Volumetric analysis of perimetry tests to guide central testing: the functional vulnerability zone

ABSTRACT Purpose To use volumetric analysis applied to total deviation perimetric data to characterise the presence of residual sensitivity measurable using the 10-2 in lieu of the 24-2 test grid in glaucoma (the functional vulnerability zone, FVZ) and develop a prototype end-user application for clinical guidance. Design Cross-sectional study Subjects 626 pairs of Humphrey Field Analyzer 10-2 SITA-Fast and 24-2 SITA-Faster test results of 160 subjects. Methods Total deviation (TD) values located within 10o from fixation (10-2, 68 locations; 24-2, 12 locations) were extracted. Volumetric analysis was performed to calculate the difference in TD "volume" (10-2 – 24-2) for each pair of results. Then, 24-2 test locations were interpolated across the central 10o to obtain "equivalent" 10-2 test locations used to calculate number of test locations gained (10-2 – 24-2). A positive TD volume difference and positive gain represented the presence of an FVZ. Linear regression and principal components analysis (PCA) were used to identify 24-2 test features that may predict the presence of an FVZ. Main outcome measure Volumetric differences in visual field total deviation as a surrogate for a functional vulnerability zone. Results VF global indices and the number of central 24-2 defects predicted TD volume difference (p<0.0001), with 24-2 mean deviation (MD) showing the highest R² (0.32). All parameters had low R² values when predicting gain. A simplified model comprising 24-2 MD and number of central 24-2 TD probability (TDP) defects (p<0.01) provided a method for clinically identifying the likelihood of an FVZ. PCA revealed two principal components (24-2 global indices and pointwise probability scores) that accounted for 79.7% of the variance, with PC1 accounting for 69.3% and comprising 24-2 mean deviation and number of central 24-2 total deviation probability scores at p < 2%. Conclusions The FVZ offers a data-driven approach to identifying residual dynamic range using the 10-2 test grid. A prototype model for predicting the likelihood of 10-2 utility in progression analysis is proposed.