Peripheral Inflammatory Indices in Major Depressive Disorder: Clinical Associations and Explainable Machine Learning
Year: 2026
Type: article
Study design: Observational Study
Abstract: Background: The clinical relevance of routinely available peripheral inflammatory indices in major depressive disorder (MDD) remains uncertain, particularly regarding recent suicide attempts, depressive symptom severity, and their discriminative performance. This study investigated these associations using conventional statistical analyses and explainable machine learning. Methods: This retrospective, single-center study included 86 patients with MDD and 80 healthy controls. Six blood-derived inflammatory indices were examined for between-group differences and associations with recent suicide attempts and depressive symptom severity. Individual-marker discrimination was assessed using receiver operating characteristic analysis, while three ensemble machine-learning models were evaluated using nested cross-validation and compared with MHR alone. Explainable artificial intelligence methods were applied to interpret model predictions. Results: Patients with MDD showed higher MHR, NHR, SIRI, and AISI values than controls. Recent suicide attempts were associated with higher SII, SIRI, and AISI, and these associations persisted after adjustment for age and sex; lower LHR was observed only in the unadjusted comparison and was not significant after adjustment. No inflammatory index was significantly associated with depressive symptom severity. MHR demonstrated the strongest individual discrimination. Machine-learning models showed modest discriminative performance without clear improvement over MHR alone, and feature-importance analyses consistently highlighted MHR. Conclusions: Here we show that peripheral inflammatory indices exhibit distinct observational associations with MDD and recent suicidal behavior. However, their diagnostic and predictive clinical utility remains unestablished and requires prospective, adequately controlled, independently validated studies.
Source: Brain Sciences
Authors: Figen Ünal Demir, Burak Tasci(Data curation, Methodology, Supervision, Visualization, Writing – original draft, Writing – review & editing), Sare Aydin, Yahya Mert Kocasaraç, İlker Güneysu, Burcu Eser, Gökhan Yılmaz(Data curation, Formal analysis, Resources, Software, Validation, Visualization), Filiz Özsoy, Esma Akpinar Aslan
Institutions: Tokat Gaziosmanpaşa Üniversitesi, Fırat University
Language: English
Subfield: Biological Psychiatry
Field: Neuroscience
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $2,654
OpenAlex ID: W7236121664
