A Simplified PBPK Modeling Approach for Prediction of Pharmacokinetics of Four Primarily Renally Excreted and CYP3A Metabolized Compounds During Pregnancy
- First Online:
- Cite this article as:
- Xia, B., Heimbach, T., Gollen, R. et al. AAPS J (2013) 15: 1012. doi:10.1208/s12248-013-9505-3
- 1k Views
During pregnancy, a drug’s pharmacokinetics may be altered and hence anticipation of potential systemic exposure changes is highly desirable. Physiologically based pharmacokinetics (PBPK) models have recently been used to influence clinical trial design or to facilitate regulatory interactions. Ideally, whole-body PBPK models can be used to predict a drug’s systemic exposure in pregnant women based on major physiological changes which can impact drug clearance (i.e., in the kidney and liver) and distribution (i.e., adipose and fetoplacental unit). We described a simple and readily implementable multitissue/organ whole-body PBPK model with key pregnancy-related physiological parameters to characterize the PK of reference drugs (metformin, digoxin, midazolam, and emtricitabine) in pregnant women compared with the PK in nonpregnant or postpartum (PP) women. Physiological data related to changes in maternal body weight, tissue volume, cardiac output, renal function, blood flows, and cytochrome P450 activity were collected from the literature and incorporated into the structural PBPK model that describes HV or PP women PK data. Subsequently, the changes in exposure (area under the curve (AUC) and maximum concentration (Cmax)) in pregnant women were simulated. Model-simulated PK profiles were overall in agreement with observed data. The prediction fold error for Cmax and AUC ratio (pregnant vs. nonpregnant) was less than 1.3-fold, indicating that the pregnant PBPK model is useful. The utilization of this simplified model in drug development may aid in designing clinical studies to identify potential exposure changes in pregnant women a priori for compounds which are mainly eliminated renally or metabolized by CYP3A4.
Key wordsPBPK modelspharmacokineticsphysiological changespregnancysystemic exposure
As the use of medications during pregnancy is common (1–3), efforts continue to focus on safe and efficacious dosing regimen allowing women and their doctors to make informed decisions about continued treatments during pregnancy. While some drugs are known to show altered pharmacokinetics during pregnancy (4), anticipation of systemic exposure for medications taken by pregnant women has been difficult to determine, in part due to the lack of readily available whole-body physiologically based pharmacokinetics (PBPK) models. PBPK models have become more popular with both pharmacokinetic and formulation scientists, and advantages for the application of PBPK models as tools have been described (5,6), along with their structure, methodology, and implication in drug development (7–10). During pregnancy, apart from the development of the placenta and fetus, anatomical and physiological changes occur such as changes in blood flow, cardiac output, protein binding, renal, and hepatic functions (11,12); all of which may have significant effect on the pharmacokinetics (PK) and pharmacodynamics (PD) of drugs. A number of recently published pregnant PBPK models (p-PBPK) incorporate comprehensive information related to anatomical, physiological, and metabolic changes which required complex and resource-intensive parameter inputs and custom programming (13,14). Although these models account for various time-variant parameter changes during gestation, key factors that influence the pharmacokinetics for pregnant women have only been partially described. Moreover, it is not always practical for clinical or pharmaceutical researchers to develop highly complex PBPK models in limited timeframes. Here, we describe an easily implementable mechanistic PBPK model that can characterize the disposition of drugs in pregnant women for four primarily renally excreted and CYP3A-metabolized compounds.
Significant changes in PK are more likely to be observed for drugs which have single elimination pathways. It is known that the CYP450 pathway involved is affected by pregnancy (15). Hence, reference drugs selected were compounds which are (a) primarily excreted via kidney, (b) known substrates of CYP450 isoenzymes and which undergo phase II metabolic pathways (e.g., N-acetyltransferase and glucoronidation, etc. (16)). Due to the regulations of the Food and Drug Administration, Clinical PK trials typically exclude both pregnant women and those of reproductive age because of ethical and practical considerations (17). For this reason, often only limited data are available for altered systemic PK parameters such as volume of distribution (Vd) and clearance (CL) change during pregnancy (17). Metformin, digoxin, midazolam, and emtricitabine were chosen as study drugs as published data were available (18–20). For metformin, digoxin, and emtricitabine, the major route of elimination is renal excretion. Midazolam is primarily metabolized and eliminated by CYP3A, which allows incorporating a specific-metabolizing enzyme change in the PBPK model. To our current knowledge, p-PBPK models have not been extensively described for prediction and analysis of PK changes for renally excreted compounds during pregnancy. Midazolam p-PBPK model has been established by Gaohua using Matlab Simulink. However, Cmax, area under the curve (AUC), and distribution phase were underpredicted (13). Here, we applied a simplified p-PBPK modeling strategy to predict the PK profiles in pregnant women by considering the essential physiologic and metabolic changes that occur during pregnancy for metformin, digoxin, midazolam, and emtricitabine. Furthermore, the simulated PK profiles were compared against published PK profiles from clinical outcomes (examples from literature) for model verification.
