Price MO, Thompson RW, Price FW. Risk Factors for Various Causes of Failure in Initial Corneal Grafts. Arch Ophthalmol. 2003;121(8):1087-1092. doi:10.1001/archopht.121.8.1087
Copyright 2003 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2003
To determine the risk factors for specific causes of initial corneal graft failure.
This study analyzed corneal graft survival rates in a longitudinal noncomparative case series of 3992 consecutive penetrating keratoplasties performed at a single large referral center. Regrafts (n = 352) were excluded from the analysis of risk factors for initial graft failure. Data were collected retrospectively from August 1, 1982, through December 31, 1986, and prospectively from January 1, 1987, through August 31, 1996. Patients were examined preoperatively, at 1, 3, 6, 9, 12, 18, and 24 months posttransplantation, and annually thereafter. Potential risk factors were evaluated individually by Kaplan-Meier survival analysis. Cox proportional hazards regression modeling was then used to investigate the impact of each independent variable, adjusted for the confounding influence of the other independent variables.
The use of topical glaucoma medications was a significant risk factor for corneal graft failure by 3 major causes: rejection, endothelial decompensation without a documented immunologic reaction, and ocular surface disease. Deep stromal vascularization was an independent risk factor for rejection failure. Diabetes mellitus, peripheral anterior synechiae, recipient race, and small trephination size were significant risk factors for endothelial failure.
Independent risk factors differentially impact specific causes of corneal graft failure.
PENETRATING KERATOPLASTY (PKP) is one of the most successful and frequently performed transplantation procedures. Recent long-term studies1- 3 report 10-year corneal graft survival rates of 59% to 80%. Because more than 40 000 corneal transplantations are performed each year, 4 the cumulative number of graft failures is significant despite the relatively high survival rates. It is particularly important to prevent initial grafts from failing, because survival rates for regrafts are greatly reduced.2,3,5- 8 The identification of risk factors for initial graft failure can help surgeons determine which eyes are at increased risk, so that those eyes can be more aggressively treated.
Graft failure can be broadly defined as any irreversible change in the graft preventing recovery of useful vision. Risk factors for graft failure in general include prior graft failure in that eye, glaucoma, 1,3,5 peripheral anterior synechiae, 3,5,6 and corneal vascularization.2,3,5- 8 However, graft failure represents a common end point that can be reached by distinctly different pathways. Among the most frequent causes of graft failure are immunologic allograft rejection, endothelial decompensation in the absence of any documented immunologic reaction episodes, ocular surface disease, glaucoma, and astigmatism.3,9,10 This study determines how different risk factors impact various failure causes in initial corneal grafts, because a better understanding of causes and effects could ultimately lead to improved treatment methods.
This study reports findings from a longitudinal noncomparative case series of 3992 consecutive PKPs performed by surgeons at a tertiary care referral center, Corneal Consultants of Indiana, between August 1, 1982, and August 31, 1996. Regrafts (n = 352) were excluded from the analysis. Data were collected retrospectively through December 31, 1986, and prospectively thereafter. Each patient provided written informed consent.
Surgical technique and postoperative care were as previously described.11 All patients underwent a complete ophthalmic examination before PKP, and medical history was documented. Defined postoperative examinations were performed at 1, 3, 6, 9, 12, 18, and 24 months and annually thereafter.
Grafts that did not clear within 2 weeks of transplantation were considered primary donor failures. These events were rare, occurring in only 21 eyes(0.5%). Because primary donor failures were caused by difficulties with donor tissue survivability unrelated to recipient characteristics, when these grafts were repeated, the repeat grafts were entered into the database as initial grafts for outcome analysis. Any graft that initially cleared but later irreversibly changed, preventing recovery of useful vision, was counted as a graft failure. The time of failure was defined as the first postoperative examination that the patient was seen with a failed graft. An eye was dropped from further analysis when graft failure was recorded. An eye was considered lost to follow-up when a patient had not been examined for more than 1 year from the first missed postoperative examination period. There were 2667 initial grafts with 2 or more years of follow-up and 1250 with 5 or more years of follow-up.
