Time course of the sensitivity and specificity of serum anti-SARS-CoV-2 IgM and IgG antibodies for the diagnosis of symptomatic COVID-19 in Japan


 The accurate and prompt diagnosis of SARS-CoV-2 infection is required for the control and treatment of the coronavirus infection disease 2019 (COVID-19). In this study, we aimed to investigate the time courses of the anti-severe acute corona respiratory syndrome coronavirus 2 (SARS-CoV-2) IgM and IgG titers and to evaluate the sensitivity and specificity of such tests according to the specific day after the onset of COVID-19 among a patient population in Japan. We measured the titers of SARS-CoV-2 IgM and IgG in sera from 100 subjects, including 26 symptomatic COVID-19 patients, using chemiluminescent immunoassay (CLIA) methods utilizing magnetic beads coated with SARS-CoV-2 nucleocapsid protein and spike protein. The results of a ROC analysis suggested the possibility that the cutoff values in Japan might be lower than the manufacturer’s reported cutoff (10 AU/mL): 1 AU/mL for IgM and 5 AU/mL for IgG. The sensitivity of the test before Day 8 after symptom onset was less than 50%; at Days 9-10, however, we obtained a much higher sensitivity of 81.8% for both IgM and IgG. At 15 days or later after symptom onset, the SARS-CoV-2 IgG test had a sensitivity of 100%. These results suggest that if the number of days since disease onset is taken into consideration, these antibody tests could be very useful for the diagnosis of COVID-19 and similar diseases.


Introduction
Since December 2019, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread throughout the world. The accurate and prompt diagnosis of SARS-CoV-2 infection is required for the control of the COVID-19 pandemic as well as the initiation of adequate individual treatment. In clinical practice, the detection of SARS-CoV-2 RNA using a reverse transcriptase-polymerase chain reaction (RT-PCR) and nasopharyngeal swabs, saliva, or other upper respiratory tract specimens is considered to be the standard method for the diagnosis of COVID-19 1 . Although the speci city of RT-PCR is relatively high, issues persist surrounding insu cient sensitivity (mainly because of the sampling technique), the timing of sample collection, and the types of specimens. Wang et al. reported that sensitivities of RT-PCR for bronchoalveolar lavage specimens, sputum, nasal swabs, brobronchoscope brush biopsy, pharyngeal swabs, feces, blood, and urine were 93%,72%, 63%, 46%, 32%, 29%, 1%, and 0% in a study of 205 COVID-19 patients 2 , and Pasomsub et al. reported that the sensitivity of RT-PCR using saliva samples was 84.2%, compared with RT-PCR using nasopharyngeal and throat swabs 3 . The insu cient sensitivity of the RT-PCR method especially for nasopharyngeal swabs and saliva, which are commonly used specimens, sometimes produces false-negative results. Moreover, the sampling of respiratory and nasopharyngeal specimens can cause secondary infection.
Another method for detecting COVID-19 infection is the measurement of serum titers of virus-speci c antibodies to SARS-CoV-2. Compared with the sampling methods required for RT-PCR, this serological assay reduces the risk of aerosol exposure, making it safer for medical staff. Therefore, anti-SARS-CoV-2 IgM and IgG antibodies have been expected to be useful as complementary tests, in addition to RT-PCR, for the diagnosis of COVID-19. However, a limitation of antibody tests is that they require a longer window period after infection than RT-PCR. Actually, the median IgM and IgG seroconversion period has been reported to be 10 days or longer after symptom onset 4-6 . However, information on the early time courses of serum antibody titers and the sensitivity and speci city of antibody tests after the onset of COVID-19 symptoms could make these tests more useful for the COVID-19.
In this study, we aimed to investigate the time courses of the anti-SARS-CoV-2 IgM and IgG titers as measured using chemiluminescent immunoassays (CLIA) and to determine the sensitivity and speci city of these tests according to the speci c day after the onset of COVID-19 symptoms in Japanese subjects.

Titers of SARS-CoV-2 IgM and IgG
The time courses for the serum titers of SARS-CoV-2 IgM and IgG at each time point (number of days after symptom onset) are shown in Figure 1. Only data for which information regarding the timing of the sampling (number of days after symptom onset) were available were analyzed. As a result, the titers of SARS-CoV-2 IgM and SARS-CoV-2 IgG were analyzed in the 125 sera collected from RT-PCR-positive patients (n = 26) and in the 68 sera collected from RT-PCR-negative patients (n = 34).

