A Systematic Review and Meta-Analysis Comparing the Diagnostic Accuracy Tests of COVID-19

In this paper, we present a systematic review and meta-analysis that aims to evaluate the reliability of coronavirus disease diagnostic tests in 2019 (COVID-19). This article seeks to describe the scientific discoveries made because of diagnostic tests conducted in recent years during the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic. Between 2020 and 2021, searches for published papers on the COVID-19 diagnostic were made in the PubMed database. Ninety-nine scientific articles that satisfied the requirements were analyzed and included in the meta-analysis, and the specificity and sensitivity of the diagnostic accuracy were assessed. When compared to serological tests such as the enzyme-linked immunosorbent assay (ELISA), chemiluminescence immunoassay (CLIA), lateral flow immunoassay (LFIA), and chemiluminescent microparticle immunoassay (CMIA), molecular tests such as reverse transcription polymerase chain reaction (RT-PCR), reverse transcription loop-mediated isothermal amplification (RT-LAMP), and clustered regularly interspaced short palindromic repeats (CRISPR) performed better in terms of sensitivity and specificity. Additionally, the area under the curve restricted to the false-positive rates (AUCFPR) of 0.984 obtained by the antiviral neutralization bioassay (ANB) diagnostic test revealed significant potential for the identification of COVID-19. It has been established that the various diagnostic tests have been effectively adapted for the detection of SARS-CoV-2; nevertheless, their performance still must be enhanced to contain potential COVID-19 outbreaks, which will also help contain potential infectious agent outbreaks in the future.


Introduction
The end of 2019 saw the first mention of the coronavirus disease 2019 (COVID- 19), which is brought on by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1]. Due in languages other than English, reviews, and meta-analyses. The titles and abstracts of the articles discovered using the search method were examined in the second stage of the screening stage. The entire studies with high relevance were retrieved and separated from the papers with titles or abstracts that lacked the necessary information to be taken into account during the eligibility stage. The diagnostic method, the quantity, variety, and clinical characteristics of the COVID-19 patients, as well as healthy controls, were all gathered from each chosen study. All studies that assessed diagnostic accuracy using sensitivity and specificity measures were included. Additionally, information about the geographical distribution, the number of studies associated by nation, and the frequency of diagnostic methods utilized were retrieved. Studies lacking information and having conflicting incomplete information about the COVID-19 diagnostic tests' sensitivity and specificity were not included in the study.

Statistical Analysis
Results were entered into Microsoft Excel (version 10.0, Microsoft Corporation, Redmond, WA, USA) spreadsheets and analyzed in the R programming environment (version 4.2.1) using the package "mada" (version 0.5.11) https://cran.r-project.org/web/ packages/mada/index.html (accessed on 24 October 2022); which employs a hierarchical model that accounts for within and between-study (heterogeneity) and the correlation between sensitivity and specificity [40]. Initially, the number of true negatives (TP), false negatives (FN), true positives (TP), and false positives (FP) were analyzed separately for each diagnostic technique; while the evaluation of sensitivity (Se) and specificity (Sp) made it possible to determine the diagnostic performance. Additionally, the positive likelihood ratio (LR+) expresses the ratio between the probability of expecting a positive test in a patient and the probability of expecting a positive test in a patient without the disease [41]; the negative likelihood ratio (LR−), which expresses the probability that a patient will test negative among people with the disease and the probability that a patient will test negative among people without disease; and the diagnostic likelihood ratio (DOR), which is the odds ratio of the positivity of a diagnostic test result in the diseased population relative to the non-diseased population [42]; and the 95% confidence interval (CI) were determined. Summary receiver operating characteristic (sROC) curves were fitted, according to the parameters of the "Reitsma" model of the "mada" package and were used to compare the diagnostic accuracy of CD diagnostic techniques [43]. The confidence level for all calculations was set to 95%, using a continuity correction of 0.5 if pertinent.

