The recent recurrence of highly pathogenic avian influenza virus A H5N1 was firstly reported in mid-December 2003 and continued through 2005. This study describes a sensitive and specific real-time RT-PCR method for the detection of influenza A subtype H5 and for monitoring virus loads. Using serial dilutions of influenza A H5N1 cultures, this assay reproducibly determined the lowest detection limit to be approximately 5 x 10(-2) 50 % egg infective doses (EID(50)). In contrast, the minimum detection limit was approximately 3 EID(50) in conventional RT-PCR with WHO primers and 10 EID(50) in antigen-capture ELISA. In tests of serial dilutions of in vitro-transcribed influenza A H5 gene RNA, there was linear amplification from 40 copies to 4 x 10(8) copies of target RNA per reaction and approximately six copies, and sometimes even as few as three copies, of target RNA tested positive in our assay. Thirty-five throat swabs from ill birds were tested: 33 samples tested positive using this assay. In comparison, 27, 13 and 19 samples tested positive using conventional RT-PCR, antigen-capture ELISA and virus isolation, respectively. To evaluate further the sensitivity of this real-time RT-PCR, a standard panel and 60 H5N1 isolates that contained different clades of influenza virus A/H5N1 were tested and all tested positive. To evaluate the specificity of the assay, 60 throat swabs from patients infected with influenza virus A H1 were tested; all were negative. Thirteen other viruses were also tested and all tested negative.