2.4.1. Type of study.

Any types of studies can be included, such as retrospective studies, prospective studies, case-control studies. We will include original researches with full text that assessed the diagnostic accuracy of NAATs for abdominal tuberculosis. The reference standard should be appropriate and precisely defined in the study. True positive (TP), false positive (FP), false negative (FN), and true negative (TN) values are provided directly in the articles or contain necessary data to calculate these values. We will exclude articles reported in languages other than Chinese and English, case reports, studies with a specimen size of less than 10, conference coverages, and studies with abstracts but no full text.

2.4.2. Patients.

We will include studies, which contain patients diagnosed with abdominal tuberculosis through NAATs. We will have no restrictions on age, gender, and nations.

2.4.3. Index tests.

NAATs will be considered as index test.

2.4.4. Reference standards.

Bacteriological confirmation of Mycobacterium tuberculosis (positive culture of Mycobacterium tuberculosis and/or microscopic identification of acid-fast bacilli on stained specimen smear) is reference gold standard.

Composite Reference Standard (CRS). Radiological characteristics (such as tree-in-bud pattern and cavity) and histopathological features of the suspected tissue specimen (features of chronic granulomatous inflammation with caseous necrosis/ caseating granuloma). Positive of reference standard test and/or positive of all CRSs mentioned will be considered abdominal tuberculosis. If all factors are negative, it will be considered as non- abdominal tuberculosis.

2.4.5. Literature screening and selection.

Primary search results matching the search strategy will be imported into the ENDNOTE X9.2 literature management software. Two investigators (Yanqin Shen and Likui Fang) will screen candidate studies independently by reviewing the titles and abstracts followed by the full text. Disagreements between two the researchers will be resolved by discussion with a third researcher (Guocan Yu).

2.4.6. Data extraction.

Name of first author; year of publication; country of study; reference standard; TP, FP, FN, and TN values of the test; method of patient selection; test method; subtypes of abdominal tuberculosis (such as omental and visceral tuberculosis), type of specimen; specimen processing procedures (e.g., homogenization) and specimen condition along with other parameters will be extracted. The same two researchers will independently extract relevant data from each included study and cross-check their respective information. Disagreements between two the researchers will be resolved by discussion with a third researcher, similar to that used during the literature selection phase. In the same study, data against different reference standards will be treated separately.

2.4.7. Quality evaluation.

The two researchers will assess the quality of the relevant literature using a revised tool for Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) [12] separately for the different reference standards and the disagreements between researchers will be solved by discussion with a third researchers (Guocan Yu).

According to the PRISMA-DTA statement, systematic review and meta-analysis of diagnostic test accuracy studies was not required to assess publication bias.

2.4.8. Data synthesis and statistical analysis.

TP, FP, FN, and TN values will be obtained from each included study, and the estimated pooled sensitivity and specificity of NAAT for abdominal tuberculosis associated with the 95% confidence interval (CI) will be calculated against culture or CRS, using bivariate random-effects models. We will generate forest plots for sensitivity and specificity for each study and calculate the areas under summary receiver operating characteristic (SROC) curves (AUC). We will assess heterogeneity between studies using I² statistics. An I² value of 0% will be indicative of no heterogeneity, while a value greater than 50% will indicate significant heterogeneity [13]. If the necessary data are available, subgroup analyses will be done to evaluate the diagnostic accuracy of NAATs for abdominal tuberculosis, such as different test methods, subtypes of abdominal tuberculosis, types of specimen, methods of patient selection (consecutive or convenience), methods of decontamination (with or without N-acetyl-L-cysteine/sodium hydroxide), conditions of sample (frozen or fresh), homogenization methods (mechanical or otherwise). If the heterogeneity is obvious, meta-regression analyses and sensitivity analysis will be used to explore the source of heterogeneity. The meta-analysis for predefined variable types will be performed using at least four published studies. We will analyze the data from studies against CRS and culture separately. We will use Stata version 15.0 (Stata Corp., College Station, TX, USA) with the midas command packages to generate forest plots of sensitivity and specificity with 95% CI and carry out meta-analyses and meta-regression analyses.