Tuberculosis (TB) is a frequent cause of pleural effusion (PE): in a series of 3077 consecutive patients undergoing diagnostic thoracentesis in our hospital, TB was the fourth most common etiology.1 However, in around 2 thirds of cases, a diagnosis of probable tuberculous pleural effusion (TPE) is made on basis of the biochemical characteristics of pleural fluid (PF) (lymphocytic exudate with adenosine deaminase [ADA] ≥35U/l) after other potential causes of PE have been ruled out.2 In Catalonia, the incidence of tuberculosis has fallen in recent years from 48.3/100000 inhabitants in 1994 to 14.4/100000 in 2015.3 Among the cases of exclusively extrapulmonary TB reported in 2015, pleural involvement was the second most common (25%) after lymphatic disease (42%).3

Microbiological tests performed to confirm the diagnosis of TPE are often negative. In 2 Spanish series, sputum and PF cultures in Löwenstein–Jensen medium showed a sensitivity of 41% and 36%, respectively.4,5 Nucleic acid amplification techniques or polymerase chain reaction (PCR) are fast and highly specific, but lack sufficient sensitivity. For example, in a meta-analysis of 24 studies, the sensitivity of an automated PCR technique (Xpert®-MTB/RIF) in PF was 23% in patients with probable TPE.6

The aim of this study, conducted in a large series of patients with TPE, was to analyze variables associated with: (1) a greater frequency of microbiological findings in PF or sputum; (2) pleural ADA <35U/l; (3) the presence of pulmonary opacities on chest X-ray, and (4) exitus.

In 1994, a database was implemented in our hospital to prospectively include the clinical characteristics of all consecutive adult patients who underwent diagnostic thoracentesis, irrespective of the medical department in which they were treated. For this study, all TPEs included in this database between January 1994 and September 2017 were retrospectively analyzed. Our center is a referral hospital with 440 beds that provides health coverage to 450000 individuals. The study protocol was approved by the local ethics committee (CEIC-1868).

We collected demographic (age and sex) and clinical (presence and duration of fever, human immunodeficiency virus [HIV] serology, tuberculin test, death during treatment) variables, radiological findings (PE size, location and loculation, and pulmonary opacities on chest X-ray), microbiological results (auramine or Ziehl–Neelsen staining, solid and liquid media cultures, and PCR techniques in sputum samples and PF), and PF biochemistry (total and differential white blood cell count, red blood cell count, protein, glucose, lactate dehydrogenase [LDH], and ADA). When more than 1 thoracentesis was performed in the same patient, only the results of the first were considered.

A total of 320 patients with TPE were analyzed, with a mean age of 33 (25–45) years, of whom 244 (70%) were men and 96 (30%) were women. Tuberculosis was classified as confirmed in 131 (41%) and probable in 189 (59%). Nine percent of patients had HIV co-infection and 78% had a positive tuberculin skin test (Table 1). Subjects with a positive tuberculin skin test had a significantly higher median age than those with negative results (37 vs 30 years, P=.03). Fever was recorded in 74% of cases over a median of 7 days (0–15) prior to thoracentesis. Fourteen patients died while receiving TB treatment, mostly in the first 2 months (64%).

This study describes the main clinical, microbiological, radiological and PF characteristics from a large series of patients with TPE, and determines the factors that are associated with a greater microbiological yield, lower levels of ADA in PF, the presence of pulmonary radiological opacities, and mortality.

It has been reported that the use of liquid culture media greatly increases the yield of mycobacterial isolates in TPE.11,12 However, despite the use of solid and liquid media, we only obtained an 18% positive culture rate in PF, a much lower figure than the 63% described by Ruan et al.,12 albeit similar to that of other studies.13 These discrepancies may be explained by factors such as the volume of PF processed (between 4 ml14 and 50ml,15 depending on the series), the type of liquid media used, the performance of a second culture,16 inoculation of PF in culture bottles at the bedside11,14,17,18 or in the laboratory, or the percentage of HIV co-infections (up to 80% in some studies).17 Specifically, in our series samples of only 5ml of PF were processed, with inoculation in specific culture media after the sample was received in the lab; serial PF culture were rarely performed (3%) and the population with HIV co-infection was significantly lower than in other studies.

