Possible Sampling Error in Corneal Specular Microscopy: Can Central Endothelial Data Generated by Specular Microscope...

To estimate the minimum number of cells required to obtain reliable data in a specular microscope, which could possibly represent the real clinical condition of the corneal endothelium.

A cross-sectional study of 122 eyes of 61 individuals submitted to noncontact specular microscope was conducted. Data generated by the manufacturer's software were uploaded to specific statistical software for sampling relative error calculation. When relative error was above 5%, new images were acquired and more cells counted until the desired relative error was reached. Data analyzed in this study for the desired relative error were number of cells marked by the examiner for each eye (marked cells), number of cells used for data analysis (analyzed cells), endothelial area used for analysis, sampling error, and absolute number of images used for each eye.

The average number of marked cells required to obtain a relative sampling error of less than 5% was 425.2 ± 102.2 cells. The average number of analyzed cells used by the specular microscope to generate the data was 247.4 ± 51.6 cells. The average endothelial area of the analyzed cells was 0.43 ± 0.08 mm. The mean sampling error was 3.7% ± 0.6%, and an average of 2.95 ± 0.74 images was needed to obtain a relative sampling error of less than 5%.

We conclude that, theoretically, a minimum of 425.2 cells from 2.95 images must be marked to obtain reliable results, which could possibly represent the real endothelial clinical condition of the whole cornea.