StochStack

prototype 07

入组预测(Monte Carlo)

基于蒙特卡罗模拟未来入组轨迹,对比 planned / actual / forecast,并给出 delay 风险预警与缓解计划。

Planned / Actual / Forecast

EnrollmentWeek
Planned (dashed)Actual (green)Forecast P50 (violet)Forecast Band P10-P90 (cyan)

延迟风险

计划结束前达标概率: 39%

预计达标周(P50): W45

当前相对计划缺口: 90 patients behind

Risk level: high

缓解计划

  • Immediate footprint expansion with additional high-yield sites.
  • Rebalance enrollment targets away from persistently delayed regions.
  • Run weekly command-center review with country-level accountability.

蒙特卡罗计算方式

1. Select trial-specific historical enrollment basis by therapeutic area and indication.

2. Compute weighted mean and volatility from historical weekly enrollment rates.

3. For each simulation path, sample weekly increment from N(mean, std), clipped at zero.

4. Aggregate paths into percentile envelopes (P10/P50/P90) and estimate delay probability.

increment_w ~ max(0, Normal(mu_weighted, sigma_weighted))
cumulative_w = cumulative_(w-1) + increment_w

Current mu=14.32/week, sigma=4.87/week from historical basis.

历史数据依据

ONC-NSCLC-214

US/EU · 74 sites

mean 14.6/week · std 4.9

startup delay P50: 32 days

ONC-NSCLC-198

APAC · 41 sites

mean 9.8/week · std 3.7

startup delay P50: 41 days

ONC-LU-177

Global · 88 sites

mean 16.2/week · std 5.4

startup delay P50: 35 days

update log

Prototype Change Log

  1. 2026-03-01 · v0.1.0

    新增蒙特卡罗入组预测界面

    • - 新增 planned / actual / forecast 的坐标轴式入组轨迹图。
    • - 新增基于蒙特卡罗模拟的 P10-P90 区间与 P50 中位预测线。
    • - 新增延迟风险预警、缓解行动计划与可解释计算方法展示。