Abstracts

Rationale: ESETT is a randomized, double-blind trial comparing fosphenytoin (FOS), levetiracetam (LEV) and valproic acid (VPA) in patients with established status epilepticus (ESE). An ancillary study will characterize how plasma drug concentrations relate to the likelihood of seizure cessation in children using exposure-response modeling. To do this, an individualized estimate of early drug exposure is required, but enrollment of children in an emergency setting limits the number of samples that can be collected. Hence, this study utilizes a sparse sampling approach: 1 sample collected within 20-50 min and the other within 60-120 min after the start of drug infusion. The objective of this work is to characterize the performance of this sparse sampling approach to predict concentration at 60 min (C60) and partial area under the curve from 20-120 min (pAUC) using a simulated patient population generated from literature-based models for phenytoin (PHT), LEV, and VPA. Methods: Literature-based population pharmacokinetic (PK) models were used to simulate 2 types of rich concentration-time profiles, without error (“true”) and with error, for 500 pediatric patients (8 to 75 kg) for each drug (20 mg/kg FOS PHT-equivalents, 40 mg/kg VPA, or 60 mg/kg LEV intravenously over 10 min). One timepoint and corresponding concentration with error was randomly selected from each sampling window (20-50 min and 60-120 min) for 100 randomly selected simulated patients. We then developed population PK models using the 200 concentrations from the 100 simulated patients (2/subject). The PK model used was 1-compartment for PHT and VPA and 2-compartment for LEV. The sparse-sampling model-predicted pAUC and C60 were correlated with the “true” pAUC and C60 values from the full set of simulated data.  As an alternative approach, the concentration at the randomly sampled timepoint in the first window (C1) was also compared with C60. R was used for simulations (mrgsolve), statistical analyses and graphing and NONMEM v 7.3 (Non-Linear Mixed Effects Modeling Software, Icon Ltd) for modeling. Results: Despite using mg/kg dosing in children, ~3-fold variability in predicted exposure measures was found for all 3 drugs. For the sparse sampling approach, good correlation between the predicted and “true” pAUC and C60 was observed with correlation coefficients (R) of 0.7-0.9 for all 3 drugs (p << 0.001). Using C1(10-50min) as an estimate of C60 was inferior to the sparse sampling modeling approach (R= 0.76, 0.54, and 0.43 for PHT, VPA, and LEV, respectively). Conclusions: We conclude that a sparse sampling approach can accurately predict metrics of early drug exposure. Simulations show an ~3-fold variability in early drug exposure, which aligns well with the results from the sparse sampling model. The use of sparse PK sampling will allow for exposure-response modeling which will enable further investigation of factors affecting drug response in children with ESE. This approach can be explored in other emergent conditions and/or in children when blood sample limitations exist. Funding: Sponsor: NIH NINDSSponsor Award#: 5R01NS099653-02