Antiretroviral drugs are a critical tool in preventing mother-to-child transmission of HIV. Yet, antiviral treatment options for pregnant women lag behind the “non-pregnant” population. In this blog post, I’ll discuss the reasons for this lag and how physiologically-based pharmacokinetic (PBPK) models can simulate PK during pregnancy and thus help optimize dosing for this special population.
Preventing mother-to-child HIV transmission
In 2016, approximately 1.4 million HIV-infected women in the world gave birth. Without the intervention of antiretrovirals, the risk of HIV transmission from mother to child is 15-40%. With antiviral treatment, the transmission risk is reduced to less than 1%. Thus, we can reduce the possible new infections from 560,000 to less than 14,000—a great achievement.
Unfortunately, the treatment options for pregnant HIV+ women lag behind the treatment options for non-pregnant women or men. For example, dolutegravir, elvitegravir, and TAF—routine treatments for non-pregnant HIV+ patients—are either not available or are not preferred treatment options for pregnant women.
Why do treatment options lag behind for pregnant women?
Pregnant women are usually excluded from clinical trials of investigational drugs because the risk posed to the unborn child is unknown. We do gain some insight into how these drugs perform during pregnancy after they are approved as the pharmaceutical companies maintain pharmacovigilance registries of pregnancy cases and outcomes. Also after drug approval, academic groups may perform pharmacokinetic studies in pregnant women.
Thus, there is a delay between FDA approval of a drug and data on its impact on pregnancy becoming available. For most of the older antiretroviral drugs, pregnancy information became available within two years after FDA approval. But the lag is much longer for newer treatments—six to eight years. And we have no data for the newest marketed compounds—dolutegravir, elvitegravir, cobicistat, and TAF.
Understanding how drugs perform in pregnancy is crucial to optimizing maternal care. Despite antiretroviral therapy, approximately 13% of pregnant women still have a detectable viral load at the time of delivery. Adequate maternal exposure to antiretroviral drugs is necessary for maximal reduction of viral loads and reduced risk of transmission.
Physiological changes during pregnancy can influence pharmacokinetics
An array of physiological changes that occur during pregnancy can affect the absorption, distribution, metabolism, and excretion (ADME) of drugs. Regarding drug absorption, gastric pH, gastric emptying, and intestinal motility change during pregnancy. As a pregnant woman gains weight, her volume of distribution increases as well. For metabolism, hepatic blood flow, some CYP enzymes’ activity, and renal excretion also increase.
The overall effect of pregnancy on drug exposure depends on the drug being administered. For darunavir, its concentrations during pregnancy are decreased. By contrast, etravirine concentrations are higher during pregnancy. The mechanism behind these changes is likely because CYP2C19, which metabolizes etravirine, is inhibited during pregnancy whereas CYP3A4, the main enzyme metabolizing darunavir, is induced.
Characterizing antiretroviral drug PK in pregnant women
To investigate the Pharmacokinetics of newly developed ANtiretroviral agents in HIV-infected pregNAnt women, my colleagues and I established PANNA, a European clinical pharmacology network. It’s a general study protocol for investigating over 18 antiretroviral drugs. Pregnant HIV+ women who are using at least one of these drugs can participate in the study.
The PANNA study protocol involves the following steps: We collect a full pharmacokinetic curve during the third trimester of pregnancy and again 4–6 weeks postpartum. We derive pharmacokinetic parameters from these curves using Phoenix WinNonlin and make an intrasubject comparison with the postpartum curve as the control curve. At delivery, we also try to obtain a cord blood sample to assess whether these drugs cross the placenta. The PANNA study is currently running in 25 hospitals in seven European countries. Unfortunately, it’s fairly burdensome for the patients to spend an entire day at the hospital to develop a PK curve.
To make it easier to study these pharmacokinetic changes, we next developed the “PIANO” project: Pharmacokinetic Investigations of Antiretroviral agents in HIV-infected pregNant wOmen. The aims of the PIANO study are:
- To develop a PBPK model that simulates maternal pharmacokinetics in pregnancy
- To support better dose predictions for this patient population and anticipate the effects of drug interactions and co-morbidities on exposure
- To identify knowledge gaps that limit the accuracy of PBPK modeling
The PIANO approach
To develop the PBPK model, we used the Simcyp Simulator 13.2 pregnancy model. In vitro ADME parameters for input into the PBPK models were determined experimentally or were based on literature. Model predictions were validated with the results from the PANNA study.
We chose to model darunavir (DRV) PK as it is a preferred antiretroviral for use in pregnancy. DRV is a CYP3A4 substrate, and the antiretroviral ritonavir (RTV) inhibits CYP3A4. Thus, co-administration of DRV and RTV increases the DRV concentration. For this reason, DRV is always combined with RTV as a booster. So, we had to model the interaction of these two drugs in addition to the effect of pregnancy.
Building the model
The first step was to develop a PBPK model for a single dose of un-boosted DRV. Clinical data were available to validate this model. Then, we included RTV as a booster to the model and simulated the interaction between a single dose of DRV + RTV. Next, we simulated the steady state exposure of DRV + RTV in a non-pregnant population before performing this simulation in the pregnant population. Lastly, we simulated some dose adaptations in a virtual pregnant population.
When we simulated a single dose of darunavir with our first model, we over-estimated the exposure. What could be the reason for this? This model didn’t include transporters. Both uptake and efflux transporters play a role in darunavir pharmacokinetics. But, we didn’t have quantitative data available that described these darunavir transporters like KM, Vmax, or intrinsic clearance. When we included transporters in the model, its predictions were much closer to observed data.
The next step was to include ritonavir in the model. Then we performed simulations for both dosing regimens for darunavir: 600/100mg DRV/RTV BID and 800/100mg DRV/RTV QD and calculated DRV’s Cmax and AUC for both regimens. This model had reasonably accurate predictions; DRV PK parameters were within a twofold difference from observed data.
Finally, we modeled DRV + RTV PK in pregnant patients in their third trimester as well as postpartum. We again simulated both dosing regimens. For the twice daily dosing regimen, the model showed that DRV exposure was lower in the third trimester of pregnancy than postpartum. The model’s predictions were robust for both the third trimester and postpartum. The same was true for the once daily DRV+RTV regimen. The model’s fit was not as good but was still within a twofold difference from observed data. Again, the model predicted lower DRV exposure during late pregnancy compared to postpartum.
Take home points
As with all models, our model has some uncertainties and limitations. The role of hepatic uptake, efflux transporter intrinsic clearance, and intestinal transporters in DRV PK have yet to be determined. The ritonavir model was a semi-mechanistic model. For darunavir, its absorption was not fully mechanistic, but rather employed a top down approach. Importantly, while our model assessed maternal exposure, it could not assess fetal exposure.
To summarize, our data supported a clinically relevant role for hepatic transporters in darunavir pharmacokinetics. In addition, the described model could approximate boosting by ritonavir and the decrease in maternal darunavir exposure observed during pregnancy.
To improve the mechanistic basis of the model, future studies should address hepatic and intestinal transporter-mediated darunavir disposition in greater detail. This development of the model has been published in Clinical Pharmacokinetics.
To learn more about this project, please watch this webinar I presented with my Radboud University Medical Center colleague, Dr. Rick Greupink.