2026 ECNP Congress

PS01-1082

Towards an integrated staging model of bipolar disorder: preliminary results from a multicentre study

Valentina Casati — N. Girone, M. Macellaro, L. Cremaschi, C. Gesi, B. Benatti, M. Leonardi, G. dell’Oca, E. Rinaldi, P. Gallazzi, M. Cocchi, S. Saluzzi, A. Di Giorgio, I. Rosa, C. Cavallotto, C. D’Addario, A. Caporali, F. Martella, F. de Pasquale, M. Pettorruso, G. Martinotti, B.M. Dell’Osso · Università degli studi di Milano, Department of Mental Health and Addiction- ASST Fatebenefratelli Sacco, Milan, Italy

2026 ECNP Congress

Background

Bipolar disorder (BD) is a chronic and recurrent condition, with growing evidence supporting a neuroprogressive course characterised by cumulative worsening of symptoms, functioning, and treatment response over time. This progression is thought to be driven by the interaction of multiple biological processes, including inflammatory and immune dysregulation, epigenetic alterations, mitochondrial dysfunction, and impaired neuroplasticity [1]. In this context, staging models represent a valuable tool to describe the longitudinal evolution of BD and to guide stage-specific therapeutic interventions. However, despite the various approaches proposed, the application of staging models remains limited and is predominantly based on clinical data [2]. Within this framework, the BipOlAR Disorder Integrative stagiNG Progression AcrosS Stages (BOARDING-PASS) project was developed as a multicentre, prospective, longitudinal observational study [3]. The aim of this study was to develop and validate an integrated, biologically informed staging model of BD by combining multimodal data (clinical, biological, and neuroimaging) and using machine learning approaches.

Methods

A total of 126 participants (aged 18–70 years) were recruited across three secondary-level psychiatric centres (Milan, Bergamo, Chieti). Disease staging was assigned according to the Kupka and Hillegers model at four time points: baseline (T0), 6 months (T1), 12 months (T2), and 18 months (T3), alongside sociodemographic and clinical variables. At T0, T2, and T3, blood and saliva samples were collected for epigenetic and inflammatory biomarker analyses, alongside acquisition of structural and functional 3T magnetic resonance imaging (MRI) data. Multimodal data were integrated using a supervised machine learning model based on Support Vector Machine (SVM). Model performance was evaluated using accuracy and F1-score metrics.

Results

At baseline, participants were distributed across disease stages, with a predominance of stage 3 (47.6%). At 6 months, the clinical profile remained overall stable, with a slight trend towards progression into recurrent substages. Preliminary biological findings showed a distinct inflammatory profile in individuals with BD, characterised by increased expression of IL-1β, IL-2, IFN-γ, and IL-8, alongside reduced BDNF mRNA levels. Morphological connectome analyses, based on the gyrification index, revealed a progressive and stage-dependent reorganisation of brain networks, with reduced local cohesion, increased modular segregation, and decreased network resilience, while global integration remained relatively preserved. In exploratory machine learning analyses, the multimodal model (neuroimaging and biomarkers) achieved an accuracy of 83% and an F1-score of 78%, outperforming unimodal models based on gene expression (70%) or functional MRI data (68%).

Conclusions

These preliminary findings support a model of selective and regionally circumscribed brain network vulnerability in BD, in which inflammatory and epigenetic alterations, together with connectome topological changes, complement clinical staging in characterising disease trajectory. The integration of clinical, biological, and neuroimaging data, supported by machine learning approaches, may represent a promising strategy to improve the predictive accuracy of BD staging models. Further follow-up and larger samples will be required to confirm these findings and support their potential clinical applicability.

References

  1. Martella F, Caporali A, Macellaro M, Cafaro R, De Pasquale F, Dell’Osso B, D’Addario C. Biomarker identification in bipolar disorder. Pharmacol Ther. 2025 Apr;268:108823. doi: 10.1016/j.pharmthera.2025.108823. Epub 2025 Feb 17. PMID: 39965667. 2. Dell’Osso B, Cremaschi L, Macellaro M, Cafaro R, Girone N. Bipolar disorder staging and the impact it has on its management: an update. Expert Rev Neurother. 2024 Jun;24(6):565-574. doi: 10.1080/14737175.2024.2355264. Epub 2024 May 16. PMID: 38753491. 3. Girone N, Cremaschi L, Macellaro M, Gesi C, Martella F, Saluzzi S, Martinelli S, Caporali A, Rosa I, Cavallotto C, Di Giorgio A, D’Addario C, Martinotti G, Pettorruso M, Perrucci MG, de Pasquale F, Gerevini S, Bondi E and Dell’Osso B (2025) Bipolar disorder integrative staging: incorporating biomarkers into progression across stages (BOARDING-PASS) – rationale and design. Front. Psychiatry 16:1646533. doi: 10.3389/fpsyt.2025.1646533