Dissociation between the dynamics of symptom scores and network metrics in psychosis treatment: a proof-of-concept...

Antipsychotic efficacy in schizophrenia spectrum disorders (SSD) is commonly evaluated using static measures that fail to capture the dynamic evolution of symptoms and the unfolding impact of treatment over time. Network intervention analysis (NIA) is a novel approach that models pharmacological treatments as active nodes within longitudinal symptom networks.

This proof-of-concept study aimed to contrast the information about antipsychotic-symptom interactions provided using NIA with that using a simple time-based analysis of symptom progression. We hypothesised that while overall symptom improvement would be similar across pharmacological profiles, the symptom decoupling patterns (connection density) extracted from NIA would differ significantly. We also hypothesised that the rate of change for the decoupling would be significantly dissociated from the rate of change for the clinical symptoms.

Patients with SSD underwent weekly symptom evaluation over six weeks. NIA was used to characterise the evolving impact of treatment with three distinct drug classes: muscarinic, serotonergic/dopaminergic, and adrenergic/low dopaminergic. Standardised slopes from NIA density regression models and symptom mixed-effect models were directly compared using a robust non-parametric bootstrap procedure.

NIA density analysis showed a significant time*group interaction, (p = 0.016) which was not observed in symptom severity trajectories. The slope for density reduction was steeper than the slope for symptom severity reduction (Δβ̂=0.426, p = 0.002). For each receptor profile class, the treatment node within the NIA demonstrated distinct patterns of association with symptoms.

These findings highlight the potential of NIA to capture the evolving interaction between antipsychotic classes and dynamic symptom trajectories in SSD.