Sample size estimation was based on a two-way repeated-measures ANOVA (between-groups within-time interaction), assuming a standardized effect size of f = 0.25 (medium), 3 independent groups, 9 repeated measurements, power = 95%, α = 0.05, correlation among measures = 0.40, and a Greenhouse–Geisser ε = 0.50. The minimum required sample size was 51 patients, computed in G*Power 3.1.9.7.
Continuous data were expressed as mean ± standard deviation (SD) or median (first quartile [Q1] and third quartile [Q3]) for symmetrical or skewed distributions, respectively. Categorical data were expressed as the number of cases (frequencies) or percentages (relative frequencies).
The relationships between the categorical variables were investigated by the Chi-squared test for independence.
Continuous variables in the three cohorts were compared with the one-way analysis of variance (ANOVA), Welch ANOVA, or Kruskal–Wallis test for symmetrical or skewed distributions, respectively. Post hoc comparisons for Welch ANOVA were conducted by Games–Howell tests.
IOP was analyzed with a linear mixed-effects model including fixed effects for group (vitreous status), time, and their interaction. Patient clustering was modeled with a random intercept, and within-eye repeated measurements. Estimated marginal means for group (vitreous status) at different timepoints were obtained and Bonferroni-adjusted between-group comparisons at each time point were performed. This specification accommodates unequal variances over time, serial correlation within eyes, and patient-level clustering.
IOP levels at all timepoints with respect to all examined parameters such as age, sex, AL, ACD, CCT, lens, and angle status were analyzed also with a linear mixed-effects model.
The agreement between GAT and RBT was analyzed with the Bland–Altman method
Calculations were conducted using IBM SPSS Statistics version 29.0.0.0 (241). P-values <0.05 were regarded as significant.