Prediction of sustained fetal toxicity induced by ketoprofen based on PK/PD analysis using human placental perfusion and rat toxicity data

Abstract

Aim

We encountered a case of fetal toxicity due to ductus arteriosus (DA) constriction in a 36‐week pregnant woman who had applied multiple ketoprofen patches. The aim of the present study was to present the case and develop a model to predict quantitatively the fetal toxicity risk of transdermal administration of ketoprofen.

Methods

Human placenta perfusion studies were conducted to estimate transplacental pharmacokinetic (PK) parameters. Using a developed model and these parameters, human fetal plasma concentration profiles of ketoprofen administered to mothers were simulated. Using pregnant rats, DA constriction and fetal plasma drug concentration after ketoprofen administration were measured, fitted to an Emax model, and extrapolated to humans.

Results

Transplacental transfer value at the steady state of ketoprofen was 4.82%, which was approximately half that of antipyrine (passive marker). The model and PK parameters predicted almost equivalent mother and fetus drug concentrations at steady state after transdermal ketoprofen administration in humans. Maximum DA constriction and maximum plasma concentration of ketoprofen after administration to rat dams were observed at different times: 4 h and 1 h, respectively. The model accurately described the delay in DA constriction with respect to the fetal ketoprofen concentration profile. The model with effect compartment and the obtained parameters predicted that use of multiple ketoprofen patches could potentially cause severe DA constriction in the human fetus, and that fetal toxicity might persist after ketoprofen discontinuation by the mother, as observed in our case.

Conclusion

The present approach successfully described the sustained fetal toxicity after discontinuing the transdermal administration of ketoprofen.

What is Already Known about this Subject

• The use of ketoprofen, including transdermal administration, in full‐term pregnant women leads to fetal toxicity, such as ductus arteriosus (DA) constriction.

• Previous attempts at quantitative prediction of ketoprofen‐related fetal toxicity have been hampered by a lack of data on the transplacental kinetics of ketoprofen and on the DA constriction potency profile with respect to the fetal plasma concentration after maternal administration.

What this Study Adds

• Human fetal DA‐constrictive profiles of ketoprofen, after various routes of administration to the mother, became quantitatively predictable, based on the experimental results of human placental perfusion and drug administration to pregnant rats.

• The present model can successfully account for the sustained fetal toxicity that is observed after discontinuing transdermal administration of ketoprofen, as seen in the described case.

Introduction

Ketoprofen is a nonsteroidal anti‐inflammatory drug (NSAID) that is widely used both in systemic (oral, suppository or injectable) and local topical (patches, poultices, ointments, gels, creams and lotions) formulations. Systemic use of NSAIDs in third‐trimester women is contraindicated due to adverse reactions such as persistent pulmonary hypertension of the newborn (PPHN), which is related to constriction of the ductus arteriosus (DA) 1, 2. However, relatively little is known about the risks of local administration of NSAIDs, although cases of DA constriction considered to be related to the use of ketoprofen patches in third‐trimester women have been reported 3. Consequently, the use of ketoprofen patches in the third trimester of pregnancy is now also contraindicated 4, 5. However, the risks of DA arising from use of local topical formulations of ketoprofen have not been studied.

Experimental study of fetal drug transfer in pregnant women is not feasible. However, human placental perfusion studies are a useful method for elucidating fetal drug transfer characteristics, as they enable both continuous measurement of changes in drug concentration and direct observation of drug input–output in maternal and fetal blood through the placenta 6. We have established an analytical model for studying drug disposition using human placental perfusion, and have successfully applied it for transplacental kinetic studies of salicylic acid, diclofenac and antipyrine 7, 8. Although this model allows us to simulate human fetal drug concentrations without in vivo dosing and/or blood sampling, it is important to note that further validation will be required to ensure the reliability of the approach.

In the present study, we present a case of fetal toxicity due to DA constriction in a 36‐week pregnant woman who had applied multiple ketoprofen patches. We used human placental perfusion measurements to elucidate the placental transfer kinetics of ketoprofen, and to estimate the fetal drug concentration profile during third‐trimester ketoprofen administration, and we used a rat model 9, 10 to establish the relationship between the in vivo fetal rat plasma concentration of ketoprofen and toxicity (DA constriction). These results were then extrapolated to humans in order to predict quantitatively fetal toxicity resulting from third‐trimester use of ketoprofen. Finally, we compared the model prediction with observations from our case.

Materials and methods

Figure 1 shows a schematic flow chart of the experiments and the pharmacokinetic/pharmacodynamic (PK/PD) analysis used to estimate profiles of the fetal plasma concentration and fetal toxicity of ketoprofen in humans in the current study. The estimated profiles were compared with the observed values of our case.

Figure 1.

Schematic flow diagram of the current study. A human placental perfusion study (A) and pregnant rat acute toxicity study (B) were conducted. The transplacental pharmacokinetic (PK) parameters calculated in (A) were used to estimate the ketoprofen concentrations in the human fetal plasma. The rat pharmacodynamic parameters calculated in (B) were then extrapolated to human fetuses, in order to predict quantitatively the ductus arteriosus constriction profiles following ketoprofen use by pregnant women

MLAB software (Civilized Software Inc., Silver Spring, MD, USA) was used for statistical modelling and analysis, including calculation of PK and PD parameters, nonlinear least‐squares estimations of blood concentrations and DA constriction effects.

(A) Analysis of transplacental transfer of ketoprofen by human placental perfusion study

Materials

Human full‐term placentas were obtained from gravidae after Caesarean delivery. The study protocol was approved by the ethical committees of the Graduate School of Pharmaceutical Science and the Graduate School of Medicine, University of Tokyo, and written informed consent was provided by the gravidae hospitalized in the University of Tokyo hospital before delivery.

