Abstract
Schizophrenia spectrum disorders (SSDs) exhibit a marked heterogeneity in clinical course and treatment outcomes. Some individuals achieve remission and recovery, whereas others experience repeated relapses and progressive deterioration in psychosocial functioning. This variability underscores the unmet clinical need for prognostic biomarkers to predict treatment outcomes and guide personalized care. Deep phenotyping with multimodal data integration holds promise for understanding this complexity and delivering clinically relevant predictive models of treatment response in SSDs. To address this need, we initiated the Clinical Deep Phenotyping of Treatment Response in Schizophrenia (CDP-STAR) study, a prospective, naturalistic, longitudinal observational study integrating comprehensive multimodal assessments. These include clinical phenotyping, magnetic resonance imaging (MRI), electroencephalography (EEG), retinal imaging, and extensive sampling of blood and cerebrospinal fluid (CSF) for multi-omics profiling. The study aims to externally validate promising biomarker candidates and elucidate the pathophysiological mechanisms underlying treatment outcomes. This innovative deep phenotyping framework integrates data across multiple critical domains, enabling external validation of potential biomarkers and the discovery of novel ones. Ultimately, the CDP-STAR study aims to yield mechanistic insights that advance precision psychiatry and inform clinical decision-making in SSDs.
Introduction
Schizophrenia spectrum disorders (SSDs) are characterized by significant clinical heterogeneity. While some patients respond well to initial pharmacological interventions and achieve remission, a substantial portion—up to one-third—experience treatment resistance or progressive functional decline. Currently, clinical practice relies on subjective symptom assessment rather than biological markers to predict treatment trajectories. Although numerous biomarker candidates have been identified in research settings, few have undergone the rigorous external validation in longitudinal, real-world cohorts necessary for clinical translation.
Methodology: The Deep Phenotyping Protocol
The Clinical Deep Phenotyping of Treatment Response in Schizophrenia (CDP-STAR) study was designed to bridge the gap between discovery and clinical application. As a prospective, naturalistic, longitudinal observational study, CDP-STAR employs a "deep phenotyping" approach, collecting high-dimensional data across multiple biological and clinical domains.
The assessment protocol includes:
- Clinical and Cognitive Characterization: Detailed psychiatric evaluations and environmental assessments.
- Multimodal Neuroimaging: Structural MRI, resting-state functional MRI, diffusion tensor imaging (DTI), and dynamic contrast-enhanced (DCE) MRI to evaluate blood-brain barrier (BBB) integrity.
- Electrophysiology: Resting-state EEG to capture neural oscillations.
- Retinal Imaging: Utilizing the retina as a "window to the brain" via optical coherence tomography (OCT) and electroretinography (ERG).
- Multi-omics Biobanking: Collection of blood and cerebrospinal fluid (CSF) for genomics, transcriptomics, proteomics, and metabolomics.
Patients are evaluated at baseline, 4 weeks, 3 months, 6 months, and 2 years. The primary outcome measure is the predictive accuracy (ROC-AUC) of biomarkers for acute treatment response, defined as a $\ge$25% reduction in symptoms at the 4-week mark.
Cohort and Data Integration
The CDP-STAR cohort is integrated into the broader Munich Mental Health Biobank. The study aims to recruit 352 patients, targeting a final sample of at least 132 complete longitudinal datasets to account for attrition. To enhance statistical power for secondary analyses, such as unsupervised clustering, this cohort will be pooled with existing cross-sectional (N=233) and interventional (N=120) datasets, resulting in a harmonized database of nearly 500 deeply phenotyped individuals.
Strategic Insights for Precision Psychiatry
CDP-STAR adopts several innovative strategies to advance the field:
- External Validation: A primary focus is the rigorous testing of existing biomarker candidates, such as the functional striatal abnormalities (FSA) score, in an independent, real-world clinical population.
- Systemic Pathophysiology: By including DCE-MRI and CSF sampling, the study investigates SSDs as systemic conditions involving neuroinflammation and BBB dysfunction, rather than isolated brain disorders.
- Retinal Proxies: The use of retinal imaging provides a non-invasive method to assess synaptic dysfunction and neurodevelopmental markers shared between the retina and the central nervous system.
Current Status and Statistical Power
Power analyses indicate that 132 complete profiles (balanced between responders and non-responders) are required to detect a biomarker with a clinically relevant AUC of 0.7 at 80% power. As of mid-2025, recruitment is ongoing with 44 individuals enrolled. Preliminary data shows a baseline PANSS total score average of 84.2, reflecting a markedly ill inpatient population. The observed 24% dropout rate at the 4-week follow-up is consistent with projections for this clinical demographic.
Conclusion and Future Directions
The CDP-STAR study represents a significant step toward precision psychiatry in SSDs. By moving beyond cross-sectional discovery toward longitudinal validation and multi-system integration, the study aims to provide clinicians with reliable tools for predicting treatment response. Future work will leverage unsupervised machine learning (e.g., Gaussian Mixture Models) to identify novel, biologically-driven subgroups within the schizophrenia spectrum, potentially leading to more personalized therapeutic strategies.
Summary updated