Inadequate tests for Covid-19, based on poor or dodgy data, are proliferating in the UK because there are no clear rules on what companies have to prove before they can sell them, experts believe.
The Royal Statistical Society (RSS) has set up a working group to investigate and draw up guidelines for the companies making the tests and those who assess them and buy them. The organisation says there are “concerns about a lack of basic statistical evidence on clinical and analytical performance of a number of new diagnostic tests that are available”.
The group is led by the RSS president, Prof Deborah Ashby, together with Prof Jon Deeks, who has pointed out the failings of a number of tests. Deeks has cast doubt on the claim that an antibody test made by Abingdon Health and backed by the government is “99.4% accurate” and lambasted a test from iAbra – trialled at Heathrow airport – which claimed 99.8% accuracy on the basis of three samples.
Pinned to his Twitter account is a post entitled “21 ways to spin results from a Covid diagnostic test accuracy”.
The working group will look at all types of test, including the gold-standard PCR swab test for the virus, which misses a proportion of cases. “Are we evaluating just the test as it’s working in the laboratory, or the testing process as we apply it to a to a real human being?” he asked.
The lab test is good, he added, but whether people get the right result depends on whether enough virus was picked up from the nose and mouth on the swab. That needed to be factored in to the performance of the test.
“It’s not to say that test doesn’t work well. It’s just we haven’t accounted for the fact that it can miss 10% or 20% of cases, simply because of the sampling stage,” Deeks said.
It could be that we should be doing repeat tests, he said. “In a lot of countries, people aren’t allowed out of isolation till they’ve had two negative swabs,” he said. “Whereas in the UK we’ve only gone on one, which is a little bit of an outlier.” If the UK’s data from the tests is inaccurate, that could raise questions about the validity of modelling.
New rapid swab, saliva and antibody tests do not have to be approved by a regulator. Manufacturers are required only to self-certify that they have met certain standards to get a CE-mark. “The current CE-marking system is not fit for purpose. It doesn’t protect the public. It allows poor and misleading tests to be marketed. That doesn’t help the government make appropriate decisions,” said Deeks.