Ketoprofen and 4‐biphenylacetic acid were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). Antipyrine and 4‐aminoantipyrine were purchased from Nacalai Tesque Inc. (Kyoto, Japan). Human serum albumin was purchased from Kaketsuken (The Chemo‐Sero‐Therapeutic Research Institute, Kumamoto, Japan). All other reagents used were of the highest grade commercially available.

Krebs–Ringer–bicarbonate buffer (118 mM NaCl, 4.7 mM KCl, 1.3 mM MgSO₄, 24.2 mM NaHCO₃, 2.5 mM CaCl₂) containing D‐glucose (1.0 g l^–1), dextran (molecular weight 35 000–50 000, 1.0 g l^–1) and heparin (12 500 IU l^–1), and human serum albumin (2.0 g l^–1) was used as the perfusate. The maternal and fetal perfusates were aerated with 95% O₂–5% CO₂ and 95% N₂–5% CO₂, respectively, adjusted to pH 7.3 with HCl, and warmed to 37°C. Aeration was continued throughout the experiment.

Placental perfusion study

In order to analyse the transplacental transfer of ketoprofen, we employed human placental perfusion in an open double circuit, with antipyrine used as a passive marker, as previously reported 7, 8. Fetal perfusion pressure was monitored throughout the experiment and confirmed not to exceed 40 mmHg. Leakage of perfusate from the fetal side was less than 3.0 ml h^–1. Briefly, after cannulation, the cotyledon sample was perfused with drug‐free perfusate for 30 min to stabilize the preparation and then perfused according to protocol I, Ia or II, as follows. Each protocol was conducted three times, with one placenta‐derived cotyledon each time. Nine placentas were utilized for the study. The concentration of ketoprofen was selected based on the clinical plasma concentration of ketoprofen following a single dose of 20 mg or 160 mg (plaster‐type formulation) – i.e. peak plasma concentration (Cmax) 136 ng ml^–1 or 919 ng ml^–1, respectively) 5.

Protocol I

The maternal perfusate was changed to one containing ketoprofen (293 ± 21 ng ml^–1) and antipyrine (45.2 ± 1.6 μg ml^–1), and the perfusion was conducted for 60 min. The maternal and fetal effluents were sampled periodically for 60 min.

Protocol Ia

The maternal perfusate was changed to one containing both ketoprofen (261 ± 3 ng ml^–1) and antipyrine (44.9 ± 4.7 μg ml^–1), and the perfusion was conducted for 60 min. It was then changed to a drug‐free perfusate, and the perfusion was continued up to 70 min, in order to monitor the outflow of drugs from the placental tissue.

Protocol II

The fetal perfusate was changed to one containing both ketoprofen (297 ± 54 ng ml^–1) and antipyrine (44.1 ± 0.0 μg ml^–1), and the perfusion was conducted for 60 min. Maternal and fetal effluents were sampled periodically for 60 min.

In all of the protocols, maternal and fetal effluents were sampled periodically from the maternal chamber and fetal venous cannula, respectively, and perfused cotyledon was also weighed and sampled just after the last sampling of effluent.

The transplacental transfer value at steady state (TPTss), an index of permeability across the placenta 11, was calculated as the ratio of the amount of drug transferred to the fetal effluent across the placenta to that infused at steady state (20–60 min) and used to evaluate the permeability in protocols I and Ia.

PK analysis of transplacental transfer

The tranplacental transfer of antipyrine and ketoprofen was characterized by fitting compartment models developed in our previous reports (for ketoprofen, the same model as in the case of diclofenac was applied) 7, 8 (see Figure S1) to the sets of time profiles of maternal and fetal effluents and the amount of drug in the perfused placental tissues in protocols I, Ia and II simultaneously, to obtain transplacental PK parameters K1, k2, the elimination rate constant (ks) and Vp for antipyrine, and K1, K4, k2, k3 and ks for ketoprofen, using MLAB software.

The unbound fraction (fb) of ketoprofen and antipyrine in the perfusate was determined by ultrafiltration (Centricon Ultracel YM‐30; Millipore, Bedford, MA, USA). The unbound influx clearance from mother and fetus to placental tissue, K1’ and K4’, was calculated by dividing K1 and K4 by fb, respectively.

(B) Relationship between ketoprofen plasma concentration and DA constriction effect in fetal rats

Experimental rat study

Wistar female rats on day 21 of gestation (weight 300–500 g, 11–35 weeks, Saitama Experimental Animal Supply, Saitama, Japan) were used for this study. The experimental protocol was approved by the animal ethics committee of Tokyo Women's Medical University. Animal handling was compliant with the Animal Experiment Regulations of the University of Tokyo and Tokyo Women's Medical University.

Ketoprofen in aqueous suspension (0.03–10 mg kg^–1, lactose monohydrate as diluent) was administered through an orogastric tube to each rat on day 21 of gestation. Dams were euthanized by cervical dislocation 0.5–7.5 h after dosing, followed immediately by Caesarean section and removal of the fetuses. Measurement of fetal DA constriction and collection of plasma samples in mothers and fetuses were carried out as previously reported 9.

Estimation of ketoprofen plasma concentration profile in rat dams and fetuses

PK parameters were obtained by fitting a compartment model (Figure 2a) to PK data from rat dams and fetuses after a single oral administration of drugs to pregnant rats. It was assumed that the fetal distribution volume based on body weight is equal to that of the mother, with dams weighing 400 g and fetuses 5 g each. Further, it was assumed that there were 15 fetuses per dam, and that the fetuses had no metabolic capability. PK parameters were assumed to be constant across all dosages, with the exception of distribution volume, which was calculated for each dosage. Obtained PK parameters were used to estimate fetal plasma concentration of drugs in rats.

Figure 2.

