Pleuritis

Objectives: To evaluate the diagnostic use of pleural fluid adenosine deaminase (ADAPF) levels in tuberculous pleuritis (TBpl), with a special reference to HIV coinfection and a Bayesian analysis.

Methods: We investigated a total of 216 patients with pleural effusion, including 100 with TBpl, 68 with malignant effusion, 6 with transudates, 19 with empyema, 15 with miscellaneous diseases, and 8 with diseases of unknown etiology.

Results: The mean values (SE) of ADAPF were 110 (4.5) U/L in patients with TBpl vs 28 (5.3) U/L in patients with a malignancy, 18 (5.7) U/L in patients with transudates, 13 (2.1) U/L in patients with diseases of unknown etiology, 22 (5.1) U/L in patients with miscellaneous diseases, and 191 (26.3) U/L in patients with empyema (Kruskal-Wallis test, p [is less than] 0.001). The ADAPF level was 110 (4.5) U/L in 37 HIV-positive patients with TBpl vs 114 (4.1) U/L in 52 HIV-negative patients with TBpl (Mann-Whitney U test, p [is greater than] 0.05). A receiver operating characteristic curve identified the best cutoff at 60 U/L, yielding measures for sensitivity (0.95), specificity (0.96), positive predictive values (PPVs; 0.96), and negative predictive values (0.95). A Bayesian analysis showed a posttest probability of PPV ranging from 0.5 to 0.99, resulting from a pretest probability of 0.05 to 0.9. Conclusions: ADAPF is diagnostically useful across the various prevalences of TBpl, and its best diagnositic utility is in areas of intermediate prevalence of the disease. Moreover, the diagnostic value of ADAPF is independent of HIV serologic status. (CHEST 1999; 116:97-103)

Key words: adenosine deaminase; Bayesian analysis; human immunodeficiency virus; pleural effusion; pleural fluid; tuberculous pleuritis

Abbreviations: ADA = adenosine deaminase; ADAPF = pleural fluid adenosine deaminase; LDH = lactate dehydrogenase; MTB = Mycobacterium tuberculosis; 1-NPV = 1-negative predictive value; PPV = positive predictive value; ROC = receiver operating characteristic; TB = tuberculosis; TBpl = tuberculous pleuritis

The global incidence of tuberculosis (TB) has sharply increased, particularly in areas where HIV and TB are both prevalent.[1,2] Tuberculous pleuritis (TBpl) has been noted to be more common in patients coinfected with HIV and TB than in patients without HIV infection.[3] A reliable, rapid diagnostic tool for TBpl needs to be adopted, given that a closed pleural biopsy specimen typically contains granulomas in only 60% of cases,[4] and a culture of Mycobacterium tuberculosis (MTB) takes several weeks to process.

Pleural fluid adenosine deaminase (ADAPF) has been shown to be a useful biochemical marker of TBpl and provides a reliable basis for a treatment decision, particularly in prevalent areas.[5-7] However, the limited available data suggest that ADAPF may be less diagnostically useful for immunocompromised patients with TBpl.[8] Furthermore, reports from Japan[9-11] have shown lower mean values of ADAPF in TBpl, leading to a hypothesis that the test may be less useful for Asians.[12,13]

Moreover, various reports[6-9] have quoted a wide range of cutoff values from 47 to 80 U/L. The appropriate cutoff for ADAPF among Asians has not been adequately defined. Considering that Asia is one of the areas with a high incidence of both HIV and TB, further studies are urgently needed to resolve these issues.

The purpose of this prospective study is to further define the diagnostic role of ADAPF in an Asian country where HIV and TB are prevalent, with a special reference to HIV coinfection and Bayesian analysis.

MATERIALS AND METHODS

From June 1996 to January 1998, 216 consecutive patients with pleural effusion who were admitted to the Central Chest Hospital in Nonthaburi, Thailand, were studied. A detailed history, a thorough physical examination, and a chest radiograph were taken of each patient. A thoracentesis was performed in a standard manner. A closed pleural biopsy was performed using an Abrams needle,[14] and at least three samples were taken from the biopsy.

