HIV-1 genotype assay using plasma viruses has been widely applied for detection of resistant viruses in infected individuals, whereas there are only a few reports about proviral genotype in peripheral blood mononuclear cells (PBMCs). To determine which sample, plasma or PBMC, should be used for early detection of drug-resistant viruses during antiretroviral treatment, we analyzed 275 plasma-derived and 211 PBMC-derived HIV-1 protease sequences obtained from HIV-1-infected patients during protease inhibitor (PI) therapy. In 70 of 107 pairs (65.4%) of plasma and PBMC samples taken from the same blood draws, the numbers of PI resistance-associated mutations in the plasma-derived genotype were different from those in the PBMC-derived genotype. Plasma viruses had more PI resistance-associated mutations than PBMC proviruses (P = 0.0004). Analysis of serial samples showed that plasma-derived genotype assay could detect primary mutations about 425 days earlier than PBMC-derived genotype when the plasma viral load was less than 10(4 )copies/mL. Our data suggest that genetic turnover of PBMC proviruses is slower than that of plasma viruses and that time lag between emergence of mutations in plasma-derived and PBMC-derived genotypes correlates inversely with viral load. Plasma viruses should be the material of choice for early detection of drug resistance during antiretroviral treatment.