Abstract
Tuberculous pleuritis is a common manifestation of extrapulmonary tuberculosis and is the most common cause of pleural effusion in many countries. Conventional diagnostic tests, such as microscopic examination of the pleural fluid, biochemical tests, culture of pleural fluid, sputum or pleural tissue, and histopathological examination of pleural tissue, have known limitations. Due to these limitations, newer and more rapid diagnostic tests have been evaluated. In this review, the authors provide an overview of the performance of new diagnostic tests, including markers of specific and nonspecific immune response, nucleic acid amplification and detection, and predictive models based on combinations of markers. Directions for future development and evaluation of novel assays and biomarkers for pleural tuberculosis are also suggested.
Tuberculous pleuritis is a common manifestation of extrapulmonary tuberculosis (TB) and is the most common cause of pleural effusion in many countries 1–3. Pleural TB occurs as a result of a TB antigen entering the pleural space, usually through the rupture of a subpleural focus, followed by a local, delayed hypersensitivity reaction mediated by CD4+ cells 4. This process may occur during primary or re-activation TB and may or may not involve viable bacilli entering the pleural space 5.
The presence of mycobacterial antigens in the pleural space elicits an intense immune response, initially by neutrophils and macrophages 6, 7, followed by interferon (IFN)-γ-producing T-helper cell (Th) type 1 lymphocytes 4, 8, resulting in a lymphocyte-predominant exudative effusion. This cellular trafficking is facilitated by homing surface markers and chemokine gradients 9, 10. This intense but poorly understood local immune response by sensitised lymphocytes to TB antigen is synonymous with the Koch phenomenon 11. A recent study indicates that cells of an alternative T-cell profile, CD4(+)/CD25(+)/FoxP3(+) regulatory T-cells, are expanded in TB pleural effusion and can suppress some effector responses, although the precise role of these cells remains unclear.
Conventional diagnostic tests for pleural TB include microscopic examination of the pleural fluid for acid-fast bacilli, mycobacterial culture of pleural fluid, sputum or pleural tissue, and histopathological examination of pleural tissue for granulomatous inflammation. These tests have recognised limitations for clinical use, although, in combination, they have been recognised as the best reference standard for evaluation of the accuracy of novel tests 12.
Microscopy of the pleural fluid for acid-fast bacilli is positive in <5% of TB pleuritis cases, due to the paucibacillary nature of the disease 12, 13. Mycobacterial culture of pleural fluid also has low sensitivity (24–58%) 13, 14 and is limited by the lengthy delay in obtaining results: up to 8 weeks if solid culture media are used. Biopsy of pleural tissue for combined histological examination and mycobacterial culture of pleural fluid and tissue is the most sensitive of the currently available diagnostic methods, but may still be falsely negative in 15–20% of these cases 12, 15. In addition, pleural biopsy is invasive, and yield as well as complication rates are dependent on the skill of the operator because it is technically difficult, particularly in children. Hence, biopsy adds considerable risk and cost to the workup. Mycobacterial culture of spontaneous sputum or gastric lavage has a variable yield, from 0% 16 to ∼30% 17, depending on the presence of associated lung parenchymal lesions 17. Since patients with pleural TB rarely produce sputum spontaneously, routine collection of induced sputum has been proposed for the diagnosis of pleural TB. In a prospective study of 113 patients with confirmed pleural TB, induced sputum had a sensitivity of 52%, compared with 12% sensitivity of pleural fluid culture 18. Due to these limitations of conventional tests, plus the delay of several weeks for mycobacterial culture results, newer rapid tests and biomarkers have been evaluated. The present review provides a narrative overview of the literature on the tests, summarised in table 1⇓ and illustrated in figure 1⇓, and reviews their performance characteristics for the diagnosis of pleural TB.
Schematic representation of pathways and systems involved in biomarkers for pleural tuberculosis. IFN: interferon; ESAT: early secreted antigenic target; CFP: culture filtrate protein; BMI: body mass index; ADA: adenosine deaminase; TNF: tumour necrosis factor; IL: interleukin.
Novel tests and biomarkers for the diagnosis of tuberculous pleural effusions
For clarity, all tests were grouped into the following categories, although some overlap exists: 1) nonspecific inflammatory and immune response markers; 2) specific markers of immune response; 3) nucleic acid amplification tests; and 4) scoring systems based on combinations of tests.
