Invest Ophthalmol Vis Sci. 2026 Jul 1;67(8):59. doi: 10.1167/iovs.67.8.59.
ABSTRACT
PURPOSE: The purpose of this study was to characterize topographical visual field (VF) improvement using pointwise total deviation (TD) slopes referenced to baseline severity, and to evaluate structure-function and cognitive associations in eyes receiving nicotinamide and pyruvate (NAM + PYR) treatment versus placebo.
METHODS: This post hoc analysis used study-eye data from a phase 2 randomized trial of nicotinamide (3 g/day) plus pyruvate (3 g/day) versus placebo in open-angle glaucoma, with four baseline and four post-baseline VF visits. A location was improving if its Huber robust pointwise TD slope ranked in the upper decile (≥90th percentile) of the pooled slope distribution within its baseline-severity stratum. Bayesian hierarchical models assessed treatment effects, incorporating optical coherence tomography retinal nerve fiber layer (RNFL) slopes as structural priors; Montreal Cognitive Assessment (MoCA) scores tested for cognitive confounding.
RESULTS: Thirty-two study eyes (21 receiving NAM + PYR, and 11 placebo) were analyzed. NAM + PYR treated eyes showed increased identification of improving locations versus placebo (Cochran-Mantel-Haenszel pooled odds ratio (OR) = 3.73, 95% confidence interval (CI) = 2.36-5.89, P < 0.001); a TD-difference sensitivity analysis was concordant (OR = 5.08, 95% CI = 1.22-21.17, P = 0.021). Improvement clustered along arcuate pathways and within moderate-to-deep baseline damage. The Bayesian model estimated a 98.4% posterior probability that NAM + PYR treatment increased improving locations; RNFL priors corroborated modestly but did not materially refine inference. No association was found with MoCA scores.
CONCLUSIONS: Topographical, slope-based endpoints reveal localized arcuate VF improvement with nicotinamide and pyruvate that is missed by global indices and not explained by cognitive change. Bayesian modeling suits small early-phase neuroprotection trials, although structural priors added little over this short follow-up.
PMID:42517851 | DOI:10.1167/iovs.67.8.59
