The Relationship Between Bipolar Disorder And Oxidative Stress Injury

Background: Early identification of bipolar disorder(BD)between major depression disorder(MDD)is difficult because no tools exist to objectively estimate the risk of bipolar disorder(BD)using routine clinical data.The aim of this study was to develop and validate a biochemical data nomogram using oxidative stress for prediction of the risk of BD with first admission BD or MDD patients through large and complex clinical routine biochemical data.Patients and Methods: Data on 1,252 BD and 1,359 MDD patients admitted to the Shanghai Mental Health Center from January 2009 to December 2018 were used in the analysis.Data on 30 variables from a randomly selected subsample of 1,827(70%)patients were used to construct a predictive model.These variables included age,sex,and measures of oxidative stress(uric acid,bilirubin,albumin,prealbumin),sex hormones,immune function,thyroid function,liver function,and glycolipid and fat metabolism.Univariate analyses identified 24 variables that discriminated BD and MDD at the p<0.25 level.Using these variables,the Least Absolute Shrinkage and Selection Operator(LASSO)machine learning method was used for data dimension reduction and variable selection.Eleven identified variables were subsequently included in the multivariable logistic regression analysis used to develop the predictive model.The 10 statistically significant variables(p<0.05)retained in logistic regression analysis were presented as a nomogram for use in clinical settings and the calibration,discrimination,and clinical usefulness of this nomogram were assessed.Internal validation of these results was assessed using the remaining 784 patients(30%)in the original sample and independent external validation was assessed using data from 3,797 patients meeting similar inclusion criteria from five other centers in China’s mainland.Results: The 10 predictors contained in the individualized prediction nomogram included three of the four oxidative stress features considered(direct-bilirubin,uric acid and prealbumin).The model showed good discrimination in the training sample: AUC of 75.1%(95% CI,72.9-77.3%)sensitivity of 0.66 and specificity of 0.73.The discrimination was also good in the internal validation cohort(AUC=72.1%,68.6-75.6%)and in the external validation cohort(AUC=65.7%,63.9-67.5%).Decision curve analysis demonstrated that the biochemical data nomogram had good clinical application capability.Conclusion: This study used a machine learning approach to present a biochemical data nomogram that centered on oxidative stress(BIOS MODEL),which can use clinical routine biochemical data to give individualized risk prediction of BD with first admission BD or MDD patients and verified the possible oxidative injury in BD.For better use this tool in the real-world clinical practice,replication is needed in external first admission patients that initial diagnosis is BD or MDD with overlapping variables.