Modeling Approach To Characterize Intraocular Doripenem Pharmacokinetics after Intravenous Administration to Rabbits, with Tentative Extrapolation to Humans

Abstract

The aim of this study was to determine the penetration of doripenem administered intravenously into the rabbit aqueous and vitreous humors. Nineteen New Zealand White rabbits received a 20-mg dose of doripenem intravenously over 60 min. Specimens of aqueous humor, vitreous humor, and blood were obtained 30 min (n = 5), 1 h (n = 5), 2 h (n = 5), and 3 h (n = 4) after the beginning of the infusion and analyzed by high-performance liquid chromatography (HPLC). A pharmacokinetic (PK) model was developed to fit the experimental data. Doripenem concentrations in aqueous humor were lower than those in plasma ultrafiltrates at all sampling times, with an average aqueous humor-to-plasma ultrafiltrate area under the concentration-time curve ratio estimated as 8.3%. A pharmacokinetic model with peripheral elimination described the data adequately and was tentatively used to predict concentration-versus-time profiles and pharmacokinetic-pharmacodynamic (PK-PD) target attainment in patients under various dosing regimens. In conclusion, systematically administered doripenem does not seem to be a promising approach for the treatment of intraocular infections, especially since it could not be detected in the vitreous humor. However, this study has provided an opportunity to develop a new PK modeling approach to characterize the intraocular distribution of doripenem administered intravenously to rabbits, with tentative extrapolation to humans.

INTRODUCTION

Bacterial endophthalmitis (BE) is a severe intraocular disease that may follow surgery, trauma, and bacterial keratitis or may be of endogenous origin. Without appropriate and prompt treatment, between 40 and 70% of patients may have permanent loss of vision (22). Ideally, all cases of BE would be culture proven (5, 16, 24). The identification of the etiologic organisms through culture of intraocular fluids or by PCR (17) may orientate the best therapeutic strategy, but positive-culture rates vary from case to case (5, 16, 24). The bacterial species isolated most frequently from ocular specimens are Staphylococcus epidermidis, Streptococcus sp., coagulase-negative Staphylococcus, Staphylococcus aureus, Propionibacterium acnes, and Gram-negative rods, including Pseudomonas aeruginosa (7). In addition to their intrinsic antibacterial properties, a major obstacle to predicting the efficacy of antibiotics against BE is their ability to reach the infectious site. Intraocular penetration of drugs after systemic administration is limited by the blood-ocular barriers, one regulating exchanges between blood and aqueous humor (AH) (the blood-aqueous barrier) and the second regulating the exchanges between blood and retina (the blood-retina barrier) (11, 12, 26). The blood-aqueous barrier is located primarily in the ciliary iris body and has been described as having two morphological components: an epithelial barrier at the level of the nonpigmented ciliary epithelium and an endothelial component at the level of the iris blood vessels (11). The blood-retinal barrier consists of two distinct elements: an inner barrier, the endothelium of the retinal vessels, and an outer barrier, the retinal pigmented epithelium. Passive diffusion occurs from the intraretinal spaces to the vitreous humor (12). Tight junctions have been shown to be maculae occludentes at the blood-aqueous barrier and zonulae occludentes at the blood-retinal barrier (12). However, systemically administered drugs have poor access to the inside of the eye also because of the presence of drug efflux transporters, such as P-glycoprotein (30). Yet, to date the exact nature of these transporters is not fully established.

Carbapenems are highly active synthetic beta-lactam antibiotics, frequently used as an empirical systemic treatment of BE. However, only a very few studies have been conducted in humans to characterize intraocular carbapenem concentrations after systemic administration (1, 4, 8, 15, 29). With imipenem/cilastatin, 4 daily intravenous (i.v.) injections of 0.5 to 1 g are usually administered (21, 33), whereas 3 daily injections of 500 mg would be enough with doripenem (18, 25), Doripenem is active against a broad range of Gram-positive and Gram-negative bacteria (18), with a spectrum similar to those of meropenem and imipenem. Doripenem has a low molecular weight and a low plasma-protein binding rate and is relatively lipophilic (18, 25), which may favor its penetration into the intraocular fluids (11, 20, 26, 31), with a potential interest for the treatment of intraocular infections. The objective of this study was to characterize the pharmacokinetics of doripenem in the aqueous humor (AH) and vitreous fluid (VF) after i.v. administration to rabbits, with tentative extrapolation to humans.

