INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Amplification techniques have attracted considerable interest since they offer the opportunity to shorten the time required to detect and identify Mycobacterium tuberculosis complex (MTB) organisms in both respiratory and extrapulmonary specimens. A number of amplification systems have been described recently; besides in-house assays, commercial systems have been developed with the aim of providing better standardization and reducing contamination. Two systems among those developed by the industry are familiar to those working in mycobacteriology: the Amplified M. tuberculosis Direct Test (Gen-Probe Inc., San Diego, Calif.), which was recently upgraded by the manufacturer (AMTDII), and the ligase chain reaction (Abbott LCx Probe System [LCx]), a semiautomated amplification assay that was recently introduced by Abbott Laboratories Diagnostic Division (Abbott Park, Ill.).

Since the introduction of nucleic acid amplification assays into the field of diagnostic mycobacteriology, many papers describing investigations that have evaluated the performances of in-house or commercial assays in comparison with those of microscopy and culture have been published, but only a few have carried out head-to-head comparisons of different assays. Bearing this in mind, we undertook the present study with the aim of comparing the AMTDII and LCx amplification methods.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Study design. Four hundred fifty-seven clinical specimens consecutively received for culture for acid-fast bacilli (AFB) by four Italian microbiology laboratories (center 1, Ancona; center 2, Pordenone; center 3, Vicenza; center 4, Milan) were entered into this study. All laboratories received an average of 150 to 400 specimens per month for mycobacterial culture. The specimens were almost entirely obtained from patients admitted to hospitals.

Specimen collection and processing. The investigated specimens collected from 357 patients included 170 sputum specimens, 7 bronchoalveolar lavage [BAL] specimens, 96 bronchial washings, 13 gastric aspirates, 54 urine specimens, 68 normally sterile body fluid (pleural, pericardial, and synovial fluids, cerebrospinal fluid [CSF], and ascites fluid) specimens, and 49 miscellaneous samples such as pus and biopsy specimens.

Pretreatment of selected clinical specimens for amplification. (i) Pretreatment of CSF. Equal volumes (0.5 ml) of CSF and NALC-NaOH were added to a sterile tube. After 15 min the specimen was neutralized with PBS and centrifuged at 12,000 × g for 10 min. The pellet was resuspended in 1 ml of PBS and was stored in two aliquots until the amplification techniques were performed.

(ii) Pretreatment of pleural and other sterile fluids. After decontamination with NALC-NaOH, the sediment was washed twice with sterile distilled water before stocking it for the amplification assays.

Culture. The processed sediment was cultivated by the radiometric BACTEC technique (Becton-Dickinson Diagnostic Instrument Systems, Sparks, Md.) and with Löwenstein-Jensen (LJ) medium. At center 4, MB-Redox medium (Biotest AG, Dreieich, Germany) (9) and LJ medium were used for culture for AFB. All media were incubated at 36 ± 1°C and were examined twice a week for 8 weeks. LJ slants were inspected for growth, while the MB-Redox tubes were tested for the appearance of sediment at the bottom and pink to violet particles in the sediment. The acid fastness of the particles was confirmed by the Ziehl-Neelsen staining. The radiometric growth index of the 7H12 vials was recorded by the BACTEC instrument; a growth index of >30 was considered suspect, and smears were made daily to confirm the presence of AFB.

Microscopy. Smears were stained with auramine-rhodamine fluorochrome or by the Ziehl-Neelsen method to detect AFB.

Identification of mycobacteria. Isolates were identified with specific DNA probes (Accuprobe Gen-Probe Inc., San Diego, Calif.) and by standard biochemical tests (10).

Amplification procedures. Amplification assays were run in three separate areas which had been set up in two different rooms.

Gen-Probe AMTDII. The Gen-Probe AMTD assay (Gen-Probe Inc.) was performed by following the instructions supplied by the manufacturer. Each run included positive and negative amplification controls: the former being prepared from 10⁴ and 10⁵ dilutions of a no. 1 McFarland nephelometric standard suspension of M. tuberculosis ATCC 27294, while the latter was made from a similarly prepared suspension of Mycobacterium gordonae ATCC 14470. A cutoff (CO) value of 30,000 relative light units (RLUs) was used for positive specimens. Samples showing weakly positive results (between 30,000 and 300,000 RLUs) were retested and new specimens were obtained from the patients.