MATERIALS AND METHODS
Computer Hardware and Software
GastroPlus (version 8.0, Simulations Plus, Inc, CA, USA) embedded with the Advanced Compartmental Absorption Transit (ACAT) model and PBPKPlus™ module was run on a Lenovo computer with Intel® Core™ i5 processor. The ACAT model, originally established by Yu and Amidon (21), enables the prediction of rate and extent of oral drug absorption. The PBPKPlus™ module incorporates a whole-body PBPK model which defines mathematical compartments for all major tissues using a group of differential equations (22). This module can further predict the amount of drug that distributes to each tissue as a function of time as well as the PK profiles using the input parameters of physicochemical properties (e.g., solubility, permeability, LogP, pKa, and particle sizes) and disposition data (e.g., tissue-to-plasma partition coefficients (Kp), CL, and first-pass extraction ratio (FPE)).
Development and Verification of PBPK Model in Pregnant Women
Development and Validation of PBPK Model in Nonpregnant Subjects
Tissue-to-Plasma Partition Coefficients (Kp) of the Model Drugs
Reproductive organ (fetoplacental unit)
Red bone marrow
Yellow bone marrow
Rest of body
Physiological Changes Considered in the Pregnant PBPK Model
Main Anatomical, Physiological, and Biological Parameters Needed for p-PBPK Model and Equations Used to Calculate Parameter Values During Different Gestation Ages (GA) (14)
Today body weight (kg)
TBW = 61.1 + 0.2409 GA + 0.0038 GA2
Cardiac output (CO)
CO = 301 + 5.916 GA − 0.088 GA2
Total body fat mass (TFM)
TFM = 17.14 + 0.1305 GA + 0.0008 GA2
Weight of the uterus
Weight of the uterus = 80 + 8.2931 GA + 0.3546 GA2
Fetal volume = 0.01 exp(13.604(1 − exp(−0.0702GA)))
Placenta volume = 0.0 − 0.0716 + 0.9146 GA2 − 0.0122 GA2
Aminotic fluid volume = 0 + 1.9648 GA − 1.2056 GA2 + 0.2064GA3 − 0.0061 GA4 + 0.00005 GA5
Volume of fetoplacental unit
Fetoplacental volume = Uterus weight + Placenta volume + Fetal volume + Amniotic fluid volume
Blood flow of uterine
Uterine blood flow = 1.71 + 0.2068 GA + 0.0841 GA2 − 0.0015 GA3
Plasma volume = 2.5 − 0.0223 GA + 0.0042 GA2 − 0.00007 GA3
Red blood cell (RBC) volume
RBC volume = 1.49 + 0.0098 GA
Total blood volume
Total blood volume = plasma volume + RBC volume
Glomerular filtration rate (GFR)
GFR = 114 + 3.2367 GA − 0.0572 GA2
Drug absorption may be altered as nausea and vomiting during early stage of pregnancy may result in delayed gastric emptying and reduced small intestine motility (16,24). Prolonged GI transit time can potentially increase the tmax but reduce the Cmax (16). In addition, results have shown that there are no significant differences in basal gastric pH during pregnancy, compared with nonpregnancy values (14). However, gestation-induced alteration in absorption is clinically insignificant in most of the known case studies (25–27). Therefore, this was not considered in the current model. In the current PBPK model, the GastroPlus ACAT model (with default ASF optimized LogD v6.1) was used to describe drug absorption for nonpregnant and pregnant states.
