A frequently overlooked test in the diagnosis of PE is a sputum culture for mycobacteria. This is positive in between 41.7% and 52% of cases.8,35,36

PF staining for TBPE diagnosis in an immunocompetent patient does not appear to be useful (5% yield).24 However, staining can be positive in 20%37 of HIV-positive patients, and is justified in this population. PF culture yield is higher, depending on the culture medium used. The bacillus is identified in 12%–36.6% of cases24,38 when solid media (Löwenstein–Jensen) is used. Liquid media are more sensitive. Culture yield with the BACTEC system increases from 12% to 24%,38 while Ruan et al.34 achieved 63% with a significantly shorter waiting period.38,39 Two factors may play a part in the result: HIV infection and the predominant cell population in PF. Both PF staining and culture are more often positive in HIV-positive patients than in HIV-negative subjects (37% vs 0% for staining; 43% vs 12% for solid medium culture; 75% vs 24% for liquid medium culture).37,38 In these cases, TBPE may be less due to a reaction to tuberculin proteins and more to reduced mycobactericidal macrophage activity in a weakened immune system, permitting mycobacteria to reside longer in the pleural space. The second factor that could affect culture yield is the predominant cell population. The higher the proportion of neutrophils in PF, the greater the chance that the culture will be positive.35,40 This is probably because in the early stages of the disease, when neutrophils are predominant, the immune system is not yet effective against bacilli. In this situation, the mycobacterial load in PF is greater. When the activated macrophages phagocytize Mycobacterium tuberculosis (MT), the intrapleural mycobacterial load, and therefore the chance that the culture will be positive, is reduced.41

MODS (microscopic-observation drug-susceptibility) culture is a cheap, highly sensitive and relatively rapid liquid medium from which drug susceptibility data can be obtained simultaneously. The yield from this culture in both PF and PB is greater than with Löwenstein–Jensen (20% vs 7% and 81% vs 51%, respectively) and waiting time is shorter.42

As there are few bacilli in TBPE, microbiological methods lack sensitivity. Specific TB nucleic acid sequence amplification techniques are a fast way of detecting MT in PF, even if there are few bacilli.43 In a meta-analysis of 38 articles, sensitivity was 62% and specificity was 98%,44 suggesting that this method is useful for confirming diagnosis but not for ruling it out. Low sensitivity can be attributed to the low bacillus load of the samples, the probable presence of amplification-inhibiting substances in the PF, the amplified genomic sequence, and possible intracellular mycobacterial sequestration. This test is not routinely performed in the determination of PE, due to its poor sensitivity and high cost.

Xpert is an automatic nucleic acid amplification method that simultaneously detects MT and rifampicin resistance in less than 2h.45 However, its TBPE diagnostic yield is low: specificity is 100% but sensitivity ranges between 15% and 27%.46–48

Closed PB is the most sensitive method for diagnosing TBPE. Granuloma are seen in 50%–97% of cases.17,18,24,49–51 Although granulomatous pleuritis occurs with other diseases, over 95% of cases are caused by TB. TBPE diagnosis can be increased from 79.8% without systematic PB culture to 91.5% when this technique is performed.24 Thoracoscopy can also be useful, although it is usually not necessary to go to these lengths. However, the sensitivity of this technique in TB patients can be as high as 100%.52

There are therefore numerous difficulties involved in reaching a definitive diagnoses: low test yields, long waiting time for culture results, invasive techniques such as BP. The value of determining TB immunological response biomarkers in PF for improving diagnostic yield has been evaluated (Fig. 1).

Adenosine deaminase (ADA) is a widely distributed enzyme that plays a significant role in T-lymphocyte proliferation and differentiation. A meta-analysis of 63 studies of over 8000 PEs, 2796 of which were TBPE, shows that the sensitivity, specificity, positive and negative probability ratios (PPR and NPR, respectively), and the odds ratio were 92%, 90%, 9, 0.10, and 110, respectively.53 ADA has two isoenzymes: ADA1 and ADA2. The first is found in all cells, while the other is only found in monocytes/macrophages and levels increase when these cells are stimulated by live microorganisms.54 ADA2 is more common in TBPE. There are 3 methods for determining ADA isoenzymes: (a) fraction separation by electrophoresis55,56; (b) erythro-9(2-hydroxy-3-nonyl)adenine (EHNA) ADA1 inhibition,55,57–59 and (c) ADA1 and ADA2 calculation using their different affinity ratios for adenosine and 2′-deoxyadenosine substrates.60–62 The latter yields 2′-deoxyadenosine/adenosine or ADA1/ADA63 ratios, although the first method generates fewer errors, since the data are obtained from enzyme activity and not from theoretical calculations.

Although ADA2 yield is slightly higher (ADA vs ADA2: sensitivity 100% for both, specificity 91% vs 96%60; sensitivity 93.7% vs 97%, specificity 89% vs 94%64), its use in clinical practice does not seem to be justified.

ADA levels rise in TBPE with low CD4+ cell counts65 (sensitivity 94%, specificity 95%). This is probably because monocytes are not affected by retroviral infection, unlike cell populations in which neutrophils predominate.40,66 One reason for this may be that neutrophils release cytokines that attract ADA-2-producing monocytes and macrophages into the pleural space, thus contributing to enhanced ADA activity in these effusions.67 ADA in PF is routinely determined in countries in which the prevalence of TBPE is moderate to high. In these countries, ADA yield in subjects under the age of 35 years is so high that PB may not even be required for diagnosis.68 However, in countries in which prevalence is low, the positive predictive value of this test will be equally low and will only be useful for ruling out the disease (high negative predictive value)69 (Fig. 3).

