Review showed that similar methods for specimen processing were used in all three participating laboratories. However, one laboratory used a common flask for dispensing decontaminant reagent and phosphate buffer rather than an individual tube or pipette for each specimen.

During the study period, 21,835 specimens were submitted for mycobacterial culturing at the three laboratories. Of these, 988 (4.5%) from 296 different patients were positive for M. tuberculosis. Twenty-seven had only a single positive culture. Of these, specimens from 10 patients met criteria for an investigation of possible laboratory cross-contamination (Table).

After the panel’s review, laboratory cross-contamination was identified as the cause for positive culture results for six patients (2% of all patients with cultures positive for M. tuberculosis) (Table). Rates were similar at two of the three laboratories, ranging from 2.8% in both Laboratories 1 (1 of 36 patients) and 2 (5 of 179 patients) to 0% (0 of 81 patients) with a culture positive for M. tuberculosis in Laboratory 3. At Laboratory 2, a common flask was used for dispensing reagents during the study period. One of the cross-contamination incidents (involving Patient 4) probably resulted from a malfunctioning broth-culturing system (BACTEC 460) (Becton Dickinson Microbiology Systems, Sparks MD). Patient 4’s specimen was in the BACTEC instrument immediately after another patient’s specimen (not processed on the same day), and genotypes of the two isolates matched. Cultures from two patients (Patients 7 and 9) were cross-contaminated from the same source patient during sample processing. All six of the laboratory cross-contamination incidents occurred with the initial rather than follow-up specimens for mycobacterial culture. Five of these six patients were treated for TB. Of the remaining four patients whose isolates were suspected of being cross-contaminated in the laboratory, one had a false-positive culture attributed to specimen mislabeling by a health-care provider; one had either a mislabeled specimen or mixed infection; one had active TB; and one had either a mislabeled specimen or a cross-contaminated specimen (Table). Having only a single positive culture was highly associated with laboratory cross-contamination (p<0.001, Fisher exact test).

Laboratory cross-contamination was the cause for a positive culture result in 2% of all patients with M. tuberculosis–positive cultures. Cross-contamination accounted for one fifth (22%) of patients having only one culture positive for M. tuberculosis. Of the six patients who had cross-contaminated cultures, five were treated unnecessarily with multiple anti-TB medications.

The rate of cross-contaminated cultures in our study is similar to the rates in previous studies of M. tuberculosis cultures. Most population-based studies found rates of 0.9% to 3.5% (1–8). However, such studies were retrospective and did not assess the extent of the problem in different types of clinical mycobacteriology laboratories. In this study, we used predefined criteria, which were based largely on DNA genotyping, to identify suspected cases of laboratory cross-contamination prospectively in an effort to correct factors associated with its occurrence. Our study included all clinical specimens submitted during a 1.5-year period to one county public health laboratory and two county hospital laboratories. This study was possible because of the large databank of RFLP results conducted as part of being a member of the genotyping network.

Multiple factors can cause false-positive cultures, including contaminated clinical equipment (e.g., bronchoscope), clerical errors, and cross-contamination that occurs in the laboratory. The last category can be caused by batch processing, transfer of viable bacilli from the sample needle of a broth-culturing system, e.g., BACTEC (15), a faulty exhaust hood (4), and contamination from species identification procedures such as the niacin production test (6).

Five of the six cross-contamination incidents were in a single laboratory. In four of these five cases, contamination probably occurred when reagents were dispensed with a common flask. Previous studies have reported that the step of adding the phosphate buffer was likely to have been the source of the cross-contamination (1,6,9). This procedure was later discontinued on the basis of the results of this study. None of the laboratories used positive control cultures.

As in other studies, we found that a single positive culture for M. tuberculosis was a sensitive but nonspecific marker for detection of a false-positive culture, since most patients (78%) with a single positive culture had TB. In a New York study (7), 12 (44.4%) of 27 patients with a single positive culture had a false-positive culture. These findings suggest that clinicians and laboratorians should be increasingly suspicious of a single false-positive culture. Additional specimens should be collected in cases of a single false-positive culture and the patient evaluated carefully for TB and other illnesses; the laboratory should also retain the isolate and others processed that day for genotyping. Because all specimens that met the inclusion criteria in our study were from the respiratory tract, we cannot draw any conclusions about the rate of cross-contamination of nonrespiratory specimens (e.g., cerebrospinal or pleural fluid). Nor can we draw any conclusions about a single positive culture when only a single specimen is submitted to the laboratory, as is often the case with nonrespiratory specimens.

Clinical judgment is also important in raising suspicion about cross-contamination. TB classically is accompanied by symptoms of prolonged cough, fever, weight loss, and night sweats, but other diseases such as bacterial pneumonia can cause these symptoms. Therefore, no specific clinical criterion alone can be used to definitively state that TB is present. However, an inconsistent clinical course or absence of symptoms should certainly raise suspicion that cross-contamination may have occurred (3,8,16). A determination regarding the presence of cross-contamination requires a thorough evaluation of a patient’s symptoms and clinical course as well as laboratory evaluation, with additional specimens obtained if only a single culture is positive as described above. Genotyping should be performed if cross-contamination is suspected on the basis of an inconsistent clinical course or the presence of only one positive culture (8).

Our assessment of the rate of cross-contamination did not include private laboratories; thus, our results may not be generalizable to all types of clinical laboratories. In addition, our methods depended on identifying, obtaining, and genotyping an isolate from a positive source culture; thus, we may have underestimated the true rate of laboratory cross-contamination.