Precision medicine in schizophrenia: stratifying patients by inflammatory, metabolic, and gut-brain biomarker profiles using latent profile analysis
M. Couce-Sanchez, F. Dal Santo, M. Bernardo, G. Safont, M. P. Garcia-Portilla, B. Arranz, L. Gonzalez-BlancoIntroduction
Schizophrenia is a heterogeneous mental disorder with diverse clinical and biological profiles. Current diagnostic systems relying solely on symptoms may overlook the full diversity of patients’ presentations, so developing neurobiological subtypes has been of great interest in recent research (Clementz et al. Schizophr Bull 2022; 48 56-68).
Biomarkers such as inflammatory markers (Lalousis et al. Brain Behav Immun 2023; 113 166-175), features of metabolic syndrome (Saccaro et al. Front Psychiatry 2024; 15 1225693), and gut-brain axis indicators (González-Blanco et al. Eur Psychiatry 2024; 67(1) e84) provide insights into pathophysiology and could enhance precision psychiatry approaches.
Objectives
To stratify schizophrenia patients into subgroups based on inflammatory, metabolic, and gut-brain biomarkers. To investigate associations of such profiles with clinical outcomes, including symptom severity and cognitive performance.
Methods
Data from 199 patients diagnosed with schizophrenia (DSM-5) (mean age 40.9 years, 60.3% male), enrolled in an observational, cross-sectional, multicentre study from four centres in Spain (PI17/00246) were analysed.
Assessment:
Biomarkers:
Inflammatory: C-reactive protein (CRP; mg/L).
Metabolic: Abdominal circumference (cm)
Gut-brain axis: Lipopolysaccharide-binding protein (LBP; μg/mL).
Psychometrics: CGI-SCH, Screen for Cognitive Impairment in Psychiatry (SCIP), Positive and Negative Syndrome Scale (PANSS) for Schizophrenia.
Statistics:
Models based on latent profile analysis (LPA) with 1 to 6 profiles were estimated and compared using AIC, BIC, entropy, and the bootstrap likelihood ratio test (BLRT).
Differences in cognitive tests (SCIP) and symptom severity (PANSS, CGI-SCH) were assessed by ANCOVA (controlled by age). Analyses were conducted in R.4.5.1 (package: tidyLPA)
Results
A two-profile solution (AIC = 1555, BIC = 1588, entropy = 0.930, BLRT < 0.01) was selected as optimal for balancing fit, parsimony, and clinical interpretability. Two distinct biomarker profiles emerged: a normative profile (91.5% of participants) with typical biomarker levels, and a high-risk profile (8.5%) with elevated inflammatory and metabolic markers (see profiles’ representation in image 1 ).
The high-risk group showed significant impairments in working memory and verbal fluency compared to the normative group (see comparison in
figure 2
); no significant differences were found in PANSS or CGI-SCH scores.
Image 1: Long description.