In an effort to develop a rapid and specific test for the diagnosis of pulmonary tuberculosis, direct amplification tests (DATs) of Mycobacterium tuberculosis complex (MTBC) DNA have been proposed (1). DATs have proved to be significantly more sensitive than smear examination in detecting mycobacteria in respiratory tract specimens and permit the identification of MTBC in smear-positive samples with a specificity of almost 100% (1). A recent workshop of the American Thoracic Society (ATS) considered that when smear and DAT were both positive, the diagnosis of active tuberculosis could be considered as established (2). Two commercially available DATs have already been approved by the Food and Drug Administration (FDA) in this respect. When the specimens are smear-negative, DATs have shown important limitations. Several large studies have reported that a substantial proportion (30 to 50%) of culture-positive specimens were DAT-negative (1); both sensitivity and specificity of DATs are submitted to important variations from laboratory to laboratory (3); the positive predictive value (PPV) of DATs vary considerably with the disease prevalence (4), and the sensitivity of DATs for the detection of MTBC DNA in specimens of patients with culture-negative pulmonary tuberculosis is unknown (2). Because of these shortcomings, the clinical utility of DATs for the diagnosis of tuberculosis in smear-negative patients is, at present, uncertain (2).

Recently, ligase chain reaction technology has been commercially available and a semiautomatic kit has been developed for the detection of M. tuberculosis complex-specific DNA in clinical specimens (LCx-MTB assay; Abott Diagnostics Division, Chicago, IL) (5, 6). We undertook a study to evaluate the positive and negative predictive value of this new commercially developed DAT for the diagnosis of pulmonary tuberculosis by focusing on patients at high suspicion of active tuberculosis. We further analyzed to what extend the knowledge of a positive LCx result at the time of specimen collection could have hastened the start of an adequate antituberculosis chemotherapy or avoided invasive diagnostic procedures in smear-negative patients. In conformity with the recent recommendations of the ATS workshop for clinical trials carried out to evaluate DATs (2), clinical case definition of tuberculosis according to the ATS classification (7) was used for all patients as reference-standard. All consecutive clinical samples were tested with the exception of specimens provided by patients already receiving antituberculosis therapy. Additionally, in order to assess the sensitivity and specificity of the LCx-MTB test, specimens obtained from control patients were tested.

Patients

Case patients. Consecutive patients referred to the Division of Pulmonary and Critical Care Medicine of the Pitié-Salpêtrière Hospital in Paris, France, between May and December 1996 were considered for active tuberculosis if they were part of a high prevalence group, if they had a past history of tuberculosis, or if they had clinical or radiologic findings suggestive of active tuberculosis. They were considered at high risk for active tuberculosis and included in the study as case patients if their chest radiographs showed either a miliary pattern or a nodular parenchymal infiltration, with or without cavitation, predominantly affecting the upper and posterior segments of the upper lobes, or the superior segments of the lower lobes (8). Patients with a past history of tuberculosis were included as case patients if new nodular infiltrates or cavitations were present on chest radiographs when compared with a reference radiograph. If no reference chest radiograph was available, the patients were considered at high risk for active tuberculosis and included as case patients if the chest CT scan showed clustered, poorly marginated nodules producing the characteristic "tree-in-bud" appearance (8).

Patients already receiving antituberculosis medication at time of inclusion were excluded. Respiratory tract specimens were collected and processed for fluorochrome-stained smear, conventional culture, and LCx-MTB assay. Diagnostic procedures and the start of antituberculosis therapy were decided by the physician in charge of the patient without knowledge of the results of the LCx-MTB tests.

Each patient was then classified according to the ATS classification (7) of tuberculosis (Table 1) at least 3 mo after inclusion by a pulmonologist investigator (S.J.) blinded to the LCx results. Patients considered as having active tuberculosis (TB3) were further included with patients with smear-positive tuberculosis when at least one specimen was smear-positive, patients with smear-negative culture-positive tuberculosis when at least one specimen was culture-positive, and patients with smear-negative culture-negative tuberculosis in the other cases. When culture-negative, patients were considered TB3 if (1) compared with baseline, chest radiography showed improvement after 3 mo of antituberculosis therapy; or (2) the tuberculosis was histologically proven, with or without culture of M. tuberculosis from the pulmonary biopsy specimen.

Table 1 : ATS CLASSIFICATION OF TUBERCULOSIS INFECTION

TB = tuberculosis; PPD = purified protein derivative

For each patient classified TB3, the following data were collected: number of clinical specimens obtained, interval between the collection of the first clinical specimen and the start of antituberculosis therapy, performance of invasive diagnostic procedures, presence of extrapulmonary tuberculosis, HIV-serology.

