Abstract
Purpose :
To compare assessments of morphological features of meibomian glands (MGs) on infrared photographic images between clinic staff and reading center (RC) graders, and to evaluate the correlation of RC morphologic gradings with functional correlates of MGs.
Methods :
Patients with moderate to severe dry eye disease were enrolled in the multicenter Dry Eye Assessment and Management (DREAM) Study. At baseline, clinicians performed slit lamp exams to assess tear break-up time, and MG plugging and lid secretions of the central 5 MGs of the lower lid. Meibography of both lids was performed with the Oculus 5M keratograph. Percent dropout on both full lids was assessed by clinicians or technicians using the Meiboscore (from 0% to >75% by 25%), and by RC graders as a continuous value from 0% to 100%. RC graders assessed MG morphologic features of both full lids, and of the lower lid central 5. RC percent dropout included fluffy areas and ghost glands in the dropout area. A morphological score was calculated based on the morphologic features to correspond to the clinician assessment. Statistical analyses used the generalized estimating equation approach to account for the correlation among lids.
Results :
Technicians graded dropout for 873 lids from 0% (156 [18%]) to >75% (49 [6%]). Though percent dropout graded by the RC was highly correlated with clinical grading (p<0.001, Fig), the RC grades were less extreme; clinical scores of 0% and >75% corresponded to RC median scores of 16% and 66%, respectively.
Among 460 lower lids (middle 5 MGs), mean morphological abnormality score by the RC increased modestly with the number of plugged MGs assessed by clinicians (p=0.04, Fig). There was no significant difference in mean abnormality score by clinician’s secretion status (clear, cloudy, paste, obstructed, p=0.38, Fig).
Among 751 lids (all MGs), there was no significant difference in RC mean abnormality score by clinician tear break-up time (p=0.83, Fig).
Conclusions :
RC and clinic staff assessment of percent dropout was highly correlated, although clinic staff were more likely to categorize MG dropout in the extremes of the Meiboscale. A RC composite morphologic abnormality score was not strongly correlated with clinical functional correlates; individual morphologic features may show stronger associations.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.