The application of precision dosing in the use of sertraline throughout pregnancy for poor and ultrarapid metabolizer CYP 2C19 subjects: A virtual clinical trial pharmacokinetics study

Sertraline is known to undergo changes in pharmacokinetics during pregnancy. CYP 2C19 has been implicated in the interindividual variation in clinical effect associated with sertraline activity. However, knowledge of suitable dose titrations during pregnancy and within CYP 2C19 phenotypes is lacking. A pharmacokinetic modeling virtual clinical trials approach was implemented to: (i) assess gestational changes in sertraline trough plasma concentrations for CYP 2C19 phenotypes, and (ii) identify appropriate dose titration strategies to stabilize sertraline levels within a defined therapeutic range throughout gestation. Sertraline trough plasma concentrations decreased throughout gestation, with maternal volume expansion and reduction in plasma albumin being identified as possible causative reasons. All CYP 2C19 phenotypes required a dose increase throughout gestation. For extensive metabolizer (EM) and ultrarapid metabolizer (UM) phenotypes, doses of 100–150 mg daily are required throughout gestation. For poor metabolizers (PM), 50 mg daily during trimester 1 followed by a dose of 100 mg daily in trimesters 2 and 3 are required.

The implications of such changes during gestation make dose optimization challenging, and this is confounded by the paucity of the pharmacokinetic studies for sertraline use during pregnancy. In those that have reported plasma concentrations during gestation, conflicting results indicate either an increase in trough plasma levels (Westin et al., 2017), necessitating possible dose reduction, or a decrease in plasma concentrations, requiring a possible dose increase (M.P. Freeman et al., 2008;Hostetter et al., 2000;D.K. Sit, et al 2008). The conflicting reports may, in part, be due to the complex metabolism route and longitudinal changes in the abundance of these enzyme pathways during gestation, and often small sample (patient) sizes within studies. Nevertheless, the consensus within all of these studies highlights the need for careful monitoring of depressive symptoms during the perinatal period.
Furthermore, CYP 2C19 is highly polymorphic and these genetic variabilities have been implicated in the requirement for dose adjustment in the use of sertraline and other SSRIs with phenotypes of CYP 2C19 (Bråten et al., 2020;Hicks et al., 2015). Over 30 allelic variants have been identified for CYP 2C19, with the majority of patients being carriers of CYP 2C19 *1 (extensive metabolizer [EM] trait), *2 (poor metabolizer [PM] trait), or *17 (ultrarapid metabolizer [UM] trait) alleles. Further, guidelines from the Clinical Pharmacogenetics Implementation Consortium (CPIC) (https://cpicpgx.org) detail the allele definitions and phenotypic interpretations of CYP 2C19 and their clinical relevance alongside providing recommendations for genotype-guided dosing of sertraline, namely, advocating a dose increase of at least 50% in PM but no dose adjustment for UM phenotype patients. However, conflicting reports on the impact of specific CYP 2C19 genotypes/phenotypes on sertraline have highlighted the need to investigate the impact of this further on dose adjustments (Bråten et al., 2020).
In the context of the postnatal period, SSRIs have been reported to lead to Post Natal Adaptation Syndrome (PNAS). This is, in part, due to their ability to cross the placenta, which may result in increased serotonin concentrations in the developing fetus, thus: impacting fetal respiratory, cardiovascular, and neurological development (Bérard et al., 2017;Byatt, Deligiannidis, & Freeman, 2013;Zakiyah et al., 2018).
A recent study implemented a pharmacokinetic modeling approach to explore the changes in sertraline concentrations through gestation (George et al., 2020). While they also simulated a decrease in sertraline levels, their study lacked both the use of a full-body physiological model with a dedicated gestational-age dynamic fetal model and used a limited dataset for validation purposes. Given the limited pharmacokinetic data throughout pregnancy, the predominantly reported decrease in sertraline concentrations, coupled with its complex elimination pathways, we have applied, for the first time, a full-body virtual clinical trials pharmacokinetic model to assess the dosing of sertraline throughout gestation to identity necessary dose titrations.
With a focus on the existing guidelines for the use of sertraline in CYP 2C19 phenotypes, the primary aim of this study was to: (i) evaluate the influence of gestation on plasma sertraline levels, and (ii) provide a clinically relevant dosing titration strategy for CYP 2C19 phenotype status during gestation.