Control patients. Patients considered for active tuberculosis who did not meet the above mentioned radiologic criteria were considered at low risk for active tuberculosis and included as control patients. In addition, patients hospitalized at the same time for an unrelated disease with no risk factor for active tuberculosis were included to serve as control patients for the study. The whole specimens obtained were submitted to the same procedures as described above. For each of the control patients, a review of the clinical records was conducted as described above and patients were classified according to the ATS classification of tuberculosis.

Laboratory Processing

Respiratory specimens were digested and decontaminated by the NALC-NaOH procedure (9) under biosafety conditions. Briefly, 3 ml of 0.5% N-acetyl-L-cysteine-2% sodium hydroxide was mixed with an equal volume of specimen, incubated at room temperature with gentle shaking, and centrifuged. After discarding the supernatant, 2 ml of bovine serum albumin (0.2%) were added to the pellet and the total volume of the sediment was divided into four samples (at least 500 µl per sample). The first sample was used for auramine-fluorochrome staining and culture on Loewenstein-Jensen (LJ) medium with and without 0.4% sodium pyruvate. The following two samples were used for the LCx-MTB tests. The fourth sample was kept at - 0° C for retesting in case of discrepant results between LCx, culture, and clinical evaluation. Mycobacteria grown on LJ were identified by Accuprobe (GenProbe, San Diego, CA) for M. tuberculosis complex, M. avium complex, and M. gordonae, and by conventional methods for other mycobacterial species (10).

Direct Amplification Tests

LCx-MTB test. This test is based on the amplification by ligase chain reaction of a segment of the chromosomal gene of M. tuberculosis encoding for the protein antigen b (11). This gene sequence appears to be specific of the M. tuberculosis complex and has been detected in all M. tuberculosis complex strains examined to date (12).

The LCx-MTB test was performed according to the manufacturer's recommendations. All LCx tests were performed in duplicate (LCx1 and LCx2) on distinct samples of the same specimen. Sample preparation to LCx-MTB test was done in the mycobacteriology laboratory (Area 1). Preparation consisted in two washes and centrifugations to remove inhibitors, heating at 90° C for 10 min to kill mycobacteria, and sonication to lyse the cells. Fifty microliters of the lysed samples were added to the amplification vials, which contained all necessary reagents for amplification (polymerase, ligase, buffer, dNTPs). Then, vials were transferred to another room (Area 2) where they were incubated in a thermal cycler for 37 cycles, each cycle being 1 s at 94° C following by 1 s at 64° C and then 40 s at 69° C. After amplification, vials were placed in the LCx analyzer for automatic target detection by Micro-particle Enzyme Immunoassay (MEIA). Results were expressed as ratios based on a cutoff value determined as 0.3 times the mean MEIA value of the positive controls. The LCx-MTB test was positive if the ratio was >= 1,and negative if the ratio was < 1.

Two negative (salmon sperm DNA) and two positive (M. tuberculosis H37Rv DNA) controls for amplification and detection were included in each run. One positive (5 × 10  3 colonies/ml M. tuberculosis) processing control (extraction, amplification and detection) was prepared in the laboratory as recommended by the manufacturer and was tested every two runs.

GenProbe amplified Mycobacterium tuberculosis direct (MTD) Test. The MTD test is a commercial DAT (GenProbe, San Diego, CA) based on transcription-mediated amplification that detects ribosomal RNA. The MTD test was performed according to the recommendations of the manufacturer and as described previously (13).

Discrepancy Analysis

In case of discrepant results between LCx1 and LCx2 tests, the results were compared with the tuberculosis classification of the patient. The sample for which the LCx result was discrepant with the tuberculosis classification was tested again on the lysate kept at -20° C in accordance with the manufacturer. If discrepancy remained, the LCx test was considered to have given a false result.

If both LCx tests were positive in duplicate and the patient was not classified TB3, the specimen was retested on a fourth sample of the same specimen kept at -20° C with a different amplification method (MTD-GenProbe) using an alternate species-specific primer pair in order to confirm the positive result.

If both LCx tests were negative and the culture was positive, a spike-back procedure was performed in which mycobacterial DNA was added to the processed sample to determine if inhibitors of amplification were present. In parallel, the fourth sample was submitted to LCx-MTB and to MTD-GenProbe.

In patients classified TB3 whose specimens were negative for both LCx and culture, a fourth sample of the same specimens was tested by MTD-GenProbe.

