Acta Tropica 99 (2006) 67–74
Adenosine deaminase and tuberculous pericarditis—A systematic review with meta-analysis Felipe Francisco Tuon ∗ , Marcelo N´obrega Litvoc, Max Igor Banks Ferreira Lopes Infectious and Parasitic Diseases Clinic, Hospital das Cl´ınicas, Faculdade de Medicina da Universidade de S˜ao Paulo, Avenida Dr. En´eas de Carvalho Aguiar 255, 4 Andar. Sala 4028—ICHC, Cerqueira Cesar, CEP Number 05403-010 S˜ao Paulo, Brazil Received 15 February 2006; received in revised form 29 June 2006; accepted 23 July 2006 Available online 6 September 2006
Abstract Backgrounds: Adenosine deaminase (ADA) activity in pericardial fluid is a valuable aid in the diagnosis of tuberculous pericarditis (TP), but there is no systematic review performed to evaluate the benefits of ADA activity as an adjunctive test for TP diagnosis. The objective of this systematic review was to evaluate the utility of ADA activity as a diagnostic marker of TP on patients presenting with pericardial effusion. Methods: MEDLINE, LILACS and Cochrane Library databases (1980–2005) searches to identify articles related to adenosine deaminase activity on TP diagnosis. Articles with patients with at least one TP diagnostic criteria were included. The controls were patients with other pericardial diseases with moderate or large pericardial effusion. To calculate the sensitivity, specificity, as well as positive and negative likelihood ratios we extracted the total number of confirmed TP cases over all patients with pericardial effusion as well as the number of cases with ADA activity values of 40 U/L and over. Results: Thirty one studies met our initial inclusion criteria and five articles were selected. The heterogeneity limited the specificity analysis (p = 0.004). The method yielded a sensitivity and specificity of 88% and 83%, respectively. The SROC curve presented an area with a tendency towards 1 (value of 0.9539) and corroborates the diagnostic value of ADA activity. Conclusions: The present study confirms the clinical value of ADA activity as adjunctive diagnostic marker of TP among other causes of pericardial effusion. © 2006 Elsevier B.V. All rights reserved. Keywords: Mycobacterium tuberculosis; Tuberculosis; Pericarditis; Tubeculous pericarditis; Pericardial tuberculosis; Adenosine deaminase; Pericardial effusion
1. Introduction Abbreviations: TP, tuberculous pericarditis; TST, the tuberculin skin test; PE, pericardial effusion; ADA, adenosine deaminase; BAAR, alcohol-acid resistant bacilli; DOR, diagnostic odds ratio; CI, confidence interval; SROC, summary receiver operating characteristic; LR+, positive likelihood ratio; LR−, negative likelihood ratio ∗ Corresponding author. Tel.: +55 11 30696530; fax: +55 11 30697508. E-mail addresses:
[email protected] (F.F. Tuon),
[email protected] (M.N. Litvoc),
[email protected] (M.I.B.F. Lopes). 0001-706X/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.actatropica.2006.07.004
Tuberculous pericarditis (TP) is a rare form of extrapulmonary tuberculosis (Trautner and Darouiche, 2001) that is frequently associated with high mortality and serious short and long term complications especially when the diagnosis and the correct treatment are delayed (Cherian, 2004). Survival time without treatment is approximately 4 months and even when on treatment the mortality varies between 3 and 40% (Yang et al., 2005). The leading complications are: constrictive pericarditis
68
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
with or without restrictive cardiac insufficiency, and cardiac tamponade, which is associated with high mortality if not detected and properly managed (Fowler, 1991). The lack of specificity of TP symptoms makes difficult the correct diagnosis, therefore clinical evaluation offers little help to ascertain the correct diagnosis (Cherian, 2004). The tuberculin skin test (TST) is generally positive in areas of high TB incidence and in certain patients groups like immunosuppressed patients this test may be misleading (Tuon, 2004). Thorax radiography can usually demonstrate an increase in the cardiac area but if not associated with the parenchymal alterations suggestive of tuberculosis, this exam is associated with a low positive predictive value of TP (Reuter et al., 2005a). Transthoracic echocardiogram is a useful test to detect and quantify the pericardial effusion as well as a valuable tool to guide the pericardiocentesis (Liu et al., 2001). Another procedure of interest is the video-assisted thoracoscopy, which permits direct inspection of thoracic cavity and pericardium surface as well as the guidance of diagnostic biopsies and pericardial effusion drainage if indicated (Tuon, 2006). Pericardial effusion (PE) in TP is generally an exudate with a lymphocytic pattern of cellularity, although several times it can be either neutrophilic or mixed (Reuter et al., 2005b). Acid fast bacilli stains on PE sediment are generally negative and culture sensitivity no greater than 50% (Hakim et al., 2000; Becit et al., 2005; Cegielski et al., 1997). Due to the difficulty to establish the diagnosis of TP using clinical, radiological, cytological and even microbiologic evaluation, attempts to establish TP diagnosis using other markers have