ABSTRACT
The present study aims to assess the treatment outcome of patients with diabetes and tuberculosis (TB-DM) at an early stage using machine learning (ML) based on electronic medical records (EMRs). A total of 429 patients were included at Chongqing Public Health Medical Center. The random-forest-based Boruta algorithm was employed to select the essential variables, and four models with a fivefold cross-validation scheme were used for modeling and model evaluation. Furthermore, we adopted SHapley additive explanations to interpret results from the tree-based model. 9 features out of 69 candidate features were chosen as predictors. Among these predictors, the type of resistance was the most important feature, followed by activated partial throm-boplastic time (APTT), thrombin time (TT), platelet distribution width (PDW), and prothrombin time (PT). All the models we established performed above an AUC 0.7 with good predictive performance. XGBoost, the optimal performing model, predicts the risk of treatment failure in the test set with an AUC 0.9281. This study suggests that machine learning approach (XGBoost) presented in this study identifies patients with TB-DM at higher risk of treatment failure at an early stage based on EMRs. The application of a convenient and economy EMRs based on machine learning provides new insight into TB-DM treatment strategies in low and middle-income countries.
Subject(s)
Diabetes Mellitus , Humans , Comorbidity , Treatment Failure , Electronic Health Records , Machine LearningABSTRACT
Background: Deficiency vitamin D and hyperglycemia could be related to weakened innate immune response and aggravate the progression of tuberculosis (TB). This study hypothesized that DNA promoter methylation of the pivotal genes in the vitamin D metabolic pathway might be related to diabetes and tuberculosis co-morbidity (TB-DM) susceptibility. Methods: A total of 50 TB-DM and 50 healthy subjects (HS) were included in the present study. Targeted bisulfite sequencing was applied to detect the methylation of the promoter regions of candidate genes in the vitamin D metabolic pathway (CYP24A1, CYP27B1, CYP2R1, DHCR7, and VDR) in whole blood. Results: The overall methylation level of candidate genes in this study was lower in patients with TB-DM than HS, except for CYP2R1. The results of the ROC demonstrated the potential of CYP24A1, CYP27B1, DHCR7, and VDR promoter methylation as a biomarker for diagnosing TB-DM, with all the AUC above 0.7. In subgroup analysis, we found that lower circulating vitamin D is related to a low level of CYP24A1, CYP27B1, and DHCR7 promoter methylation in patients with TB-DM. With decreasing methylation level, risk of TB-DM was significantly increased (odds ratio, 95% CI 0.343, 0.144-0.821 for CYP24A1; 0.461, 0.275-0.773 for CYP27B1; 0.09, 0.015-0.530 for DHCR7; 0.006, 0.0003-0.115 for VDR). Besides, our results revealed that there was a significant correlation between DNA promoter methylation of selected genes in the vitamin D metabolic pathway and platelet indices in TB-DM. However, there was no correlation between DNA methylation of the four genes and fasting glucose and HbA1c. Conclusion: Our results could suggest that the selected genes in the vitamin D metabolic pathway may be involved in the pathological process of TB-DM, but independent of the process of hyperglycemia to impaired immune responses to Mtb.
ABSTRACT
Tracheobronchial tuberculosis (TBTB) is reported in 10-40% of patients with pulmonary tuberculosis (PTB). Due to its non-specific presentation, the diagnosis and management are frequently delayed. The aim of the present study was to investigate the incidence, predictors and laboratory diagnosis of concomitant TBTB and PTB in Chongqing, China. Bronchoscopy was performed in all patients with newly diagnosed or relapsed PTB in order to detect TBTB between January 2018 and April 2019 in a sub-tertiary hospital in Chongqing, China. The clinical characteristics and laboratory data were analyzed to identify predictors and determine the diagnostic yield of TBTB. A total of 341 (31.4%) of the 1,085 patients with PTB who underwent the bronchoscopic examination presented with concomitant TBTB. The parameters of female sex [odds ratio (OR)=2.57], clinical symptoms (OR=6.26) and atelectasis (OR=4.3) were independent predictors of TBTB. Cough (OR=32.48) and atelectasis (OR=3.14) were independent predictors of TBTB-associated tracheobronchial stenosis. The diagnostic yields of sputum smear, bronchial brush smear, sputum culture, GeneXpert Mycobacterium tuberculosis/rifampicin resistance (GX) using sputum, GX using brushings and in bronchial brush culture used for the diagnosis of TBTB were 44.2, 44.2, 63.5, 57.7, 71.2 and 75%, respectively. GX brushings had higher diagnostic yields compared with sputum or brush smears; however, there was no significant difference between sputum/brushings cultures and GX with sputum. The incidence of TBTB in PTB was 31.4% in Chongqing, China. The parameters of female sex, atelectasis and cough were the major predictors of concomitant TBTB and associated tracheobronchial stenosis. Although GX is an accurate and rapid test to detect TBTB, additional laboratory techniques should also be adopted to improve diagnostic yields in the detection of TBTB in patients with PTB.
