Performance and health economic evaluation of the Mount Sinai COVID-19 serological assay identifies modification of thresholding as necessary to maximise specificity of the assay

Rushworth, S. A.; Johnson, B. B.; Ashurst, K.; Davidson, R.; Paddy, P.; Mistry, J. J.; Moore, J. A.; Hellmich, C.; Edwards, D. R.; Afonso, D. D.; Xydopoulos, G.; Goodwin, R.; Gomez, J.; Prakash, R.; Dervisevic, S.; Fordham, R.; Bowles, K.; Smith, J.

We evaluated the FDA approved SARS-CoV-2 immunoassay (developed at Mount Sinai, by Krammer and colleagues) for the identification of COVID-19 seroconversion and potential cross-reactivity of the assay in a United Kingdom (UK) National Health Service (NHS) hospital setting. In our "set up" cohort we found that the SARS-CoV-2 IgG was detectable in 100% of patients tested 14 days post positive COVID-19 nucleic acid test. Serum samples taken from pregnant women in 2018 were used as a negative control group with zero false positives. We also analysed samples from patients with non-COVID-19 viral infections, paraproteinaemia or autoantibodies and found false positive results in 6/179. Modification of the sensitivity threshold to five standard deviations from the mean of the control group eliminated all false positive result in the set up cohort. We confirmed the validity of the test with a revised threshold on an independent prospective "validation cohort" of patient samples. Taking data from both cohorts we report a sensitivity of the Mount Sinai assay of 96.6% (28/29) and specificity of 100% (299/299) using a revised threshold cut-off, at a time point at least 14 days since the diagnostic antigen test. Finally, we conducted a health economic probabilistic sensitivity analysis (PSA) on the costs of producing the tests, and the mean cost we estimate to be 13.63 pounds sterling (95%CI 9.63 - 18.40), allowing its cost effectiveness to be tested against other antibody tests. In summary, we report that the Mount Sinai IgG ELISA assay is highly sensitive test for SARS-Cov-2 infection, however modification of thresholding was required to minimise false positive results.