The pleural fluid was examined macroscopically, cytologically, biochemically, and microbiologically, including acid-fast bacilli staining and cultures for MTB. One piece of each sample of pleural tissue was cultured for mycobacteria. The fluid aliquots were analyzed for protein, erythrocyte count, leukocyte subsets, glucose, and lactate dehydrogenase (LDH). In patients with pulmonary and pleural TB, the sputum was also smeared and cultured for mycobacteria.

The adenosine deaminase (ADA) assay was based on the method described by Giusti.[15] Fluorochrome staining with fluorescence microscopic examination was employed for the identification of tubercle bacilli. Acid-buffered Ogawa media were used for the culture of MTB. The laboratory technician was strictly blinded to the tentative diagnosis of each patient; furthermore, the clinicians were unaware of the ADAPF level when the diagnoses were assigned. After counseling, the patients who had TBpl diagnosed were offered HIV serologic testing (an enzymelinked immunosorbent assay with Western blot confirmation). Blood CD4 measurements were made in HIV-seropositive patients. All the patients were followed-up for at least 6 months.

Diagnostic Classification

For the purposes of this study, a diagnosis of TBpl required one of the following: the presence of granulomas in pleural tissue in the absence of any clinical evidence of sarcoidosis, tularemia, and fungal infection[13]; or a stained acid-fast bacilli or MTB-positive culture from the pleural tissue, pleural fluid, or sputum.

A malignancy was diagnosed when neoplastic pleural tissue and/or fluid cytology were identified.[16] Parapneumonic effusions were diagnosed when the effusion was associated with pneumonia, pulmonary abscess, or bronchiectasis, and empyema if the pleural fluid cultures were positive.[17]

Analysis and Statistical Methods

Data are expressed as mean (SE). Group comparisons were made using the nonparametric Kruskal-Wallis test or the Mann-Whitney U test as appropriate. The [chi square] test was used for the comparison of categorical variables. Data analyses were made by appropriate computer software (SigmaStat; SPSS Inc; Chicago, IL). Statistical significance was assessed at the 5% level.

Following standard definitions, the diagnostic role of ADAPF was assessed in terms of sensitivity and specificity.[18] The best cutoff value was selected by a receiver operating characteristic (ROC) curve.[19] A Bayesian analysis was performed to obtain positive predictive values (PPVs) and 1-negative predictive values (1-NPVs) using the following formulas[18]:

PPV = Pr(S) / [Pr(S) + (1 - Pr)(1 - Sp)] NPV = [(1 - Vr)Sp]/[(1 - Pr)Sp + Pr(1 - S)]

where Pr is prevalence or pretest probability, S is sensitivity, and Sp is specificity. The difference between PPV or 1-NPV and pretest probability is the positive diagnostic gain (G+) or the negative diagnostic gain (G-) of the test.

RESULTS

During the study period, a total of 916 patients with pleural effusion were investigated. A diagnosis of TBpl was made in 100 patients (46%). Of these 100 patients, a pleural biopsy exhibited granulomatous inflammation in 77, a pleural tissue culture was positive in 52, pleural fluid staining was positive in 6, and a pleural fluid culture was positive in 93.

In 68 patients (31%), effusion was malignant. In 40 patients, benign conditions other than TBpl included 19 patients with empyema, 9 with parapneumonic effusion, 1 with a pulmonary embolus, 1 with chylothorax, 1 with hemothorax, 1 with lupus pleuritis, 1 with effusion from a pneumothorax, and 1 with effusion associated with atelectasis. In six patients (four with cardiac failure, one with nephrosis, and one with hepatic cirrhosis), transudates were found. In the remaining eight patients, no definite cause could be ascertained.

Patients and Pleural Fluid Profiles

Of the 100 patients with TBpl (mean age, 43 years old), 66 were men and 34 were women (Table 1). The mean ages for patients with malignancy (56 years old) and transudates (62 years old) were older than the mean ages for the other groups. As anticipated, pleural fluid leukocytes, neutrophils, and LDH levels were higher in the empyema group. Pleural fluid leukocytes, lymphocytes, neutrophils, and eosinophils did not significantly differ among patients with TBpl, malignancy, transudates, and unknown etiologies.