| Step 1: Validation of sertraline
We utilized the Simcyp "healthy volunteer" (HV) population group for studies with baseline populations consisting of nonpregnant females.
For pregnant population groups we used the Simcyp "pregnancy" population. This population was developed previously by Simcyp researchers and includes gestation-dependent changes in physiology, cardiac output, tissue perfusion, blood volume alongside biochemistry modification (e.g., human serum albumin) and enzyme/protein expression De Sousa Mendes et al., 2015;Jogiraju et al., 2017;Lu et al., 2012). Sertraline is not available within the Simcyp Simulator; however, a previous study developed and validated a sertraline compound for use within the Simcyp Simulator (Templeton et al., 2016), with modifications made by our group to allow its use during gestation.
In order to apply this previously validated model within the  mg as a single dose with C max reported (Saletu, Grunberger, & Linzmayer, 1986); and (v) 11 male and 11 female healthy volunteers aged between 18 and 45 years old administered a 200 mg daily for 30 days, with sampling on day 30 (Ronfeld et al., 1997); The design of trials within Simcyp were matched to these clinical studies. Simcyp Simulator parameters for sertraline are detailed in the Supplementary Materials (Section 1: Table S1).

| Step 2: Validation of sertraline during pregnancy
In order to apply the developed sertraline model during pregnancy, we conducted further validation using data extracted from a retrospective analysis of therapeutic drug monitoring services in Norway (Westin et al., 2017). This study included 56 pregnant and 52 nonpregnant (female) sertraline plasma concentrations, obtained from 34 women taking an oral dose of 50 mg daily. Importantly, this study reported individualized sample data throughout gestation rather than a central tendency without variance (D.K. Sit et al., 2008), missing patient sample data throughout the study or poor sample sizes (M.P. Freeman et al., 2008). The Simcyp Pregnancy model has been utilized previously to assess changes in plasma concentration in pregnant women (Jogiraju et al., 2017;Ke et al., 2018;Olafuyi & Badhan, 2019) and this study represents its application in the context of sertraline for the first time. The Simcyp Pregnancy model changes the physiology of the mother (e.g., tissue volumes) throughout the study period, which allows the model to operate in a dynamic nature, updating the

| Step 3: Impact of CYP 2C19 polymorphism on sertraline plasma concentration during pregnancy
Sertraline plasma concentrations are known to be altered in different CYP 2C19 phenotypes (Hicks et al., 2015). In order to simulate the impact of CYP 2C19 phenotypes in pregnant women, we simulated entirely extensive metabolizer (EM), poor metabolizer (PM), and ultrarapid metabolizer (UM) populations through revision of the default phenotype distribution to ensure uniform phenotype populations. For each phenotype, CYP 2C19 enzyme abundance was also incorporated and detailed in the Supplementary Materials (Section 2).
The study design implemented a 10 � 10 trial design with a daily dose of 50 mg once-daily throughout gestation and sampling (of plasma concentration) conducted for every 5 th week and presented as the final 24 h of that period. Where appropriate, data were also presented on the final dosing day of the week during trimester 1 (T1: week 10), trimester 2 (T2: week 20), and trimester 3 (T3: week 30).
In the absence of any published data, the default value of 0 pmol/ mg was used for CYP 2C19 PM phenotypes within the Simcyp Simulator (Djebli et al., 2015;Gong et al., 2018).

| Step 4: Dose adjustment during gestation
To explore approaches to sertraline dose titration during gestation on resultant plasma concentrations, dosing was initiated at 50 mg oncedaily and increased in weekly increments by 50 mg to a maximum of 300 mg once-daily. A proposed therapeutic range was set at 10-75 ng/ ml (Bråten et al., 2020). This was based on reports from the Arbeits-

F I G U R E 1 A workflow modeling approach for sertraline
Data were reported for each phenotype studied, namely, EM, PM, and UM subjects, on the final day of each trimester and presented as the percentage of subjects possessing trough plasma concentrations outside of the therapeutic range (i.e., below 10 ng/ml and above 75 ng/ml).

