Grading and Baseline Characteristics of Meibomian Glands using Meibography images in the Dry Eye Assessment and Management (DREAM) Trial

Abstract

Purpose : To describe various morphological features of meibomian glands (MGs) in moderate to severe dry eye disease in the Dry Eye Assessment and Management (DREAM) study and introduce an additional approach to grading MG atrophy

Methods : Infrared photographic images (Oculus Keratograph 5M) of the upper and lower lid of both eyes in DREAM study participants were captured at baseline. Trained and certified non-physician graders independently evaluated each eyelid image for various MG morphological features. Each eyelid image was viewed in three equal sections using a template and Adobe Photoshop. Discrepant results between two graders assigned to each image set were adjudicated by the reading center director (ophthalmologist). Graders measured drop out areas of MG atrophy in each lid in 2 ways: including and excluding ghost glands and fluffy areas as dropped out. Intergrader and grade-regrade agreement was summarized with intraclass correlation coefficients (ICC).

Results : Meibography images were obtained from 250 DREAM participants. 7% of images could not be graded due to poor lid eversion (5%) or no lid eversion (2%). Morphological features of MGs are defined with representative images in Table 1. Intergrader agreement was good (ICC range= 0.56 to 0.89) except for tortuous, thin, thick and dropout glands. Grade-regrade agreement was good (ICC range= 0.54 to 0.88) except for thick, thin, overlap and dropout glands (Table 2). All morphological features occurred more frequently in the upper lid compared to the lower lid (p<0.001), except for dropout glands, gaps and fluffy areas that were more frequent in the lower lid. Drop out area measurements had better agreement when fluffy areas and ghost cells were included in the dropout area (Intergrader ICC 0.65 vs 0.56 and grade-regrade ICC 0.70 vs 0.39).

Conclusions : Non-physician, trained graders can evaluate MG gland features and drop out areas with good agreement and reproducibility on regrading. The upper lid had more visible MG features on meibography images. Poorly everted lids were present indicating that additional training for imagers is needed.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

Table 1: Definitions and meibography image examples of morphological features of meibomian glands in dry eye disease

 

Table 2. Inter-grader and grade-regrade agreement in evaluating meibomian gland features and drop out areas.