Validity of clinical symptoms and chest radiography in predicting pulmonary tuberculosis at the Kenyatta National Hospital, Nairobi, Kenya

Background: Tuberculosis (TB) incidence increases annually by 1% globally, and was declared a global emergency by the World Health Organization (WHO) in 1993. Diagnosis of TB infection is confirmed by microscopy of sputum smears for acidfast bacilli (AFB) while suspicion of the infection is frequently based on clinical and radiological findings. However, sensitivity of the microscopic diagnosis in identifying all cases of TB is only 40-60%, and patients with negative sputum smears for AFB represent a diagnostic dilemma. Therefore, the gold standard approach for TB diagnosis is to culture the organism from sputum or blood in a laboratory. Unfortunately this requires dedicated equipment and takes up to 8 weeks to yield results. It was, therefore, necessary to develop clinical and chest radiological algorithms that can be predictors for positive culture in Kenya. Objective: To establish predictive value of clinical symptoms and chest radiography in the diagnosis of pulmonary tuberculosis. Study Design: Prospective cross- sectional survey Setting: Accident and Emergency department and general medical wards, Kenyatta National Hospital (KNH), Nairobi Kenya. Study population: All patients seen in accident and emergency department, and admitted in the KNH medical wards within the study period with cough for more than two weeks. Methodology: Patients with cough for more than two weeks were screened and those who met the inclusion criteria were recruited into the study. Each patient was offered HIV testing after counseling. Three sputa were taken of which one was early morning sputum for examination for AFB. In addition, confirmatory sputum culture for Mycobacterium tuberculosis complex was conducted on all patients and the patients informed about the results by their primary clinician. Chest radiographs was performed on all recruited patients at the time of recruitment. Two trained and experienced radiologists blinded to the laboratory diagnosis of TB, independently assessed the chest radiographs. Results: A total of 271 were included into the study. Of these 126(46.5%) had culture proven TB. 55(44%) were HIV positive, 56(44%) were HIV negative and 15(12%) declined to be tested. Sputum smear microscopy had an overall sensitivity of 71% and specificity of 96%. Smear microscopy was frequently positive in patients without HIV infection 46(82%) compared to 33(60%) who were infected with HIV (p=0.019). Majority 68(54%) of patients with culture confirmed TB were males. The mean age was 32 years, range 15 to 63 years. The most prevalent symptoms in patients with TB were weight loss (52%), pleuritic chest pain (45%) and night sweats (40%). A combination of these three symptoms had 1.3 fold associations with culture proven infection. On univariate analysis, weight loss (p<0.001), pleuritic chest pain (p=0.001), night sweats (p<0.001), nodular alveolar infiltrates affecting upper zones (p<0.001) and cavitations affecting upper zones (p<0.001) were significantly associated with diagnosis of pulmonary tuberculosis. On multivariate analysis, weight loss (p=0.001), night sweats (p=0.001), chest pain (p=0.001), nodular infiltrates affecting upper zones (p<0.001) and cavitations (p=0.001) independently predicted diagnosis of pulmonary tuberculosis. HIV positive patients with unintentional weight loss defined as 10% presumed or measured weight loss in the last one month, were six times more likely to have TB (95% Cl, 2.73 to 14.5) while haemoptysis was significantly associated with HIV negative patients (p=0.009). Likewise HIV positive patients were less likely to have cavitations (p=0.003) and more likely to have intrathoracic adenopathy (p=0.043), miliary pattern (p<0.001), diffuse infiltrates (p=0.013) and normal chest radiographs (0.048) compared to HIV negative patients. HIV positive patients with positive sputum cultures were less likely to have had a positive sputum smear compared to HIV negative patients (p=0.019). In patients with HIV/TB co-infection, mean CD4 count was 124 cells/mm3 with a median of 98 cells/mm3 and a range of (4-501). 32(58%) had CD4 counts of less than 100cells/mm3 and were more likely to have intrathoracic adenopathy (p=0.67), miliary pattern (p=0.078), normal chest radiograph (p=0.17) and less likely to have upper lobe infiltrates (p=0.062) and cavitations (p=0.3Q). Conclusion: Night sweats, weight loss and pleuritic chest pain were positive predictors of pulmonary tuberculosis and when combined had 1.3 fold association with culture proven TB. Cavitatory disease and nodular infiltrates affecting upper zones in the chest radiograph were still typical of TB infection. In the setting of an underlying HIV infection, chest radiographic findings tend to be more atypical and could present as normal or with minimal involvement and especially in those with advanced immunosuppression. (author)