Background: to measure the adherence and non-adherence to a highly active antiretroviral therapy (HAART); to identify some of factors related to HAART non-adherence; to establish the predictive values of the factors associated with non-adherence to HAART.

Method: analytical study of prevalence. This study was accomplished at the AIDS Day Unit, Clinical Hospital, State University of Campinas (UNICAMP), Campinas, São Paulo, Brazil. Sixty patients were selected and interviewed in March 2002. Considering the three days before the interview, they were classified as adherent: intake of 95% or more of the total number of pills prescribed.

Results: adherence was 73.3% in the 95% criterion. The multivariate logistic regression analysis using the Stepwise selection indicated: the risk of non-adherence was 6.48 fold greater in the African Brazilian race; the absence of side effects presented a 7.6 fold risk that increased by 12% at each pill intake, which means that the risk was 1.12.

Conclusions: the adherence observed was higher than the one found in literature. It is evident the need for an in-depth investigation of the predictive factors and the suggestion that socio-demographic and cultural factors may interfere in treatment adherence to HAART. A professional should, therefore, be well informed about the population he treats so specific interventions that favor the improvement of adherence to the HAART treatment can be adopted.

KEY-WORDS: adherence; HIV; infectious diseases; risk factors; antiretroviral therapy.