The extent to which HIV-1 proviral DNA mutations cause clinically relevant antiretroviral resistance is still controversial. Paired plasma HIV-1 RNA and whole blood DNA were compared in patients failing HAART to investigate if the additional knowledge of archived mutations could improve the selection of potentially active drugs. Seventy-three HIV-1-infected patients with first/second HAART failure were studied before starting a new regimen based on RNA genotyping. Follow-up data after a 12-week therapy were available. DNA genotyping was retrospectively performed on stored whole blood samples and mutational profiles were compared to those from RNA. The mean number of IAS pol mutations was significantly higher in RNA (4.45 +/- 2.76) than in DNA (2.88 +/- 2.47) (P < 0.001). DNA genotyping provided a 6% increase in detection of resistance-associated mutations. Among 64/73 patients showing discordant DNA/RNA profiles, 54 (84%) also differed for predicted active drugs. 16/73 (22%) patients had >or=1 mutation revealed by DNA genotyping alone, probably affecting therapy success in 2/16. However, neither RNA/DNA discordance nor detection of isolated DNA mutations were statistically associated with outcome. In conclusion, plasma RNA remains the elective choice for HIV genotyping in patients with therapy failure, even if the detection of proviral resistance-associated mutations, not simultaneously found in RNA, is a frequent event. Therefore, in some cases DNA plus RNA genotyping might assist in choosing more accurately subsequent antiretroviral regimens.