Yamamoto Keiko, Takeda Kentaro, Maeda Hideki
Regulatory Science, Graduate School of Pharmaceutical Sciences, Meiji Pharmaceutical University, Tokyo, Japan.
Genmab K.K., Tokyo, Japan.
Clin Transl Sci. 2025 Aug;18(8):e70313. doi: 10.1111/cts.70313.
The first global approval of a microsatellite instability-high solid tumors was a landmark in oncology, paving the way for targeted therapies approved globally. However, guidance remains limited regarding the biological and methodological conditions under which such designs, statistical borrowing, are most effective. We retrospectively evaluated the feasibility of tumor-agnostic development using Bayesian modeling based on the objective response rate (ORR) as the primary endpoint, focusing on single-agent molecular targeted therapy (MTT) in Japan. Using Pharmaceuticals and Medical Devices Agency (PMDA) approval documents, we identified MTTs approved for ≥ 3 cancer types in Japan between 2001 and 2023. We analyzed whether their ORR in treatment lines without standard therapy exceeded thresholds using the beta-binomial model (BBM) and hierarchical Bayesian model (HBM). Among 97 approved MTTs, 57 were for solid tumors or sarcomas, and 14 for ≥ 3 indications. Poly (ADP-ribose) polymerase (PARP) inhibitors and human epidermal growth factor receptor 2 antibody-drug conjugate (HER2 ADC) consistently exceeded thresholds in both models. In contrast, mechanistic target of rapamycin (mTOR) and vascular endothelial growth factor (VEGF) inhibitors showed generally consistent results across both models, although some exceeded the threshold while others did not, indicating considerable variability. This study evaluates single-agent therapies using ORR as the primary endpoint; therefore, the findings may not apply to combination therapies or endpoints such as progression-free survival or overall survival. Nevertheless, integrating Bayesian models and biological understanding can clarify when statistical borrowing is appropriate, particularly in biologically similar tumor types and improve interpretability and the strategic feasibility of basket trials in tumor-agnostic development.
微卫星高度不稳定实体瘤的首个全球批准是肿瘤学领域的一个里程碑，为全球批准的靶向治疗铺平了道路。然而，关于此类设计（统计借用）最有效的生物学和方法学条件的指导仍然有限。我们回顾性评估了以客观缓解率（ORR）作为主要终点，使用贝叶斯模型进行肿瘤agnostic开发的可行性，重点关注日本的单药分子靶向治疗（MTT）。利用日本药品和医疗器械管理局（PMDA）的批准文件，我们确定了2001年至2023年期间在日本被批准用于≥3种癌症类型的MTT。我们使用贝塔二项式模型（BBM）和分层贝叶斯模型（HBM）分析了它们在无标准治疗的治疗线中的ORR是否超过阈值。在97种获批的MTT中，57种用于实体瘤或肉瘤，14种用于≥3种适应症。聚（ADP-核糖）聚合酶（PARP）抑制剂和人表皮生长因子受体2抗体药物偶联物（HER2 ADC）在两个模型中均持续超过阈值。相比之下，雷帕霉素机制靶点（mTOR）和血管内皮生长因子（VEGF）抑制剂在两个模型中的结果总体一致，尽管有些超过阈值而有些未超过，表明存在相当大的变异性。本研究以ORR作为主要终点评估单药治疗；因此，研究结果可能不适用于联合治疗或无进展生存期或总生存期等终点。尽管如此，整合贝叶斯模型和生物学理解可以阐明何时进行统计借用是合适的，特别是在生物学上相似的肿瘤类型中，并提高篮子试验在肿瘤agnostic开发中的可解释性和战略可行性。
