The Effect of Changes in Laboratory Practices on the Rate of False-Positive Cultures for Mycobacterium tuberculosis

False-positive cultures for Mycobacterium tuberculosis are relatively common. In a recent review of the literature, false-positive results were detected in 13 of 14 (95%) DNA fingerprinting studies that evaluated more than 100 patients, with a median false-positive rate of 3.1% (interquartile range, 2.2%–10.5%).1 False-positive cultures can result from clerical error,2 contamination of clinical equipment (eg, bronchoscopes),3 misidentification,4,5 and laboratory cross-contamination (during initial batch processing,6,7 from inadequate clearance of contaminated aerosol,8 or from carryover on the sampling needle of the BACTEC 460 radiometric broth culture system [BD Biosciences, Sparks, Md]).9 

In a previous study, we used DNA fingerprinting to evaluate samples processed during a 5-year period (1989–1994) and found a false-positive rate of 4% (8/199).6 After reviewing the mechanisms of cross-contamination, we made changes in laboratory policies and techniques with the goal of reducing the rate of false-positive cultures. Six years after making these changes we reevaluated the false-positive rate using the same methodology that we used in our previous study.

The Mycobacteriology Laboratory of Denver Health serves the Denver Metro Tuberculosis Clinic and Denver Health Medical Center with its associated specialty and Community Health Centers. The laboratory is staffed primarily by one dedicated mycobacteriology technologist; 7 general microbiology technologists from Denver Health Medical Center provide weekend and vacation coverage. The laboratory has used broth (BACTEC 460) and solid media (Mitchison/Middlebrook biplate, Remel, Lenexa, Kan) for all specimens processed from 1988 to the present. Specimens from nonsterile sites (eg, sputum, urine) undergo initial processing (decontamination, buffering, centrifugation) in batches of up to 12 specimens. A DNA hybridization test (AccuProbe kit, Gen-Probe, San Diego, Calif) was used to identify M tuberculosis.

Other local hospital laboratories, which were sampled in both the past study and this study, generally had low volumes of mycobacterial cultures. The equipment and staffing differed in each laboratory. None of these laboratories made specific changes in policies or procedures to decrease cross-contamination.

Because previous studies have shown that acid-fast-bacilli smear-positive specimens are more likely than smear-negative specimens to be the source of bacilli that cross-contaminate other specimens,6 we made an administrative change to reduce the number of smear-positive specimens processed in the laboratory (Table 1). After determining that the yield of culture in our laboratory of smear-positive specimens obtained from patients not on therapy was nearly 100%,10 we made a policy of accepting a maximum of 2 smear-positive specimens from any patient within a 1-month time period.

Changes in laboratory techniques were designed to minimize the possibility of cross-contamination during batch processing. These changes included processing raw, nonsterile specimens separately from specimens with a known high risk of causing contamination and paying more careful attention to the mechanisms of cross-contamination during batch processing (Table 1). Specimens with a known high risk of being the source of cross-contamination included laboratory proficiency-testing specimens (previously documented to be the source of cross-contamination6), broth from BACTEC vials being reprocessed to remove bacterial or fungal contamination,6 and specimens from patients with prior positive cultures. We followed standard guidelines for use of the BACTEC 460 apparatus. We switched to individual aliquots of buffer solution, prepared in single-use tubes. We continued to use a common container of sodium hydroxide solution (a fine-tipped spray plastic spray bottle) because the extremely caustic nature of this decontamination solution makes it an unlikely source of cross-contamination.

We fingerprinted one isolate from each anatomic site from all patients having a positive culture for M tuberculosis. In addition, we fingerprinted isolates obtained more than 90 days after patients started treatment, as well as all nonpatient specimens (laboratory proficiency-testing specimens) containing M tuberculosis that were processed in the laboratory. Specimens from other laboratories were collected as part of an ongoing effort to collect specimens from all patients with culture-positive tuberculosis in Denver. When cross-contamination was suspected, we attempted to obtain a sample of all isolates processed that same week in the laboratory.

Isolates were fingerprinted using standard techniques: IS6110 was used as the initial DNA fingerprinting technique,11 and pTBN12 was used as the secondary DNA fingerprinting technique for isolates that had fewer than 6 hybridizing bands with IS6110.12 The DNA fingerprint results were interpreted without knowledge of the clinical characteristics of the patients.

We identified patients for whom only one specimen had a positive culture (single-positive patients) as the first step in evaluating possible false-positive cultures; we used this approach because the probability of contamination of multiple specimens from a single patient is extremely small. As in our previous study, we then evaluated the clinical, radiographic, and microbiologic data of any patient with a single positive culture for which the DNA fingerprint matched that of another isolate processed in the laboratory within 42 days.6 We defined a false-positive culture as a culture from a single-positive patient for which the DNA fingerprint matched that of another isolate if the patient from whom it was obtained (1) lacked convincing clinical, radiographic, or histopathologic evidence of tuberculosis, and (2) did not have an identifiable epidemiologic link to the patient from whom the possible source specimen was obtained.6 The false-positive rate was defined as the number of patients with a false-positive culture divided by the total number of patients with a positive culture for M tuberculosis.

