Eur J Clin Microbiol Infect Dis (2000) 19 : 859–867 Q Springer-Verlag 2000 Article Clinical Evaluation of an In-House IS6110 Polymerase Chain Reaction for Diagnosis of Tuberculosis J. Almeda, A. García, J. González, L. Quintó, P.J. Ventura, R. Vidal, G. Rufí, J.A. Martínez, M.T. Jiménez de Anta, A. Trilla, P.L. Alonso Abstract The aim of this study was to clinically validate a heminested polymerase chain reaction (PCR) method, based on the IS6110 insertion segment of Mycobacterium tuberculosis complex, for the diagnosis of tuberculosis. Samples of pulmonary, extrapulmonary and blood origin were collected prospectively from 331 patients. All samples were processed to detect acid-fast bacilli by direct stain, culture and PCR. The gold standard comparison was a clinically based final case definition of tuberculosis corresponding to group 3 of the American Thoracic Society’s classification system. The sensitivities of stain, culture and PCR were 41%, 65% and 59%, respectively. Overall specificity exceeded 97% for all techniques. The combination of PCR and direct stain achieved a sensitivity similar to that of culture alone. The PCR method detected 74 of 95 (78%) culture-positive results. In a hospital setting, PCR could be a useful, reliable tool for diagnosis of tuberculosis and may be introduced as a complementary routine diagnostic laboratory method. J. Almeda, L. Quintó, P.J. Ventura, P.L. Alonso Unitat d’Epidemiologia i Bioestadística, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Hospital Clínic, Barcelona, Spain A. García Departament de Microbiologia i Parasitologia, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Facultat de Medicina, Universitat de Barcelona J. González (Y) Servei de Microbiologia, Hospital Clínic, Villarroel 170, 08036 Barcelona, Spain e-mail: jgm6medicina.ub.es Tel.: c34-93-2275522 Fax: c34-93-2275400 J. González, M.T. Jiménez de Anta Departament de Microbiologia i Parasitologia, Institut Clínic d’Infeccions i Inmunologia, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Hospital Clínic, Barcelona, Spain R. Vidal Servei de Neumologia, Ciutat Sanitària Vall d’Hebron, Barcelona, Spain G. Rufí Servei de Malalties Infeccioses, Ciutat Sanitària Universitària de Bellvitge, L’Hospitalet del Llobregat, Barcelona, Spain J.A. Martínez, A. Trilla Departament de Malalties Infeccioses, Institut Clínic d’Infeccions i Inmunologia, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Hospital Clínic, Barcelona, Spain Introduction Current policy for control of tuberculosis (TB) consists of early identification and cure of infectious cases. Most infectious cases are detected by direct-smear microscopy examination of sputum, and treatment is usually started before obtaining laboratory confirmation. Microscopy for acid-fast bacilli (AFB) is a rapid, simple diagnostic method. Although its sensitivity ranges from 22% [1] to 78% [2], its specificity remains high (99% [1] to 100% [3]). The sensitivity of direct staining may vary as well, depending on the clinical radiological findings, ranging from 52% in patients with tuberculous caverns to 32% in patients with pulmonary infiltrates only [4]. Culture is more sensitive than direct staining for detection of AFB, but it takes longer. Culture requires an average of 21.7 days to obtain results; even the Bactec radiometric technique (Becton Dickinson, USA) requires 13.4 days to produce results [5]. The sensitivity of culture ranges from 50% [6] to 81% [7], estimates that can also vary depending on the clinical presentation. Thus, a more rapid, sensitive and specific diagnostic assay, such as PCR, could contribute to the control of Mycobacterium tuberculosis [8–11]. PCR may overcome some difficulties associated with conventional diagnostic techniques, such as the large number of bacilli required for positive stained smears and the slow growth of mycobacteria in selective media [12]. 860 Several studies validating PCR as a tool for diagnosis of TB have recently been published. Clarridge et al. [13] reported a very high sensitivity using an IS6110targeted PCR system. Overall, the sensitivity and specificity of PCR ranges from 50 to 100% [14–19]. It has been suggested that this wide range may be due to the variability of PCR results in different laboratories [20, 21]. According to this data, the patients were classified as belonging to group 3 of the ATS classification. Classification was determined by two independent TB expert reviewers, based on the initial clinical history recorded, without seeing the patient, and before confirmation by any microbiological laboratory result. Undefined cases were established by lack of agreement between the two reviewers. In this study we evaluated the clinical validity and applicability of a PCR method based on the IS6110 insertion segment for the detection of Mycobacterium tuberculosis in HIV-negative patients. The PCR was compared with direct staining and culture. The results of all diagnostic methods were compared with a clinically based case definition of TB, which is our gold standard, and sensitivity and specificity values were calculated. The American Thoracic Society [23] classifies people exposed to Mycobacterium tuberculosis into six groups. Group 0 comprises persons not exposed, not infected and tuberculin skin test (PPD)negative. Group 1 comprises those people who have been exposed to tuberculous patients but who are PPD-negative. Group 2 comprises infected (positive PPD test) persons who have no clinical, radiological or microbiological