There have been several reports describing the use of IHC in the diagnosis of tuberculosis (14). However, in this study we show that using an antigen against the secretory mycobacterial antigen MPT64, it is possible to achieve consistently high sensitivity (89-93%) and specificity (95-98%) with IHC on different types of tissues. The strength of this technology is that it is robust, readily available in routine surgical pathology laboratories and can detect fragmented tubercle bacilli . Compared with ZN staining that has a sensitivity of 10–45%  and requires an intact cell wall this technique offers a major improvement in diagnostic potential and should be suited for the diagnosis of pauci-bacillary EPTB. IHC for tuberculosis has, however, been slow to catch on as a routine diagnostic method in histopathology laboratories probably due to the lack of a specific anti-mycobacterial antibody suitable for all types of tissue [5, 14, 16] and hence the exact diagnostic role of IHC for M. tuberculosis has to be assessed in appropriate control groups and with appropriate antisera in endemic areas. Our large study is the first to show that IHC with an antibody to MPT64 is sufficiently robust to establish etiological diagnosis of M. tuberculosis complex infection in different types of tissues of EPTB.
Our results show that IHC with anti-MPT64 has better specificity, sensitivity, and predictive values than anti-BCG (table 4). This was particularly clear in intestinal wall tuberculosis, where anti-MPT64 showed a 100% specificity compared with 60% for anti-BCG (table 3). Anti-MPT64 antibodies also gave sharp and strong signals with clear background compared with anti-BCG antibodies making interpretation easier and permitting a more confident diagnosis of M. tuberculosis complex organism. Lower specificity with anti-BCG could be due to cross-reactivity with other infectious organisms as described earlier [20–22].
The sensitivity of anti-MPT64 is also very high but not very different from anti-BCG. The amplification method used in IHC improves the recognition of positive fragments. The few false negatives could be caused by the length of formalin fixation prior to processing which is known to reduce sensitivity. Another explanation for the false negative results may be that the number of mycobacteria present is below the sensitivity level of IHC (about 5 × 105 to 1 × 106 organisms per gram tissue) and the lesions are exuberant inflammatory responses to a minimal number of organisms.
Four of the negative controls were positive with both PCR and IHC with both antibodies. Among these were two lymph node samples with histological changes of non-specific lymphadenitis that may represent early or latent tuberculosis infection. According to Goel et al , early tuberculous lesions may not be identified by histopathology because the formation of granulomas and emergence of the classical histopathological tuberculous picture may be a late phenomenon. They suggest that such cases might represent the transition between the incubation and development of disease. Perhaps, in some cases, our IHC method can play a role in the early diagnosis of tuberculosis when histological examination fails to provide a diagnosis. The high prevalence of tuberculosis together with parasitic infection is well known in tuberculosis endemic countries. One of our controls with intestinal parasite turned out positive with IHC and PCR  and may well be a case of tuberculosis. It is difficult to explain the positive results of both tests on foreign body granulomas from Norway, however, as suggested by Mustafa et al, the possibility of latent infection cannot be ruled out .
In endemic countries, the majority of granulomatous lesions without necrosis are considered to be tuberculosis but this may not be the case in the developed world. Interestingly, when we looked specifically for the bacilli in the different zones of the granuloma they were more frequently detected by anti-MPT64 in the epitheloid cells than in the necrotic area. Using anti-BCG antibodies antigens were also detected in necrotic area. We also found that the percentage of stained cells was higher in non-necrotic granulomas than in necrotic granuloma with anti-MPT64 compared to anti-BCG with clearer and stronger signals. Hence, non-necrotic granulomatous lesion staining with MPT64 will support a diagnosis of tuberculosis.
Ideally, culture should be used as gold standard when comparing diagnostic test performance in tuberculosis. This investigation is, however, associated with low sensitivity especially in EPTB and as in our series, the corresponding results from controls are usually not available . Histopathology remains one of the most important methods for diagnosing tuberculosis, however; it cannot differentiate changes caused by M. tuberculosis, non-tuberculous mycobacteria or other granulomatous diseases. We used nested PCR as the reference for comparison. In recent years, the sensitivity and specificity of PCR for diagnosis of tuberculosis has been well documented and is in the range of 60–98% in reported series where PCR was compared with culture as gold standard [28, 29]. Our results also showed strong association between PCR and culture with all culture positive samples also being PCR positive. While PCR is increasingly used in the detection of mycobacteria from the tissue sample, the cost of the instruments and reagents, sensitivity to contamination and technical demand limits its use in developing countries .