Abstract
Purpose :
Meibomian gland loss is a significant predictor for dry eye disease. Non-contact infrared meibography is a popular imaging technique integrated into multi-functional corneal topographers, and can be used clinically to grade the degree of meibomian gland loss in patients. In this study, we examined the inter-observer reliability of the CSO Antares, a multi-functional corneal topographer and infrared meibographer with integrated analysis software, in quantifying meibomian gland loss in upper and lower eyelids.
Methods :
Ninety-six images were taken of the meibomian glands for the upper and lower eyelids in both eyes for 24 subjects. Images were analyzed by two independent graders using the directions provided with the Antares integrated analysis software. Interrater reliability was compared for the calculated percent area loss using Bland-Altman limits of agreement, as well as categorically by a displayed 5-point Likert clinical grading scale (grade 0 as <10% loss; grade 1 as <25% loss; grade 2 as <50% loss; grade 3 as <75% loss; and grade 4 as >=75% loss) using a weighted Cohen’s kappa statistic.
Results :
Subjects ranged in age from 8-24 years old, with a mean age of 14.8 years old. When comparing the calculated percent area meibomian gland loss between graders, a mean difference of 0.856% (SD=4.117%) was observed for upper eyelids, and a mean difference of 21.41% (SD=12.56%) was observed for lower eyelids. Categorical meibomian gland loss scores using a 5-point Likert clinical grading scale displayed strong agreement between the 2 observers for grading meibomian gland loss for the upper eyelids (κ = 0.722, p < .0005), and fair agreement for grading meibomian gland loss for lower eyelids (κ = 0.557, p < .0005).
Conclusions :
Inter-observer grading repeatability for meibomian gland loss with the integrated analysis software of CSO Antares topographer displayed stronger limits of agreement between observers for both the continuous percent area loss metric and the 5-point Likert clinical grading scale when examining upper eyelids. For lower eyelids, more inter-observer variability was observed in the analysis of meibomian gland loss for both metrics. These findings suggest that the grading approach with the integrated analysis software of the Antares is better suited for consistent evaluation of meibomian gland loss for upper eyelids between eye care practitioners.
This is a 2021 ARVO Annual Meeting abstract.