Matrix metalloproteinases (MMPs), endopeptidases named for their catalytically active chelated zinc–calcium dependence, are synthesized and secreted from several cell types as inactive zymogens. They are involved in the physiological and pathological processes of corneal extracellular matrix (ECM) remodeling such as tissue repair, inflammation, angiogenesis, morphogenesis, and tumor invasion.1–4 Matrix metalloproteinases comprise a multigene family of enzymes (matrixins) divided into subgroups according to their ability to degrade different ECM substrates: collagenases (MMP-1, MMP-8, MMP-13, MMP-18), gelatinase A (MMP-2), gelatinase B (MMP-9), stromelysins (MMP-3, MMP-10, MMP-11), matrilysin (MMP-7), membrane-type (MMP-14, MMP-15, MMP-16, MMP-17, MMP-24, MMP-25), metalloelastase (MMP-12), and enamelysin (MMP-20).3–5
In normal tissue, most MMPs are secreted as soluble latent zymogens that are mainly activated during inflammation and remodeling of the wound-healing process. Matrix metalloproteinases are activated in ECM by cleavage of the propeptide at the amino terminus. In addition, MMPs are regulated through transcriptional control, cytokines, growth factors, hormones, oncogene products, and other ECM-derived signals.6 The interactions of MMPs with specific tissue inhibitors of metalloproteinases as well as prostaglandins (PGs) and cytokines, such as transforming growth factor (TGF)-β and interleukin (IL)-1, are essential for the regulation of normal corneal wound healing. Matrix metalloproteinases play an important role in ECM deposition and degradation. Altered regulation of these collagenases and gelatinases may be critical in the pathogenesis of certain corneal diseases.
Nonsteroidal antiinflammatory drugs (NSAIDs) prevent the synthesis of PGs through inhibition of the cyclooxygenase (COX) pathway.7,8 Ophthalmic NSAIDs are principally used to treat postoperative inflammation and pain associated with cataract and refractive surgery, prevent cystoid macular edema, and maintain intraoperative mydriasis. However, NSAIDs have inhibitory effects on wound healing in several tissues including the cornea.9–11 Postoperative complications have been associated with the use of certain NSAIDs after cataract and refractive surgery. These include punctate keratopathy, stromal infiltrates, conjunctival ulceration, scleral melting, and corneal melting with associated perforation.12–15
On the basis of these clinical observations, we used immunohistochemistry, zymography, and Western blot to investigate the potential effects of NSAID eyedrops on the expression of metalloproteinases capable of degradation of collagen types I, II, and III (MMP-1, MMP-8)1,2 and gelatinases also implicated in the corneal wound-healing process that are closely associated with epithelial basement membrane and collagen types IV, V, and VII degradation (gelatinase A or MMP-2 and gelatinase B or MMP-9).3
Materials and Methods
Seventy adult female Lewis rats weighing 200 to 225 g were divided into 2 experimental groups: debrided corneal epithelium in the right eye and intact corneal epithelium in the left eye. The rats were then randomly divided into 4 experimental test groups of 17 each as follows: artificial tears comprising carboxymethylcellulose sodium 0.5% (Refresh Plus® PF); diclofenac sodium 0.1% (Falcon®); diclofenac sodium 0.1% (Voltaren®); preservative-free ketorolac 0.5% (Acular® PF). All animals were treated in accordance with the Association for Research in Vision and Ophthalmology Statement for Use of Animals in Ophthalmic and Vision Research.
Anesthesia comprised an intramuscular injection of 0.5 mL/kg body weight of a 1:1 mixture of 100 mg/mL ketamine and 20 mg/mL xylazine. After 1 drop of proparacaine hydrochloride 0.5% was applied, uniform central corneal epithelial defects were created in the right eye (debrided corneal group) by a 60-second application of a 5.0 mm diameter filter paper disk soaked in ethanol 75% solution. Immediately, the treated eyes were irrigated with a balanced salt solution for 60 seconds and erythromycin 0.5% ophthalmic ointment was applied. The left eye of each rat served as an intact control. The NSAID or artificial tears eyedrops were administered 4 times a day for 1 week. The rats were killed 48 hours (n = 24) and 7 days (n = 24) after initiation of the treatment.
