Assessment of ocular redness measurements obtained with keratograph 5M and correlation with subjective grading scales

To examine correlations between ocular redness scores provided by the Keratograph 5M and those determined using two image-based grading scales.

Observational prospective cross-sectional study. Two hundred and twenty six eyes of two hundred and twenty six participants (175 patients using anti-glaucoma eye drops and 51 subjects untreated). All subjects were scored automatically using the keratograph 5M. These redness scores (RS) were then correlated with the gradings provided by the Efron and McMonnies/Chapman-Davies scale (MC-D) scales (two observers).

Excellent reproducibility was observed for both the Efron (weighted K=0.897, 95% CI 0.823-0.904) and MC-D (weighted K=0.783, 95% CI 0.752-0.795) scales. Keratograph RS and the scores obtained with both Efron (Spearman's Rho=0.43, P<0.001) and MC-D (Spearman's Rho=0.48, P<0.001) scales were significantly correlated. RS for the bulbar and limbal - nasal and temporal quadrants also correlated moderately with the two subjective scales. Through Bland Altman analysis, poor agreement was detected between the objective and subjective methods: agreement values for the Efron scale or MC-D scale (matching scorers between observers) versus overall RS showed high biases (-15.58 and -22.05 respectively) and wide limits of agreement (LOA) (-46.169 to 15.005 and -52.534 to 8.19 respectively). Lowest bias was observed between temporal limbal RS and Observer 2 Efron score (-0.04).

Although it emerged as a reliable objective method, the keratograph 5M overestimated the scores compared with the subjective grading scales when used to grade the degree of ocular redness. Therefore, they should not be interchangeable methods.