SETTING: Urban health clinic, Nairobi. OBJECTIVE: To evaluate the impact on tuberculosis (TB) case detection and laboratory workload of reducing the number of sputum smears examined and thresholds for diagnosing positive smears and positive cases. DESIGN: In this prospective study, three Ziehl-Neelsen stained sputum smears from consecutive pulmonary TB suspects were examined blind. The standard approach (A), > or = 2 positive smears out of 3, using a cut-off of 10 acid-fast bacilli (AFB)/100 high-power fields (HPF), was compared with approaches B, > or = 2 positive smears (> or = 4 AFB/100 HPF) out of 3, one of which is > or = 10 AFB/100 HPF; C, > or = 2 positive smears (> or = 4 AFB/100 HPF) out of 3; D, > or = 1 positive smear (> or = 10 AFB/100 HPF) out of 2; and E, > or = 1 positive smear (> or = 4 AFB/100 HPF) out of 2. The microscopy gold standard was detection of at least one positive smear (> or = 4 AFB/100 HPF) out of 3. RESULTS: Among 644 TB suspects, the alternative approaches detected from 114 (17.7%) (approach B) to 123 cases (19.1%) (approach E) compared to 105 cases (16.3%) for approach A (P < 0.005). Sensitivity ranged between 82.0% (105/128) for A and 96.1% (123/128) for E. The single positive smear approaches reduced the number of smears by 36% compared to approach A. CONCLUSION: Reducing the number of specimens and the positivity threshold to define a positive case increased the sensitivity of microscopy and reduced laboratory workload.