signs of TB. Group 3 comprises patients with active TB who have clinical or radiological evidence of TB. Ideally, the culture should be positive. If the culture is negative, the PPD test must be positive. Group 4 comprises persons with nonactive TB, defined as cases that are PPD-positive, without clinical signs of active TB but with positive microbiological data or radiological signs that suggest TB in the past. Group 5 comprises patients suspected to have TB but for whom confirmation is unavailable and diagnostic studies are incomplete. Materials and Methods Microbiological Processing of Clinical Samples. All clinical samples were processed at the microbiology laboratory of the Hospital Clinic. Laboratory technicians were blinded to the group of patients from which each clinical sample came. Study Population. The study was conducted at the Hospital Clinic of Barcelona from 1 March 1994 to 29 February 1996. In order to identify all TB patients promptly, a list of those patients admitted to the hospital who were started on isoniazid treatment was obtained from the Pharmacy Department every day. In addition, a list of the patients being followed-up in the TB outpatient clinic was obtained twice a week. On the basis of this information, the patients were classified into the following categories: new HIV-positive TB cases, new HIV-negative TB cases, TB cases with more than 1 week of treatment, patients undergoing prophylactic treatment, or non-TB cases. Following receipt of informed consent, all HIV-negative patients who started anti-TB treatment were included. TB treatment was a short-course, rifampicin-based 6-month chemotherapy regimen (rifampicin, isoniazid and pyrazinamide for the first 2 months, followed by 4 months of rifampicin and isoniazid). Pulmonary or extrapulmonary clinical samples for diagnostic purposes were collected, according to the clinical presentation in each case, at the time of diagnosis and before treatment was initiated. Blood samples were also obtained from all patients at the same time. Another two control groups of volunteers, who were supposed to be non-TB individuals, were also recruited. Group 1 comprised third-year students attending the School of Medicine, University of Barcelona, and group 2 comprised patients with chronic obstructive pulmonary disease (COPD) without clinical signs of TB who were admitted to the Pneumology Department. HIV antibodies were determined and tuberculin skin tests performed anonymously; signed, informed consent was obtained in all cases. Gold Standard and Case Definition. The following case definitions [22] were established before the study was started. The final definition (at the end of treatment) was determined by an investigators’ consensus, taking the following criteria into consideration: clinical picture, chest radiographic findings, tuberculin skin test, microbiology results, other laboratory findings, computed axial tomography scan (when available) and clinical response to antiTB drug therapy. According to this data, the cases were classified into group 3 of the American Thoracic Society (ATS) classification described below [23]. Doubtful cases were defined by a lack of consensus of two or more authors. Initial definition represents the situations that clinicians face the first day of treatment. Nonsterile samples were digested and decontaminated by the NaOH-NALC (N-acetyl-L-cysteine) method [24] before culture. Sterile samples were divided into two equal aliquots, one for culture and smear and the other for amplification procedures. The latter aliquot, was submitted to decontamination as a nonsterile specimen in order to obtain an appropriate sample for the amplification technique. Liquid samples such as cerebrospinal fluid, joint fluid, abdominal fluid, pleural fluid and urine were concentrated by centrifugation if more than a 1 ml volume was available [24]. The tissue biopsies were ground and processed as the remaining samples. Smears of all samples (except blood) were prepared and stained with auramine-rhodamine fluorochrome stain. Determination of semiquantitative grade was made in a 1 to 4 scale [24]. For all samples, aliquots of 200 ml were inoculated onto Löwenstein-Jensen slants and incubated at 37 7C for up to 8 weeks. In addition, 500 ml of the sediment was inoculated into radiometric Bactec 12B medium and incubated at 37 7C for 6 weeks. Blood samples were cultured directly by adding 5 ml to a Bactec 13A bottle and incubated following the same scheme. Solid medium slants were read weekly and Bactec cultures were read eight times over 6 weeks. A growth index of 200 on Bactec medium was considered positive and was further confirmed by Ziehl-Neelsen stain. Solid medium slants were considered positive when colonies grew; growth on positive slants was further confirmed as mycobacteria by Ziehl-Neelsen stain. Strains were identified using the morphological aspect on Bactec (Becton Dickinson, USA) [25] smears and the AccuProbe (Gen Probe, USA) method. When identification was inconclusive, gas chromatography [26] and routine biochemical methods were used [24]. A 0.5 ml aliquot of the pellet remaining after decontamination was obtained for each sample except blood. Blood samples were processed as follows: leukocytes were extracted from 3 ml of total blood by Ficoll-Hypaque (1119) gradient, washed and resuspended in Tris-EDTA buffer (100 mM Tris-HCl, 10 mM EDTA, pH 8.0) [27]. Two 0.5 ml aliquots were obtained. Aliquots were stored at –30 7C until amplification procedures were performed. 