Metabolic risk stratification and psychotic symptoms in first-episode drug-naïve major depressive disorder: a principal component-based cross-sectional study

Figure 1. Extraction of metabolic principal components and their correlations with clinical symptoms. (A) Scree Plot of principal component analysis for metabolic indicators with eigenvalue-based component selection. The x-axis represents the extracted principal components (PC1 to PC6), and the y-axis represents the eigenvalue of each principal component. The blue line with dots illustrates the decreasing trend of eigenvalues across principal components. The red dashed horizontal line marks the threshold of eigenvalue = 1. Principal components with eigenvalues exceeding 1 (PC1 and PC2) were retained as valid components for subsequent mediation and correlation analyses. (B) Correlation heatmap of metabolic principal components and clinical variables. Pearson correlation heatmap of key variables among patients, including metabolic principal components (PC1, PC2), Hamilton Depression Scale (HAMD) scores, Hamilton Anxiety Scale (HAMA) scores, and psychotic depression (PD) status. The x-axis and y-axis both represent the included variables. Each colored cell displays the Pearson correlation coefficient (r) between two corresponding variables. The vertical color bar on the right serves as the…

Figure 2. Serial association model of the association between metabolic dysregulation and psychotic symptoms. Bias-corrected bootstrap with 5000 resamples was used to test the significance of association effects. The independent variable (X) was metabolic principal component 1 (PC1, representing metabolic dysregulation), the sequential mediators were Hamilton Depression Scale (HAMD) score (M1) and Hamilton Anxiety Scale (HAMA) score (M2), and the dependent variable (Y) was psychotic depression (PD) status. Association coefficients are labeled on each path. All coefficients reflect associations derived from cross-sectional data. The table below presents the specific indirect associations: the indirect association via depressive symptoms (X→M1→Y) was significant, the indirect association via anxiety symptoms alone (X→M2→Y) was not significant, and the chain-mediated indirect association through both depressive and anxiety symptoms (X→M1→M2→Y) was significant. All coefficients in this figure reflect statistical associations derived from cross-sectional data. The terms “direct association”, “indirect association” and “sequential pattern” are used to describe the statistical structure of the…

Figure 3. K-means clustering for metabolic risk stratification. The scatter plot stratifies patients using two core variables: metabolic indicator levels (represented by principal component 1, PC1) and depressive symptom severity (measured by the Hamilton Depression Scale, HAMD score). The x-axis represents PC1, reflecting the relative levels of metabolic indicators, and the y-axis represents HAMD scores, reflecting depressive symptom severity; each dot represents an individual patient, colored by their assigned cluster. Three risk-stratified groups were identified: low-risk (Cluster 2, blue dots, lower metabolic indicator levels and mild depressive symptoms), moderate-risk (Cluster 1, red dots, intermediate metabolic indicator levels and depressive symptoms), and high-risk (Cluster 3, green dots, higher metabolic indicator levels and severe depressive symptoms), providing distinct patient profiles for subsequent risk-related analyses. HAMD, Hamilton depression rating scale; PC1, principal component 1.

Figure 4. Receiver operating characteristic curve analysis for discrimination of psychotic symptoms. Receiver operating characteristic (ROC) curve analysis was conducted to evaluate the discriminative performance of metabolic indicators for psychotic symptoms. The y-axis denotes sensitivity, representing the true positive rate of correctly identifying cases with psychotic symptoms. The x-axis denotes 1 − specificity, corresponding to the false positive rate. Each colored line corresponds to a distinct marker: the cholesterol-glucose metabolic component (represented by principal component 1, PC1), fasting blood glucose (FBG), total cholesterol (TC), low-density lipoprotein cholesterol (LDLC), and high-density lipoprotein cholesterol (HDLC). The diagonal reference line reflects an area under the curve (AUC) of 0.5, indicating discriminative ability equivalent to random chance. PC1 yielded the highest AUC across all included markers, suggesting relatively stronger discriminative capacity compared with any single metabolic parameter. FBG, fasting blood glucose; HDLC, high-density lipoprotein cholesterol; LDLC, low-density lipoprotein cholesterol; PC1, principal component 1; ROC, receiver…