Pregnancy‐related effect on elexacaftor, tezacaftor and ivacaftor pharmacokinetics in women with cystic fibrosis

Abstract

Aim

The number of pregnancies among women with cystic fibrosis (wwCF) has steadily increased over the past decade. However, the pharmacokinetics (PK) of elexacaftor–tezacaftor–ivacaftor (ETI) during gestation remains uncharacterized, despite its widespread use in this population. To date, no clinical trials have investigated the effectiveness, safety or PK profiles of ETI during pregnancy. Given the significant physiological changes that occur throughout gestation, which can alter drug PK, characterizing how pregnancy impacts ETI exposure represents a key step towards informing appropriate dosing strategies. Therefore, the aim of this analysis was to assess pregnancy effect on ETI PK.

Method

As part of the French national observational cohort study, therapeutic drug monitoring was performed in 145 wwCF, including 14 pregnant participants. PK profiles for each compound were characterized using the previously published ETI adult population PK model from the same cohort, with model development and evaluation performed using Monolix® software.

Results

A total of 294 plasma samples, including 47 collected from pregnant participants, were drawn during routine clinical care. Pregnancy was found to significantly increase the apparent clearance of tezacaftor (0.97 to 1.44 L/h) and of ivacaftor (9.64 to 11.36 L/h), indicating a decrease in systemic exposure.

Conclusion

These findings raise important questions regarding whether decreased total tezacaftor and ivacaftor exposure may compromise the sustained treatment efficacy. The necessity for dose adjustments during pregnancy remains to be evaluated from a PK (unbound ETI concentrations) and pharmacodynamic (clinical effectiveness) perspective, thus warranting further studies.

What is already known about this subject

• Pregnancies among wwCF are rising in the era of highly effective CFTR modulators,

• No sponsor‐led clinical trials have yet disclosed data regarding ETI clinical safety, efficacy or pharmacokinetics (PK) during pregnancy,

• Pregnancy‐related physiological changes may alter ETI exposure, highlighting the need to study its effect on ETI PK for better dosing strategies and treatment outcomes.

What this study adds

• Significant impact of pregnancy on the apparent clearance of TEZ and IVA,

• This lowered systemic exposure to TEZ and IVA in pregnant wwCF does not necessarily imply an increase in dose but rather calls for personalized dose management in certain patients

• Clinical follow‐up throughout pregnancy along therapeutic drug monitoring could be a valuable strategy to guide individualized care

1. INTRODUCTION

Cystic fibrosis (CF) is a life‐limiting, multisystem genetic disorder caused by variants in the CF transmembrane conductance regulator (CFTR) gene. Defective CFTR protein disrupt chloride and bicarbonate transport across epithelial cells, leading to thickened secretions in various organs. CF is mainly characterized by mucus plugging, chronic bacterial respiratory infections and inflammation, resulting in airway damage and accelerated loss in lung function, which may lead to respiratory failure. ¹ , ² In addition to pulmonary complications, clinical manifestations include maldigestion and malabsorption of nutrients, with exocrine pancreatic insufficiency (EPI, affecting ~80% of patients) and CF related diabetes (~20% of adults).

Infertility is observed in the vast majority of males and in 35%–50% of females. ² In women with CF (wwCF), reduced fertility is multifactorial. ³ Thickened cervical mucus, acidic uterine environment, impaired gonadotropin secretion and anovulatory cycles—particularly in the context of poor nutritional status and reduced lung function—are all contributing factors linked to subfertility in wwCF. ² , ³ Despite these challenges, the introduction and widespread use of CFTR modulators since 2012 has transformed the clinical course of CF and reproductive health outcomes. The most effective CFTR modulators to date are used in triple combination therapy, known as elexacaftor–tezacaftor–ivacaftor (ETI). These therapies have not only improved respiratory health and nutritional outcomes but have also substantially extended life expectancy. ⁴ The 25‐year increase in survival over a decade has shifted the focus of care towards long‐term quality of life, including reproductive health and family planning. ⁴

With ETI now established as the current standard of care, improved overall health and reduced infertility have led to a growing number of wwCF actively pursuing pregnancy, as they look towards possible longer and healthier lives. Concurrently, spontaneous conceptions are increasingly reported. ³ In the United States, the number of pregnancies per year among wwCF has continuously increased, from 310 in 2019 to 655 in 2024. ⁴ A similar trend has been observed in France, with 58 pregnancies reported in 2019 to 128 in 2022. ⁵ As a result, ETI is now commonly used during pregnancy, despite the absence of evidence‐based dosing guidelines for this population. In light of the evolving epidemiological landscape of the disease, these changes have prompted a growing interest in the safety of ETI during gestation. ⁶ , ⁷ , ⁸

Despite their widespread use during pregnancy, the pharmacokinetics (PK) of ETI remain poorly characterized in pregnant patients. While ETI is generally considered safe during gestation, no sponsor‐led clinical trials have yet disclosed its clinical efficacy, safety or appropriate dosing in this specific population.

Shad et al. recently reported maternal ETI plasma concentrations from pregnant wwCF, receiving reduced dosing due to adverse events such as depression or elevated bilirubin. ⁶ Although overall maternal health remained stable and no complications were reported, one woman who discontinued ETI after conception, experienced a decline in clinical respiratory status and ultimately resumed treatment. ⁶ These findings point to two important clinical dilemmas: first, discontinuation of ETI due to fetal safety concerns may compromise maternal health ⁶ ; second, dose adjustments are often made empirically, in the absence of clear guidelines or support of PK data. This is particularly relevant given the known risks of poorly controlled CF during pregnancy for both mother and fetus, including accelerated deterioration in respiratory function, increased susceptibility to respiratory infection with increased systemic inflammation, poor gestational weight gain, which may cause adverse pregnancy outcomes, and neonatal complications. This consideration is further heightened when the fetus is diagnosed with CF, ⁹ as it raises the question of whether higher maternal dosing might be required to achieve therapeutic efficacy in utero. Optimized maternal disease control may therefore confer benefits not only to maternal health but also potentially to fetal outcomes, both indirectly and in the context of fetuses affected by CF. Indeed, the efficacy or safety of any dose adjustment (dose reduction or dose increase) strategy is currently uncertain, due to lack of PK and pharmacodynamic evidence.

