Abstract
Purpose :
In vivo corneal confocal microscopy is a novel non-invasive imaging technique, in which corneal nerve morphology can be imaged in great detail. Automated software to quantify these corneal nerves is readily available, with acceptable inter-rater and inter-operator reproducibility. Recently, studies have demonstrated decreased corneal nerve density in various neuropathies such as diabetic neuropathy, idiopathic small fiber neuropathy, sarcoidosis and hereditary neuropathies. To date, there are no multi-center studies employing in vivo corneal confocal microscopy. Inter-center reproducibility has not been described in the literature, therefore we aim to assess inter-center variability in corneal nerve parameters measured by corneal confocal microscopy.
Methods :
Images of 54 eyes of 27 participants (19 male, 8 female) were obtained with corneal confocal microscopy (HRT III with Rostock Cornea Module) in two centers by two operators, 4-10 days apart. Both operators received the same training (prof. Malik and colleagues, University of Manchester). Per eye, three high-quality images of the corneal subbasal nerve plexus were selected. Images were analyzed using ACCMetrics V2.0 (University of Manchester) to determine corneal nerve fiber length (CNFL), corneal nerve fiber density (CNFD) and corneal nerve branch density (CNBD). The mean values of both eyes were averaged. Inter-center variability was calculated using Bland-Altman plots and intraclass correlation coefficient (ICC).
Results :
Using Bland Altman plots, bias ± standard deviation was -2.1 ± 3.9 for CNFD, -1.7 ± 2.0 for CNFL and -12.3 ± 11.5 for CNBD. ICC was 0.82 for CNFL, 0.83 for CNFD, 0.64 for CNBD.
Conclusions :
In our study, ICC was very good for CNFL and CNFD, but not for CNBD. Even though both operators had different experience in ophthalmic examinations (one MD/PhD student, one ophthalmic resident) we managed to obtain high intraclass correlation. Therefore we conclude that in vivo corneal confocal microscopy of the corneal subbasal nerve plexus is suitable for multi-center studies when CNFL and CNFD are used as outcome measures.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.