For changes of volume of distribution, with the simplified pregnant PBPK model (model 1), drug distribution during pregnancy was assumed to be mainly impacted by changes of body weight and plasma protein levels. The average maximum weight gain during the pregnancy period is about 12.5 kg (28), which reflects an increase of total body water (∼8 L) and adipose tissue (∼4 kg) (15,16). Plasma protein levels, including albumin and α-acid glycoprotein, generally decrease during pregnancy relative to prepregnancy values. For example, plasma protein level gradually decreases since second trimester and eventually reach ∼70 to 80% of the normal level at time of delivery (14,15,29). The fraction of fup may increase due to decreased albumin or α-glycoprotein levels, which may increase drug distribution for drugs with high protein binding. However, the quantitative correlation between plasma level and fup is not well understood. Therefore, fup changes during pregnancy cannot be directly predicted based on plasma protein levels, and this is not considered in most of pregnant PBPK models (13,30). Here, fup was assumed unaltered during pregnancy unless fup was measured in pregnant women during the clinical trials (19). If needed, the Kp values were recalculated by the TC equations when plasma protein binding data were available. In the comprehensive pregnant PBPK model (model 2), more gestation-related changes and parameters were considered. A lumped fetal and placenta compartment (fetoplacental unit) was used to represent uterus, fetus, placenta, and amniotic fluid. Because substantial morphological and physiological changes in fetoplacental unit occur during pregnancy, the mass or volume of uterus, placenta, amniotic fluid, and fetus undergoes dramatic changes, which comprises about 35% of the total gestational weight gain during pregnancy (31). Increase of body fat may account for about 55% of total weight gain during pregnancy (32). During the course of pregnancy, plasma cell volume may increase 34–70% while red blood cell volume increases to a lesser extent along with the plasma volume (14). Overall, the apparent Vd may increase during pregnancy, further resulting in lower systemic drug concentration when the unadjusted dose strength (dose used in drug labeling) is used in pregnant subjects. The change in the volume of these tissues during pregnancy can be described by using the equation summarized in Table II.
The change of metabolic enzyme activities, GFR, and renal blood flow rate can significantly alter the drug elimination during pregnancy. The metabolic activities of CYP3A4, CYP2D6, UGT1A1, and UGT1A4 are known to be induced by elevated estrogen/progesterone levels while the activities of CYP1A2 and CYP2C19 are known to be inhibited by progesterone and estradiol (12,16,33). For CYP3A metabolized compounds (e.g., midazolam), an average 35% increase of systemic CL and FPE in gut and liver were included in pregnant PBPK model across the gestational ages (14–40 weeks) based on the findings by Tracy et al. (34). For really excreted compounds, the GFR and net secretion CL can increase up to 40% and 50% during pregnancy, respectively, resulting in enhanced renal excretion of unchanged drugs (16). Based on available data, GFR gradually increases during first 20–26 weeks of gestation and then reaches a plateau or slightly declines in the late stages of pregnancy (14,35). The increase in GFR in a particular gestation week was quantitatively estimated for the current pregnant PBPK model by equations reported by Abduljalil et al. (14). The dynamic changes of net renal secretion during pregnancy have not been well characterized. The net secretion clearance significantly increases during mid and late pregnancy, while is unaltered in early pregnancy (<12 weeks) as in the case of metformin (18). This finding also explains the ∼40% increase in metformin net secretion clearance in mid and late pregnancy (>22 weeks), which was considered in current PBPK model. On the other hand, the current model does not account for the change of net secretion clearance in early pregnancy (<14 weeks). In addition, there are no detailed literature data to describe the trend of secretion clearance changes between the 14 and 22 weeks of gestation. Therefore, the current PBPK model may not be suited to predict the PK of renally eliminated compounds for pregnant women during 14–22 weeks of gestation.
Cardiac output and uterine blood flow rate can significantly increase during pregnancy. The mean maternal cardiac output has a peak of ∼7.30 L/min during the third trimester, which drops down to the baseline value during postpartum period (36,37). Uterine blood flow increases during pregnancy and peaks at late pregnancy as the uterus receives about 12% of cardiac output in comparison to 0.5% in prepregnant women. The longitudinal increase of cardiac output (models 1 and 2) and uterine blood flows (model 2 only) during pregnancy were calculated using polynomial regression equations in Table II.
