Statistical Analysis

We compared the demographic and clinical data using the independent two-sample t-test, Chi-square test, or analysis of variance (ANOVA) between subgroups when appropriate, in Statistical package for Social Sciences (SPSS 22.0 for Windows). The significance threshold was set at p < 0.05.

Because the abnormal activation characteristic/pattern of brain often results from disease (depression), onset-age factor, and their interaction, a 2 × 2 full factorial model was used in SPM8, assessing the effects of onset age (young vs. old) and diagnosis (MDD vs. HCs) with age, gender, and education level as covariates; the interaction between onset age and diagnosis and the main effects of age-onset and diagnosis were tested. Multiple comparisons were corrected using the Monte Carlo simulation-based Alphasim program (1,000 times): the initial threshold on voxel level was set as p < 0.001, the cluster level was p < 0.05, and the corresponding cluster size was >16 (connection radius = 5 mm for correction).

Post-hoc analysis was conducted on extracted values of the clusters (Figure 1) that reached significance in the interaction analysis mentioned above. This analysis was completed in SPSS; Bonferroni-corrected pairwise comparisons were calculated to further investigate significant simple effects. To investigate the relationship between abnormal ReHo and clinical features, partial correlations were analyzed to explore the correlation between these abnormal values of ReHo and HDRS score or illness duration in each patient subgroup (age, gender, and years of education as covariates).

Interaction between onset age and depression on Regional homogeneity (ReHo). Compared with young HCs and LOD patients, EOD patients had increased ReHo in the left postcentral gyrus (A) and right inferior frontal triangular gyrus (B). However, EOD patients had significantly decreased ReHo in the right superior frontal gyrus/supplementary motor area and the right medial frontal gyrus (C) than young HCs. Color bar represent the t-value.