NONSPECIFIC INFLAMMATORY AND IMMUNE RESPONSE MARKERS
Adenosine deaminase
Adenosine deaminase (ADA), released by activated lymphocytes, macrophages and neutrophils, is a nonspecific marker of inflammation. The ADA2 isoenzyme released from monocytes and macrophages is the predominant contributor to total ADA activity 49. A high diagnostic accuracy of ADA activity measurement has been reported in several studies. In a meta-analysis, Greco et al. 19 found that among 31 studies published prior to 2000, including 4,738 patients, the pooled sensitivity was 92% (range 56–100%) and pooled specificity was 89% (55–100%) using composite reference standards including culture, histology, sputum culture and response to therapy. The specificity for discriminating TB from malignant effusions, an important differential diagnosis in elderly patients, remained high (95%) but was disappointingly low for parapneumonic effusions.
Greco et al. 19 concluded that in low and intermediate TB incidence settings, the negative predictive value was sufficient that a negative ADA activity result would preclude the need for pleural biopsy. However, in the same settings, the positive predictive value was poor. In contrast, in high prevalence settings, a positive ADA would provide a 99% post-test probability of TB. This crucial point, that the predictive value of a test is highly dependent on the prevalence of the disease, explains the differences in performance of this test in different studies. Thus, the clinical context must be taken into account when interpreting the ADA result.
Many studies in the review by Greco et al. 19 included <30 patients, resulting in wide confidence intervals (CIs) around sensitivity and specificity estimates and significant heterogeneity between estimates. In addition, few studies reported blinding or described in detail the inclusion criteria for the control group.
Using histopathology as the reference standard, pleural fluid ADA activity, PCR and immunoglobulin (Ig)A-ELISA tests were evaluated in a high TB incidence country 20 among 77 patients with pleural effusion, 60 of whom had TB pleuritis. ADA activity was the only test that had a significantly higher sensitivity than histopathological examination. The study included only 17 nontuberculous effusions, and, therefore, specificity could not be adequately estimated 21.
Determination of ADA isoenzyme activity in pleural fluid may increase the accuracy of the test. ADA1 is secreted by lymphocytes and monocytes, while ADA2 is secreted only by monocytes and is found in a higher concentration in TB pleuritis 22, 23. However, because the additional yield is small, a study would require a very large sample size to demonstrate that isoenzymes have significantly higher specificity than total ADA activity.
Neopterin
Neopterin is a marker of Th1 immune activation, as it is secreted by activated macrophages. Neopterin levels in pleural fluid have been found to be higher in patients with TB than patients with malignancy 24, 25 or other 26 conditions. Very high levels of neopterin have been observed in uremic pleural effusions 27, raising doubts about its specificity for TB. In one study that used histological evidence of caseating granuloma as the reference standard for TB, the sensitivity and specificity of neopterin were 44 and 85%, respectively 25. In a second study, which used ADA as the reference standard, the sensitivity was 85% and the specificity was 93% 26. However, ADA may not be a suitable reference standard because it is also an indicator of nonspecific inflammation, so a high degree of correlation may reflect the same underlying inflammatory response as neopterin.
Leptin
Leptin, a 16-kDa product of the obese (ob) gene, may be involved in cross-regulation between nutritional status and the immune response in TB. Serum leptin levels have been shown to be reduced in patients with active pulmonary TB 50 and cancer 51. One study evaluated total leptin pleural fluid levels and leptin pleural fluid/serum ratio in 17 patients with pleural TB, and parapneumonic (n = 7) and malignant (n = 21) effusions 28. The reference standard for TB in this study was the presence of granuloma in pleural tissue. ADA activity was also measured in pleural fluid. Pleural ADA activity sensitivity and specificity were both 100% and pleural leptin (<9.85 ng·mL−1) sensitivity and specificity were 82% (p = 0.01). Thus, there is no evidence that leptin performs better than pleural fluid ADA.