This is consistent with findings from Taylor‐Cousar et al., who reported clinical deterioration of five of six women who discontinued ETI out of concern for fetal risk—highlighting the pressing need for clearer guidance on therapeutic decision‐making during pregnancy. ⁷

Importantly, toxicity studies of ivacaftor in rats reported the occurrence of infant cataracts. ¹⁰ Similarly, Jain et al. described three cases of infants who developed bilateral congenital cataracts within 6 months of birth after in utero and breastfeeding exposure to ETI, raising concerns regarding rare but serious adverse effects. ¹¹

Gestation induces profound physiological and anatomical changes, including increased maternal bodyweight and blood volume, elevated cardiac output, decreased plasma protein concentrations and upregulated hepatic enzyme activity, which can substantially impact drug PK, by altering absorption, distribution, metabolism and elimination. Indeed, reduced protein concentrations can increase the unbound fraction and distribution of highly protein‐bound drugs, and increase in hepatic enzyme activity or enhanced renal clearance may accelerate drug metabolism and elimination. ¹² Therefore, understanding how these changes affect ETI exposure is critical to ensure both maternal health and therapeutic efficacy, as well as minimizing potential risks to the mother and the fetus. To address this knowledge gap, we conducted a population pharmacokinetic (popPK) study using maternal plasma samples to estimate the effect of pregnancy on ETI PK.

2. METHODS

2.1. Study design

The present analysis used observational data from adult wwCF of childbearing age (18–46 years), including both pregnant and non‐pregnant individuals who were enrolled in the French national, prospective, multi‐centre observational cohort study. ¹³ Participants were followed across one of the 47 CF French reference centres. Data and samples were collected between 1 January 2022 and 30 June 2025, as part of routine clinical care, from women treated with the standard ETI dosing regimen, which consisted of two tablets of ETI (100/50/75 mg) in the morning, and one tablet of ivacaftor (IVA, 150 mg) in the evening. ¹⁴ , ¹⁵

All participants were at steady state with respect to ETI therapy. Patients were instructed to take the medication with a high‐fat meal, in accordance with Food and Drug Administration (FDA) and European Medicines Agency (EMA) guidelines. ¹⁴ , ¹⁵ Blood samples were collected during routine therapeutic drug monitoring (TDM) and assayed at the Pharmacology Department of Cochin Hospital (Paris, France). The date and time of both drug intake and blood sampling were recorded by trained nurses using a standardized TDM request form. Clinical and demographic variables—including height, age, sex at birth, CF‐related diabetes status, pancreatic insufficiency and CFTR genotype—were collected at baseline. Bodyweight was recorded at the time of sampling. Body mass index (BMI) was calculated as weight (kg) divided by height (m²).

CFTR genotypes were grouped to reflect ETI responsiveness on the basis of recent studies. ¹⁶ , ¹⁷ Indeed, variants known to be responsive to ETI, included F508del, as well as selected non‐F508del variants. ¹⁶ , ¹⁷ Therefore, patients were characterized as carriers of either one or two ETI‐responsive variants. Patients carrying an ultra‐rare variant with unknown ETI responsiveness were grouped separately.

Each sample was classified into a trimester according to the gestational age at the time of collection: samples collected between gestational weeks 1–13 were assigned to the first trimester, weeks 14–27 to the second and week 28 onward to the third.

Percent predicted forced expiratory volume in 1 s (ppFEV₁) measurements were collected at multiple time points: prior to pregnancy, during pregnancy and after pregnancy. For one participant, the pre‐pregnancy ppFEV1 value with ETI treatment was not available, and for another, the post‐pregnancy assessment had not yet been performed. The Global Lung Initiative Equations were used to determine the ppFEV1 values. Further details regarding the cohort and study design are available in previously published results. ¹³ , ¹⁶ , ¹⁷

2.2. Ethics

The cohort study was approved by the Institutional Review Board of the French Society for Respiratory Medicine (Société de Pneumologie de Langue Française) (2020‐003). In accordance with French regulations, written informed consent was not required for this study. However, patients were provided with written information outlining the study's objectives and the intended use of their data prior to inclusion.

2.3. Quantification method

Plasma concentrations of ETI were quantified using our laboratory‐developed liquid chromatography coupled with tandem spectrometry (LC–MS/MS) method. ¹⁸ The technique was validated according to EMA guidelines in the range of 0.075 to 20 mg/L for elexacaftor (ELX) and tezacaftor (TEZ), of 0.0525 to 14 mg/L for IVA. The analytical method underwent full validation and demonstrated satisfactory precision, with inter‐ and intra‐day coefficients of variation below 14.3%, and good accuracy, with inter‐ and intra‐day bias ranging from −13.7% to 14.7% for all analytes, confirming the reliability of the concentration measurements used in the pharmacokinetic analyses.

Proteins were precipitated from 50 μL of human plasma using acetonitrile containing internal standards. After centrifugation, the supernatant was directly injected into the LC–MS/MS system. Chromatographic separation was performed on a Waters Acquity UPLC® BEH C18 column (2.1 × 50 mm, 1.7‐μm particle size) maintained at 40°C. The mobile phase consisted of water with 0.05% formic acid (A) and acetonitrile with 0.05% formic acid (B), using a gradient elution at a flow rate of 0.5 mL/min. Detection was carried out on a Xevo TQD triple‐quadrupole‐mass‐spectrometer (Waters®, Milford, USA) using Heated ElectroSpray ionization in positive mode and multiple reaction monitoring (MRM) for each compound. The following transitions for quantification were monitored: m/z 521.3 > 103.1 for TEZ, m/z 598.3 > 97.1 for ELX and m/z 410.5 > 172.1 for IVA‐D18.

