Controlled Drug Release from an Ocular Implant: An Evaluation Using Dynamic Three-Dimensional Magnetic Resonance Imaging

Abstract

purpose. The ability of an episcleral implant at the equator of the eye to deliver drugs to the posterior segment was evaluated, using a sustained-release implant containing gadolinium-DTPA (Gd-DTPA). The movement of this drug surrogate was assessed with magnetic resonance imaging (MRI) in the rabbit eye. The results were compared with a similar implant placed in the vitreous cavity through a scleral incision at the equator.

methods. Polymer-based implants releasing Gd-DTPA were manufactured and placed in the subconjunctival space on the episclera or in the vitreous cavity in live rabbit eyes (in vivo) and in freshly enucleated eyes (ex vivo). Release rates of implants in vitro were also determined. Dynamic three-dimensional MRI was performed using a 4.7-Tesla MRI system for 8 hours. MR images were developed and analyzed on computer.

results. Episcleral implants in vivo delivered a mean total of 2.7 μg Gd-DTPA into the vitreous, representing only 0.12% of the total amount of compound released from the implant in vitro. No Gd-DTPA was detected in the posterior segment of the eye. Ex vivo, the Gd-DTPA concentration in the vitreous was 30 times higher. In vivo eyes with intravitreal implants placed at the equator delivered Gd-DTPA throughout the vitreous cavity and posterior segment. Compartmental analysis of the ocular drug distribution from an episcleral implant showed that the elimination rate constant of Gd-DTPA from the subconjunctival space into the episcleral veins and conjunctival lymphatics was 3-log units higher than the transport rate constant for Gd-DTPA movement into the vitreous.

conclusions. In vivo, episcleral implants at the equator of the eye did not deliver a significant amount of Gd-DTPA into the vitreous, and no compound was identified in the posterior segment. A 30-fold increase in vitreous Gd-DTPA concentration occurred in the enucleated eyes, suggesting that there are significant barriers to the movement of drugs from the episcleral space into the vitreous in vivo. Dynamic three-dimensional MRI using Gd-DTPA, and possibly other contrast agents, may be useful in understanding the spatial relationships of ocular drug distribution and clearance mechanisms in the eye.

Advances in biomedical engineering and ocular surgical techniques have encouraged the development of sustained-release drug delivery implants to treat a variety of ophthalmic diseases. ^{1 2} Polymer-based drug delivery implants placed in the vitreous cavity have efficacy in treating posterior segment diseases such as cytomegalovirus retinitis and diabetic macular edema ^{3 4} ; however, complications include vitreous hemorrhage, retinal detachment, and endophthalmitis. ^{5 6} To avoid these sight-threatening complications, investigators have suggested placing drug delivery implants in the subconjunctival space at the equator of the eye where the sclera in humans is relatively thin and easily accessible, and through which drug molecules may readily diffuse into the eye. ⁷ However, the movement of drugs from an episcleral location at the equator of the globe to reach the macula has not been well investigated. The standard preclinical techniques of assessing ocular drug distribution—autoradiography and fluorescein labeling—require euthanasia of the animals at various time points, with serial sectioning of the eye to reenact the movement of the drug molecules in vivo. ⁸ Standard techniques do not permit serial time studies and determination of drug transport gradients in the eyes of individual animals. Fluorophotometry is limited to assessments of the movement of fluorescent compounds in the aqueous and vitreous humor. ⁹ We sought to develop a technique to evaluate the ocular distribution over time and elimination pathways of a drug surrogate from a sustained-release implant in normal rabbit eyes using magnetic resonance imaging (MRI). To this end, we developed a sustained-release implant that releases the contrast agent gadolinium-diethylenetriaminopentaacetic acid (Gd-DTPA, molecular mass 938 Da; Magnevist, Berlex, Inc., Montville, NJ). These implants were evaluated separately in the subconjunctival space on the episclera at the equator and in the vitreous cavity. The movement of the drug surrogate was assessed using dynamic three-dimensional MRI. 

