Hillary Husband, Ph.D.

Research Scientist I

Hillary joined MetrumRG in 2021 after earning her MS in Mathematics and PhD in Engineering from Louisiana Tech University. Her thesis work centered on utilizing physiologically based pharmacokinetic modeling for therapeutic drug monitoring. She has research experience in PK/PD modeling, survival analysis, item response theory modeling, and R package development. Hillary is excited to join Metrum because of a shared enthusiasm for open science and cross-disciplinary collaboration.

Recent publications by this scientist

Reinforcement Learning for Pharmacometrics: A Proof of Concept and Future Directions.

December 6, 2024

Presented at ACoP 2024. A proof-of-concept was conducted using deep reinforcement learning to optimize vancomycin dosing based on predicted PK profiles. The reinforcement learner was able to produce dose recommendations that (on average) exceeded the standard-of-care recommended dose, and future opportunities for using reinforcement learning within pharmacometrics are vast with a large potential for impact through personalized dosing.

Download PDF

Data Gaps, Model Mishaps: Quantifying the Impact of Missing Pharmacometrics Data on Pharmacodynamic Projections.

December 6, 2024

Presented at ACoP 2024. In this analysis aimed to evaluate the ability to estimate robust population PD parameters and a landmark endpoint in different scenarios. In the primary workflow different levels of missing PK data were tested. Additionally, varying degrees of of individual-level and residual variability were tested. Finally, scenarios with smaller populations were tested. In all scenarios, PD parameters and endpoints could be estimated with adequate accuracy and precision.

Download PDF

Illustrating Integration and Interpretation of the Deep Compartment Model Approach using Keras and R in a Population PK Modeling Analysis

November 14, 2023

Presented at ACoP14. Deep compartment models (DCMs) are a proposed alternative to traditional nonlinear mixed effect (NLME) pharmacometrics approaches [1]. DCM uses neural networks to represent estimated pharmacokinetic parameters which can then be used in either closed-form or ordinary differential equation (ODE)-based representations of pharmacokinetic models

Download PDF