HRT demonstrates high reproducibility in an inter-surgeon comparison

Rim- and cup-related parameters showed greater reproducibility than RNFL-related parameters.

ByMichael Piechocki

MILAN — A retinal imaging device showed a high degree of reproducibility in a study performed here.

According to Stefano Miglior, MD, first author of the study, in clinical practice the Heidelberg Retinal Tomograph (Heidelberg Engineering) is used both for identifying early glaucoma patients and for obtaining objective disc measurements that are useful for follow-up purposes.

To obtain cross-sectional information, the HRT uses a dedicated software program that considers a series of parameters, including overall disc area, rim area and cup shape. Dr, Miglior said these have shown a high degree of reproducibility in their study.

Dr. Miglior said their results showed rim- and cup-related parameters, with the exception of cup volume, all had higher reproducibility than RNFL-related parameters.

“For serial comparison of longitudinal studies I would suggest the analysis of rim and cup areas more than cup and rim volumes. I would discourage the clinician in using RNFL-related parameters, given the higher variability and the unclear clinical significance,” Dr. Miglior said.

(Editor’s note: Officials at Heidelberg Engineering point out that this research was performed on the HRT, which is no longer commercially available. The company now manufactures the HRT II.)

‘Almost perfect’ results

The researchers looked at 12 parameters in their reproducibility analysis, including disc area, cup area, cup-to-disc area, rim area, height variation contour, cup volume, rim volume, mean cup depth, maximum cup depth, cup shape measure, mean retinal nerve fiber layer (RNFL) thickness and RNFL cross-sectional area.

According to the study, the intra-observer intra-image reproducibility analysis resulted in a mean interclass correlation coefficient (ICC) score of “substantial” to “almost perfect” for each parameter evaluated. Intra-observer inter-image reproducibility analysis resulted in a mean ICC score ranging from “moderate” to “perfect” for each parameter.

Inter-observer intra-image reproducibility analysis resulted in a mean ICC score from “moderate” to “almost perfect” and inter-observer inter-image reproducibility analysis resulted in a mean ICC score from “substantial” to “almost perfect” for each parameter.

According to the report, all mean expected measurements were within ±10%, with most being within ±5%.

Standard deviations of the differences between the intra-observer inter-image evaluations were not statistically significant, Dr. Miglior said.

Reproducibility study

Creating a mean image

Taking a series of images to create a mean image using the Heidelberg Retinal Tomograph induces more variability than retracing the contour line of the optic disc, a study indicates.

The study results show that rim- and cup-related parameters had higher reproducibility than RNFL-related parameters. The first author of the report said the difference was small, but it is difficult to explain why the difference was seen.

The first author suggests that analysis of rim and cup areas be used in preference to cup and rim volumes and that RNFL-related parameters have a higher degree of variability.

Dr. Miglior and colleagues with the Universita’ degli Studi di Milano Biococca here conducted the study to assess the reproducibility of HRT optic disc measurements and to estimate the HRT measurements that can be expected in eyes believed not to have progressive optic disc changes.

The study evaluated the observer-related and image-related sources of variability using the HRT when two independent observers analyzed both the same image and different images (intra-image reproducibility and inter-image reproducibility). They also evaluated the reproducibility between the two observers (inter-observer reproducibility).

The study included 55 patients between 25 and 79 years of age. To be included, patients had to have 20/20 visual acuity, a clear lens and a normal retina.

Patients were excluded if they had myopia greater than –6 D, optic disc abnormalities and a history of neurophthalmic disease.

Five images were taken each day for 3 consecutive days by the two independent observers, yielding a total of 165 mean topographic images. The mean standard deviation of these was 27.87 ±10.56. No statistically significant difference was seen between the observers for any parameters evaluated.

Eleven eyes were affected by primary open-angle glaucoma at different stages of disease. A wide range of disc, rim and cup sizes and a wide range of morphologic changes were also seen.

A mean topographic image (MTI) was computed each day from three of the five images taken. For intra-image comparison, each of the observers manually drew a contour line on one of the three MTIs chosen at random. For the inter-image comparison, each observer drew a contour line on the same MTI. The two contour lines were then superimposed using HRT software onto the main two MTIs of each series.

Software variable

According to the report, tracing of the contour line by the observers was less influential as a source of variability than mean image construction in terms of intra-image comparison.

The authors note there was no variability in disc area in terms of intra-observer evaluations between images because the first contour line was automatically superimposed upon the subsequent images and thus, theoretically, did not affect the analysis.

According to the report, the disc area and stereometric parameter findings agree with other similar studies.

In the case of longitudinal studies, HRT provides software that has been shown to be able to detect progressive topographic disc changes. Dr. Miglior said the software, however, does not take into account any specific HRT stereometric parameter, but compares the position of each pixel along the Z axis of the reconstructed images.

“Surgeons, clinicians or technicians need to be extremely careful in taking the images, possibly after dilating the pupils. When possible, I would suggest taking a series of images to obtain a well-centered and sharp mean topography image based on at least three single images,” he said.

For Your Information:

Stefano Miglior, MD, can be reached at Clinica Oculistica, Ospedale S. Paolo, Via di Rudini 8, 20142 Milan, Italy; +(39) 335-6256-762. Dr. Miglior has no direct financial interest in the products mentioned in this article, nor is he a paid consultant for any companies mentioned.

Heidelberg Engineering, makers of the Heidelberg Retinal Tomograph, can be reached at 1499 Poinsettia Ave., Suite 160, Vista, CA 92037 U.S.A.; +(1) 760-598-3770; fax; +(1) 760-598-3060.

Reference:

Miglior S, Albé E, et. al. Intraobserver and interobserver reproducibility in the evaluation of optic disc stereometric parameters by Heidelberg Retinal Tomograph. Ophthalmol. 2002;109:1072-1077.