S L Graham1, S M Drance. 1. Department of Ophthalmology, University of British Columbia, Vancouver, Canada.
Abstract
BACKGROUND: The interpretation of high-pass resolution perimetry (HRP) fields can be difficult. An age-related probability plot was derived from a known data base to determine whether this improved specificity and sensitivity in early glaucoma detection. METHODS: Forty glaucoma patients with minor field loss and 40 normals of equivalent age and sex underwent HRP. All had previous Humphrey field data available for comparison. The detection of previously confirmed field defects by HRP was examined using different parameters. RESULTS: Using the age-corrected 95% confidence levels gave the best results, with sensitivity of 82.5% (33/40) and specificity of 85% (34/40). Using the contour plot yielded lower sensitivity (67.5%), while subjectively "eyeballing" the ring printout gave higher sensitivity (90%), but led to poor specificity (72.5%). CONCLUSION: The use of a probability plot enhances the performance of HRP. A certain proportion of cases may be missed using one criterion alone, but combining the information with other indices can increase the yield.
BACKGROUND: The interpretation of high-pass resolution perimetry (HRP) fields can be difficult. An age-related probability plot was derived from a known data base to determine whether this improved specificity and sensitivity in early glaucoma detection. METHODS: Forty glaucomapatients with minor field loss and 40 normals of equivalent age and sex underwent HRP. All had previous Humphrey field data available for comparison. The detection of previously confirmed field defects by HRP was examined using different parameters. RESULTS: Using the age-corrected 95% confidence levels gave the best results, with sensitivity of 82.5% (33/40) and specificity of 85% (34/40). Using the contour plot yielded lower sensitivity (67.5%), while subjectively "eyeballing" the ring printout gave higher sensitivity (90%), but led to poor specificity (72.5%). CONCLUSION: The use of a probability plot enhances the performance of HRP. A certain proportion of cases may be missed using one criterion alone, but combining the information with other indices can increase the yield.