Pleural Fluid Soluble Interleukin-2 Receptor as a Biomarker for the Diagnosis of Tuberculosis Pleural Effusion: A Systematic Review and Meta-Analysis

Background Several studies have assessed the diagnostic accuracy of pleural fluid soluble interleukin-2 receptor (sIL-2R) for tuberculous pleural effusion (TPE) but with varied results. Therefore, we conducted this systematic review and meta-analysis to evaluate the accuracy of sIL-2R for TPE. Methods PubMed, Ovid, and Web of Science databases were searched from inception to 23 March 2021 to identify eligible studies concerning the diagnostic accuracy of fluid sIL-2R for TPE. The sensitivity and specificity of sIL-2R for TPE were pooled with a bivariate model. We estimated the global diagnostic accuracy of PE sIL-2R with a summary receiver operating characteristic (sROC) curve. The revised Quality Assessment for Diagnostic Accuracy Studies tool (QUADAS-2) was used to assess the quality of eligible studies. Results A total of nine studies with 270 TPEs and 586 non-TPEs were included in the final analysis. The pooled sensitivity and specificity were 0.81 (95% CI: 0.76–0.86) and 0.92 (95% CI: 0.77–0.98), respectively. The area under the sROC curve (AUC) was 0.82 (95% CI: 0.79–0.86). No significant publication bias was observed. Conclusions Pleural fluid sIL-2R is a useful diagnostic marker for TPE. However, the diagnostic accuracies of already available biomarkers such as pleural fluid adenosine deaminase, interferon-γ, and interleukin-27 appear to be superior relative to sIL-2R. Therefore, it might not be preferable to use sIL-2R for diagnosing TPE.


Introduction
Tuberculosis pleural effusion (TPE) is a common cause of pleural effusion (PE) [1], and its diagnosis is a challenge for clinicians. e gold standard for diagnosing TPE is Mycobacterium tuberculosis (Mtb) culture from pleural fluid or pleural tissue on Ziehl-Neelsen staining [2]. However, the Mtb culture is time-and labor-consuming, which did not allow rapid diagnosis. In addition, the sensitivity of Mtb culture is less than 20% [3]. Ziehl-Neelsen staining is a rapid diagnostic tool with high specificity, but its sensitivity is less than 5% [3]. Imaging-guided or thoracoscopic pleural biopsy is an effective diagnostic method with high specificity, but its sensitivity is approximately 70% [4]. In addition, biopsy is an invasive procedure and can be associated with complication [5], and its diagnostic accuracy is greatly affected by the operator and observer. Nucleic acid amplification tests (NAATs) are promising molecular diagnostic tools for TPE [6]. Xpert MTB/RIF is the most widely-used commercial NAAT in clinical practice with high specificity. However, its sensitivity is approximately 50% [7]. e low sensitivity may be due to the lack of Mtb in pleural fluid [7,8]. Soluble pleural fluid biomarkers represent another promising diagnostic tool for TPE diagnosis. ere are some pleural fluid biomarkers such as adenosine deaminase (ADA), interferon-gamma (INF-c), and interleukin-27 (IL-27) [9]. However, meta-analysis evidence shows that their sensitivities and specificities are around 0.90 [10][11][12]. erefore, it remains essential to develop novel pleural fluid diagnostic markers for rapid diagnosis of TPE.
Interleukin-2 (IL-2) is a growth factor that promotes T lymphocyte proliferation in an autocrine manner [13]. Its effect on helper T-cells is mainly mediated by binding to an IL-2 receptor protein (IL-2R). Some studies have shown that IL-2R can be released from the cell surface in a soluble form, termed soluble IL-2 receptor (sIL-2R) [14]. Helper T-cells are a dominant cell population in TPE [15]. ese cells can be recruited into the pleural space and involved in the protective immunity against Mtb [15,16]. Several studies have revealed that pleural fluid sIL-2R could be used for the diagnostics of TPE, but the results varied. Herein, we conducted a systematic review and meta-analysis to evaluate the accuracy of sIL-2R for the diagnosis of TPE.

Search Strategy.
We searched the PubMed, Ovid, and Web of Science databases to identify eligible studies since inception to 23 March 2021. e search algorithm in the PubMed database was ("Tuberculosis, Pleural" (mesh) or "tuberculosis pleural effusion" or "tuberculosis pleurisy" or "tuberculosis pleuritis" or "tuberculous pleurisy" or "tuberculous pleuritis" or "tuberculous pleural effusion" or "pleural effusion * " or "pleural fluid * ") and ("Receptors, Interleukin-2" (mesh) or "soluble interleukin-2 receptor * " or "soluble interleukin-2 receptor * " OR sIL-2R OR "Receptors, Interleukin-2" (nm)). We used a similar strategy in searching the Web of Science database. In addition, all references listed in eligible studies were also manually searched.
is manuscript is reported by following the PRISMA-DTA (Preferred Reporting Items for a Systematic Review and Meta-Analysis of Diagnostic Test Accuracy Studies) guidelines, which is recommended by the Cochrane Collaboration [17].