Pharmacokinetic models to predict the fetal plasma concentration of ketoprofen in rats (A) and humans (B). Ce, drug concentration in the effect compartment (μg ml^–1); Cf, drug concentration in the fetal venous compartment (μg ml^–1); Cm, drug concentration in the maternal intervillous compartment (μg ml^–1); Cpf, drug concentration in the fetal plasma compartment (μg ml^–1);Cpm, drug concentration in the maternal plasma compartment (μg ml^–1); K1” and K4”, influx clearance in vivo (l h^–1/placenta); k2 and k3, efflux clearance (h^–1); k1e, transition rate constant to effect compartment (h^–1); ka, absorption rate constant (h^–1); ke0, elimination rate constant from the effect compartment (h^–1); kef, elimination rate constant from the fetal compartment (h^–1); kem, elimination rate constant from the maternal compartment (h^–1); kmf and kfm, mother–fetus transfer rate constant (h^–1); ks, elimination rate constant from the placenta (h^–1); ktr, transition rate constant from the skin to plasma compartment (h^–1); Qf, fetal blood flow rate (3 l h^–1/placenta); Qm, maternal blood flow rate (18 l h^–1/placenta); Vdf, drug distribution volume in the fetal plasma compartment (l kg^–1); Vdm, drug distribution volume in the maternal plasma compartment (l kg^–1); Ve, drug distribution volume in the effect compartment (l kg^–1); Vf, fetal vascular volume in the placental tissue (0.024 l/placenta); Vm, maternal intervillous volume in the placental tissue (0.113 l/placenta); Xg, amount of drug in the gastrointestinal tract (mg kg^–1 or mg); Xp, amount of drug in the placental compartment (mg/placenta); Xsk, amount of drug in the skin compartment (mg); Xvh, amount of drug in the vehicle compartment (mg)

Estimation of unbound ketoprofen concentration in rat plasma

Eq. (B‐1) for the relationship between the bound (Cb) and unbound (Cf) concentration of ketoprofen presented by Satterwhite and Boudinot 12 was used to estimate Cf.

$$\mathit{Cb}=\frac{\mathrm{N}\cdot \mathrm{K}\cdot \mathit{Alb}\cdot \mathit{Cf}}{1+\mathrm{K}\cdot \mathit{Cf}}$$ (B‐1)

(N = 2.137, K = 3.109 × 10⁵ 12).

The albumin concentration in plasma for rat dams and fetuses was measured using the AssayMax Rat Albumin ELISA Kit (Assaypro LLC, St Charles, MO, USA). The drug plasma concentration in fetal rat estimated above was taken as total ketoprofen concentration in plasma (Call = Cb + Cf).

Description of the relationship between unbound plasma concentration and DA constriction in fetal rats

The effect compartment model shown in Figure 2A was used to describe the relationship between the DA constriction profile in rats and the unbound plasma concentration in fetal rats estimated above. It was assumed that the drug concentration in the plasma compartment was equivalent to that in the effect compartment at steady state (k1e/Ve = ke0/Vd,f). It was also assumed that the unbound drug concentration in the effect compartment is expressed by Eq. (B‐2), similarly to that in plasma. The relationship between unbound drug concentrations (Ce,u) and degree of DA [the inner diameter ratio of the DA to pulmonary artery (PA)] in the effect compartment model was then described by an Emax model (eq. (B‐2)). In Eq. (B‐2), ‘Emax’ indicates maximum DA constriction, ‘EC _50,u’ indicates unbound drug concentration in the effect compartment at half‐maximum DA constriction, and ‘ γ’ is the Hill coefficient.

$$\mathit{DA}/\mathit{PA}=1-\frac{Emax}{1+exp\left\{\gamma \left(ln\mathit{EC}50,u-ln\mathit{Ce},u\right)\right\}}$$ (B‐2)

(C) Quantitative prediction of drug concentrations and DA constriction effects in human fetuses following ketoprofen use in the third‐trimester of pregnancy

Calculation of plasma clearance based on protein binding

The estimated influx clearance from the maternal plasma to placental tissue, K1”, was calculated by multiplying K1’ by the unbound fraction in maternal plasma (fp,m). Similarly, the estimated influx clearance from the fetal plasma to placental tissue, K4”, was calculated by multiplying K4’ by the unbound fraction in fetal plasma (fp,f).

K1’ and K4’ were obtained in the placental perfusion study mentioned above. The values of fp,m and fp,f were estimated to be 0.017 and 0.014, respectively, using the equation for the relationship between Cb and Cf of ketoprofen in healthy adults 13 and values of albumin concentration for immediately postpartum mothers and neonates (20.8 g l^–1 and 26.6 g l^–1, respectively) as reported by Nau et al. 14, together with the assumption that 1/k1 + Cf = 1/k1.

Estimation of drug concentration in human fetal plasma

PK parameters for healthy adults were obtained by fitting conventional two‐compartment models with first‐order absorption (oral or rectal administration), or with first‐order absorption and first‐order transition from skin to plasma (transdermal administration) to PK data in healthy adults given ketoprofen as follows: single intravenous bolus (100 mg), single and repeated oral administration (150 mg sustained‐release capsule), single intrarectal administration (50 mg and 75 mg suppository), single and repeated topical dermatological administration (20 mg patch, 30 mg poultice, and 100 mg and 300 mg gel) 15, 16, 17, 18. It is known that PK changes generally occur during pregnancy, such as increasing apparent volume of distribution, decreasing plasma concentrations and increasing half‐life 19. However, no data are available for ketoprofen in pregnant women, so in the present work we assumed that the PK parameters were the same as those for adult nonpregnant women.