Of the 13 940 specimens processed in the Mycobacteriology Laboratory of Denver Health from July 1994 to December 1999, 630 (4.5%) from 184 patients grew M tuberculosis. In addition, 48 nonpatient specimens grew M tuberculosis. There were 62 single-positive patients; none of the isolates from these patients matched an isolate processed within 42 days. Therefore, the false-positive rate was 0% (0/184). This rate was significantly lower than the 4% rate (8/199) determined in our previous study (P = .008). Of the 184 patients who had a positive culture in our laboratory during the study period, 31 (17%) had isolates that matched another isolate processed in the laboratory.

Three of 22 specimens (13.6%) obtained from other laboratories in Denver had false-positive cultures. This rate was similar to the false-positive rate of 11.9% (5/42) determined for other laboratories in Denver in our previous study (P = .84). An additional patient was strongly suspected of having a false-positive culture by the treating clinician, the tuberculosis control program, and the laboratory involved. However, the isolate from this patient and the other culture-positive specimens processed that day were not available for fingerprinting. The patient was not treated for tuberculosis and did well.

The mechanisms of false-positive cultures were probable contamination during initial batch processing of nonsterile specimens (2 of the documented false-positive cultures and the single suspected false-positive culture) and misidentification of an isolate as M tuberculosis (Table 2). The misidentification was due to a false-positive DNA hybridization test (AccuProbe). This isolate was positive on repeated AccuProbe testing in our laboratory but did not have the conventional microbiologic characteristics of M tuberculosis (no cording and negative results on tests for niacin and nitrate). Furthermore, the fatty acid profile of this isolate by high-pressure liquid chromatography (done at the Mycobacteriology Laboratory of the Centers for Disease Control and Prevention, Atlanta, Ga) was not characteristic of M tuberculosis or any other currently recognized mycobacterial species.

Only 2 of the patients with false-positive cultures were treated for tuberculosis, and these 2 patients were treated for less than 2 weeks before the false-positive culture was suspected and treatment was stopped.

After relatively simple administrative and procedural changes in our laboratory, the rate of false-positive cultures for M tuberculosis decreased from 4% (8/199) to 0% (0/184). The study compared the rate of false-positive cultures in 2 different time periods, but it is unlikely that the decrease in false-positives was due to changes other than those outlined above. The laboratory used the same culture methodology during both time periods (BACTEC 460 plus solid media), and we used a consistent, rigorous algorithm to detect and evaluate possible false-positive cultures. Furthermore, it is notable that the false-positive rate among samples processed by other laboratories in Denver, which did not make changes in procedures to reduce the rate of false-positive cultures, did not change. Therefore, we conclude that the specific administrative and procedural changes made in our laboratory were effective in reducing the rate of false-positive cultures.

We made changes in laboratory policies and procedures based on a review of the mechanisms of false-positive cultures noted in previous studies. Our first concern was to limit the number of smear-positive specimens processed in the laboratory. Smear positivity is associated with high concentrations of mycobacteria, generally 105/mL,13 and therefore a greater likelihood of being the source of a droplet or splash that results in cross-contamination. In a previous study, we demonstrated that the yield of mycobacterial culture from acid-fast-bacilli smear-positive specimens obtained from patients not on treatment was 99.2% (436/439).10 Therefore, we were able to limit the number of high-risk specimens processed without compromising the diagnostic yield of mycobacterial culture.

Our changes in laboratory procedures focused on batch processing because cross-contamination at this step appears to be the most common mechanism of false-positive cultures.1 The first change was that specimens with a high likelihood of being culture positive were processed separately and not in a batch with other nonsterile specimens. Such specimens were those from laboratory proficiency programs, BACTEC vials being reprocessed for bacterial or fungal contamination, and specimens from patients with previous positive cultures. The second set of changes in laboratory technique was specifically designed to reduce cross-contamination. Contamination of multiuse containers of buffer is the most common mechanism for large outbreaks of cross-contamination,7,14 so we began using individual aliquots of buffer. Centrifugation and agitation of specimens to mix decontamination or buffering solutions with the specimen can result in the generation of an aerosol that can cross-contaminate other specimens,8 so we added a 5-minute waiting period to allow aerosolized particles to settle before opening the processing tube.

Our study highlights misidentification as a cause of false-positive culture reports. The use of DNA probes to identify the M tuberculosis complex (M tuberculosis, Mycobacterium bovis, or Mycobacterium africanum) reduces the time required for species identification, but other mycobacterial species can hybridize with the sequence in the ribosomal RNA gene sequence used in the AccuProbe kit. For example, Mycobacterium celatum,5 ,Mycobacterium terrae,4 and the unnamed mycobacterial species from one of our patients can cause a false-positive result on a DNA probe test. Therefore, laboratories should consider the inclusion of alternative speciation methods for isolates that are positive by DNA probe analysis but that do not have the colony morphology or cording that are characteristic of the M tuberculosis complex.

The number of patients with false-positive cultures in the present study was small. Nevertheless, it is notable that none of these patients received more than 2 weeks of unnecessary treatment, whereas most patients in our previous study of false-positive cultures received a complete course of treatment.6 Our experience is that once clinicians know of the possibility of false-positive cultures, it is not difficult to recognize cases of misdiagnosis of active tuberculosis due to a false-positive culture.

The restriction fragment length polymorphism analyses in this study were supported by the Centers for Disease Control and Prevention, National Genotyping Network cooperative agreement.