Transscleral Diffusion of Carboplatin:  An In Vitro and In Vivo Study FREE

RETINOBLASTOMA IS the most common primary intraocular malignancy of childhood, with approximately 200 new cases each year in the United States.^1- 2 Focal treatments, such as cryotherapy, laser photocoagulation, and thermotherapy, have evolved as excellent treatment options, but are limited to small or single tumors.³ Large tumors or vitreous seeds have necessitated enucleation or external beam radiation.^4- 5 External beam radiation effectively controls some tumors when used alone, but not without significant morbidity, including an increased rate of vitreous hemorrhage, cataract formation, midface hypoplasia, vision loss, and second tumor formation in the field of radiation.^6- 8 Because of these factors, chemotherapy (systemically and locally) is the treatment of choice for eyes with significant tumors.

Retinoblastoma has been shown to respond to newer drug regimens and carboplatin has become a cornerstone in the treatment of this disease. However, when given systemically, the physician must be alert to potential adverse effects, including leukemogenic effects, immune incompetence, and death. Localized periocular delivery of carboplatin is an option that could maintain treatment benefits while avoiding these serious adverse effects. Several in vitro studies^9- 10 have confirmed that isolated sclera is permeable to various compounds, such as dexamethasone, methotrexate, and dextran 10, 40, and 70 polymers. Murray et al¹¹ were also able to show that intraocular tumor growth could be inhibited in transgenic murine retinoblastoma using subconjunctival delivery of carboplatin. In addition, serial subconjunctival injections of carboplatin are somewhat effective in the treatment of human intraocular retinoblastoma.¹²

For local cancer treatment, an increased concentration and a longer duration of drug exposure are ideal. This might be achieved by incorporating chemotherapeutic drugs into a sustained-release vehicle. One such medium that might aid in the local delivery of carboplatin is a human-derived fibrin protein sealant, which, on polymerization, provides a semisolid medium for drug delivery. MacPhee et al¹³ showed that human fibrin can serve to deliver fluorouracil concentrations sufficient to kill renal tumor cells. When the anhydrous form of the chemotherapeutic agent was used, delivery was extended up to 120 hours. Typically, the amount of carboplatin available at the injection site is limited by its volume and concentration. However, because the anhydrous form of a drug can be incorporated as a slurry directly into the fibrin sealant, at a concentration above its solubility limit, more drug can be delivered from a smaller volume of vehicle. Approved by the Food and Drug Administration, fibrin sealant is also advantageous as a delivery vehicle because it is formulated to contain human proteins, which minimizes immunogenicity and foreign body reactions and eliminates the need for physical removal of the clot, which is naturally degraded by the body.

It is postulated that retinoblastoma may be better treated by delivery of carboplatin in fibrin sealant. Therefore, one arm of this study involved a series of in vitro experiments intended to evaluate the transscleral permeability of carboplatin from a fibrin sealant vehicle and the commonly used balanced salt solution (BSS) vehicle. The design of the in vitro scleral perfusion chambers used in these experiments simulated a static depot of drug adjacent to the sclera. This controlled setup eliminated variable factors, such as drug dispersion, tear fluid turnover, and blood flow. This allowed for a more direct comparison of the drug diffusion kinetics from each medium and for the determination of whether using a sustained-release vehicle might provide a benefit in delivering carboplatin locally to the eye. The goal in the other portion of this study was to determine carboplatin concentrations in various ocular tissues after subconjunctival administration to the rabbit eye. The aim was to achieve a maximum release of carboplatin by "overloading" the fibrin sealant with significantly more drug than can be ordinarily dissolved in the usual BSS vehicle.

Fibrin sealant (HEMASEEL APR 4-mL kit; Haemacure Corp, Sarasota, Fla) was prepared according to insert instructions with slight modification. The fibrin sealant was supplied as a 2-part kit composed of blood-clotting factors that worked similar to an epoxy glue. One part was a human sealer protein concentrate that was reconstituted with 2 mL of bovine fibrinolysis inhibitor solution, and the other part was human thrombin, 1000 IU, that was reconstituted with 2 mL of calcium chloride solution, 80 µmol. To prepare the sealant, all parts were heated to 37°C and reconstituted with their respective solutions. The protein concentrate was stirred for 10 minutes. The 2 solutions were then loaded into a syringe unit (Duploject; Immuno-US, Rockville, Md), which consisted of a clip for 2 identical syringes and a common plunger that ensured that equal volumes of the 2 components were forced through a common joining piece before being mixed in the application needle. On injection, the 2 solutions polymerized to form a clot at the injection site. For the in vitro experiments, 50 mg of carboplatin (Paraplatin; Bristol-Myers Squibb Co, Princeton, NJ) was incorporated into the fibrin sealant where the thrombin and protein concentrate were reconstituted. For the in vivo experiments, 150 mg of carboplatin was added. This yielded a calculated initial concentration of 12.5 and 37.5 mg/mL for the in vitro and in vivo experiments, respectively.