Data Sources and Study Selection
In this study, a systematic review followed by meta-analysis was performed to measure the accuracy of diagnostic tests for COVID- 19. A flowchart of the study strategy was prepared and presented ( Figure 1). To this end, a search was made in the PubMed database with the MeSH terms "COVID-19", and a co-occurrence network map of MeSH terms was developed; Through the search, 981 scientific articles were obtained between the years 2020 and 2021. The minimum number of occurrences of keywords was set at a value of five, and a network map with 2.518 keywords was generated (Figure 2A). The formation of five main clusters was found in the analysis of the network map, in the cluster related to diagnostic techniques (purple color) terms such as "Reverse Transcription-Polymerase Chain Reaction", "Reverse Transcriptase Loop-Mediated Isothermal Amplification", "Clustered Regularly Interspaced Short Palindromic Repeats", "Microarrays", "Next-Generation Sequencing", "Enzyme-Linked Immunosorbent Assay", "Antiviral Neutralization Bioassay", "Biosensor", and "Immunoassay". In addition, terms such as "COVID-19", "SARS-CoV-2", "adult", China", "disease outbreaks middle-aged", and "female" were common denominators ( Figure 2A). Chain Reaction", "Reverse Transcriptase Loop-Mediated Isothermal Amplification", "Clustered Regularly Interspaced Short Palindromic Repeats", "Microarrays", "Next-Generation Sequencing", "Enzyme-Linked Immunosorbent Assay", "Antiviral Neutralization Bioassay", "Biosensor", and "Immunoassay". In addition, terms such as "COVID-19", "SARS-CoV-2", "adult", China", "disease outbreaks middle-aged", and "female" were common denominators ( Figure 2A).    The number of studies chosen for RT-PCR, RT-LAMP, CRISPR, MA, NGS, ELISA, ANB, BS, and immunoassays were: 303, 3,14,3,16,145,35,95, and 367, respectively ( Figure 2B). The three-step eligibility criterion allowed 303, 369, and 215 articles to be excluded, from the criteria for identification, screening, and eligibility, respectively. Therefore, 99 articles were selected for meta-analysis ( Figure 3). It is observed that in most studies, the diagnostic techniques used were immunoassays (CLIA, LFIA, CMIA, ECLIA, and FIA) ( Figure 3A). Additionally, China, the United States of America, and India are the countries that have carried out a higher number of studies related to diagnostic tests for COVID-19 ( Figure 3B,C). ure 2B). The three-step eligibility criterion allowed 303, 369, and 215 articles to be excluded, from the criteria for identification, screening, and eligibility, respectively. Therefore, 99 articles were selected for meta-analysis ( Figure 3). It is observed that in most studies, the diagnostic techniques used were immunoassays (CLIA, LFIA, CMIA, ECLIA, and FIA) ( Figure 3A). Additionally, China, the United States of America, and India are the countries that have carried out a higher number of studies related to diagnostic tests for COVID-19 ( Figure 3B,C).
The correlation between sensitivities and false positive rates was analyzed and a negative result is shown r = 0.313, 95%CI (−0.502, 0.834  Figure 5 and Supplementary Figure S2.

Enzyme-Linked Immunosorbent Assay for IgM
Five studies were selected using IgM-detecting ELISA as a diagnostic tool [74,84,87,89,90], in which a total of 1585 subjects were studied. Sensitivity ranged from 46 Figure 8 and Supplementary Figure S5.

Enzyme-Linked Immunosorbent Assay for IgA
Five studies were selected using IgA-detecting ELISA as a diagnostic technique [74,75,77,89,90]. A total of 1632 subjects were studied. Sensitivity ranged from 79.

Chemiluminescence Immunoassay for IgM
Five studies were selected using CLIA as an IgM detection technique [77,79,103,104,106]. A total of 1240 subjects were studied. Sensitivity ranged from 58.

Chemiluminescence Immunoassay for IgM-IgG
The analysis identified 28 published studies that used CLIA to detect IgM-IgG antibodies for COVID-19. After analysis, only five studies [77,85,103,106,108] were selected. A

Chemiluminescence Immunoassay for IgM-IgG
The analysis identified 28 published studies that used CLIA to detect IgM-IgG antibodies for COVID-19. After analysis, only five studies [77,85,103,106,108] were selected. A total of 1008 subjects were studied. Sensitivity ranged from 64.

Lateral Flow Immunoassay for IgM
Six studies were selected using LFIA as an IgM detection technique [75,77,79,111,112,115], in which a total of 2704 subjects were studied. Sensitivity ranged from 23.