The corneas of 48 rats were removed at the scleral ring, bisected, embedded in OCT compound, and frozen at −86°C. Cryostat sections of 6.0 to 8.0 μm from all corneas were stained with hematoxylin and eosin or immunohistochemically using antibodies specific for MMP-1, MMP-2, MMP-8, and MMP-9. To assess the possible role of corneal collagenases and gelatinases in association with NSAID eyedrops, 22 rats (5 from each treatment groups and 2 with normal rat corneal tissue) were killed at 48 hours for immunoblotting and zymography studies.
Immunohistochemical staining was performed using an avidin–biotin–peroxidase complex (Vector Laboratories) technique. Primary antibodies consisted of polyclonal antibodies recognizing epitopes of both the latent and active forms of MMP-1 and MMP-8 (Chemicon) as well as monoclonal antibodies against MMP-2 and MMP-9 (Oncogene and Chemicon, respectively). The primary antibodies were applied to the corneal sections and incubated at room temperature for 1 hour. After they were washed with Tris base (TBS) (20 mM Tris-hydrochloride, pH 7.5, 150 mM sodium chloride [NaCL]), a biotin-labeled secondary antibody, goat anti-rabbit or horse anti-mouse immunoglobulin (Vector Laboratories), was used. Finally, after incubation with the avidin–biotin–peroxidase complex, slides were developed in 3,3′ diaminobenzidine and counterstained with methyl green 1% in methanol. All samples were stained in parallel to minimize specimen variation, and cell staining intensity was graded by a masked observer.
Analysis of immunostaining was performed to determine whether there was statistical significance between the artificial tears group and the NSAID groups. A P value less than 0.05 by the Fisher exact test was considered significant.
To clarify the relationship between early corneal metalloproteinase expression after the application of NSAID eyedrops and the quantitative level of each MMP, the 4 test groups and the untreated normal rat corneas were analyzed by Western blot and zymography.
The MMP-2 and MMP-9 levels were also assayed by zymography. Debrided (n = 20) and intact (n = 20) rat corneas from the 4 treatment groups and the undebrided, untreated control corneas (n = 4) were dissected and trephined (3.0 mm) at 48 hours and immediately cultivated in a 24-well culture dish with 500 μL of serum-free medium (modified Eagle medium, minimal essential amino acid, and antibiotic–antimycotic) at 37°C in 5% carbon dioxide atmosphere for 72 hours.
Each conditioned media aliquot was subjected to gel electrophoresis (10% SDS polyacrylamide gel containing 1 mg/mL gelatin) under nonreducing conditions. Fifteen microliters of each conditioned medium and zymography sample buffer were combined and incubated at room temperature for 5 minutes. Electrophoresis was performed at 100 to 150 volts for 2 to 4 hours. The gels were washed in renaturing buffer (Bio-Rad Laboratories) for 30 minutes at room temperature, incubated with development buffer (Bio-Rad) at 37°C for 16 hours, stained for 3 hours with Coomassie brilliant blue R-250, and destained with 3 changes of destain solution (Bio-Rad Laboratories) at 15, 30, and 60 minutes.
Western Blot Assays
The levels of MMP-1 and MMP-8 in the cornea-conditioned culture medium were assessed by Western blot. The blot was probed with polyclonal antibodies against MMP-1 or MMP-8 (Chemicon). Fifteen microliters of conditioned medium were mixed with lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 10 mM ethylenediaminetetraacetic acid, 1% Triton X100), heated at 90°C for 5 minutes, and then kept on ice. The samples were run on 10% polyacrylamide SDS gel at 100 volts for 2 hours and transferred to nitrocellulose paper at 25 volts for 120 minutes. The nitrocellulose paper was agitated in blocking buffer (phosphate-buffered solution, 0.05% Tween 20, 0.5% nonfat dry milk) at room temperature for 1 hour. The primary antibody (1:5000 dilution) and secondary antibody (1:2000 dilution) incubations were carried out for 1 hour. After 3 washes with TBS and 0.05% Tween-20, the membranes were incubated in enhanced chemilluminescence ECL solution (Amersham Life Science) followed by exposure to ECL film for 15 seconds.
The Western blots and zymograms from 2 experiments with each 1 in triplicate were analyzed by densitometry. Statistical analysis was by the Student t test. A P value less than 0.05 was considered significant (NSAID groups compared with artificial tears group).
