Abstract
BACKGROUND: Fluorescence microscopy (FM) has not been implemented widely in TB endemic settings and little evaluation has been done in HIV-infected patients. We evaluated diagnostic performance, time and costs of FM with light-emitting diodes technology (LED-FM), compared with conventional (Zieh-Neelsen) microscopy in a hospital in Indonesia which acts as referral centre for HIV-infected patients. METHOD: We included pulmonary tuberculosis suspects from the outpatient and HIV clinic. Direct and concentrated sputum smears were examined using LED-FM and ZN microscopy by two technicians who were blinded for the HIV-status and the result of the comparative test. Mean reading time per slide was recorded and cost of each slide was calculated. Mycobacteria culture served as the reference standard. RESULTS: Among 404 tuberculosis suspects from the outpatient clinic and 256 from the HIV clinic, mycobacteria culture was positive in 12.6% and 27%, respectively. The optimal sensitivity of LED-FM was achieved by using a threshold of ≥2 AFB/length. LED-FM had a higher sensitivity (75.5% vs. 54.9%, P<0.01) but lower specificity (90.0% vs 96.6%, P<0.01) compared to ZN microscopy. HIV was associated with a lower sensitivity but similar specificity. The average reading time using LED-FM was significantly shorter (2.23±0.78 vs 5.82±1.60 minutes, P<0.01), while costs per slide were similar. CONCLUSION: High sensitivity of LED-FM combined with shorter reading time of sputum smear slides make this method a potential alternative to ZN microscopy. Additional data on specificity are needed for effective implementation of this technique in high burden TB laboratories.