861 A single-tube heminested PCR with unlabeled primers was used for amplification [28, 29]. This method is a modification of the tube-nested technique described by Wilson et al. [30]. Aliquots were heated at 95 7C for 30 min and concentrated by centrifugation at 12,500!g for 15 min. The pellet was resuspended in tenfold TE buffer with 50 mg/ml of lysozyme. This suspension was incubated for 90 min at 37 7C, after which it was centrifuged for 10 min. Proteinase K and sodium dodecyl sulfate (SDS) were added to the final concentrations of 250 mg/ml and 1% (w/v) and incubated for 30 min (or overnight) at 43 7C. The suspension was extracted with phenol chloroform (25 : 24 : 1) twice and chloroform-isoamyl alcohol (24 : 1) twice and was precipitated with two volumes of ethanol 100% and 0.2 M NaCl and stored at –20 7C overnight. It was then washed with 70% ethanol, dried and resuspended in 100 ml of distilled water. Ten microliters was used for PCR amplification. The DNA target used was the repetitive sequence IS6110, which is found in the members of Mycobacterium tuberculosis complex at varying copy numbers [31, 32]. The outer primers used were Tb294 (5b-GGACAACGCCGAATTGCGAAGGGC-3b) and Tb850 (5b-TAGGCGTCGGTGACAAAGGCCACG-3b), which generate a product of 580 base pairs (bp), and the inner primer used was Tb505 (5b-ACGACCACATCAACC-3b) [30], which generates a final fragment of 369 bp. The outer primers Tb294 and Tb850 were both used at a concentration of 10 nM, and the inner primer Tb505 was used at a concentration of 500 nM. The amplification was performed in 0.5 ml PCR tubes with a final volume of 50 ml using a 480 thermal cycler (Perkin-Elmer, USA). Each reaction tube contained 2.5 U of Taq DNA polymerase (Gibco BRL, USA), 2.5 U of uracil-Nglycosylase (UNG) (Boehringer Mannheim, Germany) in order to avoid contamination with the products of previous reactions [33], 200 mM each of dATP, dCTP, dGTP and 600 mM dUTP (Boehringer Mannheim), final buffer 1!, 2 mM of MgCl2 and 20 ml of sample. After incubating for 15 min at 25 7C to allow UNG to act, the temperature was raised to 94 7C for 5 min to inactivate the enzyme. The thermal cycler is programmed to perform the PCR in two stages that are distinguished by different annealing temperatures. The first stage of amplification involves 30 cycles of denaturation at 94 7C for 45 s, with primer annealing and extension carried out in one step at 72 7C for 1 min 30 s. The second stage included 30 cycles of denaturation at 94 7C for 45 s, primer annealing at 55 7C for 1 min and extension at 72 7C for 30 s, after which the reaction mixture was held at 72 7C. To detect sample inhibitors, a duplicate tube of 50 ml of PCR mix for each sample was spiked with 5 ml of an aqueous solution containing 43 pg of purified DNA target. In order to maintain reagent proportions, 5 ml of the original sample was added to these duplicate tubes instead of 10 ml. To avoid cross-contamination, DNA purification, amplification and detection was carried out in a separate room from that in which the buffers were prepared. Samples were processed in lots of 20, including one extraction-negative control and one amplification-negative control. A positive control was also included in each lot of samples. Twenty microliters of amplified product was analyzed by ethidium bromide (0.5 mg/ml; Sigma) electrophoresis in 2% (w/v) agarose gel. The gels were visualized and photographed on an ultraviolet transilluminator. The presence of a band in the same position as that shown by a positive control and spiked duplicates, namely 369 bp, was considered positive. Specificity Testing of the Amplification Method Using Bacteria and Mycobacterial Strains. The following 13 strains of six mycobacterial species and 18 strains of eight bacterial species were assayed: Mycobacterium tuberculosis (np5), Mycobacterium xenopi (np3), Mycobacterium kansasii (np2), Mycobacterium avium (np1), Mycobacterium scrofulaceum (np1), Mycobacterium gordonae (np1), viridans group streptococci (np2), coagulase-negative staphylococci (np3), Proteus mirabilis (np1), Klebsiella pneumoniae (np2), Enterococcus spp. (np2), Staphylococcus aureus (np3), Escherichia coli (np2) and Pseudomonas aeruginosa (np3). All strains were isolated previously in clinical samples. For each strain a suspension of 5 ml of distilled water was prepared for 1 McFarland bacterial density. Dilutions were prepared, and 0.5 ml aliquots for several concentrations were obtained. For Mycobacterium tuberculosis the concentrations studied were 10 7 cfu/ml, 10 5 cfu/ml, 10 3 cfu/ml and 10 2 cfu/ml. For the remaining species the concentrations used were 10 7 cfu/ml, 10 5 cfu/ml and 10 3 cfu/ml. The 0.5 ml aliquots were processed for lysis, DNA extraction, amplifications and detection following the same protocol as that used for the clinical samples. Statistical Methods. Results were expressed as proportions, and 95% confidence intervals were calculated. Sensitivity and specificity were calculated using standard methods [34], comparing the results of each diagnostic method with the case definitions. Statistical differences between sensitivity and specificity of different diagnostic methods were tested by the McNemar exact binomial test. Degree of agreement between the case definitions was measured through Kappa statistic index. A Kappa value of 1 indicates perfect agreement and a value of zero means no agreement. Cases not well defined due to lack of case definition agreement were not considered