A novel risk factor for predicting anti-tuberculosis drug resistance in patients with tuberculosis complicated with type 2 diabetes mellitus
Mengyuan Lyu, Dongmei Wang, Junwei Zhao, Zengwei Yang, Weelic Chong, Zhenzhen Zhao, Liang Ming, Binwu Ying
Abstract
OBJECTIVES: This study aimed to explore the relationship between glycosylated hemoglobin (HbA1c) and the risk of anti-tuberculosis (TB) drug resistance for TB-type 2 diabetes mellitus (T2DM) patients. METHODS: From March 2014 to June 2019, medical records from multiple centers were searched. Logistic regression analyses were performed. A predictive model for multidrug-resistance (MDR) was developed and validated. Calibration and discrimination of the model were assessed. RESULTS: Inconsistent results were found in the systemic review. A multicenter chart review with 657 records was thus conducted. The HbA1c <7% group and HbA1c ≥7% group had 390 and 267 patients, respectively. The HbA1c<7% group had a lower risk of developing rifampicin resistance, isoniazid resistance and MDR, with odd ratios (ORs) of 1.904 (p=0.001), 2.896 (p<0.001) and 3.228 (p<0.001), respectively. The between-group differences in the risk of anti-TB drug resistance were analyzed based on data from three provinces in China. After adding HbA1c grading, the predictive model for MDR (https://mengyuan.shinyapps.io/Shinyapp/) showed excellent capacity with an AUC of 75.4% in the training set (Sichuan and Gansu) and 73.9% in the internal validation set (Henan). The performances in calibration, prediction probabilities and net clinical benefit were significantly improved by HbA1c grading. CONCLUSIONS: HbA1c grading was an independent risk factor for isoniazid resistance and MDR in TB-T2DM patients.