Statistical Analysis

Statistical analyses were performed with GraphPad Prism version 9.3.1, SPSS Version 25, and Microsoft Excel Version 2203. Statistical significance was considered as P < 0.05. Continuous values were expressed as mean (95% CI) unless otherwise specified. Non-continuous variables, for example, sex, presence of pigmentary abnormalities, and study group, were dummy-coded for regression analyses.⁸⁶ Normality was assessed using the D'Agostino-Pearson test. Thicknesses (by grid or cluster) were derived from each participant as a single unit of observation (i.e. independent data). Hence, single comparisons of continuous data between groups used unpaired Student's t-test or Mann-Whitney U test depending on normality. As each comparison was individually important, there was no statistical adjustment for multiple comparisons (within each individual retinal layer) and instead results were considered contextually.⁸⁷ Multiple comparisons of continuous data between groups used one-way ANOVA and Tukey's multiple comparisons test. Comparisons of continuous data within groups (i.e. between clusters within each individual retinal layer) used paired t-test or mixed-effects repeated measures model with non-assumption of sphericity (equal variability of differences and Geisser-Greenhouse correction) and Holm-Sidak's multiple comparisons test, depending on number of comparisons. Comparisons of categorical data used χ² test. Comparisons of paired proportions (i.e. between clusters subdivided into quadrants and/or central, peri-central, para-central, peripheral, and extra-macula) used McNemar's and Cochran's Q tests.⁸⁸ Effect sizes (Z-scores) were interpreted according to Cohen et al.⁸⁹ ≥0.2 = small, ≥0.5 = medium, and ≥0.8 = large.