Interface Haze Associated with Topical Nonsteroidal Anti-Inflammatory Drug Use After Descemet Stripping Automated Endothelial Keratoplasty

Introduction

Posterior corneal lamellar keratoplasty by means of Descemet stripping automated endothelial keratoplasty (DSAEK) ¹ has become the procedure of choice for corneal disorders originating from an abnormal endothelium, such as Fuchs dystrophy, pseudophakic bullous keratopathy (PBK), and failed penetrating keratoplasty (PK). ² DSAEK appears similar to PK in regard to complication rates, graft survival/clarity, final visual acuity, and endothelial cell loss. It appears superior in regard to earlier visual recovery, refractive stability, postoperative refractive outcomes, wound and suture complications, and intraoperative and late suprachoroidal hemorrhage risk.^2,3

The most common complications in published reports include graft dislocations—14% (0%–82%); endothelial graft rejection—10% (0%–45%); primary graft failure—5% (0%–29%); and iatrogenic glaucoma—3% (0%–15%). Other less common complications reported include endophthalmitis, retinal detachment, cystoid macular edema (CME), epithelial ingrowth, suprachoroidal hemorrhage, choroidal effusion, pupillary block, and interface issues.^2–6

Interface haze is not uncommon after DSAEK, but it is usually a transient phenomenon with a minimal effect on the final visual outcome.^4–9 In some cases, however, interface abnormalities may persist and decrease best corrected visual acuity (BCVA), cause graft failure, and even require a repeat graft.^4–6 Haze, keratitis, epithelial downgrowth, retained descemet's membrane, and calcareous degeneration ¹⁰ have all been reported as interface complications after DSAEK.^2–8 At least 50% of failed grafts demonstrate some form of interface fibrosis on histopathologic examination.^11–13 The cause is not known in many cases, but our experience suggests that topical nonsteroidal anti-inflammatory drug (NSAID) use may be related. In a prospective, randomized trial comparing ketorolac tromethamine 0.4% with nepafenac sodium 0.1% on the degree of postoperative pain and rate of healing after superficial laser-assisted in situ keratomileusis (epi-LASIK), increased haze was found in the nepafenac-treated group, ¹⁴ and NSAIDs are associated with keratolysis and alteration of matrix metalloproteinase (MMP) regulation. ¹⁵ The purpose of this study was to determine if topical NSAID use was associated with the development of interface haze after endothelial keratoplasty.

Methods

We performed a retrospective study of 100 eyes that did not receive topical NSAIDs postoperatively and compared these to 61 eyes that did receive topical NSAIDs and followed them for 12–24 months. This research was conducted in accordance with the tenets of the Declaration of Helsinki and with the approval of the University of Oklahoma Health Sciences Center Institutional Review Board. Information was collected regarding diagnosis, type of surgery performed, postoperative medication regimen, complications, comorbidities, and development of CME. Primary outcome measures included the presence or absence of interface haze and final BCVA. All surgeries were performed or supervised by a single surgeon (D.U.S.), and cases and controls were selected from the same time period to decrease the effect of changes in surgical techniques or other factors. The technique was as described elsewhere, ¹⁶ with key features, including stripping of Descemet membrane with maintenance of the anterior chamber with hyaluronate 1% (Healon; Abbott Medical Optics, Santa Ana, CA), scraping of the peripheral stroma, use of microkeratome precut endothelial tissue, and manual insertion using forceps through a 5 mm wound. All patients received postoperative topical fluoroquinolone antibiotics and prednisolone acetate 1%. In the patients receiving NSAIDs, nevanac dosing was twice a day or 3 times a day. Patients receiving diclofenac or ketorolac were dosed 4 times a day. All patients received NSAIDs for a duration of 3 to 4 weeks. For patients who underwent triple procedure or DSAEK alone, there was no difference in the dosing of prednisolone acetate: 1% 4 times a day for 2–3 months, then tapered 1 drop per day each 2–3 months thereafter (3 times a day, twice a day, then once daily). All patients received the fluorquinolone for 1 week for all surgeries. If present, interface haze was documented as mild, moderate, or severe. Statistical significance was determined by chi-squared analysis with P<0.05.

Results

In the group that did not receive postoperative NSAIDs, 9/100 (9%) developed interface haze. In the group that received postoperative NSAIDs, 31/61 (51%) developed haze. Postoperative topical NSAID use was found to be significantly associated with the development of interface haze (P=0.0001). The risk ratio was 5.7 and the odds ratio was 10.5. Other characteristics of each group are displayed in Table 1; there was no statistically significant difference in these variables or outcomes.

In the patients who developed haze, all but one was a transient phenomenon, lasting at most 4 months. The patient who developed moderate-severe haze had prolonged haze even at the last follow-up visit (2 years postoperatively), but was still correctable to 20/30. One patient did not demonstrate haze at his initial postoperative visit, but did show transient mild haze after being started on ketorolac to treat CME. In this series, no repeat grafts were performed to treat intractable haze.

Of the patients who developed haze, 24/31 (77%) were on nepafenac; 11 dosed twice a day (BID), and 13 dosed 4 times a day (QID). Ketorolac-treated patients made up the remaining 7/31 (23%), all dosed QID. Overall, 49 of the 61 patients who received postoperative NSAIDs were given nepafenac. Of the patients placed on nepafenac, 24/49 (49%) developed haze. Of the patients placed on ketorolac, 7/10 (70%) developed haze. Of the patients placed on diclofenac, 0/2 (0%) developed haze. The difference in the incidence of haze between the NSAID-treated groups was not statistically significant (P=0.31). Only 5/100 (5%) in the non-NSAID group developed CME and 12/61 (19%) in the NSAID group developed CME.

Discussion

Lamellar corneal surgeries have long been known to be associated with interface effects on vision; modern endothelial keratoplasty may reduce, but does not eliminate this issue. Histopathologic studies in successful DSAEK eyes demonstrated that successful grafts appear to heal without abundant scarring or keratocyte proliferation. ¹⁷ The determining factors associated with the development of interface haze have not been definitively elucidated. One of the obvious drawbacks of the current study is that it is a retrospective study and susceptible to bias; interface haze was also not measured objectively. The incidence of interface haze reported herein is also higher than in other studies; this difference could be attributed to differences in subjective determination of haze, differences in surgical techniques (such as retained viscoelastic), differences in the studied patient population, or other undetermined factors. Another possible confounder of this study is that patients who received NSAIDs were more likely to be undergoing a triple procedure (Table 1). Other potential factors that we hypothesize may contribute to interface haze include UV exposure, ¹⁸ fluoroquinolone activation of MMP, ¹⁹ relative vitamin C deficiency, ²⁰ and other patient, surgical, or donor characteristics that were not assessed or controlled in this study.

Conclusions

In our study, topical NSAIDs were associated with the development of interface haze after DSAEK. Most cases tended to be mild and transient; however, there was only a small risk of CME if prophylactic postoperative NSAIDs were not used. Interestingly, the patients using NSAIDS had a higher rate of postoperative CME. Given the retrospective nature of the data, this may represent selection bias for a group that was at higher risk for CME. The group with interface haze had no significant difference in postoperative visual acuity. Further study is needed to determine the factors that contribute to interface haze in lamellar corneal endothelial surgeries; future prospective, randomized studies with objective determination of interface characteristics are needed to definitively answer the questions raised herein.