Sputum smear-microscopy is the most widely used laboratory test for diagnosing tuberculosis but in poorly equipped settings can expose laboratory staff to the infectious pathogen Mycobacterium tuberculosis . Consequently, the risk of tuberculosis disease has been found to be 7-79 times greater in laboratory staff than the general population . Bleach is bactericidal and adding bleach to sputum may sterilise it, potentially protecting staff from tuberculosis infection during processing although this would also prevent subsequent culture based testing. However, the sterilising activity of bleach is poorly characterised for M. tuberculosis and the bleach concentrations and exposure times required during bleach-sedimentation to sterilise sputum and prevent biohazard to staff are unknown [3–6].
Smear-microscopy fails to diagnose patients who have low concentrations of M. tuberculosis in their sputum, hampering tuberculosis control. Conventional smear-microscopy involves smearing sputum on a microscope slide that is then stained and examined by high power microscopy to detect the causative acid-fast bacillus M, tuberculosis. For a 50% probability of finding a single acid-fast bacillus in 100 microscopy fields, approximately 5, 000 acid-fast bacilli must be present per ml of sputum . Consequently the sensitivity of this technique is typically only 30-70% of the sensitivity of culture [8, 9]. Tuberculosis patients who have AIDS and/or are children usually have lower concentrations of M, tuberculosis bacilli in their sputum, so the diagnostic sensitivity of smear-microscopy is lower in these patients [10–12]. Thus, reliance on smear-microscopy may cause missed or delayed tuberculosis diagnosis, potentially increasing morbidity, mortality and tuberculosis transmission. Increasing the sensitivity of tuberculosis diagnostic testing is a public health priority.
Diagnostic sensitivity increases if acid-fast bacilli are concentrated into the small volume that can be visualised by microscopy. Bleach-sedimentation has been hypothesised to concentrate acid-fast bacilli in sputum specimens and in support of this hypothesis a recent meta-analysis reported that bleach-sedimentation caused a 9% increase in tuberculosis diagnostic sensitivity compared to conventional smear-microscopy . Centrifugation concentrates M. tuberculosis and is used in some bleach-sedimentation protocols but centrifuges are expensive, may create biohazardous aerosols and are infrequently available in resource-poor settings. We therefore restricted our research to gravity bleach-sedimentation techniques that do not involve centrifugation [14–18].
Most studies of bleach-sedimentation reported that it slightly increased diagnostic sensitivity of smear-microscopy [6, 19]. Variations between these studies may be explained by failure to record the number of microscopy fields examined and/or time spent performing microscopy and by difficulty making blinded comparisons because bleach-sedimentation changes the appearance of sputum smears . There were also differences in protocol: 5 published bleach-sedimentation techniques share a common initial step of mixing sputum with an equal volume of 5% bleach, which is then either stained without further dilution [17, 18] or after dilution in water [14–16]. Dilution in water after adding bleach may reduce bleach-mediated damage to M. tuberculosis that can inhibit subsequent acid-fast staining . All bleach-sedimentation techniques involve some dilution of sputum and it is unknown whether they cause overall concentration or dilution of visible acid-fast bacilli [6, 19].
Most microscopy studies have compared either rates of microscopy positivity or alternatively the numbers of slides in each categorical microscopy grade (negative, weakly positive '+', positive '++', or strongly positive '+++'; see figure legends for definitions). These approaches are clinically relevant but are insensitive for assessing bleach-sedimentation because few specimens contain concentrations of acid-fast bacilli close to the threshold between microscopy grades. Consequently, when this categorical approach is used large numbers of specimens must be studied and small effects of bleach-sedimentation may be missed. The use of a more precise assessment of acid-fast bacilli concentration such as the number visible per 100 high-powered microscopy fields should facilitate characterisation of bleach-sedimentation effects.
Bleach-sedimentation lyses human cells within sputum, which clears the field of view during microscopy and may accelerate slide reading speed but these effects do not appear to have been quantified  and confound assessment of acid-fast bacilli concentrations. Consequently, it is unclear from published research whether bleach-sedimentation increases the concentration of visible acid-fast bacilli, increases the amount of sputum examined in the available time, neither or both of these effects. To overcome these limitations we developed a protocol using triplicate slides from each specimen before and after bleach-sedimentation to characterise effects on smear-microscopy for each specimen.
We used these methodological refinements to characterise the effect of bleach-sedimentation on the safety, sensitivity, speed and reliability of smear-microscopy. This novel methodology clarifies the specific effects of bleach-sedimentation and provides an explanation for the discrepant results from previous studies.