impact health-related quality of life (HRQoL)[1, 2]. Relapsing FL can further place a burden on patients, both experientially as well as financially[4].

Mosunetuzumab is a novel CD20xCD3 T-cell engaging bispecific antibody that has received accelerated approval by the US Food and Drug Administration (FDA) for the treatment of adult patients with relapsed or refractory (R/R) FL who have received two or more lines of systemic therapy[5, 6]. The FDA accelerated approval was based on the results from the Phase II GO29781 study (NCT02500407) which showed high and durable response rates for fixed-duration mosunetuzumab and a favorable safety profile[7].

Mosunetuzumab joins multiple approved therapeutic options for the treatment of multiply relapsed follicular lymphoma. The objective of this study was to assess the cost-effectiveness (CE) of mosunetuzumab and other treatments for R/R FL patients from a US payer perspective.

Methods

Model overview

A three-health state partitioned survival model (PSM), including progression-free survival (PFS), post-progression disease (PPS), and death (Figure 1), was developed to simulate the lifetime costs and benefits of mosunetuzumab against relevant comparators in adult patients with R/R FL.

The PSM approach is commonly used in oncology cost-effectiveness analyses for interventions that are expected to prolong life expectancy and impact quality of life[8-15]. PSMs use time-to-event survival data to model the proportion of patients who are in the PFS, PPS, and death states dependent on time since trial initiation. All patients begin the model in the PFS state. Patients who have experienced disease progression but have not died yet are in the PPS state, and death is an absorbing state. In our model, patients in the PFS state could be either on or off treatment to account for patients who finished therapy but remained in the PFS state. Upon entering the PPS state, patients were assumed to move to a further line of treatment and accrue a one-off cost for post-progression treatments. The ‘transitions’ of patients from the progression-free and post-progression health states into the death state were determined by the OS curves as well as using age- and gender-specific US background mortality[16].

The perspective of the CE analysis was that of a private payer in the US. A lifetime time horizon (60 years) was used along with a weekly model cycle. A half-cycle correction was applied to the model. Cost and health benefits were discounted at an annual rate of 3%[17] and a willingness-to-pay (WTP) threshold of $150,000 per quality-adjusted life year (QALY) was used to assess cost-effectiveness aligned with other published oncology CEAs [10, 12, 13, 18-21]. The model was developed in Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA, USA) in accordance with best practices, including the CHEERS guidelines[22].

Modelled treatments

Clinical guidelines[23] and clinician opinion identified the following treatments to be included in the analysis: axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), tazemetostat (taz, EZH2 wild-type population only), rituximab in combination with lenalidomide (R-Len) or bendamustine (R-Benda), and obinutuzumab in combination with bendamustine (O-Benda). A real-world cohort (RW) derived from a US electronic health record database (Flatiron Health)[24] was also included as a comparator representing a proxy for real-world patterns of care prior to the new therapies becoming available.

Data and patient selection

Individual patient data (IPD) for mosunetuzumab were available from the GO29781 study (NCT02500407), specifically the R/R FL subgroup of the group B expansion[7]. All patients were aged 18 years or older with histologically confirmed follicular lymphoma (grade 1–3a) and an Eastern Cooperative Oncology Group performance status of 0–1. Patients had disease that was R/R to two or more lines of systemic therapy, including an anti-CD20 therapy and an alkylating agent[7].

IPD were also available for R-Benda (NCT02187861, data on file, Genentech Inc.), O-Benda (NCT01059630, data on file, Genentech Inc.), and RW[24]. Clinical efficacy data (PFS and OS), baseline characteristics, Kaplan Meier (KM) curves and other inputs, for other comparators except the RW cohort were taken from the clinical trials programs of the respective treatments (Supplemental Table 1).

Treatment efficacy

The relative efficacy of mosunetuzumab vs. the comparators was estimated from indirect treatment comparisons (ITCs) due to a lack of comparator arm in the GO29781 study [updated analysis of Bosch et al., 2023[25]]. For R-Benda, O-Benda, and RW, inverse-probability of treatment weighting (IPTWs)[26] from internal data were used, given that IPD were available for these treatments. For the other comparators, a matching-adjusted indirect treatment comparison (MAIC)[27] accounting for differences in trial population baseline characteristics was employed. KM curves were estimated for all PFS and OS curves for each treatment. Survival curves were extrapolated beyond the trial period either using independently fitted parametric distributions or by applying a hazard ratio (HR) from the ITCs to a reference survival curve (when the proportional hazards [PH] assumption was not violated). Several parametric distribution functions (Exponential, Weibull, Log-normal, Generalized gamma, Log-logistic, Gompertz, and Gamma) were fitted to the PFS and OS KM data for mosunetuzumab and each of the comparators using the results from the ITCs (MAICs or IPTWs). The base case for each treatment followed the model selection process algorithm suggested by Latimer[28] taking into account the fit of the models to the observed data and the plausibility of the extrapolated portion of the models. Goodness of fit was based on the Akaike Information Criterion (AIC) and the Bayesian information criterion (BIC). Visual inspection by clinical experts validated the plausibility of the extrapolations. A frequentist model was used in the base case OS extrapolations.

