One-Year Follow-Up of COVID-19 Patients Indicates Substantial Assay-Dependent Differences in the Kinetics of SARS-CoV-2 Antibodies

Abstract

Determination of antibody levels against the nucleocapsid (N) and spike (S) proteins of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) are used to estimate the humoral immune response after SARS-CoV-2 infection or vaccination. Differences in the design and specification of antibody assays challenge the interpretation of test results, and comparative studies are often limited to single time points per patient. We determined the longitudinal kinetics of antibody levels of 145 unvaccinated coronavirus disease 2019 (COVID-19) patients at four visits over 1 year upon convalescence using 8 commercial SARS-CoV-2 antibody assays (from Abbott, DiaSorin, Roche, Siemens, and Technoclone), as well as a virus neutralization test (VNT). A linear regression model was used to investigate whether antibody results obtained in the first 6 months after disease onset could predict the VNT results at 12 months. Spike protein-specific antibody tests showed good correlation to the VNT at individual time points (r_S, 0.74 to 0.92). While longitudinal assay comparison with the Roche Elecsys anti-SARS-CoV-2 S test showed almost constant antibody concentrations over 12 months, the VNT and all other tests indicated a decline in serum antibody levels (median decrease to 14% to 36% of baseline). The antibody level at 3 months was the best predictor of the VNT results at 12 months after disease onset. The current standardization to a WHO calibrator for normalization to binding antibody units (BAU) is not sufficient for the harmonization of SARS-CoV-2 antibody tests. Assay-specific differences in absolute values and trends over time need to be considered when interpreting the course of antibody levels in patients. IMPORTANCE Determination of antibodies against SARS-CoV-2 will play an important role in detecting a sufficient immune response. Although all the manufacturers expressed antibody levels in binding antibody units per milliliter, thus suggesting comparable results, we found discrepant behavior between the eight investigated assays when we followed the antibody levels in a cohort of 145 convalescent patients over 1 year. While one assay yielded constant antibody levels, the others showed decreasing antibody levels to a varying extent. Therefore, the comparability of the assays must be improved regarding the long-term kinetics of antibody levels. This is a prerequisite for establishing reliable antibody level cutoffs for sufficient individual protection against SARS-CoV-2.

Keywords: SARS-CoV-2; antibody kinetics; assay comparison; neutralizing antibodies; predictive modelling.

FIG 1

(a to j) Method comparison plots of all quantitative S protein tests evaluated. In each panel, the x axis (logarithmized) indicates the geometric mean of both methods (BAU/mL), and the y axis (logarithmized) represents the ratio of the two methods. The solid line indicates median ratio; the dashed line at 1.0 represents the equivalence between both methods; dotted lines show the predefined boundaries of acceptance (±30% of equivalence). Color codes: pink, V0; olive green, V1; green, V2; blue, V3. ▴, observation above the upper limit of quantification for at least one of the two measurements; n_neg, number of observations not displayed due to a negative measurement in at least one assay.

FIG 2

Comparison of quantitative antibody tests with a live virus neutralization test. Scatterplot of antibody tests (x axis, BAU/mL) evaluated against the virus neutralization titer (y axis) for all time points and each individual time point (V0, V1, V2, and V3). The Spearman rank correlation coefficient is given in the top left corner of each panel; the dashed line represents the cutoff for positivity—in the case of Roche, a higher cutoff is also stated by the manufacturer for correlation with neutralizing antibodies (dotted line). Pink, V0; olive green, V1; green, V2; blue, V3. ▴, observation above the upper limit of quantification for at least one of the two measurements.

FIG 3

Boxplots of the absolute (V0) and relative (V1 to V3) values of tests evaluated, showing the kinetics of the antibody levels over time. Pink, V0; olive green, V1; green, V2; blue, V3. Patients without detectable antibodies against SARS-CoV-2 at V0 (n = 7) were not considered for the later visits.

FIG 4

Predictability of long-term virus neutralization titer. Observed and predicted neutralization titer 12 months after disease onset, based on a single measurement of antibodies or the neutralization titer 100 days after infection. The y axis is shown on a shifted logarithmic scale in order to include values of zero. Blue areas, 95% confidence limits for the mean model; dashed oblique lines, 95% prediction limit; dashed horizontal/vertical lines, cutoff for positivity for the respective test; ○, samples (neutralization titer at V3 on the y axis; intercept 100 for the respective assay on the x axis).