(*) Misc = miscellaneous disease; Trans = transudates. Values are given as mean (SE).

Values of ADAPF

ADAPF levels were significantly higher in patients with TBpl than in other patients, except in those with empyema (Fig 1). The mean values (SE) of ADAPF were 110 (4.5) U/L in patients with TBpl vs 28 (5.3) U/L in patients with a malignancy, 18 (5.7) U/L in patients with transudates, 13 (2.1) U/L in the group of patients with diseases of unknown etiology, 22 (5.1) U/L in patients with other benign conditions (the group with miscellaneous diseases), and 191 (26.3) U/L in patients with empyema (p [is less than] 0.001).

Diagnostic Use of ADAPF

The ROC curve identified 60 U/L as the best cutoff value (Fig 2). The sensitivity of ADAPF in patients with TBpl was 0.95. Excluding empyema, which is clinically distinguishable, specificity was 0.96, PPV was 0.96, and 1-NPV was 0.95.

[Figure 2 ILLUSTRATION OMITTED]

Bayes' Theorem

The sensitivity of 0.95 and the specificity of 0.96 were employed to calculate the posttest probabilities of the patient having or not having TBpl for all of the pretest probabilities or prevalences. Figure 3 shows (for ADAPF in diagnosing TBpl) that a pretest probability of 0.05 to 0.9 results in a posttest probability PPV of 0.5 to 0.99. The posttest PPV is 0.89 for the pretest probability of 0.3. The posttest PPV is [is greater than or equal to] 0 .92 for the pretest probability of [is greater than or equal to] 0.4. The maximum gain for a positive result (G+) on the test is obtained with a pretest probability of 0.2 (PPV, 0.82).

[Figure 3 ILLUSTRATION OMITTED]

Profiles Between HIV-Seropositive and HIV-Seronegative Patients With TBpl

Among the 100 patients with TBpl, 37 were HIV seropositive and 52 were HIV seronegative; the remaining 11 patients declined anti-HIV testing. The HIV-positive patients (mean age, 32 years old) were younger than the HIV-negative patients (mean age, 42 years old; p = 0.008; Table 2). There were more men in the HIV-positive group (30 of 37 patients) than in the HIV-negative group (29 of 52 patients; p = 0.02). The patients coinfected with HIV (70%) manifested combined pleuropulmonary TB more frequently than the HIV-negative patients (17%; p [is less than] 0.001). Blood CD4 was 168 (21) cells/[micro]L in the HIV-positive patients. The mean blood lymphocyte count was significantly lower in the HIV-positive group (1,509) than in the HIV-negative group (2,515; p = 0.02). Likewise, the mean pleural fluid leukocyte count was lower in the HIV-positive group (1,372 cells/ [micro]L) than in the HIV-negative group (4,634 cells/ [micro]L; p = 0.01). However, there was no significant difference in the pleural fluid percentage of lymphocytes, neutrophils, eosinophils, LDH, protein, or glucose between the HIV-positive group and the HIV-negative group.

Table 2--Clinical and Pleural Fluid Profiles of Patients With Tuberculous Pleuritis According to HIV Serologic Status(*)

(*) All cellular and biochemical parameters are from pleural fluid except for blood lymphocytes and CD4. Values are given as mean (SE), unless otherwise indicated.

ADAPF Between HIV-Seropositive and HIV-Seronegative Patients With TBpl

The ADAPF level among the HIV-positive patients, 110 (4.5) U/L, did not differ from the ADAPF level among the HIV-negative patients, 114 (4.1) U/L. When applying the cutoff at 60 U/L, the sensitivity was 0.95 in both the HIV-positive group and the HIV-negative group (Fig 4).

[Figure 4 ILLUSTRATION OMITTED]

DISCUSSION

We have evaluated the diagnostic role of ADAPF in TBpl, particularly with reference to HIV coinfection, and a suitable cutoff for ADAPF has been further delineated. Based on a Bayesian analysis, the usefulness of ADAPF has been explored in terms of a diagnostic gain across the prevalence of TBpl. The results have substantiated the diagnostic value of ADAPF in an Asian country where both TB and HIV infections are prevalent. Several practical points arising from our study merit discussion.