Appendix Supplemental Figure 1 to Supplemental Figure 7 show the PFS and OS for mosunetuzumab matched to each comparator whereas Table 1 includes the PFS and OS HRs for mosunetuzumab matched to each comparator. Additional parametric forms for survival time were tested in scenario analyses if there was substantial uncertainty in base case estimates. Also, for the comparison of mosunetuzumab against chimeric antigen receptor (CAR) T-cell therapies, a scenario assuming no difference in OS (i.e., HR = 1) was tested as no marked differences between the survival curves were observed.

Finally, waning of the relative treatment effect was not considered on the basis that patients have been off mosunetuzumab treatment for at least 20 months at the end of follow-up in the GO29781 study.

Time on treatment/discontinuation

Time to off treatment (TTOT) dictated the proportion of patients who are progression-free and on vs. off treatment at each model cycle. For mosunetuzumab, survival analysis using clinical trial data estimated TTOT. For all other treatments, TTOT was set equal to the selected parametric distribution for PFS, capped at the treatment-specific maximum number of cycles based on their respective labels and/or trial publications[29-32] as a proxy due to the lack of data. Note that axi-cel and tisa-cel are both administered as one-off treatments[29, 30] at the beginning of the model, such that TTOT is almost immediate.

Health state utilities

Benefits were measured by QALYs derived from the EuroQol 5 Dimension-5 Level (EQ-5D-5L) data collected in the GO29781 study. Utilities were estimated for PFS and PPS (Table 1) using a mixed regression model utilizing patient-level random effects and controlling for centralized baseline utilities using a US-based value set[33] and occurrence of Grade ≥ 3 adverse events (AEs). As data were relatively immature to model post-progression utility, scenario analyses were conducted using published values from Wild et al., 2006[34] and Cognet et al., 2014[35].

Resource use and costs

The model accounted for seven cost components: drug, administration, management of Grade ≥ 3 AE and cytokine release syndrome (CRS) of any grade, routine care, post-progression therapy, and terminal care. All costs are presented in 2022 US Dollars (USD), except drug acquisition costs which were current as of March 2023. Where applicable, the medical care component of the consumer price index (CPI) was used to inflate costs to 2022 USD[36].

Drug costs

Drug costs were calculated based on the dosing schedule of each regimen ascertained from US package inserts (PIs) and/or trial publications (Supplemental Table 2), drug acquisition costs based on wholesale acquisition cost (WAC) prices sourced from AnalySource®[37] (Table 1), and TTOT. Vial sharing was excluded in the base case but was tested in scenario analyses.

Administration costs

Administration costs were applied for intravenous (IV) treatments and CAR T-cell therapies.

For IV treatments, resource use was calculated based on the time needed to administer the drug, taking into account information on dosing and infusion rate found in US PIs[6, 32, 38, 39]. The time required for administering each drug compound was then matched to its relevant Current Procedural Terminology (CPT) code with unit costs based on the 2022 Centers for Medicare and Medicaid Services (CMS) physician fee schedule (Table 1)[40].

As indicated in their US PIs[29, 30], the administration of CAR T-cell therapies requires additional resources (Table 1) which were applied once in the model at the beginning of treatment. Resources accounted for leukapheresis, administration, lymphodepleting chemotherapy, and hospitalization.

Adverse event costs

Adverse events of Grade ≥3 severity and occurring in at least 5% of patients treated with any regimen were included, except those not expected to have significant cost impacts. The total cost of managing adverse events for each regimen is outlined in Table 1 and was applied once in the model at the beginning of treatment. Details on the AE rates and associated unit costs were based on the study by Lin et al., 2023[41] except for RW where the same AE cost as mosunetuzumab was assumed due to the lack of data.

CRS costs

Aside from the AE rates, patients treated with mosunetuzumab and CAR T-cell therapies face the additional risk of CRS. This was modelled separately in the model and all grade events were included for mosunetuzumab, axi-cel, and tisa-cel. Aligned with recommendations by the National Comprehensive Cancer Network® (NCCN)[42] and the US PIs of each drug [6, 29, 30], the cost of CRS reflected the use of tocilizumab, ICU admission, and non-ICU hospitalization for each regimen. The CRS costs stratified by regimen weighted by the proportion of patients experiencing CRS are tabulated in Table 1 and were applied once at treatment initiation.

Routine care costs

Routine care costs associated with the management of FL were included. These were sourced from expert opinion and included routine visits with lab work (i.e., complete blood count [CBC], comprehensive metabolic panel [CMP], and lactate dehydrogenase [LDH]) every 3 months as well as a computed tomographic chest abdomen pelvis (CT-CAP) every 6 months. For CAR T-cell therapies, a positron emission tomography (PET) scan at day 90 would be added and visits for lab work would occur monthly for the first 6 months. At progression, an additional office visit was assumed to occur for lab work, PET scan, and a biopsy, irrespective of the treatment. Table 1 summarizes the routine care costs applied in the model. In scenario analysis, an alternative value for routine care costs was tested applying the health state costs used by Oluwole and al., 2023[13] (i.e., $292/month).

