Using PBPK Models to Assess Fetal Drug Exposure

Using PBPK Models to Assess Fetal Drug Exposure

The tragedy of thalidomide provides a cautionary tale about the potential for birth defects resulting from fetal exposure to drugs. Thalidomide was used to treat morning sickness in pregnant women. By the time it was banned in 1962, more than 10,000 children had been born with thalidomide-induced birth defects. “Phocomelia,” wherein babies were born with malformations of the limbs, was the most notorious consequence of in utero exposure to this drug.

But, fetal exposure to drugs can also confer important therapeutic benefits. For instance, HIV-infected pregnant women are treated with antiretroviral agents to prevent mother-to-child viral transmission. Adequate fetal drug exposure might contribute to the efficacy of the antiviral treatment. In this blog post, I’ll discuss a novel approach to predict fetal exposure following maternal dosing and assess whether this exposure contributes to drug efficacy.

Quantifying fetal drug exposure

To assess the impact of fetal drug exposure, you need to link the dose administered to the mother to fetal exposure (eg, Cmin, CMax or AUC) and the fetal response (eg, the EC90 for inhibition of viral replication).

Unfortunately, the fetus is not easily accessible. Thus, quantifying fetal drug exposure is challenging.

Fetal drug exposure at term can be assessed to some extent by measuring the umbilical cord-to-maternal (C:M) drug plasma concentration ratio. The problem with this metric is that it is a static parameter. Therefore, the C:M concentration ratio depends on when you collect the samples during the dosing interval.

If you sample early after dosing, you might get a C:M ratio of zero because the maternal plasma concentration is very high while the cord concentration is almost nonexistent because of the time delay of transferring drugs to the fetus.

When you sample during the middle of the dosing interval, the C:M ratio might equal 1. But if you sample at the end of the dosing interval, the C:M ratio might approach infinity as maternal plasma concentrations are approaching zero and there is still measurable cord plasma concentrations.

So only measuring cord to maternal plasma concentration ratios might not provide a good estimate of overall fetal exposure.

Predicting fetal drug exposure: A case study using antivirals

PBPK modeling based on ex vivo placenta perfusion data may enable better estimation of fetal exposure throughout the dosing interval. These models can then be used to assess if the predicted fetal exposure may contribute to drug efficacy.

In this case, we used a PBPK approach to assess the impact of fetal drug exposure on antiviral activity. We started with a previously developed darunavir/ritonavir maternal PBPK model. But now we included the placental transfer of darunavir in the model. We used an in vitro input to parametrize such a PBPK model. This required assessing the degree to which drugs cross the human placenta.

The human placental barrier

The placenta is comprised of functional units called cotyledons. Maternal blood flows into the placenta where they empty into the intervillous spaces. Then, the blood flows along the placental villi, which contain the fetal blood vessels. In this way, fetal and maternal blood come into close contact with each other although they are separated by some cell layers.

For drugs to get from the maternal space to the fetal blood vessels, they have to pass the syncytiotrophoblast cell barrier. P-glycoprotein is a transport protein present in the syncytiotrophoblasts. Thus, the syncytiotrophoblasts don’t just prevent passive passage of large molecules or very hydrophilic molecules across the placenta. The barrier also includes transport proteins that can efflux drugs that initially diffused into the syncytiotrophoblasts and transport them back into the maternal blood. This is how the placenta protects the fetus.

Placental perfusion

We used placental perfusion experiments to attain input data to parameterize the PBPK model. In brief, this is the experimental design. First, we rushed the placentas from the hospital delivery rooms to the laboratory. Then, we reestablished the fetal and maternal circulations and perfused one cotyledon of the placenta with oxygenated Kreb’s Henseleit buffer at physiological temperature and fetal/maternal flow rates. This process keeps the tissue alive so you can perfuse these placental cotyledons for approximately three to six hours.



To assess the transfer of darunavir from the maternal to the fetal circulation (M->F) and back again (F->M), we did two sets of perfusions. The compartment to which the drug is added recirculates, but the flow on the other side empties into a beaker to allow sampling. These conditions allowed us to establish intrinsic clearance values from (M->F) and (F->M).