2.4. Pharmacokinetic analysis

A popPK approach was chosen for its ability to characterize drug exposure using real‐world clinical data, accommodate sparse sampling and capture true inter‐individual variability (IIV), thereby enabling robust estimation of the magnitude of pregnancy‐related changes of drug PK. The popPK models were analysed using non‐linear mixed‐effects modelling software Monolix® (Version 2021R2, https://lixoft.com/products/monolix/). Monolix® estimates PK parameters using the stochastic approximation expectation maximization (SAEM) algorithm, combined with Markov Chain Monte Carlo (MCMC) methods to account for complex variability in the data. As a foundation for model development and characterizing ETI PK profiles in wwCF, we used our previously published ETI popPK model in adults with CF from the same cohort. ¹⁹

To apply this framework to our cohort of pregnant and non‐pregnant wwCF, we retained the same structural models as the princeps models and implemented Bayesian priors on all previously estimated parameters. Previously set parameters in the princeps model were also set to the same values here. ¹⁹ This approach ensured model stability using prior PK knowledge. ¹⁹ When a parameter could not be reliably estimated from our data—indicated by relative standard error (RSE%) > 50%—the parameter was fixed to its corresponding value from the princeps model. ¹⁹ This is expected when applying a previously developed population pharmacokinetic model to a small dataset, where the available information may be insufficient to precisely re‐estimate all structural parameters. In exploratory runs where these parameters were estimated freely, the resulting estimates remained consistent with those reported in the princeps model but were associated with poor precision, supporting the decision to retain the established values. Within the nonlinear mixed‐effects modelling framework, fixing poorly informed parameters primarily reduces uncertainty in those components and does not materially influence the estimation of other parameters that are supported by the observed data.

Different residual variability models, additive, proportional or combined, were investigated. An exponential model was used for IIV.

The influence of pregnancy on ETI PK parameters was investigated using two different approaches: (i) first as a binary covariate, with the following equation using apparent clearance (CL), for example,

$$\log \left(\mathrm{CL}\right)=\log \left({\theta }_{\mathrm{CL}}\right)+\text{PREG}\times {\beta }_{\text{PREG}}^{\text{logCL}}$$

where ${\theta }_{\mathrm{CL}}$ is the typical clearance value for a typical patient, PREG equals 1 when pregnant and 0 otherwise and ${\beta }_{\text{PREG}}^{\text{logCL}}$ is the estimated effect of pregnancy on the log‐clearance, and (ii) for an exploratory purpose by evaluating a ‘trimester’ effect of pregnancy on clearance as a categorial covariate: first (T1), second (T2) and third trimester (T3).

Genotype was previously described in the ELX princeps model as a categorial covariate: carrier of two ETI‐responsive variants or one ETI responsive variant.

The number of iterations was set to the default values, with 500 iterations for the exploratory phase and 200 for the smoothing phase. Data below the lower limit of quantification (LLOQ) was managed using left censoring to ensure continuous data measurements. The LOQ values for ELX, TEZ and IVA are 0.075, 0.053 and 0.05 mg/L, respectively.

Individual PK parameters were then derived from the ETI models for each wwCF: maximum plasma concentration (C _max), minimum concentration (C _min) measured at 24 h after drug intake for ELX and TEZ, and at 12 h for IVA; and the area under the concentration‐time curve (AUC_0–24h for ELX and TEZ, AUC_0–12h for IVA).

2.5. Model development and evaluation

The residual variability models and the addition of a covariate to the structural model were compared using changes in the objective function value (OFV), requiring a significant drop of at least 3.84 points (i.e., following a chi‐squared distribution with one degree of freedom, p < 0.05, allowing statistical inference on covariate inclusion). The inclusion of a covariate was further examined by the reduction in the IIV of the corresponding PK parameter. Goodness‐of‐fit and quality of the models were based on SAEM algorithm convergence, visual assessment of observed‐predicted concentration scatter plots, corrected visual predictive checks (pcVPC) and normalized prediction distribution error (NPDE) metrics. These criteria guided the selection of models which provided model accuracy and parsimony.

2.6. ETI therapeutic efficacy target concentrations

Individual total plasma concentrations from our cohort were graphically compared to the reported half maximal effective concentration (EC₅₀) values of each molecule. The clinical efficacy endpoint data for TEZ and IVA (in combination therapy without ELX) were established upon reduction in sweat chloride levels. ¹⁴ , ²⁰ The EC₅₀ value for ELX (in combination with TEZ and IVA) was estimated, using a graphical method with the aid of Plotdigitizer; from Tsai et al. ²⁰ The EC₅₀ value was determined based on clinical trial efficacy endpoint data, described as sweat chloride reduction or an improvement in ppFEV_1. ²⁰

2.7. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in https://www.guidetopharmacology.org/ and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/22. ²¹

3. RESULTS

3.1. Population

The analysis included 145 adult (≥18 years) wwCF receiving ETI therapy, of whom 14 were pregnant at multiple sampling times.

All pregnancies followed in the cohort study progressed without complications. No adverse events were reported by clinicians that could be directly attributed to in utero ETI exposure. None of the women discontinued ETI treatment during gestation, and both maternal and neonatal outcomes were favourable, with no health concerns raised for either mothers or infants. Importantly, all infants underwent ophthalmologic evaluation at 9 months post‐delivery. No ocular abnormalities, including congenital cataracts previously flagged as a potential safety signal in post‐marketing surveillance, were identified.