| Predictive performance
To ensure appropriate predictive performance (Steps 1-2), predictions of pharmacokinetic metrics that were within 2-fold (0.5-2.0-fold) of published data was accepted as part of the "optimal" predictive performance (Edginton et al., 2006;Ginsberg et al., 2004;Parrott et al., 2011). Furthermore, predictions in Steps 1-2 were also validated using a visual predictive checking (VPC) strategy (U.S. Food and Drug Administration, 2012) when compared to reported data.
This approach compared the Simcyp Simulator predicted concentration-time profiles, which consisted of either a mean or median and the 5 th and 95 th percentiles, against the observed data. A successful validation approach was assumed when Simcyp-predicted results overlapped with the observed datasets (Almurjan et al. 2020; Olafuyi & Badhan, 2019).

| Data and statistical analysis
Retrospective (observed) clinical data were extracted from reported studies using WebPlotDigitizer v. 3.10 (http://arohatgi.info/Web-PlotDigitizer/). Tabulated (observed) clinical data were utilized as reported in studies, namely, mean and standard deviation (Steps 1 and 2). Exploratory studies (Steps 3 and 4) were reported as median and range, unless otherwise stated. Statistical analysis was conducted using a nonparametric Kruskal-Wallis test with a Dunn's multiple comparison post-hoc test. Significance was confirmed with p < 0.05.
All statistical testing was conducted using GraphPad Prism v. 8.00 for Windows (GraphPad Software, La Jolla, CA, www.graphpad.com).

| Step 1: Validation of sertraline
A previously reported sertraline model (Templeton et al., 2016) was adapted, implementing a full-PBPK model in order to appropriately model physiological changes during gestation and their impact on Vss. The model was validated against four single-dose studies, one multiple-dose study, and a dose escalation study. The resulting predicted plasma concentration-time profiles successfully predicted F I G U R E 2 Simulated sertraline plasma concentrations. (a) Single 50 mg oral doses of sertraline (X. Chen, Duan, et al., 2006, pp. 2483-2489K.M. Kim et al., 2002;Niyomnaitham et al., 2009); (b) Multiple daily 50 mg oral doses reported on day 30 (Ronfeld et al., 1997) for males (red) and females (green); (c) 100, 200, and 400 mg single doses of sertraline (Saletu et al., 1986); (d) Forest plot showing the predicted mean ± SD over the observed ratio of pharmacokinetic parameters in subjects, with the dotted and shaded area representing the 2-fold range (0.5-2) and solid black line the line of unity. For  Table S1).
The impact of pregnancy on sertraline trough (C min ) plasma concentrations for CYP 2C19 EM and UM in nonpregnant females (baseline) and throughout pregnancy following a 50-mg once-daily dose to 100 subjects per phenotype. Data represented by box-andwhisker plots with median, 5 th and 95 th percentiles detailed.  Table S3).

| Step 4: Sertraline dose optimization
In order to address changes in sertraline concentrations during gestation for CYP 2C19 phenotype subjects, we quantified the percentage of subjects with plasma concentrations outside of the therapeutic range (i.e., below 10 ng/ml and above 75 ng/ml) across a dosing range of 50-300 mg daily.
Regardless of the phenotype, the daily sertraline dose required to maintain trough concentrations within the therapeutic window was above the usual 50 mg/day throughout pregnancy. When attempting to identify an optimal dose, we ensured a balance of a low percentages of subjects outside of this window, with an optimal dose defined as where no more than 20% of subjects possessed concentrations outside of the window (Figure 6) (Supplementary Materials: Section 5 Table S4).
For EM and UM, a dose of 100-150 mg daily is suggested to be optimal throughout pregnancy. For PM, a starting dose of 25 mg once-daily resulted in >60% of subjects with trough levels below 10 ng/ml across pregnancy ( Figure 6). However, a dose of 50 mg oncedaily resulted in 24% of subjects possessing trough levels below 10 ng/ml ( Figure 6). During trimesters 2 and 3, an increase in dose to 100 mg once-daily resulted in less than 10% of the subjects demonstrating trough levels below 10 ng/ml ( Figure 6) (Supplementary Materials: Section 5 Table S4).
Doses were titrated in increments of 50 mg every 3 days over a range of 50-300 mg once-daily throughout pregnancy. Trough plasma concentrations were reported for the final dosing day of each trimester in specific EM, PM, or UM pregnancy population groups.
Percentages of subjects with plasma concentration (trough) outside of the therapeutic range (below 10 ng/ml [left panels] and above 75 ng/ml [right panels]) are reported.

