Integrating Population Pharmacokinetic Modeling and a Genome-Wide Association Study of Sulindac in Pregnancy: An Exploratory Study in a Small Cohort

Abstract

Sulindac is an anti-inflammatory drug (NSAID) that can be used as a second line tocolytic therapy in pregnant women facing the risk of preterm labor, particularly when first-line intravenous treatment fail to produce effective response. Although sulindac is used in clinical practice, we still don’t fully understand why it behaves differently from one person to another, especially in pregnant women, whose bodies undergo major changes that can affect how drugs are absorbed, distributed, and eliminated. This variation is also possible to be caused by genetic differences. We believe that understanding these variations is not just a scientific curiosity, it is key to finding the right dose for each patient , reducing the risk of side effects and increasing efficacy. The two primary objectives of our study were to come up with a population pharmacokinetic (popPK) model of sulindac and its two metabolites (sulfide and sulfone) in pregnant women with preterm labor and also to investigate the hypothesis of genetic variation that may alter individual PK parameters in a population genome-wide association (GWAS). In the initial phase of the study, we developed a nonlinear mixed-effects (popPK) model using Monolix software which incorporates the SAEM (Stochastic Approximation Expectation-Maximization) algorithm to ensure robust and reliable convergence. The modeling was based on a one-compartment model in which the absorption follows a first order with the conversion of sulindac to its metabolites by a parent-metabolite structure. Several (popPK) models were tested, and the most appropriate model was chosen based on diagnostic measures including goodness-of-fit plots, shrinkage, relative standard error (RSE), visual predictive check (VPC) and bootstrap analysis. Following this, the second part of our research involved extracting individual empirical Bayes estimates (EBEs) of eight important pharmacokinetic parameters [Tlag, Ka, V/F, Cl1, Kred, Kre-ox, Kox, Cl2] from the final model. These parameters were treated as quantitative traits for association analysis. Association tests between single nucleotide polymorphisms (SNPs) and PK parameters were performed using a generalized linear regression model (GLM) under an additive allelic effect, implemented in PLINK v2.0.0, where maternal age, gestational age, weight and the first 10 PCs were incorporated in the model to account for potential hidden structure. Following the completion of QC, The analysis identified several SNPs with suggestive associations to model-derived PK parameters, highlighting the potential value of this approach. Through integrating pharmacokinetics and genomics, our study holds the hope of contributing to the transformation of pharmacological research and the improvement of individualized therapies.