Assessment of Performance Current Diagnostic Tools of Drug-Resistant Tuberculosis: A Comprehensive Meta-Analysis

The main target of this research is to determine the best diagnostic tool for MDR TB which is appropriate to use in daily practice in Indonesia. So far, based on Indonesia's TB Guideline, the diagnostic for MDR TB is phenotypic and genotypic method using Xpert MTB/RIF and MTBDRplus. There are various new diagnostic tools, such as Abbott RealTime, Microscopic-Observation Drug Susceptibility (MODS), pncA-Seq, and Wayne test. The specific target for this research including meta-analysis comparison of the sensitivity which provides information about how well the test can catch true positive cases and specificity which provides information about how well the test can exclude individuals who are actually negative; AUC values; PPV, NPV, PLR, NLR, DOR, and Accuracy diagnostic tools for Rifampicin resistant TB, Isoniazid, Pyrazinamide, and multidrug resistant TB. MTBDRplus superior than Xpert MTB/RIF with sensitivity 97% (95% CI: 96-98, I2: 66,92%) and specificity 97% (95% CI: 96-98, I2: 79,06%); Superior than Abott RealTime with sensitivity 94% (95% CI: 92-96, I2: 82,18%) and specificity 97% (95% CI: 95-99, I2: 93,00%); and excellent to detect MDR TB with sensitivity 96% (95% CI: 93-98, I2: 62,02%) and specificity 98% (95% CI: 95-99, I2: 95,75%). The best Pyrazinamide resistance tool is pncA-Seq with sensitivity 94% (95% CI: 79-98, I2: 95,65%) and specificity 99% (95% CI: 90-100, I2: 87,50%). The best diagnostic tool applied in Indonesia is Xpert MTB/RIF for Rifampicin (RIF), MTBDRplus for Isoniazid (INH), and MGIT for Pyrazinamide (PZA).

本研究的核心目标是确定适用于印度尼西亚日常临床实践的最佳耐多药结核病（Multidrug Resistant Tuberculosis, MDR TB）诊断工具。截至目前，依据印度尼西亚结核病诊疗指南，耐多药结核病的诊断采用表型与基因分型检测方法，具体使用Xpert MTB/RIF与MTBDRplus。当前已有多款新型诊断工具问世，包括Abbott RealTime、显微镜观察药物敏感性试验（Microscopic-Observation Drug Susceptibility, MODS）、pncA-Seq以及Wayne试验。本研究的具体评估内容涵盖：针对利福平耐药结核病、异烟肼（Isoniazid, INH）耐药、吡嗪酰胺（Pyrazinamide, PZA）耐药及耐多药结核病的各类诊断工具，开展荟萃分析比较其敏感性（sensitivity，反映检测方法检出真阳性病例的能力）、特异性（specificity，反映检测方法排除真阴性个体的能力）、受试者工作特征曲线下面积（AUC）值、阳性预测值（Positive Predictive Value, PPV）、阴性预测值（Negative Predictive Value, NPV）、阳性似然比（Positive Likelihood Ratio, PLR）、阴性似然比（Negative Likelihood Ratio, NLR）、诊断比值比（Diagnostic Odds Ratio, DOR）及诊断准确率。研究结果显示，MTBDRplus的性能优于Xpert MTB/RIF，其敏感性为97%（95%置信区间：96-98，I²=66.92%），特异性为97%（95%置信区间：96-98，I²=79.06%）；同时亦优于Abbott RealTime，敏感性为94%（95%置信区间：92-96，I²=82.18%），特异性为97%（95%置信区间：95-99，I²=93.00%）；且在检测耐多药结核病方面表现优异，敏感性为96%（95%置信区间：93-98，I²=62.02%），特异性为98%（95%置信区间：95-99，I²=95.75%）。针对吡嗪酰胺耐药性检测的最优工具为pncA-Seq，其敏感性为94%（95%置信区间：79-98，I²=95.65%），特异性为99%（95%置信区间：90-100，I²=87.50%）。最终适用于印度尼西亚临床的最佳诊断方案为：利福平（RIF）检测选用Xpert MTB/RIF，异烟肼（INH）检测选用MTBDRplus，吡嗪酰胺（PZA）检测选用MGIT。