| DISCUSSION
Depression is a leading cause of disability worldwide (World Health Organization, 2008), and is thought to affect more than 20% of pregnant women (Fisher et al., 2012;Gaynes et al., 2005;Vigod et al., 2016). A key challenge for healthcare professionals is the use of pharmacological interventions during pregnancy, which is often informed by balancing the expected benefits for the mother's mental health with the possible risks to the fetus. This decision is further complicated by gestational-related alterations in maternal physiology (Isoherranen & Thummel, 2013), which can impact the pharmacokinetics of drugs. Often the combined impact of these, in addition to the longitudinal nature of these alterations, makes it difficult to extrapolate their impact during clinical practice (Tracy et al., 2005).
To augment the existing empirical approaches to treatment interventions, the application of robust and well-validated pharmacokinetic models offers a unique opportunity to apply virtual clinical trials to support medicine optimization in mental health for special population groups.
In this study we applied virtual clinical trials dosing of sertraline throughout pregnancy, to identity suitable dose titration necessary to support therapeutically maintained sertraline plasma concentrations in the mother throughout pregnancy.
We adapted a previously published sertraline model (Templeton et al., 2016) Table S1) and spanning a similar range within the population studies ( Figure 2). However, a wider AUC range in the predicted-observed ratios (Figure 2d), although still within 2-fold, are thought to be a reflection of the complexity associated with the metabolism of sertraline, namely, CYPs 2C9, 2C129, 2B6, 2D6, and 3A4, and hence the associated contribution towards interindividual variability. The variance in AUC from clinical studies (measured as mainly the standard deviation) was broadly similar to those simulated within our studies ng/ml ± 10.32 ng/ml) were within 2-fold of those reported (Westin et al., 2018) and also demonstrated a similar predicted range to that reported ( Figure 3). Furthermore, we demonstrated a decrease in mean plasma concentration throughout pregnancy, with a significant decrease in GW35 (p < 0.05) compared to baseline (Figure 3). On the contrary, a 10%, 36%, and 68% increase in plasma concentration were reported by Westin et al. (2017) Table S3).
This decrease in clearance was expected to increase sertraline trough plasma concentrations, as observed by Westin et al. (2017).
On the contrary, trough plasma concentrations for EMs and UMs decreased during gestation, with the greatest significant decrease occurring in trimester 3 (Figure 6), which concurred with a range of other reports (M.P. Freeman et al., 2008;Schoretsanitis et al., 2020; D.K. Sit et al., 2008;Tracy et al., 2005;Ververs et al., 2009). This decrease has been associated with an increase in the key female hormones estradiol and progesterone throughout pregnancy, with concentrations reaching up to 100 nM and 1 μM for estradiol and progesterone, respectively, at term. These levels are significantly greater than those during menstruation (<50 nM) (Cunningham et al., 2014;Holinka, et al., 2008). Such female hormones are known to be activators for basic helix-loop-helix transcription factors (e.g., aryl hydrocarbon receptor; AhR) or nuclear hormone transcriptional regulators (constitutive androstane receptor, CAR; pregnane X receptor, PXR; estrogen receptor, ER), which contribute to the induction of a variety of CYP isoforms and enhanced drug clearances (H. Chen et al., 2009;Jeong et al., 2008). However, the metabolic breakdown of sertraline is complicated, and includes CYPs 2B6, 2C9, 2C19, 2D6, and 3A4. The contribution of each isozyme has proven difficult to determine in vivo; however, the variable up-or downregulation of CYP isozyme expression during gestation (Abduljalil & Badhan, 2020) may contribute to the disparity observed in some studies (Westin et al., 2017). For example, the approximate 2-fold decrease in 2C19 activity coupled with approximately 2-fold increase in 2B6 activity by trimester 3 may negate the overall impact of F I G U R E 6 Dose optimization of sertraline during pregnancy in CYP 2C19 phenotyped subjects 258each pathway, in preference to changes in other physiological factors such as increases in total body water. Furthermore, the concomitant decrease in albumin is likely to cause the observed increase in sertraline plasma unbound fraction, and hence increase the volume of distribution, extending the half-life and reducing sertraline plasma levels. To confirm this, a global sensitivity analysis (GSA) was implemented to examine the combined influence of albumin levels, CYPs 2C19 and 2B6 abundance on C max , AUC, clearance (Cl), and Vss  Table S6). Given that sertraline is highly protein-bound, the decrease in albumin during pregnancy would be a significant driver for reduced plasma levels and an extension of the half-life (Little & Gynecology, 1999), potentially more so that the impact of CYP isozyme gestational changes.
