key: cord-0858416-mb4c7nci authors: Chua, Pearleen Ee Yong; Gwee, Sylvia Xiao Wei; Wang, Min Xian; Gui, Hao; Pang, Junxiong title: Severe Acute Respiratory Syndrome Coronavirus 2 Diagnostic Tests for Border Screening During the Very Early Phase of Coronavirus Disease 2019 Pandemic: A Systematic Review and Meta-Analysis date: 2022-02-14 journal: Front Med (Lausanne) DOI: 10.3389/fmed.2022.748522 sha: 7f337b6674b37ab05b1a28e52a566f523aaac1dd doc_id: 858416 cord_uid: mb4c7nci Diagnosis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) during border screening among returning residents and prioritized travelers during the early phase of a pandemic can reduce the risk of importation and transmission in the community. This study aimed to compare the accuracy of various SARS-CoV-2 diagnostics and assess their potential utility as border screening for infection and immunity. Systematic literature searches were conducted in six electronic databases for studies reporting SARS-CoV-2 diagnostics (up to April 30, 2020). Meta-analysis and methodological assessment were conducted for all included studies. The performance of the diagnostic tests was evaluated with pooled sensitivity, specificity, and their respective 95% confidence intervals. A total of 5,416 unique studies were identified and 95 studies (at least 29,785 patients/samples) were included. Nucleic acid amplification tests (NAAT) consistently outperformed all other diagnostic methods regardless of the selected viral genes with a pooled sensitivity of 98% and a pooled specificity of 99%. Point-of-care (POC) serology tests had moderately high pooled sensitivity (69%), albeit lower than laboratory-based serology tests (89%), but both had high pooled specificity (96–98%). Serology tests were more sensitive for sampling collected at ≥ 7 days than ≤ 7 days from the disease symptoms onset. POC NAAT and POC serology tests are suitable for detecting infection and immunity against the virus, respectively as border screening. Independent validation in each country is highly encouraged with the preferred choice of diagnostic tool/s. Rapid and accurate diagnosis of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection is crucial to contain the spread and inform clinical decisions to manage the patients' coronavirus disease 2019 (COVID-19) disease severity (1, 2) . Diagnostics play an essential role during the border screening to prevent importation and transmission in the community. Many countries have implemented travel restrictions, such as the closure of borders or limiting entry into the country, to control the spread (3) . However, these are unsustainable measures as the pandemic evolves over time (4, 5) . Active border screening of SARS-CoV-2 contributes to minimizing potential community transmission from the importation of cases and assist in priority travel. These screening tests may come in different forms. First, it could be via nucleic acid amplification tests (NAAT) which involve detection of viral genome in nasopharyngeal and oropharyngeal swabs through reverse-transcription PCR (RT-PCR). This is currently the reference standard for the diagnosis of SARS-CoV-2 infection. Tests typically target the envelope (E), nucleocapsid (N), spike (S), RNA-dependent RNA polymerase (RdRp) , and open reading frame 1 (ORF1) genes (6) . Second, it could be via serology tests, which involve the detection of IgM and IgG antibodies against SARS-CoV-2 in the blood typically 6-7 days after disease onset (7) . These antibodies are likely to remain detectable in the blood even after at least 8 months (8, 9) . Third, it could also be via chest imaging, screening suspected patients for features of COVID-19 infection as RT-PCR was limited during the early pandemic (10) . As PCR performance and capacity improves, imaging can be reserved to diagnose and monitor patients with severe condition or poorer prognosis (10) . Incorporating artificial intelligence into imaging gives rise to another potential diagnostic test as it increases diagnosis efficiency and accuracy, especially during the very early pandemic phase (10) . Currently, there are limited systematic and comprehensive assessments on the progress and status of diagnostic tests development at the very early phase of the COVID-19 pandemic for border screening (11, 12) . Therefore, this review aimed to compare the accuracy of different diagnostic tests (molecular, serology, clinical features, point of care testing, and imaging) and assess their potential utility as border screening for infection and immunity against SARS-CoV-2. This study was conducted with reference to Cochrane's Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Systematic searches were conducted in PubMed, Cochrane Library, Scopus, and Embase databases for published literature, and BioRvix and medRvix databases for grey literature on April 30, 2020. Since no restriction was set on the time period, the search timeline was from the point of database inception to April 30, 2020 . The search keywords such as "COVID-19, " "2019-ncov, " "SAR-CoV-2, " "diagnos * , " "polymerase chain reaction, " "serology, " "point of care, " "computed tomography, " "sensitivity, " and "specificity" were used to identify and extract articles that assessed diagnostic accuracy of existing COVID-19 diagnostic tools as presented in Supplementary Tables S1, S2. Reference lists of relevant reviews were hand-searched to identify any additional studies. The screening was done in duplicate by three authors (PEYC, MXW, SXWG). Identified publications were hierarchically screened according to the following criteria, and included in the review if they fulfilled all criteria: 1. Population: Cases are laboratory-confirmed COVID- 19 patients with no restriction on the countries, race, age group, and severity. No restriction on control definitions, which may or may not be tested for COVID-19. Controls in this review include (i) laboratory-confirmed negative COVID-19 negative patients, (ii) pre-pandemic controls without clinical suspicion of COVID-19, (iii) controls with other confirmed infections, or (iv) healthy controls. 2. Intervention/exposure: Diagnostic tests (including Point-of-Care tests) to identify and/or confirm SARS-CoV-2 infection; no restriction on the time points of the infection and sampling sites. Diagnosis tests for prognosis will be excluded. 3. Outcome: Clinical sensitivity and specificity. 