Serologic testing for SARS-CoV-2 has the potential to augment control strategies by improving our understanding of the true burden of infection during the current pandemic. While the small number of tests performed (in comparison to the estimated UK prevalence of 2.7% as of 28th March 2020)7 during this early validation study makes it difficult to draw high-powered conclusions, the results offer useful insights into the potential utility of POC serological testing at this early stage in the pandemic response. Key concerns exist around the reliability of serology testing and identifying any potential time points for appropriate use.
Our study raises significant concern for the reliance on serology testing in the initial infection phase. Analysis of the larger 200 patient cohort shows that when the initial patient group (days 5-9 post symptom onset) are considered, there is a sensitivity of 74%, with the potential therefore to miss a significant number of positive cases during their most infectious period.8 Practically, if considering potential use of the assays in the early phase of infection (as suggested by some of the assay manufacturers, and during which period self-isolation is advised by PHE) the sensitivity drops to 66% (n=32). It is not until the end of the second week from symptom onset that the test appears to demonstrate a more practical use. In the greater than 14 day since symptom onset group there was 100% sensitivity. The latest false negative was demonstrated on day 12 with reactive results produced in all tests taken thereafter (n=67) suggesting that there could be real utility in POC serology from 14 days onwards. Where PCR testing is currently limited due to demand this could therefore have a potential role in testing those healthcare workers who have already self-isolated because of possible infection in the last few months. In addition the assay may have a role in community testing where timelines in patients with mild symptoms already self-isolating would be less time critical. A role is not seen however for an impact on public health guidance around self-isolation practices during the infectious period. A PPV of 37% for the general population precludes any consideration of deployment of serology testing for general screening. Where a case definition is applied however, as is required for meeting the PCR testing criteria, the PPV is substantially improved (95.4%) and supports use of serology testing in carefully targeted populations. While limited to detection in targeted populations this would still enable a far better understanding of the infection-fatality ratio then we are currently able to draw from predominantly hospital cases of infection.
At a time when there is considerable pressure on healthcare services it is important to be able to identify SARS-CoV-2 negative patients, streamline pathways and help guide clinical decision making for differential diagnoses. In the larger, second half of the study, serological testing of ‘negative’ patients (n=50) demonstrated a specificity of 96% suggesting an additional role in supporting suspected inpatients that return negative PCR results. In particular this could help guide management of patients presenting outside the reliable detection period for PCR swabbing of the upper respiratory tract.8 If the ‘negative’ patients across both halves of the study (total n=55) are considered, three had reactive results. Given our selection of these ‘negative’ patients from among those who had an a priori clinical suspicion of SARS-CoV-2 infection but who were PCR negative (n=2/3), it is possible or indeed likely that these represented missed cases of SARS-CoV-2. One of the patients presented in extremis with acute respiratory distress-like features (Figure 1, patient 11), one with a prolonged symptom history prior to sampling for PCR, and one following inpatient hospital transfer after a prolonged stay due to a stroke and new onset of low-grade fever. Where PCR is currently the only diagnostic test in use, the ‘missed’ cases tested with PCR raise the possibility of supporting diagnostic capability via serology testing in specific cases. In particular, this may be beneficial in patients presenting during the late phase where PCR is less sensitive or where there is a protracted or unclear history of onset and clinical suspicion is supported by other findings, such as typical radiological changes.
When considering use in outpatient settings, it is vital that a suitable time period is established for a true negative result. Our study suggests that this point is to be found at day 13 post-symptom onset at the earliest, limiting the utility of these assays in enabling return to work or in informing additional self-isolation practices. There is however a real role in helping to identify those healthcare workers, and in fact the wider general population, that have been through a self-isolation period with presumptive infection as long as there is concomitant reinforcement of advice around the stringent infection, prevention control measures.
If widely employed, demand on suppliers for POC serological assays will be high and interchangeability of kits may therefore be required. The plurality of kits currently available makes it difficult to comment on individual assay performance at this early stage and it is therefore essential to evaluate performance across the range of kits available both against the gold standard (PCR) as well as against each other as soon as possible. Of the five different assays we evaluated, there was direct concordance for detection of SARS-CoV-2 infection across seven of the PCR positive patients (35/38 of reactive results). All five kits failed to detect SARS-CoV-2 infection from two known early positive samples taken at day 5 and day 6 post-symptom onset. Our findings suggested that results consistently appeared reactive from only day 8 post-onset of symptoms onwards across all five kits (extended to day 14 onwards in the larger second half of our evaluation). We cite ongoing concern for their reliability during that early period and the significant impact that missing a positive infection could have on public health control measures.