The median (range) age of all participants was 28 (18–46) years, with a median (range) bodyweight of 56 (35–100) kg. The cohort's distribution of EPI and CF‐related diabetes were of approximately 85% and 15%–20% respectively, in both pregnant and non‐pregnant groups. This distribution is fairly consistent with the prevalence of CF‐related comorbidities in the general CF adult population. Participants who were carriers of two ETI‐responsive variants constituted between 70% and 85% of the female population. Samples were collected during gestation across all trimesters (n = 9 at T1, n = 20 at T2 and n = 18 at T3) at a median gestational age of 25.9 weeks (ranging from 7.6–39.7 weeks). The ppFEV₁ measurements remained stable before, during and after gestation, with values consistently above 60%, except for two pregnancies. A comprehensive overview of the baseline characteristics is provided in Table 1 and evolution of ppFEV1 measurements throughout 15 pregnancies are shown in Figure 1.

TABLE 1.

Population demographics and clinical characteristics.

Population characteristics	Pregnant patients (n = 14)	All patients (n = 145)
Age (years)	30.0 [27.4–32.8] (23.0–46.0)	28.0 [24.0–34.0] (18.0–45.0)
Bodyweight (kg)	63.0 [60.0–71.4] (51.0–80.0)	56.0 [51.9–62.0] (35.0–100.0)
Height (cm)	163.0 [159.5–164.0] (148.0–173.0)	161.0 [158.0–165.0] (142.0–178.0)
BMI (kg/m2)	24.2 [22.6–27.5] (19.0–29.7)	21.5 [20.2–23.2] (14.4–37.0)
CFTR genotype
Carrier of two ETI‐responsive variant	12 (85.7%)	107 (73.8%)
Carrier of one ETI‐responsive variant	1 (7.1%)	31 (21.4%)
Carrier of ultra‐rare variants	1 (7.1%)	7 (4.8%)
Exocrine pancreatic insufficiency	12 (85.7%)	124 (85.5%)
CF‐related diabetes	2 (14.3%)	34 (23.4%)
Gestational age at sampling	25.9 [17.8–34.9] (7.6–39.7)	NA
ppFEV1 pre‐pregnancy	91.5 [76.3–96.5] (44.0–116.0)	NA
ppFEV1 during pregnancy	90.0 [67.5–101.0] (46.0–146.0)	NA
ppFEV1 post‐pregnancy	95.5 [89.8–102.0] (47.0–119.0)	NA

Note: Data are presented as median [interquartile range, IQR] (min–max) or n (%). All population characteristics in the table are reported at the first sampling occasion, with the exception of gestational age and ppFEV1, which is described as the median [interquartile range (IQR)] across all collected data at multiple different routine visits.

Abbreviations: BMI, body mass index; CF, cystic fibrosis; CFTR, cystic fibrosis transmembrane conductance regulator; NA, not applicable; ppFEV1: percent predicted forced expiratory volume in 1 s.

FIGURE 1.

Evolution of ppFEV1 before, throughout and after pregnancy in 14 women with cystic fibrosis. Patient 9 had two pregnancies during the study's observational window. The dots represent ppFEV1 values from patients and the connective lines and colour indicate observations from the same individual. The grey shaded area outlines the gestational window (0–40 weeks). Abbreviation: ppFEV₁, percent predicted forced expiratory volume in 1 s.

3.2. ETI concentrations

A total of 294 plasma samples, including 47 collected from 15 different pregnancies, were drawn during routine care. There were respectively four ELX, two TEZ and seven IVA concentrations, which were below the LLOQ. The median [interquartile range; IQR] sampling time was 5.7 [3.4–23.3] h after drug intake for ELX and TEZ, and 5.61 [3.27–11.16] h for IVA. Overall, observed ETI plasma concentrations fell within the expected adult exposure ranges, ¹⁴ , ¹⁵ though data exhibited broad IIV in exposure, as anticipated for these compounds and consistent with our previous findings. ¹⁹ It is noteworthy that over 97% of total plasma ETI concentrations remained well above their respective EC₅₀ values, suggesting sustained efficacy despite this variability. The observed ETI concentrations over time following drug administration are depicted in Figure 2. The observed data are overlaid with their respective expected adult exposure ranges and EC₅₀ values derived from Tsai et al. and regulatory data. ¹⁴ , ²⁰

FIGURE 2.

Observed elexacaftor, tezacaftor and ivacaftor concentrations (mg/L) vs. time after drug intake (h) from real‐world adult pregnant vs. non‐pregnant women with cystic fibrosis (wwCF) treated with elexacaftor–tezacaftor–ivacaftor (ETI). The black dots represent drug concentrations from non‐pregnant wwCF, and the blue dots are from pregnant wwCF. The blue shaded areas represent the 5th–95th percentiles of adult elexacaftor, tezacaftor and ivacaftor expected exposure, respectively, and the purple dashed lines indicate the reported half maximal effective concentration (EC50) value for each compound (related to average concentrations for elexacaftor and tezacaftor and trough concentrations for ivacaftor) derived from published data. ¹⁴ , ²⁰

3.3. Population pharmacokinetic modelling

Based on previous ETI PK knowledge, a one compartment model with first order absorption and elimination best described the women's data. The lag time to absorption (T _lag) was set to 2.17 h for ELX, as previously reported. ²⁰ Bodyweight was allometrically scaled in all three models with exponents 0.75 and 1 for the apparent clearance and volume of distribution, respectively. ²² In the adult reference model, several significant covariates were identified and retained in the current analysis. ¹⁹ Age was associated with statistically significant reduction of ELX and IVA apparent clearance. Additionally, CFTR genotype, categorized according to the number of ETI‐responsive variants, was included with a19% decrease on ELX apparent clearance in participants carrying only one ETI‐responsive variant. Finally, EPI was found to significantly reduce bioavailability of TEZ, with a 19% decrease compared to patients with no pancreatic insufficiency. ¹⁹ These covariates were incorporated into the present model using the same structural relationships, both to account for known sources of IIV and to ensure consistency with previously validated findings. Notably, the model developed in this cohort of wwCF, re‐estimated the corresponding covariate effects, which remained consistent in both magnitude and direction with those reported in the adult population.

