Jack Phu1, Michael Kalloniatis2. 1. Centre for Eye Health and the School of Optometry and Vision Science, University of New South Wales, Kensington, New South Wales, Australia. Electronic address: jphu@cfeh.com.au. 2. Centre for Eye Health and the School of Optometry and Vision Science, University of New South Wales, Kensington, New South Wales, Australia.
Abstract
PURPOSE: We sought to determine the impact of seeding point errors (SPEs) as a source of low test reliability in perimetry and to develop a strategy to mitigate this error early in the test. DESIGN: Cross-sectional study. METHODS: Visual field test results from 1 eye of 364 patients (77 normal eyes, 178 glaucoma suspect eyes, and 109 glaucoma eyes) were used to develop models for identifying SPE. Two test cohorts (326 undertaking Swedish interactive thresholding algorithm [SITA]-Faster and 327 glaucoma eyes undertaking SITA-Standard) were used to prospectively evaluate the models for identifying SPEs. Global visual field metrics were compared among reliable and unreliable results. Regression models were used to identify factors distinguishing SPEs from non-SPEs. Models were evaluated using receiver operating characteristic (ROC) curves. RESULTS: In the test cohorts, SITA-Faster produced a higher rate of unreliable visual field results (30%-49.7%) compared with SITA-Standard (10.8%-16.6%). SPEs contributed to most of the unreliable results in SITA-Faster (57.5%-64.9%) compared with gaze tracker deviations accounting for most of the unreliable results in SITA-Standard (40%-77.8%). In SITA-Faster, results with SPEs had worse global indices and more clusters of sensitivity reduction than reliable results. Our best model (using 9 test locations) can identify SPEs with an area under the ROC curve of 0.89. CONCLUSION: SPEs contribute to a large proportion of unreliable visual field test results, particularly when using SITA-Faster. We propose a useful model for identifying SPEs early in the test that can then guide retesting using both SITA algorithms. We provide a simplified framework for the perimetrist to improve the overall fidelity of the test result.
PURPOSE: We sought to determine the impact of seeding point errors (SPEs) as a source of low test reliability in perimetry and to develop a strategy to mitigate this error early in the test. DESIGN: Cross-sectional study. METHODS: Visual field test results from 1 eye of 364 patients (77 normal eyes, 178 glaucoma suspect eyes, and 109 glaucoma eyes) were used to develop models for identifying SPE. Two test cohorts (326 undertaking Swedish interactive thresholding algorithm [SITA]-Faster and 327 glaucoma eyes undertaking SITA-Standard) were used to prospectively evaluate the models for identifying SPEs. Global visual field metrics were compared among reliable and unreliable results. Regression models were used to identify factors distinguishing SPEs from non-SPEs. Models were evaluated using receiver operating characteristic (ROC) curves. RESULTS: In the test cohorts, SITA-Faster produced a higher rate of unreliable visual field results (30%-49.7%) compared with SITA-Standard (10.8%-16.6%). SPEs contributed to most of the unreliable results in SITA-Faster (57.5%-64.9%) compared with gaze tracker deviations accounting for most of the unreliable results in SITA-Standard (40%-77.8%). In SITA-Faster, results with SPEs had worse global indices and more clusters of sensitivity reduction than reliable results. Our best model (using 9 test locations) can identify SPEs with an area under the ROC curve of 0.89. CONCLUSION: SPEs contribute to a large proportion of unreliable visual field test results, particularly when using SITA-Faster. We propose a useful model for identifying SPEs early in the test that can then guide retesting using both SITA algorithms. We provide a simplified framework for the perimetrist to improve the overall fidelity of the test result.