Direct tests

The recommended test for diagnosis of SARS-CoV-2 infection involves detection of viral RNA using nucleic acid amplification tests (NAAT), such as reverse transcription (RT)-PCR (www.ecdc.europa.eu). In areas with widespread community transmission of SARS-CoV-2 and when laboratory resources are limited, detection by RT-PCR of a single discriminatory target is considered sufficient. There are still, however, specific technical considerations for laboratory testing, including specimen collection (variable collection methods), which samples to collect (upper or lower respiratory tract biospecimens, or other samples), time of collection in relation to the course of disease, and the availability of different laboratory test methods and kits (not all of which may be standardized or approved by authorities such as the US Food and Drug Administration). Then there are infrastructure considerations: are the approved laboratory facilities and trained manpower available, can the methodology be rapidly scaled up, and how are test results interpreted and false negatives excluded?

These issues have been faced by the whole scientific community, with a collective response to develop guidance. The currently used protocol was developed and optimized for the detection of the novel CoV at the Charité University Hospital, Geneva, Switzerland in collaboration with several other laboratories in Germany, the Netherlands, China, France, the UK and Belgium (Corman et al., 2020). Additionally, the existing protocol was further optimized by the Centers for Disease Control (CDC) in the USA through the comprehensive comparison and validation of alternative kits available for nucleic acid extraction and the use of alternative probe and primer sets for efficient SARS-CoV-2 detection in clinical samples (www.cdc.gov/coronavirus). Similar approaches are being undertaken by other national authorities as they continue to scale up provision for laboratories not using CE-marked assays (www.england.nhs.uk/coronavirus/).

The importance and variability of specimen collection was initially highlighted from comparison of positive rates from pharyngeal, nasal, blood, sputum, faeces, urine and bronchoalveolar lavage fluid specimens and fibrobronchoscope brush biopsy of patients with confirmed COVID-19 (Zou L et al., 2020). At present the CDC recommend collecting and testing an upper respiratory specimen, with a nasopharyngeal specimen being the preferred choice for swab-based SARS-CoV-2 testing. When collection of a nasopharyngeal swab is not possible, the following are acceptable alternatives: an oropharyngeal specimen, a nasal mid-turbinate specimen (using a flocked tapered swab), an anterior nares (nasal swab) specimen (using a flocked or spun polyester swab) or a nasopharyngeal wash/aspirate or nasal aspirate specimen. For individuals having invasive procedures, lower respiratory tract specimens are also recommended if available. Although the virus can be detected in other specimens, such as blood and stools, these have been generally less reliable than respiratory specimens.

At present it is recommended that specimens should be collected as soon as possible once a decision has been made to pursue SARS-CoV-2 testing, regardless of the time of symptom onset. The viral load in throat swabs is greatest at the time of viral onset and decreases monotonically thereafter (To et al., 2020; Zou L et al., 2020). Analysis of these temporal dynamics suggests that viral shedding may begin 2–3 days before the appearance of the first symptoms, facilitating pre-symptomatic or asymptomatic transmission (He et al., 2020). CoV have a number of molecular targets within their +ssRNA genome that can be used for RT-PCR assays. The WHO has provided primers for the genes that encode the structural proteins of the viral envelope (E) and the nucleocapsid (N), and for the RNA-dependent RNA polymerase (RdRp), which is a key part of the virus's replication machinery that makes copies of its RNA genome (Corman et al., 2020). However, there has been no demonstration that any one of these three sequences (E, N or RdRP) might offer an advantage for clinical diagnostic testing, with different targets being preferred by different authorities. For example, the Public Health England assay employs two probes against RdRp; one is a Pan Sarbeco-probe that will detect 2019-nCoV, SARS-CoV and bat-SARS-related CoV, while the second probe is specific to 2019-NCoV. Continued refinement of these NAAT assays is ongoing to facilitate their upscaling, while maintaining laboratory safety, a low cost and a high sensitivity (Won et al., 2020).

Detection of isolated viral antigens

Great efforts have been made in order to develop tests for the rapid detection of SARS-CoV-2 antigens. Antigen detection tests are designed to directly detect viral particles in biological samples such as nasopharyngeal secretions. Several rapid antigen tests have been proposed (Diao et al., 2020); however, the principal concerns are the false-negative rate due to either a low or variable viral load, and the variability in sampling, the latter having the potential to further compound the problem in cases with low viral titres, thereby increasing the false-negative rate (Tang YW et al., 2020).