Grafts that failed because of endothelial decompensation were classified as "failures by rejection" if there was any postoperative history of possible allograft rejection, such as ciliary flush with keratic precipitates on the donor endothelium, definite rejection episodes such as an endothelial rejection line, or signs of stromal or epithelial rejection. A graft failure was classified as a "rejection failure" even if the endothelial failure occurred months or years after any immunologic reactions. Grafts that failed because of endothelial decompensation in the absence of any documented immunologic reactions were classified as "endothelial failure." Grafts that failed as a result of ocular surface-related problems, such as melts, ulcers, or exposure keratitis, were classified as "surface disease failures." Less prevalent failure modes, such as astigmatism, did not result in a sufficient number of failures to perform separate risk factor analyses.
A history of preoperative glaucoma was determined by the patient's use of medications to reduce intraocular pressure. Stromal vascularization was defined as vascularization of the middle or deep stroma, extending into the area of trephination at surgery. Eyes first seen with a diagnosis of interstitial keratitis were only classified as positive for stromal vascularization if clinically significant blood flow was present.
To adjust for varying lengths of follow-up, survival analysis was used to individually evaluate possible prognostic risk factors for graft failure. Survival curves were generated using the Kaplan-Meier product limit method, and differences between subgroups were assessed using the log-rank test. Relative risks and adjusted survival probabilities were estimated using the Cox proportional hazards regression model with forward and backward selection algorithms. Multivariate models were separately determined for any cause of graft failure and for specific causes of graft failure. Whenever many characteristics are analyzed, there is an increased risk of reporting a type I error (a false-positive error). P<.05 was considered significant. However, the greatest confidence can be placed in those risk factors having the lowest P values. The statistical power to detect risk factors increases with sample size, and was higher for the analysis of all graft failures (n = 272) compared with the separate analyses of the endothelial failures (n = 82), rejection failures (n = 76), and surface complication failures (n = 49).
The patient cohort was 63.8% female, and the mean patient age was 67 years (range, 1-98 years). The mean ± SD follow-up was 43 ± 34 months (range, 1 month to 15 years). The primary preoperative diagnoses for the 3640 eyes undergoing an initial PKP were as follows:
As described elsewhere, 12 overall graft survival rates for this patient cohort varied depending on the preoperative diagnosis, with patients with keratoconus and Fuchs dystrophy having the highest graft survival rates, 97.4% and 96.9%, respectively, at 5 years. Graft survival rates were also favorable, although somewhat lower, for patients with other preoperative diagnoses (88.1% for those with pseudophakic bullous keratopathy, 83.2% for those with aphakic bullous keratopathy, and 80.2% for those with corneal scars or ulcers) at 5 years. When the initial diagnoses were grouped into 2 categories, lower risk (keratoconus and Fuchs dystrophy) and higher risk (all others), the relative risk of graft failure was 3.5 (95% confidence interval, 2.6-4.9) times greater for the higher-risk group (P<.001).
Graft failures occurred in 272 (7.5%) of 3640 initial grafts. Rejection, endothelial decompensation without an immunologic reaction, and ocular surface disease were the primary causes of graft failure, accounting for 27.9%, 30.1%, and 18.0% of the failed grafts, respectively. The remaining graft failures resulted from astigmatism (4.8%), glaucoma (2.9%), and other less frequent causes (together, 16.2%).
Characteristics considered potential risk factors for graft failure were first individually evaluated using Kaplan-Meier survival analysis. Survival analysis adjusts for different lengths of follow-up by evaluating all time points to determine if there is a significant difference between groups. Recipient characteristics having a significant effect on initial graft failure rates are shown in Table 1, along with comparative failure rates at the 5-year point. Separate survival analyses were performed to identify risk factors for the major failure causes.
The risk factors most strongly associated with overall graft failure were a history of preoperative glaucoma medication use, deep stromal vascularization, peripheral anterior synechiae, recipient race, and unusually small or large recipient bed sizes (Table 1). A small recipient horizontal corneal diameter and diabetes mellitus were also significant risk factors for graft failure by any cause.
Most of the identified risk factors were associated with only 1 or 2 of the major categories of graft failure, with the exception that use of a glaucoma medication was a risk factor for all 3. The presence of peripheral anterior synechiae or being African American increased the risk of failure by rejection or endothelial decompensation, but not by ocular surface disease. Deep stromal vascularization specifically increased the risk of failure due to rejection. Diabetes mellitus specifically increased the risk of endothelial decompensation, as did a recipient horizontal corneal diameter of 10.75 mm or less.
