Analytical performance of MAGLUMI® SARS-CoV-2 Ag tests
The default unit for the analyte detected by the MAGLUMI® SARS-CoV-2 Ag tests was the manufacturer’s arbitrary unit (AU)/mL. A linear relationship between AU/mL and pg/mL was found (R2=0.9976) , and the conversion factor to obtain pg/mL is 1.592 (Figure 1A). The total imprecision of the CLIA quantitative assay was less than 8% (Supplemental Table 1), with the limit of quantitation (LoQ) at 0.399 pg/mL (Figure 1B) and linear range from 0.4 to 3184.00 pg/mL (Figure 1C). No interference was detected in 48 commercial virus forms and 10 common clinical virus specimens (Figure 1D). There is a remarkable linear correlation between the antigen concentration versus the infectious dose for five different SARS-CoV-2 VOCs, respectively (Figure 1E). For detailed results, please refer to the supplementary data.
Characteristics of the subjects
Nasopharyngeal samples from 709 individuals were included in this study. Table 1 reports the demographic characteristics of the study subjects. The majority of the patients had relatively high Ct values (32.76% for 30-35 Ct and 34.48% for 35-40 Ct) corresponding to low viral loads, while 35 of 232 (15.09%) patients had high viral loads with Ct < 25. The median antigen concentration in patients was 16.64 pg/mL (IQR 3.28-285.37) and in healthy individuals was negligible (less than 0.1pg/mL). Overall, the SARS-CoV-2 antigen levels of patients and that of healthy subjects differed significantly (P < 0.001).
Table 1. Demographic, SARS-CoV-2 viral loadsa, antigen concentrations and time from –first positive test for COVID-19 patients and healthy individuals.
Covariateb
|
COVID-19 patients (n=232)
|
Healthy individuals (n=477)
|
P value
|
Age (years)
|
71 (59-83)
|
69 (56-79)
|
0.378
|
Gender (female)
|
116 (50%)
|
233 (48.85%)
|
0.200
|
Antigen concentrationc
|
16.64 (3.28-285.37)
|
< 0.1
|
<0.0001
|
Ct value for ORF1 gene
|
33.015 (28.19-36.07)
|
> 40
|
<0.0001
|
Ct < 25
|
35 (15.09%)
|
0
|
<0.0001
|
25≤Ct< 30
|
39 (16.81%)
|
0
|
<0.0001
|
30≤Ct<35
|
76 (32.76%)
|
0
|
<0.0001
|
35≤Ct<40
|
80 (34.48%)
|
0
|
<0.0001
|
Ct ≥ 40
|
2 (0.86%)
|
477 (100%)
|
<0.0001
|
Days from first positive
test
|
8 (5-12)
|
0
|
<0.0001
|
Symptom-free
|
48 (20.69%)
|
NAd
|
NAd
|
Mild symptom
|
145 (62.5%)
|
NAd
|
NAd
|
Moderate symptom
|
32 (13.79%)
|
NAd
|
NAd
|
Severe symptom
|
7 (3.02%)
|
NAd
|
NAd
|
with comorbidity(ies)
|
136 (58.62%)
|
NAd
|
NAd
|
without comorbidity
|
96 (41.38%)
|
NAd
|
NAd
|
a. SARS-CoV-2 viral loads were estimated via cycle threshold (Ct) of ORF1 gene.
b. Continuous variables reported as median (interquatile range) and categorical variables reported as N (percentage).
c. Antigen concentration is shown in pg/mL.
d. NA, not applicable.
Diagnostic performance
A receiver operating characteristic (ROC) curve was plotted to determine the optimal cutoff value of the SARS-CoV-2 antigen, which allows the distinction of SARS-CoV-2 infection from healthy status (Figure 2A). By comparing the antigen results between SARS-CoV-2 positive patients (Ct value ≤ 40) and healthy individuals, results provided an area under the ROC curve (AUC) of 0.987, with 95% confidence interval (CI) ranging from 0.976 to 0.994. Based on the Youden index calculation, the sensitivity and the specificity of the test reached 95.7% (95% CI: 92.2%-97.9%) and 98.3% (95% CI: 96.7%-99.3%) respectively, when a cutoff of 0.640 pg/mL was used.
The diagnostic performance of the assay was next evaluated by stratifying the results according to Ct values of RT-qPCR (Table 2). The performance of the Wondfo 2019-nCoV Antigen Test (an LFT) was evaluated in parallel (Figure 2B, 2C). Compared with RT-qPCR, the MAG-CLIA SARS-CoV-2 Ag test was positive and showed 100% concordance for samples with Ct < 33. For samples with Ct ranging from 33 to 40, the sensitivity of the test remained over 95% concordance. Contrarily, the sensitivity of the LFT decreased gradually and remarkably when Ct values increased. The LFT showed 87.8% sensitivity in detecting samples with Ct < 30. The sensitivity dropped to 66.4% for samples with Ct < 33, and to only 34.9% for samples with Ct ≤ 40.
Table 2. The sensitivity of two antigen detection assay according to the Ct values from RT-qPCR.
