Interobserver agreement in clinical grading of vitreous haze using alternative grading scales

Abstract

Purpose: To evaluate the reliability of clinical grading of vitreous haze using a new 9-step ordinal scale versus the existing 6-step ordinal scale.

Design: Evaluation of diagnostic test (interobserver agreement study).

Participants: A total of 119 consecutive patients (204 uveitic eyes) presenting for uveitis subspecialty care on the study day at 1 of 3 large uveitis centers.

Methods: Five pairs of uveitis specialists clinically graded vitreous haze in the same eyes, one after the other using the same equipment, using the 6- and 9-step scales.

Main outcome measures: Agreement in vitreous haze grade between each pair of specialists was evaluated by the κ statistic (exact agreement and agreement within 1 or 2 grades).

Results: The scales correlated well (Spearman's ρ = 0.84). Exact agreement was modest using both the 6-step and 9-step scales: average κ = 0.46 (range, 0.28-0.81) and κ = 0.40 (range, 0.15-0.63), respectively. Within 1-grade agreement was slightly more favorable for the scale with fewer steps, but values were excellent for both scales: κ = 0.75 (range, 0.66-0.96) and κ = 0.62 (range, 0.38-0.87), respectively. Within 2-grade agreement for the 9-step scale also was excellent (κ = 0.85; range, 0.79-0.92). Two-fold more cases were potentially clinical trial eligible on the basis of the 9-step than the 6-step scale (P<0.001).

Conclusions: Both scales are sufficiently reproducible using clinical grading for clinical and research use with the appropriate threshold (≥ 2- and ≥ 3-step differences for the 6- and 9-step scales, respectively). The results suggest that more eyes are likely to meet eligibility criteria for trials using the 9-step scale. The 9-step scale appears to have higher reproducibility with Reading Center grading than clinical grading, suggesting that Reading Center grading may be preferable for clinical trials.