You can't tell diddly-squat about how good Covid antibody tests are says scientist

Antibody tests are meant to tell you if you've had Covid – and there are 200 brands approved for sale. But here a scientific expert argues there's little proof that they all work

Covid-19 antibody tests are still held up as our way out of lockdown and allowing healthcare workers to assist vulnerable patients
Lois Rogers

MORE and more people are desperate to know if they have already had coronavirus as evidence grows that the vast majority of those infected with Covid-19 have no symptoms at all.

Antibody tests are still held up as our way out of lockdown and allowing healthcare workers to assist vulnerable patients, and even cancer patients to undergo treatment.

There are now more than 200 brands of Covid antibody test approved for sale, many of them available over the internet, but they may not be a foolproof way of finding out if you're immune. There is little regulation of the industry churning out these kits and, while some are better than others, none is 100 per cent accurate.

Even the two antibody tests approved for NHS hospital use, made by the drug companies Roche and Abbott, introduced in May, are less accurate than originally claimed, with specialists scathing about the decision to buy millions of them.

One of the leading experts in this area, Jon Deeks, professor of biostatistics and head of the test evaluation research group at the University of Birmingham, said: "It's a mystery why [the British government] went with only Roche or Abbott... There are other tests that are better than them."

Professor Deeks, who has embarked on the first ever worldwide review of antibody tests, added emphatically that none of the tests available "are good enough for any of us to say for certain whether or not we have already caught the disease – and even if we have, whether we are safe from being infected with it a second time".

The need for accurate antibody tests has been highlighted by a new study, which confirms that many people can be infected with Covid but have no symptoms. When 217 people were cooped up for three weeks on a Covid-infected cruise ship en route for Antarctica in March, uniquely all their samples were genetically analysed. Around 60 per cent caught Covid, but 80 per cent of those had no symptoms, reported the journal Thorax.

Other studies have suggested that people with this kind of ‘silent' infection may not be infectious. But without reliable post-infection antibody tests, we can't know.

Antibody tests look for molecules produced by the immune system to fight off new infection. Their effectiveness is dictated by when exactly a person is tested (with Covid, antibodies are thought to start reliably showing three weeks after infection), the concentration of the test reagent (a chemical that reacts in the presence of antibodies), and whether the individual has actually raised antibodies against the virus.

Antibody tests can be done at home and typically require you to prick a finger to obtain a blood sample, dab it on absorbent fibre and then add a reagent. Minutes later the test reveals if your blood contains one or two types of Covid-19 antibodies.

An international research study led by Edinburgh University (published last week in the journal Public Health) suggests that, although the tests' ability to identify the absence of antibodies is probably accurate, their ability to detect their presence is much less certain: 68 per cent of the people told that they have the antibodies won't actually have had the disease.

This is down to the key concepts of a test's ‘sensitivity' (how accurate it is in spotting people who do actually have the antibodies) and its ‘specificity' (its accuracy spotting people who definitely don't have the antibodies. A 90 per cent specificity and sensitivity means that even just taking a tiny proportion of the population, say 100,000 people, as many as 14,000 would be wrongly diagnosed and told that they've had Covid, or they haven't, when the opposite is true.

"Giving false security and reassurance could be harmful for individuals, undermine public confidence and foster further outbreaks," the report in Public Health warns.

The UK government has declared that only two drug companies, Roche and Abbott, can supply the tests and they must all be done through hospitals using blood samples taken by a healthcare professional from people's veins, not just a finger prick. Roche and Abbott's tests were initially declared 100 per cent accurate but it quickly became apparent they are not.

In May, Prof Deeks raised concerns over the methods used for evaluations – he had been asked by the respected Cochrane organisation, a British-based global charity that reviews the claims made for medicines and diagnostic tests, to investigate such tests.

"Every week we have been looking at all the publications on Covid testing and all the data," he said, adding that 54 studies of antibody tests were published up to the end of April and another 60 studies in May.

Results from the review are expected shortly but are not expected to be promising.

"You can't tell diddly-squat about how good the tests are," Prof Deeks says. "There are more than 200 antibody tests on the market. They have CE (European conformity) marks but that only shows compliance with health, safety and environmental standards."

In basic terms, this means only that the tests will not harm consumers. The Roche and Abbott tests simply conform to these standards.

"We went to the Foundation for Innovative New Diagnostics in Geneva because they had the only list of coronavirus tests approved by licensing bodies around Europe but it is not comprehensive, and we're only picking up ones that they've listed," says Prof Deeks.

"Only 20 – less than 10 per cent of the total [with CE marks] – have published any sort of evaluation from an independent source to show how well they work."

He is equally scathing about the Roche and Abbott antibody tests, which the British government has agreed to buy, at an estimated cost of tens of millions of pounds. He said a review of Roche's evidence showed its test failed to pick up 15 known coronavirus antibody positive samples out of 93, making it only 84 per cent sensitive. The Abbott test was marginally better and identified 90 out of 96 positive samples, making it 94 per cent sensitive.

Prof Sheila Bird, a biostatistics expert from Edinburgh University, gave evidence to a House of Lords science and technology committee looking at the tests earlier this month. She is just as critical, pointing out all the evaluations used samples from severely sick hospital patients.

"We want to know how these tests perform in people who have had a level of the disease that didn't require them to go to hospital," Prof Bird said. "We don't have anywhere in the UK that can offer decent evaluations of tests that we are proposing to use on a large scale, whether it's for surveillance or for personal use. The accuracy has to be substantially higher."

Prof Karol Sikora, dean of Buckingham University medical school, is desperate for a test reliable enough to allow seriously ill cancer patients to resume chemotherapy (which destroys people's immunity) knowing they are not at risk of contracting the coronavirus. He fears that antibody tests, as they are, will not be the solution.

A spokesman for Abbott insisted its test was able to identify 100 per cent of Covid positive cases and 99.9 per cent of negative cases accurately but said: "The sensitivity of the test is affected by whether (and when) a patient develops antibodies, so it is more a question of the patients' immune response, not the performance of the test."

Roche said: "Our test is designed to be used as an aid to identify who has previously been exposed to the virus and has been through a rigorous regulated process.

"It is one of the most reliable on the market, reporting 100 per cent sensitivity 14 days post PCR confirmation [ie after a positive test for a current infection] and over 99.8 per cent specificity."

© Solo dmg media

Enjoy reading the Irish News?

Subscribe now to get full access