For estimating ketoprofen output to fetal plasma, a hybrid model (Figure 2B) was constructed, and the drug concentration profiles obtained above were used as input variables for the drug concentration profile in maternal plasma (Cp,m) in the perfusion model. The placental transfer parameters (K”1, k2, k3, K”4, ks) used in the model were either obtained or estimated from study (A). It was assumed that the rate constant of fetal elimination (kef) is 0, and that the fetal distribution volume based on body weight is equal to that of the mother, with mothers weighing 60 kg and fetuses 3 kg. Drug concentration profiles in the model for mother, placenta and fetus are shown in Figure 2B. This model was used to simulate output to fetal plasma after ketoprofen administration via various routes to the mother as follows: single and repeated oral administration of a 150 mg sustained‐release capsule [administration interval (τ) = 24 h], single and repeated intrarectal administration of 75 mg suppository (τ = 24 h), single and repeated topical dermatological administration of 20 mg, 80 mg and 160 mg patch (τ = 24 h), single and repeated topical dermatological administration of 30 mg and 120 mg poultice (τ = 12 h), and single and repeated topical dermatological administration of 30 mg and 150 mg gel (τ = 8 h). In the simulation, the PK and PD parameters other than the first‐order rate constant (ka) and the transition rate constant from the skin to the plasma compartment (ktr) were taken to have the same values, irrespective of the route of ketoprofen administration.

Estimation of fetal DA constriction after drug administration to third‐trimester women

The unbound drug concentration profile in the effect compartment (Ce,u), shown in Figure 2B, was calculated according to Eq. (C‐1), using the estimated drug concentration in fetal plasma (Cp,f) and ke0 obtained in study (B). It was assumed that the unbound fraction in the effect compartment was equivalent to that in fetal plasma (fp,f). Using the calculated Ce,u and PD parameters in rats (Emax, EC50,u, γ) obtained in study (B), the degree of DA diameter (DA) for each dosage and administration route in third‐trimester women was simulated according to Eq. (C‐2).

$$\frac{\mathit{dCe}}{\mathit{dt}}=\mathit{ke}0\cdot \mathit{Cp},f-\mathit{ke}0\cdot \mathit{Ce}$$ (C‐1)

$$\mathit{DA}\left(\%\mathit{\text{control}}\right)=100-\frac{100Emax}{1+exp\left\{\gamma \left(ln\mathit{EC}50,u-ln\mathit{Ce},u\right)\right\}}$$ (C‐2)

Determination of antipyrine and ketoprofen

Total concentrations of antipyrine were determined by using the high‐performance liquid chromatography‐ultraviolet (HPLC‐UV) method 7, 8. The detection limit of antipyrine in the effluent and perfusate was 1 μg ml^–1, and that in the placental tissue was 1 μg per gram of tissue.

The total concentrations of ketoprofen in obtained samples were determined by a modification of the HPLC‐UV method described in our previous reports 7, 8. The effluent or perfusate sample was spiked with internal standard (4‐biphenylacetic acid), and extraction was carried out with an anion‐exchange solid‐phase cartridge (SampliQ SAX, Agilent, Santa Clara, CA, USA). The homogenate of placental tissue was spiked with internal standard, and chloroform and HCl solution were then added. The organic layer was evaporated to dryness. The residue was dissolved in methanol and the solution was spiked with acetate buffer, then extracted with SampliQ SAX. Plasma samples from rat dams and fetuses were treated in the same way as described for effluent or perfusate samples. The eluate from SampliQ SAX was evaporated to dryness, and the residue was then dissolved in the mobile phase and subjected to HPLC. The mobile phase consisted of 0.02 M phosphate buffer (pH 4.0) and methanol (50:50, v/v), and the detection wavelength was set at 258 nm. The detection limit of ketoprofen was 3 ng ml^–1 in the effluent or perfusate sample, 3 ng per gram of tissue in placental tissue, and 0.03 μg ml^–1 in plasma from rat dams or fetuses.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 20, and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16.

Results

We encountered a case of fetal toxicity due to DA constriction in a 36‐week pregnant woman who had applied multiple ketoprofen patches. First, we present the case, followed by the results of experimental studies, including human placental perfusion and measurements of ketoprofen plasma concentrations and DA constriction effect in fetal rats, designed to obtain experimental parameters for our mathematical model. Finally, we examine how well the predictions obtained from our model can account for the observations from the case.

Case presentation

A 36‐year‐old woman at 36 weeks’ gestation in her second pregnancy was referred to our hospital owing to fetal cardiomegaly. She was a smoker with a complicated history of asthma, cervical vertebral sprain after a traffic accident 9 years earlier, and a herniated lumbar disc. She suffered continuous cervical pain and lower back pain, and had been using patches containing 20 mg ketoprofen, prescribed by her doctor, at her discretion during pregnancy. After 32 weeks’ gestation, she applied up to 12 patches because her pain had become severe. At 35 weeks and 5 days’ gestation, fetal heart enlargement was detected during prenatal care. However, she again applied eight patches on the following day. Two days later, at 20:00 h (36 weeks and 0 days’ gestation), assessment of the fetal heart revealed moderate to severe right ventricular hypertrophy (Figure 3A), appearing as significant right ventricular wall thickening, although she had used no patches on that day. In addition, tricuspid regurgitation (TR), causing moderate insufficiency, was detected by colour Doppler imaging (Figure 3B). The TR velocity was measured as 2.6 m s^–1, although this may have been inaccurate, as the TR in the neonatal period was faster (see below). A dilated pulmonary trunk leading to a severely constricted DA was observed (Figure 3C). The diameter of the DA was narrowed to 1.7 mm (normal average value 5 mm at 36 weeks’ gestation 21). The presence of pericardial effusion seemed to indicate cardiac failure. After counselling the patient about the risk of cardiac decompensation, we performed Caesarean delivery at 23:00 h on the same day. The female newborn had a birth weight of 3014 g, with Apgar scores of 8 and 9 at 1 min and 5 min, respectively, with an umbilical artery pH of 7.33. Postpartal adaptation was uneventful. After transportation to our neonatal intensive care unit, echocardiography revealed moderate TR, a very hypertrophied right ventricle and bidirectional shunting through the patent foramen ovale. The TRs in the neonatal period at 40 min, 12 h, 2 days and 4 days after birth were 4.0 m s^–1, 4.1 m s^–1, 3.5 m s^–1 and 2.6 m s^–1, respectively. These findings were compatible with severe neonatal pulmonary hypertension due to ductal constriction and suggestive of right‐sided hypertrophic cardiomyopathy, as previously reported 22, 23. She required only oxygen administration. The DA was closed at 12 h after birth. After 10 days of oxygen treatment, TR was no longer detectable, and 16 days after giving birth she was discharged after regression of the right ventricular thickness.