been pursued (Blake and Berman, 1982). Among those markers the measurement of adenosine deaminase (ADA) activity is one of most studied and should be highlighted due to its good accuracy for values of 40 U/L and over. ADA is an enzyme required for the conversion of adenosine to inosine. It is found in any tissue, but the largest concentration is in the lymphoid tissue, mainly in T-lymphocytes (Blake and Berman, 1982). Fundamental for the immune system, ADA deficiency is a cause of T-cell deficiency (Giblett et al., 1972). Several forms of ADA with different molecular weights exist, and some authors speculate that some of these isoenzymes could be more specific for tuberculosis (Miller, 1995). High ADA activity in tuberculosis appears to be indirectly related to the subsets of T cell lymphocytes and especially to those activated lymphocytic populations involved in the antigenic response to tuberculous bacilli (Lee et al., 2001). However in clinical practice, there is no difference in the use of total ADA or an isoform of ADA (Kataria and Khurshid, 2001).
Even though several studies suggest a valuable role of ADA activity as a marker for TP diagnosis, no systematic review has been published to synthesize the data in this issue. The objective of this systematic review was to evaluate the utility of ADA activity as a diagnostic marker of TP on patients presenting with pericardial effusion. 2. Materials and methods 2.1. Search strategy A systematic search was performed on the medical literature. The databases included MEDLINE and LILACS (January, 1980–August, 2005), as well as the Cochrane Library database until 2005. The search terms were “tuberculo*”, “pericardi*” and “adenosine deaminase” (* means the inclusion of all words with the preceding radical). Bibliographies from the included studies were reviewed as well. 2.2. Study selection The studies were initially selected by two authors (F.T. and M.L.) independently. Disagreements were resolved by consensus. A list of excluded studies and log of reasons for exclusion are available from the authors upon request. The selected articles had to be written in English. Data on ADA activity in PE from TP patients and control groups were necessary to calculate the sensitivity and specificity. The method used for ADA activity had to be the same in all the studies, avoiding the measurement bias. The cutoff value for ADA activity of 40 U/L was selected because this was the value used on the majority of the studies and seems to offer the best diagnostic threshold. When different cutoff values were used, the respective authors were contacted to acquire data regarding a cutoff value of 40 U/L. Only prospective studies were considered. To exclude repeated casuistics, authors were contacted when articles were from the same medical service with overlapping time intervals otherwise the publication with the largest casuistic was considered. 2.3. Data extraction The cases included patients with at least one diagnostic criterion for TP: (1) Mycobacterium tuberculosis positive in PE or tissue culture (gold standard); (2) histopathological exam of pericardial with granulomas containing alcohol-acid resistant bacilli (BAAR); (3) granulomas in pericardial tissue associated with active tuberculosis in another site (Trautner and Darouiche,
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
2001; Cherian, 2004; Yang et al., 2005). Patients with clinical and/or laboratory evidence of TP that had clinical improvement after empirical treatment for tuberculosis were also included as TP cases even if they lack the diagnostic criteria described above because there are no significant difference in the ADA activity level between the pericardial fluid in group of confirmed TP and the group of strongly suspected pericardial tuberculosis (Koh et al., 1997). The control groups of each article included in this meta-analysis were composed by patients with other pericardial diseases with moderate to large PE, excluding those whose effusion were cardiac surgery related (e.g. after revascularization surgery). The data used for the analysis were the number of true positives (TP with ADA ≥ 40 U/L), true negatives (controls with ADA < 40 U/L), false positives (controls with ADA ≥ 40 U/L) and false negatives (TP with ADA < 40 U/L).
69
2.4. Statistical analysis The validation of this meta-analysis was based on guidelines to guarantee quality, as well as good article selection and data extraction (Deville et al., 2002; Oxman and Guyatt, 1991; Jadad and McQuay, 1996). The χ2 -test was performed to determine the variability between studies (heterogeneity) and the significance was established at p < 0.05. The quality of the articles was classified as low, moderate or high according to criteria previously described and modified (Pai et al., 2004a). The sensitivity, specificity, positive predictive value, negative predictive value and diagnostic odds ratio (DOR) of each study were calculated with a confidence interval (CI) of 95%. A graph of the SROC (summary receiver operating characteristic) curve was constructed from the sensitivity and specificity data of each article (Walter, 2002). Positive likelihood ratio (LR+) and the
Fig. 1. Study selection process.
