Journal Logo

Original Articles: Clinical Transplantation

How Effective Is Penetrating Corneal Transplantation? Factors Influencing Long-Term Outcome in Multivariate Analysis

Williams, Keryn A.1,3; Esterman, Adrian J.2; Bartlett, Christine1; Holland, Helene1; Hornsby, Ngaere B.1; Coster, Douglas J.1 on behalf of all contributors to the Australian Corneal Graft Registry

Author Information
doi: 10.1097/
  • Free


The surgical procedure of corneal transplantation, first reported a century ago in 1906 (1), is widely used to ameliorate the blinding consequences of corneal disease, which is the second most common cause of visual loss on a worldwide scale (2). Early clinical success (1), coupled with celebrated laboratory studies describing the anterior segment of the eye as an immune-privileged site (3), led to a widely-held belief that corneal transplantation is highly successful and is exempt from the immunological limitations of essential organ transplantation. Reporting of short-term data on corneal graft survival has tended to perpetuate this view.

Registries have been used to great effect in other forms of clinical transplantation and can be credited with the quantification of long-term graft survival. The Australian Corneal Graft Register was established in 1985 to examine long-term corneal graft outcomes (4). Approximately 11,000 corneal grafts performed by over 300 ophthalmic surgeons have now been followed for periods of up to 18 years, allowing estimates to be made of overall outcome in the longer term and in disparate contexts. Here, we assess whether corneal graft survival has shown an improvement commensurate with that seen for other clinical transplants over the past few decades, and identify risk factors associated with graft failure in the preoperative, intraoperative and postoperative phases.


Data Collection

Records were submitted to the Registry by the contributing surgeon as soon as possible after the graft and follow-up was requested at intervals of 12 months. Missing data were routinely sought by follow-up letter. Each graft was followed at yearly intervals until graft failure or until the death or loss to follow-up of the patient. Individual surgeons handled the consenting process for each patient, to permit information to be lodged with the register. The host institutional committee on clinical investigations provided approval for the operations of the register, which were carried out in accordance with the Declaration of Helsinki. All information submitted was amalgamated and de-identified prior to analysis.

The study period for the present analyses was May 1985 to July 2003. Information on the graft recipient, donor, eye bank, operative procedure and postoperative course was collected as previously described (4, 5). Records were submitted by a total of 580 individuals, 333 of whom were ophthalmic surgeons who carried out the transplant procedure and 247 of whom were practitioners involved in follow-up.


At the census date the Register held records of 13,831 penetrating (full-thickness) corneal grafts, of which 10,952 (79%) had been followed on at least one occasion. Recipient age at the time of transplantation varied from less than one year to more than 100 years. Approximately equal numbers of women (49%) and men (51%) had received a graft. Recipients of 676 grafts (6% of those followed) were known to have died, and 4,189 grafts (38% of those followed) had been lost to annual follow-up by the census date. Given that the median recipient age at graft was 59 years, it is likely that a proportion of those recorded as lost to follow up had actually died. Data on ethnicity were not collected.

Statistical Analyses

Kaplan-Meier survival functions were constructed to provide a graphical record of graft survival (6–8). For surviving grafts, trial time was calculated as the time between the date of graft and the date on which the patient was last seen. For failed grafts, trial time was calculated as the time between the date of graft and the date of failure. No exclusions were applied. Kaplan-Meier plots were constructed using SPSS version 11.5. Cox proportional hazards regression analysis was performed using Stata version 8. Of 9,109 patients (as distinct from grafts) with follow-up, 22% had had more than one registered graft in the ipsilateral and/or contralateral eye. To control for potential inter-graft and/or inter-eye dependence in the multivariate analyses, the Cox model was adjusted to allow for clustering by individual patient (9). The best model was found by a non-automatic backward elimination process, removing variables not appearing to be predictors of graft failure.


Overall Graft Survival and Graft Survival by Era

Penetrating corneal graft survival for the cohort is presented in Figure 1. Kaplan-Meier probability of survival was 0.86 at 1 year, 0.73 at 5 years, 0.62 at 10 years, and 0.55 at 15 years. The major reasons for graft failure are documented in Table 1; irreversible corneal graft rejection accounted for approximately one-third of cases. Comparisons of the survival of penetrating grafts carried out within 3-4 year blocks (1985–87, 1988–90, 1991–93, 1994–96, 1997–99, 2000–03) since the register’s inception are shown in Figure 2; corneal graft survival did not show an improvement with era.

Kaplan-Meier survival plot of all penetrating corneal grafts. The numbers at risk at initially and thereafter at intervals of 3 years are shown above the plot.
Reasons for the failure of penetrating corneal grafts
Kaplan-Meier corneal graft survival plots stratified according to era. Comparisons of the survival of penetrating grafts carried out within 3–4 year blocks (1985–87, 1988–90, 1991–93, 1994–96, 1997–99, 2000–03) since the register’s inception. The numbers at risk at intervals of 3 years are shown in the table beneath the plots.