Diao and colleagues (Diao et al., 2020) have reported preliminary results from the use of a fluorescence immunochromatographic assay for detecting the N protein of SARS-CoV-2 in both nasopharyngeal swab samples and urine from 239 participants, with comparison to NAAT testing where the intersection of the amplification curve and diagnostic threshold line (cycle threshold [Ct] value) was set at either ≤30 or ≤40 (Diao et al., 2020). With a higher viral load in the sample, the prespecified Ct value may be lower, as fewer replication cycles are required to achieve a detectable signal; however, with a low viral load, a greater number of replication cycles (higher Ct value) will be required for a detectable signal to be attained. For this assay with a prevalence of 87%, although the positive predictive value (PPV) was 100%, the negative predictive value (NPV) was 32% for a Ct ≤40, increasing to 97% for patients with a higher viral load as demonstrated by a Ct ≤30. This would suggest that, at present, this assay would only be useful in excluding those with high viral loads. Whether alternative approaches as previously suggested for influenza viruses in children including the use of colloidal gold-labelled immunoglobulin (Ig) G as the detection reagent (Li et al., 2020), to increase the sensitivity of rapid antigen tests for respiratory viruses, are feasible is still under consideration, with monoclonal antibodies specifically against SARS-CoV-2 under development. Further validation of this technique and similar approaches in larger populations including asymptomatic cases is warranted. Consideration of approaches to try to concentrate antigen and amplify the detection phase are, however, likely to be needed for these methods to have any clinical utility (Loeffelholz et al., 2020).

At the time of writing (25 April 2020), the non-governmental organization Foundation for Innovative New Diagnostics (https://www.finddx.org/) has listed four CE-marked rapid SARS-CoV-2 antigen detection tests, which are primarily lateral flow immunochromatographic assays based on the presence of a colloid gold conjugate pad and a membrane strip pre-coated with antibodies specific to SARS-CoV-2 antigens on a test line. If SARS-CoV-2 antigens are present in the specimen withdrawn from a nasopharyngeal swab, a visible band appears on the test line as antibody–antigen–antibody gold conjugate complex forms. The evaluation of these diagnostic tests has, however, been limited, and their CE mark means that the manufacturers state that they conform with the relevant EU legislation, but they may still not be available to purchase. According to European Union Directive 98/79/EC for in-vitro diagnostic devices, in order to affix the CE mark to COVID-19 diagnostic devices to be used by health professionals, the manufacturer has to specify device performance characteristics and self-declare conformity with the safety and performance requirements listed in the Directive. In addition, self-tests intended to be used by patients themselves must also be assessed by a third-party body (a notified body), which for these tests has yet to happen.

Although direct antigen tests are being registered by several health authorities, the sensitivity of these tests is lower than that of RT-PCR, with previous antigen-detecting enzyme-linked immunosorbent assays (ELISA) developed for SARS-CoV having limits of detection of 50 pg/ml (Che et al., 2004; Di et al., 2005). Furthermore, clarification of their specificity for SARS-CoV-2 is awaited, given the potential for cross-reaction with other human CoV. Despite these limitations, the chief advantages of antigen tests, including their rapidity (10–30 min compared with hours for NAAT testing), ease of interpretation and limited technical skill and infrastructure required compared with NAAT-based testing, continue to make them worth pursuing. However, experience with influenza antigen testing invites caution as these tests may have low sensitivity and specificity; moreover, as noted, false–negatives rate will be critical (Tang YW et al., 2020). Their greatest utility if they come to fruition may be in symptomatic patients, when the viral load will be at its greatest, to enable accurate triage.

Building an indirect test for SARS-CoV-2: serological testing

In contrast to NAAT-based testing, where as soon as the sequence is known, a diagnostic test can be built, the diagnostic technology and methodology underlying the development of serological tests is quite different, with a substantially longer timeline to obtain a robust product that is suitable for routine deployment. The principal difference is that antibody tests require identification of distinct proteins that form the viral coat, with elucidation of which proteins are most divergent from previous CoV proteins, then identification of specific antibodies to these proteins that are part of the acquired immune response to viral exposure, and finally testing to ensure that there is limited cross-reactivity with antibodies developed to other historical CoV.

