Purpose of pregnancy PK studies

PK studies in pregnancy are essential to understand how physiological changes during pregnancy affect drug absorption, distribution, metabolism, and elimination. These changes can significantly impact drug concentrations and therapeutic efficacy, potentially putting both the mother and fetus at risk if medications are not appropriately dosed. 

Pregnancy PK study design 

Need for PK assessments during the second and third trimesters: 

The physiological changes of pregnancy (e.g. increased blood volume, altered body composition, variations in metabolic activity, delayed gastric emptying, increased cardiac output, resulting increased glomerular filtration rate, and decreased albumin concentration and protein binding) become more pronounced as the pregnancy progresses. As a result, drug absorption, distribution, metabolism, and elimination may not only differ between pregnant and non-pregnant women, but also between pregnant women in their second versus third trimester, necessitating specific dosing adjustments to ensure safety and efficacy for both the mother and fetus. 

PK sampling should ​ideally be done ​​​​​when the drug has reached steady state and ​give consideration to ​the trimester of pregnancy. Optimally, PK sampling would be done in both the second and third trimesters with sampling at 20–26 weeks and 30–38 weeks, respectively, to reduce variability as much as is possible. ​A proposed ​PK sampling schema should be evaluated with pretrial simulations that account for PK and PD variability, study design and the physiological changes of pregnancy.  

Within participant comparison between pregnancy and postpartum:  

When possible, longitudinal study designs with evaluation during pregnancy and postpartum (control) provide comparisons that reduce variability and therefore sample size. As TB treatment and prevention regimens become shorter, this may not be feasible. PK reassessment can occur as early as four weeks post-delivery as a comparator to pregnancy PK, as most physiologic factors affecting PK normalize by five weeks postpartum. However, even less than four weeks postpartum, PK can be informative and postpartum PK timing could be delayed to just before treatment cessation. Table 5.2 outlines the estimated time for these factors to return to normal post-delivery status. ​​​​It should be recognized, however, that postpartum physiology may not fully reflect the non-pregnant state, and the timeline for reversal of pregnancy-related changes may vary across physiological processes and drugs. 

Pregnancy PK modelling

Population pharmacokinetic (PopPK) modelling is a quantitative approach describing drug concentrations–time profiles while accounting for variability between and within individuals, and is particularly valuable in TB studies involving pregnant and lactating women where intensive sampling and large sample sizes are often not feasible. By pooling sparse PK data across multiple participants and studies with differing sampling schedules, dosing, and study designs, PopPK models can reliably estimate drug exposure and identify key covariates such as pregnancy-related physiological changes, body weight, renal function, or postpartum status, enabling clinically meaningful conclusions even in challenging settings. This is especially important for TB drugs with narrow therapeutic windows or long treatment durations, where inaccurate dosing may compromise efficacy and promote resistance (underdosing) or increase toxicity (overdosing).

For shorter TB regimens, it may not be feasible to obtain both pregnancy and postpartum data in the same individuals; in such cases, PopPK modelling allows integration of data from non-pregnant women (or, where appropriate, other adult populations) as external comparators, acknowledging that such designs may have reduced power but remain informative.

Collection of trimester-specific exposure data, accurate gestational age, and time postpartum, provides valuable information on the trajectory of PK changes across pregnancy and postpartum. For drugs with high protein binding, consideration should be given to measuring unbound drug concentrations, as pregnancy-related changes in protein binding may meaningfully affect pharmacologically active exposure.

Recent TB studies illustrate the utility of PopPK modelling:

1. ​​​Levofloxacin in pregnancy: analyses showed a ~38% increased clearance during pregnancy, resulting in substantially lower maternal drug exposures. Model-based simulations suggested that an additional 250 mg in pregnancy may be required to achieve exposure levels similar to non-pregnant adults. The safety and clinical impact of this proposed dosing requires further evaluation (21).
2. Bedaquiline in ​​pregnancy: Bedaquiline is characterised by extensive tissue distribution, an exceptionally long terminal half-life, and the presence of an active metabolite (M2), making traditional pharmacokinetic analyses difficult, especially in pregnant and lactating populations. Using popPK modelling, investigators were able to integrate sparse plasma samples collected across pregnancy, postpartum, and lactation, to jointly characterize bedaquiline and metabolite exposure despite limited sample sizes (22).
3. Clofazimine during pregnancy: Clofazimine has an exceptionally long half-life, is highly lipophilic, and accumulates in adipose tissue with reported sex-related ​​differences, making it difficult to ensure sampling truly reflects steady state and increasing the risk of misinterpreting apparent third trimester–postpartum differences when exposure is still evolving over time (23). Work carried out on clofazimine pregnancy noncompartmental analysis illustrates both need and opportunity (24). Third trimester–postpartum exposure differences observed with intensive clinical sampling raise concerns about potential underexposure during the third trimester and/or unquantified fetal/placental accumulation. PopPK modeling can integrate emerging in-vitro placental data and data from other studies with different design and sampling schedules to resolve these uncertainties more efficiently than designing a stand-alone intensive trial. This work is ongoing but reflects the importance of collecting data during larger clinical trials, even when sample sizes are low.

These case studies illustrate how PopPK modelling can reveal clinically important differences in drug exposure across pregnancy and lactation and support informed risk-benefit assessments of drug use during pregnancy and breastfeeding in the context of TB treatment. They also reflect how small amounts of data from various studies can be combined to inform important and challenging questions about drug dosing in pregnant and lactating women.  

Pregnancy PK study sample size 

In addition to supporting data analysis, PopPK models can be used prospectively to inform study design through simulation, help to assess feasible sample sizes, optimize sparse sampling schedules, and evaluate whether a study can address key pharmacokinetic questions, defined a priori, in pregnant and lactating populations. While not all studies will undertake formal simulation exercises, early consideration of model-based design approaches can improve study efficiency, reduce participant burden, and support discussions with ethics committees, regulators, and funders. 

In the context of TB pregnancy and lactation studies, prior pharmacometric experience has shown that clinically informative PopPK analyses can often be achieved with relatively modest sample sizes when model-based approaches are used. As an illustrative example, studies enrolling 10–30 pregnant or lactating participants, each contributing a small number of strategically timed samples, may be sufficient to characterize typical exposure, variability, and key covariate effects, and to explore clinically meaningful differences relative to non-pregnant adults (21, 22, 25). A well-defined non-pregnant comparator group is critical for reliable interpretation of pregnancy effects and should be as uniform as feasible (e.g. similar regimen, formulation, dosing conditions, disease state, etc.) to minimize confounding. 

When uncertainty exists regarding achievable sample size or expected effect size, a posteriori power or precision assessments can be valuable​​ (26). Here, a posteriori refers to analyses that are done after the data have been collected, using the final population pharmacokinetic model and the actual study design (e.g. the sample size and sampling schedule). Using stochastic simulation and estimation, the fitted model can be used to explore what size of pregnancy-related exposure difference the study would have been able to detect or rule out. This helps to interpret negative findings, by showing whether the study was likely able to identify a clinically meaningful effect if one had been present. In this way, the study can still be informative even if recruitment is lower than planned and can support decisions during study conduct.