Designing Optimal Pharmacokinetic Sampling Strategies for Ketamine in Established Status Epilepticus: A Model-Based Simulation Study to Support the Ketamine Add-On Therapy for Established Status Epilepticus Treatment Trial

Background: Ketamine is a promising adjunct in the treatment of established status epilepticus (ESE), a disease frequently requires emergence care. Yet, a lack of understanding of the ketamine pharmacokinetics-pharmacodynamics (PK-PD) relationship in ESE patients poses challenges in its clinical use. However, designing a pharmacokinetic (PK) study in the emergency setting is challenged by the impracticality of an intensive sampling strategy. In contrast, model-based sparse sampling strategies offer a viable alternative, enabling adequate PK characterization while reducing patient burden and enhancing operating feasibility in the emergency setting. The objective of this study is to develop a sparse PK sampling strategy for the Ketamine Add-On Therapy for Established Status Epilepticus Treatment Trial (KESETT), enabling practical data collection in emergent settings while adequate for PK-PD evaluations. Methods: Four hundred virtual patients were simulated to reflect the expected demographic distributions and correlations of the ESE population. A published two-compartment PK model for ketamine was used to simulate plasma concentration-time profiles following intravenous administration of 1 or 3 mg/kg over 5 minutes. D-optimal sampling times were identified using the PopED R package, incorporating clinically relevant constraints for emergency care. Sparse sampling designs with 2 or 3 samples, as well as an intensive sampling design, were evaluated for their ability to estimate early exposure (pAUC0-1h), a metric relevant to planned exposure–response analyses. Sampling time flexibility was assessed using windows of ±10, ±30, and ±30 minutes. Performance was evaluated using stochastic simulation and estimation (SSE) across 1000 replicated datasets, assessing predictive bias, precision, and correlation with pAUC estimated from the intensive design. Results: The D-optimal 3-sample design was sampled at 10 minutes and 1- and 3-hours post-infusion, while the 2-sample design was sampled at 10 minutes and 1 hour. Both strategies demonstrated robust predictive performance across 1000 simulated datasets. The 3-sample design achieved a mean prediction error of 19.6% with a strong correlation to intensive sampling-derived pAUC estimates (r = 0.91, p-value < 0.01). The 2-sample design showed slightly higher error (24.3%) but retained high correlation (r = 0.89, p-value < 0.01), supporting its feasibility in settings where reduced sampling is essential. Conclusions: D-optimal design methodology enables the development of sparse sampling strategies suitable for emergency care trials such as KESETT. The proposed 2-sample strategy offers a practical balance between accuracy and feasibility, supporting its use in the Ketamine Add-On Therapy for Established Status Epilepticus Treatment Trial (KESETT).