Construct The Diagnosis Model Of Pulmonary Tuberculosis Based On MicroRNAs And Evaluate Its Diagnosis Value

Background and purposeThe correct and rapid diagnosis of pulmonary tuberculosis(PTB)is crucial to the prevention and treatment of tuberculosis(TB).However,this problem is still plaguing clinical medical workers.The current PTB diagnostic strategy is combined with clinical manifestations,imaging data and laboratory tests,and then decide if it could be diagnosed PTB.At the present,the laboratory diagnostic methods of PTB have limitations,Traditional microbiological methods such as acid fast stain method,culture method is the main means of laboratory tests,but the positive rate is low,or the test is time consuming.Modern molecular biology methods such as Xpert MTB/RIF method,though it improves positive rate of diagnosis and has realized the rapid detection of mycobacterium tuberculosis,but it needs advanced equipment and expensive.In addition,the clinical manifestations and imaging of PTB lack of specificity,and often difficult to identify tuberculosis or other lung diseases such as pneumonia and lung cancer,so the researchers have been looking for a more ideal PTB diagnostic method.The microRNA is closely related to the disease progression,has been as indicators of the diagnosis of diseases such as tumor,and provides a new way for the laboratory diagnosis of PTB.A series of study have indicated that the expression of microRNAs in patients with TB were dysfunctional,and these dysfunctional microRNAs may be used for diagnosis of PTB.However,the existing study is just on the level of screening the level of potential diagnostic indicator,it is necessary to conduct in-depth research and exploration.This study is on the basis of the existing research results,a variety of microRNAs,which have been reported to be used to diagnostic markers of PTB,is selected to test and verify by q RT-PCR.Logistic regression analysis was used to obtain the regression equation,and verified the diagnostic performance of this equation of the independent sample.We aim to construct a model based on miRNAs of PTB diagnosis,in order to effectively supplement the existing PTB diagnostic methods.MethodsHealthy group,pneumonia group and lung cancer group were used as control group,and the patients with pulmonary tuberculosis were acted as experimental group.The experimental design is divided into three parts: the mi RNAs screening phase,the diagnosis model construction phase and the diagnosis model validation phase.1.mi RNAs screening phase: a total of 178 cases were included in the study,including the experimental group(PTB group)70 cases,108 cases of the control group(36 cases in the health group,36 cases of pneumonia group,36 cases of lung cancer group).8 mi RNAs with abnormal expression in PTB patients was selected,and their expression level in 178 cases was tested by q RT-PCR.Data processing and statistical analysis of these results,the mi RNAs in accordance with the screening criteria were selected to enter the follow-up study.2.Diagnosis model construction phase: the mi RNAs meeting the screening criteria were selected for multifactor Logistic regression analysis,using the stepwise method of screening variables to establish regression equation,and using ROC curve to evaluate the diagnostic efficacy of the equation.3.Diagnosis model validation phase: to evaluate the accuracy of the diagnosis model,the independent samples were used to verify the accuracy of the model.The research object in this stage included 200 cases in the experimental group(PTB group)80 cases,120 cases in the control group(healthy group 40 cases,40 cases of pneumonia group,lung cancer group of 40 cases).Result1.mi RNAs screening phase: the screening criteria for mi RNA expression levels are: PTB group respectively compared with pneumonia group,lung cancer group and healthy group,at least one of three couples meet the mi RNA expression level difference had statistically significant,and the relative change fold is ≥2 or ≤0.5.eight mi RNAs(mi R-21,mi R-29 a,mi R-101,mi R-146 a,mir-223,miR-361-5p,mir-378 and mi R-424)were selected,Among them,although the difference in the expression level of miR-29 a was statistically significant,but its change fold was 0.7 and was excluded,the remaining 7 kinds of mi RNAs are in line with the experimental requirements.2.Diagnosis model construction phase: Used forward stepwise method: likelihood ratio method to screen the variable and successfully constructed a diagnostic model based on 3 kinds of mi RNAs(mi R-101,mi R-223 and mi R-424)and the other four mi RNAs were excluded.The ROC curve analysis confirmed that the model has good diagnostic performance for PTB,AUC = 0.92,95% CI: 0.87-0.95;the diagnostic sensitivity was 90.0%,specificity was 84.3%.3.Diagnosis model verification phase: using independent samples,ROC curve analysis confirmed the model of PTB with good diagnostic performance,AUC = 0.89,95% CI: 0.84-0.93 and diagnostic sensitivity was 86.3%,specific of 80.8%.Taking into account that PTB often have to distinguish between pneumonia and lung cancer clinically,we also analyzed the ability of the model to identify the PTB and the above two diseases.The results showed: when distinguish with pneumonia,AUC = 0.91,95% CI: 0.86-0.96,the diagnostic sensitivity was 90.0%,the specificity was 80.0%;when distinguish with pneumonia,AUC = 0.92,95% CI: 0.86-0.96,the diagnostic sensitivity was 88.8%,the specificity was 85.0%.Conclusions1.Based on the literature research,eight kinds of mi RNAs were chosen and detected by the q RT-PCR technique.The expression level difference of 7 mi RNAs(mi R-21,mi R-101,mi R-146 a,mi R-223,mi R-361-5p,mi R-378 and mi R-424)were statistically significant in pulmonary tuberculosis with control groups,so the seven kinds of mi RNAs was used to further study.2.The seven mi RNAs(miR-21,mi R-101,mi R-146 a,mi R-223,mi R-361-5p,miR-378 and mi R-424)were selected by multivariable logistic regression analysis,and successfully established a diagnosis model that based on three kinds of mi RNAs(mi R-101,mi R-223 and mi R-424).Logit(the probability of PTB P)= 6.214-1.480*(mi R-101)-0.002*(mi R-223)-0.051*(miR-424),the diagnostic performance of this model for PTB: AUC = 0.92,95% CI: 0.87 ~ 0.95;the diagnostic sensitivity was 90.0%,specificity was 84.3%.3.Using ROC curve to verify the independent samples(PTB group 80 cases,healthy group 40 cases,pneumonia group 40 cases,lung cancer group 40 cases),it confirmed that this model had a potential vaule for diagnosed PTB: AUC = 0.89,95% CI: 0.84 ~ 0.93;the diagnostic sensitivity was 86.3%,specificity was 80.8%.