Pharmacokinetics and Toxicity of Intravitreal Chemotherapy for Primary Intraocular Lymphoma

Objective  To investigate the pharmacokinetics and toxicity of intravitreal chemotherapeutic agents in the rabbit eye for the potential treatment of primary intraocular lymphoma and other intraocular malignancies.

Methods  The ocular pharmacokinetics of intravitreal methotrexate sodium (400 µg) was studied in 10 New Zealand white rabbits, and a single-compartment, first-order elimination model was used to calculate the drug half-life. With the use of these data, a treatment schedule using serial injections of intravitreal methotrexate and single injections of fluorouracil and dexamethasone sodium phosphate was developed. This schedule was studied in 4 New Zealand white rabbits to explore the combined toxicity of these agents.

Results  Methotrexate vitreous levels, following a 400-µg intravitreal injection, remained therapeutic (>0.5µM) in the rabbit eye for 48 to 72 hours. Intravitreal methotrexate, combined with fluorouracil and dexamethasone, showed no evidence of drug toxicity as determined by electroretinography and histopathologic examination.

Conclusions  A treatment schedule for primary intraocular lymphoma consisting of methotrexate intravitreal injections every 48 to 72 hours provides therapeutic drug concentrations in the vitreous and, in combination with fluorouracil and dexamethasone, appears to be safe in the rabbit eye.

Clinical Relevance  Although responsive to conventional chemotherapy or radiotherapy, recurrence of ocular involvement with primary central nervous system lymphoma occurs in more than 50% of treated cases. Anecdotal reports of the use of intravitreal chemotherapy for primary intraocular lymphoma have been encouraging. However, animal data on the pharmacokinetics and toxicity of combined intravitreal agents for the treatment of this disease are lacking.

LOCAL OCULAR therapy results in high ocular drug concentrations and has been successful in animal models of ocular malignancy.^1-5 In a limited number of reports on potential treatment measures for primary intraocular lymphoma (PIOL), intraocular injections of methotrexate⁶ or methotrexate and thiotepa⁷ have been used. Unfortunately, the data on the long-term safety and efficacy of locally injected antineoplastic agents into the eye are lacking.

Primary intraocular lymphoma, or primary central nervous system lymphoma, is a non-Hodgkin lymphoma that may arise from the brain, spinal cord, leptomeninges, or eye. It is associated with a high mortality, with ocular involvement in 25% of patients.^8,9 Treatment of primary central nervous system lymphoma with chemotherapy, with or without radiation, improves survival. However, ocular disease recurs in 50% of cases.¹⁰ Locally injected cytotoxic agents are a viable treatment option for these patients.

The choices of antineoplastic agents available for local ocular therapy are limited because most are toxic to the retina and optic nerve.^11,12 However, some antimetabolites have been shown to be safe. Methotrexate has been shown to be clinically nontoxic when injected into human eyes at doses of 400 µg.⁶ Similarly, fluorouracil^12-16 and corticosteroids,^17,18 at doses of up to 1 and 4.8 mg, respectively, have been shown to be nontoxic when injected into animal eyes. These agents have also been used in human eyes safely.^19,20 Although known to be active against non-Hodgkin lymphoma,^10,21-26 the pharmacologic characteristics and toxic effects of local ocular injections using these agents together have not been well examined.

The pharmacologic characteristics of local injections of methotrexate in the cerebral spinal fluid has been well studied and used extensively for central nervous system disease.²⁷ Protracted exposure times of methotrexate (3 days to 4 weeks) by using frequent low-dose injections or continuous infusions have been shown to improve tumor cell kill and to reduce local and systemic toxic effects.^28-30 Intrathecal therapy using multifractionated dosing of methotrexate is now standard of care in the treatment of leptomeningeal cancers. Combining methotrexate with other agents, such as other antimetabolites and corticosteroids (so-called triple therapy), has been used intrathecally to enhance tumor response rates in patients with meningeal leukemia.³¹ Extrapolating from the experience of using combination chemotherapy injected directly into the cerebral spinal fluid, triple intravitreal therapy using methotrexate, fluorouracil, and dexamethasone sodium phosphate, may lead to more durable treatment responses in patients with intraocular malignancies such as PIOL.

