Estimating fetal exposure to the P‐gp substrates, corticosteroids, by PBPK modeling to inform prevention of neonatal respiratory distress syndrome

Abstract We have previously developed a maternal‐fetal physiologically‐based pharmacokinetic (m‐f PBPK) model to dynamically predict (and verify) fetal‐maternal exposure to drugs that passively diffuse across the placenta. Here, we extended the application of this model to dynamically predict fetal exposure to drugs which are effluxed by placental P‐glycoprotein, namely the antenatal corticosteroids (ACS; dexamethasone [DEX], and betamethasone [BET]). To do so, we estimated both the placental P‐gp mediated efflux clearance (CL) and the passive diffusion CL of the ACS. The efficacy and toxicity of the currently used maternal ACS dosing regimens to prevent neonatal respiratory distress syndrome could be improved by altering their dosing regimens. Therefore, to illustrate the utility of our m‐f PBPK model, we used it to design alternative dosing regimens of DEX and BET that could potentially improve their efficacy and reduce their toxicity. The redesigned dosing regimens are convenient to administer, maintain maternal‐fetal exposure (area under the concentration‐time curve [AUC]) or maximum plasma concentration (Cmax) or both (DEX and BET) or minimize maternal exposure while maintaining fetal drug plasma concentrations above the minimum therapeutic threshold of 1 ng/ml for 48 h (BET only; based on efficacy data in sheep). To our knowledge, this is the first study to dynamically predict fetal plasma concentrations of placental P‐gp effluxed drugs. Our approach and our m‐f PBPK model could be used in the future to predict maternal‐fetal exposure to any drug and to design alternative dosing regimens of the drug.


INTRODUCTION
Approximately 80% of pregnant women take prescription or over-the-counter medications 1 to treat a variety of maternal conditions (e.g., diabetes) or to treat the fetus (e.g., prevent respiratory distress syndrome [RDS]). To maximize maternalfetal drug efficacy and minimize toxicity, it is critical that fetal exposure to the drug be quantified. Such quantification is logistically and ethically challenging. At the most, a single blood sample from the umbilical vein (UV) and maternal vein (MP) can be obtained at the time of birth. The ratio of the plasma drug concentration in these samples (UV/MP) is not a measure of fetal exposure as it only provides a snapshot of the fetal:maternal drug concentration at a single time point. 2 If such data are available over multiple time points, covering several half-lives of the drug, the data can be "naively" pooled to arrive at a complete UV/MP concentration-time profile. From this profile can be derived K p,uu , the steady-state unbound fetal:maternal partition coefficient of the drug. When a drug passively diffuses across the placenta and provided there is minimal placental or fetal clearance of the drug, K p,uu = 1. However, many drugs administered to pregnant women are substrates (e.g., antenatal corticosteroids used to prevent RDS) of transporters that are highly abundant in the placenta (e.g., P-glycoprotein [P-gp]). When a drug is a substrate of placental efflux transporters, K p,uu < 1. While K p,uu provides an estimate of the fetal exposure to the drug at steady-state, it cannot be used to dynamically predict fetal plasma concentration time profile. To do so, one needs the actual values of the transplacental clearances. Provided there is minimal placental and fetal elimination of the drug, K p,uu is related to these transplacental clearance (CL) as follows: Where CL int,PD,placenta and CL int,Pgp,placenta are the in vivo intrinsic placental passive diffusion clearance and P-gp mediated efflux clearance, respectively.
We have previously developed and verified a maternalfetal physiologically-based pharmacokinetic (m-f PBPK) model to dynamically predict fetal:maternal exposure to drugs that passively diffuse across the placenta. 2,3 To do so, we developed an innovative approach to estimate the transplacental CL of the drugs (based on midazolam as model placental passive diffusion drug). Our m-f PBPK model incorporates all the changes in gestational-age dependent physiological parameters that are important in drug disposition. For example, this model includes gestational-age dependent changes in cardiac output, organ blood flows (including to the placenta), total body water, plasma protein concentrations (e.g., albumin), changes in hepatic CYP activity, placenta size, fetal growth, etc. 2,3 Thus, our model can predict fetal exposure to drugs at any gestational age. However, even though this model is capable of incorporating placental drug transport, it has never been used to dynamically predict fetal exposure to drugs that are transported by the placenta. Here, using antenatal corticosteroids (ACS) as an example, we extend this model to do so.
ACS, dexamethasone (DEX), and betamethasone (BET) were chosen as our model drugs because they are the most common ACS used to prevent RDS and are substrates of Pgp. These two epimers were developed in the early 1960s to treat rheumatoid arthritis; however, the efficacy (prevention of RDS) and toxicity (e.g., maternal infection, 4,5 fetal neurodevelopmental disorders, 6,7 and hypoglycemia 4 ) of ACS has not been systematically addressed. 5 To prevent RDS, the usual ACS dosing regimens administered to pregnant women of 24 to 36 gestational weeks (GWs) are intramuscular (i.m.) administration of 6 mg DEX phosphate (DEX-P) every 12 h for 48 h, or 12 mg of 1:1 BET phosphate:acetate mixture (BET-P:A) every 24 h for 48 h 8,9 (henceforth referred to as the reference dosing regimens). The efficacy of these reference regimens is modest. For example, a recent large clinical trial in low to middle income countries found that the risk of both neonatal death and RDS in preterm babies was reduced by only one-sixth when the DEX reference regimen was administered to pregnant women at imminent risk of preterm birth (1) K p,uu = CL int,PD,placenta CL int,PD,placenta + CL int,Pgp,placenta

