Ramasamy and associates raise the question of the management of missing data and censored data, which is an interesting and relevant question. Whereas different hypotheses can be made to address the outcome of transplantation in patients lost to follow-up, it may be better to analyze the data with no hypothesis and to include patients in the analysis for the time they were effectively followed. Results should be presented at a postoperative time consistent with the average follow-up time. This was the reason why we decided to show 36-month results in our paper. If results at a longer postoperative time have to be addressed, then a mathematical model must be used to predict long-term results.
The cumulative incidence of rejection episodes was calculated using the Kaplan-Meier method. In this method, follow-up of each patient is included in the survival calculation. Briefly, the survival (S) probability (ie, rejection episode–free and irreversible rejection–free graft survival) at time t + 1 is a function of the survival at time t and the proportion of failures between t and t + 1 among patients available for follow-up at time t. The log-rank test permits comparing the 2 groups for the observed survival during the complete follow-up period and not at a single postoperative time. The cumulative incidence (CI) of rejection episodes and irreversible rejection is calculated as follows: CI = 1 − S.
Our database has been refreshed. Three-year follow-up data are currently available for 98 patients from the ALK group and 121 patients from the PK group. The average current follow-up time is 46.8 ± 22.4 months (mean ± standard deviation) in the ALK group and 82.7 ± 52.0 months in the PK group. The time between transplantation and data analysis is currently less than 36 months for 27 ALK patients. The remaining 24 ALK patients lost to follow-up and the 28 PK patients lost to follow-up decided not to come back to the institution and we have no more information for these patients. Three more cases of graft failure have been observed: 2 cases related to late infectious keratitis in the PK group and 1 case related to severe ocular trauma in the ALK group. Statistical analysis of the current data does not modify the results reported in our paper (ie, significantly higher graft survival and lower incidence of rejection episodes in the ALK group compared with the PK group and absence of irreversible rejection in the ALK group). A recent study reported similar results, including absence of irreversible rejection in a series of 234 deep anterior keratoplasties performed in eyes with keratoconus. The average follow-up time was 50 months in this study. In addition, these observed results are consistent with the long-term results predicted from mathematical models.