In China, tuberculous pleurisy accounts for 49.8 % of extrapulmonary tuberculosis (EPTB) [1]. Pleural effusion occurs in nearly all tuberculous pleurisy patients, and tuberculous pleurisy is thus often termed tuberculous pleural effusion (TPE). Pleural effusion is not a specific sign of tuberculous pleurisy, and it can also be observed in patients with pneumonia, malignancy, and heart failure [2], [3]. Therefore, it is important to differentiate between TPE and non-TPE in patients with undiagnosed pleural effusion. The gold references for TPE are effusion Ziehl-Neelsen (ZN) staining, Mycobacterium tuberculosis (Mtb) culture, nucleic acid amplification tests (NAATs), and pleural biopsy [4]. However, these diagnostic tools have limitations. Although the specificities of ZN staining and Mtb culture are 100 %, their sensitivities are only 3 % and 18 %, respectively [3], [5]. Pleural biopsy has a high diagnostic yield but is invasive, and its accuracy is observer-dependent [6], [7], [8]. NAATs have specificities of nearly 100 %, but their sensitivities are only between 30 % and 50 % [9], [10], [11]. Effusion biomarkers are alternative diagnostic tools for TPE because of their low cost, short turnaround time, and less invasiveness [9]. Pleural adenosine deaminase (ADA) [12], [13], interferon-γ (IFN-γ) [14], interleukin-27 (IL-27) [15], and soluble interleukin-2 receptor (sIL-2R) [16] are the most widely-studied diagnostic biomarkers for TPE, but their sensitivities and specificities are not 100 %. Therefore, developing novel diagnostic biomarkers for TPE is necessary.

The CXC chemokine receptor 3 (CXCR3) ligands include C-X-C motif chemokine ligand 9 (CXCL9), CXCL10, and CXCL11 [17], [18], [19]. They are chemotactic factors for type 1 T helper (Th1) cells [17], [18], [19]. CXCL10, also named interferon-gamma-induced protein of 10 kDa (IP-10), has been proven to be a useful diagnostic marker for TPE [20]. However, only one study has evaluated the diagnostic value of pleural CXCL9 and CXCL11 [21]. In addition, the factors affecting their diagnostic accuracy for TPE, the cellular origin, and the biological function of CXCL9 and CXCL11, remain unknown.

In this prospective study, we evaluated the diagnostic accuracy of CXCL9 and CXCL11 for TPE in two cohorts. We also investigated the factors affecting their diagnostic accuracy, and their cell origins and biological functions. We reported this work following the Standards for Reporting of Diagnostic Accuracy (STARD) statement [22].