70
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
3. Results
Four publications were from the same medical services, two from one service and two from another, time interval and patient overlap occurred causing the elimination of two articles. Among articles with different previous cutoff value from 40 U/L, two were excluded, since data with a cutoff level of 40 U/L was not provided by these authors. Five studies were selected with a final number of 462 patients (Reuter et al., 2005b; Lee et al., 2002; Martinez-Vazquez et al., 1986; Koh et al., 1997; Dogan et al., 1999) (Fig. 1).
3.1. Description of studies included
3.2. Study validity and data quality
The authors found 1053 studies and excluded 1022 during the initial screening. Of the 31 remaining articles selected, two were not written in English. Twenty articles were not prospective reports or did not provide sufficient data for sensitivity and specificity calculations.
All the studies included were prospective with consecutive data. The Giusti technique for measuring ADA activity through calorimetry was uniformly used in all the selected studies. One study, of the moderate quality classification included less than 10 patients with
negative likelihood ratio (LR−) were included with the aim of improving the significance of the ADA activity, from a certain pre-test probability, in a clinical investigation. The determination of publication bias was analyzed using the Egger test (Pai et al., 2004b). The software program Meta-Disc 1.1.1 was used to assist in the calculations, graph construction and determination of heterogeneity.
Table 1 The quality of the works used in the meta-analysis of ADA activity in tuberculous pericarditis Author (year), country
T+/total Weight Quality cases of (%) tuberculosisa
Prospective Consecutive or random sample
>10 patients with tuberculosis
ADA measure methodology
Case control
Lee et al. (2002), Korea Koh et al. (1997), Korea Reuter et al. (2005a,b), South Africa Dogan et al. (1999), Turkey Martinez-Vazquez et al. (1986), Spain
10/12 18/21 140/162 24/24 3/3
5.41 9.46 72.97 10.81 1.35
Yes Yes Yes Yes Yes
Yes Yes Yes Yes No
Giusti Giusti Giusti Giusti Giusti
No No No No No
Total
195/222
100.00
a
High High High High Moderate
Yes Yes Yes Yes Yes
T+, true positives.
Table 2 Sensitivity, specificity, PPV and NPV of ADA activity in tuberculous pericarditis Author (year), country
T+/total Sensitivity casesa of IC (95%) tuberculosis
Specificity IC (95%)
PPVb IC (95%)
NPVc IC (95%)
DORd IC (95%)
Lee et al. (2002), Korea Koh et al. (1997), Korea Reuter et al. (2005a,b), South Africa Dogan et al. (1999), Turkey Martinez-Vazquez et al. (1986), Spain Total
10/12
0.83 (0.50–0.97)
0.78 (0.64–0.87)
0.45 (0.25–0.67)
0.95 (0.83–0.99)
17.91 (3.44–93.06)
5.41
18/21
0.85 (0.62–0.96)
0.86 (0.68–0.95)
0.81 (0.59–0.94)
0.89 (0.71–0.97)
39 (7.77–195.71)
9.46
140/162
0.86 (0.79–0.91)
0.81 (0.70–0.89)
0.91 (0.85–0.95)
0.72 (0.61–0.81)
28.39 (13.4–60.15)
72.97
24/24
1.00 (0.82–1.00)
0.78 (0.63–0.88)
0.70 (0.52–0.84)
1.00 (0.88–1.00)
170 (9.53–3043)
10.81
3/3
1.00 (0.31–1.00)
1.00 (0.88–1.00)
1.00 (0.31–1.00)
1.00 (0.88–1.00)
525 (8.964–30746.3)
1.35
195/222
0.88 (0.82–0.91)
0.83 (0.78–0.88)
0.83 (0.77–0.87)
0.88 (0.83–0.91)
32.24 (17.55–59.20)
100.00
a b c d
T+, true positives. Positive preditive value. Negative preditive value. Diagnostic odds ratio.
