There has been a sharp rise in the incidence of pulmonary tuberculosis (PTB) worldwide since the mid 1980's, particularly in the Sub-Saharan African region. This has been attributed mainly to the appearance and wide spread of Human Immunodeficiency Virus (HIV) infection on the continent [1–3]. For the diagnosis of PTB the detection of Acid Fast Bacilli (AFB) in expectorated sputum is still crucial, especially in developing countries of Sub-Saharan Africa, where other facilities including sputum culture for Mycobacterium Tuberculosis (MTB) are unavailable or are prohibitively expensive. When AFB is detected in sputum, the diagnosis of PTB is certain. However diagnostic problem start when patients with suspected PTB have a negative sputum smear [4]. It has always been recognized that a proportion of patients are sputum smear negative using the Ziehl-Nelseen (ZN) stain, the commonly used stain in most laboratories in the region to detect AFB in sputum. This is a simple, rapid and cheap test but lacks sensitivity of a single sputum test [4]. About 5000 bacilli per milliliter of sputum must be present for it to be positive. However it has been reported that multiple sputum tests in a good laboratory can give a sensitivity of about 90% [5]. Sputum smear using ZN stain for AFB seems to be even less sensitive in patients with HIV associated PTB. With the sharp rise of PTB in countries which are worst affected by the HIV epidemics, the number of patients with suspected PTB who are sputum smear negative has increased [5].

Chest radiography is not always helpful in smear negative patients. The radiographic distinction between active and inactive tuberculosis can be difficult and appearance may be atypical due to other infections in HIV positive patients [4]. In fact, substantial numbers of patients are treated for tuberculosis without definitive diagnostic criteria [5]. With the advent of HIV associated tuberculosis with more frequent smear negative tuberculosis, the role of culture in TB control programs may need to be reassessed [4]. In countries where resources are limited, and where the use of chest X-rays may be inadequate due to the cost as well as atypical presentation found in HIV infected patients, clinical and/or laboratory characteristics which are able to identify smear negative but culture positive PTB are required. The Tanzania National TB and Leprosy Programme uses a smear negative PTB diagnostic algorithm adopted from the World Health Organization (WHO) (Figure 1) [6].

This study was conducted with the aim of assessing the sensitivity of the current recommended algorithm for the diagnosis of sputum smear negative PTB.

A cross-sectional study with prospective enrollment was conducted at Muhimbili National Hospital (MNH), a university teaching and national referral hospital, and at out-patient tuberculosis clinics at the Infectious Disease Clinic (IDC), Mwananyamala, Temeke and Ilala district hospitals, from September 2000 to December 2000. All these hospitals are located in the city of Dar es Salaam. The city accounts for over 26% of all new tuberculosis cases reported each year in Tanzania [7].

Adult male and female patients aged 18 to 75 years, presenting with chronic cough (≥2 weeks); who were three times sputum smear-negative for AFBs (ZN stain); and who gave a written informed consent to participate in the study and for HIV testing were included into the study. Patients with known tuberculosis or who had PTB in the past, on anti-TB for treatment or prophylaxis; those with known chronic respiratory diseases (e.g. bronchial asthma, chronic obstructive pulmonary disease, bronchiectasis), those with misplaed HIV results, contaminated cultue results and those with heart failure were excluded. The study protocol was approved by the MUHAS Ethical Review Committee.

Over the course of the study, 467 patients were enrolled. Of those enrolled, 318 (68.1%) were HIV positive. Sputum was culture positive for MTB in 27.2% (127/467) of patients; samples of 11.1% (52/467) patients were reported to have been contaminated. In the remaining 61.7% (288/467) patients sputum was culture negative for MTB at 8 weeks. Of the 467 study subjects, 68.1% (318/467) were HIV positive. Two study subjects whose culture results were negative had their HIV results misplaced. These two, together with the 52 subjects whose culture results were contaminated were excluded from further analysis. Of the 413 samples analyzed, 30.8% (127/413) were MTB culture positive.

This study showed that of those found to have a negative result for AFB, a significant proportion (27.2%) had sputum culture positive for MTB. Therefore our data indicate that smears did not detect PTB in a very large proportion of patients. Sputum culture being the gold standard for the diagnosis of Tuberculosis disease [9] shows that sputum smear is not a very sensitive tool in the diagnosis of PTB. This has been shown by other studies where sensitivity has been described to be between 51% to 53.3% [10, 11]. One of the reasons for low sensitivity is reported to be due to the fact that 10⁴/ml are required for AFB to be seen using smear microscopy [4].

