Claims that a rapid Covid-19 antibody test the government hopes to roll out this year is more than 99% accurate cannot be trusted, says a leading expert, calling for the full trial data to be made public. Jon Deeks, a professor of biostatistics and head of the test evaluation research group at the University of Birmingham, says the data published by Abingdon Health about the performance of its fingerprick antibody test were inadequate. The government hopes to roll out the test to millions of people. Writing for the Guardian, Deeks calls for the company to publish its data in full in the interests of transparency. “Secrecy destroys trust. Restoring scientific credibility will require urgent publication of the protocols and reports for these studies.” Other established British companies say they have antibody tests validated as being high quality by UK universities that they are selling around the world, but complain that they cannot get UK government approval. Ministers and their advisers appear focused on the Abingdon Health test. The company heads a consortium asked to design a rapid antibody test by Sir John Bell, regius professor of medicine at Oxford University, who is the government’s main adviser on tests. The other partners in the UK-rapid test consortium (UK-RTC) are Oxford University, BBI Solutions, CIGA Healthcare and Omega Diagnostics. Bell, who has called the AbC-19 test “truly amazing”, assessed a range of other antibody tests bought by Matt Hancock, the health secretary, in March and rejected them all as inadequate. They were mostly made by Chinese companies. The drive since has been to get a British test. In a leak to the Telegraph in July, it was claimed that Abingdon Health and partners had carried out “successful secret tests”. It said the British government “plans to distribute millions of free coronavirus antibody tests” as a result. Ministers were said to be hoping to use the test in a “mass screening programme before the end of the year”. While the test is currently “point of care” – meaning a medical practitioner does the fingerprick in a clinic or at home – Abingdon Health is thought to be preparing a dossier to get it approved as a self-test that anyone could use. The UK regulator, the Medicines and Healthcare Products Regulatory Agency, has set a bar of 98% accuracy for self-tests. Abingdon Health says its test is 99.4% accurate. However, Deeks says that statement cannotbe trusted without the full data to support it. Deeks has previously shown how companies making tests could “game” the data in their trials, for instance by selecting blood samples from people with high levels of antibodies. According to the limited trial data available from Abingdon Health, blood samples were excluded if a lab-based Elisa test called EuroImmun did not detect Covid-19 antibodies. EuroImmun is well known to be a poor test. Public Health England says it misses 28% of positive cases because it only picks up high antibody levels, not low ones. Deeks says it is possible that up to 79 true Covid-19 cases could have been excluded from the trial, which included 203 Covid-19 samples and 450 non-Covid-19 samples. He calls for scientifically rigorous and trustworthy studies and says the current regulations “fail to protect the public from the harm that can be caused by poor, inadequately evaluated tests”. Abingdon Health said: “It would be inaccurate to suggest our numbers are lower than we’ve stated. Based on independent study by Ulster University our tests are 99.40% accurate, with a sensitivity of 98.03%.” Further results would be published shortly, it said. Bell told the Guardian he had not seen the latest trial data from Abingdon Health or played any part in its validation. What he had seen of the test results “were as good or better than anything we’d seen before and all these other tests, and it was made in the UK, which means there will be security of supply,” he said. “So I’m sure it’s a good test.” But he supported Deeks’s call for transparency. “At some level I’m sympathetic to Jon Deeks because it is true that not all the validations in this space have been made transparently accessible for people to look at and say that looks OK and that doesn’t look OK,” he said.