Weight (%)
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
71
TP, however, this study corresponded to less than 2% of the total number of patients. The remaining studies were classified as high quality, allowing adequate analysis performance (Table 1). Heterogeneity analysis calculated for sensitivity was 7.59 (p = 0.108) and for specificity was 15.63 (p = 0.004), demonstrating a greater dispersion of values for specificity. The heterogeneity values of LR+ and LR− were 4.63 (p = 0.327) and 2.38 (p = 0.667), respectively. The sensitivity values were shown to be homogenous, as were LR+ and LR−, suggesting reliable calculation (Fig. 3).
Fig. 2. Publication bias assessment plotted in log DOR.
3.3. Results of the variables The sensitivity of the method was 88% and specificity was 83%, positive predictive value was 83% and negative predictive value was 88% (Table 2). These val-
Fig. 3. Sensitivity, specificity and DOR plotted on Forest graphs for ADA activity in tuberculous pericarditis. The central point determines the mean and the line determines the extremes with a confidence interval of 95%. The lozenge in the lower portion of each graph represents the sum of the mean of the works and the respective CI of 95%.
72
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
Fig. 4. SROC curve graph for ADA activity in pericardial liquid of tuberculous pericarditis.
ues are high and demonstrate good accuracy of ADA activity with a cutoff value of 40 U/L for TP diagnosis. The result for LR+ was 5.3 and for LR− was 0.14. These data, when transferred to a Bayesian normogram, demonstrate the enormous value for post-test probability in TP diagnosis with an ADA activity ≥40 U/L. The area under the curve from SROC curve presents a tendency towards 1 (0.9539) reaffirming the high values of sensitivity and specificity and corroborating the diagnostic value of ADA activity (Fig. 4). Two publications include cases that followed other than confirmed criteria of TP (patients with clinical and/or laboratory evidence of TP that had clinical improvement after empirical treatment), but the number of patients did not exceed 25% of total patients in the largest study and the ADA activity level was not different from confirmed TP. 3.4. Publication bias The evaluation of publication bias, using the Egger test, demonstrated that there was no statistical difference among the studies, despite the small number of articles (p = 0.1433) and a non-significant statistical value corroborates the absence of publication bias (Fig. 2). 4. Discussion Tuberculous pericarditis is a rare clinical condition in developed countries. In developing countries, the number of TP cases is underestimated due to diagnostic limitations, which require high cost procedures and laboratorial tests not readily available in the majority of services. Considering only pericardial diseases, TP is the
leading differential diagnosis responding for more than 70% of pericardial effusion cases in developing countries (Reuter et al., 2005b). The difficulty on establishment of an etiologic diagnosis on TP compels the search of a reliable, quick and affordable test for TP. Several studies have demonstrated the importance of a cytological pattern, nucleic acid amplification (NAA), lysozyme, ␥-interferon among others, but the most studied up to now has been ADA activity (Aggeli et al., 2000; Lee et al., 2005; MartinezVazquez et al., 1986). Despite the large quantity of studies regarding ADA activity and TP, no meta-analysis was published. The studies included in this meta-analysis demonstrate good correlation and all of them use the same technique for measuring ADA activity. Unfortunately, the total number of articles was limited by the different cutoff limits for ADA activity, languages other than English and overlapping of data (in three articles). The exclusion of these papers may have lead to some kind of selection bias. Another bias that should be considered is the control group selection. Since the majority of studies make no distinction on several subgroups of patients on control groups, there is no way to evaluate any condition that could lead to an unexpected high ADA value on controls that could interfere with the sensibility of ADA activity cutoff (e.g. lymphoma). Only one study disclosed the inclusion of HIV infected patients, the other studies made no references to this disease. This study demonstrated no difference in the level of ADA activity in comparison with non-HIV infected patients. The same study demonstrated that the levels of ADA activity measured 48 h after the introduction of antituberculous therapy were responsible for 40% of the false negatives, which justifies the approach of testing the PE prior to the institution of treatment (Reuter et al., 2005b). Four other studies were not included due to the impossibility of obtaining further data from the authors. The contribution of the studies for evaluation was disproportional, due to the small sample base of the majority of these works and one single, extensive study with the largest number of publications in this area. Reuter’s research group in Africa has been studying TP for a long time, thus, the organization of their data and the quantity of patients in a region of high incidence favored the largest proportion of data originating from this study. This could be considered a bias of sample disproportion in the studies but once there was no difference between the values found by Reuter and by the other studies produced in different countries we consider that there is
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