Figure 3.

Fetal echocardiography in our case (36 weeks’ gestation). (A) Hypertrophic right ventricle; (B) severe tricuspid regurgitation flow visualized by continuous‐wave Doppler echocardiography; and (C) constriction of the ductus arteriosus. The arrowheads indicate the diameter of the ductus arteriosus (1.7 mm)

Analysis of transplacental kinetics for ketoprofen and antipyrine using human placental perfusion study (A)

In protocols I and II, the concentrations of antipyrine and ketoprofen in the fetal and maternal effluents reached steady state at about 10 min after the start of perfusion (Figure 4). In protocol Ia, the antipyrine and ketoprofen concentrations in the effluents gradually decreased after the perfusate was changed to a drug‐free one (Figure 4). The concentration of ketoprofen in the placental tissue at 60 min was 132 ± 41 μg per gram of tissue, which was about twice as the concentration in the fetal effluent. The TPTss values of antipyrine and ketoprofen were 9.17 ± 1.55% and 4.82 ± 0.93%, respectively, and thus the transfer of ketoprofen was 52.6% of that of antipyrine. The kinetic properties of antipyrine and ketoprofen were found to be similar to those reported previously 7, 8, 24. Concentration profiles of antipyrine and ketoprofen in the fetal and maternal effluents under protocols I, Ia and II were adequately described by the model. Transplacental kinetic parameters obtained by fitting to the data are shown in Table 1.

Figure 4.

Model analysis of the transplacental transfer of antipyrine (A) and ketoprofen (B). Upper and lower graphs represent maternal perfusion (protocols I and Ia) and fetal perfusion (protocol II), respectively. Each point represents the mean ± standard deviation (n = 6 for protocol I, n = 3 for the other protocols). The lines represent model fit

Table 1.

Transplacental pharmacokinetic parameters of ketoprofen and antipyrine

	K1	fb	K1’	fp,m	K1”	k2	K1’/k2	K1”/k2	ks	K1’/(k2 + ks) b	K1”/(k2 + ks) b
	(ml min –1 per gram cotyledon)		(ml min –1 per gram cotyledon)		(ml min –1 per gram cotyledon)	(min –1 )	(ml per gram cotyledon)	(ml per gram cotyledon)	(min –1 )	(ml per gram cotyledon)	(ml per gram cotyledon)
Ketoprofen	0.167	0.260	0.644	0.017	0.0109	0.0719	8.95	0.152	0.149	2.91	0.0495
Antipyrine	0.368	1	0.368	0.884a	0.325	0.414	0.888	0.785	0.119	0.691	0.611

	k3	K4	K4’	fp,f	K4”	K4’/k3	K4”/k3	Vp	$\frac{K{\mathit{1}}^{\prime }k\mathit{3}}{K{\mathit{4}}^{\prime }\left(k\mathit{2}+\mathit{ks}\right)}$ f	$\frac{K{\mathit{1}}^{″}k\mathit{3}}{K{\mathit{4}}^{″}\left(k\mathit{2}+\mathit{ks}\right)}$ f
	(min –1 )	(ml min –1 per gram cotyledon)	(ml min –1 per gram cotyledon)		(ml min –1 per gram cotyledon)	(ml per gram cotyledon)	(ml per gram cotyledon)	(ml per gram cotyledon)
Ketoprofen	0.152	0.136	0.523	0.014	0.00733	3.44	0.0482	–	0.845	1.03
Antipyrine	–	–	–	0.869a	–	–	–	0.809	0.805c	0.712c

‐: The velues were not applicable.

K1 and K4, influx clearance from intervillous compartment to placental compartment and from fetal venous compartment to placental compartment, respectively; K1’ and K4’, deviding K1 and K4 by fb, respectively; fb: unbound fraction in perfusate; K1” and K4”, influx clearance from mother compartment to placental compartment and from fetus compartment to placental compartment, respectively, in vivo; k2 and k3, efflux rate constant from placental compartment to fetus compartment and from placental compartment to maternal compartment, respectively; ks, elimination rate constant; fp,m: unbound fraction in human maternal plasma; fp,f, unbound fraction in fetal plasma; Vp, distribution volume of the placental compartment.

^aOhkawa et al., 2001 36

^bthe values represent (Xp/Cp,mu)ss or (Xp/Cp,m)ss

^cthe values represent (Cp,fu/Cp,mu)ss or (Cp,f/Cp,m)ss

for antipyrine, this value is calculated from K1’/(k2+ks)(Vp+Vf) or K1”/(k2+ks)(Vp+Vf), Vf: fetal vascular volume in placental tissue

Relationship between ketoprofen plasma concentrations and DA constriction effect in fetal rats (B)

DA constriction measurements in fetal rats after oral ketoprofen administration

The DA of fetal rats from rat dams at 1 h and 4 h after the administration of ketoprofen (1 mg kg^–1), as well as non‐administered rat dams, was photographed using a stereoscopic microscope (see Figure S2).

The DA/PA ratio in nondosed fetal rats (n = 6) was 0.98 ± 0.08 [mean ± standard error (SE)], which is the same as the reported value 9, and was considered to be essentially equal to 1. The dose‐dependent DA constriction (DA/PA ratio) in the fetal rats in each ketoprofen administration protocol group (n = 6 for each group) is shown in Figure 5. The DA/PA ratios at 4 h following the administration of 0.1, 1 and 10 mg kg^–1 were consistent with reported values 9.

