Protocol BMS 020 was a double-blind, prospective clinical trial in which two different doses of stavudine (20 and 40 mg twice daily) were compared in human immunodeficiency virus (HIV)-infected patients with previous exposure to zidovudine for longer than 16 weeks. Genotypic and phenotypic resistance to both zidovudine and stavudine were examined after at least 2 years of stavudine monotherapy. None of 35 tested individuals harboured the codon 50 and/or 75 mutations previously described to be associated with stavudine resistance. However, more than 80% of the individuals carried mutations associated with zidovudine resistance, despite all these patients having stopped zidovudine at least 2 years earlier. Significant phenotypic resistance to stavudine was observed only in 2 of 5 tested individuals, although IC50 values were increased only 6.6- and 9.2-fold, respectively. These two patients had suffered a decline in their CD4 count, and one of them had high levels of plasma viraemia. The sequence analysis of the reverse transcriptase (RT) gene (aa 30 to 240) in these five patients revealed no changes that could be involved in stavudine resistance. In contrast, and despite having stopped treatment with zidovudine more than 2 years before, phenotypic resistance to zidovudine was observed in all five subjects, with IC50 values raised by more than 75-fold in all of them. Moreover, all harboured codon substitutions within the RT gene associated with zidovudine resistance, and these mutations remained in viral genomes examined after virus co-culture, suggesting that they provided some biological advantage to mutants, even in the absence of drug pressure. In conclusion, both genotypic and phenotypic resistance to stavudine seem to be a rare event in patients exposed to the drug, even after long periods of exposure.