The utility of tear cytokine profiles in the diagnosis of Graft vs Host Disease- Associated Dry Eye

Abstract

Purpose: To analyze a panel of tear cytokines in Graft versus Host Disease (GvHD)-associated Dry Eye (DE) patients, to correlate them with clinical features, and to develop a tear molecule level-based predictive model

Methods: 22 GvHD-DE patients and 21 healthy subjects comparable in age and gender were evaluated under standard conditions of relative humidity (45%) and temperature (230C) within The Controlled Environmental Research Laboratory. Patients were medically controlled and had discontinued topical medications for one week before examination. DE-related symptoms and signs were recorded, and unstimulated tears were collected from the most symptomatic eye. Levels of 15 molecules (EGF, IL-1Ra, IL-1β, IL-2, IL-6, IL-8/CXCL8, IL-10, IL-12p70, IL-17A, IP-10/CXCL10, IFN-γ, VEGF, TNF-α, RANTES) were measured in tear samples by multiplex bead assay in a Luminex IS-100, and correlated with clinical parameters. Multivariate logistic regression was used to build a model for predicting GvHD-DE event on the basis of an optimal subset of molecules selected by Akaike Information Criterion (AIC) and adjusted for age and gender. Prediction accuracy was estimated by leave-one-out cross validation (LOOCV).

Results: EGF and IP-10/CXCL10 tear levels were significantly (p<0.0001) decreased, whereas IL-1RA, IL-8/CXCL8 and IL-10 levels were significantly (p=0.0007, p= 0.0003, and p=0.0253) increased in GvHD-DE patients compared to healthy subjects. IL-1RA and IL-8/CXCL8 tear levels correlated positively with OSDI questionnaire score, conjunctival hyperemia, fluorescein corneal staining and lissamine green conjunctival staining, while that correlation was negative with T-BUT and Schirmer test. EGF and IP-10/CXCL10 levels correlated positively with T-BUT and Schirmer test values and negatively with OSDI, hyperemia, fluorescein and lissamine green staining scores. A predictive model based on IL-8/CXCL8 and IP-10/CXCL10 tear levels was developed, with an area under the ROC curve of 0.9004, a sensitivity of 86.36%, and a specificity of 95.24%.

Conclusions: Correlations were demonstrated between the levels of inflammatory proteins and the clinical phenotype of the GvHD associated dry eye. A GvHD-DE diagnostic model based on IL-8/CXCL8 and IP-10/CXCL10 tear levels showed a good sensitivity and specificity. These results add further knowledge to the molecular mechanisms involved in GvHD-DE and can be helpful in its diagnosis.