little chance of distortion due to the weight of each work (Reuter et al., 2005b). Patients with clinical and/or laboratory evidence of TP that had clinical improvement after empirical treatment for tuberculosis were also included as TP cases. Reuter described this group as less than 25% of total cases. The ADA activity of this group of patients should be take with caution and it could compromise the result of several studies. As proportion of this group is less than 25% and Koh et al. demonstrated in three publications that this group of patients had ADA activity level no different from confirmed TP, probably this group did not compromise the end-point of this meta-analysis (Koh et al., 1997). The present study was able to demonstrate that ADA activity has high sensitivity and specificity, proving that it is a useful test to assist the diagnosis of TP. The positive preditive value and negative preditive value were 0.83 and 0.88, respectively. These two data should be should be interpreted cautiously because both values are directly related to the prevalence of the disease in the population tested. When the test is applied in populations with high prevalence of the disease, the positive preditive value increases due to lower proportion of falsepositive results. Therefore, a increased ADA level in a PE of patient living in a country with high prevalence of TP (developing countries) will be more reliable than one performed in a developed country. In countries with low prevelance of TP, a decreased ADA level will be more reliable than a increased level, because the number of true-negative is high. The positive preditive value increases and decreases in accordance with the prevalence of the target condition in the screened population. Thus, unlike sensitivity and specificity, the positive preditive value is not a constant performance characteristic of a screening test. If the target condition is sufficiently rare in the screened population, even tests with excellent sensitivity and specificity can have low positive preditive value, generating more falsepositive than true-positive results (Laniado-Laborin, 2005). This test will not confirms with certainty the TP diagnosis, rather it corroborates to the diagnosis substantially, independent of the predictive pre-test values, which was not evaluated (Mayosi et al., 2005). For this reason we calculated the LR+ and LR−. With these values, the clinician can associate a pre-test probability with ADA activity and verify how much the ADA activity results could rise the probability of TP (Fagan, 1975). The opposite is also true, since the specificity of ADA activity was high. Although ADA activity could be used as a sole method of diagnosis, cultures should be available due to
73
incresing incidence of multidrug resistant tuberculosis (Sharma and Mohan, 2004). In summary, the present study demonstrates the enormous clinical value of a simple and affordable of ADA activity test as a diagnostic auxiliary in tuberculosis pericardial effusions. Large usage of this test will strength these findings and contributes to validate its diagnostic efficacy. Due to low frequency of this disease, multicentric studies must be performed to establish more efficient criteria that differentiate pericardial tuberculosis among all other pericardial diseases.
References Aggeli, C., Pitsavos, C., Brili, S., Hasapis, D., Frogoudaki, A., Stefanadis, C., Toutouzas, P., 2000. Relevance of adenosine deaminase and lysozyme measurements in the diagnosis of tuberculous pericarditis. Cardiology 94, 81–85. Becit, N., Unlu, Y., Ceviz, M., Kocogullari, C.U., Kocak, H., Gurlertop, Y., 2005. Subxiphoid pericardiostomy in the management of pericardial effusions: case series analysis of 368 patients. Heart 91, 785–790. Blake, J., Berman, P., 1982. The use of adenosine deaminase assays in the diagnosis of tuberculosis. S. Afr. Med. J. 62, 19–21. Cegielski, J.P., Devlin, B.H., Morris, A.J., Kitinya, J.N., Pulipaka, U.P., Lema, L.E., Lwakatare, J., Reller, L.B., 1997. Comparison of PCR, culture, and histopathology for diagnosis of tuberculous pericarditis. J. Clin. Microbiol. 35, 3254–3257. Cherian, G., 2004. Diagnosis of tuberculous aetiology in pericardial effusions. Postgrad. Med. J. 80, 262–266. Deville, W.L., Buntinx, F., Bouter, L.M., Montori, V.M., de Vet, H.C., van der Windt, D.A., Bezemer, P.D., 2002. Conducting systematic reviews of diagnostic studies: didactic guidelines. BMC. Med. Res. Methodol. 2, 9. Dogan, R., Demircin, M., Sarigul, A., Ciliv, G., Bozer, A.Y., 1999. Diagnostic value of adenosine deaminase activity in pericardial fluids. J. Cardiovasc. Surg. (Torino) 40, 501–504. Fagan, T.J., 1975. Normograma for Bayes theorem. N. Engl. J. Med. 293, 257. Fowler, N.O., 1991. Tuberculous pericarditis. JAMA 266, 99–103. Giblett, E.R., Anderson, J.E., Cohen, F., Pollara, B., Meuwissen, H.J., 1972. Adenosine-deaminase deficiency in two patients with severely impaired cellular immunity. Lancet 18, 1067–1069. Hakim, J.G., Ternouth, I., Mushangi, E., Siziya, S., Robertson, V., Malin, A., 2000. Double blind randomised placebo controlled trial of adjunctive prednisolone in the treatment of effusive tuberculous pericarditis in HIV seropositive patients. Heart 84, 183–188. Jadad, A.R., McQuay, H.J., 1996. Meta-analyses to evaluate analgesic investigations: a systematic qualitative review of their methodology. J. Clin. Epidemiol. 49, 235–243. Kataria, Y.P., Khurshid, I., 2001. Adenosine deaminase in the diagnosis of tuberculous pleural effusion. Chest 120, 334–336. Koh, K.K., In, H.H., Lee, K.H., et al., 1997. New scoring system using tumor markers in diagnosing patients with moderate pericardial effusions. Int. J. Cardiol. 61, 5–13. Laniado-Laborin, R., 2005. Adenosine deaminase in the diagnosis of tuberculous pleural effusion: is it really an ideal test? A word of caution. Chest 127, 417–418.