Study Selection.
All retrieved studies were imported into Endnote to remove duplicate publications. In addition, the studies investigating the diagnostic accuracy of sIL-2R for TPE were included. e exclusion criteria were (i) animal studies; (ii) non-English studies; (iii) conference abstracts, literature review, editorials, and commentary; (iv) studies without sensitivity and specificity data; and (v) if a two-bytwo table could not be constructed for meta-analysis. All retrieved studies were independently screened by two reviewers, and any discrepancies were resolved by consensus and full-text reviewing.
ere is no restriction regarding country, patients age, race, gender, and publication date.

Data Extraction and Quality
Assessment. Data extraction was performed by two reviewers independently. e data extracted were name of the first author, country, publication year, components of control, sample sizes of TPE and non-TPE, type of data collection (prospective or retrospective), sIL-2R assay, the reference standard for TPE diagnosis, sensitivity, specificity, area under the receiver operating characteristic (ROC) curve (AUC), and the corresponding cutoff. We constructed a two-by-two table for each eligible study with the sensitivity, specificity, and sample sizes of TPE and non-TPE. e two-by-two table contained the numbers of true positive (TP), false positive (FP), false negative (FN), and true negative (TN). e revised Quality Assessment for Diagnostic Accuracy Studies tool (QUADAS-2) was used to assess the quality of eligible studies [18].

Statistical Analysis.
We used a bivariate model to pool the sensitivity and specificity [19]. Summary ROC (sROC) curve, a global metric of a diagnostic test's accuracy, was constructed to estimate the diagnostic accuracy of sIL-2R [20]. e inconsistency index (I 2 ) was used to detect potential heterogeneity across eligible studies [21]. e Deeks's test and funnel plot were used to estimate the degree of publication bias [22]. e Stata 13.0 (Stata Corp LP, College Station, TX, USA) with the midas command was used for all statistical analyses [23].

Study Selection Process and Summary of the Eligible
Studies. Figure 1 is a flowchart of study selection. Finally, twelve studies met the inclusion criteria, but three did not report sensitivity and specificity [24][25][26].

Diagnostic Accuracy.
e diagnostic accuracy of sIL-2R in eligible studies is summarized in Table 3. Only three studies reported the ROC curve. In three studies, the AUC of sIL-2R was 0.57 [28], 0.80 [29], and 0.96 [31], respectively. Diagnostic cut-off adopted by the eligible studies ranged from 2980 U/mL to 5000 U/mL. e eligible studies' sensitivity ranged from 0.62 to 0.91, and specificity ranged from 0.41 to 1.00. e I 2 for sensitivity and specificity was 64% and 97.5%, respectively.      Table 4 lists the positive and negative predictive values (PPVs and NPVs) of sIL-2R for TPE with an assumed prevalence of TPE in the target population. In the target population with the prevalence of TPE <5% in undiagnosed PE, the NPV of sIL-2R was 98.92%. While in the target population with the TPE prevalence >50%, the PPV of sIL-2R was 95.94%.

Journal of Tropical Medicine
Because significant heterogeneity was observed in metaanalysis, we analyzed the possible source of heterogeneity among included studies. We hypothesized that sample size (>100 vs. <100), data collection (prospective design vs. retrospective design or unknown), whether the participants were consecutively enrolled (consecutive vs. unknown and nonconsecutive), whether treatment response was used to diagnose TPE (yes vs. no) are the potential sources of heterogeneity. Table 5 lists the results of meta-regression. Although these design characteristics are possible sources of heterogeneity for sensitivity, they lost significance in the joint model. erefore, the current evidence does not support these design characteristics as sources of heterogeneity. e sROC curve of PE sIL-2R is shown in Figure 3. e area under the sROC curve was 0.82 (95% CI: 0.79-0.86).

Publication Bias.
e funnel plot indicated that publication bias was not statistically significant (P � 0.35, Figure 4).