The choice of recipient and donor tissue trephine sizes was influenced by recipient corneal diameter size, vascularization, and scarring. Overall, variation in recipient horizontal corneal diameter size accounted for 49% of the variance in recipient trephine size. Small recipient beds (≤7 mm) were associated with an increased risk of endothelial failure and rejection failure (Table 1). Results were similar for unusually small donor button sizes (≤7.25 mm) (data not shown). The sizes of the recipient bed and donor button were highly correlated (R2 = 0.89), because it was routine practice to use a donor trephine size of approximately 0.25 mm larger than the recipient trephine size for all initial diagnoses other than keratoconus. For keratoconus, the donor button was often sized the same as, or slightly smaller than, the recipient bed.
An unusually large recipient bed size (>8.5 mm) was associated with an increased risk of failure by rejection or ocular surface disease (Table 1). However, only 23 grafts had a recipient bed exceeding 8.5 mm. Thirteen were placed to treat large cones in eyes with keratoconus, and none of those 13 grafts failed during the study. In contrast, the remaining 10 grafts were primarily placed to treat large corneal scars and ulcers, and 5 of those 10 failed. This was well in excess of the 13.1% graft failure rate for any eyes first seen with corneal scars or ulcers (P<.001). These results indicate that the outcomes for grafts exceeding 8.5 mm primarily depended on the underlying corneal condition requiring the use of an unusually large graft.
The surgeon's choice of suture method was influenced by the characteristics of the recipient eye, and interrupted sutures were often selected for high-risk recipients. For example, interrupted sutures were used in 83.1% of the eyes with deep stromal vascularization, compared with only 27.1% of the eyes without such vascularization, a significant difference (P<.001). Interrupted sutures were also preferentially selected for small (≤10.75-mm) recipient corneal diameters (P<.001). Grafts secured with interrupted sutures were more likely to fail by all 3 major causes compared with grafts secured with running or combination sutures (P≤.002). However, this was likely because of underlying recipient risk factors, rather than a direct effect of suture method, because grafts secured with interrupted sutures were no more likely to fail than those secured with running sutures in eyes with keratoconus, which are generally low-risk eyes (P = .22).3,9
Recipient characteristics that did not significantly influence graft failure rates included sex, age, and hypertension (Table 2). The placement of an intraocular lens secondary to PKP had no impact on graft survival rate, nor did the size difference between the recipient horizontal corneal diameter and the recipient bed (Table 2). No donor corneal characteristics had a significant effect on graft outcome. Characteristics that were analyzed included donor age (mean, 39 years; SD, 16 years) (P =.16), sex (70.2% male) (P = .65), death to enucleation time (mean, 5.7 hours; SD, 3.7 hours) (P = .60), time to cornea preservation (mean, 6.4 hours; SD, 4.3 hours) (P = .28), time to use (mean, 85 hours; SD, 33 hours) (P = .69), corneal storage solution (McCarey-Kaufman, K-Sol, Dex Sol, Optisol, or corneal storage medium) (P = .94), and cause of death (58.2% medical, 20.1% violence, and 21.7% other trauma or unintentional injuries) (P = .59). (Kaplan-Meier survival analysis and the log-rank statistic were used for continuous variables, and a Cox proportional hazards regression analysis was used for categorical variables.) Donor age was not randomized in this study, because it was routine practice to use donor tissue younger than the recipient whenever possible.
The potential risk factors identified by survival analysis (Table 1) were correlated to some extent. For example, eyes with peripheral anterior synechiae were more likely than other eyes to also have glaucoma (P<.001). To investigate the impact of each independent variable, adjusted for the confounding influence of the other independent variables, multivariate Cox proportional hazards regression models were analyzed.
Five significant risk factors were included in the final regression model for failure by any cause (Table 3). Patients classified as having glaucoma before PKP, based on their use of glaucoma medications, were twice as likely to experience graft failure as patients who did not have preoperative glaucoma. The presence of peripheral anterior synechiae or deep stromal vascularization each independently increased the relative risk of overall graft failure by 2-fold. African Americans were twice as likely to experience graft failure compared with white persons or persons of other races. In addition, recipient bed sizes of 7 mm or less were at almost a 3-fold higher risk of graft failure compared with larger bed sizes.
