Adaptive Phase I Clinical Trial — Adaptive Dose-Escalation Design
Adaptive Phase I Clinical Trial Design · Also known as: adaptive dose-escalation trial, adaptive dose-finding study, model-based adaptive Phase I design
An adaptive Phase I clinical trial is a first-in-human or early-phase dose-finding study that continuously updates the recommended dose after each patient cohort using a prespecified statistical model, rather than following a fixed rule. The goal is to identify the maximum tolerated dose (MTD) or the recommended Phase II dose (RP2D) efficiently while minimising exposure of participants to sub-therapeutic or toxic doses. Adaptive designs — most notably the Continual Reassessment Method (CRM) — replace or augment traditional rule-based designs such as the 3+3 schema.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use an adaptive Phase I design when the primary goal is to identify the MTD or RP2D for a new drug, combination, or biologic in a small patient population where efficient dose-finding is critical. It is particularly valuable in oncology, where patient populations are limited and both under-dosing (loss of efficacy) and over-dosing (life-threatening toxicity) carry serious consequences. Adaptive designs outperform the 3+3 rule when the true MTD lies near the upper dose levels, when the dose-toxicity curve is steep, or when early dropout is unlikely to disrupt cohort-based assessment. Do not use an adaptive Phase I design when DLT assessment windows are very long and accrual is fast (late-onset toxicities require modified designs such as TITE-CRM); when institutional or regulatory capacity to implement real-time model updates is lacking; or when the number of planned dose levels is two or fewer, making model-based escalation no more informative than a simple rule.
Strengths & limitations
- Reaches the MTD with fewer total patients than the 3+3 design, reducing trial duration and cost.
- Exposes fewer patients to doses far from the target — both sub-therapeutic low doses and excessively toxic high doses.
- Explicitly incorporates all prior patient data into each dose decision rather than discarding historical cohort information.
- Transparent operating characteristics (DLT rate at MTD, percent correct selection) can be evaluated by simulation before trial start.
- Flexible to incorporate dose-toxicity-efficacy models when both endpoints are available (e.g., EffTox design).
- Requires real-time data collection and model fitting infrastructure; logistically more demanding than rule-based designs.
- Model mis-specification — if the assumed dose-toxicity shape is far from the true curve — can lead to poor MTD selection, particularly in small trials.
- Regulatory acceptance varies by agency and indication; some review bodies still prefer or require a traditional design as a reference.
- Late-onset toxicities occurring after the DLT window can bias dose decisions if not handled by a time-to-event extension (TITE-CRM).
Frequently asked
Is the CRM the same as an adaptive Phase I trial?
The CRM is the most widely known model-based adaptive Phase I design, but adaptive Phase I trials encompass a broader class including BOIN, keyboard design, EWOC, EffTox, and two-stage designs. All share the principle of updating dose decisions using accumulated data, but they differ in model assumptions, computational demand, and safety constraints.
How does it compare to the traditional 3+3 design?
The 3+3 design applies a fixed rule regardless of accumulated data; it does not fit a model and cannot estimate DLT probabilities at untested doses. Simulation studies consistently show that model-based adaptive designs identify the true MTD more accurately and treat fewer patients at dangerous doses, especially when the true MTD is at a higher dose level.
What sample size is typical?
Adaptive Phase I oncology trials commonly enroll 20–40 patients, compared to the 3+3 design which may use only 6–18 but with lower MTD accuracy. The optimal sample size depends on the number of dose levels, the target DLT rate, and the desired probability of correct MTD selection, and should be determined by pre-trial simulation.
Do regulators accept adaptive Phase I designs?
The FDA and EMA both acknowledge adaptive designs in guidance documents on adaptive clinical trials. Acceptance depends on adequate pre-specified stopping rules, prospective simulation evidence, and a data safety monitoring committee with clear escalation/de-escalation authority. Regulatory consultation before trial start is strongly advised.
What software is available?
R packages dfcrm (CRM), BOIN, and escalation are widely used. FACTS (Bayesian adaptive trial software) and East Bayes are commercial platforms. SAS macros exist for some designs. Simulation code for operating characteristics should always be shared in the statistical analysis plan.
Sources
- O'Quigley, J., Pepe, M., & Fisher, L. (1990). Continual reassessment method: a practical design for phase 1 clinical trials in cancer. Biometrics, 46(1), 33–48. DOI: 10.2307/2531628 ↗
- Chevret, S. (Ed.). (2006). Statistical Methods for Dose-Finding Experiments. Wiley. ISBN: 978-0470861608
How to cite this page
ScholarGate. (2026, June 3). Adaptive Phase I Clinical Trial Design. ScholarGate. https://scholargate.app/en/epidemiology/adaptive-phase-i-clinical-trial
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Adaptive Clinical Trial DesignExperimental design↔ compare