Lysozyme
Lysozyme is present in epithelial cells of granulomas, macrophages and activated granulocytes. In a study of patients with pleural effusion in a setting with high TB incidence, Valdés et al. 29 measured ADA activity in 405 specimens, lysozyme in 276 specimens and IFN-γ in 145 specimens. Using culture of pleural fluid or tissue and histopathological evidence of caseous granuloma in pleural tissue as the reference standard, the sensitivity of ADA, lysozyme and IFN-γ were 100, 85.7 and 94.2%, respectively, while specificities were 94.9, 61.6 and 91.8%, respectively. Thus, ADA and IFN-γ were more accurate than lysozyme levels. However, 51 patients were excluded from the analysis because a final definitive diagnosis could not be achieved, and the criteria for selection of specimens for lysozyme and IFN-γ testing were not reported. For these reasons, potential selection bias cannot be excluded.
Moriwaki et al. 30 found that lysozyme could discriminate malignant from tuberculous effusions with a sensitivity of 100% and a specificity of 83%, compared with culture and/or histopathology of pleural tissue. The pleural fluid/serum lysozyme ratio had a sensitivity of 100% and a specificity of 88%. The mean fibronectin concentration in pleural fluid with tuberculous effusion was significantly higher than that in malignant effusion, but there was a marked overlap between the groups. In this study, lysozyme and fibronectin were less accurate than pleural fluid ADA 30.
Cytokines
Of all the cytokines, IFN-γ has been the most studied. The evidence for use of IFN-γ in pleural fluid for diagnosis of TB has been reviewed by Greco et al. 19 and Jiang et al. 31. In the latter review, based on 22 studies including 782 TB patients and 1,319 non-TB patients, the summary estimate of sensitivity was 89% (95% CI 87–91%) and of specificity was 97% (96–98%) 31. IFN-γ was more sensitive and specific for the diagnosis of pleural TB than interleukin (IL)-12p40, IL-18, immunosuppressive acidic protein or soluble IL-2 receptor, in a study that directly compared these tests in the same samples 33. In another study, the accuracy of IFN-γ was superior to that of ADA 19.
The concentration of other pro-inflammatory cytokines, such as tumour necrosis factor-α and IL-1β, has been shown to be higher in tuberculous than in malignant effusions, but no cut-off value for positivity has been established 34. In another study, IL-8 and IL-6, and soluble IL-6 receptor were detected in carcinomatous but not in tuberculous pleural effusions 52.
Complement activation
Porcel et al. 53 measured SC5b-9 and C3a-desArg, two products of complement activation, in the pleural fluid of 83 patients using commercial ELISA tests. The sensitivity was 84% for SC5b-9 and 81% for C3a-desArg. Hara et al. 35 showed SC5b-9 had a higher sensitivity (100%) and ADA activity had a higher specificity than conventional tests (culture and/or histopathological examination). Discrimination of TB pleuritis from malignant effusions was better than discrimination from autoimmune or parapneumonic effusions. In fact, the SC5b-9 level seemed more useful for the diagnosis of rheumatoid pleural effusions 36 and for differentiation of empyema from uncomplicated parapneumonic effusions 37.
Cell subsets
Tuberculous effusions are usually rich in mononuclear cells 12. In one study, although pleural fluid ADA activity correlated with the number of CD4+ T-cells in the pleural space 38, the number of pleural CD4+ cells was not an accurate diagnostic test for TB.
In conclusion, among the nonspecific inflammatory biomarkers that have been evaluated, ADA and IFN-γ are the most accurate for pleural TB diagnosis, with consistently high sensitivity in many studies. Specificity of ADA is a concern because of the nonspecific nature of inflammatory response. Most studies used in-house methods for ADA activity estimation, which may demonstrate considerable variability in performance. Commercial kits have been used in a few studies, but these kits should be validated in different settings. ADA activity measurement is simple and inexpensive to perform and does not require special equipment, making it an attractive option for settings with high TB incidence and limited resources. In contrast, measurement of IFN-γ is expensive, estimated to be as much as the cost of treatment for six TB patients 32.