74
F.F. Tuon et al. / Acta Tropica 99 (2006) 67–74
Lee, Y.C., Rogers, J.T., Rodriguez, R.M., Miller, K.D., Light, R.W., 2001. Adenosine deaminase levels in nontuberculous lymphocytic pleural effusions. Chest 120, 356–361. Lee, J.H., Lee, C.W., Lee, S., et al., 2002. Comparison of polymerase chain reaction with adenosine deaminase activity in pericardial fluid for the diagnosis of tuberculous pericarditis. Am. J. Med. 15, 519–521. Liu, P.Y., Li, Y.H., Tsai, W.C., et al., 2001. Usefulness of echocardiographic intrapericardial abnormalities in the diagnosis of tuberculous pericardial effusion. Am. J. Cardiol. 87, 1133–1135. Martinez-Vazquez, J.M., Ribera, E., Ocana, I., Segura, R.M., Serrat, R., Sagrista, J., 1986. Adenosine deaminase activity in tuberculous pericarditis. Thorax 41, 888–889. Mayosi, B.M., Burgess, L.J., Doubell, A.F., 2005. Tuberculous pericarditis. Circulation 112, 3608–3616. Miller, K.D., 1995. Adenosine deaminase in pericardial fluids. Which test is best? Chest 108, 1766. Oxman, A.D., Guyatt, G.H., 1991. Validation of an index of the quality of review articles. J. Clin. Epidemiol 44, 1271–1278. Pai, M., Flores, L.L., Hubbard, A., Riley, L.W., Colford Jr., J.M., 2004a. Nucleic acid amplification tests in the diagnosis of tuberculous pleuritis: a systematic review and meta-analysis. BMC. Infect. Dis. 4, 6. Pai, M., McCulloch, M., Gorman, J.D., et al., 2004b. Systematic reviews and meta-analyses: an illustrated, step-by-step guide. Natl. Med. J. India 17, 86–95.
Reuter, H., Burgess, L.J., Carstens, M.E., Doubell, A.F., 2005a. Adenosine deaminase activity—more than a diagnostic tool in tuberculous pericarditis. Cardiovasc. J. S. Afr. 16, 143– 147. Sharma, S.K., Mohan, A., 2004. Multidrug-resistant tuberculosis. Indian J. Med. Res. 20, 354–376. Reuter, H., Burgess, L.J., Doubell, A.F., 2005b. The role of chest radiography in diagnosing patients with tuberculous pericarditis. Cardiovasc. J. S. Afr. 16, 108–111. Trautner, B.W., Darouiche, R.O., 2001. Tuberculous pericarditis: optimal diagnosis and management. Clin. Infect. Dis. 33, 954– 961. Tuon, F.F., 2004. Tuberculin skin test and tuberculous pericardial effusions in: G. Cherian. Diagnosis of tuberculous aetiology in pericardial effusions. Postgrad. Med. J. 80, 262–266, e-letters: http://pmj.bmjjournals.com/cgi/eletters/80/943/262?ck=nck. Tuon, F.F., 2006. Video-assisted thoracoscopy and tuberculous pericarditis. Ann. Thorac. Surg. 81, 2338. Walter, S.D., 2002. Properties of the summary receiver operating characteristic (SROC) curve for diagnostic test data. Stat. Med. 21, 1237–1256. Yang, C.C., Lee, M.H., Liu, J.W., Leu, H.S., 2005. Diagnosis of tuberculous pericarditis and treatment without corticosteroid in a tertiary teacher hospital in Taiwan: a-14 years experience. J. Microbiol. Immunol. Infect. 38, 47–52.