Post-progression treatment costs

Upon progression, patients are assumed to receive subsequent therapy and incur the related costs. These were computed as a weighted average cost of a basket of therapies using subsequent therapy from the GO29781 study and associated mean treatment duration. The resulting cost ($57,574) is applied in the model as a one-off cost upon progression and is the same for all comparators in the model.

Two scenarios were included to explore the impact of applying differential post-progression treatment costs for each comparator on the model’s results. The basket of therapies considered and associated market shares were informed by the most recent NCCN guidelines[43] and clinician opinion. In the first scenario, only newer therapies were considered (“newer therapies only”) and the market shares were assumed to be equally distributed among mosunetuzumab, CAR T-cell therapies (50%/50% split between axi-cel and tisa-cel), R-Len, and Taz. In the second scenario (“newer therapies + R-mono”), R-mono was added to the basket of therapies. Contrary to the base case assumption, all patients experiencing progression receive post-progression treatment costs (i.e., market shares sum-up to 100%) and re-treatment with same regimen was not permitted in both scenarios. As such, for each comparator in the model, the market shares were calibrated to exclude the initial treatment. For CAR T-cell therapies, receiving one CAR T-cell therapy precluded treatment with another CAR T-cell therapy. The resulting post-progression treatment costs for each comparator in both scenarios are detailed in Supplemental Table 3.

Terminal care costs

A one-time terminal care cost of $26,275 based on a recent relapsed/refractory diffuse large B-cell lymphoma analysis using a US payer perspective[44] was applied in the model.

Outcomes

Key outcomes of the cost-effectiveness analysis were: total costs, total life-years (LYs), and total QALYs (absolute and incremental), incremental cost-effectiveness ratios (ICERs), and net monetary benefits (NMBs). Results are presented for each pairwise comparison in line with the ITC approaches that were used to generate the comparative effectiveness inputs. To assess the impact of uncertainty in model parameters and structural assumptions, sensitivity analyses were performed, including deterministic sensitivity analyses (DSA), probabilistic sensitivity analyses (PSA) with 1,000 simulations, and scenario analyses.

Results

Base case

Base case results are presented in Table 2 and Figure 2. All newer therapies resulted in higher total costs when compared against mosunetuzumab, with axi-cel and tisa-cel accruing the highest costs (Table 2). The largest components of total costs were those associated with the cost of the drug. Mosunetuzumab generated higher LYs and QALYs in all pairwise comparisons except against R-Len. More specifically, incremental LYs ranged from 1.0 to 9.7 and incremental QALYs ranged from 0.6 to 7.7 for the comparison of mosunetuzumab against tisa-cel and taz, respectively (Table 2). Mosunetuzumab dominated taz, axi-cel, and tisa-cel with greater QALYs and lower costs and was cost-effective against the RW cohort, O-Benda, and R-Benda with ICERs of $21,434, $42,731, and $78,607, respectively (Table 2). At a WTP of $150,000 per QALY, the NMB results show that mosunetuzumab is cost-effective against all comparators except R-Len in the base case analysis (Figure 2).

Deterministic sensitivity analyses

The deterministic sensitivity analysis finds that the most impactful parameters on the NMB include parameters related to drug costs, utility in PFS and PPS, and baseline age. Other cost inputs such as administration costs, AE costs, and CRS costs, had a minor impact on the NMBs. Across mosunetuzumab comparisons, NMB values are positive indicating that mosunetuzumab remains cost-effective versus each comparator at the assumed WTP of $150,000 per QALY, with the exception that mosunetuzumab is not cost-effective against R-Len across the various one-way sensitivity analyses performed.

Probabilistic sensitivity analyses

The 1,000 PSA simulations indicated robustness of the base case results to parameter uncertainty. In the PSA, mosunetuzumab was cost-effective at the WTP of $150,000 per QALY in 100%, 98%, 96%, 87%, 66%, 61%, and 43% of the simulations against taz, axi-cel, RW, tisa-cel, O-benda, R-benda, and R-len, respectively. In addition, Supplemental Figure 8 shows that the NMB results for mosunetuzumab vs. each comparator in probabilistic analyses are consistent with the deterministic results (Figure 2).

Figure 3a-b displays the cost-effectiveness acceptability curves of mosunetuzumab vs. R-Len using the base case OS distribution (exponential) and an alternative distribution (log-logistic) to address uncertainty around OS data maturity and discordance of the most appropriate modelling approach among the clinical experts. Mosunetuzumab was cost-effective in 61% of the PSA simulations compared to 43% if the log-logistic versus the exponential distribution was chosen, respectively, at a WTP per QALY of $150,000 per QALY.

Scenario analyses

Supplemental Table 3 shows the results of additional scenario analyses on the NMB of mosunetuzumab vs. each comparator. Aligned with the base case findings, mosunetuzumab remains cost-effective against all comparators except R-Len in most scenario analyses. Three types of scenarios led to mosunetuzumab being cost-effective versus R-Len: the use of a shorter time horizon (10 years), the use of published utility values from Wild et al., 2006[34] (NMB = $21,731) and Cognet et al., 2014[35] (NMB = $48,614), and alternative parametric curves for the OS of R-Len (Weibull: NMB = $293,958; Log-logistic: NMB = $115,000).

