Observer agreement using the Heidelberg retina tomograph: the Bridlington Eye Assessment Project

This study assessed agreement in optic nerve head parameters between 2 investigators independently contouring Heidelberg retina tomograph II (HRT II) optic disc images of normal elderly subjects.

Contours were drawn by both investigators on 550 randomly selected optic discs of 550 normal elderly patients (mean age, 73.1 y) drawn from a population-based study. All subjects had normal intraocular pressures and visual fields. Coefficients of variation were calculated for interobserver and intraobserver variability.

The mean disc area was 1.97 mm, with no relationship between magnitude of disagreement and the mean. There was a significant bias between investigators with a mean difference in disc area of -0.06 mm. The 95% limits of agreement of disc area were -0.59 to +0.47 mm. Ninety-five percent coefficients of agreement (1.96 SD of the parameter differences between investigators/mean parameter value x100) were similar for global and sectoral disc areas (26.7% to 32.3%). Respective coefficients for rim area, cup area, and rim volume were much higher at 35.2%, 48.5%, and 76.5% respectively, but much lower for rim/disc area ratio (12.7%). There was no significant correlation between disagreement and grade of cataract or HRT image quality.

Disagreement in contour placement can be substantial, with disproportionate effects on some calculated HRT parameters. Image quality and lens opacity have no significant influence on the degree of disagreement. Rim/disc area ratio is the HRT parameter least affected and may be more useful in comparing normative databases and diagnostic studies between centers.