key: cord-0978328-1clql98y
authors: Chen, Mengyu; Qin, Rundong; Mei, Jiang; Yang, Zhaowei; Wen, Weiping; Li, Jing
title: Clinical applications of detecting IgG, IgM, or IgA antibody for the diagnosis of COVID-19: A meta-analysis and systematic review
date: 2021-01-12
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2021.01.016
sha: 76576affe60f685dd7bf0403fcc71b26bfbecf28
doc_id: 978328
cord_uid: 1clql98y

BACKGROUND: Coronavirus disease 2019 is a global pandemic. Serological antibody test is one important diagnostic method increasingly used in the clinic, although its clinical application is still under investigation. METHODS: We conducted a meta-analysis to compare the diagnostic performances of severe acute respiratory syndrome coronavirus 2 specific antibody tests in COVID-19 patients. Test results analyzed included (1) IgM-positive but IgG-negative (IgM(+)IgG-), (2) IgG-positive but IgM-negative (IgG(+)IgM-), (3) both IgM and IgG-positive (IgM(+)IgG(+)), (4) IgM-positive without IgG information (IgM(+)IgG(+/-)), (5) IgG-positive without IgM information (IgG(+)IgM(+/-)), (6) either IgM or IgG-positive (IgM(+) or IgG(+)), and (7) IgA-positive (IgA(+)). RESULTS: Sixty-eight studies were included. The pooled sensitivities of IgM(+)IgG-, IgG(+)IgM-, IgM(+)IgG(+), IgM(+)IgG(+/-), IgG(+)IgM(+/-), and IgM(+) or IgG(+) were 6%, 7%, 53%, 68%, 73%, and 79% respectively. The pooled specificity ranged from 98% to 100%. The IgA(+) had a pooled sensitivity of 78%, but a relatively low specificity of 88%. Tests conducted two weeks after symptom onset provided improved diagnostic accuracy. Chemiluminescence immunoassay and detection of S protein as the antigen could offer a more accurate diagnostic result. DISCUSSION: Our findings support the supplemental role of serological antibody tests in COVID-19 diagnosis. However, their capacity to diagnose COVID-19 early in the disease course could be limited.

supporting the use of antibody tests in the practice for COVID-19 is missing(Lisboa Bastos et al., 2020) . Indeed, antibody subtype, the antigen of an antibody preparation, detection time, and method of measurement varied markedly among different studies.

Some studies detected both IgM and IgG and reported positive results once either one was positive, while other studies only detected IgM or IgG individually. There is no consensus on the interpretation of antibody test results . The presence of IgM, IgG, and IgA either alone or in certain combinations may be related to disease severity and immunization, which could affect diagnostic accuracy.

Therefore, our current meta-analysis aimed to investigate the diagnostic effectiveness of SARS-CoV-2 specific antibodies stratified by different positive results, including (1) IgM-positive but IgG-negative (IgM + IgG -), (2) IgG-positive but IgM-negative (IgG + IgM -), (3) both IgM and IgG-positive (IgM + IgG + ), (4) IgM-positive without IgG information (IgM + IgG +/-), (5) IgG-positive without IgM information (IgG + IgM +/-), (6) either IgM or IgG-positive (IgM + or IgG + ), and (7) IgA-positive (IgA + ). For the first three panels, our study provided clear information of antibody types in presence, while previous meta-analysis focused on the diagnostic accuracy of IgM + IgG +/-, IgG + IgM +/-, and IgM + or IgG + that only offer vague information (Caini et al., 2020; Deeks et al., 2020; Lisboa Bastos et al., 2020; Moura et al., 2020) .

Our meta-analysis followed the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines (Moher et al., 2009) . We searched the Pubmed, Medline, Embase, Cochrane library, ICTRP, ClinicalTrials.gov, Medrxiv, Biorxiv, CNKI, Sinomed, WanFangdata, and Cqvip databases. The MESH terms and entry terms for concepts of Table S1 .

The inclusion criteria were as follows: 1) patients with COVID-19 confirmed by RT-PCR test, or by the combination of RT-PCR test and clinical manifestation; 2) serological diagnostic tests with no method, antibody, or antigen restrictions; and 3) prospective or retrospective studies reporting COVID-19 confirmed cases against other non-COVID-19 individuals or suspected COVID-19 cases that could not be confirmed by microbiological tests.