Materials and Methods

Implant Manufacturing

An episcleral implant was manufactured using the following procedure (Fig. 1) : A 10% polyvinyl alcohol (PVA; wt/vol) solution was formulated by placing 1.0 g PVA (Celvol; Celanese Chemicals, Ltd., Dallas, TX) in 10 mL molecular biology grade water (BIOfluids; Biosource International, Camarillo, CA) in a 50-mL polypropylene conical tube (Falcon; BD Biosciences, Franklin Lakes, NJ) and placing it in a water bath at 100°C for 3 hours to dissolve all the PVA. The PVA solution was cooled down to room temperature and 0.71 mL of 0.5 M Gd-DTPA solution was added and stirred into the PVA mixture to produce a solution of 10% PVA and 25% Gd-DTPA (wt/wt/vol). Eight milliliters was poured onto a glass plate that produced a thin film as it dried at room temperature. A biopsy punch (AcuPunch 4 mm; Acuderm Inc, Ft. Lauderdale, FL) was used on the dried film to make 4-mm diameter disks. Individual disks (eight total) were placed in a polytetrafluoroethylene mold, which has a 4.3-mm diameter and a 1.4-mm depth. The disks were coated with the remaining 10% PVA and 25% Gd-DTPA (wt/wt/vol) solution. The dried implant was peeled out of the polytetrafluoroethylene mold. Each episcleral implant contained a total of 6.4 mg Gd-DTPA. 

An intravitreal implant was manufactured as just described, except less Gd-DTPA was loaded per implant to prevent oversaturation in the vitreous. To accomplish this, a 15% PVA and 20% Gd-DTPA (wt/wt/vol) solution was substituted, and a smaller implant was made with 3-mm disks and mold dimensions of 3.2-mm diameter and 1.2-mm depth with a suture stub extension on one side of the implant for scleral fixation. A total of 3.7 mg Gd-DTPA was loaded into each intravitreal implant. 

Implant In Vitro Release Rate Determination

As previously reported, ¹⁰ the release of Gd-DTPA from both the episcleral and intravitreal implants was equal from all external surfaces except at the suture platform in the intravitreal implant. For the batch of implants used in this study, bulk release was determined by placing the implants in 25-mL glass vials with 20 mL phosphate-buffered saline (PBS; pH 7.4) and stirring with a magnetic bar at 150 rpm at room temperature. The solution was assayed every 10 minutes for 1 hour, every 30 minutes for the next 1 hour, and hourly for the next 6 hours. The solution was completely replaced after each assay with fresh PBS to simulate sink conditions. The Gd-DTPA assays were performed using a spectrofluorometer (QuantaMaster operated by FeliX, ver. 1.21; Photon Technology International, Lawrenceville, NJ). A calibration curve was made using a 275-nm excitation wavelength and a 312-nm emission wavelength and the lowest detectable concentration of Gd-DTPA was 0.5 × 10⁻⁵ M. The amount of Gd-DTPA in the solution at each time point was measured and summated to calculate the cumulative release of Gd-DTPA from each implant. 

Magnetic Resonance Imaging

Dynamic three-dimensional MRI was used to image the rabbit eyes. All imaging experiments were performed on a 4.7-Tesla MRI system (Bruker Instruments, Billerica, MA) and the images analyzed on computer (MatLab, ver. 6.5; The MathWorks Inc., Natick, MA; and Amira, ver. 2.3; TGS Inc., San Diego, CA). MR images shown in the figures were coronal views extracted from the three-dimensional images to include the lens and implant location. The posterior segment of the rabbit eye was defined as including all tissues in the posterior one third of the eye. 

In Vivo Experiments.

New Zealand White (NZW) rabbits weighing 2 to 3 kg were used (Covance Laboratories, Inc., Vienna, VA) according to the guidelines set forth in the ARVO statement for the Use of Animals in Ophthalmic and Vision Research. Each rabbit was maintained under general anesthesia during the 8-hour imaging period. Proparacaine 1% ophthalmic drops (Allergan America, Hormigueros, PR) were used topically on the eye, and an incision was made through the conjunctiva and Tenon’s fascia in the superotemporal quadrant at the equator of the right eye. The equator of the rabbit eye was defined as the point at which the sclera starts the downward slope toward the optic nerve, approximately 5 to 6 mm posterior to the limbus. ¹¹ The episcleral implant was placed directly on the episclera at the equator and the Tenon’s fascia and conjunctiva was reapproximated using an 8-0 Vicryl suture. For intravitreal placement, a 4-mm sclerotomy was performed at the equator, and the implant was inserted into the vitreous cavity and sutured to the sclera using an 8-0 Vicryl suture. The sclerotomy was closed using an 8-0 Vicryl suture, and the conjunctiva was reapproximated to the limbus using a 10-0 Vicryl suture. 

