Abstract
The current study aimed to optimize the measurement design for each quantity of interest, determine critical limits for clinically distinguishing glaucoma-suspect eyes from non-suspect eyes in the UGDS using the optimized design, and compare the efficiency of the measured quantities in estimating sensitivity METHODS: Data from non-glaucoma suspect eyes in the UGDS were analyzed for age and sex dependence, sources of variation, frequency distribution, and assessment of critical limits for glaucoma detection. Critical limit was defined as the extreme 95% confidence limit for the 95% one-sided tolerance. Intraocular pressure (IOP) and visual field contract sensitivity (Mean Deviation, MD), linear cup-to-disc ratio (C/D-linear), neuro-retinal rim area (NRA), nerve fiber layer thickness (cpRNFLT-Global and GDx-TSNIT) were examined RESULTS: Analysis revealed no significant age or sex dependence for the measured quantities. Variability among subjects was found to dominate, affecting the precision of measurements. Frequency distributions approximated normal distributions, enabling the estimation of tolerance limits for critical assessment. The critical limits to distinguish pathological from non-pathological were estimated as 22 mmHg, -3.7 dB, 0.84, 0.6 mm(2), 56 µm and 33 µm for IOP, MD, C/D-linear, NRA, cpRNFLT-Global and GDx-TSNIT respectively CONCLUSION: It appears preferable to estimate critical limits for small samples using the extreme confidence limit of the tolerance limit. The critical limits obtained here are consistent with previously reported values for the same device models. C/D-linear and NRA-Global metrics estimated with HRT show lower sensitivity-estimation efficiency compared with the other parameters assessed. Individualized intra-patient critical limits require a small increase of the measured quantity to identify glaucoma in the patient who has the disease.