in the analysis. A significance level of 5% was chosen for all statistical tests, using Bonferroni’s correction in multiple tests. Results Following the recruitment system described, a total of 553 patients started receiving isoniazid treatment during the study period or were attended in the TB outpatient clinic. One hundred ninety-eight (36%) were found to be HIV-positive, 39 had previously started anti-TB treatment, 53 were receiving TB prophylaxis and 25 were lost to follow-up. Therefore, a total of 238 HIV-negative TB patients were included in the study group. The mean age of patients in the study was 48 years (95% CI, 46–51), with a range of 17 to 91 years. One hundred sixty-six (69%) were male. The other TB-free cohorts comprised 46 students and 47 COPD patients. The students’ mean age was 20 years, 67% being female. The mean age of the COPD patients was 68 years, ranging from 42 to 82 years, with only 13% being female. Clinical samples were collected from 331 study and control individuals (Table 1). In total, 316 samples were available for stain, 319 for culture and 314 for PCR. Additionally, blood samples were collected from 261 individuals. Clinical samples were of pulmonary origin (sputum, bronchial aspirate, bronchoalveolar lavage and gastric fluid) in 270 individuals, including all controls, and of extrapulmonary origin (lymph nodes, feces, joint fluid, abdominal fluid, cerebrospinal fluid, pleural effusion, bone marrow aspiration, urine, vertebral bone biopsy and pus) in 50 individuals; 11 samples were insufficient 862 Table 1 Results of stain, culture and PCR, according to sample origin Type of sample PCR result Negative culture Negative stain Respiratory Gastric fluid Lymph node CSF Pleural effusion Urine Other a Insufficient Total negative positive NA negative positive NA negative positive NA negative positive NA negative positive NA negative positive NA negative positive NA negative positive NA 173 2 3 4 Positive stain Positive culture NA 3 Negative stain Positive stain 10 19 1 1 9 40 NA No culture, no stain available 3 192 65 5 5 3 2 3 3 1 1 3 1 4 1 5 1 1 17 8 1 1 16 3 2 1 5 1 3 7 217 4 Total 2 1 1 40 55 7 2 1 11 12 11 331 a Includes 2 vertebral bone samples, 1 bone marrow sample, 1 fecal sample, 1 peritoneal fluid sample, 1 joint fluid sample and 3 samples of nonspecific origin NA, not available; CSF, cerebrospinal fluid to perform any test. Laboratory results of stain, culture and PCR were negative for all the students and COPD cohort members, although one COPD patient had an isolated positive blood sample result. One case of cross-contamination during extraction was detected. New aliquots of the samples affected in the same lot were prepared and tested again, and previous aliquots were rejected. Thirty-four of 314 (10.8%) clinical samples (excluding blood samples) from patients and control individuals showed evidence of inhibition; among blood samples, this percentage was 32.1%. The presence of inhibitors was demonstrated in five (22.7%) microbiologically confirmed cases for which PCR results were negative. Likewise, 29 (13.8%) of those with negative results by PCR and no microbiological confirmation showed the presence of inhibitors. The specificity testing of the amplification method used with bacteria other than mycobacteria gave the following results: none of the 18 bacterial strains belonging to eight different species assayed at three different concentrations showed amplification product in the gel electrophoresis analysis. Moreover, no amplification was obtained from the eight strains of five mycobacterial species included in the test. All five Mycobacterium tuberculosis isolates amplified the specific band in gel electrophoresis when tested at the 1!10 7 cfu/ml, 1!10 5 cfu/ml and 1!10 3 cfu/ml concentrations. However, only one of the five strains amplified the 1!10 2 cfu/ml concentration. According to these results, the sensitivity limit of this PCR method is around 1,500 to 3,000 cfu/ml. On the basis of the final case definition, there were 141 TB cases, 85 non-TB cases and 12 poorly defined cases (Table 2). Overall, PCR was more sensitive than stain, reaching statistically significant differences for extrapulmonary (Pp0.035) and total samples (P~0.0001) but not for pulmonary samples. Compared with culture, PCR had a lower non-statistically significant sensitivity. All methods showed a high and similar specificity. Culture appeared to be more sensitive than stain for both pulmonary and total samples (P~0.0001), but the difference was not significant for extrapulmonary samples. Table 3 shows the results of PCR and the PCR-stain combination compared with the results of culture. A total of 96 of 331 culture results were positive for AFB. Overall sensitivity values were below 87%. PCR was more sensitive than stain, reaching statistically significant differences for extrapulmonary (Pp0.024) and Table 2 Sensitivity and specificity of each test according to sample origin, compared with the final case definition of tuberculosis Test Total samples Sensitivity (%) Stain 57/138 (41.3) Culture 92/140 (65.7) PCR (main samples) c 82/139 (59.0) Stain & PCR combined 89/137 (65.0) PCR (blood samples) 13/104 (12.5) Pulmonary samples Extrapulmonary samples 95% CI Specificity (%) 95% CI Sensitivity (%) 95% CI Specificity (%) 95% CI Sensitivity (%) 33.0–50.0 75/77 (97.4) 74/78 a (94.9) 76/76 (100) 74/76 (97.4) 52/52 (100) 90.9–99.7 53/109 (48.6) 80/110 (72.7) 68/109 38.9–58.4 57/59 (96.6) 55/59 a (93.2) 58/58 88.3–99.6 4/29 (13.8) 12/30 (40) 14/30 57.2–73.5 50.3–67.3 56.4–72.9 6.8–20.4 87.4–98.6 95.3–100 b (62.4) 90.8–99.7 93.2–100 b 74/108 (68.5) 63.4–80.8 52.6–71.5 (100) 58.9–77.1 56/58 (96.6) 83.5–98.1 93.8–100 b (46.7) 88.1–99.6 32.5–70.6 (51.7) 95% CI 3.9–31.7 22.7–59.4 28.3–65.7 (100) 18/18 Specificity (%) 95% CI 18/18 100 19/19 (100) 18/18 81.5–100 a 82.4–100 b 81.5–100 b 81.5–100 b (100) a Two cases of Mycobacterium xenopi and 2 cases of Mycobacterium kansasii infection One side 97.5% confidence interval c Samples other than blood are included b Table 3 Sensitivity and specificity of each test by sample origin, compared with the results of culture Test Total samples Sensitivity (%) Stain 55/95 (57.9) PCR 74/95 (77.9) Stain & PCR combined 82/94 (87.2) Pulmonary samples Extrapulmonary samples 95% CI Specificity (%) 95% CI Sensitivity (%) 95% CI Specificity (%) 95% CI Sensitivity (%) 47.3–68 126/130 (96.9) 121/129 (93.8) 119/128 (93) 92.3–99.2 52/83 (62.7) 63/83 (75.9) 71/82 (86.6) 51.3–73 91/94 (96.8) 87/92 (94.6) 87/92 (94.6) 91–99.3 3/12 (25) 11/12 (91.7) 11/12 (91.7) 68.2–85.8 78.8–93.2 88.1–97.3 87.1–96.7 65.3–84.6 77.3–93.1 87.8–98.2 87.2–98.2 95% CI 5.5–57.2 61.5–99.8 61.5–99.8 Specificity (%) 95% CI 35/36 (97.2) 34/37 (91.9) 32/36 (88.9) 85.5–99.9 78.1–98.3 73.9–96.6 863 864 total samples (P~0.047) but not for pulmonary samples. Both methods showed a high and similar specificity. Based on the initial case definition (data not shown), 120 cases were classified as TB cases, 108 as non-TB cases and 10 as undefined cases. Sensitivity, for total samples, was as follows: stain, 35% (95% CI, 26.5–44.4); culture 59% (95% CI, 49.4–67.8), PCR, 53% (95% CI 44.0–62.6); and combined stain-PCR, 58% (95% CI 49.1–67.7). The difference between PCR and stain reached statistical significance for total samples (P~0.0001) but not for extrapulmonary and pulmonary samples. PCR and culture had similar sensitivity. No differences in specificity were shown for any method. In the samples submitted to the three assayed tests, the agreement between culture and both (initial and final) case definitions was as follows: (i) when comparing culture and the final case definition, there was 83.2% agreement and a Kappa index of 0.65 (P~0.001); (ii) when comparing culture and the initial case definition, there was 76.7% agreement and a Kappa index of 0.49 (P~0.001); (iii) when comparing the initial and final case definitions, there was 87.4% agreement and a Kappa index of 0.74 (P~0.001). Discussion This study shows that PCR is not more sensitive than culture, but that it is more sensitive than stain for extrapulmonary but not for pulmonary samples. Overall, culture is the most sensitive technique, However, the combination of PCR and stain, both of which are faster than culture, has a sensitivity similar to that of culture. When the non-TB groups were considered, the same pattern was seen. By adding “healthy” individuals, the estimates for specificity were more precise and slightly higher, but the sensitivity parameters remained the same as those presented in Tables 2 and 3. These results are comparable to those published recently and based on in-house methods or commercial kits; sensitivity rates reported ranged from 58 to 64% [14, 35–43]. Reports comparing commercial and in-house methods [44, 45] showed in-house techniques to be more sensitive. The lack of standardization and a poor reproducibility have been argued as limitations for the routine use of in-house techniques [14], but the same phenomenon has been demonstrated with commercial kits [46]. We also studied the value of heminested PCR in blood samples for diagnosis of TB in nonimmunosuppressed individuals (as far as testing for HIV is concerned). Thirteen of 104 blood samples collected from patients with confirmed TB (final definition) were positive by PCR, whereas only one of 23 classified control samples was positive. The presence of inhibiting substances found in this study is similar to that reported by others, ranging from 5 to 16% [47, 48]. Blood samples showed a higher percentage of inhibition (23%). Inhibition caused a problem of misclassification, which was reflected in the calculation of the sensitivity rate for PCR; the sensitivity of PCR might be higher if inhibiting substances are avoided. Blood is considered to contain a high level of inhibiting substances such as hemoglobin, although recently it has been demonstrated that pretreatment with SDS may remove the hemoglobin, at least partially [49]. We conducted this prospective study within a hospital setting to test the possibility of introducing PCR as part of the routine diagnosis. It is impossible to generalize the results for other settings. Nevertheless, these results are similar to those found elsewhere and also correspond with those found in an epidemiological study of TB in Barcelona [50]. In the design stage, we intended to include some negative controls to validate the PCR test. However, we excluded them from the primary analysis on the basis of the likelihood of introducing bias. Indeed, the total number of controls was arbitrary and we found that the bigger the number, the higher the specificity (data not shown). Nevertheless, given the fact that specificity is, in any case, high, this approach seems to be valid for evaluating sensitivity, the estimates of which did not change at all. The gold standard remains the critical issue. Our data shows that culture and the final definition seem to have similar value as a gold standard for validating a new