The rabbit was positioned in the scanner, and the head was centered in a 10-cm diameter volume coil and imaged for 8 hours. Complete T₁-weighted fast spin echo three-dimensional images were acquired over 20-minute intervals. The imaging parameters were recovery time/echo time, (TR/TE) 267/9 ms; field of view (FOV), 9 × 9 × 9 cm; acquisition matrix, 256 × 128 × 128, two averages; and echo train length, 8. A total of 256 adjacent image slices, each with a width of 350 μm, were made in the FOV. An in vivo control eye (no implant) was imaged and a representative coronal section was recorded. 

Because local toxicity to the eye from the implant may alter the pharmacokinetics of Gd-DTPA, a separate group of anesthetized rabbits had either an episcleral or intravitreal Gd-DTPA implant inserted at the equator of the right eye, using the same methods described earlier. After 8 hours, the rabbits were euthanatized, the enucleated eyes fixed in 10% formalin, and the eyes processed for histopathology, using our previously described methods. ¹²  

Ex Vivo Experiments.

NZW rabbits were anesthetized and euthanatized with an intracardiac pentobarbital overdose. The right eye including the conjunctiva, Tenon’s fascia, extraocular muscles, and intraconal fat was removed with sharp dissection. The implants were placed on the episclera or vitreous cavity as described earlier. The eye was immediately wrapped with a 4 × 4-in. gauze pad moistened with PBS and placed in a sealed 50-mL conical tube. The tube was positioned in a 7.2-cm diameter volume coil and imaged every 10 minutes for 8 hours. The parameters were TR/TE, 259/6.6 ms; FOV, 5 × 5 × 4 cm; acquisition matrix size, 256 × 128 × 128, 1 average; and echo train length, 8. A total of 256 adjacent image slices, each with a width of 160 μm, were made in the FOV. An ex vivo control eye (no implant) was imaged, and a representative coronal section was recorded. 

Quantitative Analysis Gd-DTPA from MR Images

To measure Gd-DTPA concentrations directly from gray scale T₁-weighted MR images, standard solutions of Gd-DTPA were scanned along with the rabbit eye. The following concentrations of Gd-DTPA were prepared in a 2% hydroxypropyl methylcellulose (Methocel; Dow Chemical Co., Midland, MI) solution: 1.0 × 10⁻¹ M, 0.5 × 10⁻¹ M, 1.0 × 10⁻² M, 0.5 × 10⁻² M, 0.25 × 10⁻² M, 1.0 × 10⁻³ M, 0.5 × 10⁻³ M, 1.0 × 10⁻⁴ M, 0.5 × 10⁻⁴ M, 1.0 × 10⁻⁵ M, 0.5 × 10⁻⁵ M, and 1.0 × 10⁻⁶ M. The 12 standard solutions were poured into 12 individual wells cut from a standard 96-well plate culture chamber and sealed with a silicone adhesive. Gray scale MR images of the standard solutions were developed and average intensity value of each concentration was determined using ImageJ (ver 1.27z; available by ftp at zippy.nimh.nih.gov/ or at http://rsb.info.nih.gov/nih-imageJ; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD). 

Results

In Vitro Release Rate

The in vitro release of both the episcleral (n = 4) and intravitreal implants (n = 4) demonstrated an initial burst with a declining diffusion phase (Fig. 2) . This release pattern is typical of a matrix implant with release kinetics governed by diffusion from dispersed drug in a polymer. ¹³ The episcleral implant released more total drug because of the increased drug loading in the manufacturing process. A total of 4.7 ± 0.49 mg and 3.8 ± 0.14 mg were released over 8 hours by the episcleral and intravitreal implants, respectively, and 99% of the release occurred in the first 5 hours of the assay. 

