Regulatory Update

FDA Draft Sets Out QSP Evidence for First-in-Human Dose Selection

The draft explains how mechanistic models can support minimum anticipated biological effect level estimates. It emphasizes human-relevant inputs, model risk, uncertainty and the evidence needed to justify an initial clinical dose.

NAMs InsightsSource publication · 24 Jun 2026United States

Issuing Authority: Food and Drug Administration
Jurisdiction: United States
Document Covered: Quantitative Systems Pharmacology (QSP)-Based Dose Selection for Minimum Anticipated Biological Effect Level (MABEL) in First-in-Human (FIH) Trials, June 2026 draft; notice 24 June 2026.

Rows of computing equipment in a server room. — Computing equipment. (Vlsci, CC BY-SA 3.0)

A Defined Role for Mechanistic Modeling

FDA released draft guidance in June 2026 on quantitative systems pharmacology (QSP)-based dose selection for the minimum anticipated biological effect level (MABEL) in first-in-human trials. A Federal Register notice dated 24 June announced its availability. The document is a draft for consultation, with recommendations directed at the scientific support for a model-informed starting-dose decision.

MABEL estimates the dose associated with the lowest anticipated biological activity in humans. It can be particularly relevant when pharmacology itself creates a substantial safety concern, including immune activation. QSP offers a way to connect drug disposition, target engagement and downstream biological responses while bringing together experimental and prior clinical information.

Start with the Question and Human Biology

The draft asks developers to state the question of interest and context of use. A model may support a starting dose, an anticipated target-dose range or a dose-escalation scheme. Its intended weight in the decision, together with the potential consequences of an incorrect decision, determines the model-risk assessment and expectations for evaluation.

Model structure should represent the drug’s mechanism and relevant disease biology. Important components can include target expression, biomarker turnover, protein binding, metabolism, transport and the tissues in which effects occur. Parameters should come from reliable sources where possible. For human-specific biological products, binding information from human cells or tissues is especially relevant because target interactions and immune responses can differ between species.

Check the Implementation and Its Predictions

FDA recommends verification of equations, calculations, units and software implementation, with documentation sufficient to follow the model. Basic behavior should be examined, including the absence of a pharmacological response without drug exposure, target saturation at sufficiently high exposure and appropriate variability in stochastic models.

Calibration and validation serve distinct purposes in the draft. Parameter adjustment should be transparent, with the rationale and resulting agreement with observations documented. Evaluation against clinical and nonclinical data should reflect the context of use and model risk. Information from drugs sharing a mechanism or pathway can provide additional support when direct experience with the new product is limited.

Use Uncertainty in the Dose Decision

Sensitivity analyses should identify influential parameters, while uncertainty analyses should also examine alternative biological assumptions and model structures. Simulating responses around the proposed starting dose can reveal whether a small dose increase produces a gradual or steep change in activity. These analyses are intended to inform both the initial dose and subsequent escalation.

For novel targets without clinical precedent, the draft recommends the most conservative starting dose when QSP and traditional MABEL estimates disagree. It also recognizes the different benefit–risk considerations in life-threatening disease, where prolonged exposure to subtherapeutic doses can be undesirable. A higher minimal active starting dose may be justified in such circumstances through a documented assessment of safety and potential benefit.

Document and Update the Analysis

The proposed submission should explain model purpose, structure, equations, data sources, parameter choices, calibration, validation and uncertainty. Developers should show how the analysis supports the proposed dose and any safety margin, including the effect of important assumptions. The draft encourages early discussion through pre-IND or model-informed drug-development interactions.

Modeling continues as clinical observations accumulate. Early pharmacokinetic and biomarker data can test the initial assumptions and guide refinement, particularly when observations differ from predictions. The draft therefore treats first-in-human dose selection as the start of a learning process in which the model and the clinical evidence are assessed together.