The exclusion criteria were as follows: 1) case reports with a sample size less than 10;

2) publications without 2 × 2 contingency table; 3) reviews, meta-analysis and systematic analysis; 4) studies focused on ineligible population, such as pregnant women, elderly and children; and 5) studies carried out in a population surveillance setting.

Two independent researchers screened the publications by reading titles and abstracts.

In the case of disagreement, researchers discussed the full-text publications with a third researcher.

Data were independently extracted by two authors. Data extracted from each study included the first author's last name, age and sex of COVID-19 patients, days since symptom onset, test kit manufacturer, study design, reference standard, RT-PCR sample type, blood sample type, methods, antigen, and antibody types of antibody detection. We extracted the numbers of true positive, false positive, false negative, and true negative to construct the 2x2 contingency table and to estimate sensitivity and specificity. We attempted to investigate the diagnostic accuracy of different antibody combinations, including (1) IgM + IgG -, (2) IgG + IgM -, (3) IgM + IgG + , (4) IgM + IgG +/-, (5) IgG + IgM +/-, (6) IgM + or IgG + , and (7) IgA + .

The methodological quality of included studies was assessed independently by two authors, using the quality assessment of diagnostic accuracy studies 2 (QUADAS-2) tool (Whiting et al., 2011) . The QUADAS tool consists of four key domains, including patient, index test, reference standard, and flow and timing.

J o u r n a l P r e -p r o o f

For the individual studies, we calculated estimated sensitivity and specificity based on the 2x2 contingency table. All results are reported with 95% confidence intervals (CI).

The data are summarized as paired forest plots. We used a bivariate random-effects model for the meta-analysis because different studies had different cut-off values.

We plotted summary ROC curves to estimate the overall diagnostic efficiency of each antibody during different clinical courses. We used a random effects logistic regression model to compare the diagnostic accuracy between different antibodies, different antibody detection methods, and different antigens.

Unobserved heterogeneity was quantified by I 2 . An I 2 above 50% was considered to be of significant heterogeneity. We investigated sources of heterogeneity by performing univariable meta-regression and subgroup analysis. Covariates including country, antigen, antibody detection method, detection time, blood sample type, disease severity, and whether tests were built in-house were considered to contribute to the heterogeneity.

All of the statistical analyses were performed using meta-analysis modules of Stata (Ver 16; Stata Corp. College Station, TX, USA) (metandi and midas) (Dwamena, 2007; Harbord, 2008) . Summary receiver operating characteristic (SROC) curves were plotted using Review Manager 5 (Revman 5, Version 5.3, Copenhagen: Nordic Cochrane Centre, The Cochrane Collaboration). A P < 0.05 was considered to indicate a statistically significant difference.

J o u r n a l P r e -p r o o f PROSPERO CRD42020195898

A total number of 4002 records were identified through the database search. After The detailed results are presented in Table 1 . The sensitivities of IgM + IgG -, IgG + IgM -, and IgM + IgG + were 6% (95%CI 4%-8%), 7% (95%CI 5%-11%), and 53% (95%CI 46%-60%), respectively, while the sensitivities of IgM + IgG +/-, IgG + IgM +/-, and IgM + or IgG + rise to 68% (95%CI 62%-73%), 73% (95%CI 69%-77%), and 79% (95%CI 76%-83%), respectively. In contrast, the specificity of the serological test was high, ranging from 98% to 100%. For IgA, the sensitivity was comparable to IgM + or IgG + (78%), but the specificity was only 88%. Figure 3 compares the SROC curves of IgM + IgG -, IgG + IgM -, IgM + IgG + , and IgM + or IgG + , with stratification by days from symptom onset. Except for IgM + IgG -, the area under the receiver operating characteristic (AUROC) curve peaked after 14 days from the symptom onset regardless of the antibody type. Figure 4 shows the SROC curve of IgA.

(1) Meta-regression

Given that the results of pooled diagnostic accuracy presented a high I 2 , we further performed a meta-regression analysis to investigate the potential source of heterogeneity. The factors, including characteristics of included patients, blood sample type, detection method, antigen types, and commercial availability, were included in the meta-regression models for each antibody panel. The results suggested that the marked heterogeneity might have resulted from all factors that were included in the J o u r n a l P r e -p r o o f analysis. The detailed results are presented in Table S3 .

