Basket / Umbrella / Platform trial (Master Protocol)

一行要約

Master protocol は単一試験骨格の下で複数 hypothesis を並行検証する設計枠組み。Basket = 1 drug × 多 tumor (共通 biomarker, e.g. larotrectinib NTRK basket), Umbrella = 1 tumor × 多 biomarker × 多 drug (LUNG-MAP, BATTLE), Platform = 多 drug が長期間 perpetual に開閉 (I-SPY 2, STAMPEDE)。Driver mutation 中心の精密医療時代に rare driver の efficacy を効率的に評価する gold standard で、tissue-agnostic 承認 (larotrectinib, entrectinib, dostarlimab MSI-H, pembrolizumab MSI-H/TMB-H) の基盤となった。

原理

設計パターン

設計構造代表 trial
Basket1 drug, 多 tumor (共通 marker)LIBRETTO-001 (selpercatinib RET), TRIDENT-1 (repotrectinib), VE-BASKET (vemurafenib BRAF), KEYNOTE-158 (pembrolizumab MSI-H)
Umbrella1 tumor, 多 biomarker → 多 drugLUNG-MAP, BATTLE, NCI-MATCH (一部)
Platformperpetual, multi-drug, shared controlI-SPY 2 (breast), STAMPEDE (prostate), REMAP-CAP (sepsis), ProBio (prostate)

Statistical considerations

  • Bayesian hierarchical model で sub-cohort 間 borrowing (Berry method)
  • Type-1 error 補正 (multiplicity, family-wise vs FDR)
  • Master protocol governance (DSMB, biomarker test standardization)
  • Adaptive enrollment: 効果群拡大 / 無効群早期 closure
  • Tissue-agnostic 承認には ≥3 tumor type, prespecified ORR threshold が FDA standard

主要エビデンス / 適用領域

  • Tissue-agnostic 承認: larotrectinib (NTRK), entrectinib (NTRK + ROS1), pembrolizumab (MSI-H, TMB-H), dostarlimab (dMMR), repotrectinib (NTRK)
  • NSCLC rare driver: LIBRETTO-001 (selpercatinib), GEOMETRY-mono-1 (capmatinib MET ex14), KRYSTAL-1 (adagrasib)
  • LUNG-MAP: NSCLC second-line で biomarker arm 並行
  • I-SPY 2: neoadjuvant breast platform で adaptive randomization、>20 drug graduate / drop
  • STAMPEDE: prostate cancer 多 arm long-running platform
  • REMAP-CAP: COVID-19 で oncology 外の platform trial 成功例

適用分野と限界

  • 強み: rare biomarker 効率的評価、tissue-agnostic approval pathway、shared control で N 削減、infrastructure 再利用 (perpetual platform)
  • 限界: tumor-tumor heterogeneity (basket で同じ marker でも tumor 間で response 差: BRAF V600 melanoma vs CRC)、biomarker assay standardization、regulatory landscape は country dependent、small subset で statistical precision 限定、operational complexity (multi-site, multi-sponsor)

Open Questions

  • Tumor-tumor heterogeneity within basket: 共通 marker でも tumor type で response 差の予測モデル
  • Biomarker assay standardization (NGS, IHC) の cross-platform reproducibility
  • Regulatory framework: tissue-agnostic 承認の post-marketing commitment
  • Co-clinical trial: PDX / organoid と並行 master protocol で前向き integration

関連エンティティ