The cytotoxic effects of methotrexate are enhanced when fluorouracil is given 24 hours after methotrexate exposure, and this antineoplastic synergism may improve treatment responses with intravitreal chemotherapy.^32-34 Methotrexate and fluorouracil are highly cell-cycle dependent, acting primarily during the S phase (DNA synthesis) of the cell cycle.³⁵ In contrast, corticosteroids are cell-cycle nonspecific and are cytotoxic to lymphoma cells at all stages of the cell cycle.^36-39 Therefore, for tumor cells not treated by methotrexate and fluorouracil (ie, cells in resting phase or G₀), a single intravitreal injection of dexamethasone was given in our study following the last methotrexate injection. The intravitreal dose of dexamethasone sodium phosphate used was 500 µg, since higher doses, or multiple dosing, are associated with ocular toxic effects.^18,20

The purpose of this study was to investigate the pharmacokinetics of intravitreal methotrexate in the rabbit eye to optimize the dosing interval. Furthermore, this study examined the safety of a treatment schedule using serial injections of intravitreal methotrexate and single injections of fluorouracil and dexamethasone in the rabbit eye.

The in vitro activity of methotrexate for 63 different cell lines has been investigated at the Developmental Therapeutics Program, National Cancer Institute, Frederick, Md, and results indicate that the therapeutic levels of methotrexate range from 0.1 to 1.0µM with a mean 50% growth inhibition of 0.32µM.⁴⁰ Based on the treatment of one patient with PIOL, a single intravitreal injection of 400 µg of methotrexate yielded cytotoxic levels in the vitreous for approximately 72 hours.⁷ We investigated the pharmacokinetics of intravitreal methotrexate sodium in the rabbit eye. We then used these data to design a treatment schedule combining intravitreal methotrexate with intravitreal fluorouracil and dexamethasone.

Ten New Zealand white rabbits (20 eyes) of either sex weighing 2 to 3 kg (Covance Laboratories Inc, Vienna, Va) were used, and the procedures adhered to the guidelines from the Association for Research in Vision and Ophthalmology for animal use in research. Food and water were supplied to the rabbits ad libitum. Animals were anesthetized with 35 mg/kg of intramuscular ketamine hydrochloride (Fort Dodge Inc, Fort Dodge, Ind) and 5 mg/kg of intramuscular xylazine hydrochloride (Phoenix Scientific Inc, St Joseph, Mo), and 1% proparacaine hydrochloride ophthalmic drops (Allergan America, Hormigueros, Puerto Rico) were used topically on the eye. After adequate anesthesia and akinesia were obtained, a lid speculum was placed and each eye was injected 3 mm behind the surgical limbus in the superotemporal quadrant separately with 32 µL of a 1:1 dilution in balanced salt solution (BSS) (Alcon Laboratories Inc, Fort Worth, Tx) of methotrexate sodium (preservative free, 25 mg/mL in 2-mL vials; distributed by Immunex, manufactured by Lederle Parenterals Inc). The total dose for each injection was 400 µg of methotrexate sodium. One rabbit was used as a control and injected in both eyes with 32 µL of BSS. Two rabbits receiving methotrexate injections in both eyes were euthanized with a pentobarbital overdose (Beuthanasia-D Special; Schering-Plough Animal Health Corp, Kenilworth, NJ) at each of the following intervals after injection: 1.5, 5.5, 7.5, 23.5, and 47.5 hours. The eyes were enucleated and immediately frozen at –70°C for later dissection and drug extraction. The time from enucleation to freezing was rapid (<10 seconds), which limited postmortem drug redistribution. The control rabbit was euthanized and both eyes were enucleated 24 hours after injection.