The effect of pregnancy was statistically significant on the apparent clearance of TEZ and IVA but not of ELX. Indeed, the addition of pregnancy significantly improved model fit, as evidenced by reductions in the OFVs (−19.05 and −4.36 for TEZ and IVA, respectively). Mothers with CF were estimated to have a 48% point increase in TEZ apparent clearance rate (β _{pregnancy/Cl, TEZ} [RSE%] = 1.48 [9.3%]) and a 18% point increase in IVA elimination (β _{pregnancy/Cl, IVA} [RSE%] = 1.18 [8.6%]) compared to non‐pregnant wwCF. Although, model criterions were not significantly improved in the ELX PK model, a pregnancy trend was identified with a 16% increase in apparent ELX clearance in mothers with CF. The final popPK estimates of ETI are summarized in Table 2.

TABLE 2.

Population pharmacokinetic parameter estimates of the final ETI models in adult women with CF.

	Elexacaftor	Tezacaftor	Ivacaftor
	Estimate (%RSE)
T lag (h)	2.17 (fixed)
k a(h−1)	0.59 (fixed)	2.38 (48.2%)	0.79 (36.8%)
CL (L/h/70 kg)	1.21 (5.1%)	0.97 (6.6%)	9.64 (5.2%)
V (L/70 kg)	81.11 (fixed)	25.46 (9.3%)	200.38 (18.8%)
β age/Cl	−0.43 (39.6%)	‐	−0.36 (52.2%)
β genotype/Cl (elexacaftor)	−0.19 (51.6%)	‐	‐
β pancrea.insuff/F (tezacatfor)	‐	−0.20 (29.1%)	‐
β pregnancy/Cl	‐	1.48 (9.3%)	1.18 (8.6%)
ω Cl	0.41 (10.6%)	0.24 (12.1%)	0.51 (8.3%)
RUV, proportional	‐	0.39 (5.3%)	‐
RUV, combined, a	2.46 (12.6%)	‐	0.23 (33.5%)
RUV, combined, b	0.18 (34.1%)	‐	0.2 (30.0%)
Shrinkage	41.5%	48.6%	26.2%

Note: %RSE: percent relative standard error; T _lag: lag time; k _a: absorption rate; Cl: total apparent clearance (allometrically scaled to 70 kg); V: apparent volume of distribution (allometrically scaled to 70 kg); β _age/Cl: effect of age on elexacaftor or ivacaftor apparent clearance; β _genotype/Cl: effect of genotype (two vs. one ETI responsive variant) on elexacaftor apparent clearance; β _{pancrea.insuff/F}: effect of exocrine pancreatic insufficiency on tezacaftor bioavailability; β _pregnancy/Cl: effect of pregnancy on tezacaftor and ivacaftor apparent clearance; ω_Cl: inter‐subject clearance variability expressed as standard deviation; RUV: residual unexplained variability with a, the additive error component and b, proportional error component. Shrinkage was calculated from the empirical standard deviation of the random effects and the estimated standard deviation.

The exploratory analysis of trimesters' effect (T1, T2 and T3) on ETI PK yielded no substantial statistical differences on any of the PK parameters and in any of the three models. However, a notable trend was observed, with increased TEZ and IVA apparent clearance, which appears pronounced from the second trimester onwards (compared to a non‐pregnant state or the first trimester). The inclusion of pregnancy as a binary covariate in the ELX model and as a categorial model (trimesters) in the TEZ and IVA model are reported with all the estimated parameters in Table S1.

3.4. Model evaluation

All three final popPK models provided reliable parameter estimation, with small relative standard errors (RSE%), as shown in Table 2. While two parameters (β _genotype for ELX and β _age for IVA model) had moderately high RSEs (slightly above 50%), their estimated values remained stable across a range of model structures and covariate testing scenarios. This supports the reliability of their estimates and suggests that it does not reflect structural misspecification or overparameterization, but rather reflects the smaller range of higher age groups or diverse genotypes present in this cohort compared to the general adult population. Diagnostic plots, pcVPCs and NPDEs (displayed in Figures S3–S5), collectively supported the adequacy of the final models. NPDEs were reasonably centred around zero and showed no major trends over time or predicted concentrations, suggesting an unbiased fit. Although η‐shrinkage values were moderately high for ELX (41.5%) and TEZ (48.6%), which may limit confidence in the empirical Bayes estimates (EBE), the covariate effects and population parameters using the SAEM algorithm are not impacted and therefore remain robust and reliable. In addition, the majority of observed concentrations fell within the 95% prediction intervals in the pcVPCs, supporting the models' ability to capture the central tendency and variability of the data.

3.5. ETI exposure in pregnant women

The model‐predicted maternal PK parameters (AUC, C _max and C _min) for all three compounds in women from our cohort were consistent with the expected adult exposure ranges. ³ , ¹⁴ , ²⁰ , ²³ Importantly, the predicted TEZ AUC_0–24h during pregnancy was low, with 44.7% of predicted values falling below the expected 5th percentile of the TEZ prediction interval in adults with CF. ²³ The predicted ETI exposure (AUC) of non‐pregnant vs. pregnant wwCF derived from the final ETI models are illustrated in Figure 3, with respective expected exposure ranges in adults with CF. ²³

FIGURE 3.

