Posters

Diagnostic performance of Heidelberg Retina Tomograph (HRT 3) algorithms in a population-based setting: the Thessaloniki Eye Study (TES)

Poster Details

First Author: A.Salonikiou GREECE

Co Author(s):    E. Anastasopoulos   F. Yu   T. Pappas   P. Founti   N. Dervenis   F. Topouzis     

Abstract Details

Purpose:

To assess the diagnostic performance in glaucoma (sensitivity, specificity, accuracy) of Moorfields Regression Analysis (MRA) and Glaucoma Probability Score (GPS) incorporated in the HRT 3 in a population-based setting. To date, only two population-based studies have published MRA data in Caucasians while only one of these two has published both MRA and GPS data.

Setting:

The TES is a cross-sectional, population-based, epidemiologic study of chronic eye diseases in a Greek population over 60 years of age done in Thessaloniki, a major urban centre in Greece.

Methods:

79 glaucomatous, 1002 non-glaucomatous and 270 glaucoma suspect cases were included in this analysis. All participants had a thorough ophthalmic examination including HRT imaging. Glaucoma was defined on the basis of both optic disc and visual field damage. Glaucoma suspects were defined according to specific criteria when the criteria for glaucoma were not fulfilled. Sensitivity, specificity and accuracy were calculated for MRA and GPS. Analyses for MRA were run grouping borderline (BL) results 1)with within normal limits (WNL), 2)with outside normal limits (ONL) and 3)excluding them. Analyses were run twice; with and without including glaucoma suspects in the normal population.

Results:

For MRA, when glaucoma suspects were included, sensitivity and specificity was 70.89% and 82% (BL grouped with WNL), 88.61% and 64.23% (BL grouped with ONL), 86.15% and 78.11% (BL excluded) respectively. Highest accuracy was 81.3% when BL grouped with WNL. Glaucoma suspects excluded, sensitivity and specificity was 70.89 % and 84.03%, 88.61% and 66.67% and 86.15% and 80.68% respectively for the three groups. Highest accuracy was 83.1% (BL grouped with WNL). For GPS, glaucoma suspects included, for a specificity of 81.9% sensitivity was 73.1%. Glaucoma suspects excluded, for a specificity of 83.6%, sensitivity was 73.1%.

Conclusions:

The two diagnostic algorithms of the HRT 3 showed similar specificity and sensitivity for glaucoma diagnosis in our population. MRA specificity was higher when BL results were grouped with WNL results. Specificity dropped when BL were excluded, and dropped even more when BL were grouped with ONL. Highest MRA accuracy was achieved when glaucoma suspects were excluded and BL were grouped with WNL. Specificity of both MRA and GPS dropped by about 2% when glaucoma suspects were included in the analysis. Co-funded by the EU (European Social Fund) and Greek national funds (Act 'Aristia', Operational Program 'Education and Lifelong Learning').

Financial Disclosure:

NONE