Immunohistochemically (Figures 1 and 2), MMP-1 and MMP-8 were not detectable in undebrided corneas receiving artificial tears at 48 hours or 7 days. After corneal epithelium debridement, MMP-1 and MMP-8 expression in the artificial tears group was localized mainly in the corneal epithelium as a result of the normal wound-healing process. However, in corneas with intact corneal epithelium, the topical application of diclofenac sodium 0.1% (Falcon and Voltaren groups) consistently induced MMP-1 expression in the corneal epithelium and superficial stroma at varying levels.
The level of MMP-8 was also higher in the intact corneas in the diclofenac sodium 0.1% groups (Falcon, Voltaren) 48 hours and 7 days after treatment. Even when intact and debrided corneas treated with ketorolac (Acular) showed positive staining for MMP-1 and MMP-8, only the percentage of positively stained eyes for MMP-1 was statistically significant compared with that in the artificial tears group.
Immunohistological staining results and the statistical difference between the artificial tears group and NSAID groups at 48 hours and 7 days are shown in Table 1. With the current antibodies, minimal expression of MMP-2 was present in a few stromal cells and MMP-9 was not specifically detected in any treatment group.
Western Blot and Zymography
Electrophoretic analysis (Figure 3) of secreted cell proteins demonstrated that the production of latent and active forms of MMP-1 (52 to 42 kD) and MMP-8 (85 to 64 kD) by intact tissue in the artificial tears group was minimal and did not differ from the untreated normal control corneas. Latent and active forms of MMP-1 were expressed at similar levels in all debrided corneal epithelium NSAID groups and in NSAID groups with intact epithelium.
At 48 hours, MMP-8 was detectable in the debrided and intact corneal epithelium in NSAID test groups; however, in intact corneas, greater amounts were observed in the diclofenac sodium groups, indicating a higher level of MMP expression.
In contrast to the immunohistochemical staining, substrate-gel electrophoresis (Figure 4) showed gelatinase A (MMP-2, 72 kD) expression in the intact corneal epithelium samples after topical NSAID eyedrop treatment. This expression showed greater amounts of staining in the corneas in the diclofenac sodium groups. Conversely, MMP-2 enzymatic production expressed was less in intact corneas in the ketorolac group than in the diclofenac sodium groups, while the basal MMP-2 expression was not affected in the treated intact cornea subgroups and the artificial tears group. The latent form of gelatinase B (pro-MMP-9) was weakly identified by zymography in the intact corneal epithelium NSAID groups. The values of each MMP band from the Western blots and zymography studies (2 experiments, each in triplicate) were quantified by densitometric analysis and studied statistically to indicate the significance between the NSAID groups and artificial tears group (Figure 5).
Because of the lipophilic nature of the corneal epithelium, most lipophilic drugs penetrate the corneal layers via the intracellular route, whereas hydrophilic drugs use the paracellular route.16 Corneal tissue is capable of accumulating medication such as pilocarpine and NSAIDs; therefore, repeated doses may have a cumulative effect.16 Many ophthalmic studies report toxic effects associated with the long-term use of ophthalmic drugs. In particular, toxicity has been associated with the presence of certain preservatives in high concentrations, which is not the case with NSAID formulations.17–19 A recent study found that some preserved eyedrops induce corneal epithelial damage and an inflammatory reaction with recruitment of immunocompetent cells at the limbal area.20
In the past century, NSAIDs have been widely used as therapeutic agents for their antiinflammatory and analgesic properties. The therapeutic properties of NSAIDs rest primarily in their ability to inhibit the COX enzyme. Cyclooxygenase is a key regulatory enzyme that catalyzes the first steps in the biosynthesis of PGs from the precursor arachidonic acid. Two isoforms of COX have been recognized, COX-1 and COX-2.8 Recent research shows that COX-1 is a constitutive “housekeeping” enzyme responsible for homeostatic functions such as regulating PGs that protect the tissue from damage and accelerate wound healing.
