Background
Tuberculous pleural effusion (TPE), one of the most common forms of extrapulmonary TB, has a presentation spectrum from fully absorbed benign to complicated pleural thickening and even serious complications such as empyema and fibrothorax, which may have a lasting effect on lung function. Early and effective diagnosis could minimize hospital days and maximize quality of life. At present, the most direct evidence for
Mycobacterium tuberculosis (MTB) infection is aetiology [
1], which has suboptimal sensitivity. Therefore, patients whose pleural effusion is characterized by lymphocytic exudates [
2] combined with high ADA levels are frequently diagnosed as having TPE by clinicians, resulting in high prevalence.
Adenosine deaminase (ADA), an enzyme produced from lymphocytes that is involved in purine metabolism, has been used in the diagnosis TPE for a long time due to its excellent sensitivity and specificity [
3,
4]. However, recent studies have shown that patients with empyema, malignancy, or rheumatoid pleurisy can also have high ADA levels [
5], and negative results in older TPE patients [
6] and fluctuation were obviously affected by the patient profile and local tuberculosis (TB) prevalence [
7].
The interferon gamma release assay (IGRA) is a commercially available cost-effective assay that detects changes in interferon gamma (IFN-γ) caused by
MTB infection. The WHO guidelines [
8] rejected the recommendation of the IGRA for differentiating active TB, especially in high-burden countries, but the guidelines published by the European Centre for Disease Prevention and Control (ECDC) [
9] proposed that the IGRA could contribute supplementary information for patients who test negative for acid-fast bacilli (AFB) and
MTB culture in sputum. In recent years, an increased number of studies have researched the utility of the IGRA to diagnose TPE. A recent meta-analysis [
10] assessing the performance of the IGRA for diagnosing TPE exhibited satisfactory outcomes (PPV = 82%, NPV = 87%), but the results were heterogeneous (I
2 = 92.0–82.5%), suggesting that the results are still controversial and polarizing. Theoretically, tuberculosis antigen-specific responses by IGRA should provide a discriminatory value superior to non-specific inflammatory biomarkers (e.g., unstimulated IFN-γ or ADA), but this is not the case. We boldly presume that the superiority of the IGRA could be concealed by discrepancies in population characteristics, causing inconsistencies between studies.
Currently, there are two commercial kits for IGRA; one is an enzyme-linked immunosorbent assay (ELISA), which detects the IFN-γ in the whole blood and is represented by the QuantiFERON-TB Gold test [
9], which was approved by the FDA in 2004; the other test is the enzyme-linked immune-spot assay (ELISpot), which detects IFN-γ released by mononuclear cells isolated from whole blood under the stimulation of specific antigen, represented by the T-SPOT assay [
9] developed by Oxford University. Both methods have similar principles, but they are slightly different in detection technology and concrete operation. To further to clarify the diagnostic role of the IGRA for TPE in patients with a high TB burden, this prospective study was conducted to investigate the utility of the IGRA assay (T-SPOT.TB) in the discrimination of TPE and to compare the difference in potency with ADA in TPE subjects with different characteristics.
Methods
Participant population and study procedure
A prospective study was performed at Beijing Chest Hospital, Capital Medical University, from June 2015 to October 2018, in which all suspected pleural effusion (PE) patients were enrolled consecutively. The enrolled patients met the following criteria: (1) age ≥ 14 years; (2) presentation with PE on chest ultrasonic examinations; and (3) tolerated thoracic puncture and had more than 100 ml pleural effusion. The exclusion criteria were as follows: (1) HIV positive; (2) a history of immunodeficiency, autoimmune disease or use of immunosuppressive drugs; and (3) previous anti-tuberculosis treatment for more than 2 weeks.
Clinical samples (including sputum, peripheral blood (PB) and pleural fluid (PF)) from all participants were processed for diagnostic purposes after obtaining written informed consent (P.S. The consent of the participants below age 18 years were obtained from their guardian); tests included routine clinical biochemical testing for each PF sample, which contains total protein, glucose, lactate dehydrogenase (LDH) and ADA, and smear microscopy, culture and Gene-Xpert for each sputum and PF sample. All participants’ clinical data were extracted by the investigators, and the treatment process and discharge diagnoses were tracked. All participants were followed up for 12 months to verify the final diagnosis, and patients with negative outcomes after anti-TB treatment in the last 12 months were deemed indeterminate diagnoses.
Clinical categories of pleurisy
Patients were divided into three groups according to the composite reference standard (CRS), which was composed of clinical, laboratory, and radiological examinations and follow-up data of diagnostic treatment. (1) Bacteriologically confirmed TPE was represented by the isolation of MTB in PE, sputum or pleural tissue by culture, microscopy or Gene-Xpert, or a pleural biopsy that demonstrated caseating granulomas. (2) Probable TPE lacked bacteriological confirmation, but all patients were treated empirically for TB based on clinical suspicion (e.g., typical clinical symptoms, remarkable radiological imaging and positive outcome of anti-TB treatment during follow-up). (3) Non-TPE indicated cases were diagnosed definitively as other diseases, such as malignancy or empyema (non-tuberculous disease).
