Predictive comparison of specific depression symptoms for peripheral inflammation

Peripheral inflammation has been associated with depression but may not map uniformly onto all depression symptoms and rather be associated with specific neurovegetative symptoms. The current work aims to extend previous research by applying benchmark comparison in a causal random forest framework, which allows to statistically compare depression sum scores with specific symptoms in their relation to inflammation. The analysis was run relying on data of 311,474 individuals from the UK Biobank. Circulating C-reactive protein, white blood cell counts, the neutrophil-to-lymphocyte ratio and the systemic immune inflammation index were used as markers of peripheral inflammation. Depressive symptoms were assessed with a self-report questionnaire. Across multiple analytic settings, we found lethargy to perform significantly better in predicting peripheral inflammation levels compared to depression sum scores. These findings point to a specific phenotype of depression in which fatigue-related symptoms are most closely tied to peripheral inflammation, underscoring the value of symptom-level approaches for biomarker discovery and precision intervention. However, the performance metrics for all models were very low (e.g., R2 < 1-2 %), making the practical significance of these differences difficult to interpret.