4. Comparator: SARS-CoV-2 samples or patients confirmed by nucleic acid tests or next-generation sequencing. Disagreements were discussed with the fourth author (JP) to reach a final consensus. This review defines the reference standard test as either PCR or sequencing. Point-of-care (POC) diagnostics are defined primarily as tests that can be conducted at the point of patient care, outside laboratory settings. Titles and abstracts of extracted studies were first assessed for relevance before full texts of relevant studies were retrieved further screening with the above criteria. If available, published versions of included preprint studies were used for data extraction. A PRISMA flow diagram of the study selection process is shown in Figure 1 . A developed data extraction form was pilot tested with a subset of eligible studies and the finalized form consolidated the following information from each study: author, country of conduct, study and population characteristics, the type of index diagnostic test, and their criteria/cut off, and outcome measures. Outcome measures include true positive (TP), true negative (TN), false positive (FP), false negative (FN), sensitivity, specificity, and any other key findings. Each index test assessed in included studies was extracted as an individual study, i.e., if multiple index tests were examined in a single study using the same/overlapped population/samples, each index test was considered an independent study with its own dataset. In studies assessing outcomes with control groups, controls tested by the reference test will be preferred, otherwise, a generalizable control such as a healthy population will be selected. In addition, the per-patient and high-performance outcomes will be preferred if alternative data were reported. Lastly, only validation and test set results were extracted from studies assessing diagnostic models or artificial intelligence (AI) The methodological quality of included studies was evaluated with Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2). The original tool includes 14 items assessing the risk of bias, sources of variation, and reporting quality; possible responses to each item were "yes, " "no, " or unclear" (13) . Signaling questions and their scoring were adapted from Deeks et al. (14) and McInnes et al. (15) and modified to suit this review. The questions are presented in Supplementary Table S4 . Studies were assessed for risk of bias in four domains: patient selection, index test, reference test, and flow and timing. The risk of bias in each domain will be rated for each study; possible domain ratings are low, unclear, or high risk of bias. Quality assessment was conducted in duplicate by PEYC, MXW, and SXWG. Discrepancies were resolved by consensus with J.P. before moving on to the next stage of analysis. The main outcome compared the accuracy of diagnostic tests with different working principles: (i) nucleic acid amplification test excluding sequencing (NAAT), (ii) NAAT POC tests, (iii) sequencing, (iv) serology, (v) serology (POC), (vi) imaging, (vii) imaging with artificial intelligence (AI), (viii) clinical and/ or laboratory features model, and (viii) combination of diagnostic tests. Serology tests (including POC tests) were analyzed based on the detection of immunoglobulin G (IgG), immunoglobulin M (IgM), IgG and/or IgM, and antibodies (Ab). Subgroup analysis was further conducted according to the following: (i) duration of samples collected from the onset of symptoms, early (≤ 7 days) and late (≥7 days) phase for serology and serology (POC); (ii) gene targets of NAAT and NAAT (POC); (iii) sample collection from different specimen sites for NAAT and NAAT (POC); (iv) differential performance of diagnostic tests on symptomatic and asymptomatic individuals; (v) geographical regions where studies were performed, based on Asia, China, America, and Europe. For studies with incomplete data, TN, TP, FN, and FP were calculated based on the 2 × 2 table or as much as possible, the inbuilt calculator in Review Manager 5.3. Bivariate analysis was conducted with the provided TN, TP, FN, and FP to generate sensitivity, specificity, diagnostic odds ratio (DOR), and the summary receiver operating characteristic curves (SROC) with their corresponding 95%CI. In the comparison of diagnostic accuracy across different working principles, forest plots included studies reporting both sensitivity and specificity. Subgroup analyses included studies reporting either one or both sensitivity and specificity. In the subgroup analysis on symptomatic and asymptomatic patients, specificity was not pooled as controls did not display COVID-19 symptoms and cannot be stratified. Continuity correction was performed for DOR and SROC, and sensitivity and specificity when necessary. Sensitivity, defined as TP/(TP+FN), indicates the proportion of positive cases that the test correctly identifies in COVID-19 subjects (12) . Specificity, defined as TN/(TN+FP), indicates the proportion of negative results a test correctly identifies in non-COVID-19 samples (12) . The calculation of DOR and SROC utilized both sensitivity and specificity (16) . A high DOR indicates good diagnostic accuracy. The area under the curve (AUC) of SROC-which reflects the overall performance of the test-was also calculated. An AUC of one indicates a perfect test. The I 2 statistic and Cochrane test were used to evaluate statistical heterogeneity. Heterogeneity was characterized as minimal (<25%), low (25-50%), moderate (50-75%), or high (>75%) and significant if p < 0.05. The publication bias for the included studies was assessed through Deek's funnel plot asymmetry test. The slope coefficient with p < 0.10 indicated a significant asymmetry. Meta-analysis was conducted. However, this study prioritized and only discussed pooled estimates derived from three or more studies. Full meta-analysis results can be found in the Supplementary Material. Publication bias was conducted when there were more than two studies available. The A total of 5,416 unique studies were screened for relevance with their titles and abstracts. Subsequently, 5,221 studies and 100 studies were respectively excluded in the primary and the full-text screening. Studies excluded during full-text screening can be found in Supplementary Table S5 . Of the 95 studies eventually included in this review, 85 studies were included in the meta-analysis. The 