MATERIALS AND METHODS

Animals.

All the experiments of this study were carried out according to the statements of the Association for Research in Vision and Ophthalmology for the use of animals in ophthalmic and vision research under agreement no. 86 051 (3). New Zealand White rabbits (n = 19) weighing between 1,648 and 2,440 g were acclimatized in wire cages for a minimum of 5 days before the beginning of the experiment.

Doripenem administration.

During experiments, rabbits were housed in contention cages. Topical anesthesia (Xylocaine 5% spray; AstraZeneca, France) was applied on the rabbit's ear before the infusion set was implanted (Microflex infusion set, 25 gauge, 0.5-mm outer diameter; Vygon, Ecouen, France) in the marginal ear vein. Doripenem (Doribax; Janssen-Cilag, Issy-les-Moulineaux, France) was infused intravenously over 60 min at a flow rate of 4 ml · h⁻¹ and at a dose of 20 mg.

Blood, AH, and VF sampling.

Blood and ocular samples were collected after general intramuscular anesthesia with ketamine hydrochloride (Panpharma, Fougères, France) at 35 mg · kg of body weight⁻¹ and xylazine hydrochloride (Bayer, Puteaux, France) at 5 mg/kg. Arterial blood samples from the marginal ear artery (contralateral from the one receiving the antibiotic) and ocular samples (AH and VF) were collected at 30, 60, 120, or 180 min after the beginning of the infusion. Each animal was sampled once. Five rabbits were sampled 30, 60, and 120 min after the beginning of the infusion, and four were sampled at 180 min. Peripheral blood (3 ml) was drawn with an i.v. catheter (BD Insyte-W, 22 gauge, 0.9 by 25 mm; Becton, Dickinson S.A., Madrid, Spain) into heparinized vacuum tubes. Blood samples were immediately centrifuged for 10 min at 1,500 × g at room temperature, and plasma was divided into two aliquots; one was immediately frozen at −20°C, and the other was ultrafiltrated for 30 min at 1,500 × g at room temperature (Centrifree YM-30 systems; Millipore SAS, Molsheim, France). Ultrafiltrates (UF) with unbound antibiotic were stored at −20°C until analysis. Before AH and VF sampling, a topical ocular anesthetic solution of oxybuprocaine (Théa, Clermont-Ferrand, France) was applied. Povidone iodine ophthalmic solution (Meda Pharma, Paris, France) was topically used for disinfection, followed by washing with saline. One hundred microliters of AH of the left eye was sampled by limbal access of the anterior chamber with a 27-gauge needle. Immediately after, 100 μl of VF was sampled from the same eye with a 20-gauge needle. AH samples were immediately frozen at −20°C, and VF samples were first diluted by half in Hanks' buffered salt solution buffer (HBSS; PAN Biotech, Dutscher, Brumath, France) before freezing. Animals were then sacrificed with sodium pentobarbital, 2 ml i.v. (Ceva Santé Animale, Libourne, France).

Quantification of doripenem in UF, AH, and VF.

