TRANSACTIONS OFTHE ROYAL SOCIETY OFTROPICAL MEDICINE AND HYGIENE (1997) 91,420-421

Diagnosing tuberculosis in a resource-poor setting: the value of sputum concentration

David Wilkinsonl* and A. Wim Sturm3 ‘Centre for Epidemiological Research in Southern Africa, South African Medical Research Council and Hlabisa Hospital, Hlabisa, South Aji-ica;2Department of Medical Microbiology, University of Natal, Durban, South Africa

Abstract Diagnosis of tuberculosis in resource-poor settings is hampered by the insensitivity of the direct Ziehl-Neelsen (ZN) smear. Liquefaction and concentration of sputum before preparing a ZN smear has been proposed as a way of increasing diagnostic sensitivity. A field trial of this technique was done in a district hospital in South Africa among 166 consecutive tuberculosis suspects. Correlation between the 2 types of smear was high, but the extra cases diagnosed after concentration was offset by a similar number that, initially positive, were negative after concentration. Overall diagnostic sensitivity of smear microsco- py was not increased by sputum liquefaction and concentration. The value of this technique may lie in combining it with direct microscopy. Limiting specimen examination to one per patient and making an initial direct smear with subsequent concentration only if the direct smear was negative increased sensi- tivity from 43% to 55% without any reduction in specificity. However, overall diagnostic sensitivity remained disappointing.

Keywords: tuberculosis, Mycobacterium tuberculosis, diagnosis, Ziehl-Neelsen smear, sputum concentration

Introduction Accurate diagnosis of tuberculosis is essential for both

optimal disease control and individual patient manage- ment. While smear-positive cases are more infectious than smear-negative cases, and hence more important from a public health perspective (GRZYBOWSKI et al., 1977), smear-negative/culture-positive cases also de- serve accurate diagnosis and treatment, at least in those countries that can afford to provide it. Unfortunately, the sensitivity of a Ziehl-Neelsen (ZN) smear compared to culture is typically less than 50% (ABER et al., 1980).

GEBRE et al. (1995) renorted that liauefaction and concentration of‘sputum increased smea; sensitivity by 100%. They did not, however, report details of the pa- tients sampled and their statistical analysis was per spe- cimen, not per patient: this is important because the need is to diagnose individuals irrespective of the number of smears produced. It is possible, for example, that doubling the number of positive smears increases the proportion of patients diagnosed with tuberculosis by a smaller amount if the ‘extra’ positive smears tend to be produced by the same patients. Nevertheless, en- couraged by their findings we decided to test the tech- nique in a busy district hospital in rural South Africa.

Methods Subjects and setting

The study took place in Hlabisa hospital, South Africa (WILKINSON, 1994). All patients with suspected tuberculosis (CROFTON et al., 1992) that present to the hospital or its clinics are admitted for investigation. All have a chest X-ray taken and are asked to produce a sputum specimen each morning for direct ZN smear. A single positive ZN smear is considered diagnostic (CROETON et al.. 1992). Not more than 3 snecimens are , , examined from each patient if initial smears are nega- tive. One hundred and ninety-six consecutive adults admitted over a period of 3 months were enrolled in the trial.

Specimen examination Each patient was provided with 2 specimen contain-

ers (labelled A and B) each morning and was asked to expectorate into each (A first, then B). The laboratory was given one of the pair, and a direct ZN smear was made, in standard fashion. Sputum was transferred to a clean glass slide by swab, air-dried, heat-fured, stained by the ZN method, and examined at 1000 x magnifica-

*Address for correspondence: CERSA/MRC, P.O. Box 187, Mtubatuba 3935, South Africa.

tion. Smears showing alcohol- and acid-fast bacilli were reported as positive. The container given to the labora- tory (A or B) was altered for each subsequent specimen to help control for possible differences in specimen quality depending on the order in which they were pro- duced.

The remaining sputum was stored at 4°C and then transported to the Department of Medical Microbiology in the University of Natal, where the sputum specimens were liquefied by the N-acetyl-L-cysteine/sodium hy- droxide (NALC-NaOH) method and concentrated bv centrifugation for 20 mm at 3000 g and 4°C before be- ing cultured on Lowenstein-Jensen and Middlebrook 7Hll agar.

The second specimen of each nair was liquefied and concentrated in- Hlabisa hospital with laboratory staff ‘blinded’ to the identity of the patient producing the specimen. Sputum was transferred to a 10 mL screw- top tube and mixed with an equal volume of 45% NaOCl for 15 min. The mixture was shaken by hand every few minutes before 8 mL of distilled water were added and the mixture was centrifuged at about 15OOg for 15 min. The suuematant was discarded and the nel- let suspended in a ‘few drops of the remaining fluidbe- fore a ZN-stained smear was prepared in the standard way (GEBRE et al., 1995).

Analysis Final diagnoses made by the attending medical officer

were recorded for all patients. Comparison was made between both the direct and concentrated ZN smears and the culture results. Sensitivitv, specificity, and pos- itive and negative predictive values for each-(with their 95% confidence intervals [CIs]) were calculated directly with EpiInfoTM version 6.02.

Results Thirty patients (62 specimens) were excluded from

the analysis. A concentrated ZN smear was not made from 15 specimens, 9 cultures were contaminated, and 3 culture results were lost. All results obtained from the 30 patients with any missing result were excluded, to al- low comparison between patients who had results for di- rect ZN smear, concentrated ZN smear, culture, and the final diagnosis (330 specimens; 166 patients).

Results from direct and concentrated ZN smears agreed for 305 (92%) specimens. However, 13 (3 1%) of 42 specimens positive by direct ZN smear were negative following liquefaction and concentration, and 12 (29%) of 41 specimens positive by concentrated ZN smear were negative by ZN smear (Table 1).