Cyclooxygenase-2 is induced by inflammatory stimuli associated with cytokines (IL-1, IL-2, tumor necrosis factor-α) as well as the growth factors (TGF-β, fibroblast growth factor) released by migratory and other cells.21,22 Several studies support the hypothesis that the unwanted side effects of standard therapy with NSAIDs result from their ability to inhibit COX-1, whereas the therapeutic antiinflammatory effects are from the inhibition of COX-2.23,24 Another potential explanation for their deleterious effect is that blocking COX diverts precursors to the lipooxygenase pathway, which may yield well-known potent proinflammatory intermediate metabolites (12-hydroxyeicosatetraenoic acid [HETE], leukotriene B4 [LTB4], 5-lipooxygenase products).8 Further work in this chemical series led to the discovery of the high intrinsic potency of the LTB4 receptor on neutrophils from human and animal species, providing convincing evidence that LTB4 and the 12-HETE metabolite are directly involved in the pathogenesis of the delayed wound-healing process.25,26
Although topically administered NSAIDs are increasingly used in ophthalmology practice, there is growing evidence that they may induce changes in the ocular surface and may produce conjunctival and corneal epithelial cell damage.27 Takahashi et al.28 report that diclofenac sodium eyedrop solution inhibits the growth of cultured conjunctival epithelium by inducing changes in the cytoskeleton structure. This has been associated with lower prostaglandin levels. Diclofenac sodium and other NSAIDs may cause metabolic and morphologic changes in the corneal stroma and epithelium.9,28,29 Experimental studies demonstrate that diclofenac sodium and ketorolac decrease corneal sensitivity by inhibiting prostaglandin synthesis.30 An alternate explanation for corneal hypoesthesia is a direct blockade of ionic channels at the nerve endings with a decrease of sensory inflow from corneal nociceptor nerve fibers to the central nervous system.31 Long-term treatment with NSAID eyedrops may have adverse effects as a result of neurotrophic epitheliopathy and increased healing time.9,12,32
The American Society of Cataract and Refractive Surgery issued an alert based on a preliminary study it conducted among its members indicating that specific formulations of topical ophthalmic NSAIDs (generic diclofenac) may contribute to severe corneal complications that include ulceration with stromal keratolysis and even perforation after routine anterior segment surgeries. Previously reported data show that, in general, NSAIDs such as indomethacin can enhance collagenase expression and fail to inhibit matrix biosynthesis.33,34 The experiments in our study were conducted to determine the potential association and roles of interstitial collagenase (MMP-1), PMN leukocyte collagenase (MMP-8), gelatinase A (MMP-2), and gelatinase B (MMP-9) in response to the topical application of various NSAIDs in normal and debrided corneal epithelium.
The corneal ECM plays an active, complex role in the regulation of epithelial and stromal cells, influencing their development, migration, proliferation, and metabolic functions to stabilize the physical structure of the tissue. Regulating the balance between degradation and regeneration of the corneal ECM is crucial for the repair and maintenance of proper tissue architecture and clarity. Several enzymes, including MMPs, could be involved in the ECM remodeling associated with corneal wound healing. Degradative processes are carried out mainly by collagenases and gelatinases present in the corneal cells and ECM. These groups are responsible for the specific breakdown of collagen types I, II, and III (collagenases MMP-1 and MMP-8) and collagen types I, IV, V, VII, and X, fibronectin, gelatins, and galectin-3 (gelatinases A and B) implicated in ECM and corneal epithelial basement membrane degradation.4,35,36
During corneal injury and inflammation, MMP expression plays an important role in the proteolytic process and wound healing; however, in pathological situations, MMP up-regulation may lead to ulcerative keratolysis.15 The possibility that the collagenases (MMP-1 and MMP-8) and gelatinases A and B are involved in the development of ulcerative keratolysis associated with the topical application of NSAIDs has not been considered previously. Several studies report the toxic effects on corneal and conjunctival epithelial cells after treatment with ophthalmic solutions containing certain preservatives in high concentrations or their long-term use. Thus, we decided to focus our study on the first 48 hours after the application of commercially available NSAID eyedrops. An important aspect of the study was the identification of MMP-1, MMP-2, and MMP-8 expression in normal corneal tissue after early NSAID eyedrop use with minimal or negative expression in the control and artificial tears groups. Matrix metalloproteinase expression may play a role in the delayed corneal epithelial healing associated with some NSAID eyedrops.15
These results provide evidence that some undesirable effects on corneal wound healing from the use of specific NSAID eyedrops (or their use in vulnerable corneal epithelium) may be secondary to the overexpression of the collagenases MMP-1 and MMP-8 and the gelatinase MMP-2. Our paradoxical data concerning the 3 MMP-1 bands in the epithelial defect groups are explained by the enzymatic cleavage mechanism of the latent MMP-1, which generates an intermediary form and then the finally activated MMP-1.
These, in turn, may impair normal corneal healing by destroying newly deposited ECM (hydrolyze fibrillar collagen and elastin), disrupting epithelial cell interaction and the epithelial basement membrane. These findings may have clinical significance in the pathophysiology of ulcerative keratolysis associated with topical NSAID use as we and other authors15,37–39 report.