With the previous two CoV, a variety of assays encompassing different methodologies were developed, including ELISA, chemiluminescence, western blotting, protein microarray and immunofluorescence platforms. However, only ELISA and chemiluminescence were deemed suitable for clinical application because of costs, time-to-results, relative simplicity and ability to scale to very large throughput. It is these platforms that are once again being examined for detection of antibodies to SARS-CoV-2.

Dynamics of seroconversion

Understanding viral and host interactions during the acute and convalescent phases is critical to be able to understand both the timing of initial seroconversion after exposure to SARS-CoV-2 and the subsequent duration of antibodies. However, at present the studies regarding seroconversion are being developed in parallel to the assays, limiting some conclusions. The data do suggest that seroconversion after exposure to SARS-CoV-2 is very similar to that after other acute viral infections, with IgG concentration beginning to rise as IgM concentrations reach a plateau (Figure 1 ). However, observations have shown that IgM and IgA growth is relatively slow related to other respiratory viruses, which has been suggested to contribute to the heterogeneous pathogenicity of SARS-CoV-2 in COVID-19 patients (Zhao J et al., 2020).

Open in a separate window

Figure 1

The time relationship between viral load, symptoms and positivity on diagnostic tests. The onset of symptoms (day 0) is usually 5 days after infection (day –5). At this early stage corresponding to the window or asymptomatic period, the viral load could be below the RT-PCR threshold and the test may give false-negative results. The same is true at the end of the disease, when the patient is recovering. Seroconversion may usually be detectable between 5–7 days and 14 days after the onset of symptoms; therefore, in the first phase of the disease, the serological tests are more likely to give false-negative results. The dotted black line in the graph illustrates the sensitivity of the chemiluminescent assay as derived from the data sheet of a commercial test (Abbott Diagnostics, USA). Ig, immunoglobulin; RT-PCR, reverse transcription-PCR; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

The most comprehensive study to date of seroconversion assessed 173 patients affected by COVID-19 using an assay developed to detect antibodies against the receptor binding domain of the S protein of SARS-CoV-2 (Zhao J et al., 2020). The median seroconversion times of total antibody, IgM and IgG were 11, 12 and 14 days (Zhao J et al., 2020). The respective seroconversion rates for total antibody, IgM and IgG were 93.1%, 82.7% and 64.7% (Zhao J et al., 2020), with the cumulative seroconversion curve suggesting that the rate for total antibody and IgM reached 100% 30 days after the onset. These studies have also highlighted the temporal nature of testing as, despite all patients subsequently being confirmed as COVID-19 positive, in the early phase of illness (within 7 days since the onset) the NAAT test only exhibited 66.7% sensitivity, with the antibody assays having an even lower positive rate of 38.3% (Zhao J et al., 2020). However, the sensitivity of antibody overtook that of RNA testing from day 8 after symptom onset and reached over 90% across day 12 after onset. Among samples from patients in the later phase (day 15–39 after onset), the sensitivities for total antibody, IgM and IgG were 100.0%, 94.3% and 79.8%, respectively. In contrast, RNA was detectable in only 45.5% of samples from days 15–39. In a separate small series of nine cases, seroconversion occurred after 7 days in 50% of patients (after 14 days in all) but was not followed by a rapid decline in viral load (Wolfel et al., 2020). An analysis of 285 patients further supports IgG seroconversion within 19 days after symptom onset (Long et al., 2020). Collectively, these data suggest that there is a role for both tests depending on where the patient is on their infection journey, with the combined use of NAAT and antibody tests markedly improving the sensitivity of a pathogenic diagnosis for COVID-19 patients in different phases of the illness.

With respect to antibody titres and disease severity, critically ill hospitalized patients have been reported to exhibit significantly higher antibody titre values than non-critical patients in some (Long et al., 2020; Zhao J et al., 2020) but not all studies. In the previous epidemics of SARS-CoV and MERS-CoV, antibody titres were positively associated with disease severity (Choe et al., 2017; Okba et al., 2019). In a limited case series (n = 57 confirmed SARS-CoV-2 cases), six patients with detectable viral RNA in the blood were at increased risk of severe disease progression compared with those with low titres, but unfortunately the authors did not measure antibody titres (Chen W et al., 2020b). Clarification is awaited of whether even in previously healthy individuals a high viral titre and/or high antibody titre can predict disease severity and likely progression.

