Objectives. To compare the fever course after starting therapy in pati
ents diagnosed of active tuberculosis with and without HIV infection a
nd evaluate the usefulness of empiric antituberculous therapy in diagn
osing the disease. Methods. Review of clinical records from all patien
ts meeting the following criteria for three years: recovery of Mycobac
terium tuberculosis from any clinical sample, knowledge of serological
status to HIV, initial therapy of tuberculosis, absence of ther cause
s of fever identified, and not being treated with drugs which potentia
lly could interfere with the course of fever during their hospital sta
y. Results. At admission HIV-positive patients with tuberculosis were
afebrile in a significantly lower proportion than HIV-negative patient
s (17% vs. 54%, respectively; p<0.001). After initiating antituberculo
us therapy, the median time to fever resolution was similar in both HI
V-positive and HIV-negative patients (6 and 4 days, respectively). Aft
er two weeks of therapy, 25% of HIV-positive patients and 23% HIV-nega
tive patients still had fever. No factor was identified which could pr
edict the delay in resolution of fever. Conclusions. The course of fev
er was similar in both HIV-positive and HIV-negative patients after in
itiating antituberculous therapy. This empirical therapy may be useful
in diagnosing tuberculosis, as fever resolved in the first two weeks
of therapy in most patients.