Calibration Curve for MRI Quantitative Analysis

The relationship of Gd-DTPA concentration and image intensity values were determined during each experiment by scanning standard Gd-DTPA solutions along with the eye. The lower detection limit of the Gd-DTPA was 0.5 × 10⁻⁵ M and the intensity signal initially increased with concentration up to 0.5 × 10⁻² M (Fig. 3) . The intensity signal decreased with Gd-DTPA concentration beyond 0.5 × 10⁻² M because of T₂ shortening effects. ¹⁴ Therefore, areas of very high drug concentrations immediately around the implant could not be quantified. The high concentration of Gd-DTPA and the solid nature of the implant combined to render it black on the images. 

MR Images of In Vivo and Ex Vivo Control Eyes

The in vivo MR image of a control eye showed little signal intensity in the vitreous and aqueous, reflecting good T₁ saturation of the tissue water molecules (Fig. 4A) . The ex vivo MR image of a control eye showed a higher signal relative to the surrounding tissue (Fig. 4B) , due in part to postmortem hardening of the tissue and a slightly different TR/TE for the image. The effective background, however, was within experimental error for detection of Gd-DTPA and showed a less than 1% difference from the calibration standard in the image. 

Episcleral Implant: Ocular Drug Distribution

In the in vivo experiments with episcleral implants placed at the equator of the eye (n =2), Gd-DTPA was first observed in the vitreous and aqueous humor after a mean of 84.3 ± 12.7 and 115.7 ± 28.4 minutes, respectively (Figs. 5A 5B) . Over the 8-hour scan, the total mean cumulative amount of Gd-DTPA measured in the vitreous and aqueous humor was 2.7 ± 0.1 and 3.1 ± 2.1 μg, respectively. The mean cumulative amount that entered the vitreous and aqueous (i.e., 5.8 μg) was only 0.12% of the total cumulative amount released by the implant over an 8-hour period in vitro (Fig. 2) . No Gd-DTPA was present in the posterior segment of the eye. A gradual accumulation of Gd-DTPA was detected in the right buccal lymph node in both rabbits (Fig. 6) . In the ex vivo experiment with episcleral implants (n = 2), one rabbit had an implant placed at the equator, the other between the equator and the optic nerve. With the implant placed at the equator, Gd-DTPA was rapidly detected in the vitreous within 30 minutes and in the anterior chamber in 60 minutes (Figs. 5C 5D) . Gd-DTPA was present throughout the vitreous cavity and retina with a cumulative amount of 81.3 μg in the vitreous and 67.8 μg in the aqueous humor, after 8 hours of scanning. In this same time frame, the more posterior implant delivered proportionately more Gd-DTPA into the vitreous (196.6 μg) than into the aqueous humor (8.7 μg; Figs. 5E 5F ). 

Compartmental analysis of the ocular drug distribution from an episcleral implant at the equator in vivo was performed. Based on first-order elimination kinetics, a three-compartment model of the eye was made using previously described methods, with the following rate constants ¹⁵ :

•  
  k ₁: rate constant for transfer of Gd-DTPA from the subconjunctival space directly into the aqueous humor;
•  
  k ₂: rate constant for transfer of Gd-DTPA from the subconjunctival space directly into the vitreous humor;
•  
  k ₃: rate constant for elimination of Gd-DTPA from the aqueous humor through Schlemm’s canal to the aqueous veins;
•  
  k ₄: rate constant for elimination of Gd-DTPA from the vitreous (i.e., movement through the retina or anteriorly into the aqueous humor);
•  
  k ₅: rate constant for elimination of Gd-DTPA from the subconjunctival space (i.e., movement into the episcleral veins and conjunctival lymphatics); and
•  
  R(t): the in vitro release rate of Gd-DTPA from the implant.