diagnostic technique. Although this study had a rather long follow-up, its prospective design allowed us to compile all the essential data required to determine the case definition, on which the analysis was based. We used a diagnostic method developed by our group [28, 29], a one-tube heminested PCR based on a modification of a nested PCR described by Wilson et al. [21]. This technique targeted the most commonly used fragment, the insertion sequence IS6110 [16–19, 31]. In order to test the specificity of the amplification technique described, several bacterial and mycobacterial strains were tested at different concentrations. None of these strains amplified the 369 bp band. This was confirmed when samples from non-TB patients were tested, showing the high specificity of the amplification technique used. In addition, five Mycobacterium tuberculosis strains at four different concentrations of mycobacteria were tested in order to approximate the threshold of this technique. The results of this test showed a sensitivity of around 1,000 to 3,000 cfu/ml, which may be lower than that reported by several authors in the initial studies applying this methodology to the diagnosis of TB [16–18]. However, the clinical evaluation of the method described in the present study showed a sensitivity comparable to that of other methods that were evaluated using clinical cases, 865 including a representative proportion of smear-positive and smear-negative cases, despite the different protocols applied in extraction, purification and amplification [14, 35–45]. This could mean that, with the current methodology, the sensitivity of PCR techniques is similar to that described here. This conclusion corresponds with the results obtained in a multicentric study using artificial samples with a known number of mycobacteria [21]. On the other hand, it could be argued that our PCR sensitivity estimates could be lower because Mycobacterium tuberculosis strains lack the insertion segment IS6110, as has recently been found in Asian patients. Nevertheless, the spread of this strain appears to be limited [51–53]: most Mycobacterium tuberculosis strains have 2–20 copies, and only 1 to 5 copies are present in Mycobacterium bovis strains [16, 31]. In order to overcome the well-known PCR-related sensitivity and specificity problems, the heminested technique includes a second step that enhances the sensitivity by amplifying a second set of primers [54] and that enhances specificity by reducing the need for manipulation and, therefore, decreasing the risk of contamination. This technique has the advantage of a much lower cost when compared with a conventional two-stage nested PCR, which requires more reagents. In the present study, the sensitivity and specificity of the amplification technique used were the main focus, and therefore, the extraction technique used was considered a reference, despite being laborious. However, from a theoretical point of view, other, faster extraction methods could be applied to obtain results the same day. In conclusion, we consider that the PCR cannot, at present, replace culture for the final confirmation of clinical cases because of its lower sensitivity and the additional need to have growth of Mycobacterium tuberculosis colonies for susceptibility testing and other special studies. However, the use of stain and PCR combined greatly increases the number of cases detected before culture results are available; therefore, it is recommended that PCR be introduced into the hospital laboratory as part of the standard procedures for the diagnosis of tuberculosis. The PCR technique is a fast, reliable method for the diagnosis of TB, but its sensitivity, total automation and cost should be improved in the future. In the future, PCR results will probably need to be considered for the classification and treatment of TB, just as the results of stains are now considered in developed countries. Acknowledgements We thank the Departments of Infectious Diseases, Microbiology, Internal Medicine, Pneumology, Pharmacology and Hematology of the Hospital Clinic (Barcelona) for their support and collaboration. Finally, we are very grateful to all of the patients and medical students (course 1992–1998 School of Medicine, Barcelona University) who participated in this project. Financial support was provided by the following institutions: (i) Fondo de Investigaciones Sanitarias de la Seguridad Social (FIS), grant 96/0028, (ii) Sociedad Española de Neumo- logía y Cirugía Torácica (SEPAR), grant 96/444, (iii) Departament de Microbiologia i Parasitologia, Divisió de Ciències de la Salut, Universitat de Barcelona, predoctoral grant (A.G.). References 1. Boyd JC, Marr JJ: Decreasing reliability of acid-fast smear techniques for detection of tuberculosis. Annals of Internal Medicine (1975) 82 : 489–492 2. Strumpf IJ, Tsang AY, Sayre JW: Reevaluation of sputum staining for the diagnosis of pulmonary tuberculosis. American Review of Respiratory Disease (1979)119 : 559–602 3. Burdash NM, Manos JP, Ross D, Banister ER: Evaluation of the acid-fast smear. Journal of Clinical Microbiology (1976) 4 : 190–191 4. Greenbaurn M, Beyt BE, Murray PR: The accuracy of diagnosis tuberculosis at a large teaching hospital. American Review of Respiratory Disease (1980) 121 : 477–481 5. Pierre-Audigier C: Techniques diagnostiques recentes pour le diagnostic des infections mycobacteries. Press Medicale (1994) 23 : 665–670 6. Krasnow I, Wayne LG: Comparison of methods for tuberculosis