CDER researchers and collaborators at NCTR are developing pharmacometric models that can support software tools that can be used by clinicians to guide drug dosing for pregnant women.

The Scientific Challenge

Pregnant women are a special and vulnerable population with respect to drug therapy. They experience a multitude of time-varying physiological and metabolic changes, many of which are mediated by pregnancy-related hormones. These fluctuations can have a direct and dynamic effect on drug exposures. Therapeutic drug use in this population is increasingly common; however, due to legal, ethical, and practical concerns, pregnant women have historically been excluded from clinical trials. Information specific to the safety, efficacy, and dosing of drugs during pregnancy is therefore scarce, and such drugs are often prescribed without adequate clinical knowledge. Failure to treat major depressive disorder (MDD) during pregnancy can lead to impaired self-care, failure to follow prenatal guidelines, suicidality, and impulsivity that can endanger the health of both mother and child. However, many expectant mothers are still hesitant to take necessary medications due to a lack of reassuring clinical safety information. Select observational pharmacokinetic studies that have been conducted in pregnant women taking antidepressant medications report there is increased drug clearance as pregnancy advances. However, these studies do not provide a quantitative measure to aid in dose adjustment. Quantitative predictive modeling tools, such as physiologically-based pharmacokinetic (PBPK) modeling, can help ensure that the treatment is maximizing maternal therapeutic benefit and minimizing fetal risk at any given gestational age.

Quantitative Modeling to Predict Optimal Sertraline Dosing During Gestation

To predict exposures and dosing to the antidepressant drug sertraline during pregnancy, CDER’s Office of New Drugs (OND) worked with FDA’s National Center for Toxicological Research (NCTR) research scientists to develop an appropriate time-varying PBPK model. PBPK modeling is a tool that mathematically integrates physiological, physicochemical, and drug-dependent information to predict a drug’s pharmacokinetics. PBPK models define tissue and organ compartments as fixed volumes that are connected through a blood circulation loop. A set of differential equations is established to describe drug concentrations within each compartment. Examples of drug-related parameters that are inputs for the PBPK model include tissue-plasma partition coefficients (the ratio of drug concentration in the tissue to the drug concentration in the plasma at steady state), blood-to-plasma ratio, fraction of unbound drug in the plasma, rate of clearance, and rate of absorption. Examples of physiological parameters include blood flow rate, cardiac output, body weight, and plasma volume.

The workflow for model development for sertraline in pregnancy is depicted in Figure 1. Initially, published studies were mined to understand the relevant processes involved in the absorption, distribution, metabolism, and excretion of sertraline in nonpregnant and pregnant women. Physicochemical properties of sertraline, physiological changes in pregnancy, and sertraline pharmacokinetic studies were collected. A minimal nonpregnancy PBPK model structure was developed with five compartments corresponding to the liver, gut, plasma, richly perfused tissues, and slowly perfused tissues. Drug distribution into a compartment can either be limited by perfusion (blood flow) or by the cell membrane (permeability). A perfusion-limited model was chosen as it best describes small molecules such as sertraline. Tissue-plasma partition coefficients were estimated using the tissue composition-based technique described by Poulin and Haddad for highly lipophilic compounds and were adjusted to fit the observed plasma concentrations of the calibration dataset derived from nonpregnant women. Quantitative estimates for clearance of sertraline through metabolism by individual liver cytochrome P450 enzymes were derived from in vitro studies using human recombinant enzymes. The in vitro clearances were scaled to in vivo hepatic clearance using the in-vitro-in-vivo extrapolation (IVIVE) method. Sensitivity analysis was performed to identify parameters with influence on sertraline exposure. The sensitive parameters were assigned either known distributions or a coefficient of variation of 20-30% for unknown distributions. Monte Carlo Simulations were used to estimate the effects of uncertainty in parameter values and variability in input parameters on model simulations. The nonpregnancy model was calibrated using observed plasma concentrations of sertraline from a pharmacokinetic study in nonpregnant females (N=11) taking sertraline daily for 30 days. Following satisfactory predictions in nonpregnant women, the model was converted to a minimal pregnancy model with the addition of a lumped placental-fetal compartment. Gestational-dependent physiological changes in the form of polynomial equations were adapted from the published literature and incorporated in the model for various parameters such as cardiac output, plasma volume, body weight, and pregnancy-specific tissues volumes and blood flow.

