Abstract Objective: The incidence of tuberculosis worldwide has emphasized the need for better assays designed to diagnose the disease, principally the extrapulmonary form. The objective of the present study was to validate the performance of an automated method for the determination of adenosine deaminase (ADA) activity in pleural fluid (PF) and cerebrospinal fluid (CSF), comparing it with a conventional method (the modified Giusti method). Methods: In total, 134 samples were selected from among those tested in our laboratory: 94 PF samples and 40 CSF samples. The ADA activity was determined using the two methods. Inter- and intra-assay precision was determined, linear regression analysis was performed, simple concordance tests were conducted, and the means of the differences were calculated. Results: The correlation coefficients for PF and CSF samples were, respectively, 0.96 and 0.95. Inter-assay precision was determined using 21 replicates at 3 different activity levels: low, medium and high. The percentage coefficient of variation (%CV) was, respectively, 5.9, 8.1 and 5.8 for PF samples, compared with 21.9, 18.6 and 13.8 for CSF samples. Intra-assay precision in %CV was 1.3 and 11.7, respectively, for PF and CSF samples. The concordance between the methods in PF and CRF samples was, respectively, 96.8% and 100%, considering the reference values for the diagnosis of TB to be 40 U/L (conventional) and 30 U/L (automated) in PF samples, versus 9 U/L (for both methods) in CSF samples. Conclusions: The results validate the use of the automated method of determining ADA activity in PF and CSF samples as an alternative to the conventional method.

Keywords: Adenosine deaminase; Tuberculosis/diagnosis; Pleural effusion; Cerebrospinal fluid.