To our knowledge, a study of the diagnostic value of ADAPF specifically for HIV-infected patients with TBpl is novel. In a study by Hsu et al[8] of 10 immunocompromised hosts with TBpl, the sensitivity of ADAPF was only 0.40 (based on a cutoff at 80 U/L), leading to a premonition that ADAPF was less diagnostically useful in immunocompromised hosts.[8] In sharp contrast, our study showed a sensitivity of 0.81 at the cutoff of 80 U/L in 37 HIV-infected patients (Fig 4). Furthermore, the sensitivity rose to 0.95 in both the HIV-positive group and the HIV-negative group at the cutoff of 60 U/L, which was more appropriate in our study.

In a study by Niwa et al[9] of 28 patients with TBpl, the mean ADAPF level was only 42.9 U/L and the sensitivity at the cutoff of 60 U/L was only 0.14 (4 of 28 patients), leading to the notion that ADAPF may be less useful for Asians.[12,13] On the contrary, our study has proved ADAPF to be extremely useful, yielding a sensitivity of 0.95 at the cutoff 60 U/L. The discrepancy of the results between the two studies is most likely attributed to the different methods of ADA analysis. Niwa et al[9] employed the method described by Hayashi et al,[20] whereas the method of Giusti[15] was used in our study. Therefore, when interpreting results, physicians should be aware of the different cutoff levels that can occur with the differing methods of ADAPF analysis.

Previous reports[5-9] have quoted a wide range of cutoff values (from 47 to 80 U/L) of ADAPF. For patients with TBpl, some investigators[5,7] employed the lower range of ADAPF as a cutoff. It is noteworthy that in our study, the ADAPF level was [is less than] 47 U/L in 4 of 100 patients with TBpl and the same measure ([is less than] 47 U/L) was seen in a considerable number of patients studied by Burgess et al.[6] This contrasts with the study by Valdes et al[5] in which none of the 81 patients with TBpl had ADAPF levels [is less than] 47 U/L. This difference likely reflects the differences in the study population and the diagnostic criteria for TBpl since all three studies reportedly employed a similar method of ADAPF analysis.

Choosing a cutoff value for a diagnostic test is often not a simple task. We opted to employ an ROC curve, a graphic approach preferable when there are many possible cutoff values.[19] Based on the ROC curve, the ADAPF level at 60 U/L is the most suitable cutoff, yielding a very high sensitivity (0.95) and specificity (0.96). Nonetheless, one always has to consider the consequences of misinterpreting a diagnostic test, and an overdiagnosis may be less harmful than an underdiagnosis in cases of TBpl. On these grounds, it may be argued that an ADAPF level as low as 40 in our study may be a better cutoff; however, this cutoff merely increases the sensitivity to 0.97 but lowers the specificity to an unacceptable level of 0.83.

More importantly, the use of an ROC curve in the present study permits a meaningful comparison of the sensitivity and specificity at different cutoff values between our study and other studies.

A diagnostic test cannot be interpreted fully with information limited to sensitivity and specificity. The prior estimate of the likelihood of the disease (the pretest probability or prevalence) must also be taken into account.[21] Therefore, we decided to utilize a Bayesian analysis (a formula that to our knowledge has received no attention in the literature on ADAPF). The distance between the posttest curves and the diagonal expresses the positive and negative gains after the test results. It should be mentioned that the prevalence in this context refers to the prevalence of TB among patients with pleural effusions, not the prevalence of TB in the population.

Our findings show that ADAPF is useful across the prior probability of TBpl from 0.05 to 0.9. The test is highly useful at a prior probability of [is greater than or equal to] 0.3, yielding a posttest PPV of [is greater than or equal to] 0.89. The largest diagnostic gain on the test is obtained with a prior probability of 0.2. The best utility of the test is obtained with an intermediate prior probability (from 0.4 to 0.7) yielding PPV values of [is greater than or equal to] 0.93. Likewise, if the prior probability is [is greater than or equal to] 0.8, the test result does not substantially increase the prior probability of the disease.