Fig. 3.

Effect of prevalence of tuberculous pleural effusion on positive and negative predictive values of diagnostic adenosine deaminase (sensitivity 90%, specificity 93%). When disease prevalence is low, the positive predictive value is also low, so this parameter is useful for ruling out the disease (high negative value). If prevalence is high, the positive predictive value is also high, so it may be useful for confirming the disease. NPV: negative predictive value; PPV: positive predictive value.

(0.08MB).

ADA determination has several limitations. As various methods are available, results may not be comparable and cutoff points can vary, while other factors, such as ethnicity, may also play a part. The cutoff point, therefore, should be established according to the results of each site, or at least, on the basis of similar population studies using the same methodology.70 In 3 studies of 630 patients with non-TB lymphocytic PE, 2.8% of the cases (18 patients) had high ADA levels.71–73 Raised ADA levels may be found in one third of parapneumonic effusions and in two thirds of empyemas,74 as well as in other types of effusion.23,24 ADA also produces false negatives, but repeatedly low levels seem to rule out the disease.75 ADA remains an inflammatory biomarker and thus cannot be viewed as a substitute for culture nor does it provide information on anti-TB drug sensitivities. In regions with high mycobacterial resistance, therefore, a lymphocytic exudate and a high ADA result would suggest the need for PB, culture, and drug sensitivity testing.76

Interferon gamma (IFN-γ) is a cytokine released by CD4+ T-lymphocytes that increases macrophage mycobacterial activity. IFN-γ in PF or released by mononuclear cells in PF after stimulation from MT-specific antigens can be determined. In a meta-analysis of 22 articles and 2101 patients (782 with TBPE), sensitivity, specificity, PPR, NPR, and odds ratio were 89%, 97%, 23.45, 0.11, and 272.7, respectively.77 As with ADA, there is no universal cutoff point, because laboratory methods are not always the same. Malignant effusions, particularly if they are hematological, and empyemas may also present high IFN-γ levels.78 IFN-γ yield in TBPE is slightly higher than that of ADA, but the difference is not significant.79

Determination of IFN-γ stimulated with MT-specific antigens (ESAT-6 and CFP-10) may contribute to TBPE diagnosis.80 In a meta-analysis of 366 patients (213 with TBPE) from 7 studies, sensitivity, specificity, PPR, NPR, and odds ratio in PF were 75%, 82%, 3.5, 0.24, and 19, respectively. Levels in blood were slightly lower.81 The quality and cost of these results do not justify routine determination in daily clinical practice.82

Determination of other biomarkers in PF for the diagnosis of TBPE had a lower yield less than that of ADA and IFN-γ (see summary in Table 1).83–94

Although differential cell count in the first 2 weeks can reveal a predominantly polymorphonuclear cell population,95 the predominant population is normally leukocytes (>50%).34 If the leukocyte count is higher than 80%, the differential diagnosis would include TB, lymphoma, and some rare cases of metastatic PE.96 Other possibilities can be ruled out due to low prevalence (for example, rheumatoid pleuritis, and sarcoidosis),97 clinical history (aorto-coronary bypass surgery,98 acute lung rejection, and yellow nail syndrome), or the appearance of the fluid (chylothorax). Mesothelial cell count is rarely >5%,99 although a higher figure would not rule out TB.100 The suggestion that a PF eosinophil count >10% all but rules out TB still more controversial.101,102

Several studies have developed diagnostic tests based on a combination of variables (Table 2).82,103–108 In general, all manage to improve the yield of individual parameters, but since they rely on complex, often arbitrary, statistical calculations, they are of little use in daily clinical practice. In contrast, Burgess et al.108 obtained a high diagnostic yield with the combination of ADA >50U/l and a lymphocyte/neutrophil ratio of >0.75.

In a study of patients under the age of 40 (218 cases, 165 with TBPE), a regression tree was used to classify PE as TB or non-TB. Twenty variables (clinical, radiological, laboratory tests, and tuberculin results) were taken into consideration, while the statistical program only selected ADA values (as the principal variable) (>35U/l) and the lymphocyte percentage (>31.5%) (sensitivity 99.4%, specificity 98.1%) (Fig. 4).109 The results of a second model, without ADA, selecting (in this order) lymphocyte percentage, fever and cough were significantly worse than the ADA model (sensitivity 95.2%, specificity 94.3%) (P<.001).

Fig. 4.

Regression tree for predicting tuberculous pleural effusion. Adenosine deaminase (U/l) and lymphocytes (%) were the variables selected for the final regression tree. ADA: adenosine deaminase; TBPE: tuberculous pleural effusion.

(0.1MB).

Source: Valdés et al.109

In a study designed to determine whether TBPE can be diagnosed solely by PF analyses, ADA >45U/l and lymphocytes >80% had a sensitivity of 58.4% and specificity of 99.5%. Specificity of 100% was achieved if proteins >5g/dl were added to these criteria (sensitivity 34.9%).110