Predictors of functional outcome in patients with major depression and bipolar disorder: A dynamic network approach to identify distinct patterns of interacting symptoms

• Published in: PloS one Vol. 18; no. 2; p. e0276822
• Main Authors: Platania, Giuseppe Alessio, Savia Guerrera, Claudia, Sarti, Pierfrancesco, Varrasi, Simone, Pirrone, Concetta, Popovic, Dina, Ventimiglia, Andrea, De Vivo, Simona, Cantarella, Rita Anna, Tascedda, Fabio, Drago, Filippo, Di Nuovo, Santo, Colliva, Chiara, Caraci, Filippo, Castellano, Sabrina, Blom, Johanna M C
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 15-02-2023; Public Library of Science (PLoS)
• Subjects: Age; Biology and Life Sciences; Bipolar Disorder; Care and treatment; Clusters; Cognition; Cognitive ability; Cognitive therapy; Computer and Information Sciences; Covariance matrix; Depression; Depressive Disorder, Major; Diagnosis; Domains; Drug therapy; Executive Function; Humans; Major depressive disorder; Medicine and Health Sciences; Memory; Mental depression; Mental disorders; Methods; Network analysis; Neural networks; Nodes; Patient outcomes; Patients; Pharmacology; Phenotypes; Prospective Studies; Psychiatric services; Psychotropic drugs; Recovery of function; Signs and symptoms; Social interactions; Social Sciences; Therapeutic applications; Italy
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0276822

The purpose of this study is to use a dynamic network approach as an innovative way to identify distinct patterns of interacting symptoms in patients with Major Depressive Disorder (MDD) and patients with Bipolar Type I Disorder (BD). More precisely, the hypothesis will be testing that the phenotype of patients is driven by disease specific connectivity and interdependencies among various domains of functioning even in the presence of underlying common mechanisms. In a prospective observational cohort study, hundred-forty-three patients were recruited at the Psychiatric Clinic "Villa dei Gerani" (Catania, Italy), 87 patients with MDD and 56 with BD with a depressive episode. Two nested sub-groups were treated for a twelve-week period, which allowed us to explore differences in the pattern of symptom distribution (central vs. peripheral) and their connectedness (strong vs weak) before (T0) and after (T1) treatment. All patients underwent a complete neuropsychological evaluation at baseline (T0) and at T1. A network structure was computed for MDD and BD patients at T0 and T1 from a covariance matrix of 17 items belonging to three domains-neurocognitive, psychosocial, and mood-related (affective) to identify what symptoms were driving the networks. Clinically relevant differences were observed between MDD and BD, at T0 and after 12 weeks of pharmacological treatment. At time T0, MDD patients displayed an affective domain strongly connected with the nodes of psychosocial functioning, while direct connectivity of the affective domain with the neurocognitive cluster was absent. The network of patients with BD, in contrast, revealed a cluster of highly interconnected psychosocial nodes but was guided by neurocognitive functions. The nodes related to the affective domain in MDD are less connected and placed in the periphery of the networks, whereas in BD they are more connected with psychosocial and neurocognitive nodes. Noteworthy is that, from T0 to T1 the "Betweenness" centrality measure was lower in both disorders which means that fewer "shortest paths" between nodes pass through the affective domain. Moreover, fewer edges were connected directly with the nodes in this domain. In MDD patients, pharmacological treatment primarily affected executive functions which seem to improve with treatment. In contrast, in patients with BD, treatment resulted in improvement of overall connectivity and centrality of the affective domain, which seems then to affect and direct the overall network. Though different network structures were observed for MDD and BD patients, data suggest that treatment should include tailored cognitive therapy, because improvement in this central domain appeared to be fundamental for better outcomes in other domains. In sum, the advantage of network analysis is that it helps to predict the trajectory of future phenotype related disease manifestations. In turn, this allows new insights in how to balance therapeutic interventions, involving different fields of function and combining pharmacological and non-pharmacological treatment modalities.