Abstract
Purpose :
Early detection of change in meibomian glands (MG) may allow for earlier intervention in MG disease. Subjective and semi-objective grading of MG compare the area of glands visualized to the tarsal plate area. Semi-objective grading of meibography images by tracing the MG area of individual glands may improve sensitivity, however, inter-observer agreement of this method is unknown. This study aimed to evaluate the agreement of traditional subjective and semi-objective MG assessment methods to individual gland tracing between two observers.
Methods :
Randomly selected MG images of the upper eyelid of 30 normal participants (mean±SD age 16.57±5.89 years, range 7-24 years, male=15) taken with commercially available meibographers (Keratograph 5M, Oculus; Antares, CSO) were evaluated by two masked examiners. MG area of the central 12mm of the eyelid was assessed using three approaches: 1) subjective MG grading (degree 0 = no meibomian gland loss (MGL); 1 = <25% MGL; 2 = 26–50% MGL; 3 = 51–75% MGL; 4 = >75% MGL), 2) semi-objective ImageJ tracing of overall MG border area to tarsal plate area (Fig. 1) and 3) ImageJ individual gland tracing compared to total tarsal plate area (Fig. 2). Inter-observer agreement for semi-objective measures was analyzed using paired sample t-test, Intraclass Correlation Coefficient (ICC), and Bland-Altman analysis. Wilcoxon signed-rank and Cohen’s weighted kappa were used to analyze subjective measures.
Results :
There was no significant median difference in subjective grading (p =.257) or in mean difference in semi-objective grading of the % MG border area between observer 1 and observer 2 (mean±SD; 89.57±7.33% to 89.88±6.44%; p =.695). Semi-objective grading of % MG area of individual glands was significantly different between observer 1 and observer 2 (mean±SD; 65.24±6.36% to 55.47±6.38%; p<.001). There was significant agreement between the two observers for subjective grading (κw=0.64, 95% CI: 0.39 to 0.89, p<.001), and % MG border area (ICC=0.82, Bland-Altman LOA: –8.54 to 7.93); however, poor agreement was observed with individual gland tracing (ICC=0.24, Bland-Altman LOA: –2.39 to 21.93).
Conclusions :
Although an individual gland tracing approach to MG assessment could potentially detect early changes in MG morphology, the significant difference in % MG area for individual glands suggests that this method may not produce effective agreement between observers.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.