SPECIFIC MARKERS OF IMMUNE RESPONSE
T-cell response to specific antigens
Recently, in vitro, T-cell-based IFN-γ release assays (IGRAs) have been developed and licensed for diagnosis of latent TB infection. Normally, these tests use peripheral blood mononuclear cells (PBMCs), but they can be used with pleural fluid mononuclear cells. These assays detect IFN-γ secreted by mononuclear cells in response to in vitro stimulation with the Mycobacterium tuberculosis-specific antigens early secreted antigenic target-6 and culture filtrate protein-10. The genes that encode these antigens are not present in any of the M. bovis bacille Calmette–Guerin (BCG) strains or certain common nontuberculous (environmental) mycobacteria 54. Thus, in theory, the test should not cross-react with antigens present due to BCG vaccination 54, 55. In a recent study 39, the T-SPOT.TB test (Oxford Immunotec Ltd, Oxford, UK) was performed on PBMCs and mononuclear cells from pleural fluid from 20 patients clinically suspected to have TB pleuritis and 21 subjects with other diagnoses. The sensitivity of T-SPOT.TB was 90% using the blood assay and 95% for pleural fluid, but the specificity was only 67% for blood and 76% for pleural fluid. This poor specificity may reflect positive reactions due to coincidental latent TB infection, coexisting or transient infection.
An alternative commercially available T-cell assay using an ELISA platform is the QuantiFERON®-TB Gold in-tube test (Cellestis Ltd, Carnegie, Australia), which is more amenable to high throughput and flexibility with analysis. However, unpublished preliminary observations indicate that high background IFN-γ levels in the negative control tube precludes the use of ex vivo unprocessed pleural fluid in the assay. Pre-specified numbers of mononuclear cells in culture medium may overcome this drawback.
IGRAs, designed to detect latent TB infection, will only be useful for the diagnosis of pleural TB disease if it can be shown that persons with latent TB infection and pleural effusions due to nontuberculous causes, such as cancer, have positive IGRA using serum, but negative IGRA using pleural fluid. This finding would suggest that TB-sensitised lymphocytes remain in peripheral blood but are not present among the lymphocytes found in the pleural fluid. This hypothesis as yet remains unproven. Until this issue is settled, these assays may not offer any additional value for the diagnosis of pleural effusion, beyond the measurement of free IFN-γ in the pleural fluid.
B-cell response (antibody detection)
Although detection of serum antibodies against TB antigens is known to have poor and highly variable sensitivity and specificity 56, 57, attempts have been made to detect antibodies in pleural fluid specimens using techniques such as ELISA. Among pleural fluid antibody tests, detection of IgA against MPT-64 and MT-10.3 (Rv3019c), two recombinant protein antigens, had the best sensitivity 20, 40. Anti-A60 IgM and IgG antibodies in pleural fluid had suboptimal sensitivity for TB 41. Using antibodies against five different antigens, Chierakul et al. 42 found the performance of serology of pleural fluid to be disappointing, with sensitivity and specificity <60%. Anti-P32 levels were reported to be higher among five patients with pleural TB, but the sensitivity and specificity were not reported 43. Sensitivity to anti-lipoarabinomannan (LAM) and to anti-TB glycolipid antibodies was very low in other studies, although specificity was high 44–47.
One study reported a simple and cost-effective diagnostic tool (an in-house TB screen test) for pulmonary and extrapulmonary TB, using a liposome agglutination assay to detect serological responses to purified mycobacterial glycolipid antigens 58. The assay was able to detect very low anti-glycolipid antibody concentrations in the serum of individuals with TB. The status for latent TB in the control group was not reported. To date, the test has not been applied to pleural fluid specimens, although four subjects with pleural TB were serum tested 58.
In conclusion, serologic tests using antibodies against mycobacterial protein and glycolipids show some potential for the diagnosis of pleural TB because of their high specificity, but are limited by very poor sensitivity.
DETECTION OF M. TUBERCULOSIS NUCLEIC ACID SEQUENCES BY AMPLIFICATION TESTS
Several commercial and in-house assays exist for the amplification and detection of M. tuberculosis nucleic acids from specimens such as sputum. These tests have also been used with specimens such as pleural fluids. In a meta-analysis of 40 studies of nucleic acid amplification tests (NAATs) for pleural TB, Pai et al. 48 reported that commercial NAATs have a potential role in confirming TB pleuritis because of high specificity (98 (95% CI 96–98)%). However, these tests had low and variable sensitivity (62 (43–77)%) and, therefore, were not useful in excluding the disease. The accuracy of in-house NAATs was poorly defined because of heterogeneity across studies, presumably reflecting the heterogeneity of in-house test protocols. Similar results have been reported with NAATs for TB meningitis 59 and TB lymphadenitis 60.