The eyes were dissected while frozen by first removing the cornea and lens from the globe. The vitreous was then expressed and isolated for drug extraction. The methotrexate was extracted from the tissues by the addition of an equivalent weight of high-performance liquid chromatography (HPLC) grade acetonitrile (Burdick & Jackson Inc, Muskegon, Mich), sonicated for 3 minutes at a level of 3.5 with an ltrasonic processor (GEX 600; Thomas Scientific, Swedesboro, NJ), and incubated for 24 hours at room temperature. The samples were spun down in a centrifuge (TOMY MTX-150; Peninsula Laboratories Inc, Belmont, Calif) for 15 minutes at 10 000 rpm, and the supernatants were submitted for HPLC analysis.

Samples were analyzed by using a computerized HPLC system (HP1100; Hewlett Packard, Agilent Technologies, Palo Alto, Calif) equipped with a UV detector, an autosampler, a gradient pump, and a workstation (HP Kayak; Hewlett Packard) that controls the operation of HPLC and analyzes the data. A chromatography column (5 µm, 250 × 4.6 mm) (Ultrasphere C-18; Beckman Coulter, Inc, Fullerton, Calif) was used for separation and detection was set at 302 nm. The flow rate used was 1.0 mL/min, with a mobile phase of 85% of 63mM sodium phosphate dibasic and 19mM citric acid and 15% acetonitrile. A five-point standard curve of methotrexate (0.5-150 µg/mL) was constructed with correlation coefficient 1.000, and methotrexate concentrations of samples were calculated on the basis of peak areas.⁴¹

Methotrexate concentration was determined for each vitreous specimen(20 eyes total for all time points, 4 eyes per time point or 2 eyes for each of 2 rabbits per time point). All data were then fit to a single-compartment, first-order elimination model as shown in equation 1 with 1/C²_pred weighting (where C_pred is the concentration predicted by the best fit of the data to a single exponential decay; and the weighting scheme assumes that the error in the concentration value is proportional to the concentration) to obtain estimates of the volume of distribution (V_d)and elimination rate constant (k):

where C indicates concentration, t is time after administration of the drug, exp(−k · t) is equivalent to e^{(−k · t)} (e to the power of [−k · t]) in which e is Euler's number with a numerical value equal to 2.71828. The clearance was calculated as the product of V_d and k, and the half-life was calculated as 0.693/k.

Since at each time point methotrexate concentrations in both eyes of 2 rabbits were measured, the usual calculation of SD, which is based on the statistical independence of each measurement, cannot be applied. Therefore, the SD for each time point was derived from the variance components model, which recognizes that the variance of these data was the sum of 2 components: that between rabbits and that within rabbits.

Data from part 1 of the study indicated that an intravitreal injection of methotrexate sodium (400 µg) can yield therapeutic levels for 48 to 72 hours (see "Results" section). Using this information, a treatment schedule of serial injections of intravitreal methotrexate and a single injection of fluorouracil and dexamethasone (Table 1) was developed. The intravitreal doses of fluorouracil and dexamethasone suggested in the table were recommended based on the reported ocular toxic effects data in the literature.^14,16-18

Part 2 of this study examines the safety of using serial injections of intravitreal methotrexate and single injections of fluorouracil and dexamethasone in the rabbit eye. Doses used in this ocular toxicity test were increased by 100% to explore the effects of maximum therapy. A total of 3 cycles, given 1 month apart, were administered to study the long-term ocular toxic effects of combination intravitreal chemotherapy.

Four New Zealand white rabbits of either sex weighing 2 to 3 kg (Covance Laboratories Inc) were used for this part of the study, and anesthesia was given before injections as described herein. Drug injections were given in the right eye of each rabbit, whereas control injections with BSS (Alcon Laboratories Inc) of equal volume were administered in the left eye. Methotrexate sodium(25 mg/mL, preservative free) was injected undiluted in a volume of 32 µL for a total of 800 µg per injection. Fluorouracil (50 mg/mL, ICN Pharma ceuticals Inc, Costa Mesa, Calif), diluted to a concentration of 25 mg/mL with BSS, was injected in a volume of 40 µL for a total of 1000 µg per injection. Dexamethasone sodium phosphate (Decadron, 24 mg/mL; Merck & Co Inc, West Point, Pa) was injected undiluted in a volume of 42 µL for a total of 1000 µg per injection. The schedule of injections was performed as shown in Table 1. A total of 3 cycles were administered, each cycle separated by 28 days.

