Differential Diagnosis Models Of Multiple Myeloma With Renal Injury And Chronic Kidney Disease Or Nephrotic Syndrome

ObjectiveAs one of the most common malignant tumors in the blood system,the incidence of multiple myeloma has been increasing in recent years.When kidney injury occurs in patients,most show decreased glomerular filtration rate,which may lead to the initial diagnosis in the department of nephrology and being misdiagnosed as kidney disease.In this study,differential diagnosis models of multiple myeloma with renal injury and chronic kidney disease or nephrotic syndrome were established based on clinical information,so as to find more effective models and explore the difference indexes,finally providing the methodological basis for the accurate diagnosis and improving the understanding of the differences in the examination indicators between the diseases.MethodsThe study subjects were from a hospital in zhengzhou in 2019,among which 77 cases with multiple myeloma with renal injury from hematology department were recruited asthe case group,and additional 112 cases with kidney disease patients from nephrology department,including 30 cases with chronic kidney disease and 82 cases with nephrotic syndrome,were recruited as the control group.Meanwhile,the clinical information of the subjects,including demographic characteristics,initial clinical symptoms,serum biochemical indexes and immunological indexes,were collected,and the subjects were randomly divided into training sets and testing sets at a ratio of 3:1.First,the collected clinical information was analyzed by univariate analysis to compare the differences between the two groups and to determine the variables for modelling.Second,based on the training sets,the support vector machines(SVM),decision tree(DT)and artificial neural networks(ANN)were established with the significant variables of univariate analysis as the input variables and the group as the output variables,respectively.Besides,based on the testing sets,the prediction effect of the models was evaluated by accuracy and area under receiver operating characteristic curve(AUC).Finally,the subgroup analysis and the sensitivity analysis of the model were carried out by using the optimal model.Statistical analysis were conducted by IBM SPSS Statistics 21.0,SPSS Clementine 12.0 and MedCaLc V11.6.0.0 software.All statistical tests were two-sided,and the level of statistical signficance was set at?=0.05.Results1.Univariate analysis was performed on the clinical information from the two groups.For demographic characteristics,the differences of age and hypertension history were statistically significant between the two groups(P<0.05).For initial cinical symptoms,the differences of proteinuria,anemia,osteodynia and lower extremity edema were statistically significant between the two groups(P<0.05).Among serum biochemical indexes,the differences of creatinine,beta 2 microglobulin,albumin/globulin,lactate dehydrogenase(LDH)and estimated glomerular filtration rate were statistieally significant between the two groups(P<0.05).Among immunological indexes,the differences of the M-protein typing,serum free light chain difference,IgA,IgG and IgM were statistically significant between the two groups(P<0.05).2.Based on the significant variables of univariate analysis,the accuracy of SVM,DT and ANN was 0.843,0.902 and 0.941,and the AUC was 0.822,0.879 and 0.932,respectively.Besides,the prediction effect of the ANN model was better than that of the SVM model,and the difference was statistically significant(P=0.022).3.The subgroup analysis were analyzed by the ANN model.The results showed that the predictive accuracy and AUC of the ANN model between the case group and the control with chronic kidney disease were 0.972 and 0.960,respectively.The prediction accuracy and AUC of the ANN model were 0.962 between the case group and the control with nephrotic syndrome.4.Based on the common important variables of the three models,including lower extremity edema,osteodynia and LDH,the sensitivity analysis of ANN was carried out.The results showed that compared with the initial ANN model,the accuracy and AUC of the ANN model were 0.836 and 0.843 when lower extremity edema was removed,and the prediction effect was significantly decreased(P<0.001).When osteodynia was removed,the accuracy and AUC of the ANN model were 0.847 and 0.836,and the prediction effect was also significantly decreased(P<0.001).When LDH was removed,the prediction effect of the ANN model was significantly decreased(P=0.003),and the accuracy and AUC were 0.873 and 0.869,respectively.Conclusions1.The ANN model has the best prediction effect for the differential diagnosis of multiple myeloma with renal injury and chronic kidney disease or nephrotic syndrome,which provides the methodological basis for the accurate diagnosis of multiple myeloma with renal injury and chronic kidney disease or nephrotic syndrome.2.Lower extremity edema,osteodynia and LDH are more important indicators to differentiate multiple myeloma with renal injury and chronic kidney disease or nephrotic syndrome,which improves the understanding of the differences in the examination indicators between the diseases and optimizes the selection of characteristic indexes in differential diagnosis.