Intravitreal Voriconazole: An Electroretinographic and Histopathologic Study

Background  Voriconazole, a novel triazole antifungal agent, presents potent activityagainst a broad spectrum of yeast and molds.

Objective  To determine whether voriconazole could be safely used as an intravitrealagent in the treatment of fungal endophthalmitis.

Methods  Retinal toxicity of voriconazole was examined in a rodent animal model.Voriconazole solutions were serially diluted and injected intravitreally intothe eyes of normal adult Sprague-Dawley rats so that the final intravitrealconcentrations were 5 μg/mL, 10 μg/mL, 25 μg/mL, 50 μg/mL, and500 μg/mL (n = 3 for each concentration group). Saline was injectedinto the fellow eyes of all animals as controls. Three weeks after injections,electroretinograms were measured, and eyes were subsequently enucleated forhistologic examination.

Results  In electroretinographic studies, maximum scotopic b-wave, intensityneeded for half saturation, and saturated a-wave amplitude were measured.There was no statistically significant difference in these parameters recordedbetween control eyes and voriconazole-injected eyes in any concentration groups.Histologic examination with light microscopy did not reveal any retinal abnormalityin the eyes with 5 to 25 μg/mL of intravitreal voriconazole. In the eyeswith 50 μg/mL and 500 μg/mL of voriconazole, small foci of retinal necrosiswere occasionally observed in the outer retina, especially in the eyes with500 μg/mL of voriconazole.

Conclusions  Our results demonstrate that intravitreal voriconazole of up to 25 μg/mL causes no electroretinographic change or histologic abnormality in rat retinas.This indicates that voriconazole is a safe antifungal agent for intravitrealinjection in rodents and may be used in the treatment of human fungal endophthalmitisfollowing further study.

Fungal endophthalmitis, although uncommon, remains a serious ophthalmologicchallenge owing to its limited available treatments and potentially devastatingocular consequences. Fungal endophthalmitis can be caused either by exogenousorigin, such as ocular trauma or surgery, or by endogenous infection spreadingto the eye, such as those in immunocompromised patients. Until recently, intravitrealinjection of amphotericin B has been the principal treatment for fungal endophthalmitis,¹ although other potential intravitreal antifungal agentshave been investigated.^2-4 However,intravitreal amphotericin B, even at low concentrations, 4.1 μg/mL or 8.3 μg/mL(5-μg or 10-μg injection into 1.2 mL of rabbit vitreous),^4,5 cancause focal retinal necrosis.^6,7 Furthermore,resistance to amphotericin B has been documented in a variety of human systemicfungal infections.⁸ Fluconazole, a triazoleagent, has been used systemically as a supplement or alternative to amphotericinB to treat fungal endophthalmitis because it can reach effective concentrationin the vitreous after oral administration,^9,10 butit lacks a broad spectrum of coverage against many of the most commonly encounteredorganisms found in fungal endophthalmitis.^11,12 Thus,ophthalmologists have had a very limited number of effective antifungal agents,and the current treatment protocols for fungal endophthalmitis are far fromoptimal.

Recently, a new antifugal agent, voriconazole, has been approved bythe US Food and Drug Administration for systemic fungal infection. Voriconazoleis a second-generation synthetic derivative of fluconazole, and it differsfrom fluconazole by the addition of a methyl group to the propyl backboneand by the substitution of a triazole moiety with a fluoropy rimidine group.The structural changes in voriconazole result in a higher affinity for thefungal 14-α-demethylase, leading to more potent activities.¹² Like fluconazole, voriconazole exerts its effectsprimarily by inhibiting the fungal cytochrome P450 CYP3A enzyme lanosterol14-α- demethylase, preventing the conversion of lanosterol to ergosterol.This in turn causes depletion of ergosterol, which disrupts the integrityand function of the fungal cell membrane, eventually leading to cell lysis.¹³ Voriconazole also inhibits 24-methylene dihyfrolanasteroldemethylation in certain yeast and filamentous fungi, explaining its increasedactivities against molds.^14,15