Electroretinograms (ERGs) were obtained at baseline; at days 3, 5, and 9 of each treatment cycle; and at every 7 days between cycles. The ERGs were continued for 3 months after the last cycle to examine for any latent toxic effects. The rabbits were anesthetized using the same procedures detailed herein, and the pupils were dilated with 1 drop of 2.5% phenylephrine hydrochloride(Akorn Inc, Decatur, Ill) and 1% tropicamide (Alcon Inc, Humacao, Puerto Rico). The ERGs were recorded from each eye separately after 30 minutes of dark adaptation. A monopolar contact lens electrode (ERG-Jet, La Chaux-des-Fonds, Switzerland) was placed on the cornea and served as a positive electrode. Subdermal needle electrodes inserted in the forehead area and near the outer canthus served as the ground and negative electrodes, respectively. The ERGs were elicited by flash stimuli delivered with a photostimulator (PS22; Grass Instruments, Quincy, Mass) at 0.33 Hz. Responses were amplified, filtered, and averaged with a signal averager (Spirit; Nicolet Instruments Corp, Madison, Wis). Averages of 20 responses were measured to obtain amplitude and implicit time values of a and b waves.

Rabbits were euthanized and both eyes were enucleated 2 weeks after the last ERG. Enucleated eyes were fixed in 10% formalin immediately after removal. Paraffin sections through the pupillary optic nerve head axis, including the injection sites, were stained with hematoxylin-eosin for light microscopic examination.

To determine whether there was a difference between treated and placebo eyes, the difference of treated minus placebo means of logarithm-transformed ERGs at the end of the study was tested for difference from zero, adjusted for the difference at the beginning of the study, by analysis of covariance. The mean of treated eyes over all times minus the mean of placebo eyes was tested by the 1-sample t test.

Following a 400-µg injection, methotrexate concentration in the vitreous declined monoexponentially, demonstrating only an elimination phase without an observable distribution phase. The volume of distribution was 2.16 mL compared with a vitreous volume of 1.50 mL.⁴² Thus, methotrexate appears to distribute rapidly into a volume 1.44 times the volume of the vitreous within the first 2 hours. The half-life of methotrexate in the rabbit vitreous was 7.6 hours. Methotrexate levels remained above a therapeutic level in the rabbit eye for 48 hours after an injection of 400 µg (5.25µM) (Figure 1). The level is predicted to drop below 1µM by 66 hours. At 72 hours, the predicted concentration is 0.58µM (0.26 µg/mL). The clearance of methotrexate was approximately 0.20 mL/h.

The ERG analysis was performed by examining the dark-adapted a- and b-wave amplitudes of the experimental and control eyes of each rabbit at each time point (Table 2 and Figure 2). Fluctuations in the a- and b-wave amplitudes are noted in both the experimental (right) and control (left) eyes. These fluctuations do not correspond specifically to the timing of injections. A large decrease in a-wave amplitude in control and experimental eyes is noted at 44 days after completion of the second cycle. Full recovery is noted, however, with no statistically significant difference in the mean of the a- and b-wave amplitudes between the treatment eyes (right) and control eyes (left) after the final cycle (P = .11). Furthermore, there were no significant differences in the mean ratio of the a- and b-wave amplitudes, averaged over all the study points, between the treatment eyes (right) and control eyes (left) (P>.20).

Histopathologic examination of the enucleated eyes showed no photoreceptor or ganglion cell layer damage in the experimental eyes compared with the control, away from the site of injection (Figure 3). The optic nerve and medullary rays appeared intact in both the treated and control eyes.

A cycle of intravitreal injections of methotrexate, fluorouracil, and dexamethasone was developed for the treatment of intraocular malignancy such as PIOL. The dosing interval selected for serial methotrexate injections will deliver therapeutic drug levels in the vitreous for 8 days. This protracted exposure of lymphoma cells to methotrexate would allow most of the tumor cells that are active in the cell cycle to pass through the S phase and be exposed to the drug. Antineoplastic synergism is enhanced when methotrexate levels are at least 10µM and fluorouracil is given 24 hours after methotrexate exposure.⁴³ According to equation 1, the concentration of methotrexate in the vitreous is 46µM 24 hours after the methotrexate injection; therefore, a fluorouracil injection on day 2 of the cycle would be appropriate. Dexamethasone is given at the end of the cycle to potentially treat cells not progressing through the S phase. Corticosteroid application at the end of a cycle may also be useful in treating an inflammatory response secondary to tumor cell death.

