Pharmacokinetic Modeling and Model-Based Hypothesis Generation for Dose Optimization of Clonidine in Neonates With Neonatal Opioid Withdrawal Syndrome

Abstract

The No-POPPY study (NCT03396588), a double-blind, randomized trial compared morphine with clonidine therapy for neonatal opioid withdrawal syndrome (NOWS) and found that the duration of treatment was similar across groups. This is significant because perinatal use of morphine has the potential for neurodevelopmental consequences. Still, the clonidine group reached symptom stabilization (Finnegan score (FS) < 8) later than the morphine group and had a more significant number of patients who required adjunct therapy. However, the mean FS was consistently lower in the clonidine group after day 6. This prompted us to use pharmacokinetic (PK) and parametric time-to-event (TTE) modeling to simulate dosage schedules that may decrease the time to stabilization and reduce the need for adjunct therapy. Population PK (popPK) analysis was conducted, and the final model was a one-compartment model with first-order absorption and elimination, incorporating allometric scaling and age effect on apparent clearance (CL/F) and apparent volume (V/F). The population estimates for CL/F and V/F were 13.6 L/h/70 kg and 416 L/70 kg, respectively, similar to the reported values. A Weibull model described the TTE data best, followed by incorporating predicted average concentrations to yield the final Weibull accelerated failure time model. Simulations of dosing strategies showed that increasing both the starting and maximum doses could potentially shorten the time to stabilization, and thus, length of treatment and hospital stay. Given the hypothesis-generating nature of this analysis, the recommended dosing regimens should be tested prospectively to evaluate their benefits.

Figure 1

Diagnostic plots of the final clonidine population pharmacokinetic model. The blue lines represent the loess regression line. In the DV vs. PRED/IPRED plots, the solid gray lines represent the line of unity. CWRES, conditional weighted residuals; DV, dependent variable (clonidine concentration) in ng/mL; IPRED, individual predicted value in ng/mL; PRED, population predicted value in ng/mL; TAD, time after the most recent dose; TIME, time after the first dose. An influential data point at TIME = 850 was removed from the CWRES vs. TIME plot as it skews the loess regression line.

Figure 2

Simulated pharmacokinetic profiles of protocol and alternative dosing regimens are listed in Table 1 . The solid lines represent the simulated median of individual predicted concentrations from 500 virtual patients.

Figure 3

Overlaying of the Weibull baseline model and nonparametric survival estimates. The curved solid line and dashed lines represent the model estimate and its confidence interval. The shaded area represents the confidence interval for the Kaplan–Meier curve.

Figure 4

Simulated proportion of patients achieving symptom stabilization over time for protocol and alternative dosing regimens in the absence of adjunct therapy. Solid lines and shaded areas represent the median and 90% confidence intervals from the simulations.