Many recent studies report that this novel triazole antifungal agentpresents potent activity against a broad spectrum of yeast and molds. Whencompared with amphotericin B, fluconazole, itraconazole, and flucytosine against6970 isolates of Candida species obtained from morethan 200 medical centers worldwide, voriconazole and ravuconazole (anothernew triazole agent) were each more active than amphotericin B against all Candida species and were the only agents with good activityagainst Candida krusei.¹⁶Candida albicans is generally the most susceptible yeast,with a voriconazole MIC₉₀ (the minimal inhibitory concentration[MIC] of drug causing a 90% growth inhibition of organisms) of only 0.06 μg/mL,while Candida glabrata is the least sensitive, withan MIC₉₀ of 2.0 μg/mL.¹⁷ Otherstudies showed that voriconazole was more active than amphotericin B againstfilamentous fungi, such as Aspergillus species, witha mean MIC of 0.19 to 0.58 μg/mL, and Pseudallescheriaboydii,^18,19 an especiallyinvasive Aspergillus,²⁰ witha minimum fungicidal concentration (at tissue concentrations approximatelytwice the MIC) of 0.7 to 1.0 μg/mL.¹² Theendemic fungal pathogens Fusarium species, Histoplasma capsulatum, Coccidioidesimmitis, Blastomyces dermatitidis, Penicillium marneffei, Scedosporium apiospermum, Paracoccidioides brasiliensis, as well as Cryptococcus neoformans, and the dermatophytes are also fully susceptibleto voriconazole.^12,21 Voriconazolealso has good activity against those fungi that are resistant to the othercommonly used antifungal agents, such as amphotericin B and fluconazole.^13,22,23 Although it is sensitiveto miconazole, Paecilomyces lilacinus–causedexogenous endophthalmitis is usually resistant to the first-line antifungalagent amphotericin B (MIC>16 μg/mL),^24,25 butit should be and was treated successfully with voriconazole²⁵ becauseits MIC is 0.12 to 0.5 μg/mL and its MIC₉₀ is 0.5 μg/mL.¹⁹ Voriconazole does not appear to be cross-resistantwith amphotericin B, likely owing to the different sites of action of the2 agents.²⁶

Because the treatment for fungal endophthalmitis is very limited andvoriconazole shows potent broad-spectrum coverage for fungal infections, thisstudy was designed to examine whether voriconazole could be safely used asan intravitreal agent in the treatment of fungal endophthalmitis. Rats wereused as animal models in our study. Intravitreal voriconazole injections wereperformed, and retinal function and anatomy were subsequently examined usingelectroretinographic (ERG) and histologic studies.

Sprague-Dawley albino rats 6 to 7 weeks old (approximately250 g) wereobtained from Charles River Laboratories (Wilmington, Mass). Animals werefed ad libitum with Purina laboratory chow (Ralston Purina, Atlanta, Ga) andwater, with room lighting consisting of a 12-hour light/12-hour dark cycle.The experiments were carried out in accordance with the Association for Researchin Vision and Ophthalmology principles of animal maintenance and care, andwere approved by the institutional review board at Baylor College of Medicine,Houston, Tex.