WHAT QUESTION DID THIS STUDY ADDRESS?
Here, we extended our m-f PBPK model to dynamically predict fetal exposure to drugs which are effluxed by placental P-glycoprotein, namely the antenatal corticosteroids.

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?
To our knowledge, this is the first study to dynamically predict fetal plasma concentrations of placental transported drugs.

HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?
Our approach and our m-f PBPK model could be used in the future to predict maternalfetal exposure to any drug (including those that are transported) and to devise dosing regimens of the drug to maximize maternal-fetal drug efficacy and minimize drug toxicity.

| 1059
FETAL EXPOSURE TO CORTICOSTEROIDS BY PBPK MODELING AND SIMULATION (within 48 h). 5 Alternatively stated, despite ACS therapy, the relative risk of RDS or infant mortality within 28 days of birth remains high (relative risk of RDS of 0.81 and 0.84, respectively). Collectively, these data suggest a need to further refine DEX and BET dosing regimens to prevent RDS. ACS safety and efficacy are linked to their fetal (and maternal) drug exposure, but defining this relationship is difficult. First, as indicated above, UV/MP drug concentration at a single time point does not provide a measure of fetal exposure to ACS. Second, maternal drug exposure is not a good surrogate of fetal exposure as these drugs are substrates of P-gp 10,11 and acts to variably diminish fetal ACS exposure relative to that in the mother. [12][13][14] The extent of this diminution and the variability therein has not been well-defined, but likely changes as pregnancy progresses due to the gestationalage dependent change in placental P-gp protein abundance. 15 Finally, even though DEX and BET are epimers, they differ in their pharmacokinetic characteristics. For example, BET has lower hepatic clearance and longer half-life than DEX. 16,17 In addition, the formulation of the two ACS used for intramuscular administration differ, resulting in different pharmacokinetics. BET is administered as the 1:1 mixture of the phosphate (BET-P; fast release) and the acetate (BET-A; slow release) resulting in a sustained release of BET from the i.m. depo site compared to the rapid release of DEX following DEX-P i.m. administration. 18 To overcome the aforementioned challenges in determining fetal ACS exposure, methods to accurately predict (rather than measure) are needed. Therefore, we applied our recently developed m-f PBPK model to dynamically predict fetal exposure to the ACS. To do so, we needed to estimate both the CL int,Pgp,placenta and CL int,PD,placenta of the ACS (see Figure 1 for workflow). The latter was estimated using our previously adopted innovative approach that is based on the passive diffusion CL of midazolam (a model passive diffusion drug). The former was based on the in vivo K p,uu of the ACS estimated using the observed UV/MP values obtained from multiple maternal-fetal dyads. That is, we optimized the magnitude of placental P-gp efflux clearance (CL int,Pgp,placenta ) of each ACS to match (as closely as possible) the observed UV/MP ratios.
Because maternal plasma drug concentrations drive fetal plasma drug concentrations, to accomplish the above goals, we first optimized ACS plasma concentration in the nonpregnant population (top panel, Figure 1). Second, without changing any of the pharmacokinetic (PK) parameters of the ACS, they were input into our m-f PBPK model. This model incorporates all the physiological changes caused by pregnancy (including the 2-fold induction in hepatic CYP3A activity). Only the k a and t lag of the ACS were optimized to match the observed plasma concentration-time (C-T) profile of the ACS in pregnant women (second panel of Figure 1). Then, the fetal-placental ACS PK parameters were estimated (CL int,PD,placenta ) or optimized (CL Pgp,placenta ) to match the observed UV/MP ratio after i.m. administration of the ACS (third panel of Figure 1). Finally, to illustrate the utility of our approach, all the above ACS maternal-fetal PK parameters were input into our m-f PBPK model to dynamically simulate fetal exposure to several ACS dosing regimens alternative to those used in the clinic. 19 These alternative dosing regimens have the potential to increase the efficacy and reduce the toxicity of the ACS.

