Cost-Effectiveness of Immune Checkpoint Inhibitor Therapy plus BCG for High-Risk Non-muscle Invasive Bladder Cancer: Analyses of CREST, POTOMAC, and ALBAN

Abstract

Purpose:

Three recent trials evaluated immune checkpoint inhibitors (ICI)+BCG for treatment of high-risk non-muscle invasive bladder cancer (HR-NMIBC). Two of these trials (CREST, POTOMAC) demonstrated that ICI+BCG improved event-free survival compared with BCG alone but resulted in higher rates of treatment-related adverse events. Herein, we evaluated the cost-effectiveness of ICI+BCG compared with BCG alone. We then created a publicly available cost-effectiveness calculator to facilitate future NMIBC drug value comparisons.

Materials and Methods:

We used a Markov model to compare sansalimab+BCG to induction and maintenance BCG alone for BCG-naïve high-risk NMIBC. Efficacy and toxicity probabilities were extracted from the CREST trial. One-way and probabilistic sensitivity analyses were performed. Incremental cost-effectiveness ratios (ICERs) were compared using a willingness-to-pay threshold of $100,000/quality-adjusted life year (QALY). Analyses were repeated using POTOMAC and ALBAN data.

Results:

From a U.S. Medicare payer’s perspective, the combination of sasanlimab+BCG resulted in 0.03 additional QALYs (6.12 vs 6.09) at an additional cost of $145,940 relative to BCG alone. Combination therapy was found not to be cost-effective over a lifetime horizon (ICER = $6,316,217/QALY). On one-way sensitivity analysis, the combination of sasanlimab+BCG became cost-effective only if the cost of sasanlimab was reduced by > 94% (to $1399/treatment). Similar findings were found from a U.K. perspective and with data from POTOMAC/ALBAN.

Conclusions:

ICI+BCG is not cost-effective as a combination therapy relative to BCG alone. Further efforts are needed to improve the efficacy/toxicity profile of novel therapies, while continued scrutiny of the health system cost implications of new agents remains warranted.

Introduction

Bladder cancer is the ninth most common cancer worldwide, with an estimated 614,298 new cases and 220,596 related deaths annually.¹ Approximately 75% of new diagnoses represent non-muscle-invasive bladder cancer (NMIBC), half of which are classified as high-risk.² Standard of care management for high-risk NMIBC (HR-NMIBC) remains transurethral resection of the bladder tumor (TURBT) followed by induction plus 3 years of maintenance intravesical Bacillus Calmette-Guérin (BCG).^{3, 4} While BCG is initially highly effective for most patients, approximately 15–40% of treated patients will experience a recurrence within five years.^{5, 6} Risks of disease progression and mortality increase with greater lines of therapy,⁷ emphasizing the value of improving cancer control outcomes in the first-line treatment setting. Moreover, the associated intensive surveillance and the high likelihood of multiple treatments over a patient’s disease course have made bladder cancer one of the costliest malignancies to treat over a patient’s lifetime.^{8, 9} The adoption of novel therapies—which currently cost between $200,000 to $600,000 per year in the U.S.—is likely to drive these costs even higher.¹⁰

Immune checkpoint inhibitor (ICI) therapy is now approved for treatment of BCG-unresponsive NMIBC¹¹, and efforts have been directed to assess ICI utility earlier in the disease course.¹² Specifically, the recent CREST trial was the first to demonstrate the efficacy of ICI+BCG in patients with BCG-naïve HR-NMIBC.¹³ In this trial, the subcutaneously administered PD-1 antibody sasanlimab, combined with BCG, was compared with BCG alone and resulted in a 7.3% improvement in three-year event-free survival (EFS). However, 29.1% of patients receiving sasanlimab experienced grade 3 or higher adverse events. Similarly, the POTOMAC trial studied durvalumab+BCG relative to BCG alone for HR-NMIBC and found a 4.9% improvement in 24-month disease-free survival with a 21% risk of grade 3 or higher events. Given the high cost of ICIs, the long-term sustainability of this therapeutic strategy for healthcare systems remains uncertain. As such, thoroughly understanding the tradeoffs among quality of life, cancer control, and cost for ICI+BCG is essential to inform patient-centered decision-making as well as to develop value-based policy and reimbursement strategies. The primary aim of this study was to assess the value of sasanlimab+BCG versus BCG alone using a Markov model based on CREST trial data. As a secondary aim, we sought to determine whether using similar agents—such as durvalumab and atezolizumab—influenced the value of ICI+BCG using data from the POTOMAC and ALBAN trials, respectively. Finally, we created a publicly available cost-effectiveness tool based on our model to help facilitate the incorporation of cost and toxicity considerations in future drug development and trial design.

