Intraobserver and interobserver reproducibility in the evaluation of optic disc stereometric parameters by Heidelberg Retina Tomograph

Purpose: To assess intraobserver and interobserver reproducibility of the measurement of stereometric parameters of the optic disc by means of the Heidelberg Retina Tomograph I (HRT). Study design: Observational study, with interobserver variability. Participants: Fifty-five volunteers (healthy subjects and patients with glaucoma). Methods: HRT examination of the optic disc was repeated on 3 consecutive days on 1 eye of each of the 55 subjects. During each session, five single images were randomly acquired by two independent observers. One mean topography image (MTI), based on three single images, was then built at each session. For the intraobserver intraimage evaluation, the two observers traced their own contour line on one randomly chosen MTI. This procedure was repeated three times. For the intraobserver interimage and interobserver intra/interimage evaluations, the two observers traced their own contour line on the MTI of the first session, which was then automatically superimposed on the MTIs of the other two sessions. Main Outcome Measures: Reproducibility of the 12 stereometric parameters was calculated for each observer by means of the intraclass correlation coefficient (ICC). The expected range of variability between two independent evaluations was calculated by the scatter-plots of each test-retest difference versus their mean. The standard deviation of the mean test-retest score differences was used to describe the spread of score differences. Results: The ICC ranged between 0.79 and 0.99 for intraobserver intraimage and between 0.56 and 1 for intraobserver interimage evaluation. The ICC ranged between 0.54 and 0.99 for interobserver intraimage and between 0.65 and 0.97 for the interobserver interimage evaluation. ICC was almost perfect to perfect for planimetric measures (0.81 < ICC ≤ 1), substantial to almost perfect for volumetric and cup measures (0.61 < ICC ≤ 0.99), and moderate to almost perfect for retinal nerve fiber layer related measures (0.41 < ICC < 0.99). The expected variability was low (95% confidence interval, <±9%). Interimage evaluation showed a higher variability than intraimage evaluation in both of interobserver (P = 0.012) and intraobserver evaluation (P = 0.028 and P = 0.031 for the two observers). Conclusions: Measurement of optic disc stereometric parameters by HRT is highly reproducible. However, the use of retinal nerve fiber layer-related parameters should be taken cautiously. The image acquisition-induced variability seems larger than the operator-induced variability