In many resource-limited countries, diagnosis of TB is based on microscopy of ZN-stained smears. However, fluorescence microscopy of Auramin-stained smears may be more sensitive and more quick. LED microscopes have replaced conventional halogen fluorescent microscopes and have brought fluorescent microscopy and Auramin staining into the reach of resource-limited countries 
. In 2010, the WHO recommended that LED microscopy using Auramin staining be phased in as an alternative for conventional ZN microscopy in both high- and low-volume laboratories 
. However additional data in high TB burden settings are needed to define the optimal technical conditions and its performance, especially among HIV infected patients.
We evaluated the performance of LED-FM in a hospital setting in Indonesia. LED-FM was more sensitive but less specific than ZN for the diagnosis of pulmonary TB both in unselected patients with TB and in HIV-infected patients. The diagnostic accuracy of LED-FM was better if a threshold of ≥2 AFB/length of the smear was used instead of the WHO threshold of ≥1 AFB/length. When sputum was concentrated before smear preparation, sensitivity and specificity of LED-FM increased. In HIV-infected patients, most of whom had advanced disease, LED-FM showed a slightly higher sensitivity and lower specificity compared with conventional ZN-microscopy. Finally, reading time of LED-FM was less than half from ZN-microscopy, while running costs were similar.
In line with previous studies we found that LED-FM had a higher sensitivity, but a lower specificity compared to ZN-microscopy 
. By using the new definition of a positive smear (≥1 AFB/length) from WHO 
, LED-FM in our study showed a relatively low specificity, resulting in a significant number of false-positive results based on ‘scanty smears’. Using ROC analysis, ≥2 AFB/length appeared to be the optimal threshold for LED-FM.
Although the number of studies is limited, fluorescence microscopy has shown promising results for diagnosing TB in HIV-infected individuals. Two studies 
included in the systematic review 
reported increased sensitivity of conventional FM. In a study from Kenya, FM was twice as sensitive as ZN for HIV-infected cases using culture as a gold standard 
. A second study in India showed an incremental yield of 26% for FM over ZN staining among 200 patients with clinically and radiologically diagnosed TB, including 15% HIV-positive patients 
. These studies were performed with conventional FM.
In our study among 256 HIV-infected patients, by using the new cut-off (≥2 AFB/length) LED-FM had a somewhat higher sensitivity (65.2% vs 58%), albeit lower specificity compared to ZN-microscopy (90.4% vs 96.3%). This is in line with one study among 426 patients from Uganda, 85% of whom were HIV-infected, published after Steingart’s systematic review 
. This study showed that conventional FM was more sensitive than ZN (49% vs 38%), but less specific (81% vs 96%) due to many scanty FM results. In that study the optimal performance of FM was seen when cut-off >4 AFB/100 fields used to define positive smears 
. Whitelaw et al compared the accuracy of LED-FM with ZN in 345 patients, including 88 who were HIV-infected patients 
. Similar to our study LED-FM performed better than ZN staining when using unprocessed samples, with a lower sensitivity of LED-FM in HIV-infected patients compared to HIV-uninfected patients. To our knowledge our study is the second to evaluate the performance of LED-FM in HIV-infected patients.
The lower specificity of LED-FM compared with ZN in our study may have several explanations. First, as a gold standard we used solid culture, which may be false-negative, especially among HIV-infected patients with paucibacillary TB. The probability of obtaining a positive culture is related to the number of AFB in the specimen, with only about 50% of cultures of specimens with 1–2 AFB per-100 fields being identified as positive, increasing to 80% and 96,7% for specimens with scanty (1–9 AFB/100 fields) and “positive” AFB grades, respectively 
. This suggests that at least some LED-FM positive, culture-negative patients may have incorrectly been classified as false-positive LED-FM. Of course, false positive LED-FM results may also have been due to technical errors and lack of experience or training of technicians. Technicians in this study were experienced, but the technique and LED-microscope had not been used prior to the study.
Most microscopic examinations for pulmonary TB are performed directly on smeared and stained preparations of unprocessed sputum 
. Various methods of concentrating sputum based on centrifugation have been shown to increase diagnostic yield when used prior to microscopy 
. Our results showed that decontamination and concentration of sputum may improve the diagnostic accuracy of both ZN-microscopy and LED-FM. This is a feasible option if culture is performed routinely for all samples received in a laboratory. However, this is not the case in our hospital, like in many other settings in Indonesia and elsewhere. As such, it seems unsuitable to be implemented in high-burden laboratories.
In addition to higher sensitivity, shorter reading time and the simplicity of the Auramin-O staining method also indicate the potential benefit of making LED-FM more widely implemented in settings with a high burden of TB. However, the appropiate and feasible external quality assurance (EQA) procedures are needed for TB laboratories planning to implement LED-FM because the fluorochrome-based stained fade over time 
WHO has recently endorsed the implementation of GeneXpert MTB/RIF assay for national TB programmes in developing countries as the first initial screening of TB in HIV 
. To our knowledge, there is only one study that comprehensively compared the implementation of LED-FM with GeneXpert MTB/RIF assay. The results showed that the Xpert MTB/RIF assay outperformed LED microscopy in all type of specimens, and that all sputum smears reported as “scanty AFB” had a positive Xpert result. However, it is unlikely in the short term, that Xpert can be implemented widely to replace smear microscopy as the initial diagnostic test in Indonesia like in many other low-resource or high-volume settings. LED fluorescent microscopy may therefore be used as a reasonable and cost-effective alternative 
Our study has several limitations. First, our microscopist had experience with ZN microscopy and never used fluorochrome-based technique before the study was conducted, but we observed the increase of technical skill after training until they were confident to start this study. Second, we used solid culture as the reference standard, knowing that solid culture may be false-negative, especially in HIV-infected patients with paucibacillary disease. Specificity of LED-FM might in fact have been higher if liquid culture was used as a reference standard.
In conclusion, our study showed that LED-FM is more sensitive and more quick than conventional ZN-microscopy for TB diagnosis, but that a higher threshold for positivity than the current WHO-threshold may improve its specificity. This makes a suitable method and reasonable alternative to more sensitive but much more expensive molecular tests, especially in low-resource settings. However, implementation of LED-FM must be followed by adequate training, standardized protocols for staining and smear examination, and sustainable EQA systems.