Cord Formation in MB/BacT Medium Is a Reliable Criterion for Presumptive Identification of Mycobacterium tuberculosis Complex in Laboratories with High Prevalence of M. tuberculosis

Abstract

We evaluated cord formation in MB/BacT broth as a rapid method for presumptive identification of the Mycobacterium tuberculosis complex. Kinyoun acid-fast-stained smears from 370 positive MB/BacT bottles were examined for the presence of serpentine cording. The smears were examined independently by two observers. Observer 1 (the supervisor of the mycobacteriology laboratory) examined all of the smears while observer 2 (a clinical microbiologist not familiar with acid-fast bacillus [AFB] microscopy) examined 148 randomly chosen smears that were read by observer 1 without knowledge of which smear was which. The sensitivity, specificity, and positive and negative predictive values of cording for the presumptive identification of M. tuberculosis read by observer 1 were 88.2, 97.4, 99.2, and 69.7%, respectively. These values were reported at 90.6, 52.3, 82.8, and 69.7%, respectively, by observer 2. Our laboratory prevalence of M. tuberculosis among positive cultures was 78% during the time this study was conducted. At the time of positive signal of the MB/BacT bottles, the broth of the bottles had sufficient cell mass to allow for observation of the presence or absence of serpentine cording. The presence of cords in MB/BacT broth is a reliable criterion for rapid, predictive identification of the M. tuberculosis complex for laboratories with a high proportion of the M. tuberculosis complex when the smears are examined by a microbiologist who has experience with AFB staining.

It has been reported that cord factor plays a fundamental role in the genesis and persistence of the chronic and granulomatous lesions caused by mycobacteria (9). Virulent strains of the Mycobacterium tuberculosis complex (M. tuberculosis) often grow as characteristic ropes, bundles, or serpentine cords of acid-fast bacilli (AFB) in liquid media and are influenced by the composition of the culture medium and the conditions of culture (3, 5). Avirulent variants of M. tuberculosis and most mycobacteria other than M. tuberculosis (MOTT) grow in liquid media in a nonoriented, dispersed fashion (10). In general, MOTT grows in broth media as dispersed single cells or as small clumps of bacilli with random orientation. However, MOTT strains produce loose, incomplete pseudocords (4, 7, 10). Cording or pseudocords have previously been reported with other Mycobacterium species: M. kansasii, one M. avium complex, M. szulgai, M. chelonae, M. phlei, M. terrae, and M. gordonae (1, 4, 10).

Several researchers have reported that cord formation in the BACTEC medium is a reliable criterion for rapid presumptive identification of M. tuberculosis; however, the sensitivity, specificity, and positive and negative predictive values were reported between 23 and 95, 89 and 99, 83 and 99, and 80 and 98%, respectively (1, 4, 6, 10). It can be discerned from these reports that the laboratory prevalence of M. tuberculosis among positive cultures was between 25 and 36%. These researchers urged other laboratories to develop their own experience and determine the reliability of cord formation for presumptive identification of M. tuberculosis in their own laboratories. In our laboratory, we use the MB/BacT system (Organon Teknika, Turnhout, Belgium) which relies on a continuous colormetric CO₂ detection device to indicate the mycobacterial growth in a closed, automated system as a broth culture medium for mycobacterial cultures. Our laboratory prevalence of M. tuberculosis among mycobacterial isolates is between 75 and 78%. One report has defined the MB/BacT system as a well-automated system for detection of mycobacteria in clinical specimens, compared to the BACTEC 460TB system and solid media (8). There has been no report on the utility of cord formation in positive MB/BacT broth. We therefore evaluated the presence of cord formation in positive MB/BacT bottles as a method of predictive identification of M. tuberculosis.

In this study, 370 positive mycobacterial cultures from 201 patients were examined for the presence or absence of cording in MB/BacT bottles from April 1996 to November 1998. Clinical specimens included 262 respiratory specimens (sputum, tracheal aspirate, pleural fluid, lung biopsy samples, and bronchoscopy samples) and 108 nonrespiratory specimens (27 urine, 3 bone marrow, 24 tissue and pas, 11 cerebrospinal fluid, 16 joint fluid, 1 blood, 14 gastric aspiration, and 12 other body fluid samples). All nonsterile specimens were digested and decontaminated by the NALC–4% NaOH method and were concentrated by centrifugation (2). Sediments were cultured into one MB/BacT bottle according to the manufacturer's instructions and onto two Loweinstein-Jensen (LJ) slants prepared media (Oxoid). All LJ cultures were incubated for 8 weeks at 35°C in an atmosphere of 10% CO₂ and were examined weekly. MB/BacT bottles were incubated for 8 weeks at 35°C in the system incubator and were analyzed every 10 min by the system software. When a bottle flagged positive, one smear was made on the MB/BacT broth without vortexing of the broth. The smears were stained by the Kinyoun staining method (Difco-Bacto TB Carbolfuchsin KF primary satin) (2). Smears were examined independently by observer 1 (supervisor of the mycobacteriology laboratory) and observer 2 (a clinical microbiologist not familiar with AFB microscopy). The entire smears, which were stained by the Kinyoun method, were first scanned at a magnification of ×100 for the presence of AFB, and then they were evaluated at a magnification of ×1,000 to differentiate serpentine cords from clumps of bacilli with random orientation or to visualize single dispersed AFB. Serpentine cords were defined as tight, rope-like aggregates of AFB in which the long axis of the bacteria paralleled the long axis of the cord. Smears with only a few organisms on the slide were not evaluated for the presence of cording, due to insufficient quantity of organisms present, and were not included in this study.

