The clinical description and RT-PCR results for the 11 patients with probable SARS from whom postmortem lung tissue samples were examined are summarized in Table 1. The mean age of the 11 patients was 70 years (range 43–99). Six of the 11 patients were men. All but 1 of the 11 patients had underlying coexisting conditions, the most common of which was diabetes mellitus in 6 patients. The mean duration of illness was 20 days (range 8–32). Seven patients had been intubated and mechanically ventilated before death. Three patients had requested not to be intubated (information on ventilation was not available for one patient). Ten of the 11 patients were treated with ribavirin; 6 of the 11 patients were treated with steroids.

SARS-CoV was detected in all 22 postmortem lung tissue samples collected from all 11 patients who died with a diagnosis of probable SARS. All 13 postmortem lung samples from the seven non-SARS fatalities that occurred during the SARS outbreak were negative for SARS-CoV, as were all 10 lung-tissue samples collected from patients with lung cancer 5 years before the outbreak (Table 2). The corresponding sensitivity and specificity of the RealArt HPA-Coronavirus LightCycler RT Reagent assay are both 100% (95% confidence interval [CI] for sensitivity 84.6% to 100%; 95% CI for specificity 85.1% to 100%) for the detection of SARS-CoV.

The SARS-CoV viral load in postmortem lung tissue ranged from 2.7 x 10⁴ copies/g tissue to 3.8 x 10⁹ copies/g tissue. Higher viral loads (>10⁶ copies/g tissue) were associated with patients who had a shorter duration of illness (<21 days) (p=0.002, Fisher exact test). The use of ribavirin or steroids was not significantly associated with viral load levels (Table 2).

Twenty-five randomly selected amplicons from the sample pool were sequenced to assess specificity and possible cross-reactivity to other pathogens. A BLAST (available from: URL: http://www.ncbi.nlm.nih.gov/BLAST/) search performed against the SARS-CoV genomes in GenBank, European Molecular Biology Laboratory, DNA Data Bank of Japan, and Protein Data Bank on the National Center for Biotechnology Information Web site (available from: URL: http://www.ncbi.nlm.nih.gov/), indicated that all amplicon samples contained SARS-CoV polymerase gene sequence.

By using a standardized RT-PCR assay, SARS-CoV has been unequivocally identified in the lung tissue of all patients who died with probable SARS but not in any of the controls. These observations support a primary role for this virus in patients with SARS who have fatal outcomes and provide additional, strong evidence to fulfill Koch’s postulates regarding SARS-CoV as the cause of SARS (8). SARS-CoV was found in different lung samples from the same patient, suggesting that the virus is widely disseminated throughout the lung at the time of death. Previous studies suggested that progression of disease to respiratory failure may be primarily mediated by host immune response rather than viral replication (4). Although viral RNA in lung tissue does not necessarily indicate replicating virus, virus in multiple lung lobes, often in high copy number, at the time of death suggests that SARS-CoV may also be contributing to disease progression. The fact that higher viral loads were significantly associated with patients with a shorter duration from onset of illness to death supports the role of viral replication as a contributor to death. Ten of the 11 patients had received therapy with ribavirin, and 6 patients were treated with steroids. The failure to eradicate SARS-CoV despite ribavirin therapy and the lack of association between the use of ribavirin and SARS-CoV viral load are consistent with in vitro data showing that ribavirin has no activity against this agent (9).

Global efforts to contain SARS have been severely impeded by the lack of a standardized, sensitive, and specific diagnostic test for SARS-CoV. Different diagnostic strategies, including culture, serologic assays, and molecular detection methods, have been described, but each of these tests has limitations. In-house RT-PCR assays have been associated with sensitivities as low as 50% in patients with SARS (10), which raises uncertainty about the role of CoV versus co-pathogens in mediating severe or fatal SARS. By contrast, the sensitivity and specificity of the RealArt HPA-Coronavirus RT-PCR assay for detecting CoV in lung tissue samples appear to be excellent. In addition, with the real-time LightCycler system, the assay generates quantitative results within 1 hour, which is much shorter than traditional PCR reactions.

The type of specimen tested, the timing of sample collection, (i.e., acute versus convalescent phase) the method of specimen collection, as well as the method of sample preservation may have substantial impact on the results obtained from a diagnostic test. The lower sensitivity of SARS-CoV detection reported by Peiris et al. (10) may be a consequence of these confounding factors. Our study design of examining lung biopsies from clearly defined patient populations overcame confounding issues, such as sampling technique, nonspecific case definitions, and possible undocumented exposure to SARS. Given the predominance of respiratory symptoms in patients with SARS, lung samples have perhaps the highest viral titers of all specimen types; yet in nonfatal cases, obtaining routine lung biopsies is not practical. Other respiratory tract specimens may be satisfactory substitutes for biopsies, but further studies examining the prevalence of SARS-CoV in these other specimen types and in a larger population are needed. With the use of standardized commercially available assays, comparison of results from different centers may be facilitated.