Materials and Methods

Patient Population

In this IRB-exempt study, the index patient was based on CREST trial eligibility¹³: BCG- naïve, HR-NMIBC (high-grade Ta, T1 and/or CIS), with no prior administration of PD-1, PD-L1, PD-L2, or CTLA-4 inhibitors or immunostimulatory agents. The median patient age was set at 67 years, based on the age distribution of clinical trial participants. CREST trial data were used in the primary analyses, as this was the first trial to demonstrate efficacy. However, model parameters were altered to match the ALBAN and POTOMAC trials in the secondary analyses.

Markov Model Structure

A health-transition state Markov model informed by the CREST trial design was created comparing induction plus 2 years maintenance intravesical BCG with or without sasanlimab. A Markov model was chosen to simulate how a typical patient moves between various health states, such as surveillance, disease progression, and death, within each treatment arm. Re-induction BCG was allowed for patients who experienced NMIBC recurrence after the initial induction course. The CREST secondary endpoint of comparing sasanlimab+induction only BCG with BCG induction+maintenance was not included in the model, given the lack of statistical difference in EFS between these arms in the trial. A time cycle length of 3 months was chosen to replicate the surveillance schedule of CREST. Model health states are detailed in Figure 1 and include no progression/recurrence (surveillance), disease progression (metastatic, muscle-invasive bladder cancer), disease recurrence (NMIBC), mild and severe treatment toxicity, and death. The model was run to a lifetime horizon. The model was programmed, and all analyses were performed using R, version 2024.12.0+467.

Figure 1:

Markov model schematic.

EV: enfortumab vedotin, Pembro: pembrolizumab, NMIBC: non-muscle-invasive bladder cancer, BCG: Bacillus Calmette-Guérin, MIBC: muscle-invasive bladder cancer

Probabilities

Probability values for initial treatment toxicity, cancer-specific survival, and EFS (time from randomization to recurrence of high-grade disease, progression of disease, persistence of CIS, or death due to any cause, whichever occurred first) were derived from published CREST trial data.¹³ Toxicity and disease progression probabilities for subsequent treatments were obtained from the literature and are detailed in Supplemental Table 1.

Costs and Utility Values

To provide a global perspective, costs for each health state were calculated from two diverse health system perspectives: U.S. Medicare (base case) and the United Kingdom (Supplemental Tables 2 and 3). Since the cost of subcutaneous sasanlimab is currently unknown, we used the cost of subcutaneous nivolumab, which is also comparable in price with subcutaneous pembrolizumab, as a surrogate for our model.

Utility values, ranging from 0 (equivalent to death) to 1 (perfect state of health), were extracted from the literature (Supplemental Table 4). The utility value and toll (duration of time experiencing that utility) were used to calculate quality-adjusted life years (QALYs), which is a metric that accounts for both length and quality of life. We assumed that short-term toxicities occurred after first-line treatment to intentionally bias the model in favor of the sasanlimab plus BCG arm, thereby disregarding any potential lifelong adverse events from ICIs. The most common minor and major (grade 3 or higher) adverse events reported from CREST were used, and utility values were based on these specific events. Future costs and QALYs were discounted at a 3% annual rate.

Primary Outcomes

The primary outcome was the comparison of sasanlimab+BCG vs BCG alone in terms of treatment costs, effectiveness (measured in QALYs), and incremental cost-effectiveness ratio (ICER). The ICER represents the quotient between the difference in costs and the difference in QALYs between BCG and ICI+BCG (ICER=Δcosts/ΔQALYs). In other words, an ICER quantifies how much it costs to gain one additional year of perfect health (QALY). A treatment ICER below a willingness-to-pay threshold, which specifies the maximum amount of money a healthcare system is willing to pay for one additional QALY, of $100,000/QALY and £20–30,000/QALY was considered cost-effective from a U.S and U.K. payer’s perspective, respectively.

Sensitivity Analyses

All model variables were varied within plausible ranges in one-way sensitivity analyses to determine the influence of each parameter uncertainty on model outcomes. Since the cost of sasanlimab was the predominant driver of model outcomes, we performed two-way sensitivity analyses to understand how efficacy and cost may simultaneously impact cost-effectiveness. Multivariable probabilistic sensitivity analyses using 100,000 Monte Carlo simulations were used to assess the robustness of the model. A gamma distribution was used for costs. Beta distributions were used for probability and utility values.