Animals were anesthetized with intraperitoneal injections of a solutioncontaining ketamine (95 mg/mL) and xylazine (5 mg/mL) in a dose of 0.2 mLper 100 g of body weight. Additional topical anesthesia was provided by 0.5%proparacaine hydrochloride (Alcon Laboratories Inc, Fort Worth, Tex). An ophthalmicsolution of 0.3% ofloxacin (Allergan Inc, Irvine, Calif) was applied to theocular surface before injection and a bacitracin ophthalmic ointment of 500u/g (E. Fougera & Co, Melville, NY) after injection to prevent infection.Intravenous voriconazole, a white lyophilized powder, was obtained (Vfend;Pfizer Inc, New York, NY). Based on a study of microbial keratitis causedby a variety of fungal pathogens, the MIC of voriconazole is 0.5 to 5.0 μg/mL.²⁷ Because this is a retinal toxicity study, we choseto use 5 μg/mL as MIC although it is much higher than those in literature(see introduction). Voriconazole solutions were serially diluted with balancedsalt solution (Alcon Laboratories Inc) so that the final intravitreal concentrationswere 5.0 μg/mL, 10 μg/mL, 25 μg/mL, 50 μg/mL, and 500 μg/mL(1-, 2-, 5-, 10-, and 100-fold of MIC, respectively) based on prior data thatadult-rat vitreous volume is 56 ± 4 μL.²⁸ Seriallydiluted voriconazole solutions of 6 μL were injected intravitreally intorat eyes under a dissecting microscope using a Hamilton microinjector (HamiltonCo, Reno, Nev) (n = 3 for each concentration group). A 30-gaugeneedle was first used to make a punch incision 0.5 mm posterior to the temporallimbus, and the Hamilton needle was then inserted through the incision, approximately1.5 mm deep, angled toward the optic nerve until the tip of needle was visualizedin the center of vitreous. Balanced salt solution of equal volume (6 μL)was injected into the fellow eyes of all animals as controls. Following intravitrealinjection, animals were kept under ambient light on a 12-hour light/dark schedule.Three weeks after injection, animals were processed for electroretinogramrecordings and subsequent retinal histologic examinations.

Prior to testing, rats were allowed to dark-adapt overnight. Under dimred light, rats were anesthetized with a solution of ketamine (95 mg/mL) andxylazine (5 mg/mL). Pupils were dilated with a single drop of 0.5% mydryaciland 2.5% phenylephrine. A drop of 0.5% proparacaine hydrochloride was appliedfor corneal anesthesia. Rats were placed on a heating pad maintained at 39°Cinside a Ganzfeld dome coated with highly reflective white paint (MunsellColor Laboratory, New Windsor, NY). A small amount of 2.5% methylcellulosegel was applied to the eye, and a platinum electrode was placed in contactwith the center of the cornea. Similar platinum reference and ground electrodeswere placed in the forehead and tail, respectively. After placement in thedome, rats were kept in complete darkness for several minutes. Signals wereamplified with a Grass P122 amplifier (Grass Instruments, West Warwick, RI)(bandpass 0.1 Hz-1000 Hz). Data were acquired with National Instruments Laboratory-PCDAQ board (National Instruments Corp, Austin, Tex) (sampling rate 10 000 Hz).Traces were averaged and analyzed with custom software written in Matlab (TheMath Works Inc, Natick, Mass). Flashes were calibrated in a manner similarto that described²⁹ and detailed elsewhere.³⁰ Flashes for scotopic b-wave measurements were generatedby a Grass PS-33+ photostimulator (Grass Instruments). Light was spectrallyfiltered with a 500-nm interference filter. Flashes varied in intensity from−3.85 to −0.76 log scotopic candela-sec/m². For analysisof the a-wave and cone function, we used a 1500-W Novatron xenon flash lamp(Novatron of Dallas, Dallas, Tex), which produced approximately 2.92 log scotopiccandela-sec/m².

Following ERG tests, animals were euthanized with an overdose of intraperitonealketamine and xylazine. The eye was enucleated, an incision was made in thecornea, and the eye was fixed immediately in 4% formaldehyde in 0.1M phosphatebuffer (pH 7.4). After 15 minutes in the fixative, the lens was removed andthe eye was cut along the cornea–optic nerve axis into halves. Grossexaminations of the tissues were performed. Tissues were further fixed overnightin 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1M phosphate buffer (pH7.4). Tissues were then embedded in paraffin, sectioned at a thickness of6 μm, and stained with hematoxylin-eosin. Light microscopy was used forhistologic examinations.

