Purpose: Rolando's classification system for tear mucus ferning patterns subjectively assigns grades based on the size and spacing between ferns. Grade 1 and 2 patterns are considered "normal," whereas grade 3 and 4 patterns are often associated with keratoconjunctivitis sicca. This study was designed to examine the intraobserver and interobserver repeatability of Rolando's system.
Methods: Photographic slides (N = 418) of portions of tear ferning patterns were randomly assembled, numbered, and graded by two investigators (I1 and I2). I1 graded the slides twice; the slides were remixed under masked conditions between the first and second runs. I2 graded the slides once, independent of I1.
Results: For the four-grade system, intraobserver agreement was 85.41% (simple kappa = 0.75; 95% confidence interval ]CI] = 0.69-0.82). Interobserver agreement was 80.62% (kappa = 0.67; CI = 0.60-0.74) and 86.12% (kappa = 0.75; CI = 0.68-0.83) for the first and second runs of I1, respectively. Analysis of the ability to simply classify normal (grade 1 and 2) from abnormal (grade 3 and 4) patterns revealed intraobserver repeatability of 94.50% (kappa = 0.76; CI = 0.67-0.86). Interobserver agreement was 92.10% (kappa = 0.65; CI = 0.56-0.75) and 94.26% (kappa = 0.71; CI = 0.62-0.81) for the first and second runs, respectively.
Conclusion: Based on the high rate of agreement between intraobserver and interobserver trials, Rolando's grading system appears to be an easy and consistent method for the classification of tear ferning patterns.