The reasons for the low sensitivity of NAATs in pleural fluid specimens are unclear. The presence of inhibitory substances in pleural fluid is an unsatisfactory explanation, as most commercial assays have an internal positive control to account for inhibition. The small amount of mycobacteria in pleural effusion may play some role, but the low sensitivity is more likely to be related to a technical aspect of nucleic acid extraction. Thus, caution is needed when interpreting negative NAAT results in pleural fluids, and performance will depend largely on pre-test probability.
In conclusion, the evidence is consistent that NAATs have modest sensitivity but excellent specificity for pulmonary and extrapulmonary TB. At present, no commercial kit has been approved by the US Food and Drug Administration for the diagnosis of extrapulmonary TB, and NAATs cannot be used in isolation to rule in or rule out pleural TB.
SCORING SYSTEMS BASED ON COMBINATIONS OF MARKERS
Due to the limitations of individual tests, scoring systems have been developed, based on the results of multiple tests (table 2⇓). A scoring system that makes use of simple clinical and pleural fluid data was created by Porcel and Vives 61 to discriminate between TB and malignant pleural effusions. Two models were developed, one with and one without ADA activity measurement. In the first model, four variables predicted a tuberculous aetiology: ADA ≥40 U·L−1 (five points); age <35 yrs (two points); temperature ≥37.8°C (two points); and pleural fluid red blood cell count <5×109·L−1 (one point). In the second model, TB was predicted by: age <35 yrs (two points); temperature ≥37.8°C (two points); pleural fluid red blood cell count <5×109·L−1 plus no history of malignancy (three points); pleural protein ≥50 g·L−1 (one point); and pleural fluid/serum lactate dehydrogenase ratio ≥2.2 (one point). Summated scores of at least five in model one and at least six in model two yielded measures of sensitivity (95 and 97%, respectively) and specificity (94 and 91%, respectively) for discriminating TB from malignant effusions.
Combination of tests and biomarkers for the diagnosis of tuberculous pleural effusions
Villegas et al. 62 also proposed a combination of clinical and biological markers. In their study, ADA activity, IFN-γ levels and PCR were tested in the same samples for the diagnosis of microbiologically and/or histologically confirmed pleural TB. The sensitivities of ADA activity, IFN-γ levels and PCR were 88, 86 and 74%, respectively, and the specificities were 86, 97 and 90%, respectively. The combination of PCR, IFN-γ and ADA activity allowed an increase of sensitivity and specificity compared with individual methods in isolation. However, in other studies using different tests in the same samples, in which ADA activity had a higher accuracy, no other test significantly added sensitivity to the ADA 20, 65.
In a logistic regression model, a combination of duration of disease, protein levels, total leukocyte count, percentage of lymphocytes and ADA activity measurement was >95% sensitive and specific for the diagnosis of TB in a high prevalence setting 63. The first four variables added specificity to ADA activity, without loss of sensitivity. The lymphocyte/neutrophil ratio has also added to ADA specificity in another study of 472 pleural effusions 64.
More models, which take into account clinical features and simple, inexpensive and rapid diagnostic tests, can be developed with the use of modelling tools. These models are highly relevant to clinicians practising in poor resource settings. Even when ADA is unavailable, use of demographic and clinical information in combination with a cell differential count, in high incidence settings, can yield a diagnosis of TB with a high predictive value.
EMERGING TECHNOLOGIES THAT NEED TO BE EVALUATED FOR PLEURAL TB
In the past few years, several new TB diagnostics have been developed. These include tests such as loop-mediated isothermal amplification, detection of LAM (a TB antigen) using ELISA, phage-based assays and rapid culture systems 66, 67. To date, none of these have been adequately evaluated for TB pleuritis.