Thus, we used this recirculating system to obtain data to study fetal drug disposition. We added darunavir to the maternal circulation and measured it disappearing from the maternal reservoir. From the disappearance curve, we estimated (M->F) and (F->M) intrinsic clearance values for one cotyledon.

Scaling cotyledon clearance

To make PBPK predictions, we needed to scale intrinsic clearance from one cotyledon to clearance of a complete placenta. We scaled it based on the average number of cotyledons present in the placenta—approximately 30. Scaling could also be based on placental volume or weight. Then we corrected for protein binding as protein is present in our perfusion system. Thus, we also determined placental inbound clearance and entered it into our PBPK model.

Maternal and fetal PBPK model structure

The PBPK model that represented a pregnant woman in the third trimester contained both maternal and fetal features. The maternal model included a number of organs and tissues: brain, heart, kidney, muscle, skin, liver, lung, adipose, and bone.

The placental disposition model included the clearance value which describes flux from the maternal arterial blood to the fetal blood and the clearance value that described flux back from the fetal blood to the maternal venous blood.

There’s also some disposition included from fetal blood to the rest of the fetal body and also rate constants that described transport from the fetal blood to the amniotic fluid and back again. When the fetus swallows amniotic fluid, this recirculates the drug from the fetal blood to the amniotic fluid. We did not include fetal clearance data other than passage across the placenta. Thus, no hepatic clearance was included in the model.

Predicted maternal and fetal plasma concentrations

Using this model, we simulated two dosing regimens: the 600/100 milligram darunavir/ritonavir twice daily dosing regimen and the 800/100 milligram darunavir/ritonavir regimen once daily dosing regimen. Our predictions of both maternal and fetal exposure were reasonably close to clinical data from maternal and fetal cord blood samples.

Fetal drug exposure in relation to antiviral response

Now that we confirmed that our model could predict maternal and fetal drug exposure, we wanted to know whether it was possible to link exposure to viral drug sensitivity.

We performed simulations of different dose levels of darunavir administered either once or twice daily. Then we compared the fetal drug exposure relative to darunavir’s EC50. The standard dosing regimen for both once and twice daily darunavir yielded a fetal exposure that was above darunavir’s EC50. These data suggested that fetal exposure might contribute to viral inhibition and thus, aid preventing viral mother-to-child transmission.

Model limitations and takeaways

Like any model, our fetal model had limitations. We managed to simulate fetal exposure, but we were restricted to late-stage pregnancy. Likewise, no parameter variability was included, so we simulated mean values. For our estimation of placental clearance, we used term placentas not preterm placentas. Finally, the maternal disposition model was based on the third trimester of pregnancy. Ideally, we would be able extrapolate to earlier stages of pregnancy. The model could be enhanced by including more complex placental and fetal models that describe transporters, tissue composition, and CYP activity in the fetal liver.

The first pregnancy PK data generally doesn’t become available until after market approval of a drug. By combining in vitro placental drug perfusion studies with modeling and simulation techniques before regulatory approval, we can better inform dose optimization of maternal pharmacotherapy and, potentially, fetal drug treatments.

To learn more about this project, please watch this webinar I presented with my Radboud University Medical Center colleague, Dr. Angela Colbers.

Rick Greupink

About the Author

Rick Greupink

Assistant Professor of Pharmacology, Radboud University Medical Center

More Posts by This Author

Dr. Rick Greupink holds a PharmD from the University of Groningen, the Netherlands, obtained a PhD in pharmacokinetics and drug delivery from the same university, and further specialized as a pharmacologist during postdoctoral fellowships in both clinical and pre-clinical settings in pharmaceutical industry and academia. At the Radboud University Medical Center, Dr. Greupink is faculty in the department of Pharmacology & Toxicology, where his research focuses on investigating the pharmacological roles of drug-transporting membrane proteins in barrier and excretory tissues. The aim is to better understand and predict the impact of drug transporters on clinical pharmacokinetics, drug efficacy, and drug-induced toxicity. In this context, current projects include mechanistic studies on drug disposition in pregnancy, placental transfer, accumulation and effects of small and large molecule pharmaceuticals.