For the comparison of mosunetuzumab against axi-cel and tisa-cel, the various scenario analyses corroborated the base case results that mosunetuzumab is cost-effective vs. both CAR T-cell therapies. These findings persisted even after assuming no difference in OS between treatments. Increasing the costs accrued during progression by testing alternative differential post-progression treatment costs or relying on another source for the estimation of routine care cost also had a modest impact on the NMBs.

Discussion

Older treatments for patients with R/R FL, including chemoimmunotherapy, have associated toxicities that may impact quality of life, while newer treatments appear to have higher efficacy, but still are associated with significant, although different, toxicities as well as higher costs. Modelling allows for combining up-to-date data on costs, benefits, and risks to assess which therapies for R/R FL after two or more lines of systemic therapy are the most cost-effective.

This model compares mosunetuzumab with potential comparators that are either approved or recommended for the treatment of R/R FL, consistent with guidelines and expert clinical opinion. Mosunetuzumab appears cost-effective versus all treatments, except for R-Len in the base case analysis.

This US-based R/R analysis contributes uniquely to the FL cost-effectiveness literature for several reasons. First, our study relied on IPD for mosunetuzumab, and two comparators included in the model (i.e., O-benda and R-benda). Second, both approved CAR T-cell therapies in R/R FL (i.e., axi-cel and tisa-cel), have been incorporated as comparators in the model. Lastly, a comparator reflecting real-world treatment patterns informed by data from Flatiron Health, a US electronic health record database, was also included. In a recent systematic review of the literature[45], 24 cost-effectiveness studies were identified in FL, 5 of which were in R/R patients. Only one study was identified [46] that assessed cost-effectiveness of treatments in R/R FL patients in the third line setting. The study compared ibritumomab tiuxetan – production of which has been subsequently discontinued – with 4-dose and 8-dose rituximab regimens in patients with relapsed FL in the Netherlands. Since then, a few additional cost-effectiveness studies comparing newer treatments for patients with R/R FL and focusing on US payers have been published employing either a PSM[9,13,15] or a Markov model[18, 19]. Guzauskas et al., 2018[9] used a 3-state PSM to compare O-benda followed by obinutuzumab monotherapy versus bendamustine monotherapy for the treatment of FL patients who relapse after or are refractory to a rituximab-containing regimen (R/R-rituximab). The study relied on the GADOLIN trial which recruited patients with rituximab-refractory indolent NHL and was adjusted to reflect the R/R-rituximab population using data from the National LymphoCare Study. The authors found that O-benda followed by obinutuzumab monotherapy was likely to be cost-effective compared to bendamustine monotherapy in R/R-rituximab FL patients. Potnis et al., 2023[18] developed a Markov model to assess the cost-effectiveness of axi-cel vs standard of care (SOC) therapies in adults with R/R FL. Patients entered in the model in third line and received either axi-cel or SOC. Transition probabilities were informed by the ZUMA-5 trial and the Lymphoma Epidemiology of Outcomes Consortium for Real World Evidence (LEO CReWE) cohort study, respectively. Following progression, patients could initiate another line of therapy (up to 2 lines of targeted therapy are modelled and included taz and copanlisib) or move to best supportive care. At a WTP of $150,000 per QALY, the base case analysis reported that axi-cel was not cost-effective vs. SOC with an ICER of $182,127 per QALY, a finding that was supported by the majority of their PSA iterations. The authors conducted a series of scenario analyses in which axi-cel was either dominated, not cost-effective, or cost-effective against SOC, and thus highlighting the sensitivity of the results to key modelling assumptions (e.g., extrapolation assumptions, time horizon, costs assumptions, data to inform SOC). By contrast, Oluwole et al, 2024[15] employed a 3-state PSM to compare axi-cel to SOC in R/R FL who have had at least two prior lines of systemic therapy. The authors found that axi-cel would be considered cost-effective compared to SOC with a base case ICER of $88,300. They highlighted the use of SCHOLAR-5 to inform SOC as opposed to the LEO CReWE as one of the main reasons that may explain the discrepancies between both studies. At the American Society of Hematology conference held in December 2023, two studies reported on the cost-effectiveness of mosunetuzumab versus CAR T-cell therapies in R/R FL[13, 19]. Lin et al., 2023[19] developed a Markov model with a time horizon of one to 10 years to compare mosunetuzumab, axi-cel, and tisa-cel. The authors reported that mosunetuzumab was cost-effective against both CAR T-cell therapies at all time points with positive NMBs. The authors acknowledged that longer term follow up data may impact future cost-effectiveness and QALY comparisons. Oluwole et al, 2023[13], using a 3-state PSM and a lifetime horizon, found that axi-cel was cost-effective compared to mosunetuzumab at a WTP of $150,000 per QALY. These findings are contradictive to those reported by Lin et al., 2023[19] and the present analysis. Although various reasons may explain discrepancies in findings between our model and the model by Oluwole et al., 2023, one major difference is worth noting. Contrary to our model, Oluwole and colleagues assumed that 40% of axi-cel patients alive at 5 years experienced long-term remission (i.e., functional cure). The authors relied on a single center trial with 14 FL patients[47] to support this assumption as the model used 24-month ZUMA-5 trial data which is insufficient to properly assess the plausibility of reaching a “functional cure” at 5 years. Without reliable internal/external trial evidence on the plausibility of a functional cure in R/R FL, and in light of past HTA appraisals where a “cure” was deemed plausible independently of the technology[48], exploring a “cure” scenario for axi-cel only may have biased the results. Unfortunately, the authors did not present the cost-effectiveness results without imposing a cure.

