Reliability, Convergent Validity and Time Invariance of Default Mode Network Deviations in Early Adult Major Depressive Disorder

Abstract

There is substantial variability across studies of default mode network (DMN) connectivity in major depressive disorder, and reliability and time-invariance are not reported. This study evaluates whether DMN dysconnectivity in remitted depression (rMDD) is reliable over time and symptom-independent, and explores convergent relationships with cognitive features of depression. A longitudinal study was conducted with 82 young adults free of psychotropic medications (47 rMDD, 35 healthy controls) who completed clinical structured interviews, neuropsychological assessments, and 2 resting-state fMRI scans across 2 study sites. Functional connectivity analyses from bilateral posterior cingulate and anterior hippocampal formation seeds in DMN were conducted at both time points within a repeated-measures analysis of variance to compare groups and evaluate reliability of group-level connectivity findings. Eleven hyper- (from posterior cingulate) and 6 hypo- (from hippocampal formation) connectivity clusters in rMDD were obtained with moderate to adequate reliability in all but one cluster (ICC's range = 0.50 to 0.76 for 16 of 17). The significant clusters were reduced with a principle component analysis (5 components obtained) to explore these connectivity components, and were then correlated with cognitive features (rumination, cognitive control, learning and memory, and explicit emotion identification). At the exploratory level, for convergent validity, components consisting of posterior cingulate with cognitive control network hyperconnectivity in rMDD were related to cognitive control (inverse) and rumination (positive). Components consisting of anterior hippocampal formation with social emotional network and DMN hypoconnectivity were related to memory (inverse) and happy emotion identification (positive). Thus, time-invariant DMN connectivity differences exist early in the lifespan course of depression and are reliable. The nuanced results suggest a ventral within-network hypoconnectivity associated with poor memory and a dorsal cross-network hyperconnectivity linked to poorer cognitive control and elevated rumination. Study of early course remitted depression with attention to reliability and symptom independence could lead to more readily translatable clinical assessment tools for biomarkers.

Keywords: cognitive control network; default mode network; depression; reliability; resting-state fMRI; rumination; time-invariance.

Figure 1

Investigation Strategy for DMN Reliability and Time Invariance in MDD. Yellow DMN viewed from medial sagittal slice. White circles denote PCC and HPF seeds. Analyses examine DMN connectivity differing between rMDD and HC for (A) reliability and specificity by examining ICCs after controlling for numerous potential confounders; (B) disease course by examining the remitted phase and clinical features; and (C) exploratory convergent validity through correlation with cognitive features of the illness.

Figure 2

Days Between fMRI Scans Boxplot Separated by Diagnostic Group. Boxplots of number of days between two fMRI scan sessions for healthy control and remitted major depressive disorder participants. Dark colored line represents the median, open circles are outliers and asterisks represent significant outliers. There were a total of 4 outliers for healthy controls and 2 outliers for remitted MDD individuals.

Figure 3

Comparison of Prefrontal Cortex Clusters with Active and Familial Risk for Depression. Comparison of 3 left PCC seed rMDD-different connectivity clusters (red), 3 right PCC seed rMDD-different connectivity clusters (yellow), bilateral SFG connectivity from precuneus seed (also showing connectivity with our PCC seed) found in active depression (18) (blue) and bilateral MFG connectivity from PCC found in familial-risk for depression (19) (green). (A) Axial view of each significant cluster and ROIs with 8 mm radius found in studies of active and familial-risk for depression. (B) t-test comparisons of mean connectivity in first scan from bilateral PCC to prefrontal cortex clusters found in the current study, combined left and right active depression SFG, and combined left and right familial-risk depression MFG. Bars represent ±1 standard error from the mean. ^*Significantly different between groups (p < 0.005, k > 57). aMDD, active major depressive disorder; f-MDD, familial-risk for major depressive disorder.

Figure 4

Connectivity Clusters Different Between rMDD and HC from Bilateral PCC and HPF. White circles denote bilateral seeds on medial sagittal view of each hemisphere with DMN highlighted in yellow. Branches to clusters are weighted by level of reliability, with solid branches representing hyperconnectivity in rMDD and dotted branches representing hypoconnectivity. Cluster size is scaled according to number of significant voxels. Colors indicate component-belonging and significant/trend-level convergent validity and disease course associations. COM, Component. ^†p < 0.10, ^*p < 0.05, ^**p < 0.01.