New pieces for the puzzle: Nonsteroidal anti-inflammatory drugs and corneal ulcers

The cause of infiltrated corneal ulcers after topical nonsteroidal use is puzzling. Recently, there was an epidemic of ulcers after cataract surgery associated with the generic diclofenac (DSOS) by Falcon. However, this was not the first occurrence of marginal corneal ulcers with topical nonsteroidal use. Marginal corneal ulcers have been reported in eyes having photorefractive keratectomy (PRK) when the eyes were treated for 48 hours with bandage contact lenses and topical nonsteroidal anti-inflammatory drugs (NSAIDs) but not topical corticosteroids.1 The addition of topical corticosteroids eliminated the marginal ulcers.2 The ulcers in PRK patients occurred with the brand-name drugs Voltaren® (diclofenac sodium) and Acular® (ketorolac tromethamine). However, ulcerations after cataract surgery were not reported with either of these drugs with or without the combined use of topical corticosteroids. Could there be an association between the ulcers after PRK and cataract surgery? More important, is there anything in ophthalmic and nonophthalmic research to explain what occurred in each type of ulceration? In 1993, Phillips et al.3 reported significantly lower prostaglandin levels in corneas treated with Voltaren than with the vehicle or fluorometholone (FML®) in rabbits having PRK and receiving drops every hour for 10 hours. This finding was expected; not expected was the finding that the Voltaren eyes had significantly more lymphocyte filtration than the FML or vehicle eyes. Voltaren caused an increase in cellular inflammation. If prostaglandins are such potent mediators of inflammation, why would eyes with lower levels of prostaglandins show increased cellular inflammation? Why did humans receiving only NSAIDs after PRK develop marginal infiltrates—essentially localized areas of increased cellular inflammation? Nonophthalmic researchers have shown that high nonphysiological doses of NSAIDs cause apoptosis in a variety of tissue-cultured cells. The cell lines with these changes include monocytes and fibroblasts.4,5 These findings appear to occur after 48 hours of incubation with higher doses of NSAIDs in the range of 50 to 500 μmol/L concentration. Associated with the apoptosis are less complete stages of cell death, including blebbing, production of apoptotic bodies, and nucleosomal ladders indicative of cell death. These partial stages of apoptosis may lead to liberation of inflammatory precursors, which in the in vivo situation would be expected to cause or stimulate cellular inflammation. In cases of widespread apoptosis, the apoptotic process blurs with necrotic cell death.6 Associated with the induction of apoptosis with high levels of NSAIDs is the paradoxical upregulation of a novel COX-2-type enzyme—the inflammatory enzyme blocked by NSAIDs.4 The exact role of this COX-2 enzyme is unknown, but it is distinct from the COX-2 enzyme usually stimulated by bacterial and inflammatory stimulants. Different NSAIDs induce apoptosis and COX-2 enzyme production at varying doses, with diclofenac appearing to be the most potent, producing these secondary effects at the lowest concentrations.4 The active site of the COX-2 enzyme appears to differ from that of the more commonly recognized enzyme. Four features distinguish the novel COX-2 enzyme: diclofenac washes out easily from its active site; acetaminophen is a more potent inhibitor; aspirin does not inhibit the diclofenac-induced variant enzyme; and it is less sensitive to inhibition by NSAIDs. The dose of NSAIDs necessary to produce widespread apoptosis and increased COX-2 enzyme may be in the range of what is normally nonphysiological systemically but within the range possible with high-dose ocular use. Studies with topical antibiotics demonstrate widely varying drug levels in the cornea, with some being quite high.7,8 Frequent use of the drug, as in the study by Phillips et al.,3 or the usually higher concentrations achieved with topical dosing may be enough in some cases to attain the “nonphysiological” doses used in the cell-line studies. Perhaps the use of a contact lens coupled with the epithelial defect after PRK lead to local concentrations high enough to damage keratocytes and white cells in post-PRK eyes, leading to the development of marginal infiltrates. Corticosteroids also block COX-2 enzymes9 and therefore would be expected to block some of the paradoxical effect of high nonphysiological doses of NSAIDs. The addition of corticosteroids did indeed block the development of ulcers associated with NSAID use after PRK. The occurrence of infiltrates and ulcers with the use of DSOS may be the result of increased penetration of the drug into the corneas or an unknown increased propensity of this particular form of diclofenac to stimulate apoptosis and increased COX-2 enzyme. The ulcers with DSOS seemed to occur primarily at the incision sites for cataract surgery, where penetration into the cornea could be increased. Moreover, the vehicle for the DSOS may have been formulated to allow more prolonged contact or penetration of the drug into the cornea. In the gastrointestinal tract, nonselective NSAIDs, such as those currently used topically, may cause ulcerations by blocking a COX-1 enzyme responsible for promoting both stem-cell survival in the crypts of Lieberkuhn and regulating the vascular flow in the gastric mucosa.10 Could a COX-1 enzyme that regulates similar functions along the limbus also be contributing to the development of marginal ulcerations in some cases? The etiology of these sterile corneal ulcers is undoubtedly complex, and other biochemical explanations have been proposed. Alternative explanations to the one I have given contend that pharmacologic inhibition of prostaglandin production by topical NSAIDs may have a detrimental effect on the cornea. Overexpression or excessive activation of matrix-degrading metalloproteinase enzymes may be essential in the pathogenesis of sterile corneal ulceration.11 In certain tissues, cyclo-oxygenase inhibitors stimulate matrix metalloproteinase production.12 It is also possible that normal corneal homeostasis requires low constitutive levels of prostaglandin production. Prostaglandin E2 receptors are present on the corneal epithelial cells.13 Prostaglandin E2 has been reported to stimulate corneal epithelial proliferation and maintain corneal epithelial cell dimension and morphology.14 Subepithelial infiltrates (SEIs) after adenoviral infections may result from an altered epithelial barrier function and increased production of chemotactic factors.15 However, it should be noted that while Teal and coauthors2 termed the corneal ulcerations after NSAID use and PRK as SEIs, these lesions and others noted after NSAID use are clinically different from the lesion after adenoviral infections. It will be interesting to see how these apparently conflicting findings eventually come together to explain the myriad corneal ulcers and inflammation we see clinically. At this point, we need animal studies to determine whether the high levels of NSAIDs are being attained in eyes with frequent dosing of NSAIDs or with contact lenses. Studies are also needed to determine whether the DSOS formulation caused increased levels of diclofenac in corneas or whether this particular form of diclofenac is more potent for causing cellular destructive reactions.