Scleral tissue was obtained from 10 human donor eyes (Georgia Eye Bank, Atlanta) that had been stored in moist chambers for no longer than 8 days. Scleral preparation and in vitro diffusion setup were performed according to Rudnick et al.⁹ The perfusion apparatus clamped the sclera between a donor chamber on the episcleral side and a receiver chamber on the choroidal side. Approximately 300 µL of carboplatin in either fibrin sealant or BSS (Alcon Laboratories, Inc, Ft Worth, Tex) was applied to the episcleral surface. The choroidal side was perfused continuously with BSS. The perfusate was collected in a fraction collector every hour for 24 hours and analyzed by atomic absorption spectroscopy to determine the concentration of carboplatin that had diffused through the sclera. Based on these data, 2 time-dependent diffusion curves were created. One depicted the absolute amount of carboplatin that diffused across the sclera, and one showed the diffusion of carboplatin as a percentage of the peak. The diffusion constant, K_trans (measured in centimeters per second), was measured at the point of peak flux for carboplatin in each delivery vehicle and was calculated as follows:

K_trans = [R_total/(A × t)] × [1/D],

where R_total equals the total moles through sclera in time t; A, the surface area of the sclera (measured in centimeters squared); t, time (measured in seconds); and D, concentration of original solution in the donor chamber (measured in moles per milliliter).

Dutch Belted rabbits (N = 44; weight, 1.5-2.4 kg) were used to determine ocular tissue levels of carboplatin following subconjunctival injection using either a fibrin sealant or a BSS delivery vehicle. The animals were anesthetized with intramuscular injections of 0.6-mL xylazine hydrochloride, 20 mg/mL, and 0.6-mL ketamine hydrochloride, 100 mg/mL, before injection procedures. One group received 300 µL of fibrin sealant containing carboplatin (approximately 37.5-mg/mL calculated concentration) injected subconjunctivally in the left eye. The injection was made in the superior temporal region, carefully avoiding extraocular muscles. Samples of fibrin sealant from the injection syringes were saved and analyzed to verify the initial carboplatin concentration. The right eye served as a control and remained untouched. Formation of the fibrin sealant clot adherent to the sclera was evaluated immediately after injection by visual and tactile inspection. A second group of rabbits received 300 µL of BSS containing carboplatin (approximately 10-mg/mL calculated concentration) in the subconjunctival space of the left eye. A slitlamp examination was performed and photographs were taken immediately after injection and 2 weeks thereafter. Rabbits were anesthetized as previously described and killed with an overdose of pentobarbital sodium (Euthanasia-5 solution; Henry Schein, Inc, Port Washington, NY), 324 mg/mL, at 1½ hours, 48 hours, and 2 weeks after injection. All animals were handled according to the Association for Research in Vision and Ophthalmology statement for the use of animals in ophthalmology and visual research.

Blood samples were obtained via direct cardiac tap. After enucleation, the aqueous humor was removed from each eye with a 25-gauge needle inserted into the anterior chamber. Eyes were immediately frozen in an acetone and dry-ice bath and stored at −70°C until dissection. The frozen eye was divided with a scalpel into the portion of the sclera directly exposed to the fibrin sealant injection (exposed sclera) and the portion that was not exposed. The remaining fibrin clot was removed from the sclera to which it was adherent. Next, choroid and retina together were peeled away from the attached sclera. The frozen vitreous was then divided in 2, one closest to the fibrin sealant (exposed vitreous) and one further from the fibrin sealant(unexposed vitreous). Blood samples were placed in a heparinized tube and spun down at 1000g for 10 minutes, and 0.5 mL of plasma was removed. All samples were weighed and then dissolved in nitric acid.

