First-generation reverse transcription-PCR (RT-PCR) assays for severe acute respiratory syndrome-associated coronavirus (SARS-CoV) gave false-negative results in a considerable fraction of patients. In the present study, we evaluated two second-generation, replicase (R) gene-based, real-time RT-PCR test kits--the RealArt HPA coronavirus LC kit (Artus, Hamburg, Germany) and the LightCycler SARS-CoV quantification kit (Roche, Penzberg, Germany)--and a real-time RT-PCR assay for the nucleocapsid (N) gene. Detecting the N-gene RNA might be advantageous due to its high abundance in cells. The kits achieved sensitivities of 70.8% (Artus) and 67.1% (Roche) in 66 specimens from patients with confirmed SARS (samples primarily from the upper and lower respiratory tract and stool). The sensitivity of the N-gene assay was 74.2%. The differences in all of the sensitivities were not statistically significant (P = 0.680 [analysis of variance]). Culture cells initially contained five times more N- than R-gene RNA, but the respective levels converged during 4 days of virus replication. In clinical samples the median concentrations of R- and N-gene RNA, respectively, were 1.2 x 10(6) and 2.8 x 10(6) copies/ml (sputum and endotracheal aspirates), 4.3 x 10(4) and 5.5 x 10(4) copies/ml (stool), and 5.5 x 10(2) and 5.2 x 10(2) copies/sample (throat swabs and saliva). Differences between the samples types were significant but not between the types of target RNA. All (n = 12) samples from the lower respiratory tract tested positive in all tests. In conclusion, the novel assays are more sensitive than the first-generation tests, but they still do not allow a comprehensive ruling out of SARS. Methods for the routine sampling of sputum without infection risk are needed to improve SARS RT-PCR.