With the three-compartment model, three first-order ordinary differential equations were developed and solved simultaneously to find k. We assumed the in vivo release rate was the same as the in vitro release rate of the implant and the compartments were homogenous  

\[\frac{dM_{\mathrm{A}}}{dt}\ {=}\ k_{1}M_{\mathrm{S}}\ {-}\ k_{3}M_{\mathrm{A}}\]

 

\[\frac{dM_{\mathrm{V}}}{dt}\ {=}\ k_{2}M_{\mathrm{S}}\ {-}\ k_{4}M_{\mathrm{V}}\]

 

\[\frac{dM_{\mathrm{S}}}{dt}\ {=}\ {-}\ (k_{1}\ {+}\ k_{2}\ {+}\ k_{5})M_{\mathrm{S}}\ {+}\ R(t)\]

 

where M _S is the total mass of Gd-DTPA in the subconjunctival space, M _V is total mass of Gd-DTPA in the vitreous humor, M _A is the total mass of Gd-DTPA in the aqueous humor, and R(t) is the in vitro release rate of Gd-DTPA from the implant. 

The rate constant k essentially describes the facility of drug movement between compartments—the higher the number, the greater the mass movement. The k values reported were the best fit for the experimental data of M _A and M _V:

•  
  k ₁: 0.0046/h
•  
  k ₂: 0.0047/h
•  
  k ₃: 0.0043/h
•  
  k ₄: 0.0713/h
•  
  k ₅: 2.9900/h

Intravitreal Implant: Ocular Drug Distribution

In vivo, Gd-DTPA released into the vitreous and a concentration gradient was present in the vitreous cavity with higher levels present behind the lens and lower levels at the vitreoretinal interface (Fig. 7A 7B) . There was no difference in the concentration of Gd-DTPA over the medullary rays, where retinal vessels are present, compared with the surrounding avascular retina. This Gd-DTPA concentration gradient was absent in the images of the ex vivo eyes. In vivo, there was a slightly higher Gd-DTPA signal at the vitreoretinal interface than in the choroid, with minimal compound observed in the sclera (Fig. 8) . In comparison, the ex vivo eyes showed relatively higher concentrations of Gd-DTPA at the vitreoretinal interface and low signals in the posterior retina and choroid (Fig. 8) . In vivo, there was less Gd-DTPA present in the aqueous humor than in the ex vivo eyes. In the ex vivo eyes, the only clear passage of Gd-DTPA anteriorly appeared through the zonules immediately adjacent to the lens into the aqueous humor in the posterior chamber and tended to pool between the posterior iris and the anterior lens capsule. 

Histopathology to Evaluate for Implant Toxicity

Paraffin-embedded sections of the eye stained with hematoxylin and eosin were examined in six rabbits: three with an episcleral implant and three with an intravitreal implant. The histopathologic appearance shown by light microscopy in all eyes showed a normal cornea, anterior chamber, iris, lens, ciliary body, and vitreous. In the superotemporal quadrant where the implant was inserted, the conjunctiva showed a mild mixed inflammatory response, and the retina showed the expected mechanical injury around the site of the incision in the eyes with the intravitreal implant. However, the remaining quadrants showed normal retina, choroid, and sclera. The optic nerve and medullary rays were normal in all eyes. 

Discussion

Episcleral implants at the equator of the eye did not deliver a significant amount of Gd-DTPA into the eye, and no compound was detected in the posterior segment. A 30-fold increase in vitreous Gd-DTPA concentration occurred in the enucleated eyes, suggesting that reducing the barriers to flux can increase the movement from the episclera into the vitreous. These barriers include elimination pathways such as episcleral and conjunctival venous blood flow, ¹¹ conjunctival lymphatic drainage, ¹⁶ choroidal blood flow, ¹⁷ and counterdirectional convection currents caused by the difference in hydrostatic pressures between the suprachoroidal tissue and the episcleral veins. ^{18 19} Our study did not specifically identify the principal elimination pathway of Gd-DTPA movement from the episclera, but collectively, these pathways have an rate constant for elimination (k ₅) 3 log units higher than the rate constant for transfer into the vitreous (k ₂). In our study, the presence of a high concentration of Gd-DTPA in the buccal lymph node, the proximal portion of the cervical lymph drainage system, ²⁰ was consistent with previous work showing radiolabeled compounds being eliminated from the subconjunctival space. ²¹ The relatively low flow rates of the lymph fluid in this region enables imaging of pooled Gd-DTPA in the lymph node. ¹⁶ Entry and elimination of Gd-DTPA from the episcleral veins is difficult to image because of the rapid dilution of Gd-DTPA as it enters the blood stream. 