Figure 5.

(A) Model analysis of ketoprofen plasma concentration profiles in rat mothers and fetuses. Each point represents the mean ± standard error (SE) (n = 3). (B) Ductus arteriosus/pulmonary artery (DA/PA) ratio profiles in rat fetuses. Each point represents the mean ± SE (n = 6). In both (A) and (B), the lines were calculated as described in the text

Plasma concentration of ketoprofen in rat dams and fetuses

The plasma concentration profiles of ketoprofen in rat dams and fetuses for each dosing group (n = 3) are shown in Figure 5. Cmax in rat dams and fetuses was observed at 0.5–1 h after administration. Plasma concentrations at 4 h after administration were linearly related to dose in the range 0.03–10 mg kg^–1, both for rat dams and fetuses.

The rat mother–fetus PK parameters obtained by fitting the compartment model (Figure 2A) to the data are shown in Table 2. Ketoprofen concentration profiles were accurately described in both rat dams and fetuses (Figure 5).

Table 2.

Rat mother–fetus pharmacokinetic parameters (A) and rat pharmacodynamic parameters of DA constriction (B) in response to ketoprofen

(A)
Parameters	Unit	Estimate	±	SD
ka	h–1	5.54 × 105	±	1.69 × 106
kmf	h–1	0.0652	±	0.0429
kfm	h–1	0.749	±	0.437
kem	h–1	0.325	±	0.077
Vd	l kg–1	0.134	±	0.047

(B)
Parameters	Unit	Estimate	±	SD
ke0	h–1	0.0545	±	0.3154
Emax		0.898	±	0.289
γ		0.624	±	0.325
lnEC 50,u		−5.56	±	5.32
(EC 50,u )	μg ml–1	3.87 × 10−3	(1.89 × 10−5 – 7.91 × 10−1)

It is assumed that Vd = Vdm = Vdf.

DA, ductus arteriosus; EC _50,u, ibuprofen plasma unbound concentration giving 50% of Emax; Emax, maximum DA constriction; ka, absorption rate constant; kem, elimination rate constant from the mother compartment; ke0, elimination rate constant from effect compartment; kmf and kfm, mother‐to‐fetus and fetus‐to‐mother transfer rate constant, respectively; SD, standard deviation; Vd, drug distribution volume in the mother plasma compartment; γ, Hill's coefficient.

Enzyme‐linked immunosorbent assay measurements of albumin in rat plasma showed concentrations of 19.8 ± 1.4 for dams (n = 10) and 10.3 ± 1.0 g l^–1 for fetuses (n = 10) (mean ± SE). These results were similar to reported values 25, and were equivalent to 62.1% (for pregnant rats) and 32.2% (for fetuses) of those in nonpregnant rats 12.

Description of the relationship between unbound drug plasma concentration and degree of DA constriction in fetal rats

The PD parameters for the effect of ketoprofen on fetal DA constriction obtained by fitting are shown in Table 2. In this experiment, maximum DA constriction potency was observed at 4 h, whereas the maximum plasma concentration of ketoprofen was observed at 0.5–1 h after administration to rat dams. The models accurately described the rat DA/PA ratio profiles, especially the delay of the DA constriction profile with respect to the fetal ketoprofen concentration profile (Figure 5).

Quantitative prediction of drug concentration and DA constriction in human fetuses following ketoprofen use in the third trimester of pregnancy (C)

Estimation of the unbound plasma concentration profile of ketoprofen in humans

The human in vivo influx clearance of ketoprofen in plasma (K1” and K4”), as calculated using fp,m and fp,f, was 0.0109 ml min^–1 per gram of cotyledon and 0.00733 ml min^–1 per gram of cotyledon, respectively.

Figure 6 shows the simulated profiles of ketoprofen concentration in the fetal plasma after maternal use of a 150 mg sustained‐release capsule, 75 mg suppository, 20/80/160 mg patch, 30/120 mg poultice and 30/150 mg gel formulations of ketoprofen in both single and repeated administrations, using the developed model (Figure 2B) and the obtained transplacental PK parameters (Table 1). It can be seen that the concentration profiles in the fetus were delayed compared with those in the mother, although the degree of delay depended on the administration route and dosage form. These findings suggest that the fetal/maternal ketoprofen concentration ratio at steady state would be almost 1 when the plasma concentration reaches a plateau after transdermal administration.

Figure 6.

Model‐predicted profiles of ketoprofen plasma concentration and ductus arteriosus (DA) constriction after single or repeated administration of ketoprofen to human mothers as oral or rectal administration (A); and transdermal administration (in plaster form (B), poultice form (C), and gel form (D). The lines were calculated as described in the text

The influence of the value of kef on the ketoprofen concentration profile in the fetus was examined after the single administration of a 75 mg suppository or a 80 mg patch, assuming kinetic constants for kef of 1/10, 1/5, 1/2 and 1 relative to the kinetic constant of maternal elimination (kem) (see Figure S4). The results indicated that the kef value has little influence on the fetal concentration profile after patch administration to the mother but it significantly influences the fetal concentration profile after suppository‐based administration to the mother.

Estimation of fetal DA constriction after drug administration to third‐trimester women

Figure 6 shows the simulated profiles of fetal DA constriction after maternal use of a 150 mg sustained‐release capsule, 75 mg suppository, 20/80/160 mg patch, 30/120 mg poultice and 30/150 mg gel formulations of ketoprofen in both single and repeated administrations using estimated human PK parameters for each formulation of ketoprofen (Table 3). The results suggest that the DA was continuously constricted to 40% of the normal average inner diameter following repeated use of oral or suppository formulations. They also suggest that the DA was continuously constricted to 60% of the normal average diameter following repeated use of topical dermatological formulations, even at the usual dosage levels, while at larger doses, such as the simultaneous use of eight patches (20 mg per patch), the DA was continuously constricted to less than 50%.

