Detection of SARS-CoV-2 antibodies is insucient for the diagnosis of active or prior COVID-19

We assessed the performance of Abbott's SARS-CoV-2 IgG assay and the PanbioTM COVID-19 IgG/IgM rapid test device for the diagnosis of active and past/cured COVID-19. Three cohorts of patients were chosen. Cohort 1, patients (n=65) who attended the emergency department on March 30, 2020 with clinical suspicion of active COVID-19 (n=56 with proven/probable COVID-19). Cohort 2, hospital workers (n=92) who had either been (n=40) or not (n=52) diagnosed with proven/probable COVID-19 and were asymptomatic at the time of the sampling. Cohort 3, patients (n=38) cared at the hospital before the start of the COVID pandemic. Detection of serum antibodies was done using Abbott´s SARS-CoV-2 IgG assay and the PanbioTM COVID-19 IgG/IgM device. Both methods showed 98% agreement for IgG detection. No antibodies were detected in the 38 samples from hospitalized pre-COVID subjects. The diagnostic performance of IgGs detected by Abbott´s SARS-CoV-2 assay in Cohorts 1/2 was: sensitivity (60.7%/75%) and specicity (100%/84.6%). The diagnostic performance of IgM by PanbioTM COVID-19 in Cohorts 1/2 was: sensitivity (16%/17.5%) and specicity (100%/98.1%). We show that IgG detection alone is insucient for the diagnosis of active or past COVID-19. IgM detection has a limited diagnostic value.


Introduction
The COVID-19 epidemic, which started in late December 2019 in China, rapidly spread to many countries affecting millions of people worldwide [1,2]. The clinical presentation may range from asymptomatic or mild illness to severe pneumonia, acute respiratory distress syndrome, and progress to severe and fatal respiratory failure and death. Symptoms usually emerge 2-14 days after viral exposure and COVID-19positive patients may experience fever, cough, fatigue, muscle pain, and shortness of breath, among the most commonly reported clinical manifestations [3,4].
To date, the diagnosis of COVID-19 is mainly based on the detection of SARS-CoV-2 virus RNA by PCR from samples of patients under clinical suspicion. However, PCR sensitivity is 80% at best, with lower values in early and late stages of the infection [5]. Furthermore, and given the large number of infected patients who probably have not been tested or reported as having PCR false-negative results, detection of viral RNA is insu cient to assess the current number of people affected by COVID-19 [6].
The detection of IgG antibodies against SARS-CoV-2 is possible in serum samples of patients with active COVID-19 in the rst week after the onset of the infection [7]. This potentially allows to improve the microbiological diagnosis along with RNA detection from nasopharyngeal swabs in active cases and also detect past/cured cases [8]. The serological detection of IgG against SARS-CoV-2, mostly using inhouse procedures, has shown a sensitivity that ranges between 78% and 100% in patients requiring hospital admission [7][8][9][10][11][12][13][14][15]. Abbott´s SARS-CoV-2 IgG assay is a recent commercial chemiluminescence immunoassay (CLIA) platform approved by the FDA that allow detecting IgG in serum samples and has been validated in samples from hospitalized patients with COVID-19 [16]. On the other hand, the PanbioTM COVID-19 IgG/IgM rapid test is a point-of-care lateral ow device that may be a good alternative to CLIA-based procedures. However, the role of IgG detection for the diagnosis of past/cured COVID-19 remains an enigma, and, to the best of our knowledge a comparison of the two diagnostic platforms has not been done yet.
We will also assess the diagnostic performance of the IgG and IgM detected for the diagnosis of active and past/cured COVID-19.

Materials And Methods
Hospital setting. The study was conducted at Gregorio Marañón hospital, a 1,550-bed large tertiary Clinical data. The following demographic and clinical data were collected: age, sex, presence of symptoms compatible with COVID-19 (fever >38 °C, headache, cough, asthenia, myalgia, sore throat, runny nose, dyspnoea, anosmia, dysgeusia, abdominal pain, and diarrhoea), date of symptom onset, hospital admission, radiological images, immunosuppression, and outcome (favourable or death during admission). Clinical data were collected in a pre-established protocol that was either retrospectively lled out retrieving the information from electronic charts for Cohort 1 patients or lled out at sample collection for Cohort 2 patients.
Patients with clinical signs and symptoms of COVID-19 in the epidemic scenario were classi ed considering PCR results on nasopharyngeal samples and radiological ndings as either having a proven (positive PCR result) or probable disease (negative PCR but a suggestive radiological pattern) [17]. Asymptomatic subjects by the time of sample collection were considered as not infected (if PCR proved to be negative) or asymptomatic carriers (if PCR proved to be positive).
Data analysis. Categorical variables were described and compared using Chi-square or Fisher's exact tests and continuous variables were compared using the t-student and the Mann-Whitney U tests (IBM SPSS Statistics, version 26; Armonk, NY, USA).
Samples from patients in Cohort 3 were used to rule out antibody detection cross-reactivity. Sensitivity, speci city, positive predictive value, and negative predictive value were calculated separately for Cohorts 1 and 2. Independent analyses were carried out considering or excluding probable cases at different stages of the disease (less or more than seven days from clinical disease onset) for patients in Cohort 1; no further analyses of diagnostic performances were done for subjects in Cohort 2 given that only one patient had probable COVID-19 and all samples were taken >8 days after the onset of the symptoms. We

