Methods

Setting and patients. Participants were drawn from the urban outpatient clinic supported by the Kenyan Ministry of Health and Médecins Sans Frontières (Doctors Without Borders), which provides HIV and tuberculosis treatment to the community living in the slum of Mathare (Nairobi, Kenya). In 2001, the HIV prevalence in the adult population of Nairobi was 15%, and 50% of patients with tuberculosis were infected with HIV [13]. In Mathare, 76% of patients with smear-negative suspected tuberculosis were HIV infected (M. Bonnet, personal communication). Consecutive patients aged >15 years who presented with a cough for >2 weeks were eligible for the study [14]. Intake of antituberculosis drugs or quinolone in the 4 weeks before the screening was an exclusion criterion. Demographic information, treatment history, and clinical characteristics were recorded.

Specimen collection and processing. Patients submitted 3 sputum specimens during 2 consecutive days, after they were given instructions on how to produce a good quality specimen [15]. The first specimen was collected in the clinic at the initial consultation, the second at home early the next day, and the third in the clinic when the patient brought the morning specimen [16]. A minimum quantity of 1 mL of specimen was required. Smears were made from each specimen for direct smear microscopy. Each smear was heat fixed and was stained using the hot Ziehl-Neelsen method (1% filtered carbol-fuchsin and 0.1% methylene blue). The remainder of the specimen was transferred to a 15-mL disposable plastic conical tube with an equal volume of neat commercial bleach (3.5% NaOCl). The mixture was agitated using a vortex mixer and was placed vertically on the bench, away from drafts, at room temperature for overnight sedimentation (15–18 h). After sedimentation, the supernatant was poured off, and the sediment was mixed with the remaining fluid. One or 2 drops were transferred with a sterile glass pipette to a slide. A bleach smear was made, was air dried, was heat fixed, and was stained by the same Ziehl-Neelsen method.

The direct smear and bleach smear specimens were examined by bright-field microscopy (magnification, ×1000) on site by the same laboratory technician who was blind to the result of previous smears from the same patient. Staff turnover of the 2 study technicians prevented the same technician from reading all smears from the same patient. The exact number of AFB observed in 100 high-power microscopic fields (HPF) of each smear was recorded on unique, separate laboratory forms and in the laboratory register by the laboratory supervisor. For both direct smear and bleach sedimentation, a positive smear was defined using 2 different grading scales: the World Health Organization-International Union Against Tuberculosis and Lung Disease scale, requiring ⩾10 AFB/100 HPF, and the scale recommended by the American Thoracic Society, requiring ⩾1 AFB/100 HPF [16–18]. The microscopic appearance of a direct smear was defined as “good” if blue cellular elements were present without debris, as “too thick” if cells were lying on top of each other, as “too thin” in cases of insufficient background elements, as “under decolorized” if background was purple or red, and as “overheated” if red crystals were seen. Similar assessment of bleach smears was not conducted because of the digestion of the cellular elements in the smears.

As part of internal quality management, 100% of positive and 10% of negative bleach smears and direct smears were blindly reexamined monthly by the laboratory supervisor. External quality assessment consisted of 100 randomly selected direct smears blindly reexamined at the end of the study by the tuberculosis laboratory of Kenya Medical Research Institute (KEMRI) (Nairobi, Kenya).

NaOCl solution. The 3.5% domestic bleach available on the Kenyan market (Jik Bleach Regular; Reckitt Benckiser East Africa; active ingredient, sodium hypochlorite at 3.5% mol/vol when packed) was used without the addition of distilled water [19]. The same batch of bleach was used during the entire 10-month study duration. To reduce the risk of oxidation, the solution was stored in 750-mL dark bottles, with minimum head-space in a dark area [20]. The room temperature was monitored. Each week, a 250-mL aliquot of bleach “working solution” was collected to process the specimens for that week. All remaining working solution was discarded at the end of the week. The date and hour of aliquoting of the working solution were recorded on the weekly container. Before use, the presence of free chlorine was assessed using a swimming pool tester with diethyl-p-phenylene diamine 1 (pocket type; chlorine range, 0.1–6.0 mg/L) after prior dilution at d=2.10⁴. The quality of NaOCl was considered adequate if the concentration was in the range 1.5–2 mg/L, in accordance with manufacturer's recommendations.