Carboplatin concentrations were measured by flameless atomic absorption spectroscopy, using a graphite furnace (model HGA-600; PerkinElmer Inc, Shelton, Conn), with the use of a graphite tube atomizer and an automatic sample dispenser, as described by Madden et al,¹⁴ with minor modifications. A standard curve covering the range of 0 to 1080 µg/L was run at the start and end of each group of samples and after each change to a new graphite tube. Samples were diluted to achieve concentrations within the standard curve. Sample concentrations were determined by comparison of the peak area of the signal with that of the external standards.

All average values are reported as mean ± SEM. The t test was used to determine the significance between mean values. P<.05 was considered significant.

Analysis of presamples from the episcleral side (donor chamber) of the in vitro scleral permeability experiments showed that the carboplatin concentrations were 8.0 ± 0.3 and 6.9 ± 0.2 mg/mL for fibrin sealant and BSS, respectively. Figure 1, A, shows the diffusion of carboplatin across the sclera from each vehicle as a function of time. Figure 1, B, uses the same values to illustrate the amount that diffused across the sclera as a percentage of the peak. At the end of 24 hours, 33.2% ± 1.8% of the initial carboplatin remained in the fibrin sealant and 5.5% ± 1.0% remained in the BSS(Table 1). The peak K_trans of carboplatin from BSS was 27.0 ± 1.7 × 10⁻⁶ cm/s, while the peak K_trans of carboplatin from fibrin sealant was 13.7 ± 2.3 × 10⁻⁶ cm/s (Table 1).

The average initial concentration of carboplatin in fibrin sealant was 24.5 ± 0.7 mg/mL, vs 8.1 ± 0.1 mg/mL in BSS. Table 2 shows concentrations of carboplatin delivered by either fibrin sealant or BSS to the ocular tissues of the treated eyes and found in the blood plasma at 1½ hours, 48 hours, and 2 weeks.

Because injection concentrations were different for BSS and fibrin sealant, carboplatin levels in each tissue were also shown standardized to an initial injection concentration of 10 mg/mL. These corrected drug levels for the treated and untreated eyes are shown in Table 3. Carboplatin as delivered by fibrin sealant was significantly higher in the exposed sclera at 48 hours (P<.01) and in the plasma at 1½ hours (P = .02). Carboplatin levels were higher from the BSS in the unexposed sclera (P = .01) and in the plasma (P = .03) at 48 hours. Also, carboplatin levels from BSS in the untreated eye were significantly higher in the aqueous humor at 1½ (P<.01) and 48 hours (P = .01), the choroid and retina at 48 hours (P = .01) and 2 weeks (P = .01), the sclera at 48 hours (P = .04) and 2 weeks (P = .03), and the vitreous at 1½(P = .03) and 48 hours (P= .01). After enucleation, at all time points, a fibrin sealant bleb could still be noted in the superior-temporal region of the eye. However, with treatments of BSS, this was not observed given that the vehicle disseminated promptly on injection.

The in vitro portion of this study was a preliminary analysis that demonstrated that fibrin sealant could indeed provide a transscleral release of carboplatin. The perfusion chambers used in these experiments held an isolated piece of sclera that kept the carboplatin/vehicle solution static on one side of the tissue. Factors such as drug dispersion, tear fluid turnover, and blood flow were not taken into account and, therefore, these experiments were not designed to be a quantitative predictor of carboplatin levels that could be achieved in vivo. However, they assessed the value of the sealant as a local drug delivery vehicle for carboplatin. To gain an accurate understanding of the levels that could be achieved in vivo, a series of rabbits were given subconjunctival injections of carboplatin in fibrin sealant.

The in vitro diffusion profiles in Figure 1, A, show that the amount of carboplatin that diffused through the sclera from the BSS vehicle was initially higher than the amount that diffused from the fibrin sealant. However, those amounts eventually decreased below the carboplatin levels being released from the sealant. Figure 1, B, depicts how quickly each vehicle was depleted of carboplatin by expressing the values from Figure 1, A, as a percentage of the peak. The curves from Figure 1, B, which were extrapolated to 48 hours, show that the levels of carboplatin were exhausted more rapidly from BSS than from fibrin sealant, illustrating how the sealant was providing a more stabilized drug release. For the in vitro studies, the concentration of drug in each vehicle was chosen to establish a diffusion curve over 24 hours, rather than to maximize drug release. Figure 1, B, shows that, at the current concentration and volume of injection, the projected release of BSS would be 0.2% of the peak, while the fibrin sealant would be at a 22-fold higher level, 4.4%. However, one benefit of the sealant over BSS is that it can be loaded with drug above its solubility limit, to provide higher levels of drug for extended periods. Rather than simple diffusion out of an aqueous drug delivery vehicle, the fibrin sealant becomes a durable reservoir of undissolved drug available for delivery.

