Objective: To compare the frequency of unfavorable outcome (death or default and treatment failure) between tuberculosis (TB)/HIV co-infected patients treated for TB after laboratory confirmation of the diagnosis and TB/HIV co-infected patients who were so treated without diagnostic confirmation. Methods: A retrospective cohort of TB/HIV co-infected patients who started TB treatment between July of 2002 and June of 2004 at an HIV/AIDS referral center in Recife, Brazil. The main exposure variable, laboratory confirmation of TB, was adjusted for three different sets of variables: sociodemographic variables; HIV/AIDS-related variables; and TB-related variables. In order to evaluate the statistical significance of the results, we calculated odds ratios, with 95% confidence intervals, and p values (from chi-square tests and likelihood ratio tests). Results: A total of 262 patients were studied. No association was found between laboratory confirmation of the diagnosis of TB at treatment outset and unfavorable outcome, even after adjustment for confounders. In the final multiple logistic regression model, the following variables remained: the presence of other opportunistic diseases; CD4 lymphocyte count below 50 cells/mm3; viral load between 10,000 and 100,000 copies/mL; dyspnea; the disseminated form of TB; and change in the TB treatment regimen due to adverse reactions or intolerance. Conclusions: Our results suggest that TB treatment in TB/HIV co-infected patients without etiologic confirmation of TB, at the discretion of experienced physicians in referral centers, did not increase the risk of unfavorable outcomes. In addition, it allowed the identification of groups that should be closely monitored due to a greater risk of unfavorable outcomes.

Keywords: Tuberculosis; Diagnosis; Therapeutics; Treatment outcome; HIV infections.