Introduction: Light-emitting diode fluorescence microscopy (LED-FM) has been endorsed by the World Health Organization (WHO) for tuberculosis diagnosis, but its accuracy in HIV-infected patients remains controversial, and only some few studies have explored procedural factors that may affect its performance. Objective: To evaluate the performance of LED-FM for tuberculosis diagnosis in patients with and without HIV infection using a newer, less expensive LED lamp. Materials and methods:We compared the performance of LED-FM and Ziehl-Neelsen (ZN) microscopy on respiratory specimen smears from tuberculosis (TB) suspects and patients on treatment examined by different technicians blinded for HIV-status and for the result of the comparative test. We analyzed the effect of concentrating specimens prior to microscopy using different examination schemes and user-appraisal of the LED device. Results: Of the 6,968 diagnostic specimens collected, 869 (12.5%) had positive Mycobacterium tuberculosis cultures. LED-FM was 11.4% more sensitive than ZN (p<0.01). Among HIV-positive TB patients, sensitivity differences between LED-FM and ZN (20.6%) doubled the figure obtained in HIVnegative patients or in those with unknown HIV status (9.3%). After stratifying by direct and concentrated slides, the superiority of LED-FM remained. High specificity values were obtained both with LED-FM (99.9%) and ZN (99.9%).The second reading of a sample of slides showed a significantly higher positive detection yield using 200x magnification (49.4 %) than 400x magnification (33.8%) (p<0.05). The LEDdevice had a very good acceptance among the technicians. Conclusion: LED-FM better performance compared with ZN in HIV-infected patients and user-appraisal support the rapid roll-out of LED-FM. Screening at 200x magnification was essential to achieve LED-FM increased sensitivity.

No abstract

Moderate training (a three-day course) is not enough to make technicians proficient in LED-FM; however, great ability can be reached after a short "in situ" training phase even without the presence of experienced staff available in field to review doubtful results. Training was more effective in services with a minimum workload of 750 slides/year.