MATERIALS AND METHODS
Please, see Supplementary Information.

Verification of Simcyp PBPK model of ACS using the observed data from the non-pregnant Indian population
When the SimCYP Simulator was populated with observed clearance, volume of distribution at steady state (V ss ), renal clearance (CL R ), and other PK parameters obtained in the non-pregnant White population after i.v. administration of DEX and BET, the model recapitulated the observed ACS plasma concentration-time profiles within the a priori defined acceptance criteria (data not shown). Then, when the model, populated with these parameters, was used to predict area under the concentration-time curve from zero to infinity (AUC 0-∞ ) and CL in the Indian population after BET-P:A i.m. administration, the predicted values fell within 0.

Verification of m-f PBPK model of ACS in the pregnant population
Consistent with our previous observations, 20 to predict the i.v. plasma concentration-time profiles of the ACS in the White population at term, we assumed that CYP3Amediated hepatic intrinsic CL of the ACS was induced twofold by pregnancy (i.e., DEX CL hep,int = 106 L/h, BET CL hep,int = 62 L/h). Using our acceptance criteria, these modelpredicted plasma concentration-time profiles were successfully verified by comparing them with the observed profiles after i.v. administration of BET-P (Figure 3a

OpƟmizaƟon of PBPK Model of ACS in the Non-pregnant PopulaƟon
1. Populated our non-pregnant PBPK model with observed literature values of ACS physicochemical parameters and IV pharmacokineƟcs (e.g., MW, LogP, CL hep,int , CL R , V ss , f m , f u ) and confirmed that the model recapitulates the observed data aŌer IV administraƟon of the ACS. 2. Fixed the above parameters and then opƟmized PK parameters relevant to IM drug administraƟon (e.g., k a and t lag ) unƟl the plasma C-T profiles of the ACS predicted by the PBPK model matched those observed in the non-pregnant populaƟon aŌer IM administraƟon of the ACS.

VerificaƟon of m-f PBPK model of ACS in the Pregnant PopulaƟon
1. Verified the model with data observed aŌer IV administraƟon of the ACS to pregnant women. 2. OpƟmized PK parameters relevant to IM drug administraƟon (e.g., k a and t lag ) unƟl the plasma C-T profiles of the ACS predicted by the m-f PBPK model matched those observed in the pregnant populaƟon aŌer IM ACS administraƟon. T lag (0.2 h) were optimized to better describe the observed data (absolute average fold error [AAFE] = 1.3). In contrast, simulating a twofold induction of BET CL hep,int failed to predict maternal concentrations after i.m. administration of the BET-P:A mixture (AAFE = 2.03; one data point per subject; Figure 5a). Surprisingly, the CL hep,int that accurately described maternal BET concentrations with an AAFE of 1.41 was 11.2 L/h, a value much lower than that in the nonpregnant population (Figure 5b).

