[Regression analysis of corneal endothelium after nonmechanical penetrating keratoplasty]

The corneal endothelial cell density is a crucial parameter for the pump function and the transparency of grafts after penetrating keratoplasty (PK). The purpose of this study was to assess corneal endothelial cell density with different regression models after nonmechanical penetrating keratoplasty and to check for differences between diagnoses and two different storage methods.

Two-hundred ninety-six eyes (195 keratoconus, 101 Fuchs' dystrophies, 148 each with short-term preserved and organ-cultured donor corneas) of 268 patients were included in this prospective study. Donor and recipient trephination was performed using nonmechanical trephination technique with the excimer laser 193 nm along metal aperture masks from the epithelial side. The time course of the endothelial cell density (specular microscope EM 1100, TOMEY, Erlangen) after PK was assessed. Endothelial cell density was first analyzed in a cross sectional manner at the 3, 6, 12, 18 and 24 months follow-up and, secondly in a longitudinal manner with linear, polynomial and exponential regression models in the sense of minimizing the residuum (distance between observed and predicted endothelial cell count). The mean donor postmortem time was 10.3 +/- 6.8 hours for short-term-preserved and 19.6 +/- 9.5 hours for organ-cultured corneas (p < 0.0001). The storage time was 63 +/- 49 and 19 +/- 7 days (p < 0.0001), respectively.

In a cross section, overall mean endothelial cell density decreased from 2145 +/- 599 cells/mm2 at the 3 months to 1751 +/- 605 cells/mm2 at 2 years follow-up (p > 0.05). Cell density did not differ significantly between different diagnoses or storage methods at any postoperative stage. In a longitudinal section, the linear regression model estimated an annual decrease of 214 cells/mm2. In a polynomial model the decrease expressed by a tangent to the regression line at 24 months was 175 cells/mm2. The exponential regression model yielded a relative decrease of 9.5% annually. The so-called residuum as a measure for the validity of the regression model was maximal in the linear and minimal in the exponential estimate. With keratoconus and short-term preserved donor material the endothelial cell loss was less in the regression analysis.

During the first two years after nonmechanical trephination in PK, a non-significant decrease in endothelial cell density was observed. The exponential regression model seems to be predestinated for analysis of the time course of corneal endothelium in a longitudinal manner. The annual cell loss ranged around 9.5% without significant differences between diagnoses and storage methods.