Secondary outcomes

Two additional trials evaluating ICI+BCG among patients with BCG naïve HR-NMIBC were analyzed. The POTOMAC trial compared BCG (induction and two years maintenance) + one year of durvalumab intravenous therapy (IV) with BCG (induction and two years maintenance) alone.¹⁴ Similarly, the ALBAN trial compared BCG (induction + one year maintenance) plus one year of IV atezolizumab with BCG (induction + one year maintenance) alone.¹⁵ We ran separate models with data informed by the POTOMAC and ALBAN trials. Again, costs, overall effectiveness (QALYs), and ICERs were calculated using the same methodology employed in the evaluation of sasanlimab.

Interactive Web-Based Interface

An interactive web-based interface was created to host our BCG-naïve model (https://geratcliff.shinyapps.io/bladder_cancer_model/). This model allows for user-adjustable input parameters to improve stakeholder engagement and understanding how candidate drugs perform in the BCG-naïve space in terms of costs, QALYs, and ICER relative to induction+maintenance (2 years) BCG alone (details on model creation are provided in the supplemental material).

Results

Base Case

Over a lifetime horizon, from a U.S. Medicare payer’s perspective, treatment with sasanlimab+BCG was found to be associated with slightly higher effectiveness than BCG alone (6.12 vs. 6.09 QALYs), equivalent to roughly 11 additional days in perfect health. The EFS curves from our simulated model closely mirrored those observed in the CREST trial (Figure 2). Total costs for patients treated with sasanlimab+BCG were approximately four times higher than for patients treated with BCG alone ($192,581 vs. $46,640). As a result of the much higher cost and limited incremental effectiveness, sasanlimab was noted to be not cost-effective compared to BCG (ICER: $6,316,217/QALY, Table 1). Similarly, sasanlimab+BCG was also not cost-effective when analyzed from a U.K. payer’s perspective (ICER: £2,663,309/QALY; Supplemental Figures 1, 2, and 3).

Figure 2:

Model validation of event-free survival probability compared to CREST trial findings. Blue lines represent EFS of sasanlimab+BCG, and grey lines represent EFS of BCG alone. Bolded lines represent actual outcomes from the study Markov model, while non-bolded lines represent outcomes from the CREST trial.

BCG: Bacillus Calmette-Guérin, I+M: induction plus maintenance

Table 1:

Primary and Secondary Cost-Effectiveness Analyses Results.

	Total Cost ($)	Effectiveness (QALYs)	ICER ($/QALY)	Cost-Effective
CREST
BCG	$46,640	6.09
BCG + sasanlimab	$192,581	6.12	6,316,217	No
POTOMAC
BCG	$50,245	6.07
BCG + durvalumab	$161,443	6.10	3,391,655	No
ALBAN
BCG	$47,794	6.06
BCG + atezolizumab	$192,565	6.02	Absolutely dominated	No

BCG: Bacillus Calmette-Guérin, QALY: quality adjusted life year, ICER: incremental cost-effectiveness ratio

One-way Sensitivity Analysis

The model was predominantly sensitive to the cost of sasanlimab. In fact, sasanlimab cost was the only parameter that resulted in sasanlimab+BCG becoming cost-effective when varied at extreme values. Nevertheless, the cost of sasanlimab would have to be lowered to less than $1,399 per cycle in this scenario (an approximately 94% decrease from the base case of current subcutaneous nivolumab pricing). All other parameters had substantially less influence on model conclusions (Figure 3).

Figure 3:

Tornado diagram of net monetary benefits 1-way sensitivity to model parameters. The bars on the right correspond to non-sasanlimab cost variables and are magnified to better visualize how these variables influence model outcomes when ignoring the cost of sasanlimab.

BCG: Bacillus Calmette-Guérin, s: sasanlimab, MIBC: muscle-invasive bladder cancer, EV: enfortumab vedotin, Pembro: pembrolizumab, NMIBC: non-muscle-invasive bladder cancer, RC: radical cystectomy

Two-way Sensitivity Analysis

The two-way sensitivity analyses (Figure 4) also indicated that the per-cycle cost of sansanlimab is the primary driver of its lack of cost-effectiveness. Thresholds for the utility of sansanlimab and the probability of EFS were highly sensitive to cost. If the drug eliminated all EFS events (0% probability), the maximum allowable cost per dose would still be approximately $6,000. Similarly, if treatment with sasanlimab incurred no disutility (i.e., equivalent to perfect health), the intervention would only become cost-effective at marginally higher cost thresholds ($2,000) compared to the base-case scenario ($1,399).

