Jack Phu1, Sieu K Khuu2, Ashish Agar3, Ireni Domadious2, Anika Ng2, Michael Kalloniatis4. 1. Centre for Eye Health, University of New South Wales, Kensington, Australia; School of Optometry and Vision Science, University of New South Wales, Kensington, Australia. Electronic address: jack.phu@unsw.edu.au. 2. School of Optometry and Vision Science, University of New South Wales, Kensington, Australia. 3. Centre for Eye Health, University of New South Wales, Kensington, Australia; Department of Ophthalmology, Prince of Wales Hospital, Randwick, Australia. 4. Centre for Eye Health, University of New South Wales, Kensington, Australia; School of Optometry and Vision Science, University of New South Wales, Kensington, Australia.
Abstract
PURPOSE: To use factor analysis to visualize and assess the reproducibility and consistency of clinical quantitative parameters that can optimally distinguish among healthy, glaucoma suspect, and manifest glaucoma patients at a cross-sectional level and thus to describe the transition of quantitative change among the diagnostic categories. DESIGN: Retrospective cross-sectional study. PARTICIPANTS: The medical records of healthy, glaucoma suspect, and manifest glaucoma patients (diagnosed by expert clinicians) seen at the Centre for Eye Health in 2015 (n = 148, n = 664, and n = 129, respectively) and 2018 (n = 242, n = 464, and n = 126, respectively) were reviewed. One eye was selected for the study. METHODS: Quantitative clinical measures (intraocular pressure [IOP], central corneal thickness [CCT], visual field [VF], and OCT) were extracted and binary logistic (backward stepwise) regression was performed to identify factors that dictated separation between diagnostic pairs. These were used systematically as inputs for factor analysis to determine a final model that could potentially predict a clinical diagnosis. MAIN OUTCOME MEASURES: Intraocular pressure, CCT, VF (mean deviation and pattern standard deviation) indices, and OCT optic nerve head parameters and thickness values (retinal nerve fiber layer [RNFL] and ganglion cell-inner plexiform layer). RESULTS: Few clinical parameters were identified commonly as significant across all diagnostic pairings for 2015 (3 of 23: IOP, pattern standard deviation, and 7-o'clock RNFL thickness) and 2018 (1 of 23: vertical cup-to-disc ratio). Few parameters overlapped when comparing 2015 and 2018 results, highlighting inconsistencies in the models between years. Factor analysis showed good separation between healthy persons and glaucoma patients. Using biplots to visualize the data in 2-dimensional clusters, glaucoma suspect patients demonstrated substantial overlap with healthy and glaucoma cohorts. The contributions of each parameter to diagnostic separation changed between groups and years. CONCLUSIONS: Despite advances in quantitative ocular imaging and perimetry, the transition among healthy, glaucoma suspect, and manifest glaucoma patients remains confounded by a lack of consistent, reproducible combinations of quantitative clinical criteria. These results highlight the nebulousness (at patient-, instrument-, and clinician-related levels) of glaucoma diagnosis that remains contingent on individual clinical expertise and assessment.
PURPOSE: To use factor analysis to visualize and assess the reproducibility and consistency of clinical quantitative parameters that can optimally distinguish among healthy, glaucoma suspect, and manifest glaucomapatients at a cross-sectional level and thus to describe the transition of quantitative change among the diagnostic categories. DESIGN: Retrospective cross-sectional study. PARTICIPANTS: The medical records of healthy, glaucoma suspect, and manifest glaucomapatients (diagnosed by expert clinicians) seen at the Centre for Eye Health in 2015 (n = 148, n = 664, and n = 129, respectively) and 2018 (n = 242, n = 464, and n = 126, respectively) were reviewed. One eye was selected for the study. METHODS: Quantitative clinical measures (intraocular pressure [IOP], central corneal thickness [CCT], visual field [VF], and OCT) were extracted and binary logistic (backward stepwise) regression was performed to identify factors that dictated separation between diagnostic pairs. These were used systematically as inputs for factor analysis to determine a final model that could potentially predict a clinical diagnosis. MAIN OUTCOME MEASURES: Intraocular pressure, CCT, VF (mean deviation and pattern standard deviation) indices, and OCT optic nerve head parameters and thickness values (retinal nerve fiber layer [RNFL] and ganglion cell-inner plexiform layer). RESULTS: Few clinical parameters were identified commonly as significant across all diagnostic pairings for 2015 (3 of 23: IOP, pattern standard deviation, and 7-o'clock RNFL thickness) and 2018 (1 of 23: vertical cup-to-disc ratio). Few parameters overlapped when comparing 2015 and 2018 results, highlighting inconsistencies in the models between years. Factor analysis showed good separation between healthy persons and glaucomapatients. Using biplots to visualize the data in 2-dimensional clusters, glaucoma suspect patients demonstrated substantial overlap with healthy and glaucoma cohorts. The contributions of each parameter to diagnostic separation changed between groups and years. CONCLUSIONS: Despite advances in quantitative ocular imaging and perimetry, the transition among healthy, glaucoma suspect, and manifest glaucomapatients remains confounded by a lack of consistent, reproducible combinations of quantitative clinical criteria. These results highlight the nebulousness (at patient-, instrument-, and clinician-related levels) of glaucoma diagnosis that remains contingent on individual clinical expertise and assessment.
Authors: Janelle Tong; Jack Phu; David Alonso-Caneiro; Sieu K Khuu; Michael Kalloniatis Journal: Transl Vis Sci Technol Date: 2022-04-01 Impact factor: 3.048