Multivariate Analysis

Following on from the results of Kaplan-Meier analyses, which were used to indicate variables of interest, Cox proportional hazards regression modeling was used to investigate the joint effects of a subset of 27 variables on penetrating corneal graft failure. The assumption of proportional hazards appeared reasonable as assessed by the Kaplan-Meier plots. Table 2 tabulates the parameter estimates resulting from the fit of the best clustered Cox model. The first level of each categorical variable was taken as the referent. The hazard ratios for a given variable were adjusted for all other variables in the model.

Multivariate analysis of factors influencing corneal graft survival: parameter estimates from final Cox regression model

The variables that were retained in the final model were transplant centre (16 surgeons with high case-loads compared with all others combined), indication for graft (stratified according to the most common preoperative diagnoses), number of previous ipsilateral grafts, corneal neovascularisation at the time of graft (stratified according to the number of quadrants of the cornea containing vessels), history of inflammation in the grafted eye, history of raised intraocular pressure in the grafted eye, donor age (considered as a continuous variable), graft diameter (in mm), lens status (phakic, pseudophakic or aphakic), requirement for anterior vitrectomy at the time of graft, early removal of graft sutures, postoperative neovascularisation of the graft, occurrence of immunological rejection episodes, postoperative rise in intraocular pressure, type of treatment for glaucoma after graft (none, medication only, surgical intervention only, medication plus surgical intervention), and arrangements for recipient follow-up.

The factors that did not influence penetrating corneal graft survival significantly in multivariate analysis were era of graft (examined in 5 year blocks), surgeon case-load (<25 grafts/year compared with ≥25 grafts/year), cause of donor death (stratified into major categories of death from: cardiac/circulatory disease; cerebrovascular disease or event; malignancy; trauma or accidental death; respiratory disease), retrieval of the donor cornea from a multi-organ donor (by comparison with a cadaveric donor), identity of eye bank providing the donor cornea (each of 5 licensed eye banks in Australia), recipient age at graft (considered as a continuous variable), history of recipient blood transfusion prior to corneal transplantation, history of recipient pregnancy prior to corneal transplantation, development of postoperative uveitis, development of synechiae after transplantation, or occurrence of a postoperative stitch abscess.


Despite the accumulated clinical experience of almost a century (1), and notwithstanding the immune-privilege enjoyed by the cornea and anterior segment of the eye (3), we report that the survival of human corneal allografts is no better in either the shorter or longer term than that of many other organ transplants (10–12). Corneal graft survival falls steadily from the time of surgery to 55% at 15 years. Furthermore, the results have failed to improve over the past 15 years, a period in which outcomes for solid organ grafts have generally (13–15), although not invariably (16, 17), improved steadily. A similar constancy in corneal graft survival rates with time has previously reported in a series of 1,681 corneal grafts performed by a single surgeon from a single centre over a twenty-year period (18). The ever-increasing specificity and efficacy of systemic immunosuppression probably accounts for some of the improvements observed in solid organ graft survival in recent years. Unfortunately similar strategies are not necessarily applicable to corneal transplantation. Systemic immunosuppression has only a limited place in corneal transplantation because of the associated morbidities. A patient requiring a corneal graft is not in a life-threatening situation and the serious side-effects which can result from long-term immunosuppression are seldom justified (19).

Multivariate analysis, clustered for repeat and bilateral grafts (9), was used to identify those independent variables that were associated with corneal graft failure. The so-called centre effect exerted a significant influence on corneal graft survival that appeared to be independent of both case-mix and case-load, with hazard ratios varying from 0.3 to 2.55. The differences in outcomes observed amongst centres are not readily explained, but are possibly influenced by differences in postoperative management (20). It may be of interest in this respect that a previous survey of Australian ophthalmologists revealed great variation in preferred regimens of immunosuppression for the prophylaxis and treatment of corneal graft rejection (21). Topical glucocorticosteroids are almost always prescribed, but the evidence-base to support one formulation over another, or one regimen of administration over another, is lacking. Our database contains records of only 68 followed corneal grafts (of a total of 10,952) in patients for whom immunosuppression with cyclosporin A, tacrolimus, azathioprine, cyclophosphamide, or methotrexate (sometimes in combination) was used; 27 of these grafts have failed. Differences in patient ease of access to expert ophthalmic care in the postoperative period may also account for the striking centre effect. A centre effect has also previously been reported to operate for visual outcome (as distinct from graft survival) after corneal transplantation (22).