In view of the rigorous diagnostic criteria for TBpl in our study, one may have expected some eases to be missed, particularly when considering that none of the eight patients with an unknown diagnosis had an elevated ADAPF. Thus far, the median follow-up among these patients is 13 months (range, 7 to 16 months), and none of the patients have developed TB. Nonetheless, it is very unlikely that the major results on the diagnostic value of ADAPF would have been significantly affected by the diagnostic pitfall, if any, in this small group of patients.

ADA is the enzyme that catalyzes the conversion of adenosine to inosine, and ADA is found in most cells, particularly in lymphocytes.[15] In pleural effusions, ADA levels significantly correlate with the number of CD4+ lymphocytes[22]; therefore, false-positive test results are found in pleural effusions due to lymphomas, adenocarcinomas, rheumatoid pleuritis, and systemic lupus erythematosus.[23,24] In the present study, the ADAPF level was [is greater than or equal to] 60 U/L in four patients with neoplastic effusions (one with lymphoma and three with adenocarcinoma). A further analysis of ADA isoenzymes[25,26] should help differentiate these false-positive results. However, this procedure is highly elaborate and not yet feasible in our laboratory.

As in our study, previous reports[5,6] have found increased ADAPF levels in patients with empyema, wherein the immune response involves polymorphonuclear cells and macrophages rather than lymphocytes. Therefore, the high ADA activity results chiefly from large phagocytic populations.[27] We are unaware of any other explanation for elevated ADA levels in patients with empyema. Nonetheless, empyema is readily distinguishable on a clinical basis.

The authors, by no means, advocate the use of ADAPF as a replacement for histopathologic or mycobacteriologic study. The use of ADAPF should be viewed as a rapid and accurate method of suggesting a diagnosis of TBpl, thereby expediting the initial decision-making on therapeutic and management plans.

It is noteworthy that reports from Spain,[5,7] Portugal,[22] South Africa,[6] and Thailand (our study) have been very positive on the use of ADAPF. It appears that the diagnostic value of ADAPF is not confined to any particular ethnic group or geographic area. The evidence in the literature strongly suggests that the usefulness of ADAPF is related to the prevalence of TB; therefore, the Bayesian analysis was used in the present study.

In order to standardize the ADAPF determination, laboratories should use the colorimetric method as described by Giusti[15] because the values in previous studies[5-7,22] have been in agreement, albeit with different cutoff levels due largely to the aforementioned reasons.

In conclusion, the present study has substantiated the diagnostic value of ADAPL levels in patients with TBpl in an Asian country where TB is prevalent. The very high diagnostic yield is independent of HIV coinfection. Moreover, the ADA level is useful across a wide range of prevalence of TBpl, with its best diagnostic utility being in areas of intermediate prevalence.

REFERENCES

[1] Sudre P, ten Dam G, Kochi A. Tuberculosis: a global overview of the situation today. Bull World Health Organ 1992; 70:149-159

[2] Hongthiamthong P, Riantawan P, Subhannachart P, et al. Clinical aspects and treatment outcome in HIV-associated pulmonary tuberculosis: an experience from a Thai referral center. J Med Assoc Thai 1994; 77:520-525

[3] Frye MD, Pozsik CJ, Sahn SA. Tuberculous pleurisy is more common in AIDS than in non-AIDS patients with tuberculosis. Chest 1997; 112:393-397

[4] Levine H, Metzger W, Lacera D, et al. Diagnosis of tuberculous pleurisy by culture of pleural biopsy specimen. Arch Intern Med 1970; 126:269-271

[5] Valdes L, Alvarez D, San Jose E, et al. Value of adenosine deaminase in the diagnosis of tuberculous pleural effusions in young patients in a region of high prevalence of tuberculosis. Thorax 1995; 50:600-603

[6] Burgess LJ, Maritz FJ, Le Roux I, et al. Use of adenosine deaminase as a diagnostic tool for tuberculous pleurisy. Thorax 1995; 50:672-674

[7] Ocana I, Martinez-Vazquez JM, Ribera E, et al. Adenosine deaminase activity in the diagnosis of lymphocytic pleural effusions of tuberculous, neoplastic and lymphomatous origin. Tubercle 1986; 67:141-145

[8] Hsu WH, Chiang CD, Huang PL. Diagnostic value of pleural adenosine deaminase in tuberculous effusions of immunocompromised hosts. J Formos Med Assoc 1993; 92:668-670

