Comparison of corneal endothelial cell density estimated with 2 noncontact specular microscopes

To assess the repeatability of endothelial cell density (ECD) measurements by the EM-935 endothelial microscope and their agreement with those of the Konan-Noncon Robo SP 6000 (Noncon Robo) specular microscope.

We assessed the agreement between automated and semiautomated methods of analysis of the EM-935 and the Noncon Robo by measuring ECD in 40 eyes of 20 healthy subjects and calculating the 95% limits of agreement (LoA) and plotting Bland-Altman graphs. We then evaluated the repeatability of both the methods of the EM-935 by calculating the intraclass correlation coefficients (ICC) and coefficient of variation (CV).

The mean ECD of the 40 eyes was 2531+/-244 cells/mm2 for Noncon Robo and 2483+/-159 cells/mm2 for EM-935-automated. The difference between these 2 methods was not statistically significant. The mean ECD for EM-935-semiautomated was 2635+/-190 cells/mm2 and was statistically significantly higher than the 2 other methods (p<0.001). The 95% LoA were -435 to 339 cells/mm2 for the Noncon Robo and EM-935-automated, -230 to 438 cells/mm2 for the Noncon Robo and EM-935-semiautomated, and -347 to 43 cells/mm2 for the EM-935-automated and EM-935-semiautomated. Repeatability was better for EM-935-semiautomated method compared with automated method as expressed by ICC (95% CI) of 0.80 (0.52-0.95) vs 0.50 (0.09-0.84) and coefficient of variation of 2.43% vs 2.85%, respectively.

The EM-935 specular microscope showed better repeatability for the semiautomated mode compared with the automated mode. Although measurement agreement with the Noncon Robo was somewhat better for the semiautomated mode, agreement was only moderate for both methods. This leads us to recommend that these instruments should not be used interchangeably.