Factors associated with undetectable viral load (<80 copies/ml) were investigated among non-pregnant adults in antiretroviral treatment in a specialized service for HIV/AIDS in Southern Brazil. Use of antiretrovirals was investigated in two interviews (one month interval). Clinical data were collected from the clinical records; viral load previous to adherence measurement was defined the viral load previous to the first interview; the final viral load, the viral load subsequent to the second interview (interval between measures approximately 6 months). Undetectable final viral load occurred in 48% of the patients and was positively associated with levels of treatment adherence (p<0.001), being 19% for less than 60% of adherence and about 60% for adherence greater than 80%. In the multivariate model, the odds of undetectable final viral load was four times greater for 80-94% and > or =95% of adherence (CI 95% 1,80-13,28; CI 95% 1,73-9,53), compared with less than 60% adherence; it was greater for less than 6 months in treatment (OR = 3.37; CI 95% 1.09-10.46); and smaller for viral load previous to adherence measurement 5.2 log10 (OR = 0.19; CI95% 0.06-0.58), adjusted for these variables and sex, age, clinical status, current immune status, group of drugs and interval between the two measurements of viral load. The crude odds were lower for age 16-24 years and use of Nucleoside Analog Reverse Transcriptase Inhibitors only, but these effects were not significant in the multivariate model. There was no evidence of effect of sex, clinical status, current immune status, and changes in treatment regimen. Treatment adherence gave the largest effect. Motivational interventions directed at adherence may improve treatment effectiveness.