LAUSR

BACKGROUND Although several pharmacological treatment options for depression are currently available, a large proportion of patients still do not achieve a complete remission or respond adequately to the initial antidepressant prescribed for reasons that remain relatively unknown. This study explored the application of serum biomarkers to the predict the efficacy of escitalopram for treating depression, to guide clinical drug selection. METHOD In this study, 306 patients suffering from depression were treated with escitalopram (10 mg) for 6 weeks. After 6 weeks of treatment, the patients were divided into an escitalopram-sensitive group (ES, n = 172) and an escitalopram-insensitive group (EIS, n = 134) according their HAMD-24 scores after 6 weeks of treatment. Serum samples from all participants were collected on the first day, and 10 different serum biomarkers were analysed. Data from 100 patients in the ES group and 100 patients in the EIS group were then used to build a logistic regression model, and a receiver operating characteristic (ROC) curve was drawn. To validate the accuracy of our model, another 72 patients in the ES group and 34 patients in the EIS group were studied. RESULTS Of the 10 selected serum biomarkers, 4 were screened to build the regression model. BDNF, FGF-2, TNF-α and 5-HT. The regression equation was Z = 1/[1 + e-(-5.065+0.145 (BDNF)+0.029 (FGF-2)-0.368 (TNF-α)+0.813 (5-HT))], and the 4 biomarkers-combined detection achieved an AUC (area under the ROC curve) of 0.929 and a predictive accuracy of 88.70%. LIMITATION Decision support tools based on our combined biomarker prediction models hold comparatively great promises; however, they need to be validated on a much larger scales than current studies provide. CONCLUSION The logistic regression model and ROC curves based of the serum biomarkers used in this study provide a more reliable means to predict the efficacy of escitalopram in patients with depression, and provide clinical evidence for drug selection.