Lateral Flow Immunoassay for IgM-IgG
Nine studies were selected using LFIA as an IgM-IgG detection technique [75,77,85,90,108,109,115,117,118]. A total of 9629 subjects were studied. Sensitivity ranged from 44.  Sensitivity and specificity are reported as mean (95% confidence limits). The forest plot represents the estimated sensitivity and specificity (black squares) and their 95% confidence limits (horizontal black line) [75,77,79,111,112,115].

Lateral Flow Immunoassay for IgM-IgG
Nine studies were selected using LFIA as an IgM-IgG detection technique [75,77,85,90,108,109,115,117,118]. A total of 9629 subjects were studied. Sensitivity ranged from 44.    Sensitivity and specificity are reported as mean (95% confidence limits). The forest plot represents the estimated sensitivity and specificity (black squares) and their 95% confidence limits (horizontal black line) [75,77,85,90,108,109,115,117,118].

Chemiluminescent Microparticle Immunoassay
The analysis identified 13 published studies that used CMIA as a diagnostic technique for COVID-19. After analysis, only five studies [81,88,105,107,133] were selected. A total of 939 subjects were studied. Sensitivity ranged from 62.8 to 95.7%, with a median of 90.3%, and 95%CI (76.4, 96.4). Test for equality of sensitivities analysis showed: χ 2 = 51.

Fluorescence Immunoassay
Three studies were selected using the FIA technique [121,134,135], in which a total of 829 subjects were studied. Sensitivity ranged from 38.0% to 92.6%, with a median of 64.4%,  Figure 19 and Supplementary Figure S16.

Summary ROC Curves (sROC)
Comparison of the diagnostic techniques data for COVID-19 (RT-PCR, RT-LAMP, CRISPR, ELISA IgG, ELISA IgM, ELISA IgA, ABN, BS, CLIA IgG, CLIA IgM, CLIA IgM-IgG, LFIA IgG, LFIA IgM, LFIA IgM-IgG, LFIA N protein, CMIA, and FIA) was performed through an sROC curve analysis ( Figure 21) and due to implicit or explicit alterations between studies and variation in the cut-off points of the test, differences in sensitivity and specificity may occur [141,142]. The area under the curve (AUC) calculated for the diagnostic techniques for COVID-19 is shown in Figure 21, showing better performance for ABN. Furthermore, when the AUC was limited to the observed false positive rates (FPR) (AUCFPR), results revealed the relatively better performance of the ABN diagnostic test for COVID-19 ( Figure 21).

Summary ROC Curves (sROC)
Comparison of the diagnostic techniques data for COVID-19 (RT-PCR, RT-LAMP, CRISPR, ELISA IgG, ELISA IgM, ELISA IgA, ABN, BS, CLIA IgG, CLIA IgM, CLIA IgM-IgG, LFIA IgG, LFIA IgM, LFIA IgM-IgG, LFIA N protein, CMIA, and FIA) was performed through an sROC curve analysis ( Figure 21) and due to implicit or explicit alterations between studies and variation in the cut-off points of the test, differences in sensitivity and specificity may occur [141,142]. The area under the curve (AUC) calculated for the diagnostic techniques for COVID-19 is shown in Figure 21, showing better performance for ABN. Furthermore, when the AUC was limited to the observed false positive rates (FPR) (AUC FPR ), results revealed the relatively better performance of the ABN diagnostic test for COVID-19 ( Figure 21).

