Prevalence of non-pneumonic infections with SARS-correlated virus.

Sir

Patrick Woo and colleagues (Mar 13, p 841) suggest that non-pneumonic severe acute respiratory syndrome (SARS)-virus infections are more common than SARS-virus pneumonia in Hong Kong on the basis of a serological survey of about 1000 non-pneumonic healthy individuals of different groups and periods. We query the validity of the authors' experimental approach, which involved western blots only, on the confirmation of the ELISA positive samples.

Woo and colleagues used SARS-virus nucleocapsid protein expressed in Escherichia coli as the coating antigen in their diagnostic ELISA, and the positive samples were confirmed by two “independent” western blots against E coli-expressed nucleocapsid and spike polypeptides. These confirmatory assays were aimed at eliminating the false positives potentially caused by the cross-reactivity between antibodies against the nucleocapsid proteins of other human coronaviruses and those of SARS virus. However, the authors overlooked the possibility of an interaction between residual E coli antigens and naturally occurring antibodies against E coli in the hosts as another potential source of false positives. They did not discuss the purity of the purified E coli-expressed antigens, which is a crucial factor in the interaction because a tiny amount of residual E coli antigen is sufficient to interfere with the diagnostic assay. Moreover, the presence of naturally occurring antibodies against E coli in serum from healthy individuals has been reported previously.2, 3, 4 Such antibodies are mainly present because of infection with common E coli strains and the natural intestinal flora in healthy individuals.

Woo and colleagues' confirmation of the four positive samples by use of two western-blot assays is not persuasive. First, the authors did not present the corresponding western-blot results to show the binding specificity—ie, a single prominent band of the target antigen—which is needed to exclude the possibility of cross-reactivity between the host's antibodies and the residual E coli antigens. Second, the molecular size of both target antigens (ie, spike and nucleocapsid) are virtually the same (about 50 kDa), which could create a problem with respect to the validity of the diagnostic assay if contaminating E coli antigens of a similar size are present. In our opinion, Woo and colleagues should use other independent diagnostic systems to control for various sources of false positives—eg, western-blot assays with antigens of significantly different molecular sizes or antigens expressed in another system, ELISA with viral cell-culture extract as the coating antigen, or indirect immunofluorescence.

A similar study was done by the Center for Emerging Infectious Disease of the Chinese University of Hong Kong. The seroprevalence of asymptomatic, or non-pneumonic, SARS-virus infection in that study was 0·009% (one in 12 000), which is significantly lower than that reported by Woo and colleagues (0·48%). The differences between these two studies are unknown, but bearing in mind their importance with respect to the possible human reservoir for SARS-virus infection in Hong Kong, such serological studies must be carefully designed to eliminate false positives.