Our model has several limitations worth noting. First, due to the lack of comparator arm in the GO29781 study, the model relied on ITCs where residual bias from substantial differences in study design and population across trials persisted in some comparisons, especially for R-Len, despite efforts to adjust for cross-study heterogeneity[25]. As with any ITC, the extent to which the seven comparators are suited for head-to-head comparison should be taken into account, and how well the methods used (MAIC and IPTW) can ameliorate potential comparability issues. For R-Len, the ITC failed to control for key prognostic factors mainly due to differences in trial populations with ∼53% and 0% of the patients in AUGMENT trial [49] being 2L and refractory to rituximab compared with 0% and ∼79% in the GO29781 trial[7], respectively. This may have significantly biased the results against mosunetuzumab. The statistical uncertainty around the ITC results has been incorporated in the probabilistic sensitivity analysis by relying on ITC-derived bootstrapped parameter estimates. Our analysis showed that the probabilistic and deterministic results were consistent. Additionally, a scenario analysis using the unadjusted trial data (i.e., no ITCs) did not change the conclusions drawn from the base case analysis.

Second, the validity and robustness of partitioned survival models beyond the observed trial duration is dependent on the maturity of the survival data used. In the ongoing mosunetuzumab trial GO2971, the OS data cut used in the current model is considered immature, with only 11% events recorded at the data cut off used (data on file, Genentech Inc.). The observation period for several of the comparators is also quite short, meaning that the parametric estimates are arguably uncertain. In any type of model, immature survival data represent an area of uncertainty in predictions made. We therefore tested various time horizons in scenario analyses and assessed the robustness of the results through extensive deterministic and probabilistic sensitivity analyses. The results are relatively robust across all sensitivity analyses performed except for one scenario. For R-Len, the AIC and BIC statistics were inconclusive, with all models being within 5 points by both measures, indicating that no single model was clearly the best fit. In a scenario analysis using a different parametric specification for the OS survival curve of R-Len, mosunetuzumab became cost-effective vs. R-Len (NMB = $115,000) and this result held in 61% of the PSA simulations based on a WTP of $150,000 per QALY. Nevertheless, uncertainty remains as both mosunetuzumab and R-Len have immature survival data and the ITC could not match on important covariates. For the comparison of mosunetuzumab against axi-cel and tisa-cel, the results of the ITC favoured the CAR T-cell therapies in terms of PFS (HR > 1) but favoured mosunetuzumab (HR > 1) in terms of OS. Although it may be expected that a PFS benefit be associated with an OS benefit (and vice-versa), a recent systematic literature review of clinical trials in R/R FL[50] found a weak PFS-OS correlation in this condition. Therefore, it is plausible to observe treatments being associated with a PFS benefit (HR <1) coupled with a loss in OS (HR > 1) and vice-versa. In our ITC, although the point estimates indicated a more favourable OS for mosunetuzumab relative to the CAR T-cell therapies, no marked differences between the survival curves were observed. Consequently, a scenario analysis was conducted assuming no difference in OS (i.e. HR = 1) between mosunetuzumab and CAR T-cell therapies. The results from this scenario analysis indicated that mosunetuzumab remained cost-effective against both CAR T-cell therapies.

Third, the distribution of subsequent treatments upon progression may impact the clinical and economic outcomes across the comparators under comparison which is an area of uncertainty. It was not possible to explicitly model the impact of different subsequent therapies on the clinical outcomes of a given treatment, as this cannot be reliably informed by the available data. As such, our model assumed that any survival differences are only driven by the comparator survival curves aligned with other published CEAs[11, 13, 14]. However, we have explored in scenario analyses the impact of applying differential post-progression treatment costs for each comparator on the economic outcomes. Ideally, specific baskets of subsequent therapies based on actual market share information would be used to calculate the post-progression treatment costs of each comparator in the model. Unfortunately, this is rarely feasible, particularly when comparisons are informed by MAICs, as full information on this in the indication of interest is almost never available from published articles. To mitigate this uncertainty, two distinct scenarios were tested, and the results were aligned with the base case findings.

