The sensitivity and specificity of various severe acute respiratory syndrome coronavirus (SARS-CoV) PCR primer and probe sets were evaluated through the use of commercial kits and in-house amplification formats. Conventional and real-time PCR assays were performed using a heat-block thermocycler ABI 9600, the Roche LightCycler version 1.2, or the ABI 7000 Sequence Detection System. The sensitivity of all primers was between 0.0004 and 0.04 PFU with viral cell lysate and between 0.004 and 0.4 PFU in spiked stool specimen per PCR assay. The primer sets for real-time PCR assays were at one least 1 log more sensitive than the primer sets used in the conventional PCR. A panel of viruses including swine gastroenteritis virus, bovine coronavirus, avian bronchitis virus (Connecticut strain), avian bronchitis virus (Massachusetts strain), human coronaviruses 229E and OC43, parainfluenza virus (type III), human metapneumovirus, adenovirus, respiratory syncytial virus, and influenza A were tested by all assays. All real-time PCR assays used probe-based detection, and no cross-reactivity was observed. With conventional PCR, analysis was performed using agarose gel electrophoresis and multiple nonspecific bands were observed. Two commercial extraction methods, magnetic particle capture and silica-based procedure were evaluated and the results were comparable. The former was less laborious with shorter time for completion and can easily be adapted to an automated system such as the MagNa Pure-LC, which can extract nucleic acid from clinical samples and load it into the sample capillaries of the LightCycler. As exemplified by this study, the continued refinement and evaluation of PCR procedures will greatly benefit the diagnostic laboratory during an outbreak of SARS.