The final regression model for failure by immunologic rejection contained 3 independent risk factors (Table 3).Eyes with preoperative glaucoma were at a 2-fold increased risk of failure due to rejection compared with eyes without preoperative glaucoma (Table 3). Eyes with deep stromal vascularization were almost 3 times more likely to experience rejection failure compared with eyes without such vascularization. In addition, the risk of rejection failure was 3-fold higher in eyes with unusually small recipient beds (≤7 mm) compared with those with larger recipient beds (>7 mm).
Five independent risk factors were identified in the final model of failure by endothelial decompensation without an immunologic reaction (Table 3). The use of glaucoma medications increased the risk of endothelial failure by almost 2-fold. Peripheral anterior synechiae conferred a 3-fold increased risk of endothelial failure. The risk of endothelial failure was twice as high in patients with diabetes mellitus as in patients without diabetes mellitus. African Americans had a 3.6-fold increased relative risk of endothelial decompensation compared with white persons and others. Unusually small recipient beds (≤7 mm) were associated with an almost 5-fold increased relative risk of endothelial decompensation compared with recipient beds exceeding 7 mm.
The only significant risk factor for ocular surface-related failures was use of glaucoma medications, which conferred a 3-fold increased risk (Table 3).
The purpose of this study was to identify the risk factors associated with distinct causes of graft failure in initial corneal transplantations. We individually examined risk factors for rejection, endothelial decompensation without evidence of rejection, and ocular surface disease, because these were the most prevalent causes in the patient cohort and together accounted for 76.1% of the initial graft failures. We believe it is particularly important to anticipate and prevent failure of initial grafts, because regrafts have a much poorer prognosis. Although prior studies5,10,13 have identified risk factors for rejection failures, to our knowledge, this is the first study to determine risk factors not only for rejection but also for endothelial decompensation and ocular surface-related failure causes.
Patients using glaucoma medications were at an increased risk of graft failure by all 3 major causes, even after adjusting for other confounding risk factors (Table 3). The incidence of preoperative glaucoma was 15%, comparable to published rates of 9% to 28%.1,14,15 Overall, grafts in eyes with preoperative glaucoma were twice as likely to fail as those in eyes that were not being treated for glaucoma, supporting findings in other studies.1,14
Preoperative glaucoma was an independent predictor of endothelial failure, consistent with clinical observations by one of us (F.W.P.) that a high intraocular pressure can lead to rapid endothelial decompensation, perhaps through direct damage to endothelial cells.16 Interestingly, patients being treated with glaucoma medications were also at a significantly increased risk for rejection and ocular surface-related failures. Increased leukocyte and fibrocyte accumulation in conjunctival and limbal tissue has been reported secondary to the use of topical glaucoma medications.17,18 This inflammatory cell response may increase the likelihood of recipient immunologic recognition of donor tissue, leading to immunologic graft rejection. The inflammatory reaction seems to be primarily related to the preservatives, such as benzalkonium chloride, rather than the medications themselves.18,19 Benzalkonium chloride has cumulative toxic effects on the corneal-conjunctival surface, and its detergent properties can also adversely alter tear film fluidity and breakup time.20 This may be particularly problematic following PKP, because the donor corneas are neurotrophic and, therefore, more susceptible to exposure and toxicity. Therefore, in patients with glaucoma, transplantation surgeons should consider the use of glaucoma medications with noninflammatory preservatives (such as sodium chlorite [Purite]), the use of noncorticosteroid immunosuppressants after PKP to minimize corticosteroid-responsive increases in intraocular pressure, or earlier intervention with surgical alternatives for intraocular pressure control. In the United States, there is only 1 approved topical glaucoma medication using sodium chlorite, brimonidine tartrate (Alphagan-P; Allergan, Inc, Irvine, Calif). The most potent topical antiglaucoma medications are the prostamides, 21 but they have inflammatory preservatives, and this group of drugs seems to have an inherent propensity to cause increased inflammation in some eyes. None of the eyes in our patient cohort received topical prostamides, and further studies will need to be performed to evaluate their effect on graft survival.