Table 3.

Estimated human pharmacokinetic parameters for each formulation of ketoprofen

Formulation	ka (h –1 )	ktr (h –1 )	Vdm (mg kg –1 )	kem (h –1 )	k12 (h –1 )	k21 (h –1 )
Intravenous	–	–	0.0729 ± 0.0069	0.931 ± 0.060	0.519 ± 0.077	0.472 ± 0.061
SR capsulea , b	0.158 ± 0.012	–	0.121 ± 0.006	0.931 (Fixed)	0.519 (Fixed)	0.472 (Fixed)
Suppository	0.729 ± 1.540	–	0.147 ± 0.298	0.687 ± 1.349	–	–
Plastera	0.028 ± 0.001	0.912 ± 0.226	0.073 (Fixed)	0.931 (Fixed)	0.519 (Fixed)	0.472 (Fixed)
Poulticea	0.007 ± 1.5 × 10−4	0.282 ± 0.016	0.073 (Fixed)	0.931 (Fixed)	0.519 (Fixed)	0.472 (Fixed)
Gela	0.007 ± 4.1 × 10−4	0.083 ± 0.008	0.073 (Fixed)	0.931 (Fixed)	0.519 (Fixed)	0.472 (Fixed)

SR, sustained release. The values indicate estimate ± standard deviation

^aSome parameters were fixed to those of intravenous administration, as indicated

^bZero‐dimensional dissolution rate in gastrointestinal compartment

The precise details of ketoprofen patch usage in our case remain unclear but the patient reported using a 160 mg ketoprofen patch repeatedly every 24 h up to the day before giving birth, and DA constriction (1.7 mm inner diameter) was observed as long as 20 h following discontinuation of use. Figure 7 shows a comparison between the observed data in our patient and simulated profiles of maternal and fetal plasma ketoprofen concentration and DA diameter. The estimated plasma ketoprofen concentrations at delivery (24 h following discontinuation) were 0.79 ng ml^–1 for the mother and 1.5 ng ml^–1 for the fetus, which were not inconsistent with the lower limit of quantification (2.5 ng ml^–1) in our case. The 1.7 mm DA measurement represented constriction to 34% of the standard inner diameter of 5 mm at week 36 of pregnancy 21, while the simulations predicted constriction to 56% at 20 h after discontinuation (Figure 7).

Figure 7.

Simulated profiles of maternal and fetal plasma ketoprofen concentration and ductus arteriosus (DA) diameter in the case presented in the present study. (A) Plasma concentration. The lines indicate predicted concentrations in the case presented. Maternal and fetal plasma ketoprofen concentrations were under the limit of quantification (0.0025 μg ml^–1) at delivery (24 h after drug discontinuation) in the case presented. (B) DA constriction. The line indicates predicted DA diameter (% control) in the case presented

Discussion

We successfully calculated the transplacental PK parameters for ketoprofen and antipyrine using the model constructed for the study (Table 1). The influx clearance from mother to placenta (K1’), corrected for fb in the perfusate, was 0.368 ml min^–1 per gram of cotyledon for antipyrine and 0.644 ml min^–1 per gram of cotyledon for ketoprofen, indicating that the influx clearance of ketoprofen is about twice that of antipyrine. Existing reported values for the K1’ of salicylic acid and diclofenac are 0.0451 ml min^–1 per gram of cotyledon and 6.27 ml min^–1 per gram of cotyledon, respectively 7, 8, making the K1’ value of ketoprofen about 1/5 to 1/10 that of diclofenac and about 15 times that of salicylic acid. The membrane permeability of drugs that are not substrates of a specific transport mechanism is proportional to the n‐octanol/water partition coefficient and the −1/2 power of the drug's molecular weight 26. For such drugs, we observed a positive correlation (R = 0.942, P < 0.05) between K1’ and [(partition coefficient) / (molecular weight)^½] (see Figure S3). However, for antipyrine and ketoprofen, the K1’ value is greater than the membrane permeability, indicating that a mechanism other than passive diffusion is contributing to the transplacental transfer of ketoprofen from mother to fetus. The involvement of transporter(s) in the transplacental transfer of ketoprofen remains to be investigated in detail.

In the present analysis, the value of the kinetic parameter that represents the elimination of drug from the placenta (ks) was derived from the mean recovery of antipyrine (97.1%). Although this was less than 100%, it was similar to values obtained in our previous report 8. The mean recovery of ketoprofen (96.7%) was also less than 100%. The amount of ketoprofen in cotyledon at the end of the perfusion studies was also included in the recovery. Ketoprofen is known to be glucuronidized in the liver 16, and the placenta is known to express a glucuronyl transferase (UGT2B family) 27. Therefore, ketoprofen may undergo glucuronidation in the placenta, although there are no reports dealing directly with the metabolism of ketoprofen in the placenta, and we did not detect any evidence of metabolism, such as metabolites in the effluent.

Model development required various assumptions because relatively little information is available regarding the fetus, and this was a limitation of the present analysis. The rationale for key assumptions was as follows. Firstly, it was assumed that the fetuses have no metabolic capability (kef = 0). This is considered plausible because the expression levels of UGT isozymes in the fetus are much lower than in the mother 19. An advantage of this assumption is that it should make a false‐negative assessment of a fetal risk unlikely, as shown in Figure S4. This figure also shows that the ketoprofen concentration profile in the fetus after suppository‐based administration is more susceptible to the influence of the value of kef than that after patch administration, probably because the drug is continuously released from the patch. Secondly, it was also assumed in this model analysis that the concentration–response relationship for constriction of the DA in humans is equivalent to that in rats. This may be reasonable because the response is mediated by the prostaglandin E2 receptor 28, for which the amino acid sequence homology between rats and humans is as high as 82% 29, although no comparative data are yet available regarding prostaglandin E2 receptor activity between rats and humans. In addition, the half‐maximal inhibitory concentration (IC ₅₀) value of ketoprofen for cyclooxygenase (COX)‐2, which is more involved than COX‐1 in the DA constriction 30, was reported, in an ex vivo assay using whole human blood, to be 0.88 μM 31 or 2.9 μM 32, both of which values are similar to the unbound EC ₅₀ value [0.276 μg ml^–1 = 1.09 μM after correction for the unbound drug ratio (fu,f = 0.014) in fetal plasma] obtained in the current study. These considerations support the validity of extrapolating PD parameters obtained from the present in vivo animal experiments to humans.