Figure 4:

Two-way sensitivity analyses comparing costs of sasanlimab to event-free survival and utility (preference-based quality of life) while receiving sasanlimab.

BCG: Bacillus Calmette-Guérin, s: sasanlimab

Probabilistic Sensitivity Analysis

Using a willingness-to-pay threshold of $100,000/QALY, Monte Carlo probabilistic sensitivity analysis at a lifetime horizon found sasanlimab+BCG to be cost-effective in 0% of simulations (Figure 5). Even at a willingness-to-pay threshold of $500,000/QALY, BCG + sasanlimab was cost-effective in just 6% of simulations.

Figure 5:

Cost-effectiveness acceptability curve at a lifetime horizon.

BCG: Bacillus Calmette-Guérin, s: sasanlimab

Secondary Analyses

When the model was re-run using data from the POTOMAC trial, results were similar. That is, treatment with one year of IV durvalumab+BCG resulted in 6.10 QALYs compared with 6.07 QALYs for BCG alone. Although the cost of durvalumab+BCG was lower than that of sasanlimab+BCG at $161,443, durvalumab+BCG was still not cost-effective compared to BCG alone, with an ICER of $3,391,655/QALY (Table 1).

Since the ALBAN trial found no statistically significant difference in EFS between BCG alone and atezolizumab+BCG (no improvement in effectiveness), not surprisingly, this approach was absolutely dominated by BCG when adjusting our model to these outcomes. BCG was both more effective (6.06 vs. 6.02 QALYs) and less costly ($47,794 vs. $192,565, Table 1). Details regarding one-way, two-way, and probabilistic sensitivity analyses using POTOMAC and ALBAN trial designs and data are provided in Supplemental Figures 4, 5, 6, 7, 8, and 9.

Discussion

In a Markov model based on CREST trial data, we found that adding an ICI to BCG for HR-NMIBC was not cost-effective. This finding was confirmed in analyses of the POTOMAC and ALBAN trials and was maintained across diverse healthcare systems. While combining ICI with BCG slightly improved treatment effectiveness, the high cost of ICIs and the already strong efficacy of BCG alone limit the value of this treatment strategy.

The rationale for adding ICI to BCG for patients with treatment-naïve high-risk disease is to enhance cancer control at the time of initial therapy and thereby improve the rates of bladder preservation, and potentially, at a health system level, reduce costs by avoiding second-line treatments after BCG failure. However, with relatively modest improvements in cancer control (7.3% absolute improvement in EFS at three years) and high cost of ICIs, this strategy necessitates scrutiny. Most EFS events were driven by disease recurrence rather than progression, for which an increasing number of bladder-preserving salvage treatment options are now available. Moreover, bladder preservation does not necessarily equate to improved quality of life. Indeed, severe immune-related toxicities remain a serious concern with ICI use, as grade 3 or higher adverse events occurred in 29.1% and 34% of patients treated with sasanlimab (CREST) and durvalumab (POTOMAC), respectively.^{13, 14} In an attempt to intentionally skew our model in favor of ICI, we assumed these adverse events were transient and had limited long-term impact. However, even after excluding long-term toxicities in our analyses, the incremental effectiveness of adding ICI to BCG was modest (0.03 QALYs, or about 11 days of perfect health). Moreover, chronic immune-related toxicities are well described and can persist for years in up to half of patients treated with ICI therapy.¹⁶

Given the substantial treatment-related toxicities, it has been proposed that ICI+BCG be reserved for patients at the highest baseline risk of progression, including those with high-grade T1 disease, bulky or multifocal tumors, concomitant CIS, or variant histology.¹⁷ In a post hoc subgroup analysis of the CREST trial, patients with CIS with or without papillary tumors and those with high-grade T1 disease without CIS appeared to derive greater benefit from ICI+BCG compared with BCG alone (3-year event-free survival: CIS, 83.0% vs 71.8%; T1, 80.5% vs 71.3%).¹⁸ However, our findings suggest that even this highly selective strategy is unlikely to be cost-effective and yields limited incremental value. Notably, across a wide range of EFS assumptions between treatment arms, there was no scenario in which ICI+BCG emerged as the highest-value treatment option. Reducing total treatment duration is one potential approach to improving value. However, neither POTOMAC nor CREST demonstrated that adding ICI to induction-only BCG was superior to induction + maintenance BCG. Similarly, the ALBAN trial—which compared 1 year of BCG plus atezolizumab to BCG alone—found no statistically significant difference in EFS between treatment arms. These findings suggest that the duration of BCG may influence ICI efficacy. Supporting this hypothesis, a large prospective multicenter study showed that full-dose BCG for three years was associated with fewer recurrences compared with one year of BCG.¹⁹

