Estimating SARS-CoV-2 seroprevalence in Canadian blood donors, April 2020 to March 2021: improving accuracy with multiple assays

Pillar 4
Functional Genomics & Structure-Function of VOCs

Seroprevalence data can provide key insights about incidence of infection and exposure to pathogens during epidemics. Serological studies have affirmed that an antibody response is correlated with clearance of viral shedding, have helped confirm true infections, and have helped identify populations with elevated risks of infection.  Early in the pandemic, getting an estimate of the true number of infections was difficult.  Testing for active infections was only available to symptomatic individuals and priority groups, so asymptomatic and mild cases that went untested were likely contributing to underestimates of the true levels of SARS-CoV-2 infections.  Antibodies generated in response to an infection may last for weeks, so serology assays are well-suited to enable estimation of the true rate of infection in a population.

The main objective of this study was to estimate seroprevalence of SARS-CoV-2 infection in a low seroprevalence setting, using leftover blood samples from Canadian blood donors collected between April 2020 and March 2021.  Blood donors are a subpopulation that allows rapid, repeated evaluation of seroprevalence during a public health crisis. At the time of publication, multiple commercially available and in-house serology assays were available for high-throughput serosurveillance. Unfortunately, no “gold-standard” serology assay existed because existing assays exhibited imperfect sensitivities and specificities (because false-positives and false-negatives become harder to identify when disease prevalence is low). Previously successful approaches to measure Norovirus seroprevalence in the absence of “gold-standard” detection method relied on performing a combination of assays.  Therefore, the authors asked whether they could likewise combine existing SARS-CoV-2 assays to improve accuracy of seroprevalence measurements.  Four serology assays were used to measure IgG antibodies against SARS-CoV-2.  Two statistical methods to estimate seroprevalence were used, both of which rely on combinations of assays: a latent class analysis (LCA) method and composite reference standards (CRS) method.  The LCA method classifies samples based on their testing outcome(s) and known information about the specificity and sensitivity of the diagnostic tests to predict a probability of true infection.  The CRS method uses constellations of results from tests (for example, a positive result from a highly specific test, combined with positive results from several sensitive but less specific tests) as a substitute for a gold standard assay. Using these approaches, it was estimated that seroprevalence increased from approximately 0.8% in April 2020 to approximately 6.3% in March 2021.  As expected, being vaccinated was strongly associated with seropositivity.  This study provides support for using a combination of serology approaches to identify true seroprevalence during public health crises, epidemics, and pandemics when a “gold-standard” is not available.

Read article