The model was verified using a published pharmacokinetic study in pregnant women (N=8) with observed plasma concentrations of sertraline from the second and third trimester. Although there was considerable variability in the observed plasma concentrations, potentially due to the observational nature of the study, the PBPK pregnancy model predictions were within 2-fold of the observed concentrations. The PBPK modeling predicted that, in order to attain comparable sertraline exposures, pregnant women would have to receive greater drug doses—increased by 9%, 52%, and 117% in trimesters 1, 2, and 3, respectively—relative to the dose given to non-pregnant women (Figure 2a-c).

The Significance of a PBPK Pregnancy Model for Sertraline

The pregnancy PBPK model was converted into a prototype web-based interactive dosing tool for clinicians to potentially use at the point-of-care, using the R ‘shiny’ package (Figure 2d). This tool is an initial step towards a software application with which a clinician would be able to predict the dose adjustments needed to maintain an effective exposure as pregnancy progresses. The tool is currently in the early stages of development and is not fully validated for current clinical practice. Future versions will incorporate variability and uncertainty in model predictions as well as fetal drug exposure predictions.

Quantitative prediction of drug exposure in pregnancy using PBPK modeling promises to provide guidance about drug dosing in pregnant women in the absence of clinically observed pharmacokinetic data. The PBPK model developed for sertraline dosing in pregnancy is one of the very few studies to empirically suggest quantitative dose adjustments during the trimesters of pregnancy. The relevance to the FDA mission is, as pregnant women are rarely included in research studies, PBPK can be used as a guide by FDA reviewers to propose effective drug dosages for pregnant women in drug labeling. This research has the potential to strengthen public health guidelines for the safe and effective dosing of drugs during pregnancy, in the interests of improving neonatal outcomes and protecting maternal health.

Highlight Box:

Select Publications

George B, Lumen A, Nguyen C, et al. Application of physiologically based pharmacokinetic modeling for sertraline dosing recommendations in pregnancy. NPJ Syst Biol Appl. 2020;6(1):36. Published 2020 Nov 6. doi:10.1038/s41540-020-00157-3

Figure 1: Workflow for pregnancy physiologically based pharmacokinetic model.

Workflow for the development of the pregnancy physiologically based pharmacokinetic (PBPK) model. A nonpregnancy PBPK model representing an average human adult was established using the mean value for parameters and calibrated with pharmacokinetic data in nonpregnancy from a calibration dataset. Following satisfactory calibration, population prediction was achieved by performing sensitivity analysis and Monte Carlo simulations. The population nonpregnancy model was extended to pregnancy by incorporating physiological changes in pregnancy. The pregnancy model was used to predict exposure to sertraline during the second and third trimester of pregnancy and compared to a verification dataset. Following verification, an interactive pregnancy dosing tool was created using ‘Shiny.’

Figure 2. Predicted change in steady-state sertraline plasma concentrations with gestational age and interactive PBPK dosing tool for sertraline.

a–c. Predicted sertraline plasma concentrations in pregnant women (blue) and nonpregnant women (red) according to gestational age (GA) is shown in graphs [GA 7 (a), GA 20 (b), GA 34 (c)]. Lines represent the mean concentration while the colored areas represent the 95% prediction interval (2.5th–97.5th percentile range of a virtual population [N = 1000]). d. A screenshot for the web-based interactive PBPK dosing tool. Users can adjust various parameters including gestational age, body weight, dose and number of doses. Please note that the current version of the tool is a prototype and includes mean plasma concentration versus time profiles for illustrative purposes and does not include estimates of computed population variabilities.