Identification of all mycobacteria was determined either by conventional biochemical testing or with AccuProbe Mycobacteria Culture Identification Probe Kits (Gen-Probe, San Diego, Calif.). Also, the broth of positive MB/BacT bottles was subcultured onto a 7H11 agar plate and two LJ slants for purity and growth control.

Kinyoun-stained smears were prepared from 370 positive MB/BacT broth cultures and were examined by observer 1 for the presence of cords of AFB. Fifteen (4%) smears were not evaluated for cording, due to an insufficient number of organisms on the slides, and they were not included in the study. Their cultures grew 10 M. tuberculosis, 4 M. gordonae, and 1 Mycobacterium fortuitum species. The remaining 355 positive cultures yielded 357 mycobacteria. Of these 355 cultures, 353 were pure growth cultures, 1 was a mixture of M. tuberculosis and M. fortuitum, and 1 contained M. tuberculosis with M. gordonae. Cording was not seen on the slides for these two bottles. Table 1 lists the 357 mycobacteria recovered and their cording results. Observer 1 examined all of the 355 smears (275 of the 355 smears were prepared from cultures in which M. tuberculosis was grown, 2 smears were prepared from mixed cultures, and 76 smears were prepared from cultures in which MOTT was grown). Observer 1 read all of the smears at a sensitivity of 88.2%, specificity of 97.4%, positive predictive value of 99.2%, and negative predictive value of 69.7% for serpentine cording. Cords were reported by observer 1 in 88.2% (n = 246) of the M. tuberculosis isolates, but cording was not reported by observer 1 for 11.8% (n = 33) of M. tuberculosis isolates. Both of the false positive smears that were read as cord positive by observer 1 were prepared from cultures that grew M. gordonae.

TABLE 1.

Mycobacteria isolated from positive MB/BacT cultures and cord formation results reported by observer 1

Species	No. (%) of cultures	Cord formation
		No. positive (%)	No. negative (%)
M. tuberculosis complex	279 (78.1)	246 (68.9)	33 (9.2)
M. gordonae	40 (11.2)	2 (0.6)	38 (10.6)
M. fortuitum	6 (1.7)		6 (1.7)
M. scrofulaceum	6 (1.7)		6 (1.7)
M. abscessus	6 (1.7)		6 (1.7)
M. chelonae	5 (1.4)		5 (1.4)
M. flavescens	4 (1.1)		4 (1.1)
M. xenopi	3 (0.8)		3 (0.8)
M. szulgai	3 (0.8)		3 (0.8)
M. terrae complex	3 (0.8)		3 (0.8)
M. triviale	2 (0.6)		2 (0.6)
Total	357 (100.0)	248 (69.5)	109 (30.5)

We also wanted to determine whether the observer's experience with reading acid-fast smears is important to report true cording. Thus, observer 2 read 148 randomly chosen smears that were read by observer 1 (who has experience with AFB microscopy) without knowledge of which smear was which. Of these smears, 106 were prepared from cultures in which M. tuberculosis was grown and 42 were prepared from cultures in which MOTT was grown. The sensitivity, specificity, positive predictive value, and negative predictive value of serpentine cording were reported at 90.6, 52.3, 82.8, and 68.7%, respectively, by observer 2. Observer 2 reported the cords in 90.6% (n = 96) of the M. tuberculosis isolates. Cording was not reported for 9.4% (n = 10) of the M. tuberculosis isolates read by observer 2. In retrospect, the two M. fortuitum cultures and one Mycobacterium flavescens culture that were read as cording positive by observer 2 contained too many organisms per microscope field to adequately assess the arrangement of the individual bacilli. The remaining 17 smears that were read as cord positive by observer 2 were prepared from mycobacterium cultures that grew M. gordonae (nine cultures), M. scrofulaceum (three cultures), M. szulgai (two cultures), M. xenopi (one culture), M. abscessus (one culture), and the M. terrae complex (one culture). When they were examined for a second time after the final identification, these cultures did not show typical cord morphology and the arrangement of the bacilli illustrated pseudocords. The results of the smear examination by the two observers are given in Table 2. The kappa statistic value (chance corrected measure of agreement) of the two observers gave only a 82.6% (P = 0.000) agreement for M. tuberculosis and a 10.4% (P = 0.12) agreement for MOTT.