95 included studies involved a total of at least 29,785 patients/samples. At least 11 studies recruited asymptomatic patients. Studies included in this review were mostly conducted in China (n = 56), Italy (n = 10), and the United States (n = 6). There were three studies each from Hong Kong and the Netherlands, and two studies each from Japan, the United Kingdom, and Germany. One study was conducted each in South Korea, Slovenia, Spain, Taiwan, France, Canada, Belgium, Sweden, and Denmark. There was one study conducted across two countries, in China and US. Serology tests were generally less sensitive (range 55-98%), while specificity remained high (range 95-100%), depending on the test's targeted antibody. The number of tests with the respective targeted antibody/antibodies was summarized in Table 1 . Serology tests concurrently detecting for IgG and/or IgM antibodies (n = 11; 89%, 95%CI: 82-93%) was the most sensitive, i.e., accurate in identifying true positives, followed by detecting for IgM only (n = 16; 81%, 95%CI: 71-88%), IgG only (n = 16; 78%, 95%CI: 64-88%), and for both IgG and IgM (n = 4; 55, 95%CI: 43-67%). Sensitivity for tests detecting unspecified Ab antibodies (96%, 95%CI: 91-98%) was also relatively high. Two serology studies involving IgA antibodies had high sensitivities of 93 and 99%, with a high specificity of 93 and 98% respectively. Overall, serology (POC) has low pooled sensitivity ranging from 40 to 69%, albeit high pooled specificity ranging from 95 to 99%. Antigen test (POC) has a comparative sensitivity 98%, 95%CI: 90-100% and specificity 100%, 95%CI: 89-100% with NAAT (POC). However, this was based on a single study. The incorporation of AI into diagnostic imaging (n = 10; 89%, 95%CI: 84-93%) resulted in superior sensitivity compared to conventional imaging alone (n = 8; 82%, 95%CI: 65-91%). Likewise, imaging (AI) had relatively high specificity of 93%, 95%CI: 87-96% as compared to conventional imaging (62%, 95%CI: 47-75%). Diagnostics based on clinical and/or laboratory features (n = 4) ranked between imaging (AI) and imaging with its pooled sensitivity of 86% (95%CI: 75-92%) and specificity of 84% (95%CI: 72-92%). While outperforming conventional imaging, clinical/laboratory feature-based diagnostics still fell below AI incorporated imaging. Nonetheless, imaging methods generally had better sensitivity than most serological methods detecting IgG/IgM. The use of laboratory-based serology methods together with POC serology for IgG, IgM, and Ab respectively increased sensitivity to 94-99%. However, only one study reported the use of this approach and results should be interpreted with caution. Sequencing results from two studies were highly sensitive at 100%, but specificity was inconclusive with one study at 43% and the other at 0%. Inconclusive specificity resulted from the noninclusion of controls in the study sample, while 0% specificity resulted from the novel test's inability to accurately identify true negatives, i.e., controls from the sample. This could be due to the low number of controls (n=4) present in the small study sample of 13 individuals. In the subgroup analysis according to disease onset shown in Figure 3 , Supplementary The performance of NAAT and NAAT (POC) tests on specimens collected from different sampling sites were compared in Figures 5E-G Figure S8 and Supplementary Table S11 . The applicability of studies to this review was assessed across the domains of patient selection, reference standard, and index test used. Studies were rated as low, high, or unclear levels of concern regarding their applicability. In terms of the reference standard, there was little concern about the applicability of all studies utilized in this review. While most studies (58%) were of low level of concern in the index test domain, 40.2 and 1.8% of studies were of high and unclear levels of concern, respectively. This is attributed to the fact that multiple diagnostic kits reported were non-commercialized kits and may be conducted and interpreted differently from commercialized products. The applicability of included studies to our review fared poorly in the domain of patient selection, with only 29.2% graded of low concern. There was a high level of concern for 62% of studies due to the case-control design utilized, which is known to falsely increase sensitivity and specificity through spectrum bias (19) . The level of concern for applicability of 8.8% of studies cannot be ascertained. Our study affirms the better performance of NAAT and NAAT (POC) over other diagnostic tests. NAAT, which detects active infection, has been recommended by the National Institutes of Health (NIH) (20) and Centers for Disease Control and Prevention (CDC) (21) . Interestingly, NAAT (POC) retained high sensitivity and specificity despite a shorter workflow. NAAT (POC) included in the review were mostly commercialized test kits using automating RT-PCR, reverse transcription loopmediated isothermal amplification (RT-LAMP), and an in-house test using mRT-LAMP-LFB. An alternative to RT-PCR, RT-LAMP amplifies RNA under isothermal conditions (22), enabling a simpler, cheaper, and smaller performance device than a thermal cycler (22) . There are variants of RT-LAMP utilizing different detection methods such as iLACO assay, One-pot RT-LAMP assay, Integrated RT-LAMP, and CRISPR-Cas12. There is mounting evidence on the potential of RT-LAMP as a POC test as it is simpler, fast, and as sensitive as RT-PCR (22, 23) . Nonetheless, NAAT and NAAT (POC) performance should be interpreted with caution. While not many studies in this review reported cycle threshold (Ct) values, NAAT and NAAT (POC) performance may be influenced by the variation in Ct values used to determine positive cases across studies. Thus, cases/controls could be categorized as false negatives/positives depending on the Ct values used in the index tests. Serology tests detect antibodies against SARS-Cov-2, usually detectable after 1-3 weeks of symptom onset (24). They are not recommended for the diagnosis of acute COVID-19 infections and can delay infection control efforts in the community (20, 21) . However, they can be important for epidemiological surveillance since the detection of antibodies against SARS-Cov-2 can indicate past and asymptomatic infections (21) . In a typical humoral response, the body first produces IgM within 5-7 days of infection (25) . However, for SARS-CoV-2 infection, all three isotypes, IgM, IgG, and IgA can be detected in a narrow timeframe of seroconversion (25) . The detection of IgG and IgA before IgM, pointing to weak IgM response in SARS-CoV-2 infections, has been evidenced (25) . A study detected specific IgG during the early phase of illness in some individuals, 4-6 days after symptoms (25) . These collectively suggest that IgG or IgA detection could be more sensitive than the conventional IgM detection in the early stage of infection (25) . As IgA is in charge of mucosal immunity, it is a crucial first-line defense against such respiratory viruses. Studies have reported its detection as early as 3 days after symptoms appear before class-switching to IgG (26) . However, CDC refrained from concluding any distinction in assay performance based on immunoglobulin classes, IgG, IgG, and IgM, or total antibody. CDC further recommends against the use of IgA assay due to insufficient information on the dynamics of IgA detection in serum (24). Serological methods are also limited by disease prevalence in the community, which varies with the outbreak duration and virus strain in the country and the effectiveness of mitigation measures (27) . Low prevalence in the population, with the mainly asymptomatic general population, may challenge the serology test's accuracy in determining past infection (27) . Our findings for serology (POC) corroborate those by Ricco et al., who found commercially available serology (POC) tests with a moderate sensitivity of 64.8%, and high specificity of 98% in IgG/IgM (28) . Likewise, WHO does not recommend the use of serology (POC) tests for patient care purposes (29) . Serology (POC) tests are based on the lateral flow immunoassay (LFIA) technique-a solid-phase immunoassay combining the principles of thin-layer chromatography and immune recognition reaction (30) . Interestingly, serology (POC) and NAAT (POC) tests' sensitivities differ greatly from their laboratory tests counterparts. Our study noted a similar performance between NAAT and NAAT (POC) while serology (POC) has lower performance when compared to the laboratory method. NAAT (POC), such as Simplexa COVID-19 Direct (Diasorin Molecular LLC, Cypress, CA) (31) and GenMark ePlex SARS-CoV-2 assay (32), essentially automate steps in the workflow of a laboratory assay. The similar performance of the diagnostic tests could be attributed to the same fundamental methodology utilized in the assays. Conversely, serology (POC) relies on the immunochromatographic visualization of lines on nitrocellulose membranes. The simplicity of cassette-like serology (POC) tests is a double-edged sword, as results are visually interpreted. This increases the ease of interpretation but also the subjectivity of the interpretation. Sensitivity is reduced as results are prone to a false negative. Other factors that may interfere with serology (POC) test performance include execution by inadequately trained personnel, which is common in manpowerstrained settings, loading of insufficient sample volume due to dropper usage (33) , and delayed interpretation of serological assay cassettes (34) , which may distort the results initially obtained. The commencement of vaccination programs globally may not entirely affect the use of serology for SARS-CoV-2 diagnosis if the vaccines used do not result in the production of target antibodies (such as anti-N, E, or M) of the serology test. The four vaccines are mainly used in mass vaccination programs, namely those by Pfizer-BioNTech (BNT162b2), AstraZeneca (AZD1222), Gamaleya (Sputnik-V), and Moderna (mRNA-1273) all result in the generation of antibodies against the spike protein (35-37) Moreover, the majority of the registered diagnostic tests with the United States Food and Drug Authority emergency use authorization are detecting other structural genes/proteins, rather than the spike protein. With the potential increasing need to assess for the presence of immunity due to either prior (but not recent infection) infection or due to vaccination at the border, the POC serological tests will become highly relevant with more countries adopting the requirement of the immunity "passport" in the near future. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) primarily infects the respiratory system, resulting in lung complications (12) . Imaging methods such as CT and x-ray, albeit highly sensitive, are limited by their low specificity. While CT can identify indistinct signs of the disease and at earlier onset as compared to x-ray (38) , there is no defined feature to differentiate COVID-19 from other respiratory conditions such as pneumonia or acute respiratory distress (39) . In times of pandemic, radiologists are insufficiently exposed to clinical manifestation of the disease to accurately differentiate cases (40) . Furthermore, as patients with mild disease presentation or in the early stage of COVID-19 may not present abnormality in their chest radiography, findings might be misled by potential comorbidities (39) . Hence, the choice between using CT or xray for diagnosis ought to factor in the duration since the onset of symptoms. Nevertheless, using solely imaging to diagnose patients with COVID-19 is not recommended due to variability in chest imaging findings documented to date (41) . Imaging should ideally be used for supplementary diagnosis or resource allocation in test-kit constrained settings. This review is the first to include studies that incorporated AI in diagnostic imaging to the best knowledge of the authors. Imaging (AI) was observed to have higher specificity as compared to conventional imaging as AI algorithms were trained to differentiate COVID-19 from other pneumonia (42) . The incorporation of AI in imaging tools can shorten result turnover from minutes to seconds (43) , which is crucial in managing the large influx of patients during a pandemic. Given the novelty of AI as an emerging technology in diagnosis, it would be recommended to harness imaging (AI) as a supportive tool for radiologists or as a complementary test for COVID-19. Clinical features or laboratory parameters have been increasingly utilized in models to diagnose or predict COVID-19. This could overcome the limitations of PCR and assist in test allocation in resource-constrained settings. Clinical manifestations in patients with COVID-19 range from mild symptoms to severe