Network analysis of the Quick Inventory of Depressive Symptomatology: Reanalysis of the STAR*D clinical trial

Network analysis is yet to be used to examine patient-reported symptom severity and change during citalopram treatment for major depressive disorder. We aimed to identify: (I) network systems; (II) central symptoms; and (III) network differences, in patient-reported depression for baseline, endpoint and change scores. STAR*D data during citalopram treatment were reanalyzed to examine depression based on the Quick Inventory of Depressive Symptomatology Self-Report (QIDS-SR). Network analyses were computed from the QIDS-SR item-level severity scores at baseline and endpoint, and from estimated change scores based on mixed models, adjusted for confounding by dose and baseline severity. Centrality indices for each symptom were computed. Networks were contrasted for connectivity with permutation tests. Network analyses grouped symptoms consistently as: Sleep disturbances, cognitive and physical avolition, Affect and Appetite. Symptom centrality was highest for Energy at baseline, Mood at endpoint, and Mood and Concentration on change scores. Generally, permutation tests showed that the networks all significantly (p<.05) differed. Results demonstrated: (I) a replicable network group of the symptoms of depression that modestly mapped onto well-known mechanisms for depression; (II) symptoms with high centrality that may be future treatment targets (e.g., mood); and (III) that the form of the networks differed across treatment time-points, thereby contributing centrality as a possible mechanism to the initial severity debate. These findings highlight the utility of focusing on symptoms rather than total scores to understand how treatment unfolds, and tentative mechanisms.