Discussion
The disease COVID-19 has had a catastrophic effect, teaching future generations to examine several conditions that encourage the growth of infectious diseases [143]. The COVID-19 pandemic brought to light the gaps in disease detection, warning, and response systems. Given how interconnected the globe is and how quickly a pandemic might spread, it illustrated the necessity of restarting the global health and health security system [144]. The key factor in the spread of infectious diseases is human migration, and this factor is increasing due to globalization and transportation networks. In this environment, there is the potential for the quick and challenging management of the spread of harmful germs like SARS-CoV-2 [145]. Asymptomatic carriers may be able to transfer COVID-19 during the incubation period without displaying any symptoms or signs because the percentage of asymptomatic individuals who test positive for COVID-19 ranges from 8.44 percent to 39.00 percent [146]. In contrast, the production of antibodies starts days after the onset of the infection. As a result, the start time of the potential infection should be taken into consideration to make an accurate diagnosis, keeping in mind that molecular tests will become less sensitive over time while serological tests will become more sensitive after a few days of infection [147]. These factors make the quick and precise diagnosis of individuals with COVID-19 infections crucial. In addition, diagnostic tests required processing in laboratories with sophisticated materials, leading to a longer turnaround time and several days needed for results to be available. Some of these obstacles were eventually overcome by newly designed assays that had analytical precision that is more precisely specified than prior assays [148,149].
Since SARS-CoV-2 is an RNA virus, it can be detected using any of the various molecular tests for RNA detection [150,151]. To use DNA detection techniques, reverse transcriptase must convert the virus RNA into a DNA complement [152]. Although RT-PCR is currently the most used molecular technique for the early identification of COVID-19, several promising alternatives exist, including RT-LAMP and CRISPR [153]. Due to its high specificity, RT-PCR is regarded as the gold standard molecular diagnostic test for COVID-19 throughout the world; however, the slow detection of the virus was caused by the limited availability of kits and reagents, the use of pricy laboratory equipment, and the requirement for qualified personnel [154,155]. The RT-PCR test has also been documented to fail in suspected and confirmed patients with clinical repercussions. In this situation, further clinical and molecular testing ought to be considered when determining

