Setting: Karonga district, northern Malawi.
Objective: To compare the sensitivity and specificity of two versus three smears for the diagnosis of pulmonary tuberculosis in a setting with high HIV prevalence.
Design: A total of 1992 pulmonary tuberculosis suspects with three sputum smears taken over a 2-7 day period and at least one culture result were studied. Smears were auramine stained and examined using fluorescence microscopy, and positives were confirmed with Ziehl-Neelsen staining and light microscopy. Cultures were set up on Löwenstein-Jensen media. True negative and positive status was defined on the basis of culture. The sensitivity, specificity, and positive and negative predictive values of two and three smears were compared.
Results: Compared to culture, the sensitivity, specificity, and positive and negative predictive values of three smears were 70%, 98%, 92%, and 92%, respectively. Restriction to the first two smears gave similar results. Of those detected as smear-positive using three smears, at least 97% would have been detected by two. Among those with HIV serology results available, the sensitivity of two smears for detecting culture-positive tuberculosis was identical to that using three.
Conclusion: In this setting, using fluorescence and light microscopy, collecting two smears rather than three would only marginally reduce sensitivity and would slightly improve the specificity of diagnosis of tuberculosis; this is unaffected by HIV status. The potential for improving specificity is important because of the costs of misdiagnosis. In practice, both sensitivity and specificity may be increased due to the time saved by examining two rather than three smears.