Optimization of DEX and BET K p,uu through sensitivity analysis
For BET, in vivo placental efflux clearance yielding K p,uu = 0.5 resulted in the best match between the predicted and observed UV/MP ratio (AAFE Kp,uu = 0.5 = 1.47) versus when no CL Pgp,placenta was invoked (AAFE Kp,uu = 1 = 2.19) (Figure 5d). For DEX, the theoretical UV/MP ratio plateau was estimated as 0.59 using the simple maximum effect (E max ) model (Figure 4c

Designing alternative ACS dosing regimens (i.m.) at GW30 using our m-f PBPK model
For the alternative ACS dosing regimens, maternal and fetal C max , minimum plasma concentration (C min ), and AUC 0-48 over the entire 48 h dosing regimen were computed. DEX-P alternative regimen (Figure 6b) of 12 mg administered every 24 h (vs. the reference regimen of 6 mg i.m. q 12 h; Figure 6a) maintained fetal AUC 0-48 of the reference regimen (342 ng*h/ml). Fetal fifth percentile C min decreased from 0.06 to less than 0.01 ng/ml. This alternative regimen also resulted in a twofold increase in maternal 95th percentile C max (259.0 ng/ml) compared to the 95th percentile C max (129.6 ng/ml) for the reference regimen. As expected, the total maternal exposure (AUC 0-48 ) remained at 771 ng*h/ml equal to that for the reference regimen.  Figure 7a) decreased fetal AUC 0-48 by 74% from 724 to 191 ng*h/ml. Fetal 5th percentile C min decreased by 80% from 5 to 1 ng/ml and remained greater than 1 ng/ml for 48 h. Maternal 95th percentile C max (33.7 ng/ml) decreased 64% versus the BET reference regimen (94 ng/ml). Total maternal mean AUC decreased 74% (375 ng*h/ml).