The movement of Gd-DTPA from the episcleral space at the equator of the globe into the eye was predominantly through the ciliary body region. Gd-DTPA transiting through the ciliary body from the episcleral would less likely be eliminated by venous drainage, since flow rates are comparatively lower than the posterior choroidal flow in most species, ²² including the rabbit. ²³ Other barriers to fluid and drug flow from the episclera to the aqueous humor, such as posteriorly directed fluid currents from uveoscleral flow, are negligible in the rabbit, ²⁴ and lymphatic drainage is absent in the ciliary body and choroid. ¹⁷ Facilitators of fluid and drug movement from the stroma of the ciliary body into the aqueous humor are the nonpigmented cells of the ciliary epithelium where transport occurs through the tight junctions, ²⁵ by pinocytosis, ²⁶ through osmotic gradients, ²² and through a carrier-mediated active transport system. ²⁷ The passage of Gd-DTPA in the ciliary body region suggests that drugs targeting this region, such as aqueous humor production suppressants for glaucoma therapy, may be delivered using episcleral implants. 

Implants placed directly in the vitreous cavity at the equator in live rabbits delivered Gd-DTPA, producing concentration gradients in the vitreous that were decreased toward the vitreoretinal interface. Based on previous experiments in which fluorescein compounds were injected in the vitreous and sections of frozen eyes were examined, a gradual reduction in the vitreous concentration contours of sodium fluorescein from anterior to posterior was suggestive of a transretinal elimination pathway. ^{8 28} The Gd-DTPA concentration gradients in the vitreous were reduced when the implants were imaged in the vitreous of enucleated eyes and the concentrations were markedly reduced at the level of the RPE and the choroid. This suggests that the major transport mechanisms are inactivated when the eye is enucleated. Mechanisms that encourage transretinal fluid and drug movement include the higher hydrostatic pressures in the anterior vitreous compared with the episcleral space, ²⁹ oncotic pressure gradients created by the proteinaceous extracellular fluid of the choroid, ^{22 30} and rapid clearance through the choroidal blood flow. ³¹ The movement of Gd-DTPA released from an intravitreal implant through the neurosensory retina was observed in both in vivo and ex vivo eyes, consistent with the lack of tight junctional barriers in the neurosensory retina. ³² Although choroidal tissue fluid can leave the eye through the sclera, either in or around perivascular spaces, ^{21 24 33} in the in vivo intravitreal implant eyes, no Gd-DTPA was detected in the posterior sclera, suggesting that the elimination of Gd-DTPA was predominantly through the choroidal circulation. The minimal Gd-DTPA movement beyond the posterior retina in the ex vivo eyes suggested that the tight junctions between the RPE cells were intact and that the clearance mechanisms, such as choroidal blood flow, were inactivated. 

Gd-DTPA can move through the vitreous in vivo by flow-induced convection and by diffusion. The relative influence of these two mechanisms can be characterized by the Péclet number (Pé):  

\[\mathrm{P}{\acute{e}}\ {=}\ V\ {\cdot}\ L/D\]

 

where V is the velocity of water in vitreous (3.3 × 10⁻⁷ cm/s), ²⁸ L is the radius of the retina of rabbit (1 cm), ³⁴ and D is the diffusion coefficient of Gd-DTPA in water (3.8 × 10⁻⁶ cm²/s). ³⁵ The Péclet number is therefore 0.087. 