To date, the fetal toxicity (DA constriction) of ketoprofen in human fetuses has been evaluated by extrapolation from animal experiments, based on oral dosage based on body weight 9. The present study was the first to report that ketoprofen‐related fetal DA constriction is delayed with respect to fetal drug concentration (Figure 5), as observed in our in vivo measurements. The transfer of ketoprofen to the fetus is rapid, and the peak concentration is reached 1 h after administration. However, strong DA constriction does not appear until about 4 h after administration. The fetal DA maintains patency through the action of prostaglandin, which is both present in circulating blood and generated in the endothelial cells and smooth muscle that constitute the DA itself 33. If local prostaglandin production is inhibited by ketoprofen and this process is delayed, the transfer of ketoprofen to DA tissue may become the rate‐limiting factor. It is also possible that a decreased level of prostaglandin circulating in the blood is the rate‐limiting factor. The primary source of prostaglandin in the circulating blood is the placenta 34, and the transfer of ketoprofen is rapid, both to the placenta and the fetus, so prostaglandin production is likely to be reduced rapidly. However, it seems likely that that a slower rate of prostaglandin elimination from the circulating blood would result in delayed expression of the fetal DA constriction. In fact, prostaglandin undergoes little metabolism in fetuses, which are generally considered to have higher levels in the circulating blood than the mothers 35.

The model established in the present study predicts an inner DA diameter of about 40% following repeated third‐trimester use of oral or suppository ketoprofen formulations (Figure 6), both of which have been reported to cause DA constriction‐related fetal toxicity in humans. Further, our model predicts a continuous inner DA diameter of up to 50% following repeated use of a 160 mg patch formulation, suggesting fetal toxicity equivalent to that in the case of systemic administration.

The fetal DA diameter was 34% and 55% of the standard inner diameter (observed and estimated values, respectively) for up to 20 h following discontinuation of ketoprofen use (Figure 7). By contrast, the fetal plasma concentration of ketoprofen 20 h after discontinuation is calculated to be 4.3 ng ml^–1, significantly lower than the EC ₅₀ of 0.107 μg ml^–1 after correction for the unbound drug ratio in the human fetal plasma. Therefore, the developed model incorporating the effect compartment with a delay successfully described the fetal sustained toxicity after discontinuing transdermal administration of ketoprofen, as observed in the case presented here.

In conclusion, the human fetal DA‐constrictive profiles of ketoprofen after various routes of administration to the mother have become quantitatively predictable, based on the experimental results of human placental perfusion and drug administration to pregnant rats. Our model provides support for the view that fetal toxicity can occur in response to not only systemic, but also transdermal administration of ketoprofen, and accounts well for the sustained fetal toxicity after discontinuing the administration of ketoprofen that was observed in our case.

Competing Interests

The authors declared no conflict of interest.

This study was supported in part by grants from JSPS KAKENHI Grant Number 19390040.

Supporting information

Supplemental figure legends

Figure S1 Pharmacokinetic models of antipyrine (A) and ketoprofen (B) transfer across the placenta. In both models, a dead volume compartment was incorporated into the maternal circulation to take into account the fluid volume in the trifurcated glass pipe. Cin, drug concentration into the compartment; Cf, drug concentration in the fetal venous compartment (µg ml^–1); Cm, drug concentration in the intervillous compartment (µg ml^–1); K1 and K4, influx clearance (ml min^–1 per gram of cotyledon); k2 and k3, efflux rate constants (min^–1); ka, first‐order rate constant (1.02 min^–1); ks, elimination rate constant (min^–1), Qf, fetal perfusion flow rate (ml min^–1 per gram of cotyledon); Qm, maternal perfusion flow rate (ml min^–1 per gram of cotyledon); Vf: fetal vascular volume in placental tissue (0.06 ml per gram of cotyledon); Vm, maternal volume in the perfusion chamber (ml per gram of cotyledon); Vp: apparent volume of placental compartment (ml min^–1 per gram of cotyledon); Xp, amount of drug in the placental compartment (µg per gram of cotyledon)

Figure S2 Frontal cuts of the fetal ductus arteriosus in pregnant rats at term (day 21) without medication (control) (A), 1 h after ketoprofen (1 mg kg^–1) administration (B) and 4 h after ketoprofen (1 mg kg^–1) administration (C). AoA, aortic arch; DA, ductus arteriosus; LPA, left pulmonary artery

Figure S3 Relationship between physicochemical properties and influx clearance into the placenta of ibuprofen, salicylic acid, antipyrine, diclofenac and ketoprofen. D7.4 represents the 1‐octanol/water partition coefficient at pH 7.4. R, correlation coefficient

Figure S4 Simulation of the fetal plasma ketoprofen concentration (after a 75 mg suppository or 80 mg plaster administration) for various values of the elimination rate constant for elimination from the fetal compartment

Tanaka, S. , Kanagawa, T. , Momma, K. , Hori, S. , Satoh, H. , Nagamatsu, T. , Fujii, T. , Kimura, T. , and Sawada, Y. (2017) Prediction of sustained fetal toxicity induced by ketoprofen based on PK/PD analysis using human placental perfusion and rat toxicity data. Br J Clin Pharmacol, 83: 2503–2516. doi: 10.1111/bcp.13352.