METHODS

Sputum samples from patients with a high clinical probability of tuberculosis but who were not under treatment at the time of sampling, and from patients known to be colonised with non-tuberculous mycobacteria, were identified as they underwent normal processing in the laboratory. Where sample size permitted, part of the sample was set aside and stored at 4–8°C, pending the results of culture. Routine processing in this laboratory comprises the following steps:

1. neat sputum

2. liquefaction (dithiothreitol)

3. concentration (centrifugation)

4. decontamination (4% NaOH)
 followed by neutralisation (K₂ PO₄)

5. centrifugation
 followed by culture for mycobacteria

The liquefaction agent used is dithiothreitol (Sputasol^™; Oxoid Ltd, Basingstoke, UK), which we feel to be superior to N-acetyl-cysteine.⁴ Centrifugation is carried out at 3000 ×g for 20 minutes at both stages 3 and 5. All samples are cultured in the BacT/ALERT 3D automated mycobacterial system and simultaneously on Löwenstein-Jensen slopes, in each case for up to eight weeks. In practice, all positive samples included in our study yielded mycobacterial growth within 28 days. All mycobacterial isolates were identified by standard methods.

Once the results of culture were known, 78 sputum samples were identified from which a mycobacterium had been isolated. The stored material from these specimens was then, in each instance, re-processed as already described, aliquots being removed for staining at each of the five numbered stages, giving a total of 390 aliquots. Two smears were made from each aliquot; one of which was stained by the ZN method,⁵ and examined at ×1000 magnification by transmitted light. The other smear was stained by the AP method⁵ and examined at ×400 magnification using a fluorescence microscope equipped for epifluorescence. In both instances, the numbers of acid fast bacilli seen were converted to a rods/field count. To compensate for the difference in field size of the two methods, 40 fields were examined for AP smears and 200 fields for ZN smears. Each smear was examined by two separate workers and graded⁶ by consensus on the following scale: 0, acid fast bacilli not seen; < 1, < 1 rod/field (on average); 1–9, from one to nine rods/field; 10–90, between 10 and 90 rods/field; > 90, uncountable number, greater than 90/field.

Results were analysed and compared with the culture data.

RESULTS

Of the mycobacteria isolated from the 78 culture positive sputum samples, 52 were members of the Mycobacterium tuberculosis complex (MTBC), 12 were of the M avium complex (MAC), nine were M malmoense, four were M kansasii, and one was M xenopi. The possibility of cell wall deterioration leading to reduced smear sensitivity was considered even though samples were stored at low temperature before inclusion in our study. However, 70 of the 78 samples were smear positive by at least one method, so that storage did not appear to impact on the results. Mycobacteria isolated from the eight smear negative, culture positive samples comprised three MTBCs, one MAC, three M kansasii, and one M malmoense. Tables 1–5 detail the results of the smear comparisons.

In all instances, smears stained by the AP method showed a higher level of detection of acid fast bacilli than those stained by the ZN method, and higher counts of acid fast rods within positive smears were noted using the AP method. The highest detection rate by both methods was in the examination of liquefied and concentrated samples (table 3), and the lowest detection rate by both methods was in the examination of post-NaOH treated samples (tables 4,5). Material examined after NaOH decontamination, whether concentrated or not, gave particularly poor results in comparison with other methods, and samples were more difficult to fix to slides. Comparatively poor results were also obtained in the examination of neat (that is, untreated sputum).

Statistical analysis

The real results were compared with the expected results for the null hypothesis of no difference between the methods (and between the treatment regimens) using the χ² test. The AP method was significantly better than the ZN method with all pretreatment regimens and the differences between pretreatment regimens were also significant. The liquefied and concentrated samples performed best for both staining methods. All differences were significant at p > 0.01—for example, AP versus ZN on liquefied and concentrated samples, χ² = 15.3 with 4 degrees of freedom, p = 0.0041.

DISCUSSION

Our results strongly suggest that the optimum detection of acid fast organisms is achieved by the application of auramine/phenol fluorescent staining to digested (liquefied) and concentrated samples. The consistent superiority of the AP stain over the ZN stain in all the settings tested shows that the common practice of trying to “confirm” a positive AP smear by ZN is unscientific.

“A suboptimal smear performance may significantly compromise or delay isolation and contact tracing measures”

Two other important points must be borne in mind. First, it has been suggested that smear positivity should be used as the criterion for selecting sputum specimens for examination by continuous automated mycobacterial liquid culture,⁷ a much more rapid method for mycobacterial isolation, the inference being that smear negative samples will, by default, receive conventional solid culture. Suboptimal smear performance will mean that some patients’ samples that should receive immediate attention do not do so, and diagnosis will be delayed. Second, the smear is often used as an index of infectivity. It is evident that smear positive patients will be more infectious than smear negative patients. However, in studies in which the smear positivity versus the culture rate is over 85%,⁸ a clear risk of transmission from smear negative culture positive patients has been demonstrated. A suboptimal smear performance may significantly compromise or delay isolation and contact tracing measures.