The detailed subgroup analysis results are presented in Table 2 .

(2) Subgroup analysis by detection time

Detection time was classified into three groups: patients receiving antibody testing in

(1) the first week, (2) the second week, and (3) greater than three weeks after symptom onset. Due to the lack of adequate studies, we did not perform an analysis for IgA + . In the first week, the sensitivity ranged from 4% to 34%, with the best performance by IgM + or IgG + and the worst by IgG + IgM -.

(3) Subgroup analysis by detection methods Detection methods were classified into three groups: (1) CLIA, (2) ELISA, and (3) LFIA. In this analysis, we evaluated studies reporting IgM + or IgG + since it has the best diagnostic performance. The pooled sensitivities of CLIA, ELISA, and LFIA were 86%

(95CI% 73%-94%), 83% (95CI% 76%-88%), and 75% (95CI% 71%-79%),

respectively. Due to data limitations, we only pooled the overall results without stratification by detection timing. The 95% confidence interval overlapped between each test method. However, the sensitivity of LFIA was significantly lower than other methods (P < 0.001).

(4) Subgroup analysis by antigen protein For IgM + or IgG + , the pooled sensitivities of S protein, N protein, and both S and N protein were 90% (95CI% 84%-94%), 79% (95CI% 65%-88%), and 88% (95CI% 83%-91%), respectively. The pooled sensitivity was higher when applying S protein (P < 0.001).

The clinically suspected patients had typical epidemiological history, signs and symptoms but without positive RT-PCR results as gold standard of diagnosis. For those patients, the positive rates of IgM + IgG -, IgG + IgM -, IgM + IgG + , IgM + IgG +/-, IgG + IgM +/-, and IgM + or IgG + were 5% (95CI% 2%-14%), 19% (95CI% 8%-37%), 32% (95CI% 11%-64%), 47% (95CI% 29%-66%), 72% (95CI% 47%-89%), and 81% (95CI% 59%-92%), respectively. The results are presented in Table 2 . Figure S2 -5 presents the paired forest plots of each analysis. Figure S6 -9 shows the summary receiver operating characteristic (SROC) curve of each analysis. Figure S10 summarizes the publication bias of each analysis. Most of the analyses were carried out without risk of bias. patients.

Our current meta-analysis attempted to provide comprehensive evidence of the diagnostic accuracy of serum antibodies in detecting COVID-19. This stands out the novelty out of others because our review has addressed the question of which positive results could provide optimal diagnostic accuracy. Of the six panels of positive antibody results of IgM and IgG, all presented a favorable specificity. However, the panels exhibited a marked difference in their sensitivities. The panel of IgM + or IgG + showed the highest sensitivity at 79%, followed by IgG + IgM +/-(73%), IgM + IgG +/-(68%), IgM + IgG + (53%), IgG + IgM -(7%), and IgM + IgG -(6%). Additionally, serological tests conducted two weeks after symptom onset provided improved diagnostic accuracy compared to the earlier tests. Our subgroup analysis further found that CLIA could offer a more accurate diagnosis than either ELISA or LFIA, while S or S/N protein could also contribute to the improvement in the diagnostic accuracy. Furthermore, IgA seems to be a potential candidate for the antibody test for COVID-19 diagnosis. Our data found that detection of IgM + or IgG + antibody panel in patients with COVID-19 could provide favorable accuracy, especially when testing was conducted two weeks after IgG were detected simultaneously after SARS-CoV-2 infection. (Lee et al., 2020; Long QX, 2020; .Differences in the half-lives of IgM and IgG, distinct detection methods, and competition between IgM and IgG have an impact on the final outcome of IgG and IgM determination in the serum. Thus, the presence of IgM failed to identify acute infection, for which the diagnosis still relies on virologic tests . In addition, we found that the sensitivity of IgG at later time points improved compared to earlier time points. This finding suggested that performing early antibody detection might decreased the diagnosis accuracy. Therefore, differences in the halflives of antibodies, detection methods, and competition between IgM and IgG have an impact on the final outcome of IgG and IgM determination in the serum. To date, the estimation of prevalence in most of the area affected by the pandemic was less than 5% (Pollán et al., 2020) . Therefore, as shown by the unacceptably low PPV of the IgM + or IgG + (67.52%) panel, serological tests were not capable of identifying nearly a third of cases. Proper algorithms and the intended detection purpose for serology testing have yet to be determined.