In conclusion, our study shows that the production of MMPs in intact corneal epithelium correlates with the early use of some topical NSAIDs and therefore supports our original hypothesis of their involvement in the development of corneal ulceration. Nonsteroidal antiinflammatory drugs are frequently used in the postoperative period. The well-known metabolic and morphologic changes in the corneal epithelium and stroma after their use, associated with the up-regulation of some metalloproteinases activated in response to the proinflammatory intermediate products from the COX inhibition as well as the enhancement of chemotactic lipooxygenase products formation, result in a comprehensible potential impediment to patient recovery. Further studies to elucidate the precise role of NSAIDs in corneal wound healing are crucial to define the proper use of these drugs in ophthalmic practice. The recent discovery of selective inhibitors of COX-1 and COX-2 provides an opportunity to test hypotheses related to the effects of ophthalmic NSAIDs on corneal wound healing.
The exact mechanisms of ulcerative keratolysis after frequent administration of topical diclofenac sodium in the postoperative period warrant further investigation.
1. Fini ME, Parks WC, Rinehart WB, et al. Role of matrix metalloproteinases in failure to re-epithelialize after corneal injury. Am J Pathol 1996; 149:1287-1302
2. Fini ME, Yue BYJT, Sugar J. Collagenolytic/gelatinolytic metalloproteinases in normal and keratoconus corneas. Curr Eye Res 1992; 11:849-862
3. Ye HQ, Azar DT. Expression of gelatinases A and B, and TIMPs 1 and 2 during corneal wound healing. Invest Ophthalmol Vis Sci 1998; 39:913-921
4. Kreis T, Vale R. Guidebook to the Extracellular Matrix, Anchor, and Adhesion Proteins. Oxford, Oxford University Press, 1999
5. Massova I, Kotra LP, Fridman R, et al. Matrix metalloproteinases: structures, evolution, and diversification. FASEB J 1998; 12:1075-1095
6. Birkedal-Hansen H, Moore WG, Bodden MK, et al. Matrix metalloproteinases: a review. Crit Rev Oral Biol Med 1993; 4:197-250
7. Vane JR. Inhibition of prostaglandin synthesis as a mechanism of action for aspirin-like drugs. Nat New Biol 1971; 231:232-235
8. Vane JR, Bakhle YS, Botting RM. Cyclooxygenases 1 and 2. Annu Rev Pharmacol Toxicol 1998; 38:97-120
9. Lu KL, Wee WR, Sakamoto T, McDonnell PJ. Comparison of in vitro antiproliferative effects of steroids and nonsteroidal antiinflammatory drugs on human keratocytes. Cornea 1996; 15:185-190
10. Haws MJ, Kucan JO, Roth AC, et al. The effects of chronic ketorolac tromethamine (Toradol) on wound healing. Ann Plast Surg 1996; 37:147-151
11. Dvivedi S, Tiwari SM, Sharma A. Effect of ibuprofen and diclofenac sodium on experimental wound healing. Indian J Exp Biol 1997; 35:1243-1245
12. Shimazaki J, Saito H, Yang H-Y, et al. Persistent epithelial defect following penetrating keratoplasty: an adverse effect of diclofenac eye drops. Cornea 1995; 14:623-627
13. Tomas-Barberan S, Fagerholm P. Influence of topical treatment on epithelial wound healing and pain in the early postoperative period following photorefractive keratectomy. Acta Ophthalmol 1999; 77:135-138
14. Donnenfeld ED, Selkin BA, Perry HD, et al. Controlled evaluation of a bandage contact lens and a topical nonsteroidal anti-inflammatory drug in treating traumatic corneal abrasions. Ophthalmology 1995; 102:979-984
15. O'Brien TP, Li QJ, Sauerburger F, et al. The role of matrix metalloproteinases in ulcerative keratolysis associated with perioperative diclofenac use. Ophthalmology 2001; 108:656-659
16. Schoenwald RD. Ocular pharmacokinetics. In: Zimmerman TJ, ed, Textbook of Ocular Pharmacology. Philadelphia, PA, Lippincott-Raven, 1997; 119-138
