To evaluate the reproducibility of in vivo confocal microscopy for quantitative corneal nerve analysis in different corneal locations.
Corneal confocal microscopy was performed on 10 healthy participants, and the corneal nerve fiber length, corneal nerve fiber density, corneal nerve branch density, and tortuosity coefficient were measured at 5 predetermined locations for only the right eye. Bland–Altman plots, intraclass correlation coefficient (ICC), and coefficient of variation of all 4 corneal nerve measurements were compared between 2 visits and between readers to assess reproducibility. Two technicians performed a masked analysis of images from both visits.
Ten participants with a mean age of 31.3 ± 2.8 years were imaged at 2 different time points separated by a mean of 4.3 ± 4.3 weeks. The interobserver agreements were better than the intervisit agreements for all the 4 corneal nerve measurements as evaluated using Bland–Altman plots. The intervisit ICC ranged from 0.13 to 0.45, and the interobserver ICC ranged from 0.55 to 0.94. The differences between observers and the differences between sessions were not statistically different among all the 5 locations (P > 0.1) for each corneal nerve measurement.
Single confocal images have poor reliability for any of the 4 corneal nerve measurements, and there is no single location on the cornea that has improved reproducibility. Averaging 5 images, from different locations, improves the reproducibility and is essential for obtaining clinically meaningful data.