(M0930-05-31) Model-Informed Precision Dosing of Riluzole in Acute Spinal Cord Injury Using Pharmacokinetic, Pharmacodynamic, and Clinical Outcome Modeling

Purpose: Model-informed precision dosing (MIPD) offers a promising strategy for personalizing drug therapy in acute spinal cord injury (SCI), a condition marked by wide interpatient variability. This study aimed to develop an integrated pharmacokinetic/pharmacodynamic/clinical outcome (PK/PD/CO) modeling framework and apply three-dimensional (3D) response surface analysis to characterize the relationships among riluzole exposure, phosphorylated neurofilament heavy chain (pNFH) levels as a pharmacodynamic (PD) response biomarker, and motor recovery. The analysis was based on patient-level data from the Phase II/III Riluzole in Spinal Cord Injury Study (RISCIS) PK sub-study, with the goal of informing individualized riluzole dosing strategies for SCI.

Methods: In this Phase II/III RISCIS PK sub-study, riluzole was administered orally or via nasogastric tube at a 100 mg twice daily loading dose on Day 1, followed by 50 mg twice daily for 13 additional days. Plasma samples were collected at admission (Day 0), and on Days 3, 7, 10, and 14. Plasma riluzole concentrations were quantified using a validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) assay, with a lower limit of quantification (LLOQ) of 1 ng/mL. Pharmacokinetic parameters were estimated using a one-compartment, time-varying PK model developed in Certara Phoenix (v8.5.2). Serum pNFH levels, a biomarker of axonal damage, were quantified using an enzyme-linked immunosorbent assay (ELISA) with the NS170 pNFH ELISA Kit (Sigma-Aldrich). Motor scores, including total motor score (TMS), upper extremity motor score (UEMS), and lower extremity motor score (LEMS), were collected at admission, Day 3, Day 14, Day 84, and Day 180. Missing data were addressed using multiple imputation via the MICE package in R (version 4.2.3). Three-dimensional response surface plots were generated in Design Expert (v13) to explore the relationships among riluzole AUC (A-axis), baseline pNFH levels or motor scores (B-axis), and longitudinal motor recovery outcomes (response axis). Final PK/PD/clinical outcome (CO) models were used to identify optimal riluzole exposure ranges associated with maximal motor recovery and to inform personalized dosing simulations.

Results: A total of 288 three-dimensional response surface plots were generated to evaluate the relationships among riluzole exposure, baseline biomarkers or motor scores, and motor recovery outcomes. In the first series of plots, riluzole AUC_D1-DX (where x=Day 3, 7, 10 or 14) was plotted on the A-axis, admission pNFH levels were plotted on the B-axis, and six-month TMS improvement was used as the response. These plots consistently exhibited curved, vault-like surfaces, indicating an optimal AUC range associated with greater motor recovery. Lower admission pNFH levels were associated with better motor outcomes. In the second series of plots, riluzole AUC remained on the A-axis, but baseline motor scores were used on the B-axis. These plots exhibited dome-like surfaces rather than vault-like patterns; however, the identified optimal riluzole AUC ranges remained similar to those derived from the pNFH-based models. Response surfaces based on Day 3 motor scores closely mirrored the trends observed with admission pNFH levels, suggesting that early clinical assessments may serve as surrogates for biomarker-based prognosis. Across models, riluzole AUC values over the first 7 days were identified as the earliest reliable exposure markers for predicting six-month motor outcomes. Longitudinal motor improvements were most reliably predicted when calculated as changes from Day 14 to Day 180, indicating that Day 14 motor scores serve as the most stable baseline reference for evaluating treatment response. A case simulation applying the final PK/PD/clinical outcome model proposed an adaptive dosing strategy of 50 mg twice daily for 7 days followed by 100 mg twice daily for 7 days, achieving a mean 14-day AUC of 37.05 μg·h/mL across a 100 simulated virtual patient population.

Conclusion: This integrated PK/PD/CO modeling and 3D response surface analysis framework demonstrated the feasibility of optimizing riluzole therapy in SCI patients based on individualized characteristics. Early pNFH levels and Day 3 motor scores emerged as valuable predictors for long-term recovery, while Day 7 riluzole AUC provided an actionable window for dose optimization. Importantly, longitudinal motor improvements were most reliably predicted when calculated as changes from Day 14 to Day 180, minimizing the influence of acute-phase variability and reducing noise in motor function assessments. These findings lay the groundwork for implementing model-informed precision dosing strategies to improve neuroprotective treatment outcomes in acute SCI. Prospective validation and incorporation of adaptive dose adjustments based on real-time PK monitoring will be critical for future clinical translation.

Acknowledgements: This project was sponsored by Mission Connect, IDER, US Army Medical Research Acquisition Activity under Contract No. W81XWH-16-C-0031, AOSpine North America Research Network, and the Christopher Reeve Foundation. We also extend our gratitude to the patients, their families, the study coordinators, and the technicians who contributed data from the clinical trials in NACTN. Lastly, we honor the late Dr. Robert Grossman, founder of NACTN, whose visionary leadership continues to inspire advancements in spinal cord injury research.


Modeling and Simulation Workflow for Riluzole Dose Optimization


Final Response Surface Plots Depicting Riluzole PK, PD, and Clinical Outcomes in SCI with the Earliest Riluzole AUC Time Point, for Determining Optimal Riluzole AUCD1-D7


Case Study: Simulated Dosing Strategy to Achieve Target Riluzole Area Under the Curve Based on Pharmacokinetic, Pharmacodynamic, and Clinical Outcome Modeling