TABLE 2.

Cord formation results of M. tuberculosis and MOTT reported by observer 1 and observer 2

Cord formation	No. of stained smears identified as the following by observer 1 or 2
	M. tuberculosis		MOTT
	Observer 1	Observer 2	Observer 1	Observer 2
Positive	246	96	2	20
Negative	33	10	76	22
Total	279	106	78	42

Morris and Reller (6) found that the sensitivity of the exam for cord formation is lower than that observed in other studies (22.9% versus 89.2, 95.2, and 93.2%) (1, 4, 10). The author mentions that the staining method of the smears may affect the test results. In our study, to investigate the effect of the staining method, two smears were prepared from 85 bottles that grew M. tuberculosis. One smear was stained by the Kinyoun method, and the other was stained by the fluorochrome (Auromine-Rhodamine) staining method. The examination by observer 1 found no difference in the cord morphology. The sensitivity of both stains was 100%.

Morris and Reller (6) also investigated the effect of vortexing on cord morphology where each BACTEC vial was vortexed before sampling for the smear. No difference in cord morphology for the 10 vortexed vials was reported. In our study, we evaluated the effect of vortexing on cord morphology for 57 randomly chosen MB/BacT bottles with growth characteristics suggesting M. tuberculosis. Since the MB/BacT manual does not detail how positive MB/BacT broth should be handled before the smear is made, one smear was made before vortexing of the bottles and another was made after vortexing of the bottles for 5 to 6 s at setting 10 with a Vortex Stuart SA2 (Scientific Automixer). The smears were read by observer 1 without knowledge of which smear was which. Fifty-three (93%) smears were read as positive for cording. Loss of cord formation was seen on only four smears. When these smears were examined under microscopy for a longer period of time and more fields were included, it was seen that all four smears very rarely had cord formation on the entire field. There was no significant difference in cording after vortexing, although the amount of cords decreased after vortexing.

In this study, 11.8% of M. tuberculosis isolates exhibited no cord formation. This is consistent with the findings of McCarter et al. and Yagupsky et al. (4, 10). We report the positive predictive value at 99.2%. Our laboratory prevalence of M. tuberculosis among positive cultures was 78% during the time period that this study was conducted. The numbers of MOTT were small, and many species were not represented. It is therefore possible that our negative predictive value is lower than those of other reports (69.7% versus 80 to 98%, respectively). Since some laboratories may not have such a large proportion of M. tuberculosis isolates, the positive predictive values have been recalculated by using 1, 5, 10, 15, 25, and 50% prevalence of M. tuberculosis among positive cultures. With these prevalences, the positive predictive values would be at 25.5, 64.1, 79, 85.7, 91.9, and 97.1%, respectively, while the negative predictive values would be as high as 99.9, 99.4, 98.7, 97.9, 96.2, and 89.2%, respectively. We agree with McCarter et al. (4) that in laboratories with a small proportion of M. tuberculosis isolates, the positive predictive value of cording for presumptive identification of M. tuberculosis would be unacceptably low. At the time of positive signal of the MB/BacT bottles, the broth of the bottles had sufficient cell mass (equivalent to or greater than the McFarland #1 Nephelometer Standard) to allow for observation of the presence or absence of serpentine cording. Only 4% of the positive bottles were not evaluated due to the insufficient number of organisms. If the broth has only a few organisms, it may not exhibit cord formation. These smears should not be considered negative for cording, and reevaluation of the broth must be done by concentration of the broth or after reincubation of the bottle. In this study, 15 cultures were excluded from the study, due to the presence of small amounts of bacilli, and the bottles were reincubated for 1 week at 35°C. After this incubation, 10 of the 15 bottles represented cording on microscopy and had M. tuberculosis growth.

It should be noted that recognition of true cording sometimes would be difficult for the inexperienced microscopist. The loose, incomplete pseudocords produced by some MOTT may be misinterpreted as true cording. This is reflected by the differences in observer 1 and observer 2's results.

The evaluation of the presence of cord formation with a careful examination of microscopic morphology of cords in MB/BacT broth is reliable and provides rapid preliminary information in the identification procedure. Decisions regarding susceptibility tests can be guided by the results from the exam for cord formation. It can be used as a laboratory guide for initial DNA probe selection, thereby decreasing laboratory cost and turnaround time. Further, it permits the rapid presumptive reporting of M. tuberculosis to physicians, enabling the rapid and accurate detection and treatment of patients with tuberculosis.