Abstract
Purpose: :
To evaluate the performance of algorithms of the Heidelberg Retinal Tomograph II (HRT II) in diagnosing glaucoma in a population-based survey.
Methods: :
A population-based, cross-sectional study of Malay persons aged 40-80 years residing in Singapore was conducted in 2004-6. Participants underwent standardized ophthalmic assessments. Glaucoma was defined according to the International Society for Geographical and Epidemiological Ophthalmology criteria. Optic nerve and retinal nerve fiber layer were assessed with the HRT II. Various linear discriminating functions (LDFs) published by Mikelberg et al.(LDF1), Burk (LDF2), Bathija et al.(LDF3), and the Moorfields Regression Analysis (MRA1 with "borderline" outcomes as positive; MRA2 with "borderline" outcomes as negative) were assessed. Area under the receiver operating characteristic curve (AROC), sensitivity, and specificity were used to evaluate diagnostic performance.
Results: :
Of 3280 persons participating in the survey, 3056 underwent HRT examination. Of these, 2861 persons (93.6%), including 121 cases of glaucoma, had acceptable images for analysis. Eyes with visual impairment, astigmatism, more myopic status and in the elderly were more likely to yield low quality image. The highest AROCs were 0.73, 0.70, and 0.71 for LDF1, LDF2, and LDF3; The AROC was 0.72 for MRA1 and 0.66 for MRA2. At 85% specificity, the sensitivities were 56.9, 51.7, and 51.7% for LDF1, LDF2, and LDF3 respectively. These figures decreased to 32.8, 31.0, and 30.2% when the specificity fixed at 95%. The sensitivity and specificity were 62.8 and 81.6% for MRA1, and 38.0 and 93.4% for MRA2. For all algorithms, smaller disc tended to be classified with higher specificity but lower sensitivity, when compared with larger optic disc.
Conclusions: :
The current LDFs and MRA with the use of HRT II are of limited value for population glaucoma screening.
Keywords: clinical (human) or epidemiologic studies: treatment/prevention assessment/controlled clinical trials