Case Example 1: Metformin (BDDCS Class 3)
Fold Change of Gestation Age-Dependent Parameters Used for Four Case Studies in the Developed p-PBPK Model
Gestation ages (week)
Volume of fetoplacental unit
Blood flow of fetoplacental unit
Volume of total body fat
Total blood Volume
Parameters for Model Drugs in Non-Pregnant and Pregnant PBPK Model. Parameter Values Used in the Pregnancy PBPK Model are in Italics (e.g., Nonpregnant Parameter Values → Late Pregnant parameter values)
Basic compound/formulation Information
Main clearance pathways
Parameters not influenced by pregnancy (e.g., physicochemical properties)
Log D at pH 7.4
Solubility in water (mg/mL)
Effective permeability (10−4 cm/s)
Diffusion Coefficients (cm2/s)
Parameters influenced by pregnancy (e.g., disposition parameters)
Fraction unbound in plasma (fup, %)
100 → 100
63 → 67
0.61 → 0.71
96 → 96
Vss predicted by PBPK (L)
82.7 → 101b
643 → 872b
85.8 → 125b
156 → 210b
Gut FPE (%)
40 → 54
Hepatic CL (L/h)
4.2 → 4.2
24.5 → 33.1
Renal CL (L/h)
28.5 → 40.0
6.9 → 10.2
19.1 → 26.6
9.9 → 13.8
7.4 → 10.5
6.8 → 9.4
Filtration CL (L/h)
9.9 → 13.8
4.6 → 7.01
6.6 → 9.1
Secretion CL (L/h)
18.7 → 26.2
2.3 → 3.11
12.5 → 17.6
Case Example 2: Digoxin (BDDCS Class 3)
Case Example 3: Midazolam (BDDCS Class 1)
Midazolam (MW of 325.8 g/mol) is a BCS/BDDCS class 1 compounds with high aqueous solubility and high permeability (Peff, 6.28 × 10−4 cm/s). Midazolam is a sensitive CYP3A substrate with significant gut metabolism and hepatic extraction after oral dose (19,51–53). Midazolam pharmacokinetics were studied in 13 women who received an oral dose of 2 mg immediate release formulation during trimester (28–32 weeks) and PP (6-10 weeks) (19). Drug elimination of midazolam is known to be mediated mainly by CYP3A metabolism and the hepatic CL used in the current model was estimated as the average value of the CL values obtained from multiple study reports (ESM Table S1). Kp values in different compartments were calculated by Rodgers and Rowland’s TC equation (54) and fup was reported as 0.67 during PP but decreases to 0.63 during late pregnancy (19). Due to increased enzyme activity of CYP3A during pregnancy, hepatic CL and gut FPE were increased by 35% (Table IV; 34). The changes of physiological parameters (as ratios of pregnancy vs. nonpregnancy, Table III) during late pregnancy (average 30 weeks) were estimated based on the modeling strategy described above. The ratios were used to adjust the input physiological parameters during pregnancy for models 1 and 2, respectively (see ESM Fig. S2 for detailed parameter values). PK simulations were performed for women during late pregnancy and PP.
Case Example 4: Emtricitabine (BDDCS Class 1)
Emtricitabine (MW: 247.3) is a weak basic drug with pKa of 2.65, with high permeability and high solubility. The compound is mainly excreted through kidneys and is likely an OCT-2 substrate (45,46). It is a once-daily nucleoside reverse transcriptase inhibitor that selectively and potently inhibits human immunodeficiency virus type 1 (HIV-1) replication (55). Emtricitabine is commonly used in HIV-1 infected women during pregnancy. Emtricitabine pharmacokinetics were studied in 26 women who received an oral dose of 2 mg immediate release formulation during trimester (median, 35 weeks) and PP (median, 8 weeks). Clinical PK data were digitized from literature (20) for women during third trimester and PP women after 4 days of 200 mg oral doses given once daily, and the oral CL (CL/F) and absolute oral bioavailability was observed as 20.6 L/h and 93%, respectively. Therefore, in the current model, the systemic CL of emtricitabine for healthy nonpregnant women was estimated to be 19.1 L/h. The generic GFR values (114 mL/min or 6.84 L/h) of healthy nonpregnant women were used in the model (14). The corresponding renal net secretion clearance (12.5 L/h) was calculated by Eqs. (2) and (3). Changes of physiological parameters (as ratios of pregnancy vs. nonpregnancy, Table III) during late pregnancy (average 35 weeks) were estimated based on the modeling strategy described. The ratios were used to adjust input physiological parameters during pregnancy for models 1 and 2, respectively (see ESM Fig. S3 for detailed parameter values). PK simulation for emtricitabine was performed for women during late pregnancy and PP based on the parameter input provided in Table IV.
