GRADING OF RETINAL NERVE-FIBER LAYER WITH A PHOTOGRAPHIC REFERENCE SET

PURPOSE: We developed a new quantitative approach for the evaluation o f diffuse atrophy of the retinal nerve fiber layer, METHODS: A simulta neous, visually supported grading system was designed, which consisted of a set of 25 reference photographs, numbered from 25 (broad, clearl y striated nerve fiber bundles) to 1 (no nerve fibers visible), We pro spectively evaluated this method by matching 60 retinal nerve fiber la yer photographs of patients with glaucoma or ocular hypertension and n ormal subjects to the reference photographs twice by three observers w ith varying experience levels in evaluating retinal nerve fiber layer. RESULTS: Intraobserver and interobserver reliability, evaluated by th e intraclass correlation coefficient statistics, was excellent (intrac lass correlation coefficient > 0.8). Intraclass correlation coefficien ts within observers were from 0.89 to 0.98 (lower limits 95% confidenc e interval, 0.84 to 0.97), and intraclass correlation coefficients bet ween observers were from 0.81 to 0.91 (lower limits 95% confidence int erval, 0.50 to 0.87). Clinical validity, in which the scores for upper and lower halves of photographs were compared with the mean deviation of the Humphrey 30.2 visual held program by using the Spearman correl ation coefficient, was substantial (r = -.68, P < .01 for upper photo score and lower held mean deviation, r = -.53, P < .01 for lower photo score and upper field mean deviation), CONCLUSIONS: By using a refere nce set of photographs of the retinal nerve fiber layer, we defined a method to derive a quantitative measurement of retinal nerve fiber lay er with good reliability and to extend evaluation of retinal nerve fib er layer photographs to nonspecialists.