Exploration Of Diagnostic Models Of Kawasaki Disease And The Function Of Gut Microbiome ’Fingerprint’

Kawasaki disease(KD)is a systemic vascular inflammatory lesion characterized by acute fever.Accurate early diagnosis and timely intravenous injection of globulin can effectively prevent serious complications such as coronary aneurysms.According to existing guidelines,incomplete Kawasaki disease(iKD)is difficult to differentiate from other febrile diseases due to its atypical clinical manifestations,resulting in delayed treatment.Therefore,it is a crucial problem to improve the diagnostic criteria and evaluation system of iKD.At present,with the development of data mining and processing methods,decision tree,Logistic regression and other models are increasingly used in the establishment of disease diagnostic models.It is worth discussing how to choose an appropriate iKD diagnosis model.In addition,the difference in pathogenesis between iKD and complete Kawasaki disease(c KD)is still unclear.The change of gut microbiome,as an important factor of KD,is involved in the occurrence and outcome of iKD and c KD.The role and function of gut microbiome KD remains to be explored and studied.Objective:(1)To establish the decision tree and Logistic regression diagnostic models based on clinical characterizes of KD patients;(2)To verify and evaluate the effectiveness of decision tree and Logistic regression model in iKD patients;(3)To analyze the characteristic spectrum of gut microbiome in c KD and iKD patients,and further explore its role and function.Methods: A retrospective study was conducted to collect the clinical data of 200 hospitalized children in Shanghai Children’s Medical Center from December 2018 to December 2019,who met the inclusion criteria,with c KD group(51 cases),iKD group(54 cases)and infectious fever group(IF)(95 cases).The training set was composed of the clinical examinations of all c KD patients and half of the patients with IF.The validation set was constituted with all IKD patients and the remaining of IF patients.The decision tree algorithm(Classification and regressionn tree,CART)and Logistic regression analysis were conducted to generate the clinical diagnostic models of c KD,which were evaluated in the verification set.In addition,fecal samples of 11 c KD and6 iKD patients were detected by metagenomic sequencing,and the clinical diagnostic function was further evaluated.Results:(1)A total of 24 clinical examined indicators were included in this study to construct the Logistic regression and decision tree models.In the training sets of c KD and IF,a Logistic regression model was established with platelets(PLT),erythrocyte sedimentation rate(ESR)and CD3-/CD19+%.ESR,n-terminal atrial natriuretic peptide precursors(nt-probnp),CD3-/CD19+% and the number of neutrophils granulocyte(GRA)were used to build a decision tree model.(2)In the validation set,it was found that the sensitivity of the Logistic regression and the decision tree model was 0.917 and 0.947,the specificity was 0.950 and 0.963,and the area under the ROC curve was 0.974 and 0.959,respectively.(3)Through metagenomic sequencing of gut microbiome,it was found that the intestinal microbial structures of patients with c KD and iKD were significantly different.The most distinct microbial species in c KD and iKD were Solobacterium＿moorei,Anaerostipes＿hadrus,Enterococcus＿faecalis,Klebsiella＿pneumoniae,Parabacteroides and Streptococcus＿thermophilus,which may play an important role in the occurrence and transformation of iKD and c KD through inflammatory pathway.Conclusion: The decision tree and Logistic regression diagnostic models were established based on multiple examined indictors of c KD,which are effective differential diagnostic methods for iKD.Moreover,as a unique microbial fingerprint,the characteristic changes of gut microbiome not only can reveal the differences in the mechanism of c KD and iKD,but also provide a theoretical basis for the excavation of new biomarkers of iKD.