The rabbit eye is commonly used to study the pharmacokinetics of drugs injected into the eye. However, several factors must be considered when interpreting these pharmacokinetic data. First, the development of the treatment schedule in Table 1 assumes that the injections of fluorouracil and dexamethasone do not affect the clearance of methotrexate from the eye. The drugs were not dosed simultaneously but rather on a staggered schedule separated by at least 24 hours. Therefore, the effect of fluorouracil and dexamethasone injections on the clearance of methotrexate would not likely be as large a concern as with concomitant dosing. If the clearance of methotrexate were delayed by fluorouracil, the methotrexate concentration in the vitreous would be increased. Therefore, therapeutic levels of methotrexate would be present at the time of fluorouracil injection as predicted by part 1 of our study, and the dosing and sequence of drugs in our treatment schedule would remain the same. A larger concern with delayed methotrexate clearance would be an increased potential for ocular toxic effects. However, part 2 of this study showed no drug toxic effects when the sequence of drugs in Table 1 was used at twice the dose and throughout 3 cycles.

Second, patients with PIOL often have had previous cataract surgical procedures and vitrectomies. One can expect the clearance of small-molecular-weight drugs to be faster in aphakic and/or vitrectomized eyes.⁴⁴ This, however, was not examined in this study.

Third, pharmacokinetic data reported for other drugs in rabbit eyes have shown intraocular drug accumulation in the vitreous of the fellow eye following an intravitreal injection in the opposite eye,⁴⁵ presumably from an intercommunicating vessel,⁴⁶ raising a concern regarding the independence of the right and left eye during intravitreal treatment. However, in previous experiments (Michael R. Robinson, MD, unpublished data, 1998), no detection of methotrexate was found in the fellow eye following intravitreal administration of methotrexate in the opposite eye.

Finally, the rabbit eye differs from the human eye in volume and surface area for drug clearance. It would be useful to extrapolate the pharmacokinetic measurements describing methotrexate clearance from the rabbit eye to the human eye to obtain estimates of the dosages and dosing intervals required to maintain a therapeutic level. The volume (V) of the human eye is 3.9 mL compared with the rabbit's eye of 1.5 mL. When chemotherapeutic agents are dosed systemically, clearance (CL) scales allometrically according to body surface area. By assuming that a similar principle applies to intravitreal dosing, an estimate of the CL of methotrexate from the human eye might be obtained as follows:

²

By using vitreous volumes for the scaling, CL from the human eye is estimated to be 0.38 mL/h. If the volume of distribution is 47% higher than the vitreous volume, as it was in the rabbit, the corresponding half-life is estimated to be 10.4 hours. The equation describing methotrexate concentration after injection would be as follows:

where C indicates concentration; exp, exponential; and t, time.

In Figure 4, we used equation 3 to predict methotrexate concentrations over time in a human eye following a 400-µg injection. Methotrexate levels of more than 0.5 µM are predicted throughout the cycle, and the drug level at 24 hours (31.9 µM) is adequate for the antineoplastic synergy with fluorouracil.

Ethical issues restrain investigators from conducting similar pharmacokinetic studies in human eyes. However, a treatment schedule based on the literature and the data presented herein could be safe and potentially effective for the treatment of primary intraocular lymphoma in humans. A pilot study at the National Eye Institute, Bethesda, Md, in conjunction with the National Cancer Institute, Bethesda, is being developed to examine the safety and efficacy of this regimen in patients with PIOL.

Accepted for publication March 8, 2001.

Corresponding author and reprints: Gisela Velez, MD, MPH, 85 Washington Park, Newton, MA 02460 (e-mail: brusie-velez@erols.com).