Skip to main content
. 2022 Sep 7;53(12):5558–5568. doi: 10.1017/S0033291722002768

Table 3.

Longitudinal associations of white matter connectivity measures with incident depressive symptoms

Model Incident depressive symptoms n = 373 cases Hazard ratio (95% CI) p value
Number of connectionsa
Lower global node degree (per s.d.)
Model 1 1.05 (0.95–1.17) 0.344
Model 2 1.06 (0.95–1.19) 0.264
Model 3 1.06 (0.95–1.18) 0.334
White matter organizationa
Lower clustering coefficient (per 1 s.d.)
Model 1 0.92 (0.79–1.07) 0.268
Model 2 0.91 (0.78–1.06) 0.202
Model 3 0.91 (0.78–1.06) 0.217
Lower local efficiency (per 1 s.d.)
Model 1 0.93 (0.78–1.11) 0.403
Model 2 0.91 (0.76–1.08) 0.273
Model 3 0.91 (0.76–1.08) 0.282
Higher characteristic path length (per 1 s.d.)
Model 1 0.99 (0.89–1.10) 0.806
Model 2 0.97 (0.87–1.08) 0.578
Model 3 0.97 (0.87–1.08) 0.550
Lower global efficiency (per 1 SD)
Model 1 1.01 (0.91–1.12) 0.892
Model 2 1.00 (0.90–1.11) 0.985
Model 3 1.00 (0.90–1.11) 0.944

PHQ-9, indicates 9-item patient health questionnaire; CI indicates confidence interval; s.d., standard deviation. Incident depressive symptoms is defined as no depressive symptoms (PHQ-9 < 10) at baseline and depressive symptoms (PHQ-9 ⩾ 10) on at least one follow-up moment.

n = 4417 in model 1. Longitudinal data are evaluated using Cox proportional hazard regressions.

Model 1: adjusted for global node degree (graph measures), age, sex, MRI date, educational level, and T2DM.

Model 2: additionally adjusted for waist circumference, total/high-density cholesterol ratio, lipid-modifying medication, systolic blood pressure, antihypertensive medication, history of cardiovascular disease, and history of the cardiovascular accident. Data missing n = 50.

Model 3: additionally adjusted for smoking behavior and alcohol use. Additional data missing n = 2.

a

Global node degree, clustering coefficient, local efficiency, and global efficiency are inversed (i.e. multiplying it by − 1) to reflect structural dysconnectivity.