Discussion
The disease COVID-19 has had a catastrophic effect, teaching future generations to examine several conditions that encourage the growth of infectious diseases [143]. The COVID-19 pandemic brought to light the gaps in disease detection, warning, and response systems. Given how interconnected the globe is and how quickly a pandemic might spread, it illustrated the necessity of restarting the global health and health security system [144].
The key factor in the spread of infectious diseases is human migration, and this factor is increasing due to globalization and transportation networks. In this environment, there is the potential for the quick and challenging management of the spread of harmful germs like SARS-CoV-2 [145]. Asymptomatic carriers may be able to transfer COVID-19 during the incubation period without displaying any symptoms or signs because the percentage of asymptomatic individuals who test positive for COVID-19 ranges from 8.44 percent to 39.00 percent [146]. In contrast, the production of antibodies starts days after the onset of the infection. As a result, the start time of the potential infection should be taken into consideration to make an accurate diagnosis, keeping in mind that molecular tests will become less sensitive over time while serological tests will become more sensitive after a few days of infection [147]. These factors make the quick and precise diagnosis of individuals with COVID-19 infections crucial. In addition, diagnostic tests required processing in laboratories with sophisticated materials, leading to a longer turnaround time and several days needed for results to be available. Some of these obstacles were eventually overcome by newly designed assays that had analytical precision that is more precisely specified than prior assays [148,149].
Since SARS-CoV-2 is an RNA virus, it can be detected using any of the various molecular tests for RNA detection [150,151]. To use DNA detection techniques, reverse transcriptase must convert the virus RNA into a DNA complement [152]. Although RT-PCR is currently the most used molecular technique for the early identification of COVID-19, several promising alternatives exist, including RT-LAMP and CRISPR [153]. Due to its high specificity, RT-PCR is regarded as the gold standard molecular diagnostic test for COVID-19 throughout the world; however, the slow detection of the virus was caused by the limited availability of kits and reagents, the use of pricy laboratory equipment, and the requirement for qualified personnel [154,155]. The RT-PCR test has also been documented to fail in suspected and confirmed patients with clinical repercussions. In this situation, further clinical and molecular testing ought to be considered when determining the COVID-19 diagnosis [26,154]. The identification of asymptomatic infections with high specificity and sensitivity is a key point for managing the pandemic, however, because it has been reported that asymptomatic infected people have a few copies below the detection limit of the nucleic acid within the upper respiratory tract, it has to lead to false negatives in asymptomatic infected individuals [156]. The investigation revealed no discernible difference between RT-PCR and CRISPR for these parameters, with RT-PCR having a median sensitivity and specificity of 94.5 and 94.4 percent, respectively. Both PCR and CRISPR are used. CRISPR, however, outperformed AUCFPR when compared, which can be partially attributed to the disparity in the number of studies considered and sample sizes. The use of CRISPR to diagnose COVID-19 offers several benefits, including quick detection (around 30 min), high sensitivity and accuracy, mobility, and lack of requirement for specific laboratory equipment [155]. Additionally, RT-LAMP demonstrated among the molecular assays the lowest sensitivity (median, 91.9%) and the best specificity (median, 98.8%). RT-LAMP is superior to RT-PCR in that it can be amplified at a constant temperature without the use of a thermocycler. One week before the beginning of symptoms, SARS-CoV-2 infection can be detected using molecular diagnostics. Molecular diagnoses are crucial for the early detection of COVID-19 because antibodies can only be identified 8 days after the onset of symptoms [157,158]. In general, molecular tests perform better when compared to serological testing when used to diagnose COVID-19, with AUCFPR values of 0.949, 0.936, and 0.952 for RT-PCR, RT-LAMP, and CRISPR, respectively.
To find the antibody response brought on by COVID-19, there are numerous serological tests available. ELISA, CLIA, LFIA, CMIA, FIA, and ANB are the principal techniques [159][160][161]. The performance of serological assays for the identification of SARS-CoV2 has been reported to be satisfactory, however, this performance is dependent on the patient's recovery, making it important to understand the kinetics of antibodies during SARS-CoV-2 infection. Establishing the serological result is crucial in this situation for illness diagnosis [161,162]. Serological assays based on ELISA, CLIA, LFIA, CMIA, and FIA data have modest sensitivity (median less than or equal to 90%) and high specificity (median higher than or equal to 97%). Since they cannot distinguish between individual immune responses or intrinsic immunological abnormalities, low sensitivity might result in false negative results. Additionally, it must be considered that antibodies can be generated in asymptomatic patients. Nevertheless, their titers are lower than those found in symptomatic patients [163]. The immune system is stimulated by SARS-CoV-2 invasion and antigen release to develop a variety of antibodies (IgM/IgA/IgG). Immune cells of the host produce IgM and IgA at an early stage of infection, whereas IgG is produced at a later stage [164][165][166].
Diagnostic methods such as ELISA, CLIA, and LFIA were classified based on the antibodies they identified (IgG, IgM, IgA, and IgG-IgM). These tests' ability to detect IgG produced the highest AUC FPR , indicating the antibody's superior performance. This may be because IgM levels begin to decline at week 5 and almost completely disappear by week 7, whilst IgG levels continue to rise after week 7, indicating the greater stability of IgG [167]. On the SARS-CoV-2 membrane, however, several structural proteins are anchored, primarily the spike (S), nucleocapsid (N), membrane (M), and envelope (E) [166,168]; N protein can be helpful in the diagnosis of COVID-19 since it is a highly immunogenic protein that is extensively produced during infection [169]. The median sensitivity and specificity of the LFIA approach were 74.7 percent and 99.4 percent, respectively, when its performance was compared to the detection of the N protein of SARS-CoV-2. Additionally, RNA is less stable than N protein, which can effectively make up for the limited sensitivity it exhibits [170]. The ability of numerous companies to produce serology tests, which may be administered to millions of people daily, can aid in the improvement of SARS-CoV-2 detection, particularly in nations with low resources [170]. However, it is advised to combine clinical, molecular, and serological diagnostic tests to achieve acceptable sensitivity and specificity [159].
It is important to note that the ANB diagnostic test had the highest sensitivity and specificity scores for the identification of COVID-19 when molecular and serological meth-ods were considered. It also had the highest AUC FPR , coming in at 0.984. Neutralizing antibodies can shield cells from virus invasion and provide protective immunity; they are created weeks after infection [171]. ANBs can objectively detect SARS-CoV-2 neutralizing antibodies, allowing for the analysis of the correlation between neutralizing antibody levels and disease severity. They can also forecast the likelihood of reinfection in COVID-19 patients [171,172]. The plaque reduction neutralization test (PRNT), the current gold standard for serological assays and the assessment of immune protection, is typically used to evaluate neutralizing antibodies [173].
The genetic makeup of SARS-CoV-2 can be investigated using molecular techniques like NGS, enabling its discovery [174], and MA has been utilized for genotyping and the identification of agents that cause diseases like SARS-CoV-2 [175,176]. Additionally, COVID-19 can be detected using the ECLIA serological-based diagnostic methodology [177,178]. The number of chosen studies for these methodologies, however, hindered their inclusion in the meta-analysis because the meta-analysis requires at least five studies with a p-value of less than 0.05 [142]. For "COVID-19", "Sensitivity and specificity", "Next-generation sequencing", "Microarray Analysis", and "Immunoassay", individual MeSH term searches revealed 195.931, 641.415, 49.966, 95.596, and 2.150 studies, respectively, but linking them turned up just 16, 3, and 367 research studies. For what ought to be regarded as work constraints, typical mistakes in systematic review and metaanalysis research, including study location and selection, missing crucial information about the results, improper subgroup analysis, conflicts with fresh experimental data, and duplicate publication [179]. Additionally, several issues, including the heterogeneity of the study groups, the clinical settings, and the diagnostic performance measures, were identified in the current investigation. Contrarily, inaccurate assessments of the results of diagnostic tests may result in an overestimation [142,180]. It should be remembered that combining clinical, molecular, and serological diagnostic tests is advised to achieve acceptable sensitivity and specificity [159,181].

