Predicting contact lens discomfort from meibomian gland morphology and function

To develop and validate a model to predict contact lens discomfort (CLD) using features of the meibomian glands (MG). This study consisted of three parts: (i) 50 meibography images from contact lens wearers and non-wearers were assessed to distinguish features of interest between the groups to develop an automated algorithm which was tested on 154 images from CLD participants (ii) A multivariable model was constructed with a separate population of 84 lens wearers using imaging features and ocular surface signs to predict Contact Lens Dry Eye Questionnaire-8 (CLDEQ-8) score (iii) The model was validated on two independent sets of images using Bland-Altman comparisons between actual and predicted CLDEQ-8 scores. Part I: Intergland distance and width of the gland at the lid margin, centre and distal end were identified as features distinguishing between lens wearers and non-wearers. Part II: In univariate analysis, CLDEQ-8 scores were negatively associated with gland width at the lid margin (r = -0.52, p = 0.01) and lipid layer thickness (r = -0.40, p < 0.05) and positively associated with tear break-up time (r = 0.51, p < 0.001), MG expressibility (r = 0.35, p = 0.01), and intergland distance at the lid margin (r = 0.73, p = 0.02). The multivariable regression model accounted for 68 % (R2 = 0.68, p < 0.05) of the variance in CLDEQ-8 scores. Part III: The mean difference (bias) between the predicted and actual CLDEQ-8 score was -0.1 with limits of agreement of ± 2.8 units. Morphometric features of MG can predict CLDEQ-8 score in lens wearers with reasonable accuracy. This approach may be a suitable screening tool in early detection and to understand the natural history of CLD.