Background. West Nile virus (WNV) is endemic in the United States. It is transmissible by blood transfusion, and the nation's blood supply is currently screened for WNV. Documented transmission of WNV infection through red blood cell (RBC) units in which the plasma co-component had a low viral load could be explained, in at least 1 instance, by cell-association of WNV; in this case, the RBC unit was released as negative by minipool nucleic acid testing (NAT) performed on plasma but was intermittently NAT-positive when subsequently tested as an individual sample. We hypothesized that a proportion of WNV bound to blood cells and was not measured by NAT performed on plasma samples. We have investigated whether WNV binds to RBCs, leading to reduction of WNV RNA detection by NAT performed on plasma samples.Methods. Equal volumes of leukoreduced RBCs and their corresponding plasma components from 20 blood donors with NAT results that were positive for WNV were tested in 5 replicates by reverse-transcriptase polymerase chain reaction TaqMan for WNV. In addition, aliquots from 8 of the RBC units were tested by infectivity assays using Vero cells.Results. The reverse-transcriptase polymerase chain reaction TaqMan assay showed that the viral load in the RBC components exceeded that in the corresponding plasma units by 1 order of magnitude. In addition, viruses associated with the RBCs were infectious in Vero cell cultures.Conclusions. These observations reinforce the notion that extraction of viral RNA from whole blood could improve assay sensitivity for blood donor screening and further reduce the residual risk of WNV transmission through transfusion.