Study design
Figure 1 shows the study design that involved 72 sputum specimens. First, bleach-sterilisation studies determined the bleach exposure required to increase laboratory safety by sterilising sputum. In the light of these results, 5 bleach-sedimentation protocols that would completely sterilise sputum were compared in pilot experiments. The best performing of these techniques was then assessed in detail. Sample size calculations were not performed because the concentration of acid-fast bacilli in specimens for the planned protocol was unknown. All experimentation was performed blinded to the results of all other tests, at room temperature, and all slides were read in random order.
Setting
The study was carried out over a 6-month period in shantytowns in Lima, Peru in which tuberculosis principally affects socioeconomically disadvantaged people [20]. Peru is a middle-income country with high tuberculosis incidence in which conventional Ziehl-Neelsen sputum smear-microscopy is the principal diagnostic test for tuberculosis.
Specimens
In collaboration with the national tuberculosis control program, sputum specimens were obtained on the day that they had been found by local laboratories to be microscopy-positive for acid-fast bacilli. All specimens were from untreated, newly diagnosed patients being investigated for clinically suspected tuberculosis. We recorded the volume and consistency (whether salivary or mucoid) of each specimen and whether the microscopist was moderately or very experienced.
(1) Bleach sterilisation assessment
Assessment of bleach-sterilisation utilised 31 sputum specimens that were homogenised and decontaminated with the sodium hydroxide N-acetyl cysteine method [3]. Briefly, a freshly prepared solution of 4% sodium hydroxide, 2.9% sodium citrate and 0.5% N-acetyl cysteine (Sigma, Saint Louis, Missouri) was mixed with an equal volume of sputum and left for 15 minutes. The decontamination was then stopped by adding a 7-times excess volume of phosphate-buffered saline (PBS, pH 6.8), centrifuging at 3, 000 × g for 20 minutes and discarding the supernatant. The addition of a 7-times excess volume of PBS and the centrifugation conditions are standard practices for centrifuge-decontamination in some laboratories in Peru because these conditions were found in pilot experiments to provide optimal neutralisation and concentration (data not shown). The pellet from centrifugation was re-suspended in 34 ml PBS and then split into 17 aliquots that were each 2 ml in volume. One aliquot was used as a control to which no bleach was added and 2 ml of 3%, 6%, 10% and 15% bleach were each added to quadruplet sets of each of the other aliquots. The bleach dilutions were prepared fresh from commercially available 15% bleach (sodium hypochlorite, NaOCl; Import Export Lider, Lima, Peru). Each of the bleach-sputum mixtures was treated with bleach for 1, 5, 10 or 20 minutes. After this exposure to a total of 16 combinations of bleach concentrations and exposure times, reactions were stopped by adding a 7-times excess volume of PBS and shaking by hand until homogenised. The solution was then centrifuged at 3,000 × g for 15 minutes, the supernatant was discarded and the pellet re-suspended in 0.2% bovine serum albumin (Sigma, Saint Louis, Missouri). The entire re-suspended pellet was then spread on a Middlebrook 7H11 agar plate (Difco, Detroit, Michigan) supplemented with 10% oleic acid, albumin, dextrose and catalase as described [3]. The plate was sealed in a Ziploc® bag (Johnson, Wisconsin) to prevent drying, incubated at 37°C in air and inspected for M. tuberculosis growth using an inverted microscope twice weekly for 8 weeks.
(2) Bleach-sedimentation assessment
Pilot work for selection of bleach-sedimentation technique
In pilot experiments, 5 published bleach-sedimentation protocols [14–18] were compared to select an optimal technique for further assessment. Triplicate conventional smears were prepared from 16 sputum specimens. The remainder of each specimen was then processed by these 5 protocols after which triplicate slides were produced from each bleach-sedimented specimen. Bleach-sedimentation without subsequent water dilution followed by sedimentation for 30-45 minutes [18] or 12-15 hours [17] considerably reduced the number of acid-fast bacilli visible on microscopy, possibly through bleach damaging M. tuberculosis (data not shown), so these techniques were not further assessed. The other techniques involved adding bleach to the sputum without shaking [14], shaking at regular intervals for 15 minutes [16] or continuous shaking for 10 minutes [15] before dilution with water followed by sedimentation. These 3 techniques had similar effects on the numbers of acid-fast bacilli visible on microscopy (data not shown). The last of these 3 techniques [15] had the most precisely defined methodology and was reported to have produced optimal results so was selected for the further evaluation described below.
Quantitative evaluation of bleach-sedimentation
The bleach-sedimentation method described by Gebre-Selassie [15] selected in our pilot work was further assessed as follows. Triplicate conventional smears were prepared from each of 25 sputum specimens to serve as controls. The remaining volume of each specimen up to a maximum volume of 1.5 ml was then placed into 15 ml polypropylene tubes (Falcon BD, San Jose, California). Fresh 5% bleach was prepared by dilution from a solution of 8% bleach that the manufacturer reported contained 8.09 g/100 ml free chlorine ions and had 1.125 g/ml density. One volume of 5% bleach (equal to the sputum volume) was added to each specimen and the mixture was shaken by hand continuously for 10 minutes. Eight volumes of distilled water were then added and the mixture was left to sediment for 16 hours. The supernatant was then pipetted off and the pellet, or the basal approximately 250 μl if no pellet had formed, was mixed by pipetting and used to prepare triplicate smears.
Smear preparation
In order to standardise the amount of sputum applied to all slides, a pipette was used to apply to each slide 40 μl of unprocessed or bleach-sedimented sputum that was smeared over a single area of approximately 1 cm × 2 cm. Forty μl was used because this volume was equivalent to 1 drop of sputum. Slides were air-dried, heat-fixed by passing over a flame and stained using the Ziehl-Neelsen method. Briefly, the smear was flooded with 0.3% carbol fuchsin, heated with a flame, left to stand for 10 minutes and washed with water. Acid-alcohol was applied for 2 minutes, the slide was washed with water and the counter-stain methylene blue was applied for 1 minute, washed off and slides left to dry vertically [3].
Smear-microscopy sensitivity
The number of acid-fast bacilli was counted in 100 high-power fields that were read per slide using standard oil-immersion light microscopy. If < 32 acid-fast bacilli were visible in 100 fields then an additional 200 fields were read. This cut-off was derived because it is the mid-point between 10 and 100 on a logarithmic scale and was selected arbitrarily to increase the precision of quantification of relatively low concentrations of acid-fast bacilli.
Smear-microscopy speed
Microscopists recorded the time they spent counting the acid-fast bacilli per 100 fields on each slide. To improve the blinded nature of the comparison, they were unaware of our research hypotheses and recorded the time spent on each slide as a laboratory routine.
Smear-microscopy reliability
Both microscopists cross read a random sample of 1 in 8 slides to determine the degree of agreement between their readings using the same protocol as the first slide reading.
Ethical considerations
Ethical committee approval was not required because this research did not involve human subjects or patient diagnosis and utilised anonymized, unlinked excess specimens that would otherwise have been discarded.
Statistical analysis
Data were analysed using SPSS 11.5 software (SPSS Corp., Chicago, Illinois). Acid-fast bacilli count data were non-Gaussian and were summarised using geometric means with 95% confidence intervals (95%CI) and were compared with the non-parametric Wilcoxon signed-rank test (for paired data) and the rank sum test (for unpaired data). Slide-reading time data were Gaussian and were summarised using arithmetic means with standard errors of the mean (SEM) that were compared using the paired Student's t-test. Correlations were assessed with the non-parametric Spearman's rank correlation coefficient (r). All p-values corresponded to 2-sided hypothesis testing.