DISCUSSION
Respiratory failure due to RDS is one of the most common causes of death in premature infants in US neonatal intensive care units. Even among infants who survive, RDS is associated with a twofold increased risk of cerebral palsy and a 1.4-fold increased risk of epilepsy. 22,23 Safe and effective ACS dosing is critical to reduce morbidity and mortality due to preterm delivery. In the present work, for the first time, we used our m-f PBPK model to predict and then verify maternal-fetal exposure to ACS. Then, to illustrate the utility of m-f PBPK model, we used it to propose alternative ACS dosing regimens that could potentially maximize ACS efficacy while minimizing ACS toxicity or could be more convenient to implement in the clinic due to reduced frequency of administration.
To develop our ACS m-f PBPK model, we first ensured that our PBPK model could describe ACS exposure after i.m. administration to non-pregnant Indian population ( Figure 2). It could, as evidenced by successful model  (Figure 2c,d). However, our model modestly overpredicted BET CL in this population perhaps due to ethnic differences in BET CL. This was not of concern to us as the sole reason for use of this data set was to allow us to estimate the dual k a for BET-P:A necessary to describe two distinct absorption (or release) phases of the phosphate and acetate prodrug. Once the PBPK model was verified for the non-pregnant population, the drug-dependent parameters were fixed and our m-f PBPK model was populated with these parameters. Then, we verified our m-f PBPK model after i.v. administration of ACS in pregnancy (Figure 3). To do so, we incorporated our previously reported twofold induction of hepatic CYP3A4 activity at term. 20 BET and DEX are cleared from the body predominately by CYP3A metabolism. 24,25 Indeed, this magnitude of induction was consistent with the observed twofold increase in midazolam CL (a selective CYP3A probe) during the third trimester ( Figure 3). Although others have reported different magnitudes of CYP3A induction in the third trimester, these findings have been based on studies where a selective CYP3A probe was not utilized. [26][27][28] Indeed, i.v. BET-P C-T profiles in pregnant women were well explained by this twofold induction of hepatic CYP3A activity (Figure 3a). Therefore, surprisingly, the observed clearance after i.m. BET-P:A administration in pregnancy (5.7 L/h 13 ) was lower than after i.v. BET-P administration to the non-pregnant population (16.3 L/h; Figure 5a). The reasons for this decrease (not increase) are not clear and should be explored further. Because the goal of this study was to predict fetal rather than maternal drug plasma concentrations, it was important to accurately describe maternal BET plasma concentration-time profiles. Hence, we decreased BET maternal CL hep,int to best describe BET maternal plasma concentration-time profile after intramuscular BET-P:A administration (Figure 5b).
In order to accurately predict fetal DEX/BET plasma concentration-time profiles, it is important to account for the processes that govern transplacental transfer of these drugs into the fetus (i.e., passive diffusion and active placental efflux clearances). Because both drugs are substrates of P-gp, 10,11 which is highly abundant in placenta, 15,[29][30][31] we aimed to optimize their magnitude of P-gp mediated efflux to obtain fetal drug exposure (in vivo K p,uu ) that agrees with the observed data. Our PBPK modeling showed that the in vivo CL int,PD,placenta for both drugs was indeed large, and limited by placental blood flow (~45 L/h in third trimester). Moreover, not accounting for placental efflux resulted in significant overestimation of fetal drug exposure and therefore underestimation of the dose needed to administer to the mother to achieve fetal plasma concentration of greater than 1 ng/ml.
The obtained K p,uu value for BET (0.5) was determined with greater confidence ( Figure 5) than for DEX (K p,uu = 0.48 or 0.31; Figure 4) due to excellent agreement of our model predicted BET UV/MP ratio with that observed. 13,32,33 In preliminary Transwell efflux studies of DEX and BET in P-gp and BCRP overexpressing MDCKII cells, we found similar in vitro P-gp mediated efflux ratios for the two ACS and no BCRP-mediated transport, respectively (unpublished data). These data strongly suggests that the in vivo K p,uu of these drugs should be similar. This is not surprising as these two ACS are epimers and P-gp does not readily discriminate between isomers. For this reason, to develop alternative dosing regimens for DEX, we used K p,uu = 0.48, a value close to that of BET. However, the in vivo DEX UV/MP data suggest that DEX K p,uu could just as well be 0.31. To resolve this discrepancy, additional in vivo DEX data sets are needed to better define its in vivo K p,uu .
