Introduction:

Patients with BCG-unresponsive carcinoma in situ (CIS) are treated with radical cystectomy (RCx) or salvage intravesical chemotherapy (SIC). In January 2020, the FDA approved pembrolizumab for BCG-unresponsive CIS +/- papillary tumors on the basis of KEYNOTE-057, a single-arm trial that reported a 3-month response rate of 38% and a 2-year recurrence-free-survival of ~15%. Given the costs (over $100,000 per year) and toxicities of pembrolizumab (63% risk of adverse events), it remains unclear whether its benefits are sufficient to warrant widespread use for BCG-unresponsive disease since non-responders will go on to RCx or SIC. To that end, we conducted a cost-effectiveness analysis comparing pembrolizumab with RCx and SIC (using gemcitabine-docetaxel as the prototypical regimen) for patients with BCG-unresponsive CIS +/- papillary features. 

Methods:

We used a decision-analytic Markov model to compare pembrolizumab, SIC (with gemcitabine-docetaxel), and RCx for patients with BCG-unresponsive CIS who are RCx candidates (index patient 1) or are unwilling/unable to undergo RCx (index patient 2). Each treatment option was a Markov node containing distinct variations of the following health states: surveillance, recurrence, progression to MIBC, progression to metastasis, treatment toxicity, and death. Incremental Cost-Effectiveness Ratios (ICERs) were compared using a willingness-to-pay threshold of $100,000/Quality-adjusted life year (QALY). The model used a US Medicare perspective with a 5-year time horizon for the base case, but varied the time horizon from 1 year to lifetime horizon as a sensitivity analysis. As many agents did not have outcomes reported beyond 2 years, we extrapolated probabilities using standard modeling approaches and then varied the extrapolated probabilities in sensitivity analyses over biologically plausible ranges. One-way and probabilistic sensitivity analyses were performed for all model parameters. 

Results:

Summary cost-effectiveness measures are presented in Table 1. For index patient 1, pembrolizumab was not cost-effective relative to RCx(ICER $1,403,008/QALY) or SIC(ICER $2,011,923/QALY). One-way sensitivity analysis revealed that pembrolizumab only became cost-effective relative to RCx with a >93% price reduction. Relative to RCx, SIC was cost-effective for time horizons <5 years and nearly cost-effective at 5 years(ICER $118,324/QALY). One-way sensitivity analysis revealed that SIC became cost-effective relative to RCx if its risk of recurrence or metastasis at 2 years was less than 55% or 5.9%, respectively. For index patient 2, pembrolizumab required >90% price reduction to be cost-effective(ICER $1,073,240/QALY). Probabilistic sensitivity analyses with Monte-Carlo micro-simulations revealed that pembrolizumab was unlikely to be cost-effective even at high willingness-to-pay thresholds (Figure 1). Further sensitivity analyses revealed that no two-way combination of extrapolated values resulted in pembrolizumab being favored over RCx or SIC for either index patient.

Conclusion:

Based on decision-analytic Markov modeling of treatment options for patients with BCG-unresponsive CIS, pembrolizumab was unlikely to be cost-effective without a >90% price reduction, regardless of a patient’s eligibility for, or willingness to undergo, RCx. Both RCx and SIC were more cost-effective than pembrolizumab. While RCx was most likely to be cost-effective compared with SIC, this was dependent on a time horizon of at least 5 years and several assumptions on SIC efficacy. Further studies may validate the cost-effectiveness of gemcitabine-docetaxel if the recurrence and metastasis thresholds in our model are met. Overall, our model supports the preferential use of RCx and SIC over pembrolizumab for BCG-unresponsive CIS.              

Funding: Vidit Sharma is funded by the Veteran’s Administration Health Services Research and Development Fellowship