Case Example 1: Metformin
Comparison Between Observed and PBPK Model Predicted Parameters
Pregnant Women (third trimester)
Non-pregnant Women (PP)
Fold Change (pregnant vs. non-pregnant)
Prediction Fold errorc
Prediction Fold errorc
Prediction Fold errorc
AUC12h_ss (ng h/mL)
AUC48h (ng h/mL)
AUCinf (ng h/mL)
AUC24h_ss (μg h/mL)
Case Example 2: Digoxin
The muscle Kp value (4.3) estimated by Poulin and Theil’ TC equation was dramatically lower than the reported Kp values (35.7), which resulted in underprediction of terminal t1/2 (predicted (11.3 h) vs. observed (46.5 h)). The model was refined with literature reported muscle Kp value and accurately simulate the terminal phase PK curve with a predicted t1/2 of 43.5 h. The simulated plasma concentrations vs. time profiles were in agreement with the observed profiles (Fig. 3b) with R2 of 0.81 and 0.90 for models 1 and 2, respectively. The predicted AUC and Cmax and the fold changes of the PK parameters (pregnant vs. PP) were within 1.2-fold of observed values (Table V). The predicted renal clearance (CLr, 10.4 L/h) for women during third trimesters was in agreement with observed data (10.9 L/h). The predicted fold change of Cmax (pregnant vs. nonpregnant) were 0.83 and 0.80 using models 1 and 2, respectively, similar to observed ratio, while the predicted fold change of AUC (model 1, 0.84; model 2, 0.85) has no large difference with the observed values (0.81).
Case Example 3: Midazolam
The ex vivo measured unbound drug fraction in plasma is 0.0071 and 0.0061 in PP and pregnant women, which were incorporated in the midazolam PBPK models. Midazolam is a compound that undergoes extensive metabolism in liver and gut. The simulated plasma concentration vs. time profile captured the observed profile well (Fig. 3c) with an R2 of 0.78 and 0.76 for models 1 and 2, respectively. The predicted AUC and Cmax were within 1.3-fold of observed values (Table V) using either models 1 or 2. The simulated ratio of Cmax (pregnant vs. PP = 0.54–0.56) was slightly underpredicted, compared to the observed values (0.72). On the contrary, the ratio of AUC was accurately predicted by the model (predicted (0.54–0.55) vs. observed (0.54)) (Fig. 4). The model successfully estimated the terminal phase half-life (predicted, 2.5 h; observed, 2.4 h). The PBPK model also predicted the short tmax and rapid absorption of midazolam. Cmax prediction for pregnant women was improved (prediction fold error, 1.2) compared with the model reported previously (13) which underpredicted Cmax by ∼40%. The prediction fold error of AUC during pregnancy using the current PBPK model was 1.3 regardless of the type of model, which was similar to the fold error of 1.35 (observed AUC of 9.5 ng h/mL vs. predicted AUC of 7.05 ng∙h/mL) predicted by a recently described midazolam p-PBPK model (13), indicating that the current PBPK model exhibits good prediction accuracy for pregnant women. In addition, there were no significant differences in simulated results between models 1 and 2.
Case Example 4: Emtricitabine
Oral Vd/F reported by the literature is about 4 L/kg (20) and the bioavailability with an IR capsule formulation is reported to be 93%, suggesting that the systemic Vss is about 3.7 L/kg. The Rodger and Rowland’s TC equation based on the input physicochemical parameters dramatically underpredicted Vss (0.60 L/kg). To our best knowledge, no observed tissue distribution data are available. Therefore, the Kp values in all tissues were multiplied by a scaling factor of 5 (based on nonpregnant data) and the simulated PK profiles fit to the observed data (Fig. 3d) with R2 of 0.80 and 0.84 for models 1 and 2, respectively. Predicted AUC and Cmax were 1.0- and 1.5-fold of observed clinical data (Table V), respectively. The predicted fold changes of AUC and Cmax (pregnant vs. nonpregnant) were within 1.3-fold of the observed ratios. The model predicted renal clearance (CLr, 26.6 L/h) during third trimesters was in agreement with observed oral clearance (25.3 L/h). Therefore, the emtricitabine pregnant PBPK model is well verified and can well predict the alternation of PK during late pregnancy. There were no significant differences in simulated results between models 1 and 2.