Abbott LCx. The LCx assay (Abbott Laboratories Diagnostic Division) was performed according to the manufacturer's instructions. Two calibrators and two negative controls were included in each test run. The results, reported as fluorescence rates, are compared with the calibrator rates. If the sample fluorescence rate exceeded 30% of the average calibrator fluorescence rate (CO value), the sample was considered positive. The ratio of the sample fluorescence rate to the CO value (S/CO ratio) was calculated. Samples with positive results exhibited an S/CO ratio of >1.0.

Clinical evaluation of patients. Clinical assessment of the patients included a medical history, signs, symptoms, chest X ray, microbiological results, and follow-up observations, as well as the results obtained for additional specimens collected during the follow-up. All records considered were reviewed by a tuberculosis expert, enabling us to set the combination of culture and clinical diagnosis with high suspicion as the "gold standard." After this analysis, the amplification results were reclassified as appropriate.

Statistical analysis. Statistical comparisons were calculated by using the chi-square test; a P value of <0.05 was considered significant.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Analytical performance of Gen-Probe AMTDII and Abbott LCx assays. Positive and negative results could be clearly distinguished by the magnitudes of both the RLU values and the fluorescence rates. The majority of samples with positive results had values of >1,500,000 RLUs for AMTDII and fluorescence rates of >1,000 (showing S/CO ratios of >2.5) for LCx. Negative samples exhibited results approximately 20-fold lower than the CO values, which were 30,000 RLUs for AMTDII and fluorescence rates ranging from 300 to 450 for LCx.

Clinical results. (i) Identification of NTM. Twenty-one specimens (six were smear positive) from 19 patients yielded nontuberculous mycobacteria (NTM). The mycobacterial species identified from these specimens were M. avium (n = 12), M. gordonae (n = 4), M. kansasii (n = 2), M. xenopi (n = 2), and M. fortuitum (n = 1).

(ii) Respiratory specimens. A total of 273 respiratory specimens collected from 205 patients were tested. Altogether, 75 specimens were culture positive for AFB; 61 isolates were found to belong to MTB (51 were from sputum specimens, 9 were from bronchial washings, and 1 was from a BAL specimen), while the remaining 14 strains (11 from sputum specimens, 2 from bronchial washings, and 1 from a BAL specimen) were identified as nontuberculous species. A comparison of the amplification results with the smear and culture results and clinical data is summarized in Table 1. A total of 95 specimens were from patients with a diagnosis of tuberculosis, and 178 were from patients with nontuberculous pulmonary disease on the basis of clinical and microbiological findings. Of the 36 samples which were smear and culture positive, all were AMTDII positive and 33 were LCx positive (the difference was statistically not significant). Twenty-five samples were smear negative for AFB but culture positive; 24 were AMTDII positive, but only 16 were LCx positive. The difference was statistically significant (P = 0.004). Nine samples collected from patients in whom tuberculosis was strongly suspected at that moment were smear and culture negative but were later confirmed to be positive by additional cultures. Of these, five were AMTDII positive and four were LCx positive (the difference was statistically not significant). The cumulative difference for all M. tuberculosis-positive specimens (65 positive by AMTDII and 53 positive by LCx) was statistically significant (P = 0.005). Twenty-five specimens were collected from patients with pulmonary tuberculosis who were receiving drug therapy. All of these specimens were smear and/or culture positive before therapy but were both smear and culture negative at the time that the sample was taken for amplification. Fourteen of these specimens were positive by both assays. Of the 164 samples from patients with nontuberculous pulmonary disease that were smear and culture negative for AFB, 1 was AMTDII positive and 2 were LCx positive. After resolution of discrepant results on the basis of the patients' clinical histories, these samples were considered to be false positive. NTM were grown from 14 specimens; all these specimens were negative by both amplification assays. Table 2 presents the sensitivity, specificity, and predictive values of both amplification methods for smear-positive and smear-negative specimens compared with the results of AFB smear and culture, assuming that the combination of culture and clinical diagnosis is the gold standard. Results for samples from patients with pulmonary tuberculosis receiving drug therapy were not considered in the analysis whose results are presented in Table 2.