bacteriology. Applied Microbiology (1969) 18 : 915–917 7. Levy H, Feldman C, Sancho H, Van der Meulen H, Kallenbach J, Koornohof H: A reevaluation of sputum microscopy and culture in the diagnosis of pulmonary tuberculosis. Chest (1989) 95 : 1193–1197 8. Barnes PF, Bloch AB, Davidson PT, Snider DE Jr: Tuberculosis in patients with human immunodeficiency virus infection. New England Journal of Medicine (1991) 324 : 1644–1649 9. Bloom BR, Murray CLC: Tuberculosis commentary on a reemergent killer. Science (1992) 257 : 1055–1064 10. Bloom BR: Back to a frightening future. Nature (1992) 358 : 538–539 11. Andersson N, Gillespie SH, Wilson S: Culture free microbiology. Lancet (1992) 339 : 733 12. Jeager H Jr, Lacy J, Smith LR, LeMaistre CA: Quantitative studies of mycobacterial population in sputum and saliva. American Review of Respiratory Disease (1967) 95 : 998–1004 13. Clarridge JE, Shawar R, Shinnick TM, Plikaytis B: Largescale use of polymerase chain reaction for detection of Mycobacterium tuberculosis in a routine mycobacteriology laboratory. Journal of Clinical Microbiology (1993) 31 : 2049–2056 14. Eisenach KD, Sifford MK, Cave MD, Bates JH, Crawford JT: Detection of Mycobacterium tuberculosis in sputum samples using a polymerase chain reaction. American Review of Respiratory Disease (1991) 144 : 1160–1163 15. Brisson-Noel A, Aznar C, Chureau C, Nguyen S, Pierre C, Bartoli M, Bonete R, Pialoux G, Gicquel B, Barrigue G: Diagnosis of tuberculosis by DNA amplification in clinical practice evaluation. Lancet (1991) 338 : 364–366 16. Cousins DV, Wilton SD, Francis BR, Gow BL: Use of polymerase chain reaction for rapid diagnosis of tuberculosis. Journal of Clinical Microbiology (1992) 30 : 255–258 17. Pao CC, Yen TS, You JB, Maa JS, Fiss EH, Chang CH: Detection and identification of Mycobacterium tuberculosis by DNA amplification. Journal of Clinical Microbiology (1990) 28 : 1877–1880 18. Shawar RM, El-Zaatari FA, Nataraj A, Clarridge JE: Detection of Mycobacterium tuberculosis in clinical samples by twostep polymerase chain reaction and non-isotopic hybridization methods. Journal of Clinical Microbiology (1993) 31 : 61–65 19. Wilson SM, NcNermey R, Nye PM, Godfrey-Faussett PD, Stoker NG, Voller A: Progress toward a simplified polymerase chain reaction and its application to diagnosis of tuberculosis. Journal of Clinical Microbiology (1993) 31 : 776–782 866 20. Noordhoek GT, Kolk AHJ, Bjune G, Catty D, Dale JW, Fine PEM, Godfrey-Fausset P, Cho SN, Shinnick T, Svenson SB, Wilson S, Van Embden JDA: Sensitivity and specificity of PCR for detection of Mycobacterium tuberculosis: a blind comparison study among seven laboratories. Journal of Clinical Microbiology (1994) 32 : 277–284 21. Noordhoek GT, Van Embden JDA, Kolk AHJ: Questionable reliability of the polymerase chain reaction in the detection of Mycobacterium tuberculosis. New England Journal of Medicine (1993) 329 : 2036 22. Smith PG, Morrow RH: Outcome measures and case definition. In: Smith PG, Morrow RH (eds): Methods for field trials of interventions against tropical diseases. Oxford University Press, New York (1991) pp 150–151 23. American Thoracic Society: Diagnostic standards and classification of tuberculosis. American Review of Respiratory Disease (1990) 142 : 725–735 24. Roberts GD, Koneman EW, Kim VK: Mycobacterium. In: Balows A, Hausler W Jr, Herrmann KL, Isenberg HD, Shadomy HJ (eds): Manual of clinical microbiology. American Society for Microbiology, Washington DC (1991) pp 304–339 25. Gonzalez J, Tudó G, Gómez J, Garcia A, Navarro M, Jimenez de Anta MT: Use of microscopic morphology in smears prepared from radiometric cultures for identification of M. tuberculosis complex, M avium complex, M. kansasii and M. xenopi. European Journal of Clinical Microbiology & Infectious Diseases (1998) 17 : 493–500 26. Luquin M, Ausina V, López-Calahorra F, Belda F, GarciaBarceló M, Celma C, Prats G: Evaluation of practical chromatographic procedures for identification of clinical isolates of mycobacteria. Journal of Clinical Microbiology (1991) 29 : 120–130 27. Iralu JV, Sritharan VK, Pieciak WS, Wirth DF, Maguire JH, Barker RH Jr: Diagnosis of Mycobacterium avium bacteremia by polymerase chain reaction. Journal of Clinical Microbiology (1993) 31 : 1811–1814 28. González J, García A, Almeda J, Quintó Ll, Alonso P, Tudó G, Mayor A, Jiménez de Anta MT: Utility of a PCR method for diagnosis of immunocompetent tuberculosis patients. In: Programs and abstracts of the 7th meeting of the Grupo Español de Micobacteriologia. Zaragoza, Spain (1996) Abstract SCO-01.2 29. González J, García A, Almeda J, Quintó Ll, Navarro M, Alonso P, Tudó G, Jiménez de Anta MT: The PCR test in blood samples in tuberculosis diagnosis in non-HIV patients. In: Program and abstracts of the 18th Annual Congress of the European Society of Mycobacteriology. Cordoba, Spain (1997) abstract p 62. 