Identification of analytical issues in method verification of SARS CoV-2 serology

Dear Editor,

Laboratories have recently introduced serological testing for coronavirus disease 2019 (COVID-19, SARS CoV-2), in response to tight deadlines. Locally (Southern Region), we have collaborated in order to ensure method verification was as complete as possible, taking into account the extensive work undertaken by Public Health England (PHE). ¹ We would like to report some early observations.

Within the Southern region, both Abbott and Roche analysers are available. Samples were swapped between sites, and agreement of classification between these two methods was excellent (Cohen’s Kappa 0.928 (n = 83, GraphPad QuickCalcs). ² , ³ However, there was discordant classification for three samples. Two were found to be close to the positive cut-off according to the Abbott Assay at relative light units (RLU) slightly over the 1.4 threshold; they were classified as ‘not detected’ on the Roche assay. The third was positive according to the Roche assay (1.0 being the positive cut-off for this method) but negative when analysed by the Abbott assay (but close to the 1.4 cut-off). ⁴ , ⁵

Sample	Analyser units		Classification
	Abbott	Roche	Abbott	Roche
1	1.43	0.0828	Positive	Negative
2	1.41	0.0897	Positive	Negative
3	1.25	5.75	Negative	Positive

In the PHE validation, the majority of results showed good discrimination between values. On the Roche platform, at time of writing, of 2055 tests run, the probability of getting a value <1.0 but >0.2 is <1%. Likewise, on the Abbott method, results around the cut-off are hard to find.

Results were correlated with RNA PCR results; unsurprisingly, those with confirmed PCR COVID infections were strong positives on both antibody assays. However, out of an initial verification set of 52 confirmed PCR-positive samples, one gave a ‘not detected’ antibody result on the Roche assay. During routine use, a further two PCR-positive patients on the Roche have tested antibody negative. At 62, 60 and 45 days postpositive swab, the Roche results were 0.08, 0.18 and 0.247, respectively. Similar cases have been identified on the Abbott system.

It may be that some cases were caught early, prior to seroconversion, or late when the antibody response had faded. This may also account for the rarity of intermediate values. Alternative explanations include the possibility that the PCR result may have been a false-positive, or that not all individuals produce antibodies detectable by the current methods available (confirmed in PHE validation). Further data will be reviewed locally, and trials are being set up nationally to review the natural history of antibody status. We would encourage such studies in order to identify the reasons for discrepancies either biological or analytical.