SETTING: Three busy laboratories in Nairobi, Kenya. OBJECTIVES: To determine the performance of an affordable fluorescence system (FluoreslenS) for tuberculosis microscopy, and to test an auramine-smear rechecking system. DESIGN: Alternating routine use of Ziehl-Neelsen (ZN) and fluorescence microscopy (FM) was performed to compare detection and errors found while rechecking. RESULTS: Overall, 19.5% of 25,250 ZN and 23% of 21,104 FM smears were positive (P < 10(-3)). The proportional increment of FM over ZN was 18% (range -6%-29%), with one centre detecting fewer positives (non-significant, NS). The average error frequencies were comparable (1.8% vs. 2.6% false-negative and 0.2% vs. 0.4% high false-positive for ZN and FM, respectively, NS). The superior performance of controllers and the overall equal ZN/FM quality in the laboratories could be demonstrated only after converting error percentages to relative sensitivity (RS). CONCLUSIONS: FM with the FluoreslenS system considerably improved sensitivity without loss of specificity in two of the busy routine laboratories, but the system is not sufficiently practical or user-friendly. Rechecking by FM can be done using guidelines for ZN smears, provided that routine ZN confirmation of positives is omitted. Calculation of RS allows an objective comparison of microscopy quality, independent of the variable prevalence of positives and sample size.
SETTING: Three busy laboratories in Nairobi, Kenya. OBJECTIVES: To determine the performance of an affordable fluorescence system (FluoreslenS) for tuberculosis microscopy, and to test an auramine-smear rechecking system. DESIGN: Alternating routine use of Ziehl-Neelsen (ZN) and fluorescence microscopy (FM) was performed to compare detection and errors found while rechecking. RESULTS: Overall, 19.5% of 25,250 ZN and 23% of 21,104 FM smears were positive (P < 10(-3)). The proportional increment of FM over ZN was 18% (range -6%-29%), with one centre detecting fewer positives (non-significant, NS). The average error frequencies were comparable (1.8% vs. 2.6% false-negative and 0.2% vs. 0.4% high false-positive for ZN and FM, respectively, NS). The superior performance of controllers and the overall equal ZN/FM quality in the laboratories could be demonstrated only after converting error percentages to relative sensitivity (RS). CONCLUSIONS: FM with the FluoreslenS system considerably improved sensitivity without loss of specificity in two of the busy routine laboratories, but the system is not sufficiently practical or user-friendly. Rechecking by FM can be done using guidelines for ZN smears, provided that routine ZN confirmation of positives is omitted. Calculation of RS allows an objective comparison of microscopy quality, independent of the variable prevalence of positives and sample size.