The efficacy of the ACS reference regimens in reducing RDS is modest. When the reference ACS regimens are used, the relative risk of RDS (compared with placebo) ranges from 0.6 to 1.16 [34][35][36][37] suggesting a need to optimize the ACS dosing regimens. However, increasing the ACS dosing rate could potentially enhance ACS toxicity. For example, when the reference DEX-P regimen is used, maternal infections significantly increase from 6% in the placebo arm to 10% in the ACS arm (odds ratio of 1.64). 38 Likewise, there are concerns of long-term neonatal neurodevelopmental toxicity from use of the ACS. 6,7,39 Therefore, to demonstrate the in vivo clinical application of our m-f PBPK model, we used it to devise alternative dosing regimens for ACS that could enhance their efficacy while minimizing their toxicity. It is not clear whether the efficacy and toxicity of these ACS is related to their maternal-fetal exposure (AUC) or C max or both.
In the absence of this information when designing alternative ACS dosing regimens, we took the conservative approach of not exceeding the reference regimen ACS maternal-fetal exposure (AUC) or C max or both. In addition, for BET only, we designed dosing regimens based on maintaining fetal plasma concentration greater than 1 ng/ml based on the efficacy data in sheep. 18 When designing these alternative dosing regimens, we also took into consideration the dosing frequency of the regimen so that it was convenient to implement in the clinic (not more frequent than twice a day).
To design a convenient alternative dosing regimen for i.m. DEX-P (Figure 6b), fulfilling criterion three in the Method section, we reduced the number of doses because the reference regimen is already administered q 12 h (Figure 6a). Administering a twofold higher dose, but less frequently (q 24 h) helped us maintain maternal and fetal AUCs (fulfilled criterion 1a and 2). Consequently, maternal peak concentrations rose, which in the clinic may increase efficacy but produce higher incidence of adverse events than observed after reference dosing regimen (maternal infection rate of 5%-6% 34 ). Due to the lack of animal or human data on fetal efficacy after lower i.m. DEX-P doses, we refrained from designing a regimen that would decrease fetal (and maternal) AUC. However, when such data are available, our model could be used to design such dosing regimens. F I G U R E 7 Predicted fetal and maternal UV plasma concentration-time profiles at GW 30 for the IM BET-P:A reference (a) and an alternative (b-d) dosing regimens using our final m-f PBPK model (a) The reference dosing regimen resulted in greater than 1 ng/ml fetal UV 5th percentile C min (5 ng/ml) and mean UV C min (7.3 ng/ml). Fetal AUC 0-48 was 724 ng*h/ml. The corresponding maternal values were 94 ng/ml and 1424 ng*h/ ml, respectively. (b) The alternative dosing regimen 1 decreased fetal UV 5th percentile C min by from 5 to 1 ng/ml and maintained UV plasma concentration above 1 ng/ml for the duration of drug administration (48 h). Fetal AUC 0-48 decreased to 145 ng*h/ml, maternal 95th percentile C max to 18.7 ng/ml and maternal AUC 0-48 to 285 ng*h/ml. All listed parameters as well as the total dose decreased by 80% compared to reference dosing regimen. (c) The alternative dosing regimen two decreased fetal UV 5th percentile C min by 80% (from 5 to 1 ng/ml) and maintained UV plasma concentration above 1 ng/ml for the duration of drug administration (48 h). Fetal AUC 0-48 decreased 74% to 191 ng*h/ml. The maternal 95th percentile C max (33.7 ng/ml) decreased 64% and the AUC 0-48 (375 ng*h/ml) decreased 74%. (d) The alternative dosing regimen three (currently used in BETADOSE clinical trial, see Schmitz et al., 2019) decreased fetal UV 5th percentile C min by 53% (from 5 to 2.3 ng/ml) and maintained UV plasma concentration above 1 ng/ml for the duration of drug administration (48 h). Fetal AUC 0-48 decreased 41% to 424 ng*h/ml. The maternal 95th percentile C max (75 ng/ml) decreased 22% and the AUC 0-48 (833 ng*h/ml) decreased 42%. Predicted mean plasma concentrationtime profiles are solid lines, 5th and 95th percentiles are dashed lines. Horizontal dotted line in maternal plasma concentration-time profiles denote maximum cutoff value for maternal 95th percentile C max (94 ng/ml defined by the reference dosing regimen). Horizontal dotted line in fetal plasma concentration-time profiles denote minimum cut-off value for fetal 5th percentile C min (1 ng/ml). These C min and C max values are absolute values determined over the entire 48 h period. AUC, area under the concentration-time curve; BET, BET, betamethasone; BET-P:A, betamethasone phosphate:acetate;; BET-P:A, BET phosphate:acetate; C max , maximum plasma concentration; C min , minimum plasma