DIAGNOSING TUBERCULOSIS IN A RESOURCE-POOR SETTING 421

Table 1. Comparison of results from sputum smears stained by the Ziehl-Neelsen method directly and after specimen concentration for the diagnosis of pulmonary tuberculosis

Concentrated smear Positive Negative Totals

Direct smear Positive 29 13 42 Negative 12 276 288 Totals 41 289 330

Table 2. Comparison of culture results with both direct and concentrated Ziehl-Neelsen stained smears for the diagnosis of pulmonary tuberculosis

Direct smea? Positive

Culture Positive Negative Totals

38 4 42 Negative Totals

73 I24 2 77 166

Concentrated smea& Positive 39 2 41 Negative 50 2 125 Totals 89 166

=Sensitivity 38/39=43% (95% CI 32-54%); specificity 73177 =95% (95% CI 87-98%); positive predictive value 38/42 =91% (95% CI 77-97%); negative predictive value 73/124 =59% (95% CI 50-68%). bSensitivity 39/89=44% (95% CI 34-55); specificity 75177 =97% (95% CI 90-100%); positive predictive value 39141 =95% (95% CI 82-99%); negative predictive value 77/166 =60% (95% CI 51-69%).

Comparison of results from the 2 smear techniques with culture is shown in Table 2. Sensitivity for both smear techniques was equally low (43%-44%) but spe- cificity was high (95-97%). Ofthe 124 patients who had negative direct smears, 51 (41%) were positive by cul- ture. Of these, 25 (49%) were subsequently diagnosed with tuberculosis clinically by the attending medical of- ficer, but 26 were not. The 26 misdiagnoses included 22 with chest infection and 4 with asthma. Of the 73 pa- tients with negative cultures, 16 (22%) were incorrectly diagnosed with tuberculosis.

Discussion Our results question the value of liquefaction and

concentration of sputum as a method for increasing the diagnostic yield of tuberculosis from smear microscopy in this setting. Although concentration correctly identi- fied an extra 12 cases of disease missed by direct smear, a similar number of cases that were positive by direct microscopy initially were not detected after liquefaction and concentration.

This prospective study under field conditions con- firmed the low sensitivity of direct ZN smears in the diagnosis of active tuberculosis and emphasizes the need to develop more sensitive diagnostic techniques. Al- though 49% of the smear-negative/culture-positive cases were diagnosed with tuberculosis clinically, the

rest were not and were discharged. Similarly, the lack of an accurate diagnostic test meant that 22% of patients without active tuberculosis (as defined by positive culture) were misdiagnosed and treated unnecessarily. Over-diagnosis exposes individuals to potentially toxic drugs and disrupts their lives unnecessarily, while unnecessarily burdening control programmes. Under- diagnosis also adversely affects individuals as, without a correct diagnosis, correct therapy cannot be provided.

It was encouraging that several specimens negative by direct smear became positive after liquefaction and con- centration, buy why did some specimens initially posi- tive become negative? It is possible that the method used was inappropriate and either liquefaction was incom- plete or acid-fast bacilli were disrupted by the process. However, all smears were checked for quality and the technicians preparing the smears were experienced in the process.

Is it possible to consider strategies that combine direct and concentrated smears? It would be possible to make a direct ZN smear from each of the 3 specimens first and then to make a concentrated ZN smear if the direct smear were negative. The risk here is of doubling the number of sm&rs done, but sensitivity would increase to 56% (48186; P=O*O91. with a snecificitv of 95%. An alternat&e pol&y would’he to examine only one sputum specimen per patient, and to make a direct smear first and a concentrated smear if the direct one were nega- tive. Interestingly, this would produce almost identical results (sensitivity 55%, specificity 95%), but with far fewer smears needed. This is because almost all the smear-positive specimens were positive on the first spec- imen. However, overall diagnostic sensitivity remains disappointing.

In conclusion, sputum liquefaction and concentration followed by ZN smear examination did not significantly increase diagnostic yield in our setting. It is considerably more time-consuming than making a direct ZN smear, and it requires a centrifuge. An important observation is that limiting examination to only one specimen per pa- tient, but using both direct and concentrated smears, did increase sensitivity’ (from 43% for direct smear alone to 55%) without reducing snecificitv. It will be imnor- tant to confirm this obse&ahon through further opera- tional research. More sensitive techniques will need to be developed if the patients with active tuberculosis are to be accurately diagnosed in resource-poor settings.

References Aber, V. R., Allen, B. W., Mitchison, D. A., Ayuma, I?, Ed-

wards, E. A. & Keves, A. B. (1980). Qualitv control in tuber- culosig bacteriol& 1. &boraiox$ s&dies on isolated positive cultures and the efficiency of direct smear examina- tion. Tuber&, 61, 123-133.

Crofton, J., Horne, N. & Miller, F. (1992). Clinical Tuberculosis. London: Macmillan.

Gebre, N., Karlsson, U., Jhsson, G., Macaden, R., Wolde, A., Assefa, A. & MiBrner, H. (1995). Improved microscopical diaenosis of uulmonarv tuberculosis in develooine countries. Trakmions bf the Roy;1 Society of Tropical Midicse and Hy- giene, 89, 191-193.

Grzybowski, S., Barnett, G. D. & Styblo, K. (1975). Contacts of active pulmonary tuberculosis. Bulletin of the Znternational Union Against Tubekulosis, 50,90-106.

Wilkinson, D. (1994). High-compliance tuberculosis treat- ment programme in a rural community. Lancet, 343, 647-648.

Received 22 November 1996; revised 18 February 1997; ac- cepted for publication 18 February 1997