Utility of Bayesian Single-Arm Design in New Drug Application for Rare Cancers in Japan: A Case Study of Phase 2 Trial for Sarcoma

Abstract

Investigational drugs for rare cancers are often approved based solely on a single-arm phase II trial that primarily evaluates response rate in Japan. Such trials typically use a fixed sample size determined on the basis of the frequentist manner. However, since predicting the speed of patient enrollment is challenging because of the disease rarity, the time needed to complete the enrollment of the fixed number of patients is prolonged in some cases. A Bayesian design without fixing the sample size is useful for single-arm phase II trials of rare cancers. However, the arbitrariness of prior distribution specifications and the frequentist operating characteristics are regulatory issues. We recently started a Bayesian single-arm phase II trial of nivolumab in patients with sarcoma for new drug application in Japan and examined the statistical rationale and design consideration. In the Bayesian design, we specify the minimum and maximum numbers of enrolled patients during the enrollment period and the prior distributions of response rates. Considering these parameters, we obtain the minimum number of responders needed for the positive conclusion of the efficacy of nivolumab for each sample size. Simulation studies demonstrated that the operating characteristics of this design would be acceptable from the frequentist view. The Bayesian design provided an adaptive decision rule for efficacy conclusion for the drug without fixing the sample size. We hope our trial's success will provide a new drug development option for rare cancers in Japan.

Practical basket design for binary outcomes with control of family-wise error rate

“…For each subpopulation, one interim analysis for futility was conducted when the number of accrued patients reached 10 (i.e., ). We assumed that followed a beta distribution with parameters Beta (10, 190) and a mean of 0.05 based on the analysis reported by Hirakawa et al (2018) [ 19 ].…”

“…In simulation studies, we compared the proposed method with an independent method. For the independent method, the posterior distributions of and were calculated independently as follows: where refered the distribution function of the beta distribution, and and were set to 0.6 and 1.4, respectively, based on the analysis reported by Hirakawa et al (2018) [ 19 ].…”

Regulatory issues and the potential use of Bayesian approaches for early drug approval systems in Japan

“…The fixed threshold can be reasonably determined based on historical data because the response rate has no or minimal placebo effect. To the best of our knowledge, there are a few instances of the use of such Bayesian approaches in Japan 14,15 …”

“…To the best of our knowledge, there are a few instances of the use of such Bayesian approaches in Japan. 14,15 A cancer basket trial also evaluates the response rate of multiple cancer types for one targeted therapy based on a common single molecular marker or genetic mutation. In some cases, such a trial is pivotal for CEA for tumor-specific or tumor-agnostic indications, although they are generally considered as "signal-finding" trials for screening therapies for further testing.…”

A multicenter investigator-initiated Phase 2 trial of E7090 in patients with advanced or recurrent solid tumor with fibroblast growth factor receptor (FGFR) gene alteration: FORTUNE trial

“…And it is obviously essential to have a drug delivery system as an exit strategy. In Japan, due to the limited number of patients, some situations existed where investigational drugs for a rare population were approved by merely a single-arm Phase 2 study that mainly evaluates response rates [20]. This study described here is a novel attempt to evaluate a new drug with a tumor agnostic approach by incorporating the MANO method Trial registration: Japan Registry of Clinical Trial: jRCT2031210043 (registered April 20, 2021) ClinicalTrials.gov: NCT04962867 (registered July 15, 2021).…”