Dynamic negative symptom networks following personalized iTBS in schizophrenia

Abstract

Background Negative symptoms are clinically heterogeneous and may be represented as interacting symptom domains. Whether personalized intermittent theta-burst stimulation (iTBS) is associated with changes in contemporaneous or temporal associations among negative-symptom domains remains uncertain. We therefore explored longitudinal negative-symptom network estimates following active versus sham iTBS. Methods Data were drawn from a double-blind, randomized, sham-controlled trial of 80 participants with schizophrenia receiving active or sham personalized iTBS. Negative symptoms were assessed using the Scale for the Assessment of Negative Symptoms (SANS) at baseline and weeks 4, 8, and 12. Cross-sectional networks were estimated using regularized partial correlations and cross-lagged panel networks using LASSO regression. Between-group differences were evaluated using permutation tests with false discovery rate correction, and cross-lagged estimates were assessed using participant-level bootstrap resampling. Results Cross-sectional analyses indicated time-dependent between-group differences in network organization. Sensitivity NCTs suggested greater global strength in the iTBS group at week 8 ( p = 0.015) and a difference in overall network structure at week 12 ( p = 0.030). Unadjusted edge differences involved alogia–anhedonia at weeks 4 and 8 and affective flattening–attention at week 8; after BH-FDR correction across all 40 edge comparisons, only the stronger affective flattening–avolition edge in the iTBS group at week 12 remained significant (q = 0.0066). The longitudinal networks were dominated by autoregressive effects, with six effects in the iTBS group and two in the sham group remaining significant after FDR correction. No cross-lagged effect within either group or formal between-group directed-edge difference was significant. Exploratory interaction models yielded several nominal between-group differences, but none survived FDR correction. Conclusions The most robust network difference following iTBS involved the affective flattening–avolition association at Week 12, whereas the longitudinal networks were dominated by autoregressive symptom persistence. These findings highlight a potentially relevant late coupling between expressive and motivational symptoms and warrant replication in larger longitudinal samples.