258. Metabolic connectomics and brain dynamics in schizophrenia: a novel perspective on treatment resistance and clinical response

Abstract

Abstract Background Around 30% of schizophrenia patients fail to respond to standard antipsychotic treatments, a condition named treatment-resistant schizophrenia (TRS). While altered functional connectivity has been extensively studied in this population, research has predominantly relied on fMRI-derived hemodynamic signals. However, as metabolic connectivity reflects neuronal activation more directly than fMRI, its role in schizophrenia remains underexplored. To our knowledge, the present investigation represents the first attempt to explore brain metabolic connectivity using 18F-FDG-PET, also integrating antipsychotic dose adjustments to address the specific impact of treatment on connectivity patterns. Aims & Objectives We aimed to: i) characterize brain metabolic connectivity in TRS, treatment-responsive schizophrenia (nTRS), and healthy controls (CTRL) using graph theory; ii) identify group differences in network properties measures; and iii) explore the associations between regional metabolic alterations and resistance to antipsychotics through machine learning. Method 18F-FDG-PET was performed in 70 subjects, 28 nTRS, 26 TRS, and 16 CTRL. Relative brain glucose metabolism maps were processed in the AAL-Merged atlas template, extracting data from 65 brain regions. Inter-subject connectivity matrices were derived using Gaussian Graphical Models with an arctangent-type penalty. A statistical comparison between brain networks was conducted through permutation testing. Additionally, a logistic model based on nested cross-validation was employed to estimate the associations between the metabolic signals of brain regions and treatment resistance. To account for the potential influence of antipsychotic medication, we incorporated chlorpromazine equivalents as a covariate in the network analysis during partial correlation calculations. Additionally, the machine-learning analysis employed medication dose-stratified folds. Results Globally, schizophrenia groups exhibited reduced connectivity compared to CTRL (p = 0.008 for nTRS; p = 0.001 for TRS). A different reorganization of DMN was highlighted in TRS compared to nTRS, including aberrant connections of frontal regions with cingulate (p = 0.03) and temporal areas (p = 0.02). TRS patients exhibited further alterations in the functional connectivity of the supplementary motor area (p = 0.01), left postcentral gyrus (p = 0.01), and right caudate (p = 0.01). Dose correction significantly influenced network metrics, unveiling previously unobserved disruptions in TRS, particularly in connections between the anterior cingulate and temporal gyri (p = 0.04), and of frontal gyri with the hippocampus (p < 0.05) and precuneus (p = 0.03). The machine learning model, incorporating dose stratification, achieved an accuracy up to 0.83 (AUC = 0.85; 95%C.I. = 0.74, 0.96; Brier score = 0.15) in identifying TRS patients, with improved predictions when focusing on DMN regions. Discussion & Conclusions The present study tackles a novel approach to metabolic connectivity analysis in psychiatric disorders, particularly schizophrenia. Results reveal specific disruptions in anterior cingulate, frontotemporal, and cortical-subcortical connectivity unique to TRS, offering novel insights into its neurobiological basis. Lastly, our findings suggest that metabolic connectivity could serve as a valuable stratifier for distinguishing TRS from nTRS patients.