Factors influencing SARS-CoV-2 IgG test sensitivity: A Bayesian analysis of seroconversion and seroreversion by time since infection, test, age and disease severity

Abstract

Background Antibody testing is commonly used to assess past exposure to pathogens, but the interpretation is complex. We quantified test-specific SARS-CoV-2 seroconversion and seroreversion by time since PCR-confirmed infection, age and disease severity.Methods We combined Belgian data from laboratory SARS-CoV-2 testing, prescriptions, contact tracing and hospital surveillance collected between March 2020 and June 2021 with data from published longitudinal studies on Wantai and EuroImmun IgG serological tests. We used a hierarchical Bayesian model to estimate time-varying sensitivity of serological tests following PCR-confirmed infection. The model employed a scaled Weibull-bi-exponential distribution. We accounted for disease severity (distinguishing between asymptomatic, symptomatic, and hospitalized cases), age (i.e., age groups 18-49, 50-64, and 65-74 years) and serological test used.Methods We combined Belgian data from laboratory SARS-CoV-2 testing, prescriptions, contact tracing and hospital surveillance collected between March 2020 and June 2021 with data from published longitudinal studies on Wantai and EuroImmun IgG serological tests. We used a hierarchical Bayesian model to estimate time-varying sensitivity of serological tests following PCR-confirmed infection. The model employed a scaled Weibull-bi-exponential distribution. We accounted for disease severity (distinguishing between asymptomatic, symptomatic, and hospitalized cases), age (i.e., age groups 18-49, 50-64, and 65-74 years) and serological test used.Results We included 44,262 serological test results: 10,864 obtained from published studies, 33,398 from Belgian laboratories. Seroconversion occurred during the six weeks following a PCR-confirmed infection. Age, disease severity and the test used strongly influenced seroconversion rates and the rate of the subsequent seroreversion. For the EuroImmun test, 82% (95% Credible Interval (CrI): 80%-84%) of symptomatic individuals in the youngest age group seroconverted, compared to 95% (CrI: 95%-96%) for the Wantai test. Seroconversion was associated with hospitalization, (OR = 8.17 (CrI: 5.56-13.72), compared to asymptomatic infection) and older age (OR = 1.65 (CrI: 1.41-1.97), compared to 18-49 year-olds). Slower seroreversion was associated with older age, hospitalization and the Wantai test. At 50 weeks, seropositivity among symptomatic 18-49 year-olds was 64% (CrI: 58%-70%) for the EuroImmun test and 95% (CrI: 94%-96%) for the Wantai test.Conclusion These findings highlight the need for test-specific, time-varying sensitivity adjustments in seroprevalence studies. Such adjustments are crucial for translating seroprevalence results to cumulative incidence estimates.