respiratory failure and even death (44) . A systematic review by Styurf et al. identified cough, sore throat, fever, myalgia or arthralgia, fatigue, and headache with a respective sensitivity of at least 50% (45). Fever, myalgia or arthralgia, fatigue, and headache were further identified to be at least 90% specific in diagnosing COVID-19, the combination of signs and symptoms was not explored (45) . This review identified the relatively high performance of modeling approaches incorporating clinical features or laboratory parameters. However, the result was highly heterogeneous, possibly due to the spectrum of variables included in each model. Models were either based on laboratory parameters exclusively, like the COVID-19 Assistant Discrimination 2.0 (46), or combined with demographic variables, like the COVID-19 Diagnosis Aid APP (47) . The overall variables included in each model ranged from 3 to 11, across the 4 studies included in the meta-analysis. Age was the most common demographic variable, while sex was included only on one occasion (48) , lactate dehydrogenase was the most common laboratory parameter. The model by Kurstjen et al. (48) , Corona-score, incorporated imaging results by CT and X-ray, while that by Li et al. (49) was the only model that incorporated overt clinical signs like respiratory symptoms and fever. Diagnostic modeling approaches utilizing commonplace laboratory test results could increase the efficiency of COVID-19 testing, which is especially crucial in resource-constrained settings. However, estimates from diagnostic models often risk being overly optimistic and misleading due to suboptimal methodology. A living systematic review by Wynants et al. recommended against using predictive models for current practice and further advised the use of updated patient data from the same setting to prevent miscalibration (50) . While serology tests were generally high in sensitivity and specificity; their usage should be considered in relation to the duration since symptom onset. Our review found higher diagnostic sensitivity in patients tested in the late phase as compared to the early phase of the disease. This finding echoes the review by Bastos et al., which found that pooled sensitivity across the immunoglobulin classes increases with the time of sample collection from symptom onset (51) . A possible reason is the rapid production of antibodies at later stages of the disease when the viral load decreases with increased seroconversion (52) . Zhao et al. demonstrated that the RNA test was the most sensitive during the early phase (within 7 days) of the illness, while serology tests had low positive rates in the early phase, but outperformed RNA tests 8 days following symptom onset (53) . These collectively suggest the complementary use of serology test with NAAT test in the later stage of the disease. Symptomatic patients were likely reviewed in most of the included studies because they were easier to identify and more predisposed to health-seeking behavior. Furthermore, symptombased testing may be prioritized by health authorities due to limited capacity, especially in the early phase of the pandemic. This study was not able to infer the performance of the diagnostic tests between the symptomatic and asymptomatic patients due to the small number of studies and samples in the asymptomatic group. Nonetheless, pooled results of two studies suggested higher sensitivity of serology and NAAT diagnostic tests in symptomatic patients. Serology tests might be less sensitive on asymptomatic individuals (34, 54, 55) or those with mild disease (56) as they mount weaker antibody responses. Asymptomatic individuals have been observed to produce significantly less IgG/IgM than symptomatic individuals, which potentially limits the performance of serology tests (54, 55) . This is also observed by better test performance of serology tests (both POC and non-POC) in symptomatic patients as compared to some studies with asymptomatic patients in our review. While there is ample evidence of similar viral load between asymptomatic and symptomatic patients, significantly faster viral clearance was observed in the former (57, 58) . Higher viremia causes lung damage that can be observed from radiography. Hence, a rapid turnover of the virus within the body may result in poorer performance of NAAT and imaging methods that rely on the detection of viral genomic material and lung manifestation. Although it is intuitive to assume a lack of viremia-induced lung inflammation in the absence of symptoms, Hu et al. identified CT abnormality in 71% of asymptomatic patients (59) while 54% of the asymptomatic patients in the Princess Diamond cruise ship had lung opacity in their chest CT (60) . On the other hand, Salvatore et al. had identified that Ct values vary with the presence of symptoms. Individuals with no symptoms at the time of sample collection had higher Ct values as compared to those reporting any symptoms (median Ct values 33.3 vs. 29.3) (61). Thus, holding all other factors constant, a higher Ct cut-off value increases the potential for false positives. Likewise, a lower Ct cutoff value may be more prone to false negatives, particularly in asymptomatic patients. In a 2014 study, Vivaldo et al. discussed differences in sensitivities of diagnostic tests due to the epidemiological evolution of DENV serotypes (62) . Similarly, we draw attention to the influence of SARS-CoV-2 distribution across geographical regions on diagnostic test accuracy. As of August 2020, there are at least 6 strains of coronavirus predominating in different geographical regions (63) . A genome-wide analysis found a higher frequency of amino acid mutation in Europe, followed by Asia and North America (64) . Potential mismatch between diagnostic RT-PCR assays and SARS-CoV-2 genome caused by mutations can result in false negatives by deterring primerbinding and amplification (65) . Hence, alternative diagnostic tests, e.g., serology tests may complement this potential pitfall of PCR primers being insufficiently accurate for detection in the presence of evolving variants. In this review, we could not infer significant differences between the geographical origins of the samples and the sensitivity of amplification tests. This suggests the continued relevance of current diagnostic tests in detecting SARS-CoV-2 infections globally. A reason behind this could be the