Comparison of C₁₈-Carboxypropylbetaine and Standard N-Acetyl-l-Cysteine-NaOH Processing of Respiratory Specimens for Increasing Tuberculosis Smear Sensitivity in Brazil

Subjects and setting.Study participants were drawn from the Outpatient and Inpatient Units of the Hospital Universitário Clementino Fraga Filho (HUCFF), Rio de Janeiro, Brazil. The first phase of the study included samples collected between October 1999 and March 2000, and the second phase included samples collected between July and September 2000. Consecutive respiratory samples sent to the laboratory from adult patients with suspected active pulmonary TB were included in the study. The reference standard for the diagnosis of pulmonary tuberculosis was culture growth of M. tuberculosis.

Approval for this study was granted by the Institutional Review Boards of HUCFF, the Brazilian Ministry of Health, and the Johns Hopkins Medical Institutions. Written informed consent was obtained from all study subjects.

Specimen collection.Respiratory specimens submitted to the mycobacteriology laboratory with culture request were included. A total of 236 specimens were analyzed during the initial phase, and an additional 120 specimens were analyzed using the modified protocol. Each specimen with a volume of 5 ml or greater was split and processed by the standard digestion-decontamination method that combines N-acetyl-l-cysteine (NALC) and sodium hydroxide (NALC-NaOH) (3, 10, 15) and by the CB-18 method.

NALC-NaOH processing.The respiratory samples were incubated for 20 min with 2.0% NALC-2.0% NaOH, and then centrifuged at 3,000 × g for 20 min. The samples were decanted, and the sediment was resuspended in the remaining supernatant. All specimens were then stained for AFB and cultured as described below.

CB-18 processing.The CB-18 smear kit provided by the manufacturer (Integrated Research Technology, LLC, Baltimore, Md.) included 20× Tris-citrate buffer, CB-18 stock, and NALC. In the initial manufacturer's protocol, respiratory specimens were treated with a buffered solution of 1 mM CB-18 in Tris-citrate buffer (50 mM Tris-HCl, 12.5 mM citrate [pH 7.6], 1.5 mM NaCl) with 15 mM NALC. The samples were mixed with buffered CB-18 and immediately subjected to centrifugation at 3,000 × g for 20 min at 25°C. The samples were decanted, resuspended in the remaining supernatant, and then used to prepare smears for diagnostic evaluation. In the second phase of the study, the manufacturer introduced modifications that included (i) increasing the concentration of CB-18 to 4 mM, (ii) lowering the pH of the buffer to pH 6.0, and (iii) incubating specimens in the CB-18 solution for 90 min. In addition, in the second phase of the study, samples were initially homogenized in an equal volume of 15 mM NALC-50 mM NaOH-2.9% sodium citrate for 15 min prior to addition of the CB-18 solution. Following the incubation, samples were again subjected to centrifugation at 3,000 × g for 20 min, decanted, and then resuspended in the remaining supernatant for smear analysis.

AFB smears.Smears made from the NALC-NaOH and CB-18 protocols were heat fixed and stained using the Ziehl-Neelsen (ZN) technique for AFB staining. A small number of specimens (85 specimens) processed with CB-18 were also stained with the Kinyoun staining technique as well as ZN. The smears were scanned at ×1,000 magnification with a light microscope. The results were quantified in accordance with published standards (7). For quality assurance, an experienced scientist in the Mycobacteriology Laboratory at HUCFF reread 10% of the negative smears and all the positive smears.

Culture.A portion of each processed sediment was removed for smear analysis. The specimen was then resuspended in 2.5 ml of distilled water. Löwenstein-Jensen slants were inoculated with approximately 0.4 ml of the resuspended sediment and incubated at 37°C for 8 weeks. All specimens that were culture positive for mycobacteria were tested by standard biochemical methods to distinguish M. tuberculosis from nontuberculous mycobacteria, as described elsewhere (7, 13).

Statistical analyses.Analyses were carried out using STATA 6.0 statistical software. Sensitivity, specificity, and positive and negative predictive values were calculated for each method. Fisher's test and McNemar's test were used for comparisons of results. Agreement between the two AFB staining methods was evaluated using the kappa score.

Comparison of processing methods.In the first phase, M. tuberculosis culture growth was identified in 47 samples, and in the second phase another 21 samples were culture positive. The initial CB-18 protocol was slightly less sensitive than routine ZN smears, 59.6% (28 of 47) versus 66.0% (31 of 47) (P = 0.29). During the first phase of the study, there was no significant difference in the positive predictive value between NALC-NaOH decontamination and the CB-18 processing method (89.8 and 88.7%, respectively) using culture as the “gold standard.” The modified CB-18 protocol improved the sensitivity of CB-18 to 71.4% (15 of 21) over that of the routine protocol, which was 61.9% (13 of 21); however, this difference was not significant (P = 0.16). The specificities of both diagnostic techniques (CB-18 and NALC-NaOH) were both 96.8% (183 of 189) in the initial protocol and were 97.0% (96 of 99) and 99.0% (98 of 99), respectively, in the modified protocol. Even with the modification of the CB-18 processing method, the positive predictive value for CB-18 remained close to that of routine ZN due to the high specificity of both methods (Table 1). The modified CB-18 protocol showed an improvement in smear staining compared to the initial protocol, but the number of positive cultures was small (n = 21). The most notable difference between the two versions of CB-18 processing was in the readability of the slides. In the first phase, 14% of smears were not readable versus 1% (P = 0.00) using the modified protocol.

Comparison of AFB staining techniques.A comparison of two staining techniques used in AFB smear analysis was performed on 85 of the samples processed using CB-18. Kinyoun staining was compared to ZN staining to determine if there was any difference that would affect the performance of CB-18 in diagnostic use. There was fair agreement between the two techniques (overall agreement = 72.9%; kappa test = 0.715). However, 22% of Kinyoun slides and 1% of ZN slides were unreadable due to technical problems (Table 2).