Not surprisingly, the well-established influence of recipient preoperative diagnosis on corneal graft survival (4, 5, 20, 22–24) was confirmed, with hazard ratios varying from 1.00 to 2.60 for common indications for transplantation. For patients with keratoconus, graft survival was impressive; all other patients fared less well. The percentage of penetrating corneal grafts performed for herpes simplex virus (HSV) infection has fallen slightly, from 5.5% over the first 6 years of the Registry’s operations to 4.9% (cumulative total) at present. Furthermore, the percentage of grafts that have failed from the sequelae of herpetic infection has decreased from 6.8% (1985–1991) to 3.5% (cumulative total) currently, possibly because of improvements in the prophylaxis and treatment of ocular HSV infection in recent years. However, herpetic eye disease as the main indication for corneal transplantation remains a significant risk factor for subsequent graft failure, with a hazard ratio of 2.6, compared with 1.00 for keratoconus as the main indication for graft. Graft survival was progressively reduced by repeated transplantation in the same eye. The erosion of normal corneal immune privilege by corneal neovascularisation, the sequelae of preoperative inflammation, and raised intraocular pressure was also well illustrated by the results of the multivariate analysis.

Of interest was the paucity of variables relating to the corneal donor that contributed to the final Cox model. Of all the factors relating to the donor or to eye bank procedures that were examined, only increasing donor age was shown to exert a deleterious but relatively weak influence on corneal graft survival, with a hazard ratio of 1.05 for every increasing decade of donor age. Given that fewer than 5% of corneas were retrieved from multi-organ donors and that 56% corneal donors were aged over 60 years at the time of death, the extremely weak effect of increasing donor age on subsequent corneal graft survival was reassuring. Recent evidence suggests that corneal endothelial cell count, rather than donor age, may be the more powerful predictor of long-term corneal graft function (25).

Of the intra-operative factors examined, the requirement for an anterior vitrectomy or the use of a rather small (<7.5 mm diameter) or large (≥9 mm diameter) graft were associated with increased risks of graft failure. Both have been identified as risk factors previously (4, 5, 26, 27). In the postoperative period, the absence of the crystalline lens (aphakia) or presence of an intraocular lens (pseudophakia), the removal of all graft sutures within 6 months of transplantation, neovascularization of the graft, the occurrence of inflammation, raised intraocular pressure or a rejection episode were all independently associated with an increased risk of graft failure. It was of interest that institution of treatment for raised intraocular pressure appeared to reduce the risk of failure, compared with no treatment.

Clearly survival is only one measure of corneal graft outcome, with successful visual rehabilitation and pain relief being other important factors in determining patient satisfaction after transplantation. However, graft failure compromises both visual function and patient comfort. How may corneal graft survival be further improved? Living-related donation, which consistently yields better survival rates in essential organ transplantation than cadaveric transplantation (28), is not an option for corneal transplantation. The improvements in solid organ graft survival over the past twenty years are attributable to better immunosuppression, as described above, and to effective histocompatibility matching. Opinion is currently divided over whether HLA matching in corneal transplantation has merit, but the weight of evidence now suggests a beneficial effect, at least in recipients considered to be at high risk of immunological rejection (18, 29–32).

If novel developments in essential organ transplantation cannot be adapted for corneal transplantation, then cornea-specific strategies must be developed. Such developments require a better understanding of factors underlying the centre effect, and of the fundamental mechanisms associated with corneal allograft failure, including both rejection and late corneal endothelial cell failure (33).


The authors gratefully acknowledge the voluntary contribution of records by 580 ophthalmologists and other practitioners across Australia.


