Background Glaucoma is defined by damage to the optic nerve that results in visual field (VF) loss widely considered to be irreversible. However, in a landmark clinical trial for glaucoma known as the Ocular Hypertension Treatment Study (OHTS)(1), reversal of VF loss was observed but assumed to be caused by artifact. We tested the assumption of whether vision loss is always irreversible.

Methods Given the inherent variability of VF testing, a prevailing challenge is distinguishing true improvement from random noise. Permutation of pointwise linear regression (PoPLR)(2) is a statistical method that determines whether an observed VF trend is a true phenomenon. PoPLR determined the prevalence of eyes with VF outcome (improvement, deterioration, or stability) with the threshold for random variation at 5%. Archetypal analysis (AA)(3) is an unsupervised machine learning technique that decomposes an individual VF into patterns of glaucomatous defects. Longitudinal trends of glaucomatous defect were measured with linear regression. To assess whether learning effect is confounder for improvement, reliability indices including false positive rate (FPR), false negative rate (FNR), and fixation losses (FL) were compared across VF outcomes using one-way ANOVA. We also analyzed structural differences by comparing rates of significant thinning of retinal nerve fiber layer (RNFL) using Chi-Square analysis. A p-value less than 0.05 was considered statistically significant.

Results This study isolated 886 eyes from 619 patients, of which 106 eyes (11.96%), 331 eyes (37.36%) and 449 eyes (50.68%) demonstrated VF improvement, deterioration, and stability, respectively. Eyes exhibiting VF improvement had significantly decreased representation of superior altitudinal defect (-0.05%/year, p<0.001), inferotemporal defect (-0.08%/year, p=0.013), and inferonasal defect (-0.12%/year, p<0.001). There was no significant difference in FPR (p=0.203), FNR (p=0.631), or FL (p=0.105) across VF outcomes. VF improvement was found to have a lower rate of significant RNFL thinning compared to VF deterioration (11.1% vs. 42.2%, p<0.001).

Conclusion Reversal of glaucomatous vision loss is a true phenomenon as VF improvement occurred more often than random variation and was accompanied by a decrease in classic glaucomatous defects with preserved RNFLT integrity. Furthermore, learning effect was not a significant confounder for improvement. Future research will assess which treatments (medication vs. surgery) can optimize reversal of glaucomatous vision loss and therefore aid clinical decision-making.

References 1. Keltner JL, Johnson CA, Quigg JM, et al. Confirmation of visual field abnormalities in the Ocular Hypertension Treatment Study. Arch Ophthal. 2000;118(9):1187-1194. https://jamanetwork.com/journals/jamaophthalmology/article-abstract/413586 2. O’Leary N, Chauhan BC, Artes PH. Visual field progression in glaucoma: estimating the overall significance of deterioration with permutation analyses of pointwise linear regression (PoPLR). Invest Ophthalmol Vis Sci. 2012;53(11):6776-6784. doi:10.1167/iovs.12-10049 3. Elze T, Pasquale LR, Shen LQ, Chen TC, Wiggs JL, Bex PJ. Patterns of functional vision loss in glaucoma determined with archetypal analysis. J R Soc Interface. 2015;12(103). doi:10.1098/rsif.2014.1118