Wild-type viruses from the ViroLogic phenotype-genotype database were evaluated to determine the upper confidence limit of the drug susceptibility distributions, or "biological cutoffs," for the PhenoSense HIV phenotypic drug susceptibility assay. Definition of the natural variation in drug susceptibility in wild-type human immunodeficiency virus (HIV) type 1 isolates is necessary to determine the prevalence of innate drug resistance and to assess the capability of the PhenoSense assay to reliably measure subtle reductions in drug susceptibility. The biological cutoffs for each drug, defined by the 99th percentile of the fold change in the 50% inhibitory concentration distributions or the mean fold change plus 2 standard deviations, were lower than those previously reported for other phenotypic assays and lower than the clinically relevant cutoffs previously defined for the PhenoSense assay. The 99th percentile fold change values ranged from 1.2 (tenofovir) to 1.8 (zidovudine) for nucleoside reverse transcriptase RT inhibitors (RTIs), from 3.0 (efavirenz) to 6.2 (delavirdine) for nonnucleoside RTIs, and from 1.6 (lopinavir) to 3.6 (nelfinavir) for protease inhibitors. To evaluate the potential role of intrinsic assay variability in the observed variations in the drug susceptibilities of wild-type isolates, 10 reference viruses with different drug susceptibility patterns were tested 8 to 30 times each. The median coefficients of variation in fold change for the reference viruses ranged from 12 to 18% for all drugs except zidovudine (32%), strongly suggesting that the observed differences in wild-type virus susceptibility to the different drugs is related to intrinsic virus variability rather than assay variability. The low biological cutoffs and assay variability suggest that the PhenoSense HIV assay may assist in defining clinically relevant susceptibility cutoffs for resistance to antiretroviral drugs.