The scotopic b-wave is a measurement of the extracellular field potentialthat primarily arises from rod bipolar cells in response to dim flashes oflight.³¹Figure1, top row, shows scotopic b-wave responses to increasing intensitiesof flashed light. The relationship between scotopic b-wave amplitude and intensitycan be modeled using a hyperbolic saturation function (Naka-Rushton function).This model yields 2 parameters, b_{max, scot} and I_0.5,representing the maximum b-wave amplitude and the intensity that provideshalf saturation. In eyes without any injection, b_max measured 790 ± 100μV (n = 4) and I_0.5 measured –3.01 ± 0.06log scotopic candela-sec/m². In eyes with balanced salt solutioninjection as control, b_max and I_0.5 were basically thesame as the eyes without any injection. In eyes with 5 μg/mL intravitrealvoriconazole, b_max measured 685 ± 70 μV (n = 2)and I_0.5 measured –2.93 ± 0.09 log scotopiccandela-sec/m². In eyes with 50 μg/mL intravitreal voriconazole,b_max measured 680 ± 110 μV (n = 3)and I_0.5, –2.96 ± 0.04 log scotopic candela-sec/m². In eyes with 500 μg/mL voriconazole, b_max measured700 ± 90 μV (n = 3) and I_0.5 measured–2.89 ± 0.06 log scotopic candela/m². Therewas no statistical difference in b_max and I_0.5 betweencontrol eyes and any of the voriconazole-injected eyes using 2-sided t test.

To characterize more directly rod photoreceptor function, we measuredthe ERG scotopic a-wave, which, in the rat, arises almost exclusively fromthe rod photoreceptors.³¹Figure 1, bottom row, shows the response to an intense flash, whichsaturated the rod photoreceptors. The saturated a-wave amplitude from controleyes measured 380 ± 65 μV (n = 4). The saturateda-wave amplitudes for the eyes with 5 μg/mL, 50 μg/mL, and 500 μg/mLintravitreal voriconazole were 305 ± 10 μV (n = 3),365 ± 84 μV (n = 2), and 355 ± 15μV (n = 3), respectively. There was no statistical differencein scotopic a-wave response between control eyes and any voriconazole-injectedeyes using 2-sided t test. Even in eyes with 500 μg/mLintravitreal voriconazole (100-fold MIC), the ERGs showed little differencecompared with the control eyes. In 1 of the voriconazole-injected eyes (5 μg/mL),a cataract developed owing to the needle injury during injection. The ERGshowed mild depression due to medium opacity in the eye. These data were excludedfrom analysis.

Gross examination of eye specimens showed no retinal hemorrhages orsigns of infection in any voriconazole-injected or control eyes. Histologicexamination with light microscopy did not reveal any retinal abnormality inthe eyes injected with BSS as controls. In eyes injected with intravitrealvoriconazole from 5.0 to 25 μg/mL (1- to 5-fold MIC), no abnormality couldbe observed in any area of retina (Figure 2A).In the eyes injected with 50 μg/mL intravitreal voriconazole (10-fold MIC),small focal retinal necroses were occasionally noticed in the outer retina(Figure 2B). In these necrotic areas,photoreceptor layer and inner nuclear layer were disorganized. Photoreceptordegeneration was evident, and photoreceptor inner and outer segments wereabsent. The ganglion cell layer appeared intact. In the eyes injected with500 μg/mL intravitreal voriconazole (100-fold MIC), more focal retinalnecrotic areas were found with more obvious photoreceptor degeneration anddisorganization of photoreceptor and inner nuclear layers (Figure 2C). Focal retinal detachment was noticed in these necroticareas. Inflammatory cells were also noticed in these focal retinal areas withchoroidal congestion present. In the other area where focal necrosis was notobserved, the retina appeared normal with light microscopy examination.

Our studies demonstrate that voriconazole did not cause retinal toxicityon either ERG or histologic studies when intravitreal concentrations were25 μg/mL or less. When the voriconazole concentration reached 50 μg/mLor more, focal retinal necrosis was occasionally noticed on histologic examination,but ERG was not affected because ERG is a mass electrical response from thewhole retina, and focal necrosis may not cause ERG abnormalities. Althoughthere may be species differences in retinal reaction to voriconazole, ourresults provide a solid reference level for its retinal toxicity. When theseresults are transferred to human eyes, assuming minimal species variability,voriconazole of 100 μg may be injected into the human vitreous withoutcausing long-term ERG or histologic abnormalities, based on the fact thatthe average human vitreous volume is 4 mL. Voriconazole appears to be muchsafer to the retina than amphotericin B because very low doses of intravitrealamphotericin B (4.1-8.3 μg/mL) cause focal retinal necrosis in rabbitstudies.⁷ Since voriconazole is superior orequal to amphotericin B for common and rare yeast and mold infections (seereviews^14,19), we suggest thatvoriconazole should be considered as a possible first-line intravitreal agentfor treatment of fungal endophthalmitis. A recent case report showed thatendophthalmitis caused by Fusarium solani was successfullytreated with intracameral, topical, and systemic voriconazole when the endophthalmitisfailed to respond to amphotericin B, fluconazole, or itraconazole.³²