(iii) Extrapulmonary specimens. A total of 184 extrapulmonary specimens collected from 152 patients were tested. Altogether 31 specimens were culture positive for AFB; 24 isolates were found to belong to the MTB (2 were from gastric aspirates, 11 were from urine, 4 were from sterile body fluids, and 10 were from miscellaneous specimens), while the remaining 7 strains (3 from urine, 3 from sterile body fluids, and 1 from a miscellaneous specimen) were identified as nontuberculous species. A comparison of the amplification results with the smear and culture results and clinical data is summarized in Table 3. A total of 32 specimens were from patients with a diagnosis of extrapulmonary tuberculosis and 152 were from nontuberculous patients on the basis of clinical and microbiological findings. Of the 11 samples which were smear and culture positive, all were AMTDII positive and 9 were LCx positive (the difference was statistically not significant). Thirteen samples were smear negative for AFB but culture positive; 10 were AMTDII positive, but only 5 were LCx positive. The difference was statistically significant (P = 0.047). Four samples collected from patients for whom tuberculosis was strongly suspected clinically were smear and culture negative. Of these, only one was positive by both assays. The cumulative difference for all M. tuberculosis-positive specimens (22 by AMTDII and 15 by LCx) was statistically significant (P = 0.048). Four specimens were collected from patients with extrapulmonary tuberculosis who were receiving drug therapy. All of these were smear and/or culture positive before therapy but were both smear and culture negative at the time that the sample was taken for amplification. Two of them were AMTDII positive and one was LCx positive. Of the 145 samples from patients with nontuberculous disease that were smear and culture negative for AFB, 1 was positive by both assays. After resolution of the discrepant results as reported above, these samples were considered to be false positive. NTM were grown from seven specimens; all these specimens were negative by both amplification assays. Table 4 presents the sensitivity, specificity, and predictive values of both amplification methods for smear-positive and smear-negative specimens compared with the results of AFB smear and culture, assuming that the combination of culture and clinical diagnosis is the gold standard. Results for samples from patients receiving antituberculous chemotherapy were not considered in the analysis whose results are presented in Table 4.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Rapid diagnostic methods, such as amplification methods, significantly decrease the time required for the diagnosis of M. tuberculosis infections. Two of these test methods, the Gen-Probe AMTDII assay and the Abbott LCx assay, were compared in this study. The kits contain all the reagents needed for sample amplification and detection and appeared to fit well in a routine microbiology laboratory's work flow. Both amplification assays exhibited good analytical performances; in particular, the reproducibility of AMTDII seemed to be considerably upgraded compared with that of its former release (2, 3). We admit that a correct evaluation of the reproducibility of AMTDII compared to that of the first version (AMTDII versus AMTD) can be assessed only in comparative evaluations. However, on average, we observed a drastic reduction in the number of weakly positive results, which represented a considerable drawback in the earlier release (11, 12). Generally, differences between CO values, values for positive and negative controls, and values for samples were broad enough to permit easy discrimination. The turnaround time of AMTDII is shorter than that of LCx (<3 versus 6 h), and the AMTDII procedure, although entirely manual, is easier to perform.