Data Analysis

Results from the classification scheme described above was used as reference-standard for demonstrating the usefulness of the LCx-MTB test in identifying patients with active pulmonary tuberculosis. The LCx results after resolution of discrepant duplicate results, but prior to any other discrepant analysis, were used in our analysis. For each sample, agreement between LCx-MTB, culture, and clinical classification was determined. For each TB3 patient, it was determined if knowledge of LCx results could have avoided invasive diagnostic procedures, reduced hospital stay, or hastened the start of an appropriate antituberculosis therapy.

Case Patients

Thirty-two patients considered at high risk for active pulmonary tuberculosis were included as case patients. In all of them antituberculosis therapy was started before results of culture and all were ultimately classified TB3. The results for all laboratory processings (conventional methods and LCx-MTB tests) according to clinical classification are given in Table 2.

Table 2 : Results of Labotory Tests in case and control Patients

* positive for M.tuberculosis complex

Of the 32 TB3 patients, 18 (56%) were smear-positive culture-positive, eight (25%) were smear-negative culture-positive, and six (19%) were smear-negative culture-negative. In the latter patients, the diagnosis of tuberculosis was based on isolation of M. tuberculosis in a pulmonary biopsy from two patients, based on radiologic response associated with documented extrapulmonary tuberculosis in three, and based only on radiological response to antituberculosis therapy in one. Five (16%) TB3 patients were HIV coinfected, two in the smear-positive culture-positive group, one in the smear-negative culture-positive group, and two in the smear-negative culture-negative group.

All 18 smear-positive TB3 patients had at least one specimen that was positive when tested by LCx-MTB (mean number of samples per patient, 2.8; range, 1 to 4). Seven out of the eight smear-negative and culture-positive TB3 patients had at least one specimen that was positive when tested by LCx-MTB (mean number of samples per patient, 3.9; range, 1 to 9). None of the specimens obtained from the six smear-negative culture-negative TB3 patients was LCx-positive (mean number of samples per patient, 2; range, 1 to 5).

Ten invasive diagnostic procedures have been performed in eight smear-negative culture-positive TB3 patients before the results of culture: six fiberoptic bronchoscopies, one CT-scan guided transthoracic needle aspiration, one pleural biopsy, one hepatic biopsy, and one open lung biopsy. Mean time from the collection of the first specimen to the start of antituberculosis therapy for these patients was 6.9 d (range, 1 to 17). Knowledge of the results of LCx-MTB could have made these procedures unnecessary for four patients, avoiding four bronchoscopies, the hepatic biopsy and the open lung biopsy. Early knowledge of the LCx results could have hastened the start of adequate antituberculosis therapy by 2 and 15 d in two inpatients and by 15 d in one outpatient. None of these three patients was HIV positive or had an acute form of tuberculosis.

Ten invasive diagnostic procedures have been performed in the culture-negative TB3 patients, including four fiberoptic bronchoscopies, three open lung biopsies, two hepatic needle biopsies, and one sternal puncture. None of these procedures could have been avoided by knowing the LCx results. Mean time before the start of antituberculosis therapy was 12.7 d (range, 2 to 25).

Control Patients

Two hundred four patients providing 412 specimens were included in the control group. Six patients were lost to follow-up, and consequently these patients and the seven specimens provided by them were excluded from the final analysis, leaving 198 control patients and 405 specimens. None of the control patients was ultimately classified TB3 and no patient was treated empirically for tuberculosis until being classified. Forty-five patients (88 specimens) were classified TB4. The reason for which none of them was considered at high risk for current tuberculosis at inclusion were as follows: in 19 patients the radiographic findings were documented to be stable for at least 6 mo; in nine patients no characteristic pattern of active disease could be evidenced on chest CT scan; the remainng 17 patients with a past history of tuberculosis were hospitalized for a known disease unrelated to tuberculosis. One patient was classified TB2 (one specimen) and 152 patients (316 specimens) were classified TB0-1. One TB0-1 patient gave two smear-positive specimens but was not included among the case patients because he was already known to be infected with M. abscessus. The clinical outcome of the control patients is given in Table 3.

Seven specimens provided by seven patients (two classified TB4 and five classified TB0-1) were positive in duplicate when tested with LCx-MTB. Two additional specimens provided by two patients classified TB0-1 yielded discrepant duplicate LCx results, which remained after the lysate of the sample yielding the positive LCx result was tested again, leaving nine patients with at least one specimen that was falsely positive when tested with LCx-MTB.

