Improving COVID testing

How can you catch all the false negatives?
08 September 2020

Interview with 

Ravi Gupta, University of Cambridge


a picture of a medical swab


One big problem when testing people for COVID-19 is false negative test results. This is where a person is tested for coronavirus infection and told they are negative when in reality they really are infected: one study found that this might be the case in more than half of patients. It probably happens because swabs collected from the nose and throat might not pick detectable amounts of the virus, particularly at the early or late stages of the infection. But there are ways of making sure that these people don’t slip through the net. Adam Murphy spoke to Ravi Gupta, form the University of Cambridge, who has been testing one of them…

Ravi - Yes. Well, it all stemmed from experiences we had very early in the epidemic, during the first wave in the hospital where I work at Addenbrooke's, where significant numbers of high likelihood patients were testing negative on their first swab and then being admitted to either intensive care or requiring oxygen. So these were clearly people who had a high probability of having the infection, but were coming up negative. And it just got me thinking that we really need to have a way of making a firm diagnosis. The reasons for having a negative test are multiple. One may be inadequate swapping technique or more likely is that during the later phases of illness, when you progress to moderate or severe disease, the virus actually is replicating or dividing or making copies of itself, not just up in the nose and throat, but actually down in the lungs, where it causes more severe inflammation, and that's why people get more sick.

So this study does not call into question what's happening in screening strategies. So if you screen negative, then you're not infectious. So I really want to make sure that people don't come away from this thinking that a negative test out in the community means you could still have COVID. Our study was really looking at hospitalised patients where they have virus production lower down towards their lungs. So we started asking ourselves, well, how else can you measure exposure to the virus or get a diagnosis? And of course your antibody response, your immune response, is a marker for infection. So we started looking at antibodies for coronavirus, and we were able to use a selected rapid finger prick test to tell us whether somebody had antibodies or not. And when we put the antibody test together with the swab test that looks for genetic material, we were able to come up with a highly sensitive and specific test for patients coming to hospital, so that we could get up to a hundred percent diagnoses.

Adam - Is that what you can get up to then, 100%?

Ravi - Yes, so in this small study, it was a relatively small number. It was only about 50 patients that we were able to do this in, because you needed to have stored serum, you needed to have a diagnosis at the front door and you needed to have had the correct swab taken. So you can identify everybody who has coronavirus using the PCR, combined with an antibody test. Which makes sense, because if you look at the way that the virus goes up and down and the way the antibodies come up and down, there's a very significant overlap between the two. So you should catch everybody using the two tests together.

Adam - And then what about the other side? False positives. Does this become a problem with this method?

Ravi - The nose and throat swab for genetic material of the virus, what we call PCR, has been shown to be very highly sensitive, with a very low false positive rate. You really don't get many false positives with that. But on the other hand, the antibody tests have a reputation for having a higher false positivity rate. So coming up with a positive, when the patient actually is negative. To address that, we looked at something like between 120, 150 patients with the antibody test. And we found only, I think, well, between zero and one false positive, so  this was actually better than we expected. So the tests actually performed very well. The finger prick antibody tests that people think are not doing so well, the selected ones performed very well in our hands. And that's the key thing. There are good tests and bad tests depending on the manufacturer. So if you choose the right test, you really can get high levels of accuracy.

Adam - And then how would you go about deploying this kind of system? Like, would it be test centres or hospitals, how would it work?

Ravi - In terms of direct application of our findings, I think we really should limit it to hospitals at the moment. And I think with the flu season coming in and lots of different viruses circulating, finger prick antibody test in 15 minutes upon admission would be very helpful because if it's negative and the patient has had symptoms for a week, let's say, you can confidently say they don't have coronavirus infection and probably have something else. So I think it could be very, very useful during periods of pressure on the NHS.


Add a comment