Abstract
Purpose :
ANSWERS (Zhu IOVS 2015) is a progression analysis method for the visual field (VF), which takes into account increasing VF variability during progression as well as spatial correlation amongst test locations. VF and imaging measurements of optic nerve (ON) structure are correlated. We evaluated the improvement in diagnostic precision by extending ANSWERS to incorporate spatially-related structural ON measurements .
Methods :
Two sets of VF (Humphrey Field Analyzer, Carl Zeiss Meditec) and ON rim area (RA) (Heidelberg Retina Tomograph, Heidelberg Engineering) data were used. The ‘progression’ datasets were from The United Kingdom Glaucoma Treatment Study (UKGTS) comprising 659 eyes from 437 patients. The ’stable’ dataset comprised 30 eyes from 30 glaucoma patients each of whom were retested 12 times within 2 months. Within a hierarchical Bayesian model in ANSWERS, the prior distribution of the VF progression rate at each VF location was set by the slopes and variance of the rate of change in the corresponding sectorial RA, leading to the structure-guided ANSWERS (sANSWERS). The progression indices from sANSWERS, ANSWERS and permutation of pointwise linear regression (PoPLR) were compared. Positive rate (as a surrogate for sensitivity) and false positive (FP) rate were quantified on the ‘progression’ and ’stable’ datasets, respectively, at various series lengths. Survival analysis was also performed to test the ability to distinguish the treatment and placebo arms in the UKGTS dataset.
Results :
In all series lengths, the positive rate of sANSWERS was higher than those of ANSWERS and PoPLR at equivalent FP rates (Figure). The improvement was greatest in short subseries. Specifically, at 7th months, the positive rate ratio at the 5% FP rate of sANSWERS over ANSWERS and PoPLR was 2.56 and 5.13, respectively. The survival analysis detected the difference between the treatment and placebo arms earlier than ANSWERS and PoPLR (Figure). sANSWERS needs 5 measurements (10 months) to detect significant difference while ANSWERS needs 7 (16 months) at a 5% FP rate.
Conclusions :
With the extra information from spatially-related structural measurements, sANSWERS is more efficient at detecting VF progression and can capture the treatment effect earlier than ANSWERS and PoPLR in a clinical trial.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.