This Péclet number suggests that the movement of Gd-DTPA in the vitreous was predominantly by diffusion, consistent with other studies examining the movement of small molecular weight compounds in the vitreous. ¹⁹  

There was more rapid movement of Gd-DTPA in the aqueous and vitreous humor in the ex vivo eyes with an intravitreal implant compared with the in vivo eyes because of the lack of transretinal clearance from the vitreous and loss of aqueous humor drainage. The ex vivo eyes showed a relative obstruction of Gd-DTPA diffusion from the vitreous cavity to the posterior chamber, probably due to the dense vitreous condensation in this region. ^{36 37} The pooling of Gd-DTPA in the posterior chamber of these eyes was due to the strong iris–lens apposition in rabbits where the iris acts as a one-way valve permitting anterior flow only. ³⁸  

There are several differences in the anatomy and physiology of the rabbit eye that have to be considered before extrapolating the results of this study to humans. ¹¹ Although the permeabilities of rabbit and human sclera to several of compounds are similar, ³⁹ mean scleral thickness is less in the rabbit ¹¹ than in the human, and this can have an effect on transscleral drug transport. ⁴⁰ The choroidal vessel permeability and flow is different in rabbits than in primates. ^{17 41 42} For example, the choroidal flow is 16% higher in rabbits than in monkeys, ^{43 44} which may increase elimination of drug from drug transiting to the vitreous released from an episcleral implant. Although there are minor variations in choroidal blood flow within the same eye of the rabbit, major variations occur in primates where flow in the macula is a log unit higher than in the retinal periphery, ⁴⁵ and this may have an impact on the transit of drug into the eye from an episcleral implant. Another important difference between rabbits and humans is the degree of retinal vascularization—the rabbit possessing a merangiotic (i.e., partially vascularized) retina and the human a holangiotic (i.e., completely vascularized) retina. In rabbit, the MR images in the region of the retinal vessels of the medullary rays showed no difference in signal compared with other areas of avascular retina. Because retinal blood flow is a fraction of choroidal blood flow, ⁴⁶ the amount of Gd-DTPA being eliminated through the retinal circulation may not be significant or it may be below the resolution of this MRI technique. Clearance of drugs through the retinal circulation may be more a factor in humans where the retinal vascularization is complete. Last, because the interactions of Gd-DTPA with ocular pigments were not examined in this study, our pharmacokinetic data in NZW (albino) rabbits may not have general applicability to pigmented mammals. 

There are limitations in the interpretation of the data in this study. Gd-DTPA concentrations could only be quantified between 0.5 × 10⁻⁵ M and 0.5 × 10⁻² M which did not allow for quantification of the high Gd-DTPA concentrations around the implant. Furthermore, with a micromolar detection limit for the resolution of the MRI system, Gd-DTPA in lower concentrations would not have been identified in the ocular tissues. Although there were no tissue toxicities observed by light microscopy in the implanted eyes, the postmortem effect of the ocular tissues on drug movement was not specifically examined in this study. However, in vitro studies accessing ocular tissue drug transport suggest that isolated sclera from rabbits and humans and flat mount preparations of the retinal pigment epithelium and choroid ⁴⁷ can remain structurally ⁴⁰ and functionally ³⁹ viable for several hours. Although the blood flow is terminated once an eye is enucleated, endothelial cell tight junctions can remain functionally intact for several hours after an ischemic insult to the endothelial cells in vitro. ⁴⁸ In addition, the fact that Gd-DTPA movement through the retina was significantly delayed with an accumulation of Gd-DTPA in the vitreous of the enucleated eyes with the intravitreal implants suggests that the outer blood–retinal barrier (i.e., the tight junctions between RPE cells) remained intact. 

Another limitation of this study is that we examined the movement from the implant of a single drug surrogate. Because the molecular mass, molecular radius, and the solubility of a compound in water influence the ocular penetration of a compound, ^{39 40} the pharmacokinetics of Gd-DTPA, a low molecular mass and hydrophilic compound, are not generally applicable to the pharmacokinetics of all drugs in the eye. In addition, an implant with different release kinetics may alter the pharmacokinetic profile, and this was not specifically examined in this study. However, improvements in drug labeling techniques with metal ion chelates with a variety of therapeutic compounds have begun to revolutionize assessments of drug distribution throughout the body using MRI, ^{49 50} and applying these new technologies will improve the general understanding of drug movement in the eye as a function of molecular weight, molecular radius, and solubility. 

