2.1 Design and registration

We will conduct a systematic review and meta-analysis of diagnostic test accuracy. The registration number of this protocol is INPLASY2020110076 on the International Platform of Registered Systematic Review and Meta-Analysis Protocols (INPLASY) [15]. We will follow the Preferred Reporting Items for Systematic Reviews and Meta-Analysis for Diagnostic Test Accuracy (PRISMA-DTA) guideline to report this study [16]. No ethical approval is required for meta-analysis.

2.2. Information sources

We will search candidate studies that assessing the accuracy of NAATs for diagnosis of NTM through PubMed, Embase and the Cochrane Library until May 2021. References cited in a review or meta-analysis will also be evaluated to determine additional studies.

2.3. Search strategy

GCY and YQS will designed and implemented the comprehensive search strategies. Use of English in the search process. Search strategy of PubMed will be listed as follows:

1. #1 "Nontuberculous Mycobacteria"[Mesh] OR “Non-Tuberculous Mycobacteria” OR “Atypical Mycobacterium” OR “Nontuberculous Mycobacterium” OR “Mycobacteria, Atypical” OR “Tuberculoid Bacillus” OR “Mycobacterium, Atypical” OR “Atypical Mycobacteria” OR “Mycobacterium terrae” OR “Mycobacterium duvalii” OR “Mycobacterium obuense” OR “Mycobacterium gilvum” OR “Mycobacterium gordonae” OR “Mycobacterium szulgai” OR “Mycobacterium flavescens”
2. #2 "Mycobacterium"[Mesh] OR Mycobacteria
3. #3 #1 OR #2
4. #4 "Nucleic Acid Amplification Techniques"[Mesh] OR "Polymerase Chain Reaction"[Mesh] OR "Real-Time Polymerase Chain Reaction"[Mesh] OR "Reverse Transcriptase Polymerase Chain Reaction"[Mesh] OR "Multiplex Polymerase Chain Reaction"[Mesh] OR "LAMP assay" [Supplementary Concept] OR "loop-mediated isothermal amplification"
5. #5 #3 AND #4

The other two databases will use a similar search strategy.

2.4. Eligibility criteria

2.5. Literature screening and selection

We will use ENDNOTE X9.2 literature management software to manage related literature. The search candidate studies will be imported into the software. Original studies that diagnosed NTM by the reference standard and have assessed the diagnostic accuracy of NAATs for NTM with full text will be included in this systematic review and meta-analysis. Clear reference standard in the original study.

Two independent researchers (GCY and YQS) will evaluate the candidate studies for inclusion based on inclusion criteria. They will review the title and abstract first, then the full text. If there is disagreement between the two researchers, they will consult with a third investigator (LHL).

2.6. Data extraction

We will extract information for each study such as first author; publication date; country; type of study design; patient selection method (such as consecutive and random); TP, FP, FN, and TN values for the index tests; index test type (such as PCR and LAMP); NTM type; specimen types (such as sputum and bronchoalveolar lavage fluid); specimen condition; specimen processing method (e.g., homogenization) and DNA extraction method. As with the literature screening process, the same two researchers will perform the data extraction independently, and if there is any disagreement, it will be handled in the same way as in the literature screening phase.

2.7. Methodological quality assessment

Using the revised tool for Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) [18], two researchers will independently evaluate the risk of bias and applicability of the included studies. The discrepancy will be dealt with in the same way as in the data extraction phase.

2.8. Data synthesis and statistical analysis

TP, FP, FN, and TN values for the index tests in each included study will be obtained first, then the estimated pooled sensitivity and specificity of NAAT associated with the 95% confidence interval (CI) will be calculated using bivariate random-effects models. The heterogeneity will be investigated by visual examination of forest plots [19]. When heterogeneity is observed, subgroup analyses and meta-regression analyses will be used to explore sources of heterogeneity. Subgroup analyses and meta-regression analyses will perform on different types of study design; patient selection methods (such as consecutive and random); index test types (such as PCR and LAMP); NTM types; specimen types (such as sputum and bronchoalveolar lavage fluid); specimen conditions; specimen processing methods (e.g., homogenization) and DNA extraction methods. Sensitivity analyses will be used to check the robustness of analyses. At least 4 eligible studies will be required to perform the meta-analysis and subgroup analyses for predefined variable parameters using Stata version 15.0 (Stata Corp., College Station, TX, USA) with the midas command [20], and meta-regression analyses using meqrlogit command.

2.9 Evidence evaluation

We will evaluate all the strength of the body of evidence according to The Grading of Recommendations Assessment, Development and Evaluation (GRADE) guideline [19]. The quality of evidence will be classified into 4 levels: high, moderate, low, and very low. We will assess the factors that determine and can decrease quality of evidence (such as study design, risk of bias, outcomes, patient populations, diagnostic test, important inconsistency in study results, imprecise evidence and high probability of publication bias) to determine the quality of evidence [21].