Our aim was to extend the concept of blink patterns from average interblink interval (IBI) to other aspects of the distribution of IBI. We hypothesized that this more comprehensive approach would better discriminate between normal and dry eye subjects.
Methods: Blinks were captured over 10 minutes for ten normal and ten dry eye subjects while viewing a standardized televised documentary. Fifty-five blinks were analyzed for each of the 20 subjects. Means, standard deviations, and autocorrelation coefficients were calculated utilizing a single random effects model fit to all data points and a diagnostic model was subsequently fit to predict probability of a subject having dry eye based on these parameters.
Mean IBI was 5.97 seconds for normal versus 2.56 seconds for dry eye subjects (ratio: 2.33, P = 0.004). IBI variability was 1.56 times higher in normal subjects (P , 0.001), and the autocorrelation was 1.79 times higher in normal subjects (P = 0.044). With regard to the diagnostic power of these measures, mean IBI was the best dry eye versus normal classifier using receiver operating characteristics (0.85 area under curve (AUC)), followed by the standard deviation (0.75 AUC), and lastly, the autocorrelation (0.63 AUC). All three predictors combined had an AUC of 0.89. Based on this analysis, cutoffs of #3.05 seconds for median IBI, and #0.73 for the coefficient of variation were chosen to classify dry eye subjects.
(1) IBI was significantly shorter for dry eye patients performing a visual task compared to normals; (2) there was a greater variability of interblink intervals in normal subjects; and (3) these parameters were useful as diagnostic predictors of dry eye disease. The results of this pilot study merit investigation of IBI parameters on a larger scale study in subjects with dry eye and other ocular surface disorders.