[9] Niwa Y, Kishimoto H, Shimokata K. Carcinomatous and tuberculous pleural effusions: comparison of tumor markers. Chest 1985; 87:351-355

[10] Aoki Y, Katoh O, Nakanishi Y, et al. A comparison study of IFN-gamma, ADA, and CA 125 as the diagnostic parameters in tuberculous pleuritis. Respir Med 1994; 88:139-143

[11] Tamura S, Nishigaki T, Moriwaki Y, et al. Tumor markers in pleural effusion diagnosis. Cancer 1988; 61:298-302

[12] Light RW. Diagnostic principles in pleural disease. Eur Respir J 1997; 10:476-481

[13] Light RW. Tuberculous pleural effusions. In: Light RW, ed. Pleural diseases. 3rd ed. Baltimore, MD: Williams & Wilkins, 1995; 154-166

[14] Abrams LD. New inventions: a pleural biopsy punch. Lancet 1958;1:30-31

[15] Giusti G. Adenosine deaminase. In: Bergmeyer HU, ed. Methods of enzymatic analysis. New York, NY: Academic Press, 1974; 1092-1099

[16] Sahn SA. Malignant pleural effusions. In: Fishman A, ed. Pulmonary diseases and disorders. 2nd ed. New York, NY: McGraw Hill, 1988; 2159-2170

[17] Light RW. Parapneumonic effusions and empyemas. In: Light RW, ed Pleural diseases. 3rd ed. Baltimore, MD: Williams & Wilkins, 1995; 129-153

[18] Armitage P, Berry G. Statistical methods in medical research. 3rd ed. Oxford, UK: Blackwell Science, 1994; 448-460

[19] Altman DG. Practical statistics for medical research. London, UK: Chapman & Hall, 1993; 417-418

[20] Hayashi R, Ishihara Y, Kitamura S, et al. Measurement of adenosine deaminase (ADA) activity in pleural effusion with special reference to carcinomatous and tuberculous pleuritis. Jpn J Thorac Dis 1981; 19:35-39

[21] Griner PF, Mayewsky RJ, Mushlin AI, et al. Selection and interpretation of diagnostic tests and procedures: principles of test interpretation. Ann Intern Med 1981; 94:565-570

[22] Baganha MF, Pego A, Lima MA, et al. Serum and pleural adenosine deaminase: correlation with lymphocytic populations. Chest 1990; 97:605-610

[23] Valdes L, San Jose E, Alvarez D, et al. Diagnosis of tuberculous pleurisy using the biologic parameters adenosine deaminase, lysozyme, and interferon gamma. Chest 1993; 103:458-465

[24] Ocana I, Ribera E, Martinez-Vazquez JM, et al. Adenosine deaminase activity in rheumotoid pleural effusion. Ann Rheum Dis 1988; 47:394-397

[25] Gakis C. Adenosine deaminase (ADA) isoenzymes ADA1 and ADA2: diagnostic and biological role. Eur Respir J 1996; 9:632-633

[26] Valdes L, San Jose E, Alvarez D, et al. Adenosine deaminse (ADA) isoenzyme analysis in pleural effusions: diagnostic role, and relevance to the origin of increased ADA in tuberculous pleurisy. Eur Respir J 1996; 9:747-751

[27] San Jose E, Valdes L, Sarandeses A, et al. Diagnostic value of adenosine deaminase and lysozyme in tuberculous pleurisy. Clin Chim Acta 1992; 209:73-81

(*) From the Divisions of Medicine (Drs. Riantawan and Chaowalit) and Pathology and Biochemistry (Dr. Wongsangiem and Mr. Rojanaraweewong), Central Chest Hospital, Department of Communicable Disease Control, Ministry of Public Health, Nonthaburi, Thailand.

Manuscript received June 12, 1998; revision accepted December 7, 1998.

Correspondence to: Pratheep Riantawan, MSc, MD, Central Chest Hospital, 39 Tiwanon Rd, Nonthaburi, 11000 Thailand

COPYRIGHT 1999 American College of Chest Physicians
COPYRIGHT 2000 Gale Group