ABSTRACT
Background: Seroprevalence studies of antibodies to SARS-CoV-2 are important for public health surveillance. Recent studies have shown that antibodies to SARS-CoV-2, both from natural infection and vaccination, decrease with time from exposure. Variation in the performance of antibody assays will impact the estimates of SARS-CoV-2 exposure and vaccination levels in a population. Using standardized serial dilutions, we evaluated 17 SARS-CoV-2 assays to establish an approximate limit of detection for each assay. Methods: The evaluated assays consisted of three chemiluminescent immunoassays (CLIAs), eight standard enzyme-linked immunosorbent assays (ELISAs), and six lateral flow assays (LFAs). All assays either evaluated IgG antibodies or total antibodies to SARS-CoV-2. The target antigen of 14 assays was the spike protein (S) or receptor binding domain (RBD);three assays evaluated antibodies to the nucleocapsid protein (N). A human SARS-CoV-2 serology standard with a WHO SARS-CoV-2 Serology International Standard binding antibody units (BAU) value of 764 BAU/mL to spike IgG and 681 BAU/mL to nucleocapsid IgG was obtained from the Frederick National Laboratory for Cancer Research. Half-logarithmic serial dilutions of the standard were then run in triplicate on each assay. Results: The MSD V-Plex chemiluminescent immunoassays (CLIAs) were the most sensitive by three logs, with positive results at a dilution greater than 1:106 (Figure). Standard ELISAs were less sensitive, with limits of detection ranging from dilutions of 1:20 (Euroimmun NeutraLISA) to 1:1620 (Euroimmun SARS-CoV-2 IgG and Euroimmun QuantiVac). Lateral flow assays (LFAs) were the least sensitive, with only one assay (Wondfo Colloidal Gold) having at least one positive result with a dilution greater than 1:180. Conclusion: As population seroprevalence to SARS-CoV-2 continues to rise, tests with a high limit of detection will be crucial for surveillance studies. As antibody levels decline after vaccination or infection, our data indicate that CLIAs like the MSD assay may provide the best opportunity to capture asymptomatic cases and individuals with low antibody titers.
ABSTRACT
Background: Live virus micro-neutralization (MN) is the gold standard for quantifying the neutralizing titer (NT) of antibodies to SARS-CoV-2. However, performing MN is labor intensive and requires a biosafety level 3 laboratory. We assessed the performance of 8 immunoassays which measure SARS-CoV-2 NT and compared them to gold standard MN results. Methods: Samples from 269 individuals known to previously be SARS-CoV-2 PCR+ (i.e., convalescent individuals, <10% hospitalized) and 200 pre-pandemic individuals were evaluated on 3 lateral flow immunoassays (LFAs;Wondfo Colloidal Gold, Wondfo Colored Microsphere, Wondfo Finecare) and 5 enzyme-linked immunoassays (ELISAs;ImmunoRank, GenScript, Cusabio, Euroimmun NeutraLISA, Euroimmun QuantiVac). MN was performed on all samples from convalescent individuals;results were classified as undetectable vs any detection of MN NT (NT<20 vs. NT>20), as well as high and low MN NT (NT>80 vs. NT<80). Receiver operating curve analysis was used to assess accuracy for detecting levels of NT. The area under the curve (AUC) was calculated for the manufacturer's cut off and empirically to identify the best discriminatory cut off value. Cohen's kappa statistics were calculated to assess categorical agreement and Spearman's rank statistics were calculated to assess correlations. Results: Of the 269 convalescent plasma samples, 89 (33%) had MN NT values <20 (undetectable) and 117 (43%) >80 (high NT). Using the manufacturer's cutoffs, sensitivity for detection of samples with any NT ranged from 79% to 100%, and the false-positive rate (ie, classifying samples with undetectable NT as positive) was highest for LFAs (72% to 84%) and ranged from 14% to 69% for the ELISAs. For all assays except the ImmunoRank and NeutraLISA ELISAs, discrimination to identify samples with any NT was improved by raising the cut off values (Table). AUCs of ∼0.94 to discriminate high NT samples could be achieved for all quantifiable assays using an adjusted cut off value. Cohen's kappa statistic ranged from 0.20 to 0.69. Spearman's rank correlation between each assay and NT value ranged from 0.73 to 0.86. Using the manufacturer's cutoffs, specificity on pre-pandemic samples was ≥98% for all assays except for Cusabio which was 86%. Conclusion: The performance of immunoassays using manufacturer's cutoff to discriminate samples with any NT was accurate (AUC>0.83 for all assays), but could be improved by changing the cutoff. Identifying samples with high NT could be achieved using an alternative cutoff.
ABSTRACT
Background: The performance of serological antibody tests to SARS-CoV-2 infection varies widely and little is known about their performance in Africa. We assessed the performance of CoronaCHEK Lateral Flow Point of Care Tests on samples from Rakai, Uganda and Baltimore, Maryland, USA. Methods: Samples from subjects known to be SARS-CoV-2 PCR+ (Uganda: 50 samples from 50 individuals, and Baltimore: 266 samples from 38 individuals) and samples from pre-pandemic individuals collected prior to 2019 (Uganda: 1077 samples, Baltimore: 580 samples) were analyzed with the CoronaCHEK assay per manufacturers protocol. Sensitivity by duration of infection and specificity among pre-pandemic samples were assessed for the IgM and IgG bands separately and for any reactivity. Poisson regression models were used to calculate prevalence ratios (PR) for factors associated with a false-positive test among pre-pandemic samples. Results: In Baltimore samples, sensitivity for any reactivity increased with duration of infection with 39% (95% CI 30, 49) during 0-7 days since first positive PCR, 86% (95% CI 79, 92) for 8-14 days, and 100% (95% CI 89,100) after 15 days (See Figure). In Uganda, sensitivity was 100% (95% CI 61,100) during 0-7 days, 75% (95%CI 53, 89) for 8-14 days, and 87% (95%CI 55, 97) after 14 days since first positive PCR. Specificity results among pre-pandemic samples from Uganda was 96.5% (95% CI 97.5, 95.2), significantly lower than the 99.3% (95% CI 98.2, 99.8) observed in samples from Baltimore (p<0.01). In Ugandan samples, individuals with a false positive result were more likely to have had a fever more than a month prior to sample acquisition (PR 2.9, 95% CI 1.1, 7.0). Conclusion: Sensitivity of the CoronaCHEK appeared to be significantly higher in Ugandan samples from individuals within their first week of infection compared to their Baltimorean counterparts. By the second week of infection the sensitivity appeared the same between geographic areas. The specificity was significantly lower in Ugandan samples than those from Baltimore. False positive results from pre-pandemic Uganda appear to be correlated with the convalescent disease state, potentially indicative of a highly cross-reactive immune response in these individuals from East Africa.