In addition, discounts on drug list prices were excluded owing to lack of data, as these discounts are proprietary which may impact the results. Also, utility values were based on data collected in the trial in the base case which differ from other published sources. In addition, the short-term follow-up period in the trial meant that the data were immature to estimate the utility in post-progression. Alternative published utility values were tested in scenario analyses and the results were unchanged. It is important to note that these alternative utility values do not appear to be US-based and hence, may not be suitable for our US model. It is also possible that our model underestimated the costs in the post-progression state. However, scenario analyses were conducted varying the post-progression treatment costs and the routine care costs. These scenarios confirmed the robustness of the model’s findings. Lastly, except for RW, the model relied on clinical trial data which may not reflect outcomes in clinical practice. Consequently, the inclusion of RW is an important comparison to be made but required simplifying assumptions that may have impacted the accuracy of its costing.

Conclusions

The cost-effectiveness of mosunetuzumab, assessed from a US payer perspective, can be made against seven separate comparators generally considered to represent standard of care in patients with R/R FL. Using a lifetime horizon, the results of this analysis suggest that fixed-duration treatment with mosunetuzumab is projected to be cost-effective compared with approved regimens with the exception of R-Len for adult patients with R/R FL who have received two or more lines of systemic therapy. Further research and long-term data collection (e.g., GO29781 OS immaturity, plausibility of a functional cure, availability of head-to-head trials) are warranted to reduce the uncertainty around the long-term clinical effectiveness and cost-effectiveness of the use of mosunetuzumab in the treatment of patients with R/R FL.

Transparency

Declaration of funding

Design, study conduct and financial support for the study were provided by F. Hoffmann-La Roche Ltd. and Genentech, Inc.

Declaration of financial/other interests

MM is an employee of the Rutgers Cancer Institute of New Jersey. He has appointments for consultancy or is an advisory board member to Genentech, Inc., F. Hoffmann-La Roche Ltd, Bayer, Juno, Seattle Genetics, Takeda, Teva, and Merck; has institutional research funding from Genentech, Inc., F. Hoffmann-La Roche Ltd, Bayer, GM Biosciences, Immunovaccine Technologies, Janssen, Pharmacyclics, Seattle and Genetics; has received honoraria or stipends from Genentech, Inc., F. Hoffmann-La Roche Ltd, ADC Therapeutics, AstraZeneca, Bayer, BMS, Celgene, Epizyme, Immunovaccine Technologies, IMV Therapeutics, Janssen, Kite, Pharmacyclics, Regeneron, Seagen, Seattle Genetics, Takeda, and Teva; and he owns stocks/and or options from Merck.

JSA was previously an employee of F. Hoffmann-La Roche Ltd. at the time of design and conduct of the study and has no financial conflicts of interest to disclose.

HP and EZ are employees of Medicus Economics, LLC. Medicus Economics, LLC and received consulting fees for research from Genentech, Inc.

DDM is an employee of F. Hoffmann-La Roche Ltd. and may own stocks/and or options from F. Hoffmann-La Roche Ltd.

SWL, SS, and EK are employees of Genentech, Inc. and may own stocks/and or options from F. Hoffmann-La Roche Ltd.

Reviewer disclosures

A reviewer on this manuscript has disclosed that they received GILEAD consultancy fees. Peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Author Contributions

F. Hoffmann-La Roche Ltd, Genentech, Inc., and Medicus Economics, LLC participated in the design of the research, the analysis and interpretation of finding, and in manuscript writing, review, and approval. All authors contributed to the development of the manuscript and maintained control over the final content.

Acknowledgments

The authors would like to acknowledge the medical writing support provided by Scott Johnson (Medicus Economics, LLC) funded by Genentech, Inc. and the third-party medical editorial assistance provided by Ashfield MedComms, an Inizio company, funded by F. Hoffmann-La Roche Ltd.

Previous presentations

Results from this manuscript were presented as a poster at the 2023 ISPOR Annual Meeting (May 7-10, 2023) as well as an encore at the 2023 Society of Hematologic Oncology Annual Meeting (September 6-9, 2023).

Data availability statement

The data that support the findings of this study are available in this article. The model is not publicly available due to its intrinsic commercial value and cannot be shared for legal reasons.