Ideally, the use of pharmacologic agents would be minimized for pregnant women. However, pregnant women with certain medical conditions may still require ongoing treatment. Clinical trials actively exclude pregnant women due to ethical and scientific reasons. Therefore, human PK data are seldom collected during pregnancy for appropriate dosage and frequency of administration. Even after years of marketing, product labels regarding PK and dose adjustments often provide limited information for pregnant patients. The US Food and Drug Administration and European Medicine Agency have required that application sponsors collect and maintain data on the effects of prescription medicines that are used by pregnant women with a new labeling system. Thus, regulatory agencies encourage drug companies to anticipate and predict drug exposure during pregnancy prospectively. Physiological changes during pregnancy can alter drug disposition according to clinical data reported in the literature. Currently, optimization of dosing regimen for pregnant women can be challenging, especially in the absence of clinical data. PBPK has been recently recognized as useful tool to provide valuable insights to aid clinical design. With incorporation of gestation-induced physiological changes, the pregnancy PBPK model can serve as a feasible alternative to empirical dose adjustment and guide the clinical trial design in pregnant women.
While PBPK models now have increased impact and applications in the pharmaceutical industry, the limitations in PBPK modeling need to be acknowledged and understood. In particular, Kp values can be inaccurately estimated using the in silico equations, resulting in biased prediction of Vd (56). If in vivo distribution data are not available in such situations, PBPK models cannot be readily established with high confidence. However, in some cases, if the PK profile of the compounds is characterized by a mono-exponential equation (similar to a one-compartmental model), the application of a uniform scaling factor on each Kp value likely provides good estimations for Vd and the shape of PK profiles as in the described case study of emtricitabine. The deviation in Kp predictions using in silico equations can be safely assumed to be similar across all tissues and can be corrected by applying a scaling factor to all Kp predicted by TC equations, if drug distribution can reach equilibrium rapidly (as with the assumption of a one-compartmental PK model). However, if the PK profile is described by a complicated multi-exponential equation, such assumptions may not stand because the magnitude of deviation of Kp prediction can differ between tissues (e.g., digoxin). In such cases, the PBPK modeling approach may require additional data, such as drug tissue distribution data (e.g., quantitative whole-body autoradiography; 22).
Several pregnancy PBPK models have been introduced for xenobiotic disposition in pregnant women (57–59). However, few publications focused on the investigation of maternal PK alternation caused by the effect of pregnancy. Recently, Gaohua et al. developed a first pregnancy PBPK model incorporating time-dependent activity of cytochrome P450 enzymes (e.g., CYP1A2, CYP2D6, and CYP3A4) and implementing gestational time-dependent physiological changes (13). This pregnancy PBPK model successfully described the disposition for drugs which undergoes extensive metabolism by major cytochrome P450 enzymes. Although a dynamically parameterized pregnant PBPK model can predict the PK outcome throughout the pregnancy, model development, validation, and application may be time-consuming due to its complexity. In general, most PK/PD studies in pregnant women are conducted in third trimesters, with the baseline assessment for comparison to the pregnant state during the PP period. To this end, model development efforts, especially modeling validation, mainly target the periods of late pregnancy when PK/PD endpoints are most commonly collected in clinical studies. Maternal drug clearance is often altered due to changes in metabolic activity of enzymes, GFR, and renal secretive transporters. The change in enzyme activity (induction or inhibition) plays an important role in altering PK for many drugs. Our practical and simplified PBPK model incorporates the change of essential parameters (e.g., BW, CO, GFR, CL, etc.), and successfully estimated the PK of several compounds during late pregnant and PP periods. The fetal and placenta, adipose, plasma, and blood volume have dramatic changes during pregnancy, which are likely expected to exhibit significant effects on drug distribution. Therefore, these changes are considered and incorporated in model 2. Interestingly, the predicted results were very similar for four drugs using the simple model (model 1) and the comprehensive model (model 2). This suggests that the increase of drug volume of distribution (unit, liter) during pregnancy can be explained by total body weight gain and is nearly proportional to the body weight. The individual contribution of the tissues/organs, even with substantial gestation-induced physiological changes, may be less important if body weight changes have been considered by the p-PBPK model, unless compounds distribute much more extensively into these tissues than other tissues/organs without significant alteration. For example, all four compounds are expected to have no extensive distribution into adipose tissue (Kp < 2), and thereby the proportionality between volume of distribution and body weight is not altered by fat gain, as shown by the models 1 and 2 predicted Vss values listed in Table IV. For tissue/organ gain with less composition of the total weight gain, such as fetaplacental unit which accounts for ∼35% of total weight gain (65% for adipose tissue), their volume increase may have even less influence on PK. The current p-PBPK model can be easily implemented using conventional PBPK modeling tools (e.g., GastroPlus or Simcyp), and should be sufficient to provide rational and accurate prediction for most of drugs which are mainly eliminated via the kidney or metabolized by CYP3A4 in the liver.
