Introduction Consequences of lack of viral monitoring in predicting the effects of development of HIV drug resistance mutations during HAART in resource-limited settings (RLS) is still a matter of debate. Design To assess, among HIV+ patients receiving their first-line HAART, prevalence of virological failure and genotypic resistance mutations pattern in a Médécins Sans Frontières/Ministry of Health programme in Busia District (Kenya). Methods Patients with HAART treatment for ≥12 months were eligible for the study and those with HIV-RNA ≥5000 copies/ml underwent genotypic study. Total HIV-1 RNA from Dried Blood Spots was extracted using Nuclisens method. Results 926 patients were included. Among 274 (29.6%) patients with detectable viral load, 55 (5.9%) experienced treatment failure (viral load >5.000 copies/ml); 61.8% were female and 10 (18.2%) had clinical failure. Median CD4 cell count was 116 cell/mm3 (IQR: 54-189). Median HIV-RNA was 32,000 copies/ml (IQR: 11000-68000). Eighteen out of 55 (33%) samples could be sequenced on PR and RT genes, with resistance associated mutations (RAMs) in 15 out of 18 samples (83%). Among patients carrying RAMs, 12/15 (81%) harboured RAMs associated to thymidine analogues (TAMs). All of them (100%) showed M184V resistance associated mutation to lamivudine as well as NNRTI's RAMS. Conclusions Virological failure rate in resource-limited settings are similar to those observed in developed countries. Resistance mutation patterns were concordant with HAART received by failing patients. Long term detectable viral load confers greater probability of developing resistance and as a consequence, making difficult to find out a cost-effective subsequent treatment regimen.