ADA measurement
ADA activity was determined colorimetrically at 37°C using a commercial kit (Adenosine Deaminase Assay Kit; Beijing Strong Bio-technologies, Beijing, China) according to the peroxidase assay [
11]. One unit of ADA was defined as the amount of enzyme that generated one micromole of inosine from adenosine per minute at 37°C. The results were expressed in international units per litre of PE (IU/L).
T-SPOT.TB in PF and PB
The PB (4 mL) and PF (45 mL) samples collected from each participant were tested within 6 h. The PB samples were diluted 1-fold and centrifuged at 900 g for 20 min, and the PF samples were centrifuged at 500 g for 10 min. The supernatant of both samples was discarded for T-SPOT.TB testing.
The T-SPOT.TB assay was conducted following the manufacturer’s instructions (Oxford Immunotec Ltd., Oxford, UK), which were identical for both the PB and PF samples. The pellets were resuspended in 8 mL of AIM-V medium (GIBCO, Rockville, MD, USA). Briefly, mononuclear cells (MCs) were separated using FicolleHypaque, washed, resuspended, and counted. Empty wells were used as negative controls, the T lymphocyte mitogen phytohemagglutinin was used as a positive control, and the ESAT-6 and CFP-10 peptides were in separate wells. Isolated peripheral blood mononuclear cells (PBMCs) and pleural fluid mononuclear cells (PFMCs) were added to the wells (2.5 × 105 cells per well) that were precoated with a monoclonal antibody against IFN-γ and incubated at 37 °C for 16 to 20 h. The spot-forming cells (SFCs) were read using an automated enzyme-linked immunosorbent spot (ELISPOT) reader (CTL-ImmunoSpotS5 Versa Analyser). A test was considered valid when the positive control > 20 SFCs/106 mononuclear cells and the negative control < 6 SFCs/106 mononuclear cells. The final SFCs of ESAT-6 or CFP-10 were defined as ESAT-6 or CFP-10 SFCs minus negative control SFCs. The Max SFCs of the T-SPOT assay were defined as the larger of the final ESAT-6 and CFP-10 SFCs.
Diagnosis
Smear, acid-fast bacilli (AFB) and mycobacterial culture
Specimens including sputum and PF (5 mL) were prepared for direct smear and stained with auramine and examined by light-emitting diode microscopy. The smear was read and interpreted in accordance with the WHO guidelines [
12]. The sputum and PF (5 mL) were preprocessed using N-acetyl-L-cysteine and sodium hydroxide (NALC-NaOH) and centrifuged, and the supernatant was discarded. The resuspended pellet was transferred to solid Lowenstein-Jensen medium (Encode Medical Engineering Co., Ltd., China) and liquid medium and subjected to culture in a mycobacterial growth indicator tube (MGIT) 960 system (Becton, Dickinson and Company, USA). The presence of the
MTB complex in any medium represented positive MPT64 antigen testing. The positive events and time were recorded.
Gene-Xpert
The Gene-Xpert test was performed according to the manufacturer’s instructions (Cepheid, Sunnyvale, CA, USA). Briefly, the specimen (including sputum and concentrated PF) and sample reagent were fully premixed at room temperature. The final 2 ml mixture was collected and transferred to the cartridge and loaded into the automatic Gene-Xpert instrument. Duplicate testing was performed on samples with an invalid result.
Statistical analysis
Data were analysed using IBM SPSS 25.0 (SPSS Inc. Chicago, IL, USA) GraphPad Prism 8.2.1 (GraphPad Software, Inc. La Jolla, USA). Quantitative variables are presented as the mean ± standard deviation (SD) or median (interquartile range (IQR)), and categorical variables are presented as frequencies (percentages). To identify differences between two independent groups, the chi-square test was used to detect differences between categorical variables, and the Mann-Whitney U test and unpaired t-test were used for continuous data in non-normal or normal distributions, respectively. A result was considered statistically significant when the P-value was < 0.05.
Receiver operating characteristic (ROC) curves were plotted to evaluate the diagnostic performance of ADA and T-SPOT.TB, obtaining the optimal cut-off value and calculating the corresponding areas under the ROC curve (AUCs). The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (+LR), negative likelihood ratio (−LR), diagnostic odds ratio (DOR) and accuracy were calculated.
Predictors that were related to TPE by a predetermined
P-value of 0.10 or less were selected and used in a multivariable logistic regression model (except symptoms). Stepwise backward selection using
P < 0.10 was used to derive the model. Multicollinearity was assessed, and variables contributing to the best fit of the final model, or most related and widely available in our setting, were retained in the final model. For application of the model, a bioclinical score chart was derived using the adjusted OR value of the predictors [
13]. Overall research was completed in keeping with the Standards for Reporting of Diagnostic Accuracy (STARD) template [
14].
