Use of simulated sputum specimens to estimate the specificity of laboratory-diagnosed tuberculosis

Abstract

Setting: Cross-contamination is not uncommon in mycobacteriology laboratories of high-income countries, as documented by bacterial genotyping. The extent of this problem in low-income countries is largely unknown, where this method is impractical.

Objective: To estimate the rate of cross-contamination in a high-volume tuberculosis (TB) laboratory in South Africa.

Design: Simulated sputum specimens labelled with false names were sent from a TB clinic, interspersed with patient samples, and processed for culture and microscopy. Results were interpreted in the context of the observed proportion of samples with positive microscopy and culture results.

Results: With microscopy, 6/190 (3.2%) simulated specimens were positive (estimated specificity = 96.8%). Considering the 881 positive microscopy results in 6093 clinical samples, we extrapolate that 19.3% (95%CI 7.0-42.8) of positive smears were false-positives. On culture, 2/190 (1.1%) of the simulated specimens were positive for Mycobacterium tuberculosis (estimated specificity = 98.9%). Considering the 1862 positive cultures from 6093 clinical samples, we estimate that 2.4% (95%CI 0.3-8.8) of positive cultures were false-positives.

Conclusion: Simulated specimens offer a simple means of estimating the proportion of false-positive results, providing information on all sources of potential error from the clinic, through the laboratory and to reporting of results.