All POC serology kits are qualitative and while there was general correlation with colour intensity as time from symptom onset increased, variance in intensity of positive results was noted. The test kit instructions comment clearly on expectation for varied colour intensity of positive results and advise that even very weak colour should be considered to be reactive. When considering the practicalities of widespread testing, clear instruction on interpretation will have to feature in standard operating procedures and should allow for comment on interpreting the colour intensity of reactive samples. Antibody levels analysed via validated laboratory based enzyme-linked immunosorbent assay (ELISA) platforms have suggested higher levels following severe compared to mild infections and cohort studies will now be needed to evaluate the extent of reliability for POC serology testing in mild community infection.9
Little in the way of conclusion can be drawn from the five test kits used to explore concerns around end-user difficulties. Re-sampling of reactive patients provided re-producible results and further investigation should be conducted to comment on reproducibility of POC serology kits. Of significant use however is the observation that tests completed against manufacturer protocol by transferring either too little, or too much whole blood into the well resulted in failure. Those tasked with carrying out testing must complete the test as instructed using the marked pipette in order to have reliable results. Failure to register a control line in these cases should avoid the mistake of reporting an inappropriate false result.
When comparing the utility of the SARS-CoV-2 split IgM/IgG tests (Core tests®, OrientGene, VivaDiag, Encode) versus the SARS-CoV-2 total antibody test (Wantai), the former have a potential capacity to provide greater information. Of the total reactive tests in the 200 cohort study (n=130), 119 showed both IgM and IgG reactivity. There were 11 reactive samples that differentiated between IgM (6/130) and IgG (5/130) (Table 1a). IgM/IgG reactivity was seen as early as day 5 post-symptom onset and as late as day 40 but there was insignificant data to make conclusions on the IgM/IgG relationship in SARS-CoV-2 infection at this stage. At present we have very little understanding of the immunological response to SARS-CoV-2 and the timeline for which IgM is likely to remain detectable. If SARS-CoV-2 has ongoing transmission among humans, detection of IgM may then have a role in detecting sentinel cases during subsequent periods of infection. Staged evaluation of IgM and IgG relationship in known SARS-CoV-2 positive patients over a prolonged period will be required to inform on the potential additional benefits, if any, of splitting out the IgM and IgG reactivity in POC assays.
Cross-reactivity leading to false positive results is of considerable concern, both with human seasonal coronaviruses and zoonotic beta-coronaviruses, including SARS and MERS. Okba et al. used serum samples from 3 SARS-CoV-2 positive patients and found cross-reactivity with the SARS-1 S and S1 proteins, and less so but also with MERS-CoV S protein.9 The S1 protein appears to be more specific to the SARS coronaviruses, with cross reactivity not seen against an array of PCR positive samples for example seasonal human coronavirus OC43 as well as non-CoV pathogens, including EBV and CMV. Subsequent testing with IgA and IgG specific ELISA showed some cross-reactivity with OC43 PCR positive samples and highlights the importance of assay design, where perhaps multiple protein targets may be required to ensure specificity for SARS-CoV-2.9 This urgently needs investigation in the western European population before widespread role out of POC SARS-CoV-2 serology. The role of serology as an adjunct to PCR has also recently been reported, with a strong correlation seen between IgM and IgG and the severity of disease.10 Seroconversion for detection of IgM and IgG was seen at median day 12 and 14 respectively with possibility of detection much earlier in the clinical course and is supportive therefore of our findings of improving sensitivity through the first 14 days of infection.10
If shown to be successful in the community setting POC serology testing could have significant implications for tackling the pandemic in resource poor settings which lack the required infrastructure for centralised molecular techniques. The benefits of utilising point of care testing to improve the efficiency of case identification and resource allocation in such circumstances has previously been discussed during the Ebola epidemic and could therefore be of considerable use during the SARS-CoV-2 pandemic.11
The small number of assays evaluated in this study limits the degree to which conclusions can be drawn. Formal power calculations have not been conducted, but with a sensitivity of 100% at greater than 14 days, accompanied by a margin of error of only 0.04, this exploratory study could now inform larger scale appropriately powered studies to fully characterise the utility of these assays in specific population groups. Specifically, a greater number of samples would be required in order to comment reliably on sensitivity and specificity and in turn, once the evolving prevalence becomes clearer, the positive and negative predictive value of these tests in various groups. A false negative result could have considerable consequence and the low numbers do not fully account for this. The need for rapid initial verification of these assays to augment clinical and public health interventions in this pandemic does not yet enable comment on many aspects of the assays, including the IgM/IgG relationship in the weeks and months post-infection. While our cohort groups include immunocompromised patients, pregnant individuals, and elderly patients, we are unable to sufficiently comment on reliability in these groups. This study is limited to evaluating reactivity in symptomatic adults with moderate-severe disease only and response in children has not been investigated. All positive patients in this study were considered to have moderate-severe infection by nature of the admitting policy of the hospital and evaluation in those with mild disease symptoms is needed to comment further on the population level applicability of these assays.