ROC analysis and Rede nition of the cutoff values
We performed the ROC analyses at each time point after symptom onset to determine the cutoff values ( Fig. 2 and Table 2). When the ROC analyses were performed using all samples we collected (n=330; RT-PCR-positive n=186, RT-PCR-negative n=144), the AUC of IgM and IgG were 0.948 and 0.952, respectively. The rede ned cutoff values of both of IgM and IgG were far below the manufacturer's cutoff value, that's 0.86 AU/mL and 4.97 AU/mL, respectively. When we calculated the sensitivity and speci city at these cutoff values, the sensitivity and speci city of IgM were 93.0 % and 86.8%, respectively. Those of IgG were 88.2 % and 100 %, respectively.
Next, we analyzed the data only at speci c points after symptom onset. the cutoff values of IgM test varied between 0.86 and 1.48 AU/mL. Regarding IgG test, the AUC reached 1.00 (sensitivity:100%, speci city:100%) after Day15, and the cutoff values were equal to the border values between RT-PCR-positive and RT-PCR-negative results (Fig. 2, Table 2). These results suggested the possibility that the cutoff values could be lower than 10 AU/mL in the Japanese population: 1 or 2 AU/mL for IgM and 5 AU/mL for IgG.

Changes in diagnostic performance after symptom onset
We calculated the sensitivity and speci city at each cutoff value (1-10 AU/mL). Figure 3 (A, B) shows the changes in sensitivity and speci city when the cutoff values were changed for samples obtained 9-10 days after symptom onset.
Especially for IgM, when compared with a cutoff of 5-10 AU/mL, the sensitivity increased up to about 80% at a cutoff of 1 AU/mL. Next, we set the cutoff value for IgM as 1 AU/mL, 2 AU/mL, or 10 AU/mL and that for IgG as 5 AU/mL or 10 AU/mL and compared the sensitivities and speci cities among the time points after symptom onset. For an IgM cutoff value of 1 AU/mL (Fig. 3F), the sensitivity at 9-10 days after symptom onset (81.8%) was much higher than that at 7-8 days after symptom onset (33.3%). When the IgM cutoff value was set as 2 AU/mL, the sensitivity at 9-10 days after symptom onset decreased to 54.5% (Fig. 3E). The Youden Index at cutoff values of 1 AU/mL and 2 AU/mL were 0.71 and 0.55, respectively. For an IgG cutoff value of 5 AU/mL, the sensitivity at 9-10 days after symptom onset was 81.8% (Fig. 3D). When the manufacturer's cutoff value for IgG, 10 AU/mL, was used, the sensitivity at the same time point decreased to 63.6% (Fig. 3C). These results suggest that the serum titers of IgM and IgG after 9 days after symptom onset might have a su cient sensitivity and speci city if the cutoff values are rede ned using the Youden index for the Japanese population.

