Comparison of meibomian gland loss area measurements between two computer programs and intra-inter-observer agreement

Background Meibography is a diagnostic test that allows in vivo evaluation of meibomian gland (MG). Nowadays, it is unknown whether the two available computer programs are equivalent to evaluate the glandular loss area. Methods This is a prospective, longitudinal, and observational study. A random selection of meibography photographs from healthy patients from the ocular surface clinic at Destellos de Luz foundation is made. The upper eyelid images were taken with the Antares (R) meibography (CSO, Florence, Italy); they were classified in five sessions with a week of separation between each measurement by an expert observer for each program, Phoenix (MAGL) and ImageJ (LAGB). An analysis of the meibomian gland loss area was performed, calculating it semiautomatically with Phoenix and manually with ImageJ. Intra-observer agreement was assessed through an intra-class correlation coefficient and the mean of standard deviations within subjects. Comparison between the two computational programs MG loss was made trough a nonparametric test. Results Fifty-four images from x patients (n, 67.3% female) were analyzed. The limits of concordance analysis between the two programs showed a range between - 18.55 and 9.14%. The mean MG loss area through ImageJ by observer 1 was 27.91 +/- 14.82% (IC 95% 23.87 to 31.96), and that by observer 2 was 29.05 +/- 15.17% (95% CI 24.91 to 33.19). The mean MG loss area through Phoenix by observer 1 was 24.48 +/- 13.97% (IC 95% 20.67 to 28.29), and that by observer 2 was 24.93 +/- 12.70% (95% CI 21.46, 28.40) Conclusions The comparison of the measurement of meibomian gland loss with both programs showed a statistically significant difference. Intra-observer repeatability and inter-observer repeatability were good, with no clinical or statistical difference.