Optimal tuberculosis case detection by direct sputum smear microscopy: how much better is more?

SETTING: A tuberculosis control project in Bangladesh. OBJECTIVE: To define the efficiency of numbers of microscopic fields screened and the sputum collection scheme used for diagnostic smear examination. DESIGN: Quality controllers noted cumulative numbers of acid-fast bacilli per 100 fields screened. The incremental diagnostic yield of different sputum sampling strategies was determined. Doubtful series were re-checked and/or further samples examined. RESULTS: Acid-fast bacilli were found in 99.6% of 1412 positive and in 79.3% of 576 scanty slides in the first 100 fields. Examination of a third specimen yielded a maximum of 2.7% positives incrementally. The most efficient strategy, using three morning specimens, yielded 94.2% positives on the first and 1.0% on the third sputum; although 10% of suspects did not return, only 1.5% of the positives were among them and more cases were confirmed and treated. The positive predictive value of a single positive or scanty smear was very high (99.2%). CONCLUSIONS: Reading more than 100 fields per smear or examining a third sputum has insufficient marginal returns to justify the workload. Examining morning samples only is more efficient, and their collection does not necessarily inconvenience patients. Treatment can be started on the basis of one positive smear. Provided that a well functioning system of smear-microscopy quality control is in place, we propose a strategy based on examination of two morning sputum samples for negative suspects, with the diagnosis based on a single positive result.