Rapid serological tests

POC immunoassays have also been developed for the rapid detection of SARS-CoV-2 antibodies (IgG and IgM). The primary advantage of these tests, as with a home pregnancy test, is being able to obtain a diagnosis without sending samples to centralized laboratories. This enables communities without the necessary laboratory infrastructure to detect SARS-CoV-2-exposed subjects using only finger prick testing rather than formal blood draws, thereby reducing training requirements and allowing clinicians to have a validated test at the bedside. As these devices are cheap to manufacture, store and distribute, and provided that a positive antibody test were confirmed to be an accurate surrogate for immunity to infection, they would also be able inform decision making. This would particularly be the case as secure confirmation of antibody status would reduce anxiety, provide confidence to allow individuals to relax social distancing measures, and guide policy makers in the staged release of population lockdown, potentially in tandem with digital approaches to contact tracing.

The rapid POC immunoassays are generally LFIA (Li et al., 2020). In lateral flow assays, a membrane strip is coated with two lines: gold nanoparticle–antibody conjugates are located on one line and bind antibodies on the other. The blood sample from the patient is put on the membrane, and the proteins are drawn through the membrane strip by capillary action. As it passes the first line, the antigen binds to the gold nanoparticle–antibody conjugate, and the complex flows across the membrane. Generally, rapid assays have a low diagnostic performance compared with ELISA assays, which is explained not only by the well-known technical differences between the two methodologies, but also by possible low antibody concentrations that may further contribute to the false-negative results observed with the rapid tests.

At present, 12 peer-reviewed articles (Li Z et al., 2020; Cassaniti I et al., 2020; Lee YL, et al., 2020; Shen B et al., 2020; Dohla M et al., 2020; Hoffman T et al., 2020; Imai K et al., 2020; Pan Y et al., 2020b; Spicuzza et al., 2020; Yong G et al., 2020; Demey B et al., 2020; Montesinos I et al., 2020) and 9 pre-print studies (Garcia FP et al., 2020; Lassauniere R et al., 2020; Liu Y et al., 2020; Yong G et al., 2020; Lou B et al., 2020; Bendavid E et al., 2020; Paradiso AV et al., 2020a; National COVID testing Scientific Advisory Board, 2020; Tuaillon E et al., 2020) have reported on the diagnostic performance of rapid assays (summarized in Table 1).

In the published studies, sensitivity and specificity ranged from 9% to 88.6% and from 88.9% to 91.7%, respectively (Table 1), while in the pre-print articles sensitivity and specificity ranged from 30% to 98.8% and from 89% to 100%, respectively. Of note the sensitivity of these tests performed in countries other than China were substantially lower than those reported for studies conducted in China. Extensive evaluation of manufacturers’ claims related to the performance of these tests and optimal timing will be required before they are suitable for widespread routine clinical use. For example, the performance of the VivaDiag, VivaCheck Biotech, China COVID-19 IgM/IgG Rapid Test was evaluated in 30 cases 7 days after confirmed NAAT testing, and despite this five (16.7%) cases were negative for both IgG and IgM (Cassaniti et al., 2020). Furthermore, when evaluating 50 patients with acute illness presenting in the emergency room, of whom 38 were positive on RT-PCR, the sensitivity of the VivaDiag COVID-19 IgM/IgG Rapid Test was only 18.4%, its specificity was 91.7%, while the NPV was 26.2% and the PPV was 87.5% (Cassaniti et al., 2020). The same VivaDiag test was evaluated in 525 healthcare workers in Italy, with only six testing positive; none was positive by NAAT testing or symptomatic, and only three had a confirmed positive result on the MAGLUMI chemiluminescence IgG assay (Paradiso et al., 2020b). Evaluation of six POC tests in a mix of 111 patients with COVID-19, other CoV or other viruses and negative controls revealed sensitivities ranging from 83% to 93% and NPV of 74–92% (Lassauniere et al., 2020). In keeping with other studies, the diagnostic performance of these tests reflected the duration of the illness, with the worst performance observed in the first 2 weeks after symptom onset (Lassauniere et al., 2020). Finally, formal evaluation of nine commercially available LFIA in a case-control mix of 182 samples revealed sensitives of 55–70% (National COVID Testing Scientific Advisory Panel, 2020).

For all studies to date, sample size has been limited, and further testing across a large diverse population from a range of geographical locations and ethnic groups is required, with inclusion of children and individuals with autoimmune disease and immunosuppression. With extensive evaluation, it is likely that technical performance may deteriorate. At present, evaluation of the current LFIA devices suggest that although they may provide some information for population-level surveys, their performance is inadequate for most individual patient applications.