Predictive Modeling Of Antidepressant Efficacy Based On Cognitive Neuropsychological Theory

Objective: Depression is a common mental disorder with high prevalence and high recurrence rate,and resulted in a heavy disease burden.However,the etiology and neuropathological mechanism of depression are still not fully understood.Drug therapy,as the most commonly used antidepressant treatment,has relatively limited effect.Finding objective markers of antidepressant efficacy and establishing effective clinical prediction models is a possible breakthrough to achieve personalized precision medicine.Most of the previous studies in this field were based on the monoamine hypothesis or the molecular and synaptic plasticity hypothesis of depression,which lacked the integration of individual psychological function and its interaction with social environment.The cognitive neuropsychological hypothesis supplemented this deficiency.In this study,based on the cognitive neuropsychological theory,combined with the cognitive psychological assessment method and resting-state Electroencephalogram(EEG)technology,the cognitive psychological characteristics and EEG objective markers that can predict the efficacy of SSRI(Selective Serotonin Reuptake Inhibitors)antidepressants in the treatment of first-episode depression were screened,and the clinical prediction model of multimodal fusion was constructed.Methods: A total of 69 participants with first-episode major depressive disorder(MDD)and 36 healthy controls(HC)were screened and completed clinical assessment(including depression,anxiety and anhedonia),negative bias measurement(including neuroticism and negative cognitive bias),social environment information collection and resting-state EEG recording.Participants with MDD completed clinical assessment and negative bias measurement again after an 8-week SSRI(Selective Serotonin Reuptake Inhibitor)antidepressants medication,and were classified as Responders(N=39)or Non-Responders(N=30)based on the reduction rate of HAMD(Hamilton depression rating scale).Based on inter-group comparison and correlation analysis using baseline data,cognitive psychological characteristics and objective EEG markers that can predict antidepressant responses were screened,and multi-modal Logistic model and machine learning model were established respectively to test their clinical application value in predicting the efficacy of SSRI antidepressants.Results: There were significant differences in baseline social support,clinical symptoms,negative bias characteristics between MDD patients and HC(p< 0.001),there were also significant differences in alpha asymmetry in the central region(p< 0.05),however,there was no significant difference in theta cordance.At baseline there were no significant differences between MDD＿R and MDD＿NR in social environment information and clinical symptoms,but there were significant differences in negative bias feature and alpha asymmetry: MDD＿R had significantly higher scores on negative bias than MDD＿NR(p< 0.05),the alpha asymmetry of central and central-parietal regions was significantly lower than that of MDD＿NR(p< 0.05).After 8 weeks of medication,both MDD groups showed improvement in clinical symptoms and negative bias(p< 0.05),and the MDD＿R group improved more than the MDD＿NR group(p< 0.05).Correlation analysis showed that at baseline negative bias were positively correlated with depression(p< 0.05),and the improvement of negative bias were also positively correlated with the improvement of depression(p< 0.05).Correlation analysis showed that at baseline there was a significant negative correlation between negative cognitive processing bias and alpha asymmetry in the parietal region(p< 0.05),neuroticism was negatively correlated with alpha asymmetry in central region and central-parietal region(p< 0.05).According to the results of group comparison and correlation analysis,following features were screened to enter the predictive model: self-consciousness factor in neuroticism,negative thoughts bias,alpha asymmetry in partial,central and central-parietal regions,HAMD(Hamilton Depression Rating Scale)factors,HAMA(Hamilton Anxiety Rating Scale)factors,TEPS(The Temporal Experience of Pleasure Scale),family economic level,and social support level.Multi-modal Logistic model and machine learning model were constructed using the features above respectively.Logistic model has poor predictive performance(accuracy 54%,sensitivity 90%,specificity7%).The machine learning model performed well(accuracy 75%,sensitivity 69%,specificity 83%),and the accuracy and specificity reached significant level in the permutation test(p< 0.001).Conclusion:(1)This research analyzed the outcome of 8-week SSRI antidepressant treatment and related factors in first-episode MDD patients,supporting the cognitive neuropsychological hypothesis of depression treatment;(2)The multimodal clinical prediction model constructed based on cognitive neuropsychological theory can effectively predict the efficacy of SSRI antidepressants for 8 weeks in first-episode MDD patients to some extent;(3)The performance of the prediction model constructed by machine learning is better than that of the traditional Logistic regression model;(4)alpha asymmetry indexes in baseline parietal and central regions of MDD patients as objective EEG markers play an important role in the prediction model of clinical efficacy of antidepressants.