In addition to laboratory tests, biological processes found in nature are being used to inspire the development of new technologies. Artificial intelligence represents an attempt to simulate the human brain in resolving complex questions by breaking them down into small problems. Neural network systems have been used to extract relevant information from complex databases, categorise them into groups with similar characteristics and organise them into distinct patterns 68. The neural network system can help in identifying patients’ variables that may not make biomedical sense but could be relevant for the diagnosis of diseases. It can also detect errors in databases. This tool has been useful in predicting pulmonary TB in patients with negative sputum smears 69, 70. The tool can also easily generate scoring models based on clinical and laboratory data. Research is needed to evaluate the usefulness of neural network systems for the diagnosis of pleural TB.
FUTURE DIRECTIONS
Table 3⇓ presents some suggested research directions. Diagnostic studies on pleural TB are particularly needed in specific populations, including individuals with HIV infection, children and other high-risk populations where traditional diagnostic tests have worse yield. In addition, commercially available tests for diagnosis of pulmonary TB should be validated for diagnosis of pleural TB. Better tests are needed, and/or available tests, such as NAAT or serologic tests that have good specificity, need to be modified to improve sensitivity.
Recommendations for future research
Multidrug-resistant and extensively drug-resistant TB are spreading worldwide and have become a special issue for TB control. There are no reports on the drug-resistance profile of strains isolated from pleural specimens, but given the trend of ever-increasing resistance, development of new techniques, such as NAATs, to identify drug-resistant strains causing pleural TB will be important.
Efforts are also needed to improve the quality and reporting of TB diagnostic studies, according to the Standards for Reporting of Diagnostic Accuracy recommendations 71. The main limitation of many published studies is the absence of a well-defined reference standard (culture and histopathological evidence of TB), leading to test performance results that may be biased. In addition to the lack of a definitive gold standard, diagnostic studies are plagued by methodological and design flaws that compromise validity, such as lack of a consecutive or random patient-sampling method, use of a case–control design (where severe cases are often compared with healthy controls) and lack of blinding 72–74. Statistical analysis of results should receive special attention in future studies 21, 75. Sensitivities and specificities are proportions, and their confidence intervals must be calculated and reported 75. When using statistical tests for comparison of sensitivities, only patients with the disease should be compared, while for specificity, only patients without the disease should be compared 21.
CONCLUSION
Diagnosis of pleural TB remains a challenge, as summarised in table 1⇑. Among nonconventional tests, ADA and IFN-γ have the best sensitivity and specificity, but they are biomarkers of the inflammatory process in the pleural space and do not confirm the aetiologic agent. NAATs and serology are promising but need further development to improve sensitivity. Limited evidence is available for other novel tests and biomarkers.
Combinations of tests seem to perform better than any single test, especially combinations that include adenosine deaminase, but only a few studies have evaluated scoring systems (table 2⇑). Further work is necessary to identify the best (and simplest) combination that will be most useful in clinical practice.
Support statement
For a full list of all grants and funding received, please see the Acknowledgements section.
Statement of interest
None declared.
Acknowledgments
A. Trajman received funding from an International Clinical, Operational and Health Services Research Training Award for AIDS and TB, Fogarty International Center, National Institutes of Health, no. 5U2 R TW006883-03. M. Pai is supported by a Canadian Institutes of Health Research New Investigator Award, and D. Menzies by a Fonds de la Recherche en Santé du Québec Chercheur National salary award. K. Dheda is supported by a South African National Research Foundation Research Chair award and a Medical Research Council career development award. R. Joshi is a recipient of a training fellowship from the Fogarty AIDS International Training Program, grant no. 1-D43-TW00003-17, USA.
- Received November 6, 2007.
- Accepted January 24, 2008.
- © ERS Journals Ltd
References
Jump To
- Article
- Abstract
- NONSPECIFIC INFLAMMATORY AND IMMUNE RESPONSE MARKERS
- SPECIFIC MARKERS OF IMMUNE RESPONSE
- DETECTION OF M. TUBERCULOSIS NUCLEIC ACID SEQUENCES BY AMPLIFICATION TESTS
- SCORING SYSTEMS BASED ON COMBINATIONS OF MARKERS
- EMERGING TECHNOLOGIES THAT NEED TO BE EVALUATED FOR PLEURAL TB
- FUTURE DIRECTIONS
- CONCLUSION
- Support statement
- Statement of interest
- Acknowledgments
- References
- Figures & Data
- Info & Metrics