inherent stability of coronaviruses, emphasized by genome data based on more than 90,000 SARS-CoV-2 isolates (66) . Alternatively, the advantageous use of two gene targets for diagnosis could have secured detection even if one target region has mutated sufficiently. We could not discern distinct differences in the performance of serology assays and imaging on samples from various geographical origins. As meta-analysis comprising only two studies was considered insufficiently robust, performance estimates of serology (non-POC) and imaging tests used outside China were largely excluded from the comparison. It is worthy of our attention as different strains could be characterized by distinct pathogenicity, which induce different levels of lung and clinical manifestations picked up by imaging methods. Altered immunogenicity in the various strains could translate to differential immune response, affecting serology assay performance and their effective time period. In line with our review finding, there is no literature reporting differential immunogenicity or pathogenicity of the strains to the knowledge of the authors at the point of writing. Alternatively, pre-test probabilities varying across geographical regions due to differing local prevalence may have influenced the false-positive rates, and thus the accuracy estimates (67) . Evidence backing the performance of gene targets is largely conflicting. The WHO established E gene and RdRp gene assays as first-line screening and confirmatory assays and recommended the less sensitive N gene assay as an additional confirmatory assay (68) . Conversely, Chu et al. suggested the use of E and N assay as screening assays, and RdRp and Orf1b as confirmation assays after identifying the N gene to have better sensitivity than Orf1b (69) . Others have reported that only the RdRp gene is almost specific for SARS-CoV-2. Interestingly, a preprint study by Loying et (71) . Difference in the persistence of gene positivity can also come to play a crucial role in COVID-19 diagnostics. Our subgroup analysis on different specimen types used in amplification tests found nasopharyngeal specimens as the most sensitive. Most samples were from the upper respiratory tract, albeit of unspecified location. We were unable to perform a comparative analysis between upper and lower respiratory tract specimens due to the paucity in studies utilizing lower respiratory tract specimens. The specimen collection site is important for the successful diagnosis of infection, given the reliance on amplification tests by health authorities in the community spread prevention and border reopening efforts (30) . While studies have identified lower respiratory samples to be more sensitive than upper respiratory tract samples (72, 73) , their use is limited by safety concerns and technical challenges in sample collection (72) . A review by Bwire et al. showed a moderate positive rate of 45.5% in nasopharyngeal specimens and a low positive rate of 7.6% in oropharyngeal specimens (74) . This review only included a single study by Chan et al. investigating the use of saliva specimens (17) . More recent studies have established saliva as a useful alternative sample type. Yokota et. al. reported NAAT sensitivity of 92% (90%CI: 83-97%) using saliva as compared to 86% (90%CI: 77-93%) using nasopharyngeal swabs during mass screening (23) . The sensitivity of saliva samples was at least on par with nasopharyngeal samples, as concluded by Wyllie et. al. in a separate study, and positive detection in asymptomatic healthcare workers was also higher using saliva samples (75) . Overall, saliva potentially minimized false positives, with better detection performance and lesser false negatives as compared to nasopharyngeal, oropharyngeal, and sputum samples in populations with low viral load (76) . Self-collection of saliva samples by patients can further mitigate the risk of exposure faced by healthcare workers, alleviating demands for swabs and personal protective equipment (23) . In this review, Chan (17) . Separately, Chen et al. detected SARS-CoV-2 in the stool of 67% of COVID-19 patients and concluded that viral presence was not associated with disease severity as positivity in stool specimens persisted in most samples even after pharyngeal swabs became negative (79) . On that note, the WHO recommends the use of stool specimens from the second week of symptom onset when there is clinical suspicion of SARS-Cov-2 despite negative respiratory specimens (80) . This review is constrained by the following limitations. Firstly, not all controls have received the reference test. At least 62.7% of the controls in the included studies and at least 66.4% of the controls included in the meta-analysis had received the reference test. Controls who were not tested or unclear were mostly constituted of samples obtained before the COVID-19 period, positive for other viruses or from healthy volunteers and blood donors at medical institutions. These subjects could have been misclassified as controls, resulting in lowered specificity. However, since pooled specificity in this review remains relatively high for most tests, there may be a minimal likelihood of misclassification. The second possible limitation is the interval between the novel and the reference test. Reference tests in some studies were conducted upon hospital admission, whilst novel tests may be conducted with new samples collected at a later time point. This may affect the time-sensitivity of the results, especially if the novel test in question is a genome amplification test (81) . On the other hand, re-testing old samples run the risk of false-negative if specimen degradation occurred under improper transport or storage conditions (82) . Thirdly, this review accepted either PCR or sequencing as reference tests, which can potentially cause misclassification. While we accepted two types of reference tests, sensitivity analysis was not performed since only three included studies utilized sequencing as the reference test and, hence, unlikely to yield a meaningful comparison. Fourthly, the predominant use of case-control study design by most included studies plausibly inflates both sensitivity and specificity estimates by the spectrum bias (83) . It is worthy to note that most cases were consecutively recruited patients at medical institutions despite controls usually being historic, stored