Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Pharmaceutical Sciences

Research Advisor

Bernd Meibohm, Ph.D.


Michael Lloyd Christensen, Pharm.D. John Carl Panetta, Ph.D. Fridtjof Thomas, Ph.D. Charles Ryan Yates, Pharm.D., Ph.D.


premature infants, LC-MS/MS, modeling and simulation, population pharmacokinetics, sample size estimation, dosing optimization


Premature infants (gestational age less than 37 weeks) are considered a vulnerable patient population due to their immaturity at birth. Currently, off-label prescribing is common in younger pediatric populations, especially in premature neonates and infants, which is a primary group receiving intensive care. Unique pharmacokinetic (PK) challenges—such as limited blood volume and frequency of blood sample collections, rapid growth and continuous developmental changes, complexity of pediatric studies as well as scientific, practical, and ethical concerns— lead to the current lack of PK information and empirical dosing in premature neonates and infants. In this research, several approaches were investigated to overcome these PK challenges. We first developed and validated an accurate and sensitive LC-MS/MS method that can simultaneously quantitate multiple drugs frequently used in pediatric pharmacotherapy using a small volume of plasma. Additionally, a modeling and simulation (M&S) approach was explored in the theophylline population pharmacokinetic (PopPK) study in order to get an appropriate study design with the optimized sample size. Finally, PopPK of caffeine was investigated in premature infants using clinical data. Optimized dosing regimens were developed based on the PopPK model and dose-finding simulation study.

Due to the limitation in sample volume, an assay that can simultaneously determine multiple drugs allows for gaining maximal information from PK studies while minimizing the burden of blood collection in pediatric patients. Acetaminophen, caffeine, phenytoin, ranitidine, and theophylline are widely used in the pharmacotherapy of premature and term neonates, but only limited information is currently available on the PK of these medications in premature neonates. An accurate, sensitive and reliable LC-MS/MS assay was developed and validated using 50 µL human plasma specimens to simultaneously quantitate these five drugs with the mean accuracy ranging from 87.5 to 115.0%. The intra-day and inter-day precisions ranges from 2.8% to 11.8%, 4.5% to 13.5% respectively. This assay quantifies a range of 12.2 to 25,000 ng/mL for acetaminophen, phenytoin, and ranitidine, a range of 24.4 to 25,000 ng/mL for theophylline, and a range of 48.8 to 25,000 ng/mL for caffeine. These ranges cover each drug’s therapeutically used concentrations in the neonatal group. No significant interference effects from hemolysis, lipemia and hyperbilirubinemia were noted when these factors existed separately or were combined. Additionally, no significant matrix effect was observed for the developed bioanalytical assay.

We then evaluated the impact of sample size on the robustness of PopPK parameter estimates in observational studies in premature neonates using a simulation approach with theophylline as the model drug. Simulated datasets for each sample size (9–200 subjects per study) with a mixed and unbalanced sampling design were first generated with the incorporation of changes in birth weight, body weight, and postnatal age (PNA) in premature neonates. The median PopPK parameters for theophylline estimated from the simulated datasets were generally in close agreement with those of the originating model across all tested sample sizes. While the accuracy, precision and power to parameter estimation benefit from increases in the number of subjects included in the study, an observational study designs with < 20 premature neonates and unbalanced sampling are inadequate to allow for the precise estimation of theophylline PopPK parameters. Furthermore, the results indicate that the impact of sample size on the power of the study was deeply influenced by the parameter of interest and the selected precision level. To detect all three covariate effects studied in this research with a power > 0.8, a sample size of 20, 40 and 60 subjects is required to reach the significant level of P = 0.05, P = 0.01 and P = 0.001, respectively. The application of PopPK modeling and simulation provides a useful approach to estimate the number of subjects needed to confidently detect the potential covariate effects on PK parameters under a specific sampling strategy—randomized and unbalanced blood sampling schedules, which is consistent with actual pediatric clinical settings.

Apnea of prematurity (AOP) is one of the major concerns in premature neonates. Caffeine is currently the first-line pharmacotherapy frequently used for the treatment of AOP. A PopPK model of caffeine was developed in premature neonates, and potential sources of variability of PK behavior for caffeine were also identified. A one-compartment model was chosen to describe the PK characteristics of caffeine in premature infants, covering a gestational range of 23 to 31 weeks with an age of up to 116 days. Body weight (WT), postconceptional age (PCA) and a low gestational age (GA) of < 25 weeks were found to be important predictors explaining the between-subject variability of caffeine PK in premature infants receiving caffeine treatment. The typical patient in the studied premature neonate population, i.e., a patient with WT of 1.5 kg, PCA of 32 weeks and with a GA > 25 weeks, is estimated to have a CL of 0.0164 L/hr and a V of 0.94 L. We also investigated the application of this PK knowledge to facilitate the development of optimal dosing regimens further through simulation, particularly to correlate steady state concentrations with response at the different dosing regimens for various age/body size groups using trial simulation. A dosing interval of 24 hours is shown to be successful with respect to the proposed target concentrations in all simulated groups. With the proposed dosing regimens, the predetermined target was attained and the simulated median trough plasma concentrations were between 8 and 20 mg/L throughout the treatment period. The dose-finding simulations based on the developed PopPK model may provide more benefit while allowing the clinicians to compare various dosing regimens and bridge the plasma caffeine levels with responses at different PCAs and different WTs.

In summary, different approaches were investigated in this study to overcome the unique PK challenges in the premature neonates and infants. A full model-based simulation approach was developed to determine an optimal sample size for PopPK study in premature neonates with the consideration of changes in birth weight, body weight, and PNA. In addition, a PopPK model was developed for caffeine in premature infants and optimal dosing regimens were proposed to reach the therapeutic target concentrations rapidly based on the PopPK model. Together with the developed LC-MS/MS assay, which is highly sensitive, accurate and reliable, population-based modeling and simulation are highly useful in supporting clinical PK studies in premature neonates and infants.