Table 1. Model parameters

Parameter	Base case value	Low value	High value	PSA distribution	Reference/Note
Baseline patient characteristics
Age (years)	60.0	48.0	72.0	Not varied	GO29781 study
Body surface area (m^2)	1.96	1.56	2.35	Not varied	GO29781 study
Body weight (kg)	81.4	65.1	97.7	Not varied	GO29781 study
Proportion male	61.1%	48.9%	73.3%	Not varied	GO29781 study
Treatment efficacy, OS, PH method (if PH assumption was not violated)
HR, mosunetuzumab vs. Taz	0.16	0.07	0.41	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. Axi-cel	0.61	0.14	1.68	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. Tisa-cel	0.86	0.15	1.93	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. RW	0.43	0.19	0.94	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. O-Benda	0.70	0.28	1.58	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. R-Benda	N/A	N/A	N/A	N/A	N/A
HR, mosunetuzumab vs. R-Len	N/A	N/A	N/A	N/A	N/A
Treatment efficacy, PFS, PH method (if PH assumption was not violated)
HR, mosunetuzumab vs. Taz	0.35	0.19	0.57	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. Axi-cel	2.09	1.30	3.12	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. Tisa-cel	1.73	1.10	2.75	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. RW	0.82	0.53	1.27	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. O-Benda	2.10	1.19	3.87	Bootstrapped	Data on file^a
HR, mosunetuzumab vs. R-Benda	N/A	N/A	N/A	N/A	N/A
HR, mosunetuzumab vs. R-Len	N/A	N/A	N/A	N/A	N/A
Utilities
PFS	0.8373	0.8073	0.8657	Beta	GO29781 study
PPS	0.7650	0.7197	0.8117	Beta	GO29781 study
Drug acquisition costs
Mosunetuzumab, 1mg/1mL (1mL)	$594.06	Not varied	Not varied	Not varied	AnalySource^®[37]
Mosunetuzumab, 1mg/1mL (30 mL)	$17,821.78	$14,257.42	$21,386.14	Not varied	AnalySource^®[37]
Tazemetostat, 200 mg (240s ea)	$18,850.00	$15,080.00	$22,620.00	Not varied	AnalySource^®[37]
Axicabtagene ciloleucel	$424,000.00	$339,200.00	$508,800.00	Not varied	AnalySource^®[37]
Tisagenlecleucel	$427,047.70	$341,638.16	$512,457.24	Not varied	AnalySource^®[37]
Obinutuzumab, 25 mg/mL (40 mL)	$7,775.32	$6,220.26	$9,330.38	Not varied	AnalySource^®[37]
Bendamustine, 25 mg (ea)	$400.00	$320.00	$480.00	Not varied	AnalySource^®[37]
Bendamustine, 100 mg (ea)	$1,600.00	$1,280.00	$1,920.00	Not varied	AnalySource^®[37]
Rituximab, 10 mg/mL (10 mL)	$824.24	$659.39	$989.09	Not varied	AnalySource^®[37]^b
Rituximab, 50 mg/mL (10 mL)	$4,121.21	$3,296.96	$4,945.45	Not varied	AnalySource^®[37]^b
Lenalidomide, 20 mg (21s ea)	$17,497.73	$13,998.18	$20,997.28	Not varied	AnalySource^®[37]
Administration costs
CPT code 96413	$140	$112	$168	Gamma	CMS[40]
CPT code 96415	$30	$24	$36	Gamma	CMS[40]
CPT code 96417	$68	$55	$82	Gamma	CMS[40]
CPT code 96374	$40	$32	$48	Gamma	CMS[40]
CPT code 96375	$16	$13	$20	Gamma	CMS[40]
CPT code 36511	$110	$88	$132	Gamma	CMS[40]
Axi-cel, excl. LD chemo drug costs	$30,778	$24,623	$36,934	Gamma	Calculated^c
Tisa-cel, excl. LD chemo drug costs	$25,436	$20,349	$30,523	Gamma	Calculated^c
Axi-cel, LD chemo drug costs	$1,357	$1,086	$1,629	Gamma, Beta^d	Calculated^c
Tisa-cel, LD chemo drug costs	$1,999	$1,600	$2,399	Gamma, Beta^d	Calculated^c
AE costs
Mosunetuzumab	$9,591	$7,673	$11,509	Gamma, Beta^d	Lin et al., 2023[41]
Taz	$448	$359	$538	Gamma, Beta^d	Lin et al., 2023[41]
Axi-cel	$16,512	$13,210	$19,815	Gamma, Beta^d	Lin et al., 2023[41]
Tisa-cel	$10,160	$8,128	$12,192	Gamma, Beta^d	Lin et al., 2023[41]
RW	$9,591	$7,673	$11,509	Gamma, Beta^d	Assumption, same as Mosunetuzumab
O-Benda	$6,099	$4,879	$7,318	Gamma, Beta^d	Lin et al., 2023[41]
R-Benda	$8,913	$7,130	$10,695	Gamma, Beta^d	Lin et al., 2023[41]
R-Len	$9,771	$7,816	$11,725	Gamma, Beta^d	Lin et al., 2023[41]
CRS costs
Mosunetuzumab	$7,136	$5,708	$8,563	Gamma, Beta^d	Calculated^e
Axi-cel	$28,987	$23,189	$34,784	Gamma, Beta^d	Calculated^e
Tisa-cel	$7,913	$6,330	$9,495	Gamma, Beta^d	Calculated^e
Routine care costs
CAR T-cell therapies, weeks 0-26	$44	$35	$53	Gamma	Expert opinion, CMS[40,51]
CAR T-cell therapies, subsequent weeks	$22	$17	$26	Gamma	Expert opinion, CMS[40,51]
Other therapies, weekly costs	$22	$17	$26	Gamma	Expert opinion, CMS[40,51]
One-off cost at progression	$413	$330	$496	Gamma	Expert opinion, CMS[40,51]
Other costs
Post-progression treatment costs	$57,574	$46,059	$69,089	Gamma	Data on file
Terminal care costs	$26,275	$21,020	$31,530	Gamma	Betts et al., 2020[44]

^aBased on an updated analysis of Bosch et al., 2023 [25]. Low and high values correspond to the 95% bias-corrected and accelerated bootstrap interval.