The K_trans for each drug delivery system was also an indicator that fibrin sealant provided a more controlled release of carboplatin (Table 1). The K_trans of BSS was significantly higher than that of fibrin sealant, indicating that the carboplatin diffused out of the BSS more quickly, while the lower K_trans for fibrin sealant indicated that the carboplatin was diffusing out more slowly. The drug was held in the fibrin sealant and available for an extended release. This finding was confirmed by our results that showed that a significantly higher amount of carboplatin remained in the fibrin sealant at the end of the 24-hour experiments (Table 1). Not surprisingly, during the 24-hour in vitro experiments, more total carboplatin diffused through the sclera from the BSS. However, the remaining drug in the sealant would still be available for continued diffusion beyond 24 hours. It is possible that, in vivo, with tear dilution, blood flow, and an increased surface area of nonocular tissue to absorb drug, having the carboplatin give a stabilized local release rather than an initial dump of all the available drug could result in more total carboplatin being released into the eye to treat retinoblastoma.

The in vivo portion of this study established quantitative data on the levels of carboplatin reaching several ocular tissues after a subconjunctival injection of each carboplatin delivery medium. To our knowledge, there were no comprehensive data on concentrations of carboplatin established in the rabbit eye through subconjunctival administration using either of these vehicles. However, Mendelsohn et al¹⁵ studied intraocular levels of carboplatin following intravenous vs peribulbar administration in primates. To compare carboplatin levels in the rabbit eye achieved with local administration of fibrin sealant with the levels previously detected in primates achieved with local administration of BSS, a series of rabbits were given subconjunctival injections of carboplatin in BSS at the typical concentration of 10 mg/mL. Mendelsohn et al found that after peribulbar administration of 10 mg of carboplatin (1 mL of 10-mg/mL carboplatin in BSS), aqueous humor levels peaked at 2.0 µg/mL, vitreous levels peaked at 2.38 µg/mL, and plasma levels peaked at 0.89 µg/mL in a 2-hour period. In our rabbit model, we injected approximately 2.4 mg of carboplatin in the BSS (300 µL of 8.1 ± 0.1 mg/mL) using a similar protocol to that of Mendelsohn et al. At 1½ hours, the aqueous humor had a peak concentration of 18.74± 5.83 µg/mL, and the vitreous levels in the portion exposed and unexposed to the injection site were 1.84 ± 0.56 and 0.75 ± 0.14 µg/mL, respectively. The plasma level at 1½ hours was 1.58± 0.14 µg/mL. These levels are similar to those found in primates, except for the level in the aqueous humor, which was most likely higher because of a more anterior injection site. The data imply that carboplatin in BSS is capable of diffusing into the eye in a similar fashion in rabbits and primates.

As previously discussed, the ability to overload drug into fibrin sealant and its ability to provide a semisolid drug reservoir represent the major advantages of this vehicle. Consequently, for the in vivo portion of this study, rabbits given subconjunctival injections of fibrin sealant had an increased load of carboplatin to maximize release to the eye. For the subconjunctival injections of fibrin sealant, approximately 7.35 mg of carboplatin was used(300 µL of 24.5 ± 0.7 mg/mL). With this concentration, the fibrin sealant–injected rabbits had a carboplatin level of 11.83 ± 5.16 and 3.98 ± 1.99 µg/mL in the exposed and unexposed vitreous, respectively, at 1½ hours. Although intravenous carboplatin can deliver drug to the vitreous, of major concern is the inadequacy of this route and the recurrence of vitreous and subretinal seeds despite control of primary retinal tumors. Abramson et al¹⁶ determined the intraocular levels of carboplatin 1 to 2 hours following the intravenous administration of carboplatin in children who were scheduled for enucleation. They found that the mean concentration of carboplatin in the vitreous was 4.05 µg/mL. Given subsequent data showing an efficacy of 14 to 20 mg of carboplatin delivered subconjunctivally, one might extrapolate that 4.05µg/mL is capable of causing regression of vitreous seeds.¹⁰ With the injection of carboplatin in fibrin sealant, a vitreous level of 11.83µg/mL was reached, more than twice the concentration detected by Abramson et al. It is likely that this higher level in the vitreous would be capable of effecting a more complete recession of vitreous seeds with a single injection. The data in this article show that with the fibrin sealant, carboplatin can be detected in the vitreous for up to 2 weeks. However, at this later time point, the exposed and unexposed vitreous levels were near baseline and had increased variability.