Conclusions
For a pandemic to be effectively managed and its spread to be stopped, it is critical to accurately detect emerging infectious agents like SARS-CoV-2. In the current study, the effectiveness of various diagnostic techniques reported for COVID-19 was assessed. For the detection of SARS-CoV-2, molecular tests (RT-PCR, RT-LAMP, and CRISPR) performed better than serological tests (ELISA, CLIA, LFIA, CMIA, and FIA) in terms of sensitivity and specificity. Additionally, it was discovered that serological tests had a very poor sensitivity but a high specificity, particularly when IgG was found. It should be noted that the ANB-based diagnostic tool reported the best performance among all the investigated approaches, demonstrating the potential for the diagnosis of SARS-CoV-2 infection. IgM and IgG serological diagnostic assays could be introduced based on the findings to track the COVID-19 acute phase and conduct ongoing surveillance. These tests can aid in identifying the existence of certain antibodies against the SARS-CoV-2 virus in the human body, which is helpful for detecting the illness when it is still in its acute stage. They may also be helpful in identifying those who have already contracted the virus and are immune to it. However, there is still potential for improvement in testing generally, and emphasis should be given to the creation of quick, scalable, and accurate assays for the prevention of future SARS-CoV-2 epidemics and other infectious diseases that might emerge.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/diagnostics13091549/s1, Figure S1: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of reverse transcription-polymerase chain reaction (RT-PCR) in the diagnosis of COVID-19; Figure S2: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of reverse transcriptase loop-mediated isothermal amplification (RT-LAMP) in the diagnosis of COVID-19; Figure S3: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of clustered regularly interspaced short palindromic repeats (CRISPR) in the diagnosis of COVID-19; Figure S4: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of enzyme-linked immunosorbent assay (ELISA) for IgG in the diagnosis of COVID-19; Figure S5: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of enzyme-linked immunosorbent assay (ELISA) for IgM in the diagnosis of COVID-19; Figure S6: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of enzyme-linked immunosorbent assay (ELISA) for IgA in the diagnosis of COVID-19; Figure S7: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of antiviral neutralization bioassay (ANB) in the diagnosis of COVID-19; Figure S8: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of biosensors (BS) in the diagnosis of COVID-19; Figure S9: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of chemiluminescence immunoassay (CLIA) for IgG in the diagnosis of COVID-19; Figure S10: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of chemiluminescence immunoassay (CLIA) for IgM in the diagnosis of COVID-19; Figure S11: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of chemiluminescence immunoassay (CLIA) for IgM-IgG in the diagnosis of COVID-19; Figure S12: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of lateral flow immunoassay (LFIA) for IgG in the diagnosis of COVID-19; Figure S13: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of lateral flow immunoassay (LFIA) for IgM in the diagnosis of COVID-19; Figure S14: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of lateral flow immunoassay (LFIA) for IgM-IgG in the diagnosis of COVID-19; Figure S15: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of lateral flow immunoassay (LFIA) for N-protein in the diagnosis of COVID-19; Figure S16: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of chemiluminescent microparticle immunoassay (CMIA) in the diagnosis of COVID-19; Figure S17: Study data and paired forest plot of the Positive Likelihood ratio, Negative likelihood ratio, and Diagnostic Odds ratio of fluorescence immunoassay (FIA) in the diagnosis of COVID-19; Table S1: PRISMA 2020 Checklist.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this study.