Levels of doripenem in UF, AH, and VF were assessed by high-performance liquid chromatography (HPLC) (19), in a system consisting of a Gemini C₁₈ security guard cartridge (4 by 2 mm; Phenomenex, California), a C₁₈ Xbridge column (5.0 μm, 150 by 2.1-mm inside diameter [i.d.]; Waters, St-Quentin en Yvelines, France) coupled to a VWR-Hitachi L-2200 autosampler (VWR International, Strasbourg, France), a Jasco UV-1570 (Jasco, Bouguenais, France) UV detector (300 nm), and an EZchrom integrator (EZChrom Elite 3.1; VWR International, Strasbourg, France). The mobile phase consisted of H₂O, formic acid, and acetonitrile (94/0.1/6, vol/vol/vol), and the pump flow rate was 0.25 ml · min⁻¹. The autosampler was set at 4°C, with an injection volume of 15 μl. Ten-point calibration standard curves were obtained with doripenem diluted in HBSS at concentrations ranging between 20 and 0.015 μg · ml⁻¹ (considered the lower limit of quantification in HBSS). The intraday variability of doripenem in HBSS was characterized at 0.015, 0.031, 0.5, and 10 μg · ml⁻¹ with coefficients of variation of 6.3%, 9.8%, 0.8%, and 0.3% (n = 6 for each level of concentration), respectively, and an accuracy of 4.5%, −0.8%, −2.0%, and −1.5% (n = 6 for each level of concentration), respectively. The interday variability was 0.031, 1, and 15 μg · ml⁻¹ with coefficients of variation of 9.3%, 2.1%, and 1.7% (n = 4 for each level of concentration) and accuracies of 13.3%, 0.3%, and 1.7% (n = 4 for each level of concentration), respectively. Concentrations in plasma were determined by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Reversed-phase chromatography was performed on a C₁₈ Xbridge column (5.0 μm, 150 by 2.1-mm i.d.; Waters, St-Quentin en Yvelines, France) coupled with a Gemini C₁₈ security guard cartridge (4 by 2 mm; Phenomenex, California). The mobile phase was 2.5 mM ammonium formate dissolved in 0.1% formic acid, 20% acetonitrile, and 80% water (0.1/20/80, vol/vol/vol). The LC-MS/MS system consisted of a Waters Alliance 2695 separation module equipped with a binary pump, an autosampler (4°C), and a Waters Micromass Quattro Micro-API tandem mass spectrometer. The mass spectrometer was operated in the positive-ion mode. Ions were analyzed by multiple-reaction monitoring (MRM). The transition ions with m/z of 421.2/274.4 were analyzed for doripenem. The transition ion m/z was 384.3/350.1 for meropenem, the internal standard. Eight-point calibration standard curves were prepared with doripenem diluted in rabbit plasma at concentrations ranging between 5 and 0.05 μg · ml⁻¹ and with 4 levels of control (0.05, 0.125, 1.25, and 5 μg · ml⁻¹). The estimated limit of quantification in plasma was 0.05 μg · ml⁻¹. For the preparation of the doripenem calibration standards, controls, and samples, 25 μl of an internal standard solution (5 ng · ml⁻¹) was added to 400 μl of spiked plasma-HBSS (50/50, vol/vol) or to the samples. Mixtures were loaded onto 1-ml Oasis C₁₈ HLB columns (Waters, St-Quentin en Yvelines, France). After washing (water, 0.5 ml), analytes were eluted with 2.5 mM ammonium formate dissolved in 20% acetonitrile and 80% water (20/80, vol/vol). After centrifugation (1,500 × g, 4°C, 10 min), eluates were directly injected (50 μl) into the HPLC system. The intraday variability values of concentrations in plasma were 0.05, 0.125, 1.25, and 5 μg · ml⁻¹, respectively, with coefficients of variation of 5.0%, 6.7%, 7.5%, and 3.0% (n = 6 for each level of concentration) and accuracies of 6.1%, 3.8%, −7.6%, and 1.0% (n = 6 for each level of concentration), respectively.

Pharmacokinetic analysis.

Individual UF and AH values were studied using a naïve pool data type of analysis (n = 19 for each medium), with the software WinNonLin (version 5.3; Pharsight Corporation, Mountain View, CA). Since doripenem protein binding was less than 5% on average in plasma and because protein concentrations in AH are less than 1% of those in plasma (6), it was assumed that there was no binding in AH. Visual inspection of the UF and AH concentration-time data suggested that the unbound area under the curve (AUC_U) in AH was less than that in UF. Data modeling was performed in two phases. In the first step, free plasma (UF) concentration-time data were analyzed using a one-compartment model with zero-order i.v. infusion, leading to the unbound volume of distribution (V_U) and elimination rate constant (k_eU) estimates. Due to its small volume, AH will incorporate extremely reduced amounts of drug from plasma and will consequently not affect the concentration-time profile in plasma. Therefore, V_U and k_eU estimates were fixed, and in the second step, a forcing function was used to estimate AH input and output clearances. Elimination from the AH compartment was added to account for the lower AUC than that in UF, and by the same reasoning to reduce the number of parameters, drug redistribution from AH was included within elimination. The corresponding differential equation for this model was equation 1:

$${\text{dC}}_{\text{AH}}/\text{dt}=\left[\left({\text{CL}}_{\text{in},\text{AH}}\times {\text{C}}_{\text{u},\text{plasma}}\right)-\left({\text{CL}}_{\text{out},\text{AH}}\times {\text{C}}_{\text{AH}}\right)\right]\times \left(1/V_{\text{AH}}\right)$$ (1)

where CL_in,AH and CL_out,AH corresponded to clearances in and out of AH, respectively. The AH volume (V_AH) was set at 0.3 ml in rabbits (10). Then, output clearance from AH to free plasma did not represent transport from AH back to plasma but rather other mechanisms of transport or elimination.

The clearance from the body (CL_U) was estimated as the product k_e · V_U. Corresponding half-lives (t_1/2) were estimated as the ratio ln2/k_eU. Unbound area under plasma concentration-time curves (AUC_U) were calculated as the ratio between doripenem infusion dose and corresponding clearance. At equilibrium, the rate of change of concentrations in AH was zero and therefore

$${\text{CL}}_{\text{in},\text{AH}}\times {\text{C}}_{\text{u},\text{plasma}}={\text{CL}}_{\text{out},\text{AH}}\times {\text{C}}_{\text{AH}}$$ (2)

For a single dose, this could be expressed after integration over time as

$${\text{CL}}_{\text{in},\text{AH}}/{\text{CL}}_{\text{out},\text{AH}}={\text{AUC}}_{\text{AH}}/{\text{AUC}}_{\text{U}}$$ (3)

Parameters were estimated by nonlinear regression using a weighting coefficient of 1/Cpred², where Cpred² was the predicted concentration according to the model. The structural pharmacokinetic (PK) model is presented in Fig. 1.

Fig 1.

Structural pharmacokinetic model.

Pharmacokinetic simulations.

Doripenem concentrations in human AH were simulated after multiple administrations (1-h i.v. infusion of 500 mg every 8 h) by replacing plasma PK parameters in rabbits with human plasma PK parameter values (23), keeping the same AH volume of 0.3 ml in rabbits (10) and humans (27), and the same CL_in,AH and CL_out,AH values across species. AH peak concentrations above MIC (C_max/MIC) and time for AH concentrations above MIC (t>MIC) were predicted for any MIC values. Simulations were performed with Berkeley Madonna software (v 8.3.18; University of California, Berkeley, CA).

RESULTS

The mean unbound fraction value of doripenem in plasma was estimated at 0.96 ± 0.38. Visual inspection of the data (Fig. 2) revealed a reasonably good fit. Measured doripenem concentrations in UF at the end of infusion, corresponding to maximum concentrations, were equal to 10.4 ± 2.1 μg · ml⁻¹, consistent with the value predicted by the model (10.5 μg · ml⁻¹). Doripenem body clearance (CL_U) and volume of distribution (V_U) were equal to 1.7 ± 1.1 liters · h⁻¹ and 0.754 ± 0.074 liter, respectively. The corresponding half-life was 0.31 ± 0.03 h (18.4 min). Doripenem AH concentrations were lower than corresponding UF concentrations at any time. The doripenem peak in AH, observed at 1 h (end of infusion), was measured at 0.49 ± 0.08 μg · ml⁻¹, again close to the value predicted by the model (0.60 μg · ml⁻¹). Doripenem concentrations in UF and AH are presented in Fig. 2, with the corresponding predicted concentrations. The estimated CL_in,AH/CL_out,AH ratio, equivalent to the AUC_AH/AUC_U ratio, was 0.083 (8.3%). Doripenem was not detected in any of the VF samples, with a limit of quantification equal to 0.015 μg · ml⁻¹.

Fig 2.