The specificity of the serological test is related to cross-reaction between antigens of SARS-Cov-2 and other coronaviruses. Currently, S protein and N proteins are mainly used as the targets for the antibody-based detection of SARA-CoV-2. The S protein includes S1 and S2 subunit that aid in host infection, while the N protein plays an important role in the transcription and replication of viral RNA (Dutta et al., 2020; . N protein was highly homogeneous to SARS-CoV (~90%) and J o u r n a l P r e -p r o o f shared sequence similarity to other human CoVs (19-45%). However, S1 subunit of SARS-CoV-2 share only 64% and 57% with SARS-CoV and MERS-CoV, respectively, and 9-37% with other human CoVs (Kames et al., 2020) . S2 subunit shows 88% sequence homology with the SARS-CoV . Therefore, S1 subunit of SARS-CoV-2 potentially demonstrate less cross-reactivities among the endemic coronavirus. In the current meta-analysis, we firstly provided comprehensive evidence

showing that using S1 protein as antigen for antibody preparation improved the diagnostic accuracy as compared to N protein, this finding corresponded to recent studies which reported a higher false positive rate of N protein compared to S1

protein (Jiang et al., 2020) . In addition, given that neutralizing antibodies primarily target S1 protein in SARS-CoV-2, the clinical value of serology test might expand to the evaluation of vaccine efficacy and individual immunity. Our data suggested that serology tests against S1 protein, rather than N protein, should be recommend to conduct in COVID-19 diagnosis.

Caution is also warranted because of the discrepancy between different detection methods. Our findings indicated that CLIA provide a more accurate diagnosis than either ELISA or LFIA in COVID-19 diagnosis. Similar results were reported by a previous meta-analysis where aimed to compare the diagnostic accuracy of different methods(Lisboa Bastos et al., 2020) . Although LFIA test have been rapidly developed and marketed to meet the urgent need for a low cost, rapid, and accurate point-of-care test, the low sensitivity of LFIA is of particular concern. The factors influencing the J o u r n a l P r e -p r o o f analytic detection of LFIA include the size of gold nanoparticles, the antibody concentration, the conjugation PH, and characteristics of membranes (Safenkova et al., 2012) . Besides, the considerable heterogeneity of included patients and sample types were also the potential contributors to the low pooled performance of LFIA. It is worth noting that many sample types were using in serological tests, including serum, plasma, and whole blood. However, recent study demonstrated a higher sensitivity by using serum or plasma as compared to whole blood, whereas there was no significant difference was observed between plasma and serum (De Marinis et al., 2020) . Therefore, using whole blood (or capillary blood) might cause a relatively low sensitivity for disease surveillance (Flower et al., 2020) . As the COVID-19 pandemic keeps accelerating, it is crucial to determine the suitable detection methods depending on the clinical settings.

Targeted at mucosa, SARS-CoV-2 is able to induce strong mucosal immunity, leading to the generation of secretory IgA (sIgA). The IgA system not only served as the firstline barrier to protect the pathogens from adhering to the mucosa, but also interacted with the innate and adaptive immune systems for the maintenance of homeostasis (Pabst, 2012; Renegar et al., 2004) . Studies suggested that IgA and IgG levels were markedly higher in patients with severe disease compared to patients with mild or moderate disease . Studies on IgA are limited, but there is room for future exploration. In further research, serological tests using a combination of the three antibodies could be a possible way to improve diagnostic accuracy. A2-D2 indicated that antibodies were detected in 0-7 days after symptoms onset.

A3-D3 indicated that antibodies were detected in 8-14 days after symptoms onset.

A4-D4 indicated that antibodies were detected after 14 days after symptoms onset.