Because of where retinoblastoma arises, the choroid and retina sample is clinically relevant. Our results showed that fibrin sealant delivered 396.59± 177.84 µg/mg after 1½ hours and 3.38 ± 1.97 µg/mg after 2 weeks. Unfortunately, few studies have looked at levels of carboplatin that can be reached in the choroid or retina with periorbital injection and, therefore, we have no comparison. To our knowledge, this is the first study in drug delivery that measures the target tissue level, a value that is perhaps the most significant of all.

For the in vivo experiments, the injection concentration of carboplatin was higher in fibrin sealant than in BSS. In an attempt to compare the effectiveness of the 2 delivery vehicles, the drug levels in the eye were standardized to a common initial concentration of 10 mg/mL. In this respect, the levels of carboplatin in sclera unexposed to the injection site were higher from BSS at 48 hours, while the fibrin sealant achieved a significantly higher level in the exposed sclera at 48 hours (Table 3). This is likely because carboplatin injected in BSS is free to disperse around the globe and diffuse into additional nonocular tissues. In contrast, the fibrin sealant retains the drug in a semisolid medium, which acts to reduce dispersion and allow drug diffusion through the sclera at the original injection site. The higher levels of drug found in the untreated eye are another indication that the BSS was allowing a greater dispersal of carboplatin. Forster et al¹⁷ confirmed an arterial connection between the 2 internal ophthalmic arteries of the rabbit, which may have enhanced the transmission of available drug to the untreated eye. Indeed, significantly higher levels of carboplatin from BSS were found in every tissue of the contralateral eye for at least 2 of the 3 time points(Table 3). This artery is absent in humans, and one would expect less carboplatin to reach the contralateral eye.

One potential drawback to using fibrin sealant as a local delivery vehicle for carboplatin is the possibility of reaching levels beyond the toxicity limit. In the previously mentioned systemic study by Abramson et al,¹⁶ higher intraocular concentrations were obtained in human eyes affected with retinoblastoma than previous unaffected animal models had predicted. This was attributed to possible disruption of the blood-vitreous barrier. Because our study was performed using unaffected rabbits, the amount of carboplatin in the vitreous may reach toxic levels if used to treat humans. Previous data¹⁸ show carboplatin to have a maximum nontoxic dose of 3 µg when given as an intravitreal injection. Other data¹⁹ confirm retinal toxicity at 10µg or higher when administered in a similar manner. In our study, using transscleral delivery, a peak vitreous value of approximately 11.83 µg/mL was achieved, which is slightly above the threshold shown to be toxic. Preliminary toxicity data suggest that there is a temporary decline in retinal function as seen by a transient decrease in electroretinographic amplitude at 2 days posttreatment.²⁰ More extensive electroretinographic studies are under way to determine the long-term toxicity level for increasing concentrations of carboplatin in a fibrin sealant vehicle.

Incorporating carboplatin into fibrin sealant gives a more controlled drug release in addition to providing a stable medium for the drug to reside. While the most favorable concentration of carboplatin to use within the sealant has yet to be determined, the drug can be loaded into this vehicle above its solubility limit, enabling higher amounts to be achieved in ocular tissues. With this in mind, the fibrin sealant vehicle may be a superior mode of delivery for periorbital injections of carboplatin in the treatment of retinoblastoma.

Submitted for publication September 18, 2001; final revision received April 1, 2002; accepted April 16, 2002.

This study was supported in part by core grant P30-EY06360 from the National Institutes of Health, Bethesda, Md; the Knights Templar Eye Foundation, Inc, Chicago, Ill; the Foundation Fighting Blindness, Owings Mills, Md; and Research to Prevent Blindness Inc, New York, NY.

We thank Haemacure Corp for donating the fibrin sealant; Brian Sippy, MD, PhD, for assisting with the histological determination and photography; and Machelle Pardue, PhD, for performing the electroretinographic studies.

Corresponding author and reprints: Henry F. Edelhauser, PhD, Emory Eye Center, 1365B Clifton Rd NE, Suite B2600, Atlanta, GA 30322 (e-mail: ophthfe@emory.edu).