The results for all laboratory processings (conventional methods and LCx-MTB tests) of the case and the control patients are given in Table 2.

Sensitivity and Specificity of LCx-MTB Compared with Culture in Detecting M. tuberculosis in Clinical Samples

The results of the LCx-MTB according to smear and culture results are given in Table 4. Types of specimen were as follows: 246 (49%) sputum samples, 100 (20%) gastric aspirates, 93 (19%) bronchoscopic aspirates, and 60 (12%) bronchoalveolar lavage fluids. Culture was positive for M. tuberculosis in 65 (13%) specimens of which 37 were smear-positive and 28 were smear-negative. Culture was positive for NTM in 27 (5.4%), and was negative for mycobateria in 407 (81.6%). Contamination rate on LJ during the period of the study was 4.2%.

Table 4 : RESULTS OF CONVENTIONAL LABORATORY METHODS AND LCX MTB TEST AFTER RESOLUTION OF DISCREPANCIES

Definition of abbreviations: NTM, non tuberculous mycobacteria.

Among the 499 specimens (94 specimens from the case patients and 405 specimens from the control patients) analyzed by LCx-MTB, there were 13 (2.6%) discrepant results between the duplicate samples. Two specimens were provided by smear-negative culture-positive TB3 patients and 11 by the control patients. Nine were resolved by performing a third LCx-MTB test on the lysate of the sample that gave discrepant results with the culture and the clinical classification. In the remaining four, the discrepant result was taken into account, leaving two false-positive and two false-negative LCx results.

Thirty-nine (7%) of the 499 specimens were smear-positive. LCx-MTB was positive in all but one of the 37 specimens culture-positive for M. tuberculosis and was negative in two specimens yielding M. abscessus. LCx-MTB was negative in duplicate in one smear-positive specimen, which cultured for a single colony of M. tuberculosis only. Thus, the sensitivity and the specificity of LCx-MTB on smear-positive specimens was 97 and 100%, respectively.

Twenty-eight (5.6%) specimens were smear-negative but culture-positive for M. tuberculosis. LCx-MTB was positive in 16 (57%) of them. LCx-MTB was negative in duplicate in 10 of these specimens and after discrepancy analysis in two more specimens. For each of these 12 specimens, spike-back analysis did not reveal any inhibitor. Furthermore, when a fourth sample of the same specimen was tested, LCx-MTB was positive in five specimens and MTD-GenProbe test in eight. For the 12 smear-negative and the one smear-positive specimens falsely negative when tested with LCx-MTB, culture yielded less than 20 colonies of M. tuberculosis (Figure 1).

Figure 1. Sensitivity of LCx-MTB in accordance with quantitative culture results.

Four hundred thirty-two specimens were smear-negative and culture-negative for M. tuberculosis. However, LCx-MTB was positive in duplicate in eight of them and after discrepancy analysis in two more specimens. These positive results were not confirmed when a fourth sample of the same specimen was tested by MTD-GenProbe. Consequently, there were 10 false-positive LCx tests when compared with culture. However, one of these culture-negative specimens was derived from a culture proven TB3 patient, leaving only nine false-positive results when compared with culture and clinical classification.

Thus, the sensitivity and specificity of LCx-MTB in detecting M. tuberculosis in smear-negative specimens was 57 and 98%, respectively.

We performed an evaluation of the clinical utility of a commercially available LCx-based test in detecting M. tuberculosis complex in respiratory tract specimens from patients at high risk for active tuberculosis. The distribution in smear-positive culture-positive, smear-negative culture-positive, and smear-negative culture-negative TB3 patients in our study were like those reported in similar settings (14). The incidence of 16% HIV coinfected patients among our TB3 patients is in accordance with the incidence of HIV infection among TB3 patients in France (15). Unexpectedly, all patients included as case patients turned out to have active pulmonary tuberculosis. This may be explained by several factors: (1) only patients with a high suspicion of active tuberculosis were included in this group; (2) chest CT scan was widely used to help distinguishing between active and inactive pulmonary tuberculosis; (3) all case patients were included by the same pulmonologist investigator; (4) the sample size of smear-negative TB3 patients was small. Adversely, presumably because of this latter reason, we did not encounter TB3 patients with atypical radiologic presentation who would not have been included as a case patient according to our selection criteria.

The sensitivity and specificity of LCx-MTB when compared with culture was 97 and 100%, respectively, in smear-positive specimens and 57 and 98% in smear-negative specimens. These results are similar to those of other studies evaluating the sensitivity and specificity of LCx-MTB (5, 6) and other commercially available DATs for detecting M. tuberculosis complex in respiratory tract specimens (1).

