In April 2003, the first-generation diagnostic tests for the SARS-CoV became available to clinicians caring for patients in whom SARS was considered in the differential diagnosis. Normally, new laboratory diagnostic tests are extensively evaluated and validated before they are introduced in routine clinical practice. However, in the case of SARS, a new and poorly understood disease, these first-generation test results were provided to clinicians on the understanding that the tests had not been validated and results had to be interpreted with caution.
Continued improvement of the sensitivity of RT-PCR methods (12
) makes an analysis of the sensitivity of these first-generation diagnostic methods less relevant. However, these results provide useful information on the best specimens for detection of virus at different stages of illness, the tissue tropism of the virus, and the duration of virus excretion.
Culture of SARS-CoV for preparing the virus-infected cell smears and for virus isolation was carried out under BSL3 conditions, but routine clinical specimens were processed in the clinical virology laboratory under BSL2 conditions after enhanced and reinforced education on safety and good laboratory practice. Given that up to 250 specimens per day were being processed for RT-PCR detection and serologic testing during peak periods, the workload could not be managed in a BSL3 laboratory. None of the laboratory staff became ill with SARS symptoms, indicating that clinical specimens for serologic testing and RT-PCR can be processed safely in BSL2 level conditions.
The association of SARS-CoV with the clinical syndrome of SARS is illustrated by the detection rates of viral RNA in clinical specimens (60% in patients with SARS, 0.6% in the non-SARS group, and 0.3% of controls). Viral RNA detection by these first-generation RT-PCR tests is less sensitive than serologic testing for diagnosing SARS. Correspondingly, 92% of 417 patients with clinically diagnosed SARS and none of the paired sera from 45 unrelated controls seroconverted to SARS-CoV. However, 2 of 25 patients designated as “not SARS” category from whom paired sera were available also seroconverted. Paired sera were available from only a few (25 of 379) patients in the “not SARS” group. At a time of intense pressure on the clinical front-line staff, there was little incentive to obtain convalescent-phase sera from patients believed not to have SARS. These 25 patients may represent a biased sample of the larger group of non-SARS patients. This contention is supported by the fact that a further 47 convalescent-phase sera subsequently obtained from this group of “not SARS” patients failed to show any additional antibodies to SARS. Even patients in the “not SARS” category had a febrile, respiratory, often pneumonic, illness; one of the two patients in the “not SARS” category who had evidence of seroconversion had an undiagnosed pneumonic illness, while the other had an undiagnosed febrile illness without radiologic consolidation of the lung. Overall, a clinical diagnosis of SARS is closely correlated with detection of viral RNA by RT-PCR and seroconversion supporting the etiologic association of SARS-CoV and SARS.
None of 2,400 blood donor sera collected in Hong Kong during the height of the SARS outbreak has any evidence of antibody to the virus. This finding suggests that the spread of SARS-CoV infection in the general community was minimal, with most of the infection associated with clusters and hospital outbreaks (13
The RT-PCR detection rates for SARS-CoV in respiratory, stool, and urine specimens in the 383 patients with seroconversion to SARS-CoV show that viral shedding progressively increased from onset of the illness until approximately day 11 after onset. Since the first-generation RT-PCR test has relatively low sensitivity, these results reflect the increasing viral load at different clinical sites during the illness. Whereas these data are cross-sectional, in a previous study viral load in nasopharyngeal aspirates was followed up longitudinally in nasopharyngeal specimens collected at days 5, 10, and 15 after illness onset; results of this study also indicated that viral load peaks at day 10 of illness (4
). Such a profile of a progressive increase in viral load is unusual for respiratory viral infections. Most other infections (e.g., respiratory syncytial virus, influenza) have peak viral titers in the respiratory secretions at or soon after the onset of clinical illness, after which viral titers and laboratory diagnostic yield decrease progressively (14
). This “crescendo” pattern in SARS-CoV detection rates and viral load in clinical specimens has a number of implications. The pattern explains the poor sensitivity of the first-generation diagnostic tests during the first 5 days of the illness and emphasizes the challenge in making laboratory diagnosis early in the disease. These results may also suggest a fundamental difference in the efficacy of the innate immune response in controlling SARS-CoV infection, in contrast, for example, with influenza infection. Innate immune mechanisms are the earliest host defenses that control viral replication and, in the case of many respiratory viruses, do so within the first few days of illness, even before the specific adaptive immune responses have been activated. This response does not appear to occur with SARS, and viral load in the respiratory tract (4
) begins to fall only when the antibody response appears, i.e., at approximately day 10 after onset of illness (4,5
). This finding may suggest that SARS-CoV is able to evade the host innate response and requires the adaptive immune response to bring the infection under control. Finally, the peak viral load in the 2nd week of illness would predict that virus is more likely to be transmitted later in the course of the illness. This result indeed accords with epidemiologic observations (15
). With regard to observations of viral load, the frequent use of steroid therapy in hospitals (16
) is a confounding factor that may contribute to the increase in virus load later in the illness.
The relative virus detection rates from different specimens during the illness suggests that respiratory specimens (nasopharyngeal aspirate, throat swab) are more useful in the first 4 days of the illness, while fecal samples are better later in the illness. Urine samples, on the other hand, are not useful at any stage of the illness. A productive cough is not common in the early stage of illness, but in patients who do produce sputum, this specimen provides a high diagnostic yield. Thus, nasopharyngeal aspirates, throat swabs, and sputum, if available, are the best specimens in the first 5 days of the illness.
Detecting virus in the fecal and urine samples, in addition to the respiratory tract, suggests that SARS is not restricted to the respiratory tract. The finding of diarrhea unrelated to antimicrobial drug use in a number of patients supports evidence that the disease is not a purely respiratory one (4
). A number of animal coronaviruses (e.g., mouse hepatitis virus and feline coronavirus) have tropism for multiple organs (17
). Viral shedding is detectable by RT-PCR in the respiratory, gastrointestinal, and urinary tracts for many weeks after onset of illness, reflecting continued virus replication at these sites. However, SARS-CoV cannot be readily cultured from any of these sites after week 3 of illness. The viral RNA detected by RT-PCR after week 3 of illness is unlikely to represent persistence of viral RNA in the absence of ongoing viral replication. The apparent dissociation between virus isolation and RT-PCR may reflect the mucosal antibody’s neutralizing the virus and rendering it less infectious. This observation also accords with the apparent absence of transmission of infection after week 2 of illness. The fact that virus isolation was done retrospectively may have affected the overall isolation rate. However, SARS-CoV appears relatively stable to freezing and thawing and is stable for many weeks in clinical specimens at 4°C or frozen at –70°C (K.H. Chan and J.S.M. Peiris, unpub. data). In any event, such a bias would be expected to be uniform both early and late in the disease.
In summary, SARS is closely associated epidemiologically with the novel SARS-CoV. The unusual profile of viral shedding from the respiratory tract may explain some of the observed transmission pattern of this disease, including the predilection for affecting healthcare workers.