Results
Description of study patients. Cohort 1 included 65 patients with clinical suspicion of active COVID-19 who were cared at the hospital´s emergency department. Table 1 shows the clinical characteristics of these patients. Sixty per cent were male and median age of 56 years. At the time of the sampling, all patients had symptoms compatible with COVID-19, most with (81.5%) pneumonia. Fever (84.6%), cough (81.5%), and dyspnoea (61.5%) were the most commonly observed symptoms. Chest X-rays showed lung opacities in 53 (81.5%) of the cases. Fifty-six patients had either proven (n=49) or probable COVID-19 (n=7), whereas the remaining nine patients were categorized as not infected.
Statistically signi cant differences were determined when comparing patients with proven/probable COVID-19 from Cohorts 1 and 2. Patients from Cohort 1 were older, the number of days between the onset of symptoms to serum sample collection were lower, and showed higher frequency of fever, cough, dyspnoea, pneumonia, and hospital admission. The most frequent symptoms in patients from Cohort 2 were headache, myalgia, sore throat, rhinorrhoea, anosmia, ageusia, and abdominal pain.
Performance of Abbott´s SARS-CoV-2 IgG assay for the diagnosis of COVID-19. From the 38 samples from patients with either community-acquired pneumonia of unknown etiology or coronavirus infection other than SARS-CoV-2, none generated a COVID-19 positive result (index value below 0.2). IgG crossreactivity with patients without COVID-19 was thus ruled out.
In Cohort 1, IgGs were detectable in samples from 34/65 patients; sensitivity, speci city, positive predictive value, and negative predictive value of IgG detection for the diagnosis of active COVID-19 were 60.7%, 100%, 100%, and 29.3%, respectively (Table 3). Serum samples from patients with proven/probable COVID-19 were collected at a median of seven days (IQR 5-10 days) from the onset of symptoms. There was no improvement in the sensitivity in serum samples taken from patients reporting disease onset >8 days, nor by excluding probable cases. Twenty-two COVID-19-positive patients -most of which were proven cases (n=20) -showed negative IgGs. We further compared patients with proven/probable COVID-19 with detectable or undetectable IgGs and found that patients with detectable IgGs had higher frequency of >38ºC fever and/or pneumonia (P<0.05). The remaining comparisons resulted in differences without statistical signi cance, although a trend towards higher frequency of dyspnoea in IgG-positive patients (P=0.055; Table 1) was detected. Sensitivity of IgG detection was higher in patients with probable COVID-19 than in subjects with proven COVID-19 and it was unrelated to the number of days from the onset of symptoms. All patients not ful lling COVID-19 criteria had negative serum IgGs.
In Cohort 2, IgGs were detectable in samples from 38/92 patients; sensitivity, speci city, positive predictive value, and negative predictive value of IgG detection for the diagnosis of past COVID-19 (Cohort 2) were 75%, 84.6%, 79%, and 81.5%, respectively (Table 3). Median number of days for sample collection from the onset of symptoms in patients with proven/probable COVID-19 was 26 days (IQR 22-32 days). Ten patients with proven COVID-19 resulted in negative IgG and serum samples of 8/10 of them taken one month later remained negative. Anosmia was more frequent in patients with proven/probable COVID-19 and detectable IgGs than in subjects without detectable IgGs (P<0.05) ( Table  2). No statistically signi cant differences were observed for the remaining comparisons. We detected IgGs (index values ranging between 2.08 and 6.75) in serum samples from eight patients not ful lling COVID-19 criteria; four of them reported mild symptoms in the two months before sample collection: headache or anosmia (n=3); cough, dyspnoea or ageusia (n=2); myalgia, sore throat or rhinorrhoea (n=1 each). Two women receiving prophylaxis with fosamprenavir were asymptomatic carriers who resulted PCR-positive and IgG-negative while participating in the study; one of them developed symptoms few days after the end of the study, suggesting an early detection COVID-19 case.
Index values of Abbott´s SARS-CoV-2 IgG assay are shown in Figure 1 were not infected and one patient with proven COVID-19 from Cohort 1.