Elexacaftor–tezacaftor–ivacaftor (ETI) model predicted‐exposure from real‐world non‐pregnant vs. pregnant women with cystic fibrosis. The dashed line and the blue shaded areas represent, respectively, the median area under the curve (AUC) value and the 5th and 95th percentiles of the prediction intervals of each compound. ²³

Further, predicted maternal C _max and C _min values were low for all three compounds compared to those of non‐pregnant individuals and are shown in Figures S1 and S2.

However, all ETI C _min values during pregnancy remained above their respective EC_50. ¹⁴ , ²⁰ The individual PK parameters estimated from the final ETI models of pregnant vs. non‐pregnant wwCF are presented in Table 3.

TABLE 3.

Individual pharmacokinetic parameter estimates of real‐world pregnant vs. non‐pregnant women with CF.

	Predicted AUC (mg h/L) of non‐pregnant wwCF (n = 233)	Predicted AUC (mg h/L) of pregnant wwCF (n = 47)	Predicted C max (mg/L) of non‐pregnant wwCF (n = 233)	Predicted C max (mg/L) of pregnant wwCF (n = 47)	Predicted C min (mg/L) of non‐pregnant wwCF (n = 233)	Predicted C min (mg/L) of pregnant wwCF (n = 47)
Elexacaftor	194.2 [154.5–255.5] (62.6–514.2)	160.1 [133.9–199.9] (73.0–298.5)	9.1 [7.4–11.8] (3.0–22.5)	7.5 [6.3–9.1] (4.0–13.4)	7.0 [5.4–9.4] (2.1–20.2)	5.6 [4.7–7.4] (2.0–11.3)
Tezacaftor	97.9 [82.2–110.5] (41.5–182.6)	58.4 [49.3–64.4] (40.9–80.2)	6.1 [5.3–6.7] (2.6–9.4)	4.3 [3.6–4.6] (3.2–5.1)	2.4 [2.0–2.9] (1.0–5.9)	1.0 [0.9–1.5] (0.6–1.9)
Ivacaftor	17.4 [13.5–22.8] (4.6–76.7)	11.1 [9.3–14.0] (3.3–19.0)	1.7 [1.4–2.2] (0.7–6.7)	1.2 [1.0–1.4] (0.5–1.8)	1.1 [0.8–1.6] (0.1–6.1)	0.7 [0.5–0.9] (0.1–1.3)

Note: Predicted data are communicated as median [interquartile range] (min–max). C _max: maximum concentration; C _min: minimum concentration. AUC: area under the curve between 0 and 24 h for elexacaftor and tezacaftor, and between 0 and 12 h for ivacaftor.

4. DISCUSSION

This study presents the first popPK analysis of ETI in pregnant and non‐pregnant wwCF. By leveraging a robust model structure previously evaluated and published; in adults with CF, we incorporated known ETI PK parameters and sources of variability and focused specifically on assessing the effect of pregnancy. Our analysis revealed substantial IIV, consistent with our earlier findings in adults, ¹⁹ further reinforcing the relevance of population modelling in this complex therapeutic setting. Pregnancy was found to significantly increase the apparent clearance of TEZ and IVA, respectively, by 48% and 18%, reducing the serum concentration of TEZ and IVA compared to non‐pregnant women. Interestingly, nearly half of the predicted TEZ AUCs in pregnant participants fell below the 5th percentile of the expected adult exposure range, ²³ raising important questions regarding the clinical implications of low drug levels during gestation. Conversely, the vast majority of ETI concentrations observed in pregnant wwCF from our cohort study exceeded the reported ETI EC₅₀ values for reduction in sweat chloride levels, ¹⁴ , ²⁰ which can provide reassurance regarding sustained efficacy despite considerable variability in ETI levels. Moreover, in each participant, ppFEV1 values appeared to have remained relatively stable through the periods of pre‐pregnancy, pregnancy and post‐partum, suggesting a preserved lung function in our 14 patients. Further, the results demonstrated variability in ppFEV1 during pregnancy in several patients, which may or may not provide the foundation for the hypothesis that variations in ETI concentrations during pregnancy could be responsible for variations in ppFEV1.

The current literature on ETI PK during pregnancy has been very limited, extending to a single study reporting maternal concentrations in six wwCF, with reduced or adjusted dosing. ⁶ Similar to our results, the authors highlighted considerable IIV in ETI exposure during gestation. ⁶

On another note, clinical reports and case studies have shown divergent outcomes. Indeed, from 45 ETI exposed pregnancies reported by Taylor‐Cousar et al., complications in two mothers and three infants were outlined. ⁷ The signalled cases proved to be challenging in terms of distinguishing disease‐ or treatment‐related outcomes. ⁷ In contrast, other studies have reported uneventful maternal and neonatal outcomes following ETI exposure during gestation. ⁸ For instance, Cimino et al. described a series of successful, uncomplicated pregnancies in women treated with ETI, further suggesting that continuation of therapy may be safe in many cases. ⁸

In a clinical landscape where pregnancy in wwCF is becoming increasingly common, the absence of robust PK data on ETI use during gestation represents a critical knowledge gap. The dosing decisions have relied largely on empirical judgement and sparse case reports, with no established guidelines or modelling‐based evidence to support optimal care. The extant literature points towards the pressing necessity for precise PK studies. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ No statistically significant effect of pregnancy on ELX PK was found here. However, this absence of evidence may reflect limited statistical power, and in the case of elexacaftor, its substantial IIV may mask subtle pregnancy‐related effects, such that small changes in PK may not reach statistical significance in this relatively small cohort.

While pregnancy was associated with a significant change in apparent clearance, this composite parameter reflects both systemic clearance and oral bioavailability, and therefore, the relative contribution of altered metabolic clearance vs. potential changes in bioavailability cannot be disentangled. This distinction is particularly relevant for ETI, as absorption is known to be enhanced when administered with a fat‐containing meal. ¹⁴ In the context of sparse PK sampling and a limited number of pregnant participants, variability in food intake at the time of dosing may therefore introduce additional variability in observed concentrations and partially confound the apparent magnitude of pregnancy‐related changes in exposure.