1. Zirm EK. Eine erfolgreiche totale keratoplastik. Archiv für Ophthalmol 1906; 64: 580.
2. Whitcher JP, Srinivasan M, Upadhyay MP. Corneal blindness: a global perspective. Bull World Health Org 2001; 79: 214.
3. Medawar P. Immunity of homologous grafted skin III. The fate of skin homografts transplanted to the brain, to subcutaneous tissue, and to the anterior chamber of the eye. Br J Exp Pathol 1948; 29: 58.
4. Williams KA, Roder D, Esterman A, et al. Factors predictive of corneal graft survival: report from the Australian Corneal Graft Registry. Ophthalmology 1992; 99: 403.
5. Williams KA, Muehlberg SM, Wing SJ, Coster DJ. The Australian Corneal Graft Registry, 1990-1992 report. Aust NZ J Ophthalmol. 1993; 21(suppl): 1.
6. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 475.
7. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomised clinical trials requiring prolonged observation of each patient. I Introduction and design. Br J Cancer 1976; 34: 585.
8. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomised clinical trials requiring prolonged observation of each patient. II Analysis and examples. Br J Cancer 1977; 35: 1.
9. Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics 2000; 56: 645.
10. Cecka JM. The UNOS renal transplant registry. Clin Transpl 2002;1.
11. Parsons DA, Tracy SE, Handa KA, Greig PD. An update of the Canadian organ replacement register (1998). Clin Transpl 1998;97.
12. Speich R, Nicod LP, Aubert JD, et al. Ten years of lung transplantation in Switzerland: results of the Swiss lung transplant registry. Swiss Med Wkly 2004; 134: 18.
13. Adam R, McMaster P, O’Grady JG, et al. Evolution of liver transplantation in Europe: report of the European Liver Transplant Registry. Liver Transpl 2003; 9: 1231.
14. McAlister VC, Badovinac K, Fenton SS, Greig PD. Transplantation in Canada: review of the last decade from the Canadian Organ Replacement Register Clin Transpl 2003;101.
15. Gruessner AC, Sutherland DE. Pancreas transplant outcomes for United States (US) and non-US cases as reported to the United Network for Organ Sharing (UNOS) and the International Pancreas Transplant Registry (IPTR) as of May 2003. Clin Transpl 2003;21.
16. Meier-Kriesche HU, Schold JD, Srinivas TR, Kaplan B. Lack of improvement in renal allograft survival despite a marked decrease in acute rejection rates over the most recent era. Am J Transplant 2004; 4: 378.
17. Meier-Kriesche HU, Schold JD, Kaplan B. Long-term renal allograft survival: have we made significant progress or is it time to rethink our analytic and therapeutic strategies? Am J Transplant 2004; 4: 1289.
18. Völker-Dieben HJ, Claas FHJ, Schreuder GMT, et al. Beneficial effect of HLA-DR matching on the survival of corneal grafts. Transplantation 2000; 70: 640.
19. Coster DJ, Williams KA. Management of high-risk corneal grafts. Eye 2003; 17: 996.
20. Vail A, Gore SM, Bradley BA, et al. Corneal graft survival and visual outcome. A multicenter study. Corneal Transplant Follow-up Study Collaborators. Ophthalmology 1994; 101: 120.
21. Barker NH, Henderson TRM, Ross CA, et al. Current Australian practice in the prevention and management of corneal allograft rejection. Clin Experiment Ophthalmol 2000; 28: 357.
22. Claesson M, Armitage WJ, Fagerholm P, Stenevi U. Visual outcome in corneal grafts: a preliminary analysis of the Swedish Corneal Transplant Register. Br J Ophthalmol 2002; 86: 17.
23. Sit M, Weisbrod DJ, Naor J, Slomovic AR. Corneal graft outcome study. Cornea 2001; 20: 129.
24. Randleman JB, Song CD, Palay DA. Indications for and outcomes of penetrating keratoplasty performed by resident surgeons. Am J Ophthalmol 2003; 13: 68.
25. Armitage WJ, Dick AD, Bourne WM. Predicting endothelial cell loss and long-term corneal graft survival. Invest Ophthalmol Vis Sci 2003; 44: 3326.
26. Weisbrod DJ, Sit M, Naor J, Slomovic AR. Outcomes of repeat penetrating keratoplasty and risk factors for graft failure. Cornea 2003; 22: 429.
27. Price FW Jr., Whitson WE, Johns S, Gonzales JS. Risk factors for corneal graft failure. J Refract Sur 1996; 12: 134.
28. Gjertson DW. Look-up survival tables for living-donor renal transplants: OPTN/UNOS data 1995-2002. Clin Transplant 2003;337.
29. Gore SM, Vail A, Bradley BA, et al. HLA-DR matching in corneal transplantation. Systematic review of published evidence. Corneal Transplant Follow-up Study Collaborators. Transplantation 1995; 60: 1033.
30. Baggesen K, Lamm LU, Ehlers N. Significant effect of high-resolution HLA-DRB1 matching in high-risk corneal transplantation. Transplantation 1996; 62: 1273.
31. Munkhbat B, Hagihara M, Sato T, et al. Association between HLA-DPB1 matching and 1-year rejection-free graft survival in high-risk corneal transplantation. Transplantation 1997; 63: 1011.
32. Khaireddin R, Wachtlin J, Hopfenmuller W, Hoffmann F. HLA-A, HLA-B and HLA-DR matching reduces the rate of corneal allograft rejection. Graefes Arch Clin Exp Ophthalmol 2003; 241: 1020
33. Patel SV, Hodge DO, Bourne WM. Corneal endothelium and postoperative outcomes 15 years after penetrating keratoplasty. Am J Ophthalmol 2005; 139: 311.

Human corneal transplantation; Registry; Outcomes

© 2006 Lippincott Williams & Wilkins, Inc.