• Tuberculosis

We read with interest the article by Dheda et al1 who followed our approach for a rapid diagnosis of smear-negative tuberculosis by bronchoalveolar lavage (BAL) enzyme-linked immunospot (ELISpot)2 in a country of high tuberculosis incidence, including individuals with HIV-1 infection.

The authors report a sensitivity of 88.9% and specificity of 94.7% of the BAL ELISpot (T-SPOT.TB test) for the diagnosis of tuberculosis in suspects with scarce or negative acid-fast bacilli (AFB) sputum smears. This observation is important, as it confirms the findings of other recent studies performed in low tuberculosis incidence countries where flow cytometric assays were performed with BAL cells in order to obtain a rapid diagnosis of tuberculosis.3 4 However, flow cytometry is technically more demanding and time-consuming than ELISpot.

Results from the largest study performed on this topic to date, a recent prospective multicentre TBNET study, showed that the BAL ELISpot is superior to blood ELISpot, tuberculin skin test and Mycobacterium tuberculosis-specific nucleic acid amplification to diagnose sputum smear-negative tuberculosis.5

However, an important difference between this study and that of Dheda et al is the high frequency of indeterminate BAL ELIspot test results (9.2% vs 33.7%) that could be related to different cell processing procedures. Fifty-four percent of indeterminate results in the cohort from South Africa were due to lack of sufficient numbers of cells or failure of the positive control, interestingly unrelated to the patients' HIV serostatus. In 46.4% of the South African cohort and 82.1% of the European cohort the reason for indeterminate results was a high number of cells already producing interferon γ (IFNγ) without stimulation in the negative control. These are probably prestimulated terminally differentiated, cytokine-secreting effector T cells.

Different definitions of indeterminate test results are another important explanation for the observed variability between the two studies. When we reanalysed the data set of the TBNET study with the cut-offs used by Dheda et al, the sensitivity and specificity of the BAL ELIspot for the detection of sputum AFB smear-negative tuberculosis changed from 90.9% and 79.9% to 87.2% and 88.1%, and the frequency of indeterminate test results increased to 30.5%. Therefore, it would be interesting to know whether application of the cut-offs used in the TBNET study will substantially reduce the proportion of indeterminate test results in the study by Dheda et al.