In this study, factors associated with the isolation of Mycobacterium tuberculosis were HIV co-infection and PF containing proteins <4g/dl, neutrophils >60%, and glucose <40mg/dl. The greater frequency of mycobacterial isolates in patients with HIV co-infection is well known.17,19 In a series of 174 TPEs, PF cultures were positive in 64% of patients with HIV co-infection, compared to 30% of other cases.19 Immune system changes in this population probably favors a greater burden of bacilli in PF. Other series have investigated the relationship between biochemical and microbiological characteristics of PF in subjects with PTE.12,15,20 In this respect, a greater percentage of mycobacterial isolates has been associated with a predominance of neutrophils12,15,20 and low concentrations of glucose15,20 and proteins in PF.12 Neutrophils are predominant in the early stages of tuberculous pleural infection, before active immunity against the bacillus has developed.21 Low levels of glucose in PF have been associated with neutrophilic TPE22 (in our series, 45% of the neutrophilic TPEs had glucose levels <40mg/dl, compared to 6% in lymphocytic TPEs; data not shown), and are consequently associated with frequently positive cultures. Moreover, PTE with low proteins levels can indicate hypoproteinemia caused by malnutrition, also associated with altered immunity.23

With regard to the determination of pleural ADA, various cut-off points have been proposed to identify TPE, depending on the age of the patient.24,25 Tay and Tee24 proposed using an ADA of 72IU/l in patients aged ≤55 years and 26IU/l in patients aged >55 years to maintain a diagnostic sensitivity of 95% in both situations. The cut-off point of 55 years corresponded to the median age of the study population. However, in our series, only 15% of the subjects were older than 55 years. This fact may have influenced the lack of association between age and ADA values. The positive correlation between LDH and ADA in PF was described in a series of 80 TPEs (r=0.47, P<.001).24 Elevated LDH in PF indicates inflammation.26 A greater degree of inflammation probably increases the activation of lymphocytes and, consequently, the production of ADA.

TPE was accompanied by radiological pulmonary opacities in 30% of cases, a similar figure to that reported in previous studies (15%–27%).1 Paradoxically, in line with previous series,15,27 this finding was not associated with a greater frequency of positive sputum cultures in the multivariate analysis (in contrast to the univariate analysis). In pulmonary tuberculosis, the highest rate of positive sputum cultures is obtained in patients with opacities in the upper lobes or cavitations,28,29 manifestations that only were observed in 17% and 1.5% of patients in our study, respectively. In our series, more pulmonary opacities were observed in small TPEs, probably because it is easier to evaluate the lung parenchyma in these situations. The fact that a greater proportion of pulmonary opacities were found in bilateral PTEs may be because these may represent more disseminated disease.

TPE mortality was 6%, slightly lower than in other studies (7%–8%).20,30 Age and comorbidities (including HIV co-infection) were higher among subjects with TPE who died.20 However, in this study, HIV patients died primarily during the period before highly active antiretroviral therapy was fully developed, so the validity of this prognostic factor in TPE is questionable. In fact, in a prospective study of 196 patients with TPE conducted in Cameroon between 2007 and 2010, no prognostic differences were found between patients with or without HIV co-infection.31 On the other hand, low protein levels in PF may reflect hypoproteinemia and malnutrition, factors that carry a poor prognosis in pulmonary tuberculosis.32,33

The main limitation of this study is its retrospective design, with the consequent loss of data. For example, the tuberculin test was not used in 23% of cases; no sputum and PF cultures were performed in 29% and 8% of subjects, respectively, and PCR techniques were not used in 83%. Another potential limitation is the inclusion of a significant percentage of probable TPEs, although the criteria used to define them is widely accepted in the literature. This is rooted in the small number of closed pleural biopsies that are performed in our hospital, and the fact that most study subjects were referred from the Internal Medicine department. Finally, the associations identified between several variables may be valid only under the conditions of this study. Thus, not only has the prognosis for HIV infection improved substantially in present times, but the prevalence of pulmonary opacities in TPE is also much higher in CT studies than in standard X-ray,2,34 although CT is not routinely indicated in the setting of pleural tuberculosis.

In summary, factors such as HIV co-infection and low concentrations of proteins in PF are associated with a greater percentage of microbiological findings and a higher mortality rate. Moreover, older patients had more pulmonary opacities on radiological studies and a higher rate of death. Finally, smaller TPEs are accompanied by lower levels of pleural ADA and more pulmonary opacities. These results require confirmation in prospective studies.

The authors state that they have no conflict of interests.