The interplay between affective and cognitive symptoms in difficult-to-treat depression: A symptom network analysis

J Affect Disord. 2026 Apr 26:121874. doi: 10.1016/j.jad.2026.121874. Online ahead of print.

ABSTRACT

BACKGROUND: Cognitive impairment is a prevalent and burdensome feature of difficult-to-treat depression, yet the symptom-level interdependencies between affective and cognitive domains remain poorly understood. We examined cross-domain interactions using a symptom network approach.

METHODS: We conducted an exploratory cross-sectional symptom network analysis in 120 patients with difficult-to-treat depression. Networks included seven harmonized depression symptoms (Hamilton Depression Rating Scale/Montgomery-Åsberg Depression Rating Scale) and three cognitive measures (processing speed, semantic memory, cognitive flexibility). Network structure was estimated using Mixed Graphical Models with LASSO regularization. Bridge symptoms were identified using bridge expected influence indices.

RESULTS: The network revealed selective cross-domain connections. Positive associations were observed between reduced appetite and processing speed and cognitive flexibility, as well as between depressed mood and semantic memory. A negative association was observed between suicidal ideation and semantic memory. Reduced appetite emerged as the most important bridge symptom linking affective and cognitive domains. Stability analyses indicated acceptable stability of bridge nodes, while bootstrapped confidence intervals suggested substantial uncertainty in individual edge estimates, underscoring the exploratory nature of the findings.

CONCLUSIONS: This exploratory analysis provides initial evidence for cross-domain, symptom-level relationships in difficult-to-treat depression. These findings highlight heterogeneity that may remain obscured in total-score approaches. Bridge symptoms may represent candidate adjunctive targets, as modulating such nodes could potentially influence multiple domains simultaneously. However, given the modest sample size and limited network stability, the results are hypothesis-generating and require replication in larger, independent cohorts before clinical interpretations can be drawn.

PMID:42049069 | DOI:10.1016/j.jad.2026.121874