Improving Dose Selection in Drug Development: Efficient Modeling Approaches Based on Early Data

The synergy between PKPD and statistical analyses has become increasingly crucial for timely and informed decision making in the complex drug development process. This presentation will highlight how the integration of PKPD modeling and statistical methodologies creates a robust framework for drug development decisions, driving both scientific excellence and organizations efficiency. We will present recent examples where this collaboration has enhanced our understanding of drug behavior, optimized dose selection and study designs, and improved the quality of go/no-go decisions. We will also discuss practical strategies for fostering effective and timely communication between PKPD and statistics teams, joint analyses planning, unified reporting strategies, synchronized decision making, and stakeholder communication. 

A question of interest in clinical development is whether evidence establishing that a drug is efficacious and safe in one indication can be leveraged to accelerate the development of that drug in another indication, either by minimizing the amount of efficacy and safety data to support dose selection in the second indication, or at the extreme, by avoiding a new study in that indication. Generally speaking, this kind of extrapolation can be conducted by bridging the two indications at the level of the drug exposure, of the target, or of a downstream biomarker. We argue here that such extrapolations require assumptions, which are not dissimilar to those required for surrogacy and in pediatric development, and we use direct acyclic graphs to depict those assumptions. We discuss how those assumptions could be supported, e.g., with existing data in the second indication (at the targeted dose level or at a different dose level), with data from another drug with same mechanism of action, or with informed pharmacological principles. We conclude by presenting our experience of situations where extrapolations were successfully used, e.g., extrapolation across treatment lines in chronic myeloid leukemia, or when the assumptions could not be supported, e.g., extrapolation for skin efficacy endpoints from psoriasis to psoriatic arthritis, or safety extrapolation from psoriasis to hidradenitis suppurativa.. 

Early-phase clinical studies, including single ascending dose (SAD)
and multiple ascending dose (MAD) trials, provide essential data for
defining the dose-response relationship of investigational
compounds. Standard dose-response analyses often rely on
endpoint-based assessments, which may overlook temporal trends in
response that impact later-phase dosing decisions. To address these
complexities, we applied a longitudinal dose-response model using an
additive exponential framework to capture both the up- and down-trends
in response over time, providing a nuanced understanding of drug
activity and safety throughout early development.


The additive exponential model accommodates fluctuations in response
by modeling the increase and subsequent attenuation of drug effects
following each dose. Specifically, this approach combines exponential
functions to capture the rapid initial response and the slower decay
observed at later time points. This structure is particularly
well-suited to represent the pharmacodynamic profiles observed in SAD
and MAD studies, where the magnitude and duration of response can vary
significantly between single and repeated dosing regimens. By treating
dose and time as continuous covariates and incorporating a
random-effects component, the model allows for the simultaneous
estimation of dose-response trajectories across various dose levels
and time points, accounting for inter-individual variability and
ensuring applicability across diverse patient profiles.


The additive model, while simple to implement, provides an accurate
characterization of the temporal patterns in response that has a
reasonable mechanistic interpretation, which is especially valuable for
compounds with complex pharmacokinetic and pharmacodynamic
profiles. The results enable refined predictions of efficacy and
safety profiles, facilitating improved go/no-go decision-making, dose
selection for Phase II trials, and adaptive design strategies.