Diagnostic performance of qualitative and quantitative methods of meibomian gland dropout evaluation in dry eye disease: An investigator-masked, randomised crossover study.

Abstract

Purpose: To evaluate the discriminative abilities and optimal cut-off values of qualitative meiboscale grading and percentage meibomian gland dropout measurements from each of two instruments (Keratograph 5M and LipiView II) in detecting dry eye disease, as defined by the TFOS DEWS II criteria.

Methods: A total of 227 community residents (143 females, 84 males; mean ± SD age, 36 ± 15 years) were recruited in a prospective, investigator-masked, randomised, crossover study. Dry eye symptomology, tear film parameters, and ocular surface characteristics were evaluated in a single clinical session. Qualitative meiboscale grading and quantitative percentage gland dropout assessment using Image J software were evaluated by independent masked assessors, in a randomised order.

Results: Overall, 92 (41 %) participants fulfilled the TFOS DEWS II criteria for dry eye disease. The diagnostic performances for percentage meibomian gland dropout measurements (C-statistic range, 0.629 to 0.647) were significantly greater than qualitative meiboscale grading (C-statistic range, 0.547 to 0.560) for both instruments. The Youden-optimal diagnostic cut-off percentage gland dropout was > 20 % and optimal threshold meiboscale grade was > 1 for both superior and inferior eyelid measurements from the two instruments. Inter-instrument weighted Cohen's kappa coefficients for meiboscale grading were 0.734 and 0.682, respectively, and Bland-Altman biases (95 % limits of agreement) for percentage gland dropout were -1.8 % (-28.1 % to 24.4 %) and -1.5 % (-29.3 % to 26.2 %), respectively.

Conclusions: Despite exhibiting comparable discriminative abilities, meibography parameters obtained from the Keratograph and LipiView were not directly interchangeable. Overall, percentage gland dropout measurements demonstrated superior diagnostic performance, which would support their use in clinical trials and epidemiology studies, where practicable.