When voriconazole was used for systemic fungal infections through eitheroral or intravenous administration, adverse effects were observed, includingtransient visual disturbance as well as hepatotoxicity and skin reactions.The most frequent adverse effect was transient visual disturbance, describedas enhanced light perception, blurred vision, photophobia, or color visionchanges. These visual events occurred in 23% to 35% of patients,¹² generallywithin 30 minutes of dosing, and most frequently during the first week oftherapy. These events were usually mild and resolved within 30 minutes. Electroretinographyhas shown that the retina was the site of these events, with decreased amplitudeof ERG waveforms in humans and dogs.³³ Histologicexamination showed no alterations in the retina or visual pathways in dogsas a result of voriconazole administration. No human histopathologic featureshave been found, and ocular examination has not detected any lesions.¹² No potential mechanism of this visual disturbancehas been described.¹⁴ It is possible that ratsmight experience similar early transient visual disturbance in our studies;this would be very difficult to determine because visual changes have to betested subjectively. However, we did not observe any ERG or histologic abnormalities3 weeks after intravitreal voriconazole (<25 μg/mL) injection. Thus,our results confirm prior systemic voriconazole studies showing that ERG changes,if any, occur in early stages, are transient, do not last more than 3 weeksafter voriconazole administration, and do not cause permanent damage to theretina.

We also studied oral voriconazole penetration in human vitreous andaqueous humor to determine if voriconazole can penetrate noninflamed eyes,and the results have recently been reported.³⁴ After2 oral doses of 400 mg of voriconazole, vitreous and aqueous humor samplesof 14 patients were obtained and analyzed with high-performance liquid chromatography.Intravitreal and intracameral concentrations of voriconazole were 0.81 ± 0.31μg/mL and 1.13 ± 0.57 μg/mL, respectively, representing38.1% and 53% of plasma concentration.³⁴

Since the MIC₉₀ of voriconazole for most of the yeasts andmolds is low, systemic voriconazole appears to be a good choice for treatmentof endogenous fungal endophthalmitis. For those systemic fungal infectionsknown to be sensitive to voriconazole, oral or intravenous administrationcan be used to treat the systemic and ocular infections. Intravitreal injectionof the drug can be used as supplementary if needed. For those fungi, suchas Fusarium species, in which the MIC is higher thanthe intraocular level achieved by systemic administration, intravitreal voriconazoleinjection may be required to achieve successful eradication of the infectiousorganism. Because voriconazole is metabolized primarily in the liver by cytochromeP450 isoenzymes,¹² some patients may not beable to tolerate systemic voriconazole owing to drug-drug interactions orhepatotoxicity. Under these circumstances, intravitreal injection may be thetreatment of choice.

In summary, there is significant penetration of voriconazole from systemicoral administration of the agent even in noninflamed eyes. It also appearsthat intravitreal voriconazole will offer a significant new treatment optionin the management of fungal endophthalmitis. The data from our rodent modelhas been encouraging, but further evalution is needed to determine the safetyand efficacy of voriconazole in humans.

Correspondence: William F. Mieler, MD, Departmentof Ophthalmology and Visual Science, University of Chicago, 5841 S MarylandAve, Chicago, IL 60637 (wmieler@uchicago.edu).

This article was corrected on 11/9/2004, prior to publication of the correction in print.

Submitted for Publication: May 8, 2003; finalrevision received February 16, 2004; accepted February 25, 2004.

Financial Disclosure: None.