At present, there is a paucity of studies exploring the impact of CYP 2C19 phenotypes on sertraline levels during pregnancy. A recent dosing guideline for sertraline that considered CYP 2C19 phenotypes has been published (Hicks et al., 2015). However, it is not clear whether the proposed guidelines are relevant to pregnant women.
Given the importance of the phenotype of the subject on gestational sertraline levels, we next examined the changes in the trough levels in relation to the therapeutic range of sertraline under a standard 50-mg daily dosage of sertraline (Bråten et al., 2020). As expected, the UM phenotypes demonstrated the largest number of subjects below 10 ng/ml (Supplementary Materials: Section 5 Table S4), whereas for the PM group, this was predicted to be in the range of 24-31%.
Finally, for all phenotypes (EM, PM, and UM), dose titrations were required to daily doses that were typically in excess of the 50-mg dose throughout pregnancy. For EM and UM, a dose escalation to 100-150 mg daily is suggested to be optimal through pregnancy. For PM, a dose of 50 mg during the first trimester followed by a dose increase in trimesters 2 and 3 to 100 mg is suggested to be optimal. Furthermore, the doses suggested in this study are within the range clinically utilized and significantly below the known toxicity range in adults (>4000 mg/daily) (Lau & Horowitz, 1996). The return of maternal sertraline plasma levels would be needed postnatally, and although this is not possible to simulate within Simcyp, tapering the dose of sertraline by 50 mg per 5-7 days is recommended to avoid withdrawal syndrome (Shelton & Richard, 2001). Furthermore, although there is very little published studies reporting pharmacodynamic changes during pregnancy for sertraline, the current approaches for the studies during pregnancy focus primarily on the clinician's role in the dose titration based on empirical changes in the psychiatric state of the patient (Ornoy & Koren, 2019). In addition, although clinicians routinely monitor drug pharmacodynamics by directly measuring physiological indices of therapeutic responses, the link between (unbound) plasma levels and clinical response is not well established for sertraline (Bergink et al., 2011;Cox et al., 1987;Sachs et al., 2002). Further, any attempt to relate unbound levels to a pharmacodynamic effect would need to further consider that the resultant central effects would be governed by the blood-brain barrier, which acts as a permeability barrier to any resultant central effects on reuptake of monoamines into the presynaptic neurons.
Further work is needed to address the reductions in sertraline plasma concentration through gestation on the resultant maternal pharmacodynamic effects on mood stability, in order to fully translate the results presented in this article to clinical practice.
A key benefit of the pregnancy PBPK approach highlighted in our study is the ability to incorporate key gestational changes in the physiology of the mother; for example, the highlighted reduction in plasma albumin and increase in maternal volume, which can be coupled with a mechanistic description of the activities of a metabolizing enzyme to enable disentangling what would otherwise be clinically complicated relationships.

| CONCLUSION
Any decision to withdraw or continue with antidepressant therapy perinatally is challenging for both maternal and fetal health. A key paradigm is the balance between the benefit of continuing treatment and the risk drug-related toxicity to the developing embryo/fetus.
Confounding treatment during gestation are longitudinal maternal physiological alternations that alter the requirements for dosing. Furthermore, the susceptibility of CYP 2C19 to polymorphisms only increases the complexity in prescribing decisions.
Our results demonstrated that dose titrations are required throughout pregnancy, with UM subjects being of concern and requiring at least double the standard dose by trimester 3, to support ongoing maintenance of plasma sertraline concentrations to within the therapeutic range.
This study has highlighted a key role for the use of pharmacokinetics to allow pragmatic exploration of dosing regimens within a perinatal setting, to support the reduction in risk of treatment relapse due to inappropriate dosing.

Certara UK (Simcyp Division) granted free access to the Simcyp
Simulators through an academic license (subject to conditions). This work was supported by Kuwait University.