BACKGROUND: The lack of HIV type-1 (HIV-1) viral load (VL) monitoring in resource-limited settings might favour the accumulation of resistance mutations and thus hamper second-line treatment efficacy. We investigated the factors associated with resistance after the initiation of antiretroviral therapy (ART) in the absence of virological monitoring. METHODS: Cross-sectional VL sampling of HIV-1-infected patients receiving first-line ART (nevirapine or efavirenz plus stavudine or zidovudine plus lamivudine) was carried out; those with a detectable VL were genotyped. RESULTS: Of the 573 patients undergoing VL sampling, 84 were genotyped. The mean number of nucleoside/nucleotide reverse transcriptase inhibitor (NRTI) mutations increased with the duration of ART exposure (P=0.02). Multivariable analysis showed that patients with a CD4+ T-cell count < or =50 cells/mm(3) at ART initiation (baseline) had a higher mean number of both NRTI and non-NRTI (NNRTI) mutations than those with a baseline CD4+ T-cell count >50 cells/mm(3) (2.10 versus 0.56; P<0.0001; and 1.65 versus 0.76; P=0.005, respectively). A baseline CD4+ T-cell count < or =50 cells/mm(3) predicted > or =1 NRTI mutation (adjusted odds ratio [AOR] 7.49, 95% confidence interval [CI] 2.20-32.14), > or =1 NNRTI mutation (AOR 4.25, 95% CI 1.36-15.48), > or =1 thymidine analogue mutation (AOR 8.45, 95% CI 2.16-40.16) and resistance to didanosine (AOR 6.36, 95% CI 1.49-32.29) and etravirine (AOR 4.72, 95% CI 1.53-15.70). CONCLUSIONS: Without VL monitoring, the risk of drug resistance increases with the duration of ART and is associated with lower CD4+ T-cell counts at ART initiation. These data might help define strategies to preserve second-line treatment options in resource-limited settings.