Discussion
is is the first systematic review and meta-analysis estimating the diagnostic accuracy of PE sIL-2R for TPE to the best of our knowledge. We included nine studies with 270 TPEs and 586 non-TPEs and found that the sensitivity and specificity of PE sIL-2R were 0.81 and 0.92, respectively. e sROC's AUC of PE sIL-2R was 0.82. In addition, no significant publication bias was observed. ese results indicate that pleural fluid sIL-2R has a relatively high diagnostic accuracy for TPE.
Sensitivity and specificity are two basic metrics of a diagnostic tool [36]. e pooled sensitivity and specificity of PE sIL-2R were 0.81 (95% CI: 0.76-0.86) and 0.92 (95% CI: 0.77-0.98), respectively. ese results mean that 81% of TPE patients had elevated PE sIL-2R and 92% of non-TPE had decreased PE sIL-2R. erefore, approximately 19% TPE patients will be missed, and 8% of non-TPE patients will be misdiagnosed as TPE if PE sIL-R level was used alone.
A weakness of sensitivity and specificity is that they can only reflect the accuracy of diagnosis at a specific cut-off; therefore, both of them are not a global measure of diagnostic accuracy [36,37]. By contrast, the AUC of sROC is cut-off independent and thus is a good measure to evaluate the accuracy of an index test [38]. In the traditional ROC curve, each point represents sensitivity and specificity at a certain cut-off. While in the sROC curve, each point represents the data of a single study; therefore, the AUC of sROC represents a globe measure of diagnostic accuracy [39]. e AUC of such curve ranged from 0.5 to 1.0, and a higher value means higher accuracy [20]. We found that the AUC of the sROC curve was 0.82 (95% CI: 0.79-0.86), indicating that PE sIL-2R has moderate diagnostic accuracy for TPE.
NLR and PLR are two meaningful measures used for ruling in or out target disease. PLR >10 or NLR <0.1 suggests high accuracy for ruling out or ruling in target disease [40]. e PLR of PE sIL-2R was 10.5 (95% CI: 3.2-34.3), indicating that patients with TPE have about ten times higher chance of a positive PE sIL-2R compared to non-TPE patients. On the other hand, the NLR for PE sIL-2R was 0.20 (95% CI: 0.16-0.27), indicating that non-TPE patients have an approximately five times higher chance of negative PE sIL-2R than TPE patients. Considering that the post-test probability (predictive value) of the index test was affected by both likelihood ratio and the prevalence of target disease, we summarized PPV and NPV of PE sIL-2R at different prevalence. Notably, the NPV of sIL-2R was 98.92% when the prevalence of TPE is 5%, indicating that the patients in low TPE prevalence areas with negative sIL-2R have an extremely low probability of TPE. erefore, negative PE sIL-2R can be used for ruling TPE in low TPE prevalence areas. While in high TPE prevalence areas, the PPV is only 91.01%. In our opinion, positive sIL-2R cannot be used for confirming TPE under such condition.
In this study, QUADAS-2 was used to evaluate the quality of the included studies. We observed bias in patient selection and index test. Some of the eligible studies did not report whether the participants were consecutively enrolled and did not avoid inappropriate exclusion, which may impair the representativeness of the subjects in eligible studies. Most studies did not use a prespecified cut-off to define positive results, which may overestimate the diagnostic accuracy of PE sIL-2R [18]. Future studies are needed to rigorously evaluate the diagnostic accuracy of sIL-2R.
In addition to pooling sensitivity and specificity, heterogeneity exploration is also an aim of meta-analysis. In this meta-analysis, significant heterogeneity was observed in eligible studies (I 2 � 96%). erefore, we performed a metaregression to explore the possible sources of heterogeneity. We found that the type of data collection, reference for TPE diagnosis, country source of study, and sample size were not the sources of heterogeneity. is result may be due to the fact that the number of eligible studies was small (n � 9). Further studies are needed to address this issue.
To present, many soluble markers in PE have been proposed to diagnose TPE, which have been summarized in our previous review [9]. Generally, ADA [41], IFN-c [11], and IL-27 [42] represent the most promising diagnostic marker. ADA is an immunosuppressor that can prevent excessive inflammatory response by catalyzing the deamination of adenosine [43]. IFN-c and IL-27 are released by T cells or activated antigen-presenting cells in response to Mtb [44,45]. Compared with these markers, the diagnostic accuracy of sIL-2R is inferior. Under such condition, the clinical implication of PE sIL-2R is limited. In our opinion, it only can be used in the case where ADA, interferon-c, and interleukin-27 cannot be tested. e present systematic review and meta-analysis has some limitations. e major limitation of this study is the number of eligible study is small. In addition, although we Journal of Tropical Medicine   performed a meta-regression to explore the possible sources of heterogeneity but failed to identify the possible sources.

Conclusion
In conclusion, our results reveal that PE sIL-2R seems to be a useful diagnostic marker for TPE. However, given the improper study design, further rigorous studies are needed to evaluate the diagnostic accuracy of pleural fluid sIL-2R for TPE. sIL-2R shows inferior diagnostic accuracy when compared with the most promising diagnostic markers. erefore, PE sIL-2R is not preferred for TPE diagnostics at the current stage.
Data Availability e data supporting the findings of this study are available within the article. Disclosure e funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of Interest
e authors declare no conflicts of interest regarding the publication of this paper.

Study
Regression Line