samples. Another potential gap was the lack of antigen test development, which could be another viable rapid testing tool. Since antigen tests detect viral proteins, the possibility of it picking up a case precedes serology testing, which detects antibodies developed as part of our immune response. Next, this review could not compare diagnostic test performance between different age groups due to a lack of studies involving children. There is a gap in the existing literature that is skewed toward COVID-19 in adults. Studies on children are limited possibly due to testing practices that prioritizes symptomatic patients, health care workers, and institutionalized seniors. As children tend to have milder symptoms, they are less likely to be tested and diagnosed (84) . Yet, there is no substantial evidence of differences in the viral load and persistence of virus detection over time between adults and children. The inclusion of children may shed light on different diagnostic test performances in a normal population setting. This review could not do a subgroup analysis of test performance comparing immunocompromised and healthy individuals. Burbelo et al. concluded in their preprint that immunocompromised individuals generally have a delayed antibody response compared to healthy individuals (85) . More concrete evidence comes from Zhao et al., who postulated that early incomplete clearance of SARS-CoV-2 virus caused repeated negative RT-PCR tests and delayed antibody response in a patient with COVID-19 with HIV-1 and Hepatitis C coinfection (86) . The selection of diagnostic tests for this group of patients ought to consider these differences in immune response in addition to the time point of testing from disease onset. This review was only able to compare the different timing from onset to the sample collection for the serology test including POC as there were limited studies for the other diagnostic tests. Lastly, only studies published up to April 30, 2020 were included in this review, and thus, only reflected tests developed in the very early phase of the pandemic. Major developments in COVID-19 diagnostic testing occurred since the search for relevant studies in this review, to develop more accurate and efficient tests. Hence, tests presented in this review may not encompass all tests that have been commercialized or authorized for use, especially more current modalities such as breath test (87) and quantitative antigen test (88) . Nonetheless, most of the newly developed or improved versions of existing tests are built on the same technology as those presented in this review, maintaining continued relevance of the comparison between molecular tests, serologic tests, and radiologic tests with or without incorporation of artificial intelligence made in this review. Findings presented in this review may be useful to aid policy makers in assessing the suitability of a test for border screening or rapid diagnosis during the very early stages in the case of a future pandemic-better understanding of the performance of each test type in early outbreak phases will allow quicker response to control virus spread. Nucleic acid amplification tests had the highest performance, among others. Amplifications tests should be employed as the reference standard test to detect SARS-CoV-2 infection whenever possible. Point-of-care NAAT and serology tests have high potential utility for border screening due to their ease of conduct and shorter turnaround time. However, it should be noted that the time point since symptom onset and severity of the patient at the point of testing will influence the performance. Serologic tests were more sensitive when testing is done in the later phase of infection. All diagnostic tests were more sensitive among symptomatic than asymptomatic individuals, which emphasizes the importance of quarantining at-risk individuals and mandatory post-quarantine testing during the early phase of the pandemic. As more countries are making predeparture and post-arrival PCR testing mandatory, in addition to two or three PCR testing during the quarantine period, and potentially with the immunity "passport" requirement, it would be highly reasonable to deploy POC tests for border screening to alleviate the resource and time constraint for the increasing demand of laboratory tests as we progressively reopen the borders. The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors. JP conceptualized the study, supervised screening progress, validated the final screened studies and analysis, and critically reviewed the manuscript. PC, MW, and SG drafted the manuscript, screened the studies, and extracted data. PC and SG analyzed data and prepared the figures and tables. All authors have read and agreed to the published version of the manuscript. This research received funding from the Ministry of Defense (N-608-000-065-001). The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed. 2022.748522/full#supplementary-material Early treatment of COVID-19 disease: a missed opportunity Can early treatment of patients with risk factors contribute to managing the COVID-19 pandemic? World Tourism Organization. 100% of Global Destinations Now have COVID-19 Travel Restrictions WHO Says Travel Bans Cannot Be Indefinite; Countries Must Fight Coronavirus International Tourism Down 70% As Travel Restrictions Impact All Regions Interpreting diagnostic tests for SARS-CoV-2 Analytical performances of a chemiluminescence immunoassay for SARS Rapid generation of durable B cell memory to SARS-CoV-2 spike and nucleocapsid proteins in COVID-19 and convalescence Antibody responses 8 months after asymptomatic or mild SARS-CoV-2 infection COVID-19 imaging: what we know now and what remains unknown Effectiveness of tests to detect the presence of SARS-CoV-2 virus, and antibodies to SARS-CoV-2, to inform COVID-19 diagnosis: a rapid systematic review Systematic review with meta-analysis of the accuracy of diagnostic tests for COVID-19 QUADAS-2: A revised tool for the quality assessment of diagnostic accuracy studies Diagnosis of SARS-CoV-2 infection and COVID-19: accuracy of signs and symptoms; molecular, antigen, and antibody tests; and routine laboratory markers Imaging tests for the diagnosis of COVID-19 Summary receiver operating characteristic curve analysis techniques in the evaluation of diagnostic tests Improved molecular diagnosis of COVID-19 by the novel, highly