Further studies comparing the effect of earlier surgical intervention with long-term use of glaucoma medications on corneal graft survival are also needed. Increased graft failure rates have been reported following certain types of glaucoma surgery, particularly with the use of glaucoma setons.22,23 Flat anterior chambers, which can directly damage the endothelium of the graft, often develop following standard filtration surgery with trabeculectomy; this procedure is also prone to complications from hypotony, such as suprachoroidal hemorrhages and cataract progression. Newer nonpenetrating glaucoma filtration techniques have a lower complication rate compared with trabeculectomy.24,25 Because of the reduced incidence of complications, such as flat anterior chambers and suprachoroidal hemorrhages, nonpenetrating filtration surgical procedures may provide a valuable alternative to multiple long-term topical medications for intraocular pressure control after PKP.
Peripheral anterior synechiae were associated with an increased risk of rejection failures and endothelial failures without rejection episodes. Anterior synechiae place the donor endothelium in closer proximity to blood vessels that may increase the risk of rejection.16 Furthermore, anterior synechiae may accelerate loss of endothelial cells through either direct traction or indirect increased anterior segment inflammation.16 Our findings support studies showing that peripheral anterior synechiae can increase the overall graft failure rate6,13,26 and the rejection failure rate.13,27
Deep stromal vascularization increased the risk of failure due to rejection by 2.7-fold (P<.001), but had no impact on rates of ocular surface-related failures or endothelial failures without an immunologic reaction. Corneal vascularization has been defined in different ways in various studies. Our results support the concept that vascularization can facilitate an immunologic response against the corneal graft.14,27- 29
Diabetes mellitus was specifically a risk factor for endothelial failure. Although diabetes mellitus is not typically investigated as a risk factor for corneal graft failure, it is associated with excess glycosylation of blood proteins, impaired circulation, breakdown of the blood/aqueous barrier, and deterioration of various organ systems.
African Americans comprised a relatively small minority of the patient cohort, reflecting the racial distribution in Indiana during the study period.30 African Americans experienced a 3- to 4-fold higher risk of endothelial failure, compared with white persons and others, despite similar rates of follow-up. Another study13 did not find recipient race to be a risk factor for graft failure, perhaps reflecting differences in patient demographics.
Recipient beds of 7 mm or smaller were used infrequently and only as necessary to accommodate a small recipient corneal diameter, scarring, or vascularization. Grafts with small recipient beds (≤7 mm) were at almost a 5-fold higher risk of endothelial failure compared with those with larger beds (Table 3). Smaller grafts may be more prone to endothelial failure because they have a smaller pool of intact endothelial cells to maintain corneal clarity. An unusually small recipient bed size was also associated with an increased risk of rejection failure, consistent with findings in high-risk eyes in the Collaborative Corneal Transplantation Studies.13 Recipient bed sizes of less than 7 mm were likewise associated with an increased risk of overall graft failure in the large Australian Corneal Graft Registry.3
Neither recipient age nor sex affected graft outcomes in this large cohort of initial grafts. There are conflicting reports1,3,6- 8,13,28,29,31- 33 in the literature regarding these characteristics. None of the donor corneal characteristics analyzed in this study had an impact on graft outcome. This is generally consistent with the consensus of other studies.3,13,14,34 Donor age was not randomized in our patient cohort, because it was routine practice to use donor tissue younger than the recipient whenever possible. The effect of donor age is under further investigation in the prospective randomized Cornea Donor Study.35
In conclusion, independent risk factors were differentially associated with specific causes of initial corneal graft failure. The use of topical glaucoma medications was a significant risk factor for corneal graft failure by 3 major causes: rejection, endothelial decompensation without a documented immunologic reaction, and ocular surface disease. Deep stromal vascularization was an independent risk factor for rejection failures. Diabetes mellitus, peripheral anterior synechiae, recipient race, and small trephination size were significant risk factors for endothelial failure in the absence of an immunologic reaction.
Corresponding author and reprints: Francis W. Price, Jr, MD, Corneal Consultants of Indiana, 9002 N Meridian St, Suite 207, Indianapolis, IN 46260(e-mail: email@example.com).
Submitted for publication September 5, 2002; final revision received March 7, 2003; accepted March 25, 2003.
This study was supported by a grant from the Indiana Lion's Eye Bank, Indianapolis.
We thank Barry Katz, PhD, Indiana University, Indianapolis, for his assistance in performing the analyses for this study.