Five methods for the assessment of plasma viral load (VL) were evaluated in 103 seropositive patients infected with various subtypes of HIV-1. The methods included three RNA-based assays (Amplicor Monitor 1.5, Quantiplex version 2.0, NucliSens), one ultrasensitive reverse transcriptase (PERT) assay and one "boosted" p24 antigen (Ag) enzyme immunoassay (EIA). Subtyping was based on sequencing in env. The sensitivities were, in decreasing order, Amplicor > PERT > p24 Ag > NucliSens > Quantiplex. The low sensitivity of NucliSens was related to the missing of several non-B (A, E, F, G) or recombinant strains, whereas that of Quantiplex did not depend on subtype. In the 1 group O sample and 4 group M samples, only PERT assay or p24Ag EIA produced a positive result. In the quantitative range, correlation was best between Amplicor and Quantiplex (r = 0.8848), fair between Amplicor and NucliSens (r = 0.7064) or PERT assay (r = 0.7266), lowest between Amplicor and p24Ag EIA (r = 0.3989). Amplicor underestimated VL in 1 subtype E sample. Thus, Amplicor performed best in terms of sensitivity (compared with all other assays) and accuracy (compared with NucliSens, PERT assay, and p24Ag) for non-B subtypes in group M samples. PERT assay appears useful for VL assessment in infections by group O or other highly divergent viruses.