Purchase this article with an account.
Leila Chew, Erica Su, Massood Mohammadi, Vahid Mohammadzadeh, Sepideh Heydar Zadeh, Robert E Weiss, Kouros Nouri-Mahdavi; Separating Glaucomatous vs. Aging-Related Effects in the Macular Region within a Bayesian Hierarchical Framework. Invest. Ophthalmol. Vis. Sci. 2022;63(7):3344 – F0153.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To test the hypothesis that longitudinal glaucoma-related ganglion cell complex (GCC) thinning may be separated from aging-related changes within a Bayesian hierarchical linear model.
63 eyes (45 patients) with suspected glaucoma (GS) and ≥2 years of follow-up and ≥3 macular OCT scans were recruited. The eyes in this cohort had suspicious optic discs but normal RNFL thickness and visual fields at baseline and final follow-up. The rate of RNFL thinning was required to be below 1 μm/year. GCC thickness at 36 3°×3° superpixels from the central 18° of macular volume scans were exported. Rates of GCC change at the superpixel level were estimated with a Bayesian hierarchical model with subject and superpixel-level random effects and residuals. One-sided Bayesian p-values were defined as the posterior probability that a given parameter was greater or less than zero (cutoff p <0.025). The fixed effect of age was included in the model to assess the magnitude of aging-related GCC thinning. The main outcome measure was the difference in the magnitude of the GCC slopes between the longitudinal slopes and cross-sectional aging-related effects.
Mean (SD) age, follow-up time, and number of OCT images were 61.3 (13.9) years, 3.5 (0.7) years, and 4.2 (1.0). The cross-sectional aging effect was significantly negative in 25 out of 36 superpixels with slopes of -0.56 to –0.22 μm per year in significantly changing superpixels. Longitudinal slopes were significantly negative in 34 out of 36 superpixels with slopes ranging from –1.38 to –0.36 μm per year in significantly changing superpixels. 33 of 36 superpixels had a steeper longitudinal slope compared to cross-sectional slope (Figure) and the differences were statistically significant in 9 of these 36 superpixels.
In this cohort, our proposed model was able to separate longitudinal glaucoma-related GCC thinning from aging-related effects; the longitudinal rates of thinning were significantly faster than the rates of change related to aging. To correctly estimate inner macular changes over time, longitudinal data with proper definition of the mean and variance structures are required.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.
Difference in each superpixel between longitudinal slopes (representing both aging and glaucoma effects) and cross-sectional slopes representing mostly effects of aging. The superpixels with red asterisks are statistically significant.
This PDF is available to Subscribers Only