30. Wilson SM, Nava E, Morales A, Godfrey-Fausset P, Gillespie S, Anderson N: Simplification of the polymerase chain reaction for detection of Mycobacterium tuberculosis in the tropics. Transactions of the Royal Society of Tropical Medicine and Hygiene (1993) 87 : 177–180 31. Thierry D, Brisson-Noël A, Vincent Lévy-Frébault V, Crawford ST, Bates JH, Gicquel B, Guesdon JL: Characterization of a Mycobacterium tuberculosis insertion sequence, IS6110, and its application in diagnosis. Journal of Clinical Microbiology (1990) 28 : 2668–2673 32. Hermans PWM, Van Soolingen D, Dale JW, Schuitema ARJ, McAdam RA, Catty D, Van Embden JDA: Insertion element IS986 from Mycobacterium tuberculosis: a useful tool for diagnosis and epidemiology of tuberculosis. Journal of Clinical Microbiology (1990) 28 : 2051–2088 33. Longo MC, Berninger MS, Hartley JL: Use of uracil DNA glycosylase to control carry-over contamination in polymerase chain reactions. Gene (1990) 93 : 125–128 34. Toman K: Sensitivity, specificity and predictive value of diagnostic tests. Bulletin of the International Union Against Tuberculosis (1981) 56 : 18–28 35. Ausina V, Gamboa F, Gazapo E, Manterola JM, Lonca J, Matas L, Manzano JR, Rodrigo C, Cardona PJ, Padilla E: Evaluation of the semiautomated Abbott LCX Mycobacterium tuberculosis Assay Direct Detection of M. tuberculosis in respiratory specimens. Journal of Clinical Microbiology (1997) 35 : 1996–2002 36. Vuorinen P, Miettinen A, Vuento R, Hällströ O: Direct detection of M. tuberculosis complex in respiratory specimens by Gen Probe Amplified M. tuberculosis Direct Test and Roche Amplicor M. tuberculosis Test. Journal of Clinical Microbiology (1995) 33 : 1856–1859 37. Piersimoni C, Callegaro A, Nista D, Bornizia S, De Conti F, Santini G, Desio G: Comparative evaluation of two commercial amplification assays for direct detection of Mycobacterium tuberculosis complex in respiratory specimens. Journal of Clinical Microbiology (1997) 35 : 193–196 38. D’Amato RP, Wallman AA, Hochstein LH, Colannino PM, Scardamaglia M, Ardila E, Ghouzi M, Kim P, Patel C, Miller A: Rapid diagnosis of pulmonary tuberculosis by using Roche Amplicor M. tuberculosis PCR Test. Journal of Clinical Microbiology (1995) 33 : 1832–1834 39. Cartuyvels R, De Ridder C, Jonckheere S, Verbist L, Van Eldere J: Prospective clinical evaluation of Amplicor M. tuberculosis PCR Test as a screening method in a low-prevalence population. Journal of Clinical Microbiology (1996) 34 : 2001–2003 40. Carpentier E, Drouillard B, Dailloux M, Moinard D, Vallee E, Dutilh B, Maugein J, Bergogne-Berezin E, Carbonnelle B: Diagnosis of tuberculosis by Amplicor M. tuberculosis test: a multi-center study. Journal of Clinical Microbiology (1995) 33 : 3106–3110 41. Bennedsen J, Thomsen VO, Pfyffer GE, Funke G, Feldman K, Beneke A, Jenkins PD, Hegginbothom M, Fahr A, Hengstler M, Cleator G, Klapper P, Wilkins EGL: Utility of PCR in diagnosing pulmonary tuberculosis. Journal of Clinical Microbiology (1996) 34 : 1407–1411 42. Bergman JS, Woods GL: Clinical evaluation of the Roche Amplicor PCR M. tuberculosis test for detection of M. tuberculosis in respiratory specimens. Journal of Clinical Microbiology (1996) 34 : 1083–1085 43. Kirschner P, Rosenan J, Springer B, Teschner K, Feldman K, Bottger EC: Diagnosis of mycobacterial infections by nucleic acid amplification: 18-month prospective study. Journal of Clinical Microbiology (1996) 34 : 304–312 44. Eing BR, Becker A, Sohns A, Ringelman R: Comparison of Roche Cobas Amplicor M. tuberculosis assay with in-house PCR and culture for detection of M. tuberculosis. Journal of Clinical Microbiology (1998) 36 : 2023–2029 45. Yuen KY, Yan WCH, Wong LP, Seto WH: Comparison of two automated DNA amplification systems with a manual one-tube nested PCR assay for diagnosis of pulmonary tuberculosis. Journal of Clinical Microbiology (1997) 35 : 1385–1389 46. Noordhoek GT, Van Embden JDA, Kolk AHJ: Reliability of nucleic acid amplification for detection of M. tuberculosis: an international collaborative quality control study among 30 laboratories. Journal of Clinical Microbiology (1996) 34 : 2522–2525 47. Brisson-Noël A, Gicquel B, Lecossier D, Levy-Frebault V, Nassif X, Hance A: Rapid diagnosis of tuberculosis by amplification of mycobacterial DNA in clinical samples. Lancet (1989) ii:1069–1071 48. Soini H, Skurnik M, Liippo K, Tala E, Viljanen M: Detection and identification of mycobacteria by amplification of a segment of the gene coding for the 32-kDa protein. Journal of Clinical Microbiology (1992) 30 : 2025–2028 49. Gamboa F, Manterola JM, Lonca J, Matas L, Viñado B, Giménes M, Cardona PJ, Padilla E, Ausina V: Detection and identification of mycobacteria by amplification of RNA and DNA in pretreated blood and bone marrow aspirates by a simple lysis method. Journal of Clinical Microbiology (1997) 35 : 2124–2128 867 50. Caylà JA, Galdós-Tangüis H, Jansà JM, García de Olalla P, Brugal MT, Pañella H: Evolución de la tuberculosis en Barcelona (1987–1995). Influencia del virus de la inmunodeficiencia humana y de las medidas de control. Medicina Clinica (Barcelona) (1998) 111 : 608–615 51. Das S, Chan SL, Allen BW, Mitchison DA, Lowrie DB: Application of DNA fingerprinting with IS986 to sequential mycobacterial isolates obtained from pulmonary tuberculosis patients in Hong Kong before, during and after short-course chemotherapy. Tubercle and Lung Disease (1993) 74 : 47–51 52. Van Soolingen D, de Haas PEW, Hermans PWM, Groenen PMA, Van Embdem JDA: Comparison of various repetitive DNA elements as genetic markers for strain differentiation and epidemiology of Mycobacterium tuberculosis. Journal of Clinical Microbiology (1993) 31 : 1987–1995 53. Yuen LKW, Ross BC, Jackson KM, Dwyer B: Characterization of Mycobacterium tuberculosis from Vietnamese patients by Southern blot hybridization. Journal of Clinical Microbiology (1993) 31 : 1615–1618 54. Pierre C, Lecossier D, Boussougant Y, Bocart D, Joly V, Yeni P, Hance AJ: Use of reamplification protocol improves sensitivity of detection of Mycobacterium tuberculosis in clinical samples by amplification of DNA. Journal of Clinical Microbiology (1991) 29 : 712–717