J o u r n a l P r e -p r o o f Suspected, the clinically suspected patients had typical epidemiological history, signs and symptoms but without positive RT-PCR results as gold standard of diagnosis 

Covid-19: testing times

Meta-analysis of diagnostic performance of serological tests for SARS-CoV-2 antibodies up to 25 April 2020 and public health implications

Serodiagnostics for Severe Acute Respiratory Syndrome-Related Coronavirus-2: A Narrative Review

Serology assessment of antibody response to SARS-CoV-2 in patients with COVID-19 by rapid IgM/IgG antibody test

Antibody tests for identification of current and past infection with SARS-CoV-2

The Nucleocapsid Protein of SARS-CoV-2: a Target for Vaccine Development

MIDAS: Stata module for meta-analytical integration of diagnostic test Performance 2020

Clinical and laboratory evaluation of SARS-CoV-2 lateral flow assays for use in a national COVID-19 seroprevalence survey

Stata module to perform meta-analysis of diagnostic accuracy. METANDI: Stata Module to Perform Meta-Analysis of Diagnostic Accuracy

Structural and functional properties of SARS-CoV-2 spike protein: potential antivirus drug development for COVID-19

Global profiling of SARS-CoV-2 specific IgG/ IgM responses of convalescents using a proteome microarray

Sequence analysis of SARS-CoV-2 genome reveals features important for vaccine design

Variation in False-Negative Rate of Reverse Transcriptase Polymerase Chain Reaction-Based SARS-CoV-2 Tests by Time Since Exposure

Dynamics of anti-SARS-Cov-2 IgM and IgG antibodies among COVID-19 patients

Comparison of throat swabs and sputum specimens for viral nucleic acid detection in 52 cases of novel coronavirus (SARS-Cov-2)-infected pneumonia (COVID-19)

Diagnostic accuracy of serological tests for covid-19: systematic review and meta-analysis

Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement

Testing: A Systematic Review and Meta-Analysis

COVID-19 Treatment Guidelines Panel. Coronavirus Disease 2019 (COVID-19) Treatment Guidelines

SARS-CoV-2 infection serology: a useful tool to overcome lockdown?

New concepts in the generation and functions of IgA

Viral load of SARS-CoV-2 in clinical samples

Serology testing in the COVID-19 pandemic response

Prevalence of SARS-CoV-2 in Spain (ENE-COVID): a nationwide, population-based seroepidemiological study

Dynamic changes of throat swabs RNA and serum antibodies for SARS-CoV-2 and their diagnostic performances in patients with COVID-19

Role of IgA versus IgG in the Control of Influenza Viral Infection in the Murine Respiratory Tract

Factors influencing the detection limit of the lateral-flow sandwich immunoassay: a case study with potato virus X

Temporal profiles of viral load in posterior oropharyngeal saliva samples and serum antibody responses during infection by SARS-CoV-2: An observational cohort study

Detection of SARS-CoV-2 in Different Types of Clinical Specimens

QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies

World Health Organization. WHO Coronavirus Disease (COVID-19) Dashboard 2020

Distinct features of SARS-CoV-2-specific IgA response in COVID-19 patients

IgM + IgG + 63 studies, 13344 samples 53

IgM + IgG +/-87 studies, 18924 samples 68

IgG + IgM +/-88 studies, 18597 samples 73

IgM + or IgG + 76 studies

IgA + 6 studies, 934 samples 78

95% confidence intervals

IgM + IgG + (25 studies, 3803 samples

IgM + IgG +/-(38 studies, 5710 samples)

IgG + IgM +/-(38 studies

IgM + or IgG + (41 studies

36 studies, 5705 samples

37 studies, 5876 samples) 7

IgM + IgG + (35 studies

IgG + IgM +/-(48 studies

IgM + or IgG + (52 studies

Days >14

IgG + IgM -(38 studies, 5387 samples

IgM + IgG + (36 studies, 5139 samples

IgM + IgG +/-(53 studies, 7855 samples

IgG + IgM +/-(51 studies

IgM + or IgG + (54 studies, 8225 samples) 93

Methods CLIA (10 studies

Suspected IgM + IgG -(8 studies, 1447 samples) 5

studies, 1446 samples) 19

IgM + IgG + (8 studies, 1426 samples

IgM + IgG +/-(10 studies, 1858 samples)

IgG + IgM +/-(10 studies, 1858 samples

IgM + or IgG + (16 studies, 2856 samples)

PPV5: positive predictive value at the prevalence of 5%; PPV10: positive predictive value at the prevalence of 10%

ELISA, enzyme-linked immunosorbent assay; LFIA, lateral flow immunosorbent assay

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.J o u r n a l P r e -p r o o f