sensitive and specific COVID-19-RdRp/Hel real-time reverse transcription-PCR assay validated in vitro and with clinical specimens Quantitative detection and viral load analysis of SARS-CoV-2 in infected patients Recognising bias in studies of diagnostic tests part 1: patient selection Testing for SARS-CoV-2 Infection Centres for Disease Control and Prevention. Overview of Testing for SARS-CoV-2 (COVID-19) Novel coronavirus disease (COVID-19): paving the road for rapid detection and point-of-care diagnostics Mass screening of asymptomatic persons for severe acute respiratory syndrome coronavirus 2 using saliva Humoral immune responses and neutralizing antibodies against SARS-CoV-2; implications in pathogenesis and protective immunity The role of IgA in COVID-19 Food and Drug Adminstration Point-of-Care diagnostic tests for detecting SARS-CoV-2 antibodies: a systematic review and meta-analysis of real-world data Advice on the Use of Point-of-Care Immunodiagnostic Tests for COVID-19: Scientific Brief Current diagnostic tools for coronaviruses-From laboratory diagnosis to POC diagnosis for COVID-19 Available online at Available online at Lateral flow (immuno)assay: its strengths, weaknesses, opportunities and threats. A literature survey Point-of-care serological assays for delayed SARS-CoV-2 case identification among health-care workers in the UK: a prospective multicentre cohort study Vaccines and What They Mean For COVID-19 Testing: John Hopkins Bloomberg School of Public Health Safety and immunogenicity of the ChAdOx1 nCoV-19 vaccine against SARS-CoV-2: a preliminary report of a phase 1/2, single-blind, randomised controlled trial How Gamaleya's Vaccine Works: The New York Times Rational and practical use of imaging in COVID-19 pneumonia The role of chest radiography in confirming covid-19 pneumonia Diagnosis of COVID-19 pneumonia using chest radiography: value of artificial intelligence Centres for Disease Control and Prevention. Interim Clinical Guidance for Management of Patients with Confirmed Coronavirus Disease (COVID-19) Review of artificial intelligence techniques in imaging data acquisition, segmentation and diagnosis for COVID-19 AI Tool Developed to Detect Abnormal Chest X-rays Quickly. The Straits Times Available online at Signs and symptoms to determine if a patient presenting in primary care or hospital outpatient settings has COVID-19 disease Rapid and accurate identification of COVID-19 infection through machine learning based on clinical available blood test results. medRxiv Development and utilization of an intelligent application for aiding COVID-19 diagnosis. medRxiv Rapid identification of SARS-CoV-2-infected patients at the emergency department using routine testing A simple laboratory parameter facilitates early identification of COVID-19 patients. medRxiv Prediction models for diagnosis and prognosis of covid-19: systematic review and critical appraisal Diagnostic accuracy of serological tests for covid-19: systematic review and meta-analysis Serology characteristics of SARS-CoV-2 infection since the exposure and post symptoms onset. medRxiv Antibody responses to SARS-CoV-2 in patients of novel coronavirus disease 2019 Clinical and immunological assessment of asymptomatic SARS-CoV-2 infections Antibody responses after COVID-19 infection in patients who are mildly symptomatic or asymptomatic in Bangladesh Diagnostic performance of commercially available COVID-19 serology tests in Brazil Descriptive, retrospective study of the clinical characteristics of asymptomatic COVID-19 patients. mSphere SARS-CoV-2, SARS-CoV, and MERS-CoV viral load dynamics, duration of viral shedding, and infectiousness: a systematic review and meta-analysis Clinical characteristics of 24 asymptomatic infections with COVID-19 screened among close contacts in Nanjing Chest CT findings in cases from the cruise ship "Diamond Princess Epidemiological correlates of polymerase chain reaction cycle threshold values in the detection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) A Meta-Analysis of the diagnostic accuracy of two commercial NS1 antigen ELISA tests for early dengue virus detection The Six Strains of SARS-CoV-2 ScienceDaily Genome-wide analysis of SARS-CoV-2 virus strains circulating worldwide implicates heterogeneity Presence of mismatches between diagnostic PCR assays and coronavirus SARS-CoV-2 genome The coronavirus is mutating-does it matter? Making sense of a negative COVID-19 swab test Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR Molecular diagnosis of a novel coronavirus (2019-nCoV) causing an outbreak of pneumonia Dynamics of ORF1ab and N Gene among hospitalized COVID-19 positive cohorts: a hospital based retrospective study. medRxiv Laboratory diagnosis of COVID-19: current issues and challenges CoV-2 Detection from upper and lower respiratory tract specimens: diagnostic and infection control implications Respiratory sampling for severe acute respiratory syndrome coronavirus 2: an overview Detection profile of SARS-CoV-2 using RT-PCR in different types of clinical specimens: a systematic review and meta-analysis Saliva or nasopharyngeal swab specimens for detection of SARS-CoV-2 Evaluation of saliva as an alternative diagnostic specimen source for SARS-CoV-2 detection by RT-dPCR Cautious handling of urine from moderate to severe COVID-19 patients Molecular and serological investigation of 2019-nCoV infected patients: implication of multiple shedding routes The presence of SARS-CoV-2 RNA in the feces of COVID-19 patients Available online at Estimating the false-negative test probability of SARS-CoV-2 by RT-PCR. medRxiv False Negatives and Reinfections: the Challenges of SARS-CoV-2 RT-PCR Testing Catalogue of Bias Collaboration Systematic review of COVID-19 in children shows milder cases and a better prognosis than adults Detection of nucleocapsid antibody to SARS-CoV-2 is more sensitive than antibody to spike protein in COVID-19 patients. medRxiv Early virus clearance and delayed antibody response in a case of coronavirus disease 2019 (COVID-19) with a history of coinfection with human immunodeficiency virus type 1 and hepatitis C virus S'pore to Roll Out Covid-19 Breath Test That Provides Fast Results: The Straits Times A novel strategy for SARS-CoV-2 mass screening with quantitative antigen testing of saliva: a diagnostic accuracy study The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.