Screening Of Clinical And Biological Markers And Establishment Of Diagnostic Predictive Models For Bipolar And Unipolar Depressive Disorders

Objective: To study the clinical and biological differentiators of bipolar and unipolar depressive disorder;Logistic regression model,Fisher discriminant analysis model and neural network prediction model were established to screen the best prediction model for bipolar depressive disorder.Methods: 94 patients with unipolar depressive disorder and 97 patients with bipolar depressive disorder(43 patients with bipolar typeⅠ and 54 patients with bipolar type Ⅱ)hospitalized in Hebei Mental Health Center from April to November 2020 were enrolled in the study,and 79 healthy controls were collected in the same period.Bipolar depressive disorder accorded with the diagnostic criteria of bipolar typeⅠand bipolar typeⅡin American Diagnostic and Statistical Manual of Mental Disorders 5th Edition(DSM-5),while unipolar depressive disorder met the diagnostic criteria of recurrent major depressive disorder in DSM-5.The demographic data of the patients were collected with a self-made general questionnaire,and the clinical symptoms were evaluated with Hamilton Depression Scale(HAMD),Hamilton anxiety scale(HAMA),32-item hypomania checklist(HCL-32)and Dimentional Anhedonia Rating Scale(DARS).The temperament types of patients were evaluated by self-rating temperament scale(TEMPS-A),and borderline personality disorder in SCID-II was selected to diagnose borderline personality.The levels of plasma cortisol,thyroid function,blood glucose,blood lipid and blood homocysteine were detected by chemiluminescence method.The levels of serum inflammatory cytokines and plasma BDNF were measured by ELISA method,and the level of BDNFm RNA was detected by real-time fluorescence quantitative PCR.SPSS22.0 and Medcalc statistical software were used for data analysis,model establishment and efficiency analysis.Result:1.The result shows that the average age of onset in patients with bipolar depressive disorder is 21.64 ± 10.42 years old,which is significantly earlier than that in patients with unipolar depressive disorder(38.79± 14.47 years old).People with bipolar depression eat midnight snacks is higher,the clinical concomitant symptoms show that patients with bipolar depression are more likely to be accompanied by mixed characteristics,atypical characteristics and psychotic characteristics.The rates of self-injury behavior(44.4%)and suicidal behavior(43.3%)are significantly higher than those of unipolar depression(8.5%,22.3%,P<0.001).The bipolar depression patients with borderline personality disorder(43.30%)are significantly higher than unipolar depression patients(11.7%)(χ~2=23.772,P<0.001),have a higher HCL-32 scores(Z=6.088;P<0.001)and a higher total scores of anhedonia(Z=3.121;P=0.002).The patients with bipolar depression have more obvious circulatory temperament,irritating temperament and exuberant temperament(P<0.05).The patients with unipolar depression are more likely to have anxiety traits,depression characteristics,seasonal characteristics,somatic diseases and more times of depression.Average duration of depressive episodes in unipolar depression(4 months)are longer than bipolar depression(3 months)(Z=2.892;P=0.004).HAMD-17 score in unipolar depression is higher,and the level of thyroxine(T4)is higher than that of bipolar depressive disorder(P<0.05).2.Unipolar depressive disorder and bipolar depressive disorder group of proinflammatory cytokines IL-6,IL-1β,TNF-α,IFN-γ are higher than healthy controls(P < 0.05);anti-inflammatory cytokine IL-6,IL-10,IL-13 are lower than those of healthy controls(P < 0.05);in terms of soluble inflammatory cytokine levels,in addition to s TNFR1,the differences between the two groups had no statistical significance(P >0.05),s TNFR2 and s IL-6R in unipolar depressive disorder and bipolar depressive disorder group are higher than healthy controls(P<0.05).Proinflammatory cytokines IL-6,IL-1 β,TNF-α,IFN-γ and anti-inflammatory cytokine IL-6,IL-10,IL-13 between unipolar depressive disorder and bipolar depressive disorder group have no statistically significant difference(P>0.05);as for soluble inflammatory cytokines receptors s TNFR2,there is no statistically significant difference between the two groups;as for s TNFR1 and s IL-6R,there are statistically significant difference between the two groups(P < 0.05).3.There is no significant difference in BDNF between unipolar depression group and healthy control group(P>0.05),but BDNFm RNA is lower than that of healthy control group(P<0.05).There is no significant difference in BDNF and BDNFm RNA between bipolar depression group and healthy control group(P>0.05).BDNF and BDNFm RNA levels in bipolar depression group are higher than those in unipolar depression group(P<0.05).4.Adjustment for age and gender of the single factor logistic regression analysis results show that the night snack habit,nature of work,age at first onset,with characteristics of anxiety,with mixed features,with melancholy,with seasonal characteristics,self-injury behavior,edge line personality,HCL-32,Exuberant temperament,T4,IFN-γ,s TNFR1,s IL-6R,BDNF,BDNFm RNA,17 factors have statistical relationship with the occurrence of bipolar depressive disorder.5.The above variables are screened out into multivariate logistic regression,and analysis shows that early age of first onset,NSSI,low T4,low s IL-6R and high BDNFm RNA are independent risk factors for bipolar depression disorder.Female,anxiety characteristic,depression characteristic and seasonal characteristic are independent risk factors for unipolar depression disorder.The establishment of three different prediction models show that: the total prediction accuracy,specificity,sensitivity and AUC of the Logistic regression model are 90.05%,93.62%,86.60% and 0.952,respectively.The total prediction accuracy of Fisher discriminant analysis model is 88.48%,the specificity is 93.62%,and the sensitivity is 83.51% and the AUC is 0.952.The total prediction accuracy of the neural network model is 91.10%,the specificity is92.55%,and the sensitivity is 89.69% and the AUC is 0.978.The prediction performance of neural network model is higher,and there is significant difference in the area under ROC curve between neural network model,Logistic regression model and Fisher discriminant analysis model.Conclusion:1.Patients with bipolar depression have an early onset age,a lifestyle characterized by more late-night eating habits,and more clinical manifestations accompanied by mixed characteristics,atypical characteristics and psychotic characteristics.More than 40% of the patients are accompanied by self-injury,suicide behavior and borderline personality.2.Patients with bipolar depression have high HCL-32 score,low degree of anhedonia,and more obvious circulatory temperament,irritable temperament,and exuberant temperament.HCL-32,DARS and TEMPS-A scales are helpful for the differential diagnosis of unipolar and bipolar depressive disorder.3.Patients with bipolar depression have lower T4 levels than those with unipolar depression and there is more obvious thyroid dysfunction.4.Inflammatory changes are present in both unipolar and bipolar depressive disorders.Imbalanced levels of pro-inflammatory cytokines and anti-inflammatory cytokines play a pathogenic role,and changes in soluble inflammatory cytokine receptors support the hypothesis of compensated immune regulatory system(CIRS)disorders.s TNFR1 and s IL-6R in bipolar depression are lower than those in unipolar depression.The expression levels of soluble inflammatory cytokine receptors in unipolar and bipolar depression are different,and the disorder of CIRS is more obvious in unipolar depression.5.Both BDNF and BDNFm RNA levels of bipolar depression are higher than those of unipolar depression,which may be biological indicators for differential diagnosis of these two diseases.6.Artificial neural network is used to construct a prediction model for bipolar depression,and the prediction effect is better than Logistic regression model and Fisher discriminant analysis model,providing a good prediction method for the early diagnosis of bipolar disorder.