Discussion
To the best of our knowledge, although there are several publications that evaluate the utility of the T-SPOT assay for the diagnosis of TPE [
9,
10], the majority of studies recruited a small sample population (
n < 100) for evaluation, and the results were conflicting. Our study provided the largest cohort (TPE:non-TPE = 397:145) to date, confirming the value of the T-SPOT assay for the diagnosis of TPE with high confidence and providing specific reference suggestions for the clinical application of the T-SPOT assay.
In this study, all suspected PE participants had been consecutively, unselectively enrolled, and all TPE patients had a definite diagnosis by bacteriological confirmation or positive outcome for anti-TB therapy, which highly reflects the demographic epidemiological characteristics of tuberculosis-prone areas. The recruited samples ranged from adolescents to elderly patients over 90 years of age, confirming again that the TPE patients had the clinical characteristics of younger males, combined with higher ADA activity and higher T-SPOT response for PF and PB. Comprehensively, the specificity of the T-SPOT assay for PF was much higher with approximate sensitivity than that for PB, which is similar to other studies [
17]. Moreover, different from the low incidence areas where the cut-off of PF T-SPOT was equal to PB T-SPOT’s [
18,
19], our research affirmed that the diagnostic cut-off obtained from PB was not the optimal for PF in the high prevalence areas, and PFs’ cut-off was much higher than PB,which was in line with the expectation of another high-burden settings, for example, 300 SFCs/10
6 mononuclear cells in Korea [
20].
ADA, the most common biomarker for the diagnosis of TB pleurisy, has a value of more than 40 IU/L in lymphocyte-dominated PE and carries a PPV of 98% in high TB endemic regions [
3,
4], while a retrospective analysis on ADA in 1637 subjects obtained an NPV of 100% with less than 15.0 IU/L [
21]. In our study, we found that recognizing > 40 IU/L as the sole indicator of TPE may not be the most suitable approach, as it obtained high specificity while sacrificing sensitivity; 35.5% (141/397) of TPE patients had an ADA level that was lower than 40 IU/L in this study, 29.8% of which had a definite aetiological basis. In addition, the utility of a cutoff of 22.4 IU/L derived from ROC analysis was better than 40 IU/L (Youden index 0.729 vs. 0.641), this situation was similar to the publication of Santos et al. [
22]. However, when comprehensively considering the non-specific elevation of ADA levels caused by non-tuberculosis inflammatory PE [
19] (particularly complicated parapneumonic effusions and empyemas) and lymphomas, patients whose ADA was more than 20 IU/L and less than 40 IU/L were classified as indeterminate ADA. Conversely, the PF T-SPOT assay showed excellent diagnostic utility between ADA indeterminate groups, and its accuracy was higher than 90.2%, and we predict that the result of the PF T-SPOT assay could be a considerable indicator for highly suspected TPE patients with indeterminate ADA.
Many previous studies [
23,
24] indicated that the performance of diagnosing TPE for patients aged more than 45 yrs. is still unknown. A recent study found that only 4.65% of elderly TPE subjects had levels over 40 IU/L [
25]. Our research fully demonstrated this phenomenon simultaneously. In addition, we observed that the unclear boundary influenced by age for ADA was opportunely concentrated in the ADA indeterminate groups. Among the patients in the ADA indeterminate group, 85.7% (18/21), 73.7% (19/26) and 47.7% (31/65) of patients aged more than 45 yrs. were in the non-TPE, confirmed TPE and probable TPE groups, respectively, while the superior performance of the PF T-SPOT assay in distinguishing between ADA indeterminate groups may be explained by its steady performance in all age groups, in addition to interference with ADA by other inflammatory aetiologies. Nevertheless, ADA is a widely used biomarker for screening TPE due to its simplicity, rapidity, and low financial cost, but the above results proved that overreliance on ADA differentiation may lead to missed diagnosis/misdiagnosis in clinical settings, especially in the indeterminate range. In addition, the PF T-SPOT assay could fill this gap.
In addition to age, we screened two additional high-risk factors that were significantly related to TPE, sex and BMI. There were 5 non-TPE patients scoring 11 (simultaneously having three high-risk factors: age < 45 yrs., male sex and BMI < 22), which directly demonstrated that these clinical characteristics could be an effective reference index for discerning TPE from other PEs. We often defined these patients as the population with a high incidence of TPE, also called the typical population. However, it is notable that the utility of ADA fluctuated distinctly by stratified analysis, and if it did not satisfy any one condition (defined as the unconventional population), it was inferior to the T-SPOT assay. The unconventional population is frequently the focal point and has difficulties in diagnosis; as a result, the PF T-SPOT assay can provide powerful identification evidence for these patients.
There are several limitations in the current study. First, the entire study was performed in a single centre that specialized in TB. The geography, relative single control composition and aetiological attribution error and/or bias were incalculable. Second, we obtained the optimal cutoff for the PF T-SPOT.TB assay from ROC analysis in this training cohort, and its definite accuracy would need further validation. Finally, 38.1% of the clinically diagnosed patients lacked the aetiological basis due to objective factors such as no sputum or unsatisfactory sputum with the detection standard, which may bias the sputum detection rate.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.