Predicting Repetitive Negative Thinking in Daily Life: Insights From Brain-Based Graph-Theoretical Predictive Modeling.

BACKGROUND

Abnormalities in the functional connectivity of large-scale brain networks, including the default mode network (DMN), salience network, frontoparietal network (FPN), and limbic network, have been implicated in repetitive negative thinking (RNT), a construct characterized by persistent and intrusive thoughts. However, the potential of these large-scale network abnormalities for predicting RNT in daily life remains underexplored.

METHODS

We leveraged brain-based graph-theoretical predictive modeling (GPM) to predict daily-life RNT in 54 individuals. Functional magnetic resonance imaging data were acquired during 1) resting state and 2) an RNT-induced state. RNT severity and its momentary fluctuations were assessed using ecological momentary assessments (EMAs).

RESULTS

GPM identified key functional organizational properties of the DMN, FPN, and limbic networks that differentially predicted the severity and fluctuations of RNT and its specific clinical features (intrusiveness, repetitiveness, RNT-related anxiety). Specifically, the centrality of the medial prefrontal cortex (DMN) predicted EMA fluctuations of intrusiveness and severity of anxiety. Conversely, the strength and centrality of the orbitofrontal cortex (part of the limbic network) predicted EMA fluctuations of repetitiveness, and the segregation of the temporal pole (limbic network) predicted overall severity of RNT. Last, fluctuations in total RNT were predicted from the strength of the orbitofrontal cortex (limbic network) and segregation of the posterior mid-cingulate cortex (FPN). Notably, RNT was better predicted from daily-life prospective assessments than from laboratory-assessed clinical questionnaires.

CONCLUSIONS

These findings highlight the utility of GPM for predicting the emergence of daily-life RNT and suggest specific network-level attributes (e.g., centrality, segregation) underlying RNT and its clinical features.