N Mlika-Cabanne, M Brauner, F Mugusi, P Grenier, C Daley, I Mbaga, B Larouze and JF Murray
Institut National de la Sante et de la Recherche Medicale (INSERM), Unite 13, Hopital Claude Bernard, Paris, France.

First, we evaluated the age profile and chest radiographic abnormalities in 146 patients from Dar-es-Salaam, Tanzania, with new- onset intrathoracic tuberculosis (pulmonary, pleural, or hilar/mediastinal adenopathy), to identify features that were associated with human immunodeficiency virus (HIV) seropositivity or seronegativity; then, we combined these data with those from a companion investigation in Burundi to develop a simple scoring system to predict HIV serologic status. Using agreed-upon criteria and simplified reporting forms, initial chest radiographs were reviewed by three readers, first independently and then at a consensus conference. Of the 146 patients, 80 (55%) were HIV seropositive and 66 were seronegative. More seropositive than seronegative subjects were 31 to 40 yr old (p = 0.03). Because the radiographic characteristics of the two serologic groups were similar in Tanzania and Burundi, we combined the data for stepwise logistic regression that revealed four highly significant variables: age, small lesions, location, and lymphadenopathy. From these, we obtained an equation to calculate the probability that a given tuberculosis patients was HIV seropositive and then we derived a scoring system that in its simplest form (threshold) predicted serologic status correctly in 68.1% of patients; a graded scale was even more accurate in the high (89.1%) and low (82.6%) ranges. This scoring system should be useful when serologic testing is unavailable or refused.