Miliary tuberculosis. Diagnostic accuracy of chest radiography.

STUDY OBJECTIVE To assess the efficacy of the chest radiograph in identifying patients with miliary tuberculosis. DESIGN Retrospective case-controlled review by three independent blinded chest radiologists. SETTING Provincial tuberculosis control center. PATIENTS Population-based sample, including all proved cases of miliary tuberculosis diagnosed in the Province of British Columbia, Canada, between November 1982 and November 1992. One hundred cases of miliary tuberculosis were identified, of which 71 had chest radiographs available for review. Forty-four normal chest radiographs and 20 chest radiographs of patients with localized pulmonary tuberculosis were also included as controls. MAIN OUTCOME MEASURES The primary outcome of measurements was the sensitivity and interobserver variability of the chest radiograph in the diagnosis of miliary tuberculosis. The observers were also asked to describe the pattern and extent of pulmonary abnormalities based on the International Labor Organization (ILO) classification of pneumoconioses. RESULTS The three independent observers identified 42, 44, and 49 of the 71 cases of miliary tuberculosis, respectively (sensitivity, 59 to 69%). The three observers incorrectly diagnosed miliary tuberculosis in 2, 0, and 2 of the 64 controls, respectively (specificity, 97 to 100%). There was good interobserver agreement (90%, kappa = 0.77). The nodules measured less than 3 mm in diameter in 90% of cases in which miliary tuberculosis was correctly identified. In 10% of cases, the nodules measured greater than 3 mm in diameter. The ILO profusion scores ranged from mild (profusion score 1) in 45% of cases, through moderate (profusion score 2) in 27%, and severe (profusion score 3) in 28%. CONCLUSIONS The chest radiograph allowed identification of 59 to 69% of cases of miliary tuberculosis with a high specificity and good interobserver agreement.