Direct microscopy versus sputum cytology analysis and bleach sedimentation for diagnosis of tuberculosis: a prospective diagnostic study

Tuberculosis (TB) is a major public health problem, with an estimated 2 billion people infected with tubercule bacilli worldwide. Estimated global prevalence of the disease is 139 per 100,000 population [1]. In recent years, the prevalence of TB has increased steeply, driven in large part by the HIV epidemic. TB in patients co-infected with HIV can be difficult to diagnose, especially by sputum-smear microscopy--often the only diagnostic tool available in resource-limited settings. The medical humanitarian aid agency Médecins Sans Frontières (MSF) works in settings where the health infrastructure needed to treat patients with TB, many of whom are co-infected with HIV, is either absent or extremely limited.

The gold standard for diagnosing pulmonary tuberculosis is culture of sputum on Löwenstein-Jensen medium. However, due to lack of access to culture facilities and the long turn-around times involved with sputum culture, most programmes use direct Ziehl-Neelsen microscopy for detection of acid-fast bacilli (AFB) in sputum smears. This technique, when performed optimally, has reported sensitivities ranging from 61.8% to 70% compared with the gold-standard [2–4]. Sensitivity is reduced if samples are of poor quality, which is often the case in children and in people who are HIV-positive, since sputum expectoration is more difficult [3, 5].

Sputum samples are also often inadequate in patients who are immune-compromised or who have not been given correct instructions on sputum expectoration [6]. Patients frequently provide pure saliva or very small amounts of sputum in saliva instead of an appropriate, purulent, sputum sample [7]. A negative result issued on examination of such a sample is misleading, since it implies that a correct sputum specimen has been examined. Sputum cytology analysis (SCA), in which smears are examined and graded according to an algorithm, has been proposed as a means to ensure sufficient sample quality. Further samples can be then requested if quality is inadequate [8].

Sputum concentration by homogenisation and microscopic examination of the sediment can increase the detection rate of AFB. Several concentration techniques using sedimentation or centrifugation have been reported [9–11]. Bleach homogenisation followed by overnight sedimentation can significantly increase detection of AFB compared with direct microscopy [12–14]. A systematic review of this technique found an increase in sensitivity of 33% in one of four studies where culture was used as a gold standard, and a mean incremental yield of 6% in four studies in which culture was not used [9]. Centrifugation resulted in higher yields than sedimentation. Cattamanchi and colleagues [15] found that bleach concentration resulted in small increases in sensitivity and small decreases in specificity. However, centrifugation requires electricity as well as equipment that might not be present in peripheral laboratories in resource-limited settings. Homogenisation can be achieved with various agents, such as sodium hydroxide (NaOH) and N-acetyl-L cysteine (NALC) [11]. Household bleach--sodium hypochlorite (NaOCl)--can also be used, and it has the advantage of being easily available worldwide.

Due to the variability of results to date, it has been recommended that further research is done before the adoption of bleach concentration as a standard diagnostic method [10].

We did a prospective field assessment in routine conditions in a peripheral laboratory in an MSF-supported hospital in Mindouli, Republic of Congo. We aimed to compare the proportion of smear-positive patients detected with and without bleach concentration of sputum by overnight sedimentation with validation of all samples by SCA. Our hypothesis was that the combination of SCA and bleach sedimentation would significantly improve the detection rate for smear microscopy of pulmonary TB.

Patient enrolment progressed more slowly than expected, with only 280 patients enrolled after an extension of the study from 1 year to 18 months. The total number of samples obtained was 890. Of the 280 patients enrolled, 41% (115) were male. The mean age was 35 years (SD 12.1). Nine patients (3.2%) were lost to follow up, and three patients died.

Most patients (223/280) provided three samples. 49/280 provided four or more and 8/280 provided fewer than three.

When a case definition of two positive smears was used, bleach sedimentation gave an incremental yield of 5/118, or 4.2% (95% CI 1.4-9.6; Table 2). Only five extra patients were detected--4.1% (95% CI 1.3-9.2) of all patients detected as positive with this case definition. After SCA, two extra patients were detected on sedimentation but none on direct microscopy. The incremental yield of both techniques was 7/118, or 5.9% (95% CI 2.4-11.8; Table 2).

83% (737/890) of samples, including extra samples requested after SCA, were good-quality sputum. 15% (137/890) were insufficient or degraded sputum. 2% (16/890) of samples were saliva or mucus. One patient failed to produce any sputum samples, but was retained in the study (three saliva samples, negative on direct, all positive on sediment).

All poor-quality samples (degraded, mucus, and saliva) were negative on direct microscopy. Using bleach sedimentation on 47 poor-quality samples resulted in seven positive results (18.4%: four [10.5%] for degraded, three [7.9%] for saliva), which improved the overall detection rate (p = 0.0215). The proportion of good-quality samples giving a positive result was significantly more using sedimentation (360 of 890) than direct microscopy (347 of 890; p = 0.0044).

Our results show that bleach sedimentation and SCA increased the detection of smear-positive TB patients by 12.4%, which is in line with the findings of other studies reporting incremental yields ranging from 7% to 253% [20]. A meta-analysis found that, in studies comparing overnight sedimentation with culture, the mean increase in sensitivity was 23%, whereas in those without culture, the mean incremental yield was 5% [9].

The main implication of our study is that implementation of bleach sedimentation would significantly increase the detection of smear-positive patients if routinely implemented in resource-limited settings where gold standard techniques are not available.