In summary, episcleral implants at the equator of the eye did not deliver a significant amount of Gd-DTPA into the vitreous, and no compound was detected in the posterior segment. A 30-fold increase in vitreous Gd-DTPA concentration occurred in the enucleated eyes, suggesting that there are significant barriers to the movement of drugs from the episcleral space into the vitreous in vivo. Dynamic three-dimensional MRI using Gd-DTPA, and potentially metal ion/drug complexes, may be useful in understanding the spatial relationships of ocular drug distribution and clearance mechanisms in the eye. 

 Submitted for publication January 29, 2004; revised March 15, 2004; accepted March 29, 2004.

 Disclosure: H. Kim, None; M.R. Robinson, None; M.J. Lizak, None; G. Tansey, None; R.J. Lutz, None; P. Yuan, None; N.S. Wang, None; K.G. Csaky, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Fabrication procedure for the episcleral implant. (1) PVA/Gd-DTPA solution in mold; (2) PVA/Gd-DTPA dry matrix discs inserted; (3) complex cured; (4) implant removed from the mold with dimensions; (5) the Gd-DTPA episcleral implant.

Figure 2.

 

Cumulative release (milligrams ± SD) over time of Gd-DTPA from episcleral (•) and intravitreal (▪) implants.

Figure 3.

 

Calibration curve from standard solutions showing the relationship of Gd-DTPA concentration and image intensity values in optical density units.

Figure 4.

 

Control eyes (no Gd-DTPA implants). (A) In vivo MRI scan (coronal image). (B) Ex vivo MRI scan (coronal image).

Figure 5.

 

MRI scans (coronal images) of the right eye with an episcleral implant. ( ) Location of implant. Bar at right correlates Gd-DTPA concentration with color in the image. (A) In vivo image 4 hours after implant insertion at the equator. The Gd-DTPA signal is predominantly in the subconjunctival space. (B) In vivo image 7 hours after implant insertion at the equator. Inset: A Gd-DTPA signal is present in the aqueous humor of the posterior chamber (right, large white arrow) and a weaker signal present in the anterior chamber (right, small white arrows). (C) Ex vivo image 4 hours after implant insertion at the equator. A higher Gd-DTPA signal is present throughout the vitreous and aqueous humor compared with the in vivo eye in Figure 4A . (D) Ex vivo image 7 hours after implant insertion at the equator. (E) Ex vivo image 4 hours after implant insertion between the equator and the optic nerve. (F) Ex vivo image 7 hours after implant insertion between the equator and the optic nerve. Because of the posterior location of the implant, a higher Gd-DTPA signal is present in the vitreous cavity (and less in the anterior chamber) compared with an implant placed at the equator in Figure 4D .

Figure 6.

 

In vivo MRI of rabbit’s head showing a combined axial and coronal image through the right eye. A reconstructed surface scan of the rabbit’s head (right inset) shows the image slice orientation. The episcleral implant (red asterisk) in the right eye shows an adjacent Gd-DTPA signal. A high signal is present in the right buccal lymph node.

Figure 7.

 

MRI scan (coronal images) with an intravitreal implant inserted through the equator of the right eye. Bar at right correlates Gd-DTPA concentration with color in the image. (A) In vivo image 4 hours after implant insertion showing movement of Gd-DTPA in the vitreous cavity away from the implant. (B) In vivo image 7 hours after implant insertion. The boxed area of vitreous is magnified (top right) and shows a sharp decline in Gd-DTPA vitreous concentration from the posterior lens capsule to the retinal surface. The graph shows the absolute Gd-DTPA concentrations in the outlined area. (C) Ex vivo image 4 hours after implant insertion showing that Gd-DTPA movement into the inferior vitreous is more rapid than in the in vivo images (A, B). (D) Ex vivo image 7 hours after implant insertion. The outlined area of vitreous is magnified (top right) and shows a small decline in Gd-DTPA vitreous concentration from the posterior lens capsule to the retinal surface. The graph shows the absolute Gd-DTPA concentrations in the boxed areas.

Figure 8.

 

Magnified views of the region of the posterior vitreous, retina, choroid, and sclera in the boxed areas in Figures 7B and 7D . Top: a histologic section of a normal rabbit eye to identify the different ocular tissues in the MR images. Hematoxylin and eosin; original magnification, ×100.

The authors thank Mark Szarowicz and Chris Hillman for providing veterinary technical expertise.