From a value perspective, BCG monotherapy as initial treatment for HR-NMIBC remains challenging to surpass based on its high effectiveness and manageable costs.²⁰ The strong value of BCG needs to be considered when designing clinical trials of novel therapies in this space. Similarly, drug pricing and reimbursement decisions should incorporate cost-effectiveness to prevent unchecked healthcare spending. To facilitate this assessment, we have created a publicly available online tool that allows users to interactively assess the cost-effectiveness of any potential new treatment for HR-NMIBC compared to BCG alone using the model developed for this study. With this application, users can adjust treatment toxicity, recurrence risk, and costs to evaluate how each variable influences the theoretical drug’s overall value (Figure 6a). A user-friendly description of the findings is also provided, including an explanation of clinical applicability (Figure 6b). To further expand the generalizability of this resource, we incorporated a willingness-to-pay adjuster that helps users assess cost-effectiveness at various thresholds, which may vary based on the specific funding contributor being considered. While our base-case analyses reflect the U.S. and U.K. healthcare systems, the online tool allows modification of model inputs to aid interpretation across diverse and resource-constrained healthcare settings. Guidelines panels can also input candidate drug information and use the tool’s output to determine the value a given drug provides.

Figure 6:

Publicly available interactive cost-effectiveness tool for BCG naïve high-risk NMIBC. A: Cost-effectiveness results for a theoretical drug that has a 10% risk of minor toxicity, 5% risk of major toxicity, 13% risk of recurrence, and cost of $5,000 per cycle (from a U.S. Medicare payer’s perspective). B: Simple language explanation of cost-effectiveness results for the theoretical drug assessed in A.

BCG: Bacillus Calmette-Guérin, NMIBC: non-muscle-invasive bladder cancer

We recognize that our analyses have several inherent limitations. First, although our model was primarily based on level one evidence, several assumptions were necessary to extrapolate findings beyond the trial follow-up period. As such, we assumed that recurrence and progression did not occur after three years of follow-up in either arm, which may under-represent the potential improved durability of response of ICI compared with BCG alone. To directly address this uncertainty, one-way sensitivity analyses were performed in which the probability of late recurrence was varied across a wide range of plausible values. No threshold was identified at which the model conclusions were materially altered, suggesting that even as longer-term trial data mature, the cost-effectiveness findings are unlikely to change substantially. Next, due to the limited comparative data for treatment options in BCG-unresponsive NMIBC, we assumed all available salvage agents were equally likely to be used and efficacious. Additionally, we assumed bladder-sparing therapy was the first approach undertaken when patients developed BCG-unresponsive disease, with cystectomy reserved for second recurrence or progression, which we acknowledge may not reflect real-world clinical practice. Nevertheless, these assumptions were consistently applied to both arms, limiting their impact on introducing bias. Owing to limitations in available data, we were unable to conduct the analysis from a societal perspective, which would incorporate patient out-of-pocket expenses and indirect costs. Inclusion of these elements would be particularly valuable for understanding patient-level tradeoffs between treatment strategies, and further research is needed to better characterize these outcomes. Lastly, at the time of our analysis here, drug cost data for sasanlimab are unavailable. Thus, we assumed pricing would be similar to subcutaneous nivolumab, given their comparable mechanisms and administration routes. To address this uncertainty, we tested a wide range of sasanlimab costs in sensitivity analyses.

Conclusions

ICI+BCG for treatment of HR-NMIBC is not cost-effective relative to BCG alone. The overall improvement in effectiveness is marginal, given the incidence of treatment toxicity and already high effectiveness of BCG alone. These findings, together with the high costs of ICI agents, merit consideration for treatment selection as novel therapeutic regimens are reported. To help address the exponential rise in bladder cancer treatment costs, we created a publicly available cost-effectiveness tool based on our findings to facilitate the incorporation of value assessments in clinical trial development as well as healthcare policy and reimbursement decisions.

Data Access and Responsibility:

Daniel Joyce had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis

Data Sharing:

Data are available for bona fide researchers who request it from the authors