A study on electroretinography as a biomarker for seasonal vulnerability in depression

Depressive disorders involve disruptions in light signal processing. Seasonal Affective Disorder (SAD) has been linked to abnormalities in phototransduction, including impaired retinal responses and sleep disturbances. Similar retinal anomalies have been observed in Major Depressive Episode (MDE) patients without seasonal pattern. However, no study has directly compared light-signaling biomarkers between SAD and non-seasonal MDE using sleep assessments and electroretinography (ERG) measures. This study aims to develop a model combining clinical and ERG markers to predict seasonality in MDE patients. Patients with MDE (N = 320) were classified based on their vulnerability to seasonality using the Global Seasonality Score (GSS) from the Seasonal Pattern Assessment Questionnaire (SPAQ), with a threshold of ≥ 11 indicating seasonal vulnerability. This dimensional approach provides a more nuanced reflection of underlying pathophysiological mechanisms than the categorical DSM-5-TR classification of SAD. Subjective sleep, psychiatric, and ERG biomarkers were analyzed. Significant variables were entered into a Backward Stepwise Logistic Regression (N = 35, subset of participants who had all available data including ERG measures), and model performance was assessed using sensitivity, specificity, accuracy, AUC, ROC curve, and Youden's index. The model retained six predictors: reduced bipolar cell amplitude (rods), increased cone response amplitude at 7 cd·s⁻¹·m⁻², increased daytime sleepiness, higher depression severity, younger age, and female gender. It demonstrated good discriminative power (AUC = 0.861, sensitivity = 0.905, specificity = 0.714, Youden's index = 0.619). The model effectively distinguishes seasonal from non-seasonal MDE, explaining 30.5% of the variance. ERG is a promising tool for identifying biomarkers of Seasonal Vulnerability in depression. Enhancing predictive approaches with multimodal diagnostic and longitudinal data could further improve early detection and personalized interventions.