Although this represents the largest cohort to date in which maternal ETI PK have been characterized, the number of pregnant participants remains limited, which may introduce some uncertainty around the precise magnitude of pregnancy‐related changes, while not affecting the overall identification of a pregnancy‐related effect on clearance.

While no statistical significance was found in ETI PK described by trimesters in any of the three models, an upward trend was observed in TEZ and IVA elimination, starting from the second trimester onwards. The number of samples collected during pregnancy (47 in total, including nine in the first trimester vs. 20 and 18 in the second and third trimester) and the cohort size of pregnant women (n = 14) were clearly low, highlighting the challenge of achieving well‐balanced sampling in this population, which could explain why the trimester‐specific effects did not attain statistical significance. These findings warrant confirmation in larger prospective studies.

The present findings show that physiological changes in pregnancy—likely affecting drug metabolism, distribution and elimination—lead to lower systemic exposure to these compounds. Importantly, ETI is extensively metabolized by cytochrome P450 3A4 (CYP3A4), a key hepatic enzyme whose activity is known to increase during pregnancy. ³ , ¹⁵ , ²⁴ All three components exhibit high but differential dependence on CYP3A4‐mediated metabolism, quantified by the fraction metabolized via CYP3A4 (fmCYP3A4), with IVA exhibiting near‐complete CYP3A4 metabolism (fmCYP3A4 ≈ 98%), followed by TEZ (fmCYP3A4 ≈ 73.2%) and ELX (fmCYP3A4 ≈ 67%), suggesting a potential difference in sensitivity to gestation‐associated CYP3A4 induction. ¹⁴ Indeed, it has been reported that CYP3A4 activity rises by approximately 1.25‐, 1.75‐ and 2.32‐fold relative to non‐pregnant levels by the end of the first, second, and third trimesters, respectively. ²⁴ This increase is usually attributed to hormonal changes in gestation, including increased oestrogen and cortisol levels, which upregulate CYP3A4 expression. ²⁵ We could hypothesize that, the observed increase in ETI apparent clearance during pregnancy may be, at least in part, a consequence of this enhanced metabolic capacity during gestation. Furthermore, the gradual reduction in exposure observed from the second trimester onwards aligns with the trajectory of CYP3A4 induction and reinforces the hypothesis that gestation‐driven metabolic changes contribute meaningfully to the PK alterations observed in our study.

The overall decrease in TEZ and IVA plasma concentrations observed during pregnancy does not indicate how the unbound, pharmacologically active, fraction of these drugs is affected. Indeed, interpretation of reduced ETI exposure during pregnancy and the differences observed between compounds, should also consider differences in plasma protein binding and unbound fraction (fu). IVA exhibits a very low unbound fraction (fu ≈ 0.001), compared with TEZ (fu ≈ 0.009) and ELX (fu ≈ 0.00704). ²⁰ Pregnancy induces significant physiological changes, including alterations in plasma protein levels (such as decrease in albumin and α1‐acid glycoprotein), which can modify drug plasma binding and enhance hepatic clearance, thereby contributing to reduced total drug exposure. The combination of different CYP3A4 dependence levels of the three compounds and their respective unbound fractions may render sensitivity to gestation‐related physiological changes.

Moreover, since only the free drug is exerting pharmacologic effects, total concentration changes alone may not reflect the therapeutic impact. In pregnancy, increased CYP3A4‐mediated metabolic capacity and reductions in plasma protein binding occur concurrently, such that enhanced clearance may reduce total drug concentrations while an increased unbound fraction may partially offset reductions in pharmacologically active exposure. Thus, quantifying unbound drug concentrations is essential to support rational dose adjustment strategies. ²⁶ This dynamic interplay underscores the need for further studies evaluating unbound ETI concentrations throughout gestation, to better interpret the pharmacodynamic consequences of reduced total exposure during gestation, and to appropriately guide dose adjustments. ²⁶

Our results suggest that in certain individuals, exposure—particularly to TEZ—may fall to the lower end of the therapeutic range. Although a significant increase in apparent clearance of TEZ and IVA was observed, this does not necessarily imply an increase in dose. Nevertheless, it does not preclude the potential need for personalized dose management in certain patients.

The absence of clinical complications or adverse events in our cohort, both for mothers and their newborns, is reassuring. However, exposure‐response analyses or long‐term outcome data are necessary to complete our results and the knowledge regarding potential risks for both mothers and fetus.

Beyond popPK, physiologically based pharmacokinetic (PBPK) modelling presents a complementary approach. The recent study by Hong et al., for instance, aimed to predict maternal–fetal ETI exposure using PBPK simulations, based on known physicochemical characteristics of ETI and known physiological changes during pregnancy. However, model validation was limited by sparse empirical data. ²⁷ Interestingly, to measure pregnancy effect on ETI PK, a non‐compartmental PK analysis was performed in a 34‐week pregnant woman with CF and compared to previously reported data in non‐pregnant CF patients. A reduction was observed in AUC values of 62% for ELX, 43% for TEZ and 15.4% for IVA in late pregnancy compared to the respective mean reported values in non‐pregnant individuals. ²⁷

In parallel, ex vivo studies such as cotyledon perfusion models may offer an opportunity to better understand placental transfer and fetal drug exposure and enrich PBPK models.

Prospective clinical trials are underway in the United States to further characterize ETI PK, safety and efficacy during pregnancy. ²⁸ These efforts, combined with real‐world data, will help refine recommendations and guide future care.

