164: Riluzole in Acute Spinal Cord Injury Study (RISCIS) With the Bayesian Approach to Overcome Small Sample Size & Inter-subject Variability

Statement of Purpose: Spinal cord injury (SCI) is a life-threatening and severely debilitating condition that results in permanent disabilities along with chronic pain, psychological trauma and social stigma. According to the 2013 American Association of Neurological Surgeons (AANS)/Congress of Neurological Surgeons (CNS) guidelines, currently, there is no standardized treatment for SCI. One of the main reasons for this delayed drug development in SCI stems from extreme inter-subject variability, which challenges the consistent pharmacokinetic (PK)/pharmacodynamic (PD) data analysis with a conventional approach. Furthermore, having a small population of SCI patients makes small sample sizes inevitable. Additionally, enrolling patients for PK study is challenging given the critical condition of patients upon hospital admission. To overcome the small sample size and inter-subject variability, we utilized the Bayesian approach for our Phase II/III clinical trial (NCT01597518), Riluzole in Acute Spinal Cord Injury Study (RISCIS). The Bayesian estimation is a powerful tool to assess both precision and variance in PK, by using conditional probability. It weighs the known prior knowledge (‘prior’) of a PK parameter and new data (‘evidence’), to estimate a new probability (‘posterior’) for predicting its PK parameters. This updated probability then serves as the new prior probability as additional data becomes available, effectively allowing us to adjust our understanding in response to accumulating information. Our hypothesis is that the Bayesian approach improves in estimating the PK parameters within the small-sized and heterogeneous RISCIS patients.

Description of Methods & Materials: Patient data were obtained from RISCIS, a multi-center, randomized, placebo-controlled, double-blinded clinical trial. Patients in the treatment group received 100 mg of riluzole twice on Day 1 as a loading dose, then 50 mg of riluzole twice daily for the rest of 13 days as a maintenance dose. Their plasma samples were drawn at 3 hours post-dose and before the next dose on Days 3, 7, 10 and 14. Plasma samples were quantified using a validated LC-MS/MS assay with an LLOQ of 0.5 ng/ml. For PK analysis, a population pharmacokinetic model was performed using Monolix v2023 R1. A one-compartment structural model with a first-order absorption and a linear elimination was conducted for this analysis. We utilized a Maximum A Posteriori (MAP) Bayesian estimation for the Bayesian estimation methods. Our existing PK parameters (Ka, V, CL values) on Phase I data were used as the prior information. Patients’ samples in RISCIS (Phase II/III) served as evidence to yield the posterior probability.

Data & Results: A total of 13 patients’ plasma concentrations in RISCIS was used for this population PK analysis. Before using the Bayesian estimate, our basic population pharmacokinetic model shows that Ka, V and CL were 0.071(hr-1), 350.28 (L) and 87.75 (L/hr), respectively. After a MAP Bayesian estimation, Ka was unchanged, while V and CL shifted to 233.76(L) and 88.34 (L/hr), respectively.

Interpretation, Conclusion or Significance: Accounting for posterior values from the Phase I trial, V was decreased while CL was slightly increased. Overall, interindividual variability in plasma clearance and volume of distribution was improved.

Disclosures: US ARMY MEDICAL RESEARCH ACQUISITION ACTIVITY, Contract No. W81XWH-16-C-0031, IDER (Institute for Drug Education and Research, Medicine of the Future)