A finger-prick test for Covid-19 antibodies being bought by the government is considerably less accurate than claimed by the company that makes it, a formal evaluation by Public Health England has found.
On 6 October the health minister James Bethell announced the government had bought 1m of the tests, even though the Department of Health and Social Care (DHSC) had by then seen the evaluation report from PHE.
Abingdon Health, the company making the test, claimed it was 99% accurate. But a PHE report, published on Thursday in the British Medical Journal (BMJ), says the test can confirm 92.5% of cases already known to be positive. But in real-world conditions, the accuracy falls to 84.7%.
It means that if 10% of people given the test had previously been infected, about one in five positive test results would be incorrect, ie false positive results.
The department says the tests are still useful for surveillance – finding out what proportion of communities have been infected with the virus in the past. In an editorial in the BMJ, two experts, Dipender Gill from Imperial College London and Mark J Ponsford from Cardiff University, said the company had overestimated the test’s accuracy due to the manner in which its trials were carried out and that it should not be widely used.
The PHE study “identifies notable limitations of the UK government’s antibody test of choice and provides good evidence that its specificity in a ‘real life’ setting is highly unlikely to be 100%. Apart from limited surveillance to estimate the proportion of a population that has been infected, widespread use of this assay in any other role could risk considerable harm,” they write.
Emails seen by the Guardian show that the DHSC blocked online publication of the results at the time of the ministerial announcement. PHE said there were “significant risks” to delaying publication because the findings were out of line with the claimed accuracy of the tests.
DHSC officials denied they blocked publication, saying they preferred to go down the route of peer review.