Reproducibility assessment. Random selections of direct smears and bleach smears were read a second time on the same day by a different technician and were read by the same technician 24 h after the first reading, to assess inter- and intrareader reproducibility, respectively. The laboratory supervisor masked the identification of the slide to ensure that the reading was blind.

Operational aspects. A list of parameters was measured to evaluate the feasibility of bleach sedimentation microscopy compared with that of direct smear microscopy: (1) the duration of bleaching (i.e., the time spent adding the bleach to the specimen in the tube and shaking the tube), (2) sedimentation, (3) the smearing (including the drying of the slide) and staining of a daily batch of specimens, and (4) the duration of the reading of individual slides. The temperature at the start and end of sedimentation was monitored. The cost of reagents and consumables to perform bleach sedimentation microscopy was assessed on the basis of Kenyan market price.

Sample size. The minimum sample size to detect at least a 10% difference in the percentage of smear-positive cases found by bleach sedimentation microscopy compared with direct smear microscopy was calculated [8]. With an average of 20% smear-positive cases detected routinely by direct smear microscopy in the study clinic, a risk of 0.05, a power of 80%, and a 10% dropout rate, the sample size was 690 patients. With a positive-smear detection rate of 20%, an expected κ coefficient >0.8, and a precision of 10%, the minimum sample size required to assess the test reproducibility was 220 smears [21].

Data analysis. Data were double entered using Epidata 3.1 (EpiData Association) and were analyzed using SPSS 11.0 for Windows (SPSS). Patient, specimen, and smear characteristics were described. Positive smear specimens and patient detection rates were compared between bleach sedimentation microscopy and direct smear microscopy, with use of McNemar's test for matched data. Four definitions of a sputum smear-positive case were used:

1. AFB on 2 of 3 smears examined, 1 of which had ⩾10 AFB/100 HPF detected [22].

2. ⩾1 AFB/100 HPF in 2 of 3 smears [14].

3. ⩾10 AFB/100 HPF in 1 of 3 smears, and

4. ⩾1 AFB/100 HPF in 1 of 3 smears.

The gain of bleach smears was calculated as the increase in positive smears by bleach sedimentation microscopy, compared with direct smear microscopy, divided by the total number of positive smears detected by direct smear microscopy. Likewise, the gain of direct smears was calculated as the increase in positive smears by direct smear microscopy, compared with bleach sedimentation microscopy, divided by the total number of positive smears detected by bleach sedimentation microscopy. The same calculation was made for smear-positive cases. The positive-smear—detection yield of bleach sedimentation microscopy performed on the first specimen was compared with that of direct smear microscopy performed on the first 2 specimens and on all 3 specimens, with use of case definitions 3 and 4 described above. Inter- and intrareader reproducibility were assessed by the calculation of the κ coefficient, which measures the agreement between 2 readings. A κ coefficient ⩾0.80 signifies almost perfect agreement.

The National Ethical Review Committee of KEMRI and the Comité de Protection des Personnes (Saint Germain en Laye, France) approved the study. Written informed consent was obtained from the participants.

Results

A total of 788 patients with suspected pulmonary tuberculosis were screened, and 644 (81.7%) were recruited for the study (figure 1). The mean age was 32 years (SD, 10.3), the male∶female ratio was 0.8 (287∶356; data was missing for 1 patient), 121 (18.8%) of the patients had a history of tuberculosis, 37 (5.7%) received a course of broad-spectrum antibiotics in the 2 weeks before recruitment, and 12 (1.9%) had a chest radiograph. All patients had sputum production, 593 (92.1%) had a fever in the past week, 596 (92.5%) had chest pain, 109 (16.9%) had hemoptysis, 550 (85.4%) had night sweats, 245 (38.0%) had weight loss, and 484 (75.2%) had appetite loss.

Among the 644 recruited patients, 614 (95.3%) provided 3 specimens, 7 (1.1%) provided 2 specimens, and 23 (3.6%) provided 1 specimen. A total of 1879 specimens were collected and processed. Of these specimens, 1401 (74.6%) were purulent, 414 (22.0%) were mucoid, 56 (3.0%) were blood stained, and 8 (0.4%) were salivary. Of 1879 direct smears, 1855 (98.7%) were good, 5 (0.3%) were too thick, and 19 (1.0%) were too thin.

The results of smear microscopy are presented in tables 1 and 2. In 36 (1.9%) of the 1879 specimens, bleach smears were unreadable. Regardless of the threshold used to define a positive smear, bleach sedimentation microscopy yielded significantly more positive smears than did direct smear microscopy. With the threshold of 10 AFB/100 HPF or 1 AFB/100 HPF, 62 (20.6%) of 301 positive smears and 86 (23.2%) of 371 positive smears, respectively, were detected by bleach sedimentation, compared with 10 (2.8%) of 363 positive smears and 2 (0.4%) of 460 positive smears, respectively, detected by direct smear microscopy. Significantly fewer positive smears were missed by bleach sedimentation microscopy with the threshold of 1 AFB/100 HPF than with the threshold of 10 AFB/100 HPF (P<.01). Of 363 positive smears (with the threshold of 10 AFB/100 HPF) detected by bleach sedimentation microscopy, 289 (79.6%) were detected as positive by direct smear microscopy. Of the remaining 74 positive smears detected only by bleach sedimentation microscopy, 51 (68.9%) showed scanty results (1–9 AFB/100 HPF) and 23 (31.1%) were negative by direct smear microscopy. The benefit of the bleach sedimentation method was greater for the mucoid specimens than for the purulent specimens.

The results of smear-positive case detection are presented in table 3. Regardless of case definition, bleach sedimentation microscopy detected significantly more smear-positive cases than did direct smear microscopy. Compared with direct smear microscopy performed on 2 or 3 specimens, bleach sedimentation microscopy performed on the first specimen detected as many smear-positive cases with case definition 3 and detected significantly more smear-positive cases with case definition 4 (table 4). Both direct smear microscopy and bleach sedimentation microscopy performed on the first specimen detected only 2 more cases than did bleach sedimentation microscopy alone.

Inter- and intrareader reproducibility of both methods showed κ coefficients ⩾0.8 (table 5). Quality control reported a 95%–100% agreement rate between the technician's results and the supervisor's results for the monthly internal control and a 99% rate for the external control.

Considering the operational aspects, a median of 12 specimens (interquartile range [IQR], 8.5–15) were processed daily. The mean ± SD duration of bleaching was 18.6 ± 7.6 min, and the mean ± SD duration of sedimentation was 16.8 ± 0.8 h. The median temperature at the start (late afternoon) and the end (early morning) of sedimentation was 30.0°C (IQR, 27.2–32.7) and 26.0°C (IQR, 23.7–28.0), respectively. Monitoring of the presence of free chlorine in the bleach solution showed that levels were adequate, with NaOCl concentrations in the range of 1.5–2 mg/L at weekly controls. The mean ± SD duration of smear preparation was longer for processed specimens (52.9 ± 25.6 min) than for fresh specimens (21.4 ± 8.3 min; P<.001), because of the longer drying time for slides made from processed specimens. The mean ± SD duration of staining was similar for both methods: 45 ± 10.4 min for bleach sedimentation microscopy and 47.1 ± 10.6 min for direct smear microscopy (P=.12). No statistically significant difference was observed in the mean ± SD duration for reading a positive smear by bleach sedimentation microscopy (3.1 ± 0.6 min) compared with by direct smear microscopy (3.0 ± 0.6 min). The additional cost of specimen processing was €0.20/specimen, which included the cost of bleach and of disposable products (2-mL syringes, 15-mL plastic conical tubes, and plastic pipettes).

Discussion

We report a significant increase in the number of positive smears and the number of affected patients detected using bleach sedimentation microscopy compared with conventional direct smear microscopy, regardless of the smear-positive case definition and AFB threshold used. One digested smear performed as well as 3 direct smears, which is consistent with recent results from a hospital-based study in Nigeria with a similar prevalence of concordant HIV infection and tuberculosis [23].

This study avoided the limitations associated with previous studies and reported in a meta-analysis of sputum-processing methods [3, 10]. First, it was a prospective, community-based study in a setting of high prevalence of HIV. The criteria to define a suspected case of tuberculosis were based on the World Health Organization guidelines for smear-microscopy screening in settings of high HIV prevalence [14]. The rate of smear-positive cases among suspected cases of tuberculosis is in the expected range of 5%–20% for countries in which tuberculosis is one of the most frequent causes of chronic cough [16]. Second, the quality of the smear microscopy was high, which was the condition in which the lowest benefit of the specimen-processing method would be expected [24]. Indeed, we report a significant benefit associated with the use of bleach sedimentation microscopy in a setting of care that already uses optimized direct smear microscopy and collects high-quality specimens. Third, the results were reported per specimen and per patient, by use of different standards for the AFB threshold and case definition. Fourth, our bleach method was standardized as much as possible, and different practical aspects were monitored. Unlike methods used in previous studies, the processing method did not require distilled water, which simplified the procedure and presumably avoided contamination of smears with saprophytic AFB from tap water, which is used routinely in conditions in which distilled water is not available [24]. Finally, the study was strengthened by its methodology based on recruitment of consecutive individuals with suspected tuberculosis and a blind evaluation.

The bleach sedimentation method is simple and does not require additional expertise beyond that required for conventional direct smear microscopy. Materials and reagents are affordable and are available locally in countries where tuberculosis is endemic. Specimen preparation does not require significantly extra workload, and the longer time required for the drying of slides (after smearing) involves no extra labor time. We did not experience faster or easier reading (because of AFB being more readily seen in smears with less debris) by bleach sedimentation microscopy, as was reported in previous studies [8]. Instead, technicians reported difficulties in focusing, because of the lack of background material in the smears, and complained of eye fatigue.

Bleach sedimentation can result in fragile smears [24]. The 36 (1.9%) unreadable bleach smears were mainly caused by the smear washing out. Indeed, the smear can wash off during slide staining, and care is required to avoid this problem. Overheating of slides may result in the formation of crystals of hydroxide, which might compromise readings. Another drawback of bleach sedimentation is the poor stability of bleach when stored in suboptimal conditions [3, 8]. In our study, we monitored the presence of free chlorine in the bleach solution, using a pool tester. This testing required previous dilution of the solution, because the highest concentration for a simple test to detect free chlorine is 250 mg/L (Color Comparator; Wagtech), and higher concentrations require spectophotometric titration. This monitoring requires additional work, which can introduce dilution errors and may be difficult to perform in routine conditions. Nevertheless, our experience and previous reports show that bleach solutions at concentrations of <6% available chlorine have an acceptable shelf life of at least 6 months if stored under suitable conditions (i.e., a cool place in opaque, nonreactive bottles with airtight caps) [20]. The 1-day delay of the overnight bleach sedimentation is a limitation, which may be overcome by use of a shorter sedimentation time [3].

The percentage of smears with AFB that were missed by bleach sedimentation microscopy, compared with by direct smear microscopy, was relatively low (0.4% with the 1 AFB/100 HPF threshold and 2.7% with the 10 AFB/100 HPF threshold). Despite the absence of culture (a limitation of this study), these bleach sedimentation—negative smears were likely to be false-negative results, perhaps explained by the aforementioned difficulty in focusing, the potential dispersion of AFB throughout the specimen, the breaking up of AFB clumps after homogenization, or smear fragility [24, 25]. Because of the same limitation of no culture, the proportion of positive smears caused by Mycobacterium species other than M. tuberculosis could not be reported. A recent survey conducted in a district of Nairobi reported that the isolates in 8.2% of 85 positive specimens were identified as Mycobacterium species other than M. tuberculosis [26].

In conclusion, this study suggests that the bleach sedimentation method can significantly improve the yield of microscopy for tuberculosis diagnosis when used in a peripheral clinic in a setting of a high prevalence of HIV. Its benefit remains significant even when a sensitive AFB threshold is used. The method is simple, affordable, and appropriate for laboratories that already perform microscopy. Additional evaluation, under operational conditions, is urgently required to determine its potential as a tool for tuberculosis control. Cost-effectiveness analyses of strategies that combine direct smear microscopy and bleach sedimentation microscopy are necessary to optimize services and to balance the need for increased sensitivity with the need to prevent patient dropout during diagnostic process.