Given the physiological changes in plasma protein concentrations during pregnancy, further research is warranted to characterize unbound drug exposure and its pharmacokinetic–pharmacodynamic relevance. Clinical decisions to interrupt therapy or adopt cautious dosing strategies to minimize fetal exposure raise important ethical and health equity considerations, as they may compromise maternal health, with downstream effects on pregnancy outcomes and long‐term caregiving capacity. This underscores the need for balanced, evidence‐informed clinical and pharmacological frameworks that protect fetal wellbeing while ensuring sustained maternal health, through disease control, and long‐term quality of life.

Despite the preliminary nature of our findings, they provide the first evidence that pregnancy significantly influences ETI PK and underscore the necessity for further investigation into the efficacy of ETI during pregnancy, particularly in cases where the fetus is affected by CF, to ensure that maternal dosing provides sufficient in utero exposure for therapeutic benefit. This PK analysis offers anticipated preliminary data that may help inform future dosing strategies in this sensitive and understudied population and contribute to the growing understanding of drug exposure and CF care during pregnancy.

AUTHOR CONTRIBUTIONS

Concept and design: PRB, SB, JMT. Acquisition of data: LFB, GL, VG, NC, JDS, JF, RK, IH, CM, PRB, VT, SB. Analysis and interpretation of data: PM, NB, FF, SU, PRB, SB. Drafting of the manuscript: PM, NB, FF, SU, SB, PRB. Critical revision of the manuscript for important intellectual content: All authors.

CONFLICT OF INTEREST STATEMENT

The authors report no conflict of interest directly relating to this manuscript. P.R.B. has received personal fees for lectures/advisory boards from Astra Zeneca, Chiesi, Insmed, Pfizer, Sanofi, MSD, Viatris, Vertex and support for travelling to meetings by Astra‐Zeneca and Chiesi. C.M. has received personal fees for lectures/advisory boards from Astra Zeneca, GSK, Vertex and Zambon. N.C. has received personal fee for an advisory board from MSD France. All other authors have no conflict of interest.

STUDY REGISTRY

The study was not registered as it is not mandatory in France for real‐word studies. It was approved by the Institutional Review Board of the French Society for Respiratory Medicine (Société de Pneumologie de Langue Française) (#2020‐003).

Supporting information

Table S1. Population pharmacokinetic parameter estimates of the ETI models with respective tested covariates, pregnancy (PREG) or trimesters (T1, T2, T3).

Figure S1. Elexacaftor‐tezacaftor‐ivacaftor (ETI) model predicted‐exposure from real‐world non‐pregnant vs. pregnant women with cystic fibrosis (CF). The dashed line represents the mean maximum concentration (C_max) value expected in adults with CF and the blue shaded area denotes the approximate 95% reference range (mean ± 2SD) of the C_max prediction interval in adults, for each compound¹.

Figure S2. ETI model predicted‐exposure from real‐world non‐pregnant vs. pregnant women with cytic fibrosis (CF). The dashed line represents the mean minimum concentration (C_min) value expected in adults with CF and the blue shaded area denotes the approximate 95% reference range (mean ± 2SD) of the C_min prediction interval in adults, for each compound¹.

Figure S3. Elexacaftor goodness‐of‐fit plots of the final population pharmacokinetic model. (a) Elexacaftor population predictions vs. observed concentrations; (b) elexacaftor individual predictions vs. observed concentrations; (c) normalized prediction distribution error (NPDE) vs. time; (d) prediction‐corrected visual predictive check plot. The dashed red lines (a & b) represent the line of unity. The blue dashed line (c) represents the linear regression fit. The black dots represent the observed data in non‐pregnant cystic fibrosis (CF) patients and blue dots describe observed data in pregnant CF woman (a‐c). The solid blue lines represent the median, 5th and 95th percentile of observed data (d). The dashed black lines represent the median, 5th and 95th percentiles of the predicted data (d). The shaded areas represent the 90% confidence intervals for the medians (pink) and for the 5th and 95th percentiles (blue) of the predicted data (d).

Figure S4. Tezacaftor goodness‐of‐fit plots of the final population pharmacokinetic model. (a) Tezacaftor population predictions vs. observed concentrations; (b) tezacaftor individual predictions vs. observed concentrations; (c) normalized prediction distribution error (NPDE) vs. time; (d) prediction‐corrected visual predictive check plot. The dashed red lines (a & b) represent the line of unity. The blue dashed line (c) represents the linear regression fit. The black dots represent the observed data in non‐pregnant cystic fibrosis (CF) patients and blue dots describe observed data in pregnant CF woman (a‐c). The solid blue lines represent the median, 5th and 95th percentile of observed data (d). The dashed black lines represent the median, 5th and 95th percentiles of the predicted data (d). The shaded areas represent the 90% confidence intervals for the medians (pink) and for the 5th and 95th percentiles (blue) of the predicted data (d).

Figure S5. Ivacaftor goodness‐of‐fit plots of the final population pharmacokinetic model. (a) Ivacaftor population predictions vs. observed concentrations; (b) ivacaftor individual predictions vs. observed concentrations; (c) normalized prediction distribution error (NPDE) vs. time; (d) prediction‐corrected visual predictive check plot. The dashed red lines (a & b) represent the line of unity. The blue dashed line (c) represents the linear regression fit. The black dots represent the observed data in non‐pregnant cystic fibrosis (CF) patients and blue dots describe observed data in pregnant CF woman (a‐c). The solid blue lines represent the median, 5th and 95th percentile of observed data (d). The dashed black lines represent the median, 5th and 95th percentiles of the predicted data (d). The shaded areas represent the 90% confidence intervals for the medians (pink) and for the 5th and 95th percentiles (blue) of the predicted data (d).

Magnas P, Bouazza N, Foissac F, et al. Pregnancy‐related effect on elexacaftor, tezacaftor and ivacaftor pharmacokinetics in women with cystic fibrosis. Br J Clin Pharmacol. 2026;92(9):3202‐3213. doi: 10.1002/bcp.70620

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are not publicly available but are available from the corresponding author upon reasonable request.