17. Burstein NL. Corneal cytotoxicity of topically applied drugs, vehicles and preservatives. Surv Ophthalmol 1980; 25:15-30
18. Gasset AR, Ishii Y, Kaufman HE, Miller T. Cytotoxicity of ophthalmic preservatives. Am J Ophthalmol 1974; 78:98-105
19. Baudouin C. Side effects of antiglaucomatous drugs on the ocular surface. Curr Opin Ophthalmol 1996; 7(2):80-86
20. Becquet F, Goldschild M, Moldovan MS, et al. Histopathological effects of topical ophthalmic preservatives on rat corneoconjunctival surface. Curr Eye Res 1998; 17:419-425
21. Talwar M, Moyana TN, Bharadwaj B, Yan LK. The effect of a synthetic analogue of prostaglandin E2
on wound healing in rats. Ann Clin Lab Sci 1996; 26:451-457
22. Vane JR, Botting RM. Anti-inflammatory drugs and their mechanism of action. Inflamm Res 1998; 47(suppl 2):S78-S87
23. Whittle BJR, Higgs GA, Eakins KE, et al. Selective inhibition of prostaglandin production in inflammatory exudates and gastric mucosa. Nature 1980; 284:271-273
24. Dubois RN, Abramson SB, Crofford L, et al. Cyclooxygenase in biology and disease. FASEB J 1998; 12:1063-1073
25. Showell HJ, Pettipher ER, Cheng JB, et al. The in vitro and in vivo pharmacologic activity of the potent and selective leukotriene B4 receptor antagonist CP-105696. J Pharmacol Exp Ther 1995; 273:176-184
26. Wiggins RE, Jafri MS, Proia AD. 12(S)-hydroxy-5,8,10,14-eicosatetraenoic acid is a more potent neutrophil chemoattractant than the 12(R) epimer in the rat cornea. Prostaglandins 1990; 40:131-141
27. Kitaoji H, Kitaoji S, Nakai Y. [Diclofenac sodium ophthalmic solution before and after cataract surgery] [Japanese]. Rinsho Ganka 1991; 45:171-174
28. Takahashi N, Miyakoshi M, Taka N, et al. [Cytotoxic effects of diclofenac sodium] [Japanese]. Nippon Ganka Gakkai Zasshi 1993; 97:145-149
29. Jumblatt MM, Paterson CA. Prostaglandin E2
effects on corneal endothelial cyclic adenosine monophosphate synthesis and cell shape are mediated by a receptor of the EP2
subtype. Invest Ophthalmol Vis Sci 1991; 32:360-365
30. Sun R, Gimbel HV. Effects of topical ketorolac and diclofenac on normal corneal sensation. J Refract Surg 1997; 13:158-161
31. Chen X, Gallar J, Belmonte C. Reduction by antiinflammatory drugs of the response of corneal sensory nerve fibers to chemical irritation. Invest Ophthalmol Vis Sci 1997; 38:1944-1953
32. Aragona P, Tripodi G, Spinella R, et al. The effects of the topical administration of non-steroidal anti-inflammatory drugs on corneal epithelium and corneal sensitivity in normal subjects. Eye 2000; 14:206-210
33. Dinge JT. The effect of NSAID on human articular cartilage glycosaminoglycan synthesis. Eur J Rheumatol Inflamm 1996; 16:47-52
34. Salvatori R, Guidon PT Jr, Papuano BE, Bockman RS. Prostaglandin E1
inhibits collagenase gene expression in rabbit synoviocytes and human fibroblasts. Endocrinology 1992; 131:21-28
35. Fini ME, Girard MT, Matsubara M. Collagenolytic/gelatinolytic enzymes in corneal wound healing. Acta Ophthalmol Suppl 1992; 202:26-33
36. Garrana RMR, Zieske JD, Assouline M, Gipson IK. Matrix metalloproteinases in epithelia from human recurrent corneal erosion. Invest Ophthalmol Vis Sci 1999; 40:1266-1270
37. Flach A. Topically applied nonsteroidal anti-inflammatory drugs and corneal problems: an interim review and comment [editorial]. Ophthalmology 2000; 107:1224-1226
38. Lin JC, Rapuano CJ, Laibson PR, et al. Corneal melting associated with use of topical nonsteroidal anti-inflammatory drugs after ocular surgery. Arch Ophthalmol 2000; 118:1129-1132
39. Congdon NG, Schein OD, von Kulajta P, et al. Corneal complications associated with topical ophthalmic use of nonsteroidal antiinflammatory drugs. J Cataract Refract Surg 2001; 27:622-631