Discussion
In this study, we investigated the time courses of the serum anti-SARS-CoV-2 IgM and IgG titers and evaluated the diagnostic test performance at speci c days after the onset of COVID-19. First, we con rmed the time courses of the serum antibodies. The median IgM titer increased till Day 18 after symptom onset and then declined (Fig. 1A), while that of IgG remained stable at above 400 AU/mL after Day 13 (Fig.1B). Similar dynamic changes were also reported by another study 7 . The timing when the median titers of IgM and IgG exceeded the manufacturer's cutoff value (10 AU/mL) were Day13-14 and Day9-10, respectively, which might beare similar to those for severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1) 8 and slightly earlier than those for Middle East respiratory syndrome coronavirus (MERS-CoV) 9,10 . In 14 of the 26 RT-PCR-positive patients, the seroconversion of IgG occurred earlier than that of IgM, while the seroconversion of IgM occurred earlier in one case and synchronous seroconversions were observed in 3 cases. In 8 cases, we could not determine the seroconversion timing, since both the IgM and IgG titers had already increased above 10 AU/mL even in the earliest available sample. In the usual immunological response to pathogens, IgM is expected to be produced earlier than IgG. Concordant with the present report, Long et al. reported ten subjects in whom the seroconversion of IgG occurred earlier than that of IgM 11 , as has also been reported in other studies conducted in China [12][13][14] . Although the exact underlying mechanism remains unclear, the earlier seroconversion of IgG might be due to the cross-reactivity of other coronavirus-speci c antibodies to the target antigen used in the the present study varied between 0.86 and 1.48 AU/mL over the days since onset, while the cutoff values for IgG were below 5 AU/mL. Consequently, the cutoff values in the Japanese population could be lower (1 or 2 AU/mL for IgM and 5 AU/mL for IgG) than the manufacturer's cutoff value (10 AU/mL). In previous studies in which the same CLIA-based assays were used, the false-positive rates for the IgM and IgG tests were reported to be 0% (0/33) and 9.1% (3/33) in an Italian study 7 and 9.1% (4/44) and 0% (0/44) in a Chinese study 16 , respectively, which are somewhat higher than those in the present study (2.9 % [1/34] and 0 % [0/34]). One of the mechanisms for false-positive results might be the presence of autoimmune diseases, and the prevalences of autoimmune diseases differ among population. For example, the prevalence of SLE, which reportedly causes false-positive cases, in Italy, China and Japan are reportedly 40-81, 10-30 17 and 3.7-19.1 18 per 100,000 of the population, respectively.
Regarding the time course of sensitivity, the diagnostic ability of antibodies, especially IgG, increases as time passes from the onset of symptoms. Zhao et al. reported that RT-PCR had a higher sensitivity (66.7%) than IgM (28.7%) and IgG (19.1%) within 8 days from the onset of symptoms, while the sensitivity of RT-PCR decreased to 54.0% after Day 8 and that of the antibody tests increased 4 . Concordantly, the diagnostic performances of both IgM and IgG were insu cient at the time points before Day7-8 (Fig.3 C-F), while at Day9-10 we obtained much higher sensitivity when set on lower cutoff values; For IgM, the sensitivity at the cutoff value of 10 AU/mL, 2 AU/mL and 1 AU/mL were 36.4 %, 54.5 % and 81.8 %, respectively and for IgG, the sensitivity at the cutoff value of 10 AU/mL and 5 AU/mL were 63.6 % and 81.8 %, respectively. These results suggested that setting the optimal cutoff value for Japanese population could make the serum antibody test more useful tool for diagnosis of COVID-19 infection, although further studies are necessary to prove this hypothesis.
It was of note the IgG titers of all the sera collected from RT-PCR-positive patients (n = 26) were above 10 AU/mL at 15 days or later after symptom onset (Fig. 1B), and the sensitivity was 100%. Several subjects with non-COVID-19 diseases, such as pneumonia caused by other pathogens or collagen diseases, can show similar clinical course and CT results, which sometimes makes a differential diagnosis di cult. The high speci city of IgG after Day 15 could help physicians to rule out COVID-19.
A limitation of this study is that only symptomatic patients were examined. Recently, the asymptomatic COVID-19 subjects have been reported to have antibody titers below the cutoff values 19 . According to a recent report, T cell immunity is su cient to exclude SARS-CoV-2 in asymptomatic patients 20 . Therefore, in respect to infection control, which requires the screening of asymptomatic subjects, an antibody test alone might be insu cient, and combination with a PCR-based test is desirable. Nevertheless, antibody tests could help physicians to diagnose COVID-19 or non-COVID-19 presenting with similar clinical courses among symptomatic subjects.
In summary, we investigated the time courses of the diagnostic performances of SARS-CoV-2 antibody tests in Japan and observed a high sensitivity after Days 9-10. Considering the high speci city of IgG levels at 15 days after symptom onset, the IgG test could be a very useful diagnostic tool for ruling out the possibility of COVID-19.

Samples
We collected residual serum samples remaining after routine clinical testing from 100 subjects who underwent RT-PCR testing at The University of Tokyo Hospital. Of these 100 subjects, 26 were diagnosed as having COVID-19 based on the results of RT-PCR, and the remaining 74 subjects were considered to have negative results. The characteristics of the subjects are described in Table 1. Serum specimens were stored at -80°C and were centrifuged at 2,300 xg for 5 minutes before measurement. The current study was performed in accordance with the ethical guidelines of the Declaration of Helsinki. Participants were informed about the study and informed consent was obtained in the form of an opt-out on the website. Patients who rejected to be enrolled in our study were excluded. The study design was approved by The University of Tokyo Medical Research Center Ethics Committee, which waived written informed consent because archived specimens were used and data in this retrospective study were retrieved from medical records (2019300NI-3).

Methods
The serum titers of SARS-CoV-2 IgM and IgG were measured using the iFlash3000 fully automatic CLIA analyzer from YHLO Biotechnology Company, Ltd. (Shenzhen, China). We used SARS-CoV-2 IgM and SARS-CoV-2 IgG kits containing magnetic beads coated with SARS-CoV-2 nucleocapsid protein and spike protein. The performance of this assay has been validated at 10 hospitals in China 21 . According to the manufacturer's insert, the cutoff value for the detection of both IgM and IgG is 10 AU/mL.
The area under the curve (AUC) was expressed as the median and 95% con dence interval (CI). For each ROC curve, we rede ned the cutoff values so that the Youden Index (= Sensitivity + Speci city -1) was at a maximum.     Diagnostic abilities of SARS-CoV-2 antibody tests according to cutoff values and time courses. We compared the sensitivities and speci cities for the serum diagnosis of COVID-19 each cut off value (1 AU/mL -10 AU/mL) at 9-10 days after symptom onset (A: SARS-CoV-2 IgM, B: SARS-CoV-2 IgG) and investigated the time course of the diagnostic ability when the cut off values were set as 1, 2, 5 and 10 AU/mL (C: 10 AU/mL, D: 5 AU/mL, E: 2 AU/mL, F: 1 AU/mL).