Study objectives: (1) To assess the utility of a single sputum specimen in the evaluation of HIV-infected patients who are suspected of having tuberculosis (TB). (2) To identify radiographic findings that discriminate between HIV-infected patients with TB and those with pneumonia of other causes.
Design: Retrospective cohort analysis.
Patients: All patients evaluated at Harborview Medical Center, Seattle, between January 1986 and July 1994 in whom culture of respiratory secretions grew Mycobacterium tuberculosis or Mycobacterium avium-complex. Patients who were coinfected with HIV formed the primary study group. Their chest radiographs were then compared with those of a matched group of patients with pneumonia of other causes.
Measurements and results: We identified 164 patients with TB, 20 of whom were HIV infected. The initial sputum specimen grew M tuberculosis in all HIV-infected patients and 99% of non-HIV-infected patients. Seventy percent of HIV-infected and 71% of non-HIV-infected patients had at least one positive smear. Most of these patients tested positive on their initial smear, and no significant difference was found between HIV-positive and HIV-negative patients (79% and 90%, respectively [p=0.34]). The addition of a second sputum smear identified all HIV-infected patients and all but one in non-HIV-infected patients who were ultimately determined to be smear positive. A total of 27 HIV-infected patients had a positive acid-fast bacilli sputum smear during the study period, 14 of which were attributable to TB (specificity=52%). The only radiographic findings that discriminated between HIV-infected patients with TB and those with pneumonia of other causes were the presence of cavitation or a miliary pattern (p=0.014).
Conclusions: A single sputum specimen was sufficient to establish the diagnosis in all HIV-infected patients with pulmonary TB. A single negative sputum smear made the diagnosis of TB significantly less likely. However, a minimum of two smears were necessary to achieve an acceptable early diagnostic yield. The presenting chest radiograph failed to discriminate between HIV-infected patients with TB and pneumonia of other causes in most cases.