Each of the smear-positive case patients was adequately identified as being infected with M. tuberculosis by LCx-MTB. It indicates that DATs can reliably be used for the rapid confirmation of tuberculosis infection as recommended by the ATS (2).

Only 50% of the smear-negative TB3 patients had at least one specimen that was positive when tested with LCx-MTB. Twelve smear-negative culture-positive specimens gave false-negative results when tested with LCx-MTB. No inhibitor of the amplification reaction could be demonstrated to explain these negative results. The fact that half of these specimens were positive when another sample of the same specimen was tested, and that the culture of each of these specimens yielded less than 20 colonies, suggests that the nonuniform distribution of the mycobacteria in the test sample or the low DNA copy number were the explanations for the initial LCx-negative results. A comparable detection threshold has been observed with other commercially available DATs that have been reported not to detect reproducibly less than 40 colonies of M. tuberculosis, although in some samples even lower numbers could be detected (2).

LCx-MTB was constantly negative when tested in the specimens derived from the culture-negative TB3 patients, again probably because of the absence or the low number of DNA copy. The usefulness of DAT in culture-negative TB3 patients has not been well established yet. Most of the reported studies used culture as a major end point. The trials using clinical end points did not give standardized definitions of culture-negative TB3 patients and did not separate them from culture-positive TB3 patients already receiving antituberculosis therapy, limiting the interpretation of negative culture results (16). The diagnosis of active pulmonary tuberculosis was undoubted in each of our culture-negative TB3 patients and none of them received antituberculosis medication until their specimens were collected. As none of these specimens was positive when tested by LCx-MTB our results suggest, despite the small number of patients, that DAT may be of limited interest in culture-negative TB3 patients.

Knowledge of the LCx results at the time of specimen collection could at least have hastened the start of antituberculosis therapy by 2 to 15 days in three (21%) smear-negative TB3 patients, a cut in the delays that can be of particular interest in acute forms of tuberculosis and in HIV-positive patients. Moreover, knowledge of the LCx results could have made invasive diagnostic procedures unnecessary in four (29%) patients, avoiding four (44%) of the fiberoptic bronchoscopies, the open-lung biopsy, and one hepatic biopsy. Because the only LCx-positive but culture-negative specimen provided by the case patients was obtained from a culture-proven TB3 patient, a positive LCx-MTB result in our study was 92% predictive of a positive culture result of the tested specimen and 100% predictive that at least one of the patients specimens would yield M. tuberculosis, precluding the need to collect further specimens. These results suggest that DAT, when positive, may be a useful step before performing further invasive diagnostic procedures.

The specificity of LCx-MTB in our study was among the highest reported with commercially available amplification tests (1) probably because the following safety measures were taken to limit the risks of contamination (3): the LCx-MTB assay was semiautomated, reducing the handling of the specimens; the procedure included an automatic destruction of the amplification products after detection; internal controls were used during the entire procedure, and a physical separation of the different steps of the procedure was observed in our laboratory. Despite these measures nine specimens from control patients were falsely positive when compared with culture and clinical classification. False-positive PCR results have been reported after use of contaminated bronchoscopes (22), and indeed three of these specimens were obtained by fiberoptic bronchoscopy. However, none of the three bronchoscopes had been used in the previous week for a culture-positive TB3 patient. Moreover, specimens collected with each of the three bronchoscopes before collecting the false-positive samples, were negative both for LCx-MTB and culture, making improbable a possible contamination of the specimens by the bronchoscope. Some studies have found positive PCR-results in specimens obtained from TB4 patients (23). In our study, only two (4%) of the 45 TB4 patients had a specimen that gave a false-positive LCx result, a proportion not significantly different from that of the TB0-1 patients chi 2 square test; p > 0.5). Moreover, a second and different amplification test (MTD-GenProbe) was negative in all the false-positive specimens, suggesting that laboratory DNA contamination of the tested sample as the most likely explanation for these false-positive results.

We conclude that the sensitivity and specificity of LCx-MTB in identifying MTBC in respiratory specimens when compared with culture is similar to that of the other commercially available DATs. In smear-positive patients, LCx-MTB is a reliable test for confirmation of tuberculosis infection. In smear-negative patients with a high clinical and radiologic suspicion of active tuberculosis, LCx-MTB could be beneficial as a step before performing further invasive diagnostic procedures and in hastening the start of an adequate antituberculosis chemotherapy. We shall address this issue in a prospective clinical trial.