For renally excreted compounds, such as metformin, digoxin, and emtricitabine, the predicted fold decreases in AUC and Cmax were comparable to the observed results. The simulated results suggested that the decrease of systemic exposure in these compounds were mainly attributed to an increased renal secretion and filtration during pregnancy. The increase of GFR in pregnancy is correlated to dilation of renal vasculature caused by the systemic vasodilation possibly mediated by progesterone and relaxin (35,60). Enhanced renal secretion CL has been reported for metformin, digoxin, amoxicillin, and certain endogenous compounds during pregnancy, suggesting of pregnancy-induced enhancement of expression and/or function of renal secretion transporter regulation. Within these three modeled compounds, namely metformin, digoxin, and emtricitabine, are known as the substrates for transporters expressed on apical or basolateral (luminal) membrane of the kidney proximal tubules. For example, OCT2 plays an important role in the renal clearance of metformin in humans (61). Similarly, basolateral renal transport of emtricitabine appears to proceed via OCT2 in S2 and/or S3 segments of the proximal tubule (62). Also, digoxin is well-known as the substrate of P-gp which transports drugs across the apical membrane of renal tubular epithelium (63–65). Although magnitude of enhanced activities for these transporters during pregnancy is not well-known, the alternation of net tubular secretion during pregnancy is considered in the current PBPK model by assuming 40% increase of CLsecretion based on the previous evaluation in the pregnancy PK of metformin (18). Interestingly, the predicted and observed AUC and Cmax fold changes are less than 30% in late pregnancy, compared to PP period. Accounting for the uncertainty caused by intersubject variability, the effect of pregnancy on PK appears not be very large for drugs that are predominately eliminated via kidney.
The current PBPK modeling approach was further evaluated using midazolam which is extensively metabolized by CYP3A4. To reflect the change of CYP3A activity in third trimesters during pregnancy, the liver and gut metabolism, represented by hepatic CL and gut FPE, was increased by 30% according to previous clinical results. The prediction accuracy of the current model is similar to the PBPK model which integrates the time-dependent activity of CYP3A4. Since CYP3A4 plays the most important role in metabolism of drugs and xenobiotics, the current model can be used to anticipate the PK alteration and support drug labeling in many investigational drugs and postmarketing products for the pregnant populations. This also suggests that higher midazolam dose may be required in pregnant women to obtain the optimal pharmacological effect. In addition, alternation of activities in other metabolizing enzymes can be easily integrated in the pregnancy PBPK models using a similar approach. For example, it is known that the activities of CYP1A2 and CYP2C19 decreased by 30–65% (66) and 50% (67,68) whereas the activities of CYP2A6, CYP2D6, and UGT1A4 increased by 54% (69), 50% (70), and 200–300% (71) in late pregnant period, respectively.
The degree to which drugs are bound to the plasma proteins also influences their disposition. Albumin concentration is reported to fall by 20–30% during pregnancy, which may alter the drug concentration of albumin-bound compounds (15,29,72). Minor changes of fup have been reported for midazolam and digoxin (19). Metformin and emtricitabine are negligibly bound to plasma proteins in human. Overall, the influence of alternation of plasma protein during pregnancy has minimal effects to the PK behaviors of our modeled compounds. In addition, the increased total body water/fat creates a larger space where the drugs may distribute. Therefore, minor changes in free drug concentration are expected as a result of the compensation effects caused by more extensive distribution of unbound drugs during pregnancy. Overall, there is little practical importance of plasma protein binding in the pharmacological response in pregnant patients, although the total drug concentration may alter for some drugs such as phenytoin.
A simplified p-PBPK model is described, which yields good predictions for renally cleared and CYP3A4 hepatically cleared compounds evaluated (metformin, digoxin, midazolam, and emtricitabine) when compared to clinical observations. The model is readily implemented in GastroPlus and could be applied for drugs with similar clearance pathways at various stages of drug development to evaluate if altered pharmacokinetics may occur in a pregnant population based on healthy volunteer data.