^b For rituximab, based on a weighted average of Rituxan® brand (32.34%; NDC: 50242-0051-21) and bio-similars (67.66%) which are further broken down between Ruxience® (36.74%; NDC: 00069-0238-01), Truxima® (27.51%; NDC: 63459-0103-10), and Riabni® (3.41%; NDC: 55513-0224-01). Market shares were based on Genentech internal data (data on file, Genentech Inc., 2022).

^c Calculated as the sum of leukapheresis, administration, lymphodepleting chemotherapy excluding drug costs, and hospitalization[41].

^d In the PSA, unit costs were varied using a gamma distribution and percentages were varied using a beta distribution.

^e Accounted for costs associated with tocilizumab use, ICU admission, and hospitalization (non-ICU) and weighted by the proportion of patients experiencing CRS[41].

Abbreviations: AE, adverse event; Axi-cel, axicabtagene ciloleucel; CAR, chimeric antigen receptor; Chemo, chemotherapy; CMS, Centers for Medicare and Medicaid Services; CPT, Current Procedural Terminology; CRS, cytokine release syndrome; Ea, each; Excl, excluding; HR, hazard ratio; Kg, kilogram; LD, lymphodepleting; Mosun, mosunetuzumab; N/A, not applicable; O-Benda, obinutuzumab + bendamustine; PH, proportion hazard; PI, Package insert; PFS, progression-free survival; OS, overall survival; R-Benda, rituximab + bendamustine; R, rituximab; R-Len, rituxumab + lenalidomide; Taz, tazemetostat; Tisa-cel, tisagenlecleucel; US, United States.

Table 2. Cost-effectiveness results in base case analysis: Costs, life-years, quality-adjusted life-years, and ICERs

Comparison of Mosunetuzumab vs.	Intervention (Mosunetuzumab)			Comparator			Incremental			ICER
	Costs	LYs	QALYs	Costs	LYs	QALYs	Costs	LYs	QALYs
Taz	$290,097	13.96	11.06	$394,683	4.25	3.35	-$104,586	9.71	7.72	Mosunetuzumab Dominant
Axi-cel	$293,659	11.06	8.63	$579,112	7.89	6.46	-$285,453	3.17	2.18	Mosunetuzumab Dominant
Tisa-cel	$289,213	9.18	7.18	$552,338	8.21	6.61	-$263,125	0.97	0.57	Mosunetuzumab Dominant
RW	$290,794	10.26	8.09	$210,510	5.45	4.35	$80,284	4.81	3.75	$21,434
O-Benda	$290,925	8.88	6.95	$231,769	6.77	5.57	$59,156	2.10	1.38	$42,731
R-Benda	$289,584	10.70	8.54	$163,774	8.80	6.94	$125,810	1.90	1.60	$78,607
R-Len	$283,028	12.76	10.33	$336,467	14.04	11.09	-$53,439	−1.28	−0.76	$70,557^a

^aMosunetuzumab had lower cost and lower QALYs compared to R-Len in the base case (i.e., R-Len was cost-effective vs. mosunetuzumab at a $150,000 per QALY threshold)

Abbreviations: Axi-cel, axicabtagene ciloleucel; ICER, incremental cost-effectiveness ratio; LYs, life years; O-Benda, obinutuzumab + bendamustine; QALYs, quality-adjusted life years; R-Benda, rituximab + bendamustine; R-Len, rituximab + lenalidomide; RW, real-world cohort; Taz, tazemetostat; Tisa-cel, tisagenlecleucel; vs., versus.

Figure 1. Model structure

Note: Patients' distribution among health states over time were estimated using the fitted PFS and OS curves. At each model cycle, the proportion of patients remaining alive and discontinuing treatment were estimated by means of TTOT data from the GO29871 study for mosunetuzumab whereas for all other treatments, the TTOT is set equal to the selected parametric distribution for PFS, capped at the treatment-specific maximum number of cycles ascertained from US PIs.

Abbreviations: OS, overall survival; PIs, package inserts; PFS, progression-free survival; PPS, post-progression disease; TTOT, time-to-off treatment.

Figure 2. Net monetary benefit associated with mosun (WTP per QALY of $150,000).

Abbreviations: Axi-cel, axicabtagene ciloleucel; Mosun, mosunetuzumab; O-Benda, obinutuzumab + bendamustine; QALY, quality-adjusted life year; R-Benda, rituximab + bendamustine; R-Len, rituximab + lenalidomide; RW, real-world cohort; Taz, tazemetostat; Tisa-cel, tisagenlecleucel; vs., versus; WTP, willingness-to-pay.

Figure 3a. CEAC of mosun vs. R-Len using alternative OS parametric curve (Log-logistic).

Abbreviations: Mosun, mosunetuzumab; R-Len, rituximab + lenalidomide; QALY, quality-adjusted life year; WTP, willingness-to-pay.

Figure 3b. CEAC of mosun vs. R-Len using base case OS parametric curve (Exponential).

Abbreviations: Mosun, mosunetuzumab; R-Len, rituximab + lenalidomide; QALY, quality-adjusted life year; WTP, willingness-to-pay.