Phase I Dose-Escalation Trial — 3+3 Design & MTD

Trial Controls

Hidden Ground Truth

Toxicity midpoint dose (DLT₅₀) 55 mg

Curve steepness 3.0

Reveal true toxicity curve

Trial Status

Ready — dose the first cohort to begin.

Dose level

1 / 7 · 10 mg

Patients treated

DLTs at level

0 / 0

True P(DLT) here

—

Cohort Log

How it works

A Phase I oncology trial's first job is not efficacy — it's finding the highest dose that is safe enough to study further, the Maximum Tolerated Dose (MTD). The classic rule-based design for this is 3+3: doses climb a fixed ladder, and a small cohort's toxicity outcome decides the next move.

Treat 3 patients at dose level i, count Dose-Limiting Toxicities (DLTs):
  0/3 DLT  → escalate to level i+1
  1/3 DLT  → expand cohort to 6 at level i
              ≤1/6 DLT → escalate to level i+1
              ≥2/6 DLT → STOP, MTD = level i-1
  ≥2/3 DLT → STOP, MTD = level i-1

In this simulator each patient's true DLT risk is drawn from a hidden logistic dose-toxicity curve — the biology a real trial can never see in advance:

P(DLT | dose) = 1 / (1 + exp(-k · (ln(dose) - ln(DLT50))))

where DLT₅₀ is the dose at which 50% of patients would suffer a dose-limiting toxicity, and k is the slope (a steeper curve means safety changes abruptly between adjacent dose levels). The 3+3 rule only ever sees Bernoulli coin-flips sampled from this curve — never the curve itself — exactly like a real trial only sees observed toxicities, not the underlying pharmacology.

Dose Next Cohort — treats the next 3 simulated patients at the current rule-determined dose.
Auto-Run Trial — steps through cohorts automatically until the algorithm halts.
DLT₅₀ / steepness sliders — reshape the hidden ground truth and start a fresh trial against it.
Reveal true toxicity curve — overlays the real P(DLT) curve so you can see how well 3+3's small samples estimated it.

Real-world relevance: 3+3 is still the most widely used Phase I design in oncology despite treating as few as 6 patients per dose level — its appeal is procedural simplicity, its known weakness is that a small, noisy cohort can mis-estimate the true curve, which is exactly what the reveal toggle lets you inspect.

Queued / untreated No DLT Dose-limiting toxicity

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.

Phase I Dose-Escalation Trial — 3+3 Design & Maximum Tolerated Dose