concentration; GW, gestational weight; m-f PBPK, maternal-fetal physiologically-based pharmacokinetic; UV, umbilical vein To design a convenient alternative dosing regimen for IM BET-P:A (criterion 3; Figure 7b), we could increase (q 12 h), maintain (q 24 h; Figure 7b) or decrease (single dose over 48 h; Figure 7c,d) the number of administered doses. Increasing the number of doses (to q 12 h) to maintain maternal and fetal AUC, as for DEX-P above, produced similar maternal and fetal C max and C min as in reference regimen, and was therefore not considered further. Availability of sheep data 18,40 gave us a new guideline for efficacious fetal drug plasma concentrations greater than 1 ng/ml (criterion 2b). These data allowed us to decrease BET-P:A dose from 12 to 2.4 mg q 24 h (Figure 7b) or to 5.4 mg single dose (Figure 7c) and fulfill criteria 1b and 2. These decreased doses resulted in decreases in fetal AUC and C max and relied on the assumption that human efficacious BET plasma concentrations are equal to that in sheep. This assumption and proposed dosing regimens need to be assessed in the clinic to ensure fetal therapeutic benefit. The ongoing BETADOSE clinical trial 21 is exploring the efficacy and safety of another BET-P:A dosing regimen that reduces total administered i.m. dose (12 mg i.m. BET-P:A administered as a single dose). Simulations of maternal-fetal drug plasma concentrations for this dosing regimen are provided as alternative dosing regimen three (Figure 7d). Studies to verify these predictions are urgently needed to promote the optimal dose of ACS to administer for preterm labor, especially in low-to-middle income countries, where rates of preterm birth are high. Overall, lower BET drug plasma concentrations, smaller fluctuations, and less frequent administration make it more attractive therapeutic option than DEX, but this conclusion should be further evaluated in the clinic.
There are several limitations to this study, most of which are related to the limited clinical data in maternal-fetal pairs available for these ACS. First, maternal-fetal PK data on these ACS are limited (especially for DEX-P) and therefore any inaccuracies in the published data will result in inaccuracies in the predicted dosing regimens. Second, limited clinical data (including PKs) are available for different ethnic populations, especially those from countries where preterm delivery rates and thus ACS use are high. DEX and BET CL for the White non-pregnant population was 15 L/h and 7-17 L/h, 19 higher than the value for the Indian non-pregnant women (DEX: 9-10 L/h; BET: 5-6 L/h; Figure 2b,d,f 41 ). A similar observation has been made for nifedipine, another CYP3A substrate. 42 Therefore, we used PK parameters from the White non-pregnant population and verified our m-f PBPK model with data from White pregnant women. Hence, PK/PD studies in pregnant Indian women are needed and are underway. 34 One such study, conducted by the World Health Organization, is investigating the efficacy of i.m. BET-P at 2 mg q 12 h for 48 h. Therefore, for comparison, we predicted the maternal-fetal exposure to BET for this dosing regimen ( Figure S2). If this regimen is found to be efficacious as the BET-P:A reference regimen, it has the potential to reduce maternal-fetal risks. Ideally, future studies will produce highquality data sets with maternal-fetal paired sampling, that include accurate recording of time post last dose, and stabilization of hydrolysis of the ACS prodrugs after collection. We should note here that the in vivo cleavage of BET-P to the BET is rapid and appears to be complete within 60 min. 43 Corresponding data for DEX-P or BET-A are not available. Because in vivo hydrolysis of BET-A is slower than BET-P, if a substantial amount of BET-A is present in the drawn blood sample and is subsequently hydrolyzed to BET prior to freezing the plasma, this could potentially explain the lower than anticipated clearance of BET in pregnant women when BET-P:A is administered but not when BET-P is administered. Parenthetically, the ACS UV/MP data used here were obtained at least 60 min after ACS prodrug administration.
To our knowledge, this is the first study to estimate K p,uu , CL int, Pgp,placenta , and CL int,PD, placenta of a placental P-gp effluxed drug from the observed UV/MP data and then to dynamically predict fetal plasma concentration of the drug, in this case the ACS. Then, these values were populated in our m-f PBPK model to simulate maternal-fetal ACS exposure, for various ACS dosing regimens, during pregnancy. Our approach and our m-f PBPK model could be used in the future to predict maternal-fetal exposure to any drug and to devise alternative dosing regimens (including those not described here) of the drug to guide drug therapy of the maternal-fetal dyad.