The diagnostic performances of the AMTDII and LCx assays for both respiratory and extrapulmonary specimens were noticeably different. In comparison with the gold standard, the sensitivity and specificity of AMTDII were 92.8 and 99.4%, respectively, for respiratory specimens and 78.6 and 99.3%, respectively, for extrapulmonary specimens. Our data agree with those from a recently published study by Gamboa and colleagues (4) showing sensitivities of 94.7 and 86.8% for respiratory and extrapulmonary specimens, respectively, and specificities of 100% for both categories of specimens. In our study, we found a lower sensitivity for extrapulmonary specimens. This can be explained by taking into account the fact that the AMTDII procedure has an increased specimen volume, from 50 to 450 µl. Such a change would lead to an increased number of positive results but may also enhance inhibition by those substances which are more likely to be contained in extrapulmonary specimens (4). AMTDII appeared to be significantly more sensitive than LCx for both respiratory (P = 0.005) and extrapulmonary (P = 0.048) specimens. The difference was found to be significant within the category of smear-negative and culture-positive samples (Tables 1 and 3). The sensitivity and specificity of LCx were 75.7 and 98.8%, respectively, for respiratory specimens and 53.6 and 99.3%, respectively, for extrapulmonary specimens. Data on the sensitivity of LCx presented in the literature are different: our data conflict with those of Ausina et al. (1), Lindbråthen et al. (7) (90.8 and 90.2% sensitivities for respiratory specimens, respectively), Tortoli et al. (13) (98.7 and 73.3% sensitivities for respiratory and extrapulmonary specimens respectively), and Gamboa et al. (5) (78.5% sensitivity for extrapulmonary specimens) but are very close to those of Moore and Curry (8) (77% sensitivity for respiratory specimens) and Yuen et al. (14) (36.8% sensitivity for smear-negative respiratory specimens). Discrepancies among sensitivities can be explained at least to some extent by analyzing the ratios of smear-positive specimens/smear-negative specimens for the MTB-yielding specimens enrolled in those studies. We found higher ratios in papers from investigators reporting better sensitivities (2.9, 4.5, and 6.4 for Lindbråthen et al. [7], Ausina et al. [1], and Tortoli et al. [13], respectively), while ratios from investigators who detected less satisfactory sensitivities were somewhat or considerably lower (2.6, 0.6, and 1.2 for Yuen et al. [14], Moore and Curry [8], and our study, respectively).

Negative results obtained by the LCx amplification assay for smear-positive, MTB-yielding samples may be explained by the presence of inhibitors of enzymatic amplification. We did not search for the presence of inhibitory substances; however, all smear-positive, LCx-negative samples were retested, and negative results were obtained upon repeat assay. In this context, nucleic acid extraction from mycobacterial cells may also be as important as amplification itself in the performance of commercial assays (1, 14). This step has probably been underestimated, but it deserves further attention for the optimization of sensitivity. For smear-negative, culture-positive samples, a low number of mycobacteria that are unequally distributed in the test suspension is perhaps the more likely explanation. Furthermore, the LCx procedure includes during specimen preparation two washing steps that should remove and eliminate inhibitors. We agree with Ausina and colleagues (1) that these steps may represent a potential drawback, causing sampling errors and consequent false-negative results especially when the number of AFB is low. Of the 13 samples (9 respiratory specimens and 4 extrapulmonary specimens) that were both smear and culture negative and that were obtained from patients strongly suspected of having tuberculosis, 6 were AMTDII positive and 5 were LCx positive, suggesting that AMTDII also appears to be slightly more sensitive than LCx in tests with this category of specimens. During the follow-up, tuberculosis could be confirmed by additional cultures only for those patients from whom respiratory specimens were collected. A small number of samples collected from patients found to be negative for MTB by culture and clinical criteria showed discrepant results (Tables 1 and 3); two of them were positive by AMTDII and three were positive by LCx. We concluded that these specimens had true false-positive results because they gave positive results by one amplification assay only and additional specimens from the patients were also negative by any test.

Of the 29 samples (respiratory and extrapulmonary specimens) obtained from patients receiving standard antituberculous therapy, 16 were AMTDII positive and 15 were LCx positive. Data from the literature demonstrate that, despite differences in the amplification target, neither assay is validated for use in the monitoring of therapeutic efficacy (1, 6, 13). None of the 21 specimens yielding NTM in culture was positive by any assay.

In summary, although at present amplification assays cannot replace the conventional diagnostic techniques (6), AMTDII and LCx were found to be rapid and specific for the detection of MTB in clinical samples. The assay protocols were easy to perform and were suitable for a routine microbiology laboratory's work flow. On the basis of our data, the difference between the sensitivities of AMTDII and LCx was found to be statistically significant for both respiratory specimens (P = 0.005) and extrapulmonary specimens (P = 0.048). Moreover, to significantly increase the sensitivity of the latter assay, systematic control for sample inhibition is recommended.