Discussion
Our study shows that serum IgG detection alone is insu cient for the diagnosis of active or past COVID-19, with sensitivity values that range between 60% and 75%, respectively. Detection of IgMs adds limited value to the performance of serological strategies. However, the coupled detection of SARS-CoV-2 RNA and IgG may improve the performance of any of the two procedures alone.
As PCR is insu cient for the diagnosis of active COVID-19, alternative procedures are essential. Non-PCR based procedures rely on the detection of either viral antigens or antibodies against SARS-CoV-2 [18]. Antigen detection has shown sensitivity and speci city values ranging between 50% and 100%, respectively. Low viral loads and variability in a sample collection limit sensitivity [19]. IgG detection may improve and speed up the diagnosis of COVID-19 in symptomatic patients, particularly in negative SARS-CoV-2 PCR cases. It has been shown that in up to 15% of COVID-19-positive patients IgG antibodies are detectable during the rst week following the onset of symptoms [8], although reaching 100% of the patients required two additional weeks [8,12,13,20]. Here, we examined the performance of serology as an adjuvant for the diagnosis of COVID-19 using commercially available procedures. Detection of IgG antibodies using Abbott´s SARS-CoV-2 IgG assay (Architect analyser) and the PanbioTM COVID-19 IgG/IgM rapid test device resulted in high agreement (98%). To the best of our knowledge, to date there is only one study that has assessed Abbott´s SARS-CoV-2 IgG assay (Architect analyser) in samples from patients admitted to the hospital [16]. The performance of the test was particularly good in samples collected ≥10 days from the onset of the disease. Bryan et al. report 100% sensitivity in cases for which samples were taken ≥17 days from the onset of the disease. The results communicated by these authors are signi cantly higher than the ones reported in our study [16]. Here, we tested patients who attended the emergency department with acute and active clinical presentations (Cohort 1). Our serum samples were collected at a median of seven days from the onset of symptoms, mostly <10 days. This may explain the high number of IgG-negative determinations in patients with proven COVID-19. Of note, there is a trend of positive samples that show high index values (>2), particularly in probable COVID-19 cases (Figure 1). Coupling IgG detection to PCR testing may have helped increase the number of COVID-19 cases from 49 (proven cases) to 56 (proven and probable cases).
Another added value of IgG detection is to unearth past undetected and cured COVID-19 cases to assess the proportion of the population that is protected (serological status) and understand the real burden of the disease. At this point, serology may be the only method to track down past/cured infected patients. Here, we con rm the observations by Bryan et al., who reported no IgG detection in samples taken from patients during periods before the COVID pandemic using Architect Abbott´s SARS-CoV-2 IgG [16]. Furthermore, we show the same lack of cross-reactivity in the PanbioTM COVID-19 IgG/IgM rapid test device. We chose patients with past/cured proven COVID-19 (Cohort 2) to study the performance of Architect Abbott's SARS-CoV-2 IgG, but the rate of false-negative serologic results increased 25%. This rate was surprisingly high given that serum samples were taken at a median of 26 days from the onset of symptoms. This may indicate the existence of a group of outpatients for whom the role of serology testing is lower than for hospitalised and sicker subjects. Patients in Cohort 2 may be representative of the general population who passed COVID-19 and did not attend the hospital.
Considering the data presented in this study, we conclude that IgG detection is insu cient to assess the current percentage of subjects with past infections. Other authors have also failed to detect IgGs in 15%-20% of COVID-19-positive patients [7,9,11,14]. Sensitivity of IgGs antibody detection against either the SARS-CoV-2 nucleoprotein or the spike protein receptor-binding domain could make the difference [10]. Since Architect Abbott´s SARS-CoV-2 IgG detects anti-nucleoprotein antibodies, we cannot rule out the presence of antibodies against other viral epitopes.
We compared study patients grouped by IgG detection results aiming to discover potential predictors of IgG-negative patients with COVID-19 in both cohorts. We failed to nd such predictors, which illustrates the di culties to determine which patients will not have detectable IgGs. Whether the absence of detectable IgG antibodies in patients with cured infection means a lack of protection to prevent a new infection remains to be de ned.
Our study is subjected to limitations. IgM detection using the Architect analyser was not available at the moment of the study. Moreover, it was not possible to collect additional serum samples from patients in   Figure 1 Scatter plots of index values obtained by Abbott´s SARS-CoV-2 IgG assay (Architect analyser) versus days from the onset of symptoms of all patients (1a), and patients from Cohort 1 (1b) and Cohort 2 (1c).
Plots in red indicate proven COVID-19 cases and dots in blue indicate probable COVID-19 cases