Abstract
Purpose :
Disc haemorrhages (DH) are an important sign that clinicians use to guide the care of patients. However, DHs are often missed during routine examination. Imaging techniques are becoming part of routine glaucoma care; therefore, we tested the agreement between observers to identify DH in Heidelberg retina tomograph (HRT) images.
Methods :
The United Kingdom Glaucoma Treatment Study (UKGTS) is the first randomized placebo-controlled trial in glaucoma and showed preservation of visual fields with latanoprost in newly diagnosed glaucoma patients. 516 patients were enrolled and, during the 24 months follow-up, patients were imaged with HRT3 (Heidelberg Engineering, Heidelberg, Germany; version 3.0.60, Heyex 1.6.2.0) at each of 11 visits. The images were analysed with the Topographic Change Analysis (TCA) algorithm, which automatically aligns all the images and offers the option to flicker the baseline scan with each of the follow-up reflectance scans. All images pairs were flickered and assessed by the same observer, masked to treatment allocation and outcome status, for changes in reflectivity in the optic nerve rim or peripapillary area. Twenty-five patients with DH and 25 with no DH were selected; the presence/absence of DH was validated from same-day disc photographs. Anonymized flicker pairs were converted into a GIF file and upload into an online survey manager (SurveyMonkey®). Five glaucoma specialists participated in the study. Images were classified as DH+ or DH- and the location of the DH in clock hours was chosen. 25 images were presented twice. Images were presented in a random order. Intra- and inter-observer agreement for identifying DH presence was calculated using kappa statistics.
Results :
The mean inter-observer agreement kappa was 0.64 (range 0.40–92) for DH presence and 98.9% of responses agreed on the location within one clock hour. The mean intra-observer agreement kappa was 0.79 (range 0.28–1.00) for DH presence and 100% of responses agreed on the location within one clock hour (Table 1).
Conclusions :
The agreement between glaucoma specialists for the detection of DH in flickered HRT images ranged from ‘moderate’ to ‘almost perfect’ and was at least ‘substantial’ for the majority of comparisons.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.