Phase I studies are the first step in testing new drugs, biologics, or delivery systems in humans. They come after laboratory and animal research and are designed to answer early safety and dosing questions.
These studies focus on how the body handles a drug (pharmacokinetics) and how the drug affects the body (pharmacodynamics). They also look at tolerability and provide a first look at whether the drug shows signs of activity.
- Measure bioavailability (how much of the drug reaches the bloodstream).
- Track how the drug and its metabolites distribute across body compartments.
- Identify side effects and estimate safe dose ranges.
- Test feasibility and safety of delivery systems (for example, vectors in gene transfer or implantable devices that release medicine over time).
Who Participates
Participants are often healthy volunteers when the expected risks are low. In areas like cancer, where drugs can be more toxic, participants are usually patients who have tried standard treatments without success.
Enrollment typically uses small cohorts so dose decisions can be adjusted as information accumulates.
How Starting Doses Are Chosen
Before the first human dose, researchers review animal and lab data to estimate a safe starting point. They apply safety margins and consider factors like the most sensitive species, exposure levels that caused no observed adverse effects, and differences in how species process the drug.
The starting dose is usually set low and increased carefully to avoid “overshooting” a tolerable level.
The Classic Rule-Based Approach: The 3+3 Design
The most widely known method is a simple, stepwise dose-escalation design, often called the “3+3.” It is common in oncology and other areas where safety margins may be narrow.
- Enroll a small cohort (often three participants) at a starting dose.
- If no participant has a dose-limiting toxicity (DLT), move to the next higher dose for a new cohort.
- If one participant has a DLT, expand the cohort at that dose to six participants.
- If two or more participants at a dose have DLTs, stop escalating; the maximally tolerated dose (MTD) is often the previous, lower dose.
In many protocols, the MTD is defined as the dose at which about one-third of participants experience DLTs, though the exact target can vary.
A Simple 3+3 Example
Imagine dose levels A, B, C, and D. Three participants receive dose A. If 0/3 have a DLT, the study escalates to dose B. At dose B, suppose 1/3 has a DLT; the cohort expands to 6. If only 1/6 has a DLT, escalation continues to dose C. If at dose C, 2/3 have DLTs, escalation stops and dose B is considered the MTD (subject to protocol rules).
This approach is conservative and easy to run, but it can be slow and may not use all the information collected as efficiently as model-based methods.
Step-Up/Step-Down and Convergence Approaches
Some rule-based designs allow both increases and decreases in dose as data emerge rather than stopping immediately when toxicity appears. These methods “hover” around the likely MTD.
- Escalate when no DLTs are observed at a dose.
- Step down to a lower dose if DLTs occur.
- Continue adjusting until the process appears to stabilize around a dose level.
After data collection, a dose–response model may be fit to estimate the dose linked to a chosen probability of toxicity (for example, 25% or 33%), depending on the clinical context.
Model-Based and Bayesian Designs
More sophisticated methods use statistical models to guide each dose decision. These designs can be more efficient and can incorporate prior knowledge and patient-specific factors, but they are also more complex to implement.
Continual Reassessment Method (CRM)
CRM is a Bayesian approach that estimates the relationship between dose and the probability of DLT, updating that estimate as data accrue.
- Start near the dose believed to be close to the MTD based on prior information.
- Treat a small cohort (often 1–3 participants), update the model with observed outcomes, and assign the next participant(s) to the dose estimated to be closest to the target toxicity rate.
- Repeat until the planned sample size is reached, then declare the estimated MTD as the dose where a hypothetical next participant would be assigned.
Enhancements and Variants
- Escalation With Overdose Control (EWOC): Adds explicit protection against assigning doses with a high probability of exceeding the target toxicity level.
- Time-to-Event CRM (TITE-CRM): Accounts for delayed toxicities so dose decisions can be made before every participant completes full follow-up.
- Multiple-Regimen Designs: Extend model-based escalation to compare two or more treatments or schedules within the same study.
- Covariate-Adjusted Models: Incorporate risk factors (for example, organ function, prior therapy) that may shift a participant’s tolerance.
A Brief CRM Illustration
Suppose the target toxicity rate is 25%. Prior information suggests that Dose 3 is near the target. The first participant is treated at Dose 3 and has no DLT. The model updates and estimates that the next best dose is Dose 4. The second participant receives Dose 4 and experiences a DLT. The model shifts the estimate downward and assigns the third participant to Dose 3. Over time, assignments may alternate between Doses 2 and 3 as data refine the curve, and the final MTD estimate may be Dose 3.
This approach can shorten the time to find a suitable dose and may expose fewer participants to subtherapeutic or overly toxic doses. It does, however, require careful monitoring and statistical support.
Managing Safety, Efficiency, and Practical Constraints
Regardless of design, studies are built to escalate cautiously while limiting unnecessary exposure to toxic doses. Safety rules are often embedded to prevent large, abrupt dose jumps.
- Small cohorts (for example, 1–3 participants) allow frequent reassessment.
- Stopping rules may halt escalation if excessive toxicity appears at any point.
- Sample sizes in oncology Phase I trials often range from about 15 to 40 participants, though this varies widely.
- Delayed or cumulative toxicities can complicate dose decisions; methods like TITE-CRM help address incomplete follow-up.
Devices and Gene Transfer: Special Considerations
When a device delivers the therapy (such as an implantable pump or a long-acting matrix), Phase I studies assess both the device and the active agent. The goal is to confirm that the device is safe and releases the drug at the intended rate.
In gene transfer studies, the vector (the delivery vehicle for genetic material) is central. Early trials evaluate vector behavior—where it goes in the body, how long it persists, and what immune responses it may trigger—alongside dose and safety.
From MTD to the Next Phase
Many Phase I trials seek a maximally tolerated dose or identify a recommended Phase II dose (RP2D). In some settings, the RP2D may be based on a “biologically effective” dose rather than strictly the MTD, especially when activity plateaus at lower exposures or when safety concerns arise.
Preliminary signs of activity (such as tumor shrinkage or biomarker changes) are recorded but are not the primary goal. These signals, together with safety and pharmacology data, inform whether the program proceeds to Phase II.
Putting It All Together
Phase I studies are carefully controlled, small-scale experiments in people that establish how a new therapy behaves in the body and how much can be given safely. Traditional rule-based designs like the 3+3 are simple and conservative, while model-based and Bayesian methods such as CRM and EWOC can be more efficient and flexible. Designs that allow step-up and step-down adjustments, account for delayed toxicities, or incorporate patient risk factors may improve decision-making but add complexity. Across all approaches, the central aim remains the same: escalate thoughtfully, protect participants, and use emerging data to pinpoint a dose that balances safety with the potential for benefit.