Ct value
|
n
(total)
|
MAGLUMI® SARS-CoV-2 Ag assay
|
Wondfo 2019-nCoV Antigen Test
|
N
(> 0.64 pg/mL)
|
n
(≤ 0.64 pg/mL)
|
accumulated sensitivity (%)
|
n (positive)
|
n (negative)
|
accumulated sensitivity (%)
|
< 30
|
74
|
74
|
0
|
100.0%
|
65
|
9
|
87.8%
|
< 31
|
89
|
89
|
0
|
100.0%
|
70
|
19
|
78.7%
|
< 32
|
101
|
101
|
0
|
100.0%
|
72
|
29
|
71.3%
|
< 33
|
113
|
113
|
0
|
100.0%
|
75
|
38
|
66.4%
|
< 34
|
131
|
130
|
1
|
99.2%
|
77
|
54
|
58.8%
|
< 35
|
150
|
147
|
3
|
98.0%
|
78
|
72
|
52.0%
|
< 36
|
173
|
169
|
4
|
97.7%
|
79
|
94
|
45.7%
|
< 37
|
186
|
179
|
7
|
96.2%
|
79
|
107
|
42.5%
|
< 38
|
206
|
197
|
9
|
95.6%
|
79
|
127
|
38.3%
|
< 39
|
223
|
213
|
10
|
95.5%
|
81
|
142
|
36.3%
|
≤ 40
|
232
|
222
|
10
|
95.7%
|
81
|
151
|
34.9%
|
Correlation between SARS-CoV-2 antigen and Ct values
MAG-CLIA targets for the N-terminal domain of N protein in SARS-CoV-2, which wraps coronavirus RNA through noncovalent bonds to form the nucleocapsid. This should result in a significant positive correlation between viral load and antigen concentration. Therefore, we examined correlations between SARS-CoV-2 antigen concentration and viral load determined by Ct values of RT-qPCR. Indeed, a significant (P < 0.0001) linear inverse relationship between the Ct values and log10 antigen levels was observed (correlation coefficient R2 = 0.747; Figure 3A). It is also notable that individual variations were remarkable.
Individual time lines
To explore the dynamic performance of MAG-CLIA SARS-CoV-2 Ag in the viral load time courses, and to explore the influence of comorbidities on the time kinetics, we performed longitudinal studies in 14 COVID-19 positive participants (Figure 4). Among the 14 participants, 2 of them had bacterial infection (patient 9, 13), 2 with diabetes (patient 7, 10), 3 hypertension (patient 2, 4, 5), 3 nephropathy (patient 6, 8, 14), 1 fungal infection (patient 3) and the other 3 (patient 1, 11, 12) without any underlying disease. Figure 4 shows modelled SARS-CoV-2 viral RNA (as seen by Ct values of ORF1 gene) trajectories together with the viral antigen measured for individuals. As the antigen concentration decreased, a decrease of SARS-CoV-2 viral loads was observed over time, as indicated by the increase of Ct values. This trend was observed for the majority of the participants. Moreover, as exemplified in patients 6, 8, 13 and 14, the fluctuation of SARS-CoV-2 Ag was found to be closely and dynamically consistent with that of the viral nucleic acid loads in the time courses. Notably, the underlying comorbidities remarkably prolonged the time for both viral nucleic acid loads and antigen concentrations to return to negative.
Evaluation of COVID-19 transmission with CLIA quantification
Recent studies observed a strong correlation between SARS-CoV-2 viral loads and transmission [14], and reported no cases of COVID-19 transmission with SARS-CoV-2 viral RNA loads <4 log10 copies/mL [15]. Given that, Log10 viral load was then estimated from the Ct value using the empirical formula 14.543-(Ct*0.3018) for the MA-6000 Real-Time Quantitative Thermal Cycler system, and the viral load of 4 log10 copies/mL was verified to correspond to the 35 Ct value of the current system. This was in accordance with previous viral culture studies which showed that SARS-CoV-2 is no longer contagious for samples with Ct values ≥ 35 [16,17]. The antigen concentration corresponding to 35 cycles was then calculated to be 8.71 pg/mL using the inverse linear regression model shown in Figure 3A. In addition, we also observed a consistent average antigen concentration of 8.82 pg/mL in nasopharyngeal swab samples of another 50 COVID-19 patients on the day when RT-qPCR Ct values first returned to above 35. Thus, we hypothesized that 8.82 pg/mL might be a promising antigen concentration in differentiating contagious patients from the recovering.
To further assess the diagnostic ability of the ultra-sensitive SARS-CoV-2 antigen test with the concentration of 8.82 pg/mL in differentiating contagious patients from the recovering, an ROC curve was then plotted by classifying results into Ct values less than 35 and Ct values over 35 (and including 35) (Figure 3B). The ROC curve in identifying infectious patients showed an AUC of 0.921 (95% CI: 0.890-0.946). When 8.82 pg/mL was selected as the critical value, the sensitivity and the specificity of the test were 84.5% (95% CI: 78.2%-89.5%) and 85.0% (95% CI: 79.6%-89.4%), respectively.
To explore the time nodes of detection, we tracked the RT-qPCR results from over 1,000 COVID-19 cases between April 1st and 10th May 2022 in our laboratory to describe the kinetic evidence of the SARS-CoV-2 Omicron subvariant BA.2 in Shanghai, China (Figure 3C). Overall, the viral load in nasopharynx reduced over time, as seen by the increase of Ct values. The Ct value reached 35 after an average of 9 days since the first observation of positive RT-qPCR test. Furthermore, SARS-CoV-2 was no longer detected in nasopharynx (Ct > 40) after 12 days.