Measured concentrations of doripenem in plasma ultrafiltrates (○) and in aqueous humor (□), with predicted concentrations in ultrafiltrate (full line) and in aqueous humor (dashed line).

DISCUSSION

Following their systemic administrations in humans, carbapenems have been shown to be distributed to the aqueous humor and the vitreous humor. However, only limited data are available (1, 4, 8, 15, 29), and only two studies were performed using an HPLC assay (8, 29). Bologna et al. (8) used imipenem/cilastatin in one patient prior to cataract surgery and reported an AH concentration of 2.2 μg · ml⁻¹ 1 h after a single intravenous 500-mg dose. Schauersberger et al. (29) administered 2 g of meropenem i.v. to 5 healthy subjects with a mean AH concentration 1 h postinfusion equal to 10.72 μg · ml⁻¹. They also reported a vitreous concentration of 4.3 μg · ml⁻¹ 55 min postinfusion in 1 patient.

We have determined intraocular concentrations of doripenem following systemic administration for the first time, using a pharmacokinetic modeling approach that had never been used before to describe antibiotic intraocular distribution. Yet, it was developed and validated by our group in order to characterize the restricted distribution of carbapenem antibiotics within the brain, most likely due to the presence of efflux transport systems at the blood-brain barrier level (14) or within peritoneal fluid to account for peripheral degradation (13). The much lower AUC in AH than in plasma suggests that doripenem distribution within this liquid is limited by the presence of efflux transport systems (32), and a pharmacokinetic model consistent with this hypothesis was successfully used for data fitting. Another major potential interest of this modeling approach is to allow interspecies extrapolations. Doripenem concentration-versus-time profiles in AH could be modeled in rabbits and then extrapolated to humans using previously published data (23), with the sole assumption that the CL_in,AH/CL_out,AH ratio was constant across species. In other words, it was assumed that factors responsible for restricted AH distribution, presumably active transport systems, were essentially similar in rabbits and humans, which remains to be proven. Furthermore, these experimental data obtained in noninfected rabbits may be extrapolated to patients only if efflux system activity is not affected by infection. However, it is known that during BE, the blood-eye barrier is disrupted, and it could be suggested that the concentrations achieved both in AH and in VF would be higher in such cases (26). Therefore, the same type of study should now be conducted in patients to assess the potential species and/or disease effect on doripenem intraocular pharmacokinetics.

Another potentially interesting aspect of this innovative pharmacokinetic modeling approach is its ability to assess PK/pharmacodynamic (PD) index as a function of MIC and predict treatment efficacy. Antibiotic concentrations measured at a single time point are of limited interest to predict treatment efficacy, as opposed to a PK/PD index such as C_max/MIC for concentration-dependent antibiotics and time over MIC for time-dependent compounds. As with other carbapenems, the efficacy of doripenem is correlated with the time during which its concentration at the target site exceeds the MIC (t>MIC) (2, 25). Considering a time over MIC at least equal to 35% of the dosing interval as a conservative PK-PD target (28), it can be observed that with the usual dosing regimen (500 mg/8 h), this objective could be reached only for bacteria with MIC values lower than 0.5 μg · ml⁻¹ at best, with a slight superiority for relatively long (4-h versus 1-h) infusion durations (Fig. 3). Yet, this objective (time over MIC of >35%) could be obtained by doubling this maintenance dose (1 g/8 h) as occasionally suggested (Fig. 3) (9).

Fig 3.

Fraction of dosing interval (t>MIC [%]) during which doripenem concentrations in aqueous humor remain above MIC versus MIC for various dosing regimens: 0.5 g over 1 h (dashed line), 0.5 g over 4 h (solid line), or 1 g over 4 h (line with long and short dashes).

In conclusion, based on these experimental data, systematically administered doripenem does not seem to be a promising approach for the treatment of intraocular infections, especially since it could not be detected within VF. However, this study has provided an opportunity to develop a new PK modeling approach to characterize the intraocular distribution of doripenem administered intravenously to rabbits, with tentative extrapolation to humans, and allows simulations to predict PK/PD target attainment under various dosing regimens.