Although the gold standard for the diagnosis of Tuberculosis involves the isolation and identification of Mycobacterium Tuberculosis (MTB) using cultures [9], the cost and facilities of doing cultures are prohibitive in most developing countries. Sputum smear microscopy remains the main diagnostic tool for PTB that allows initiation of treatment and monitoring of patient progress [11, 12]. As sputum smear and microscopy is not a very sensitive tool in the diagnosis of PTB, presumptive diagnosis is usually made based on an algorithm of clinical and radiological criteria. This is commonly termed as AFB negative PTB [9, 13]. In some cases when sputum smears are negative but the patient has clinical features highly suggestive of PTB, broad-spectrum antibiotics are recommended for at lest 10-14 days and sputum smears repeated. If the patient's condition does not improve while sputum smear remains negative, a chest radiograph is done and if found to be abnormal, a presumptive diagnosis of PTB is made and the patient is started on anti-Tuberculosis treatment as AFB negative PTB [9]. In this study patients whose sputum smears were AFB negative, were evaluated using the above algorithm by the treating doctors at the clinics or hospital. A presumptive diagnosis of AFB sputum smear negative PTB was made in 41.8% (173) of all study subjects, and patients were started on anti-TB treatment as recommended by the Tanzania NTLP. The remaining 58.2% (240) patients were assumed to have other forms of respiratory diseases and were treated accordingly.

Less than half (38.1%) of those who were presumed to have active TB and started on Anti TB actually had TB by sputum culture results. More than 60% of these patients they received 8-months of treatment despite having a negative culture results. This is similar to what has been reported before in Malawi were it was reproted that 40% of smear negative had TB confirmed microbiologically after taking Broncho alveolar larvage [14]. The treatment of individuals without tuberculosis adds to the cost of the TB programs in most developing countries. Likewise about 48% (61) of patients who had active tuberculosis by the results of sputum culture were missed and they received inappropriate treatment, leaving them vulnerable to developing severe disease as well as remaining source of TB infection in the community.

The current diagnostic algorithm leading to the establishment of the diagnosis of AFB smear negative PTB is inefficient; it over-diagnoses PTB and misses a lot of people with active disease. Instituting a more sensitive diagnostic tool will prevent the unnecessary cost of treating individuals who do not have TB and at the same time it will prevent the further spread of TB. This emphasizes the need of culture and the need of further research in order to identify a better diagnostic tool for diagnosis of AFB negative PTB.

In an attempt to improve on the diagnostic algorithm, the study looked at the clinical presentation of the patients to identify clinical laboratory and radiological features that are associated with smer negative PTB and which can be used to predict PTB in patients with symptoms suggestive of PTB. A multivariate analysis showed the following features to be highly predictive of AFB negative but culture positive PTB; low eosinophil counts, a mixed type of anaemia and the presence of cavities on chest radiographs. Low eosinphil seems to be an incidental finding Further studies have to be done to confirm this findings

Limitation in the current study is the inclusion of patients with cough of more than two weeks in which there may be inclusion of patients with simple chest infection that sometime may be complicated with cough for 2-3 weeks. This may be a selection bias that may explain the low sensitivity and specificity of the diagnostic algorithm.

Another limitation is the method of sputum delivery, which is delivered by the patient himself, may have affected the results as some might bring saliva.

We could not be certain that the algorithm was followed at all times because resechers were not involved in the management of these patients rather we evaluated the treatment gien to patients by the attending clinicians. In Tanzania National TB and Leprosy programme is well organized and the algorithm is well adhered by the District TB and Leprosy Coordinators and all workers of the NTLP who were the attending clinicians in this study

• The current procedures of establishing AFB negative PTB are not sensitive enough to establish the diagnosis of active tuberculosis. They under-diagnose PTB and over treat people without PTB.

• The presence of a dry cough, a high respiratory rate, a low eosinophil count, a mixed type of anaemia and presence of a cavity were found to be predictive of smear negative but culture positive PTB.

The study protocol was approved by the MNH Ethical Review Committee.

Written consent was obtained. To those who were not able to write oral consent was obtained

The strength of our study is that it evaluates very well the dignostic algorithm.