SCA is a low-workload intervention which takes roughly 15-30 seconds to perform, and so can easily be integrated into routine examination. Our QC results show that the technique is highly reproducible. However, bleach sedimentation increased workload and delayed results by 1 day compared with direct microscopy.

The combination which yielded the highest positivity (45.4%), using a maximum of two samples and incurring minimal extra workload and delay, was a direct smear for the first sample and bleach sedimentation for the second. This approach allows rapid identification of patients positive on the first direct smear without waiting for the results of sedimentation. The use of two samples (the supervised sample and the one obtained the next morning) increased sensitivity by allowing for one sample to be of poor quality, and by concentrating the sample which is likely to be of the best quality (the morning sample). This procedure is based on the 2007 WHO recommendations [19, 21].

One patient detected using direct smear microscopy was negative on sediment smear. This result could have been caused by chance, with the only AFB-containing portion of the sputum being used to make the direct smear.

The analyses of residual chlorine activity showed that there was no reduction of chlorine activity after the bleach was decanted into small bottles. Because homogenisation activity is also visible microscopically, (i.e. lack of homogenisation results in intact white blood cells after sedimentation), it should not be necessary to measure residual chlorine activity routinely.

Sediment smears were fragile and easily washed off the slides during staining. Attempts to prevent this with the addition of bovine serum albumin made smears difficult to decolourise, and the practice was discontinued. AFB in sediment smears were also found to fade; we therefore recommend that QC is performed within 6 months of staining.

Our results showed that poor-quality samples are more likely to be negative on direct and sediment microscopy than good-quality samples, and that both types are significantly more likely to be positive after bleach sedimentation than on direct microscopy. 10.5% of degraded samples were negative in direct microscopy and positive on sedimentation. Although the sample size was small (4 of 38), this result suggests a need for further investigation. With the three saliva samples that became positive after sedimentation, a small amount of sputum was probably present.

A recent study has stated that macroscopic evaluation of sputum for quality is as effective as microscopic techniques such as SCA [22]. The study highlighted that rejection of samples based on cytological criteria would have resulted in some positive samples being missed. However, SCA focuses on the use of cytological classification to request extra samples when negative samples are of poor quality, and all smears are examined for AFB irrespective of quality. Although the actual number of increased smear-positive patients detected with SCA was low in our study, it can be inferred that the negative results obtained with good-quality samples are more meaningful than those from poor-quality samples. The clinician can be reassured that an appropriate patient sample was examined, due to cytological confirmation of the presence of sputum. The QC results for SCA showed agreement of 90%, which was lower than the 99% seen for AFB sensitivity on both direct and sediment smears. This result was expected, since SCA is less objective than observing the presence or absence of AFB. However, 90% agreement indicates that the results are highly reproducible.

The patient positivity rate with direct microscopy without SCA was 43.2%. This rate is considerably higher than the recommended positivity rate of 5-20% [23] and implies that case detection was not optimal--i.e. that the suspects referred for testing were likely to be more strongly positive than in a standard patient population. This was probably because of a selection bias in the patients referred for TB diagnosis. The hospital functioned as an unofficial referral hospital for people living with HIV/AIDS, and very sick patients often self-referred from all over the country. These patients could have had advanced TB infections. A larger incremental yield for bleach sedimentation and SCA might have been observed if the positivity rate was lower, and patients had been in less advanced stages of infection.

The main limitation of our results is that, due to a lack of TB culture facilities, we were unable to incorporate the gold standard of TB diagnosis: culture on Lowenstein-Jensen medium. We are therefore unable to describe the techniques with regard to increased sensitivity and potential false smear positive results, but only with regard to increased detection rate of AFB or incremental yield. Lack of culture confirmation could have resulted in false detection of positive results, which might have contributed to the incremental yield following sedimentation. The sample size was not reached despite extending the study to 18 months; unfortunately, logistical limitations prevented extending the study further. Bleach sedimentation with a case definition of two positive smears might have significantly increased yield if the required sample size had been reached. However, most results reached significance despite the lower-than-desired number of participants.

Mindouli hospital functioned as a comprehensive HIV diagnostic and treatment centre, and a large proportion of those screened for TB infection were HIV positive. Although it would have been interesting to have stratified the results based on HIV status, this was not done due to concerns about patient confidentiality if TB results were linked with HIV results at the laboratory. We are therefore unable to provide an estimate for the proportion of HIV TB coinfected patients.

Our loss-to-follow-up rate was low (3.2%), possibly because of emphasis on patient education during enrolment and provision of accommodation during the 3-day sample collection procedure, necessary since many patients came from outside the Mindouli area. Under routine conditions, without the provision of accommodation, a higher proportion of patients might be lost to follow up. The overnight delay in results following sedimentation could also lead to a higher patient loss to follow up. The loss to follow up rate after routine implementation should be monitored.

Due to the disadvantages associated with the bleach concentration technique (i.e. fragility of sediment smears, delayed results, and increased workload), routine implementation of the technique should only be considered after the feasibility of introduction in a particular context has been assessed, preferably following a pilot implementation phase.

Implementation of bleach sedimentation will lead to a significant increase in the diagnosis of smear-positive TB patients. Implementation of SCA did not result in significantly increased diagnosis of TB. Efforts should be made to obtain good-quality samples, since bleach sedimentation of poor-quality samples is unlikely to result in a large incremental yield. With a case definition of one positive smear, we recommend that bleach sedimentation is implemented in